tomoyo: Use bin2c to generate builtin-policy.h

Simplify the Makefile by using a readily available tool instead of a custom sed script. The downside is that builtin-policy.h becomes unreadable for humans, but it is only a generated file. Acked-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Michal Marek <mmarek@suse.cz>
author: Michal Marek <mmarek@suse.cz> 2015-01-09 14:08:26 +0100
committer: Michal Marek <mmarek@suse.cz> 2015-04-07 21:27:45 +0200
commit: 7e114bbf51fbb015dc25d8123e090afcce5b5048 (patch)
tree: 4a859c5f6756e2893a00c7f283529e7abc99875f /tools/perf/scripts/python
parent: c517d838eb7d07bbe9507871fab3931deccff539 (diff)
0 files changed, 0 insertions, 0 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/Kconfig b/drivers/gpu/drm/amd/amdgpu/Kconfig
index 12adca8c7819..7f515be5185d 100644
--- a/drivers/gpu/drm/amd/amdgpu/Kconfig
+++ b/drivers/gpu/drm/amd/amdgpu/Kconfig
@@ -2,10 +2,13 @@
 
 config DRM_AMDGPU
 	tristate "AMD GPU"
-	depends on DRM && PCI && MMU
+	depends on DRM && PCI
 	depends on !UML
 	select FW_LOADER
+	select DRM_CLIENT
+	select DRM_CLIENT_SELECTION
 	select DRM_DISPLAY_DP_HELPER
+	select DRM_DISPLAY_DSC_HELPER
 	select DRM_DISPLAY_HDMI_HELPER
 	select DRM_DISPLAY_HDCP_HELPER
 	select DRM_DISPLAY_HELPER
@@ -17,10 +20,13 @@ config DRM_AMDGPU
 	select HWMON
 	select I2C
 	select I2C_ALGOBIT
+	select CRC16
 	select BACKLIGHT_CLASS_DEVICE
 	select INTERVAL_TREE
 	select DRM_BUDDY
 	select DRM_SUBALLOC_HELPER
+	select DRM_EXEC
+	select DRM_PANEL_BACKLIGHT_QUIRKS
 	# amdgpu depends on ACPI_VIDEO when ACPI is enabled, for select to work
 	# ACPI_VIDEO's dependencies must also be selected.
 	select INPUT if ACPI
@@ -37,14 +43,16 @@ config DRM_AMDGPU_SI
 	bool "Enable amdgpu support for SI parts"
 	depends on DRM_AMDGPU
 	help
-	  Choose this option if you want to enable experimental support
+	  Choose this option if you want to enable support
 	  for SI (Southern Islands) asics.
 
-	  SI is already supported in radeon. Experimental support for SI
-	  in amdgpu will be disabled by default and is still provided by
-	  radeon. Use module options to override this:
+	  SI (Southern Islands) are first generation GCN GPUs,
+	  supported by both drivers: radeon (old) and amdgpu (new).
+	  By default, SI dedicated GPUs are supported by amdgpu.
 
-	  radeon.si_support=0 amdgpu.si_support=1
+	  Use module options to override this:
+	  To use radeon for SI,
+	  radeon.si_support=1 amdgpu.si_support=0
 
 config DRM_AMDGPU_CIK
 	bool "Enable amdgpu support for CIK parts"
@@ -53,22 +61,48 @@ config DRM_AMDGPU_CIK
 	  Choose this option if you want to enable support for CIK (Sea
 	  Islands) asics.
 
-	  CIK is already supported in radeon. Support for CIK in amdgpu
-	  will be disabled by default and is still provided by radeon.
-	  Use module options to override this:
+	  CIK (Sea Islands) are second generation GCN GPUs,
+	  supported by both drivers: radeon (old) and amdgpu (new).
+	  By default,
+	  CIK dedicated GPUs are supported by amdgpu
+	  CIK APUs are supported by radeon
 
+	  Use module options to override this:
+	  To use amdgpu for CIK,
 	  radeon.cik_support=0 amdgpu.cik_support=1
+	  To use radeon for CIK,
+	  radeon.cik_support=1 amdgpu.cik_support=0
 
 config DRM_AMDGPU_USERPTR
 	bool "Always enable userptr write support"
 	depends on DRM_AMDGPU
-	depends on MMU
 	select HMM_MIRROR
 	select MMU_NOTIFIER
 	help
 	  This option selects CONFIG_HMM and CONFIG_HMM_MIRROR if it
 	  isn't already selected to enabled full userptr support.
 
+config DRM_AMD_ISP
+	bool "Enable AMD Image Signal Processor IP support"
+	depends on DRM_AMDGPU && ACPI
+	select MFD_CORE
+	select PM_GENERIC_DOMAINS if PM
+	help
+	Choose this option to enable ISP IP support for AMD SOCs.
+	This adds the ISP (Image Signal Processor) IP driver and wires
+	it up into the amdgpu driver.  It is required for cameras
+	on APUs which utilize mipi cameras.
+
+config DRM_AMDGPU_WERROR
+	bool "Force the compiler to throw an error instead of a warning when compiling"
+	depends on DRM_AMDGPU
+	depends on EXPERT
+	depends on !COMPILE_TEST
+	default n
+	help
+	  Add -Werror to the build flags for amdgpu.ko.
+	  Only enable this if you are warning code for amdgpu.ko.
+
 source "drivers/gpu/drm/amd/acp/Kconfig"
 source "drivers/gpu/drm/amd/display/Kconfig"
 source "drivers/gpu/drm/amd/amdkfd/Kconfig"
diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile
index dd8bc53411bd..c88760fb52ea 100644
--- a/drivers/gpu/drm/amd/amdgpu/Makefile
+++ b/drivers/gpu/drm/amd/amdgpu/Makefile
@@ -1,5 +1,5 @@
 #
-# Copyright 2017 Advanced Micro Devices, Inc.
+# Copyright 2017-2024 Advanced Micro Devices, Inc. All rights reserved.
 #
 # Permission is hereby granted, free of charge, to any person obtaining a
 # copy of this software and associated documentation files (the "Software"),
@@ -23,7 +23,7 @@
 # Makefile for the drm device driver.  This driver provides support for the
 # Direct Rendering Infrastructure (DRI) in XFree86 4.1.0 and higher.
 
-FULL_AMD_PATH=$(srctree)/$(src)/..
+FULL_AMD_PATH=$(src)/..
 DISPLAY_FOLDER_NAME=display
 FULL_AMD_DISPLAY_PATH = $(FULL_AMD_PATH)/$(DISPLAY_FOLDER_NAME)
 
@@ -37,12 +37,17 @@ ccflags-y := -I$(FULL_AMD_PATH)/include/asic_reg \
 	-I$(FULL_AMD_DISPLAY_PATH)/modules/inc \
 	-I$(FULL_AMD_DISPLAY_PATH)/dc \
 	-I$(FULL_AMD_DISPLAY_PATH)/amdgpu_dm \
-	-I$(FULL_AMD_PATH)/amdkfd
+	-I$(FULL_AMD_PATH)/amdkfd \
+	-I$(FULL_AMD_PATH)/ras/ras_mgr
+
+# Locally disable W=1 warnings enabled in drm subsystem Makefile
+subdir-ccflags-y += -Wno-override-init
+subdir-ccflags-$(CONFIG_DRM_AMDGPU_WERROR) += -Werror
 
 amdgpu-y := amdgpu_drv.o
 
 # add KMS driver
-amdgpu-y += amdgpu_device.o amdgpu_kms.o \
+amdgpu-y += amdgpu_device.o amdgpu_doorbell_mgr.o amdgpu_kms.o \
 	amdgpu_atombios.o atombios_crtc.o amdgpu_connectors.o \
 	atom.o amdgpu_fence.o amdgpu_ttm.o amdgpu_object.o amdgpu_gart.o \
 	amdgpu_encoders.o amdgpu_display.o amdgpu_i2c.o \
@@ -50,7 +55,8 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \
 	amdgpu_cs.o amdgpu_bios.o amdgpu_benchmark.o \
 	atombios_dp.o amdgpu_afmt.o amdgpu_trace_points.o \
 	atombios_encoders.o amdgpu_sa.o atombios_i2c.o \
-	amdgpu_dma_buf.o amdgpu_vm.o amdgpu_vm_pt.o amdgpu_ib.o amdgpu_pll.o \
+	amdgpu_dma_buf.o amdgpu_vm.o amdgpu_vm_pt.o amdgpu_vm_tlb_fence.o \
+	amdgpu_ib.o amdgpu_pll.o \
 	amdgpu_ucode.o amdgpu_bo_list.o amdgpu_ctx.o amdgpu_sync.o \
 	amdgpu_gtt_mgr.o amdgpu_preempt_mgr.o amdgpu_vram_mgr.o amdgpu_virt.o \
 	amdgpu_atomfirmware.o amdgpu_vf_error.o amdgpu_sched.o \
@@ -60,7 +66,8 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \
 	amdgpu_umc.o smu_v11_0_i2c.o amdgpu_fru_eeprom.o amdgpu_rap.o \
 	amdgpu_fw_attestation.o amdgpu_securedisplay.o \
 	amdgpu_eeprom.o amdgpu_mca.o amdgpu_psp_ta.o amdgpu_lsdma.o \
-	amdgpu_ring_mux.o
+	amdgpu_ring_mux.o amdgpu_xcp.o amdgpu_seq64.o amdgpu_aca.o amdgpu_dev_coredump.o \
+	amdgpu_cper.o amdgpu_userq_fence.o amdgpu_eviction_fence.o amdgpu_ip.o
 
 amdgpu-$(CONFIG_PROC_FS) += amdgpu_fdinfo.o
 
@@ -71,20 +78,23 @@ amdgpu-$(CONFIG_DRM_AMDGPU_CIK)+= cik.o cik_ih.o \
 	dce_v8_0.o gfx_v7_0.o cik_sdma.o uvd_v4_2.o vce_v2_0.o
 
 amdgpu-$(CONFIG_DRM_AMDGPU_SI)+= si.o gmc_v6_0.o gfx_v6_0.o si_ih.o si_dma.o dce_v6_0.o \
-	uvd_v3_1.o
+	uvd_v3_1.o vce_v1_0.o
 
 amdgpu-y += \
 	vi.o mxgpu_vi.o nbio_v6_1.o soc15.o emu_soc.o mxgpu_ai.o nbio_v7_0.o vega10_reg_init.o \
 	vega20_reg_init.o nbio_v7_4.o nbio_v2_3.o nv.o arct_reg_init.o mxgpu_nv.o \
-	nbio_v7_2.o hdp_v4_0.o hdp_v5_0.o aldebaran_reg_init.o aldebaran.o soc21.o \
+	nbio_v7_2.o hdp_v4_0.o hdp_v5_0.o aldebaran_reg_init.o aldebaran.o soc21.o soc24.o \
 	sienna_cichlid.o smu_v13_0_10.o nbio_v4_3.o hdp_v6_0.o nbio_v7_7.o hdp_v5_2.o lsdma_v6_0.o \
-	nbio_v7_9.o
+	nbio_v7_9.o aqua_vanjaram.o nbio_v7_11.o lsdma_v7_0.o hdp_v7_0.o nbif_v6_3_1.o \
+	cyan_skillfish_reg_init.o
 
 # add DF block
 amdgpu-y += \
 	df_v1_7.o \
 	df_v3_6.o \
-	df_v4_3.o
+	df_v4_3.o \
+	df_v4_6_2.o \
+	df_v4_15.o
 
 # add GMC block
 amdgpu-y += \
@@ -93,11 +103,12 @@ amdgpu-y += \
 	gfxhub_v1_0.o mmhub_v1_0.o gmc_v9_0.o gfxhub_v1_1.o mmhub_v9_4.o \
 	gfxhub_v2_0.o mmhub_v2_0.o gmc_v10_0.o gfxhub_v2_1.o mmhub_v2_3.o \
 	mmhub_v1_7.o gfxhub_v3_0.o mmhub_v3_0.o mmhub_v3_0_2.o gmc_v11_0.o \
-	mmhub_v3_0_1.o gfxhub_v3_0_3.o gfxhub_v1_2.o mmhub_v1_8.o
+	mmhub_v3_0_1.o gfxhub_v3_0_3.o gfxhub_v1_2.o mmhub_v1_8.o mmhub_v3_3.o \
+	gfxhub_v11_5_0.o mmhub_v4_1_0.o gfxhub_v12_0.o gmc_v12_0.o
 
 # add UMC block
 amdgpu-y += \
-	umc_v6_0.o umc_v6_1.o umc_v6_7.o umc_v8_7.o umc_v8_10.o
+	umc_v6_0.o umc_v6_1.o umc_v6_7.o umc_v8_7.o umc_v8_10.o umc_v12_0.o umc_v8_14.o
 
 # add IH block
 amdgpu-y += \
@@ -109,7 +120,9 @@ amdgpu-y += \
 	vega10_ih.o \
 	vega20_ih.o \
 	navi10_ih.o \
-	ih_v6_0.o
+	ih_v6_0.o \
+	ih_v6_1.o \
+	ih_v7_0.o
 
 # add PSP block
 amdgpu-y += \
@@ -120,12 +133,12 @@ amdgpu-y += \
 	psp_v11_0_8.o \
 	psp_v12_0.o \
 	psp_v13_0.o \
-	psp_v13_0_4.o
+	psp_v13_0_4.o \
+	psp_v14_0.o
 
 # add DCE block
 amdgpu-y += \
 	dce_v10_0.o \
-	dce_v11_0.o \
 	amdgpu_vkms.o
 
 # add GFX block
@@ -136,11 +149,14 @@ amdgpu-y += \
 	gfx_v9_0.o \
 	gfx_v9_4.o \
 	gfx_v9_4_2.o \
+	gfx_v9_4_3.o \
 	gfx_v10_0.o \
 	imu_v11_0.o \
 	gfx_v11_0.o \
 	gfx_v11_0_3.o \
-	imu_v11_0_3.o
+	imu_v11_0_3.o \
+	gfx_v12_0.o \
+	imu_v12_0.o
 
 # add async DMA block
 amdgpu-y += \
@@ -152,13 +168,17 @@ amdgpu-y += \
 	sdma_v4_4_2.o \
 	sdma_v5_0.o \
 	sdma_v5_2.o \
-	sdma_v6_0.o
+	sdma_v6_0.o \
+	sdma_v7_0.o
 
 # add MES block
 amdgpu-y += \
 	amdgpu_mes.o \
-	mes_v10_1.o \
-	mes_v11_0.o
+	mes_v11_0.o \
+	mes_v12_0.o \
+
+# add GFX userqueue support
+amdgpu-y += mes_userqueue.o
 
 # add UVD block
 amdgpu-y += \
@@ -182,19 +202,39 @@ amdgpu-y += \
 	vcn_v2_5.o \
 	vcn_v3_0.o \
 	vcn_v4_0.o \
+	vcn_v4_0_3.o \
+	vcn_v4_0_5.o \
+	vcn_v5_0_0.o \
+	vcn_v5_0_1.o \
 	amdgpu_jpeg.o \
 	jpeg_v1_0.o \
 	jpeg_v2_0.o \
 	jpeg_v2_5.o \
 	jpeg_v3_0.o \
-	jpeg_v4_0.o
+	jpeg_v4_0.o \
+	jpeg_v4_0_3.o \
+	jpeg_v4_0_5.o \
+	jpeg_v5_0_0.o \
+	jpeg_v5_0_1.o
+
+# add VPE block
+amdgpu-y += \
+	amdgpu_vpe.o \
+	vpe_v6_1.o
+
+# add UMSCH block
+amdgpu-y += \
+	amdgpu_umsch_mm.o \
+	umsch_mm_v4_0.o
 
+#
 # add ATHUB block
 amdgpu-y += \
 	athub_v1_0.o \
 	athub_v2_0.o \
 	athub_v2_1.o \
-	athub_v3_0.o
+	athub_v3_0.o \
+	athub_v4_1_0.o
 
 # add SMUIO block
 amdgpu-y += \
@@ -202,7 +242,9 @@ amdgpu-y += \
 	smuio_v11_0.o \
 	smuio_v11_0_6.o \
 	smuio_v13_0.o \
-	smuio_v13_0_6.o
+	smuio_v13_0_3.o \
+	smuio_v13_0_6.o \
+	smuio_v14_0_2.o
 
 # add reset block
 amdgpu-y += \
@@ -215,6 +257,8 @@ amdgpu-y += \
 # add amdkfd interfaces
 amdgpu-y += amdgpu_amdkfd.o
 
+# add gfx usermode queue
+amdgpu-y += amdgpu_userq.o
 
 ifneq ($(CONFIG_HSA_AMD),)
 AMDKFD_PATH := ../amdkfd
@@ -227,9 +271,11 @@ amdgpu-y += \
 	amdgpu_amdkfd_gfx_v9.o \
 	amdgpu_amdkfd_arcturus.o \
 	amdgpu_amdkfd_aldebaran.o \
+	amdgpu_amdkfd_gc_9_4_3.o \
 	amdgpu_amdkfd_gfx_v10.o \
 	amdgpu_amdkfd_gfx_v10_3.o \
-	amdgpu_amdkfd_gfx_v11.o
+	amdgpu_amdkfd_gfx_v11.o \
+	amdgpu_amdkfd_gfx_v12.o
 
 ifneq ($(CONFIG_DRM_AMDGPU_CIK),)
 amdgpu-y += amdgpu_amdkfd_gfx_v7.o
@@ -271,4 +317,17 @@ amdgpu-y += $(AMD_DISPLAY_FILES)
 
 endif
 
+# add isp block
+ifneq ($(CONFIG_DRM_AMD_ISP),)
+amdgpu-y += \
+	amdgpu_isp.o \
+	isp_v4_1_0.o \
+	isp_v4_1_1.o
+endif
+
+AMD_GPU_RAS_PATH := ../ras
+AMD_GPU_RAS_FULL_PATH := $(FULL_AMD_PATH)/ras
+include $(AMD_GPU_RAS_FULL_PATH)/Makefile
+amdgpu-y += $(AMD_GPU_RAS_FILES)
+
 obj-$(CONFIG_DRM_AMDGPU)+= amdgpu.o
diff --git a/drivers/gpu/drm/amd/amdgpu/aldebaran.c b/drivers/gpu/drm/amd/amdgpu/aldebaran.c
index 2b97b8a96fb4..daa7b23bc775 100644
--- a/drivers/gpu/drm/amd/amdgpu/aldebaran.c
+++ b/drivers/gpu/drm/amd/amdgpu/aldebaran.c
@@ -35,7 +35,7 @@ static bool aldebaran_is_mode2_default(struct amdgpu_reset_control *reset_ctl)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle;
 
-	if ((adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2) &&
+	if ((amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 2) &&
 	     adev->gmc.xgmi.connected_to_cpu))
 		return true;
 
@@ -48,59 +48,64 @@ aldebaran_get_reset_handler(struct amdgpu_reset_control *reset_ctl,
 {
 	struct amdgpu_reset_handler *handler;
 	struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle;
+	int i;
+
+	if (reset_context->method == AMD_RESET_METHOD_NONE) {
+		if (aldebaran_is_mode2_default(reset_ctl))
+			reset_context->method = AMD_RESET_METHOD_MODE2;
+		else
+			reset_context->method = amdgpu_asic_reset_method(adev);
+	}
 
 	if (reset_context->method != AMD_RESET_METHOD_NONE) {
 		dev_dbg(adev->dev, "Getting reset handler for method %d\n",
 			reset_context->method);
-		list_for_each_entry(handler, &reset_ctl->reset_handlers,
-				     handler_list) {
+		for_each_handler(i, handler, reset_ctl) {
 			if (handler->reset_method == reset_context->method)
 				return handler;
 		}
 	}
 
-	if (aldebaran_is_mode2_default(reset_ctl)) {
-		list_for_each_entry(handler, &reset_ctl->reset_handlers,
-				     handler_list) {
-			if (handler->reset_method == AMD_RESET_METHOD_MODE2) {
-				reset_context->method = AMD_RESET_METHOD_MODE2;
-				return handler;
-			}
-		}
-	}
-
 	dev_dbg(adev->dev, "Reset handler not found!\n");
 
 	return NULL;
 }
 
+static inline uint32_t aldebaran_get_ip_block_mask(struct amdgpu_device *adev)
+{
+	uint32_t ip_block_mask = BIT(AMD_IP_BLOCK_TYPE_GFX) |
+				 BIT(AMD_IP_BLOCK_TYPE_SDMA);
+
+	if (adev->aid_mask)
+		ip_block_mask |= BIT(AMD_IP_BLOCK_TYPE_IH);
+
+	return ip_block_mask;
+}
+
 static int aldebaran_mode2_suspend_ip(struct amdgpu_device *adev)
 {
+	uint32_t ip_block_mask = aldebaran_get_ip_block_mask(adev);
+	uint32_t ip_block;
 	int r, i;
 
+	/* Skip suspend of SDMA IP versions >= 4.4.2. They are multi-aid */
+	if (adev->aid_mask)
+		ip_block_mask &= ~BIT(AMD_IP_BLOCK_TYPE_SDMA);
+
 	amdgpu_device_set_pg_state(adev, AMD_PG_STATE_UNGATE);
 	amdgpu_device_set_cg_state(adev, AMD_CG_STATE_UNGATE);
 
 	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
-		if (!(adev->ip_blocks[i].version->type ==
-			      AMD_IP_BLOCK_TYPE_GFX ||
-		      adev->ip_blocks[i].version->type ==
-			      AMD_IP_BLOCK_TYPE_SDMA))
+		ip_block = BIT(adev->ip_blocks[i].version->type);
+		if (!(ip_block_mask & ip_block))
 			continue;
 
-		r = adev->ip_blocks[i].version->funcs->suspend(adev);
-
-		if (r) {
-			dev_err(adev->dev,
-				"suspend of IP block <%s> failed %d\n",
-				adev->ip_blocks[i].version->funcs->name, r);
+		r = amdgpu_ip_block_suspend(&adev->ip_blocks[i]);
+		if (r)
 			return r;
-		}
-
-		adev->ip_blocks[i].status.hw = false;
 	}
 
-	return r;
+	return 0;
 }
 
 static int
@@ -124,9 +129,9 @@ static void aldebaran_async_reset(struct work_struct *work)
 	struct amdgpu_reset_control *reset_ctl =
 		container_of(work, struct amdgpu_reset_control, reset_work);
 	struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle;
+	int i;
 
-	list_for_each_entry(handler, &reset_ctl->reset_handlers,
-			     handler_list) {
+	for_each_handler(i, handler, reset_ctl)	{
 		if (handler->reset_method == reset_ctl->active_reset) {
 			dev_dbg(adev->dev, "Resetting device\n");
 			handler->do_reset(adev);
@@ -157,7 +162,7 @@ aldebaran_mode2_perform_reset(struct amdgpu_reset_control *reset_ctl,
 	if (reset_device_list == NULL)
 		return -EINVAL;
 
-	if (adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2) &&
+	if (amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 2) &&
 	    reset_context->hive == NULL) {
 		/* Wrong context, return error */
 		return -EINVAL;
@@ -210,8 +215,10 @@ aldebaran_mode2_perform_reset(struct amdgpu_reset_control *reset_ctl,
 static int aldebaran_mode2_restore_ip(struct amdgpu_device *adev)
 {
 	struct amdgpu_firmware_info *ucode_list[AMDGPU_UCODE_ID_MAXIMUM];
+	uint32_t ip_block_mask = aldebaran_get_ip_block_mask(adev);
 	struct amdgpu_firmware_info *ucode;
 	struct amdgpu_ip_block *cmn_block;
+	struct amdgpu_ip_block *ih_block;
 	int ucode_count = 0;
 	int i, r;
 
@@ -249,10 +256,22 @@ static int aldebaran_mode2_restore_ip(struct amdgpu_device *adev)
 		dev_err(adev->dev, "Failed to get BIF handle\n");
 		return -EINVAL;
 	}
-	r = cmn_block->version->funcs->resume(adev);
+	r = amdgpu_ip_block_resume(cmn_block);
 	if (r)
 		return r;
 
+	if (ip_block_mask & BIT(AMD_IP_BLOCK_TYPE_IH)) {
+		ih_block = amdgpu_device_ip_get_ip_block(adev,
+							 AMD_IP_BLOCK_TYPE_IH);
+		if (unlikely(!ih_block)) {
+			dev_err(adev->dev, "Failed to get IH handle\n");
+			return -EINVAL;
+		}
+		r = amdgpu_ip_block_resume(ih_block);
+		if (r)
+			return r;
+	}
+
 	/* Reinit GFXHUB */
 	adev->gfxhub.funcs->init(adev);
 	r = adev->gfxhub.funcs->gart_enable(adev);
@@ -285,15 +304,10 @@ static int aldebaran_mode2_restore_ip(struct amdgpu_device *adev)
 		      adev->ip_blocks[i].version->type ==
 			      AMD_IP_BLOCK_TYPE_SDMA))
 			continue;
-		r = adev->ip_blocks[i].version->funcs->resume(adev);
-		if (r) {
-			dev_err(adev->dev,
-				"resume of IP block <%s> failed %d\n",
-				adev->ip_blocks[i].version->funcs->name, r);
-			return r;
-		}
 
-		adev->ip_blocks[i].status.hw = true;
+		r = amdgpu_ip_block_resume(&adev->ip_blocks[i]);
+		if (r)
+			return r;
 	}
 
 	for (i = 0; i < adev->num_ip_blocks; i++) {
@@ -307,7 +321,7 @@ static int aldebaran_mode2_restore_ip(struct amdgpu_device *adev)
 
 		if (adev->ip_blocks[i].version->funcs->late_init) {
 			r = adev->ip_blocks[i].version->funcs->late_init(
-				(void *)adev);
+				&adev->ip_blocks[i]);
 			if (r) {
 				dev_err(adev->dev,
 					"late_init of IP block <%s> failed %d after reset\n",
@@ -319,8 +333,6 @@ static int aldebaran_mode2_restore_ip(struct amdgpu_device *adev)
 		adev->ip_blocks[i].status.late_initialized = true;
 	}
 
-	amdgpu_ras_set_error_query_ready(adev, true);
-
 	amdgpu_device_set_cg_state(adev, AMD_CG_STATE_GATE);
 	amdgpu_device_set_pg_state(adev, AMD_PG_STATE_GATE);
 
@@ -333,12 +345,13 @@ aldebaran_mode2_restore_hwcontext(struct amdgpu_reset_control *reset_ctl,
 {
 	struct list_head *reset_device_list = reset_context->reset_device_list;
 	struct amdgpu_device *tmp_adev = NULL;
+	struct amdgpu_ras *con;
 	int r;
 
 	if (reset_device_list == NULL)
 		return -EINVAL;
 
-	if (reset_context->reset_req_dev->ip_versions[MP1_HWIP][0] ==
+	if (amdgpu_ip_version(reset_context->reset_req_dev, MP1_HWIP, 0) ==
 		    IP_VERSION(13, 0, 2) &&
 	    reset_context->hive == NULL) {
 		/* Wrong context, return error */
@@ -346,8 +359,12 @@ aldebaran_mode2_restore_hwcontext(struct amdgpu_reset_control *reset_ctl,
 	}
 
 	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
+		amdgpu_set_init_level(tmp_adev,
+				AMDGPU_INIT_LEVEL_RESET_RECOVERY);
 		dev_info(tmp_adev->dev,
 			 "GPU reset succeeded, trying to resume\n");
+		/*TBD: Ideally should clear only GFX, SDMA blocks*/
+		amdgpu_ras_clear_err_state(tmp_adev);
 		r = aldebaran_mode2_restore_ip(tmp_adev);
 		if (r)
 			goto end;
@@ -358,7 +375,30 @@ aldebaran_mode2_restore_hwcontext(struct amdgpu_reset_control *reset_ctl,
 		 */
 		amdgpu_register_gpu_instance(tmp_adev);
 
-		/* Resume RAS */
+		/* Resume RAS, ecc_irq */
+		con = amdgpu_ras_get_context(tmp_adev);
+		if (!amdgpu_sriov_vf(tmp_adev) && con) {
+			if (tmp_adev->sdma.ras &&
+				tmp_adev->sdma.ras->ras_block.ras_late_init) {
+				r = tmp_adev->sdma.ras->ras_block.ras_late_init(tmp_adev,
+						&tmp_adev->sdma.ras->ras_block.ras_comm);
+				if (r) {
+					dev_err(tmp_adev->dev, "SDMA failed to execute ras_late_init! ret:%d\n", r);
+					goto end;
+				}
+			}
+
+			if (tmp_adev->gfx.ras &&
+				tmp_adev->gfx.ras->ras_block.ras_late_init) {
+				r = tmp_adev->gfx.ras->ras_block.ras_late_init(tmp_adev,
+						&tmp_adev->gfx.ras->ras_block.ras_comm);
+				if (r) {
+					dev_err(tmp_adev->dev, "GFX failed to execute ras_late_init! ret:%d\n", r);
+					goto end;
+				}
+			}
+		}
+
 		amdgpu_ras_resume(tmp_adev);
 
 		/* Update PSP FW topology after reset */
@@ -368,6 +408,8 @@ aldebaran_mode2_restore_hwcontext(struct amdgpu_reset_control *reset_ctl,
 							tmp_adev);
 
 		if (!r) {
+			amdgpu_set_init_level(tmp_adev,
+					      AMDGPU_INIT_LEVEL_DEFAULT);
 			amdgpu_irq_gpu_reset_resume_helper(tmp_adev);
 
 			r = amdgpu_ib_ring_tests(tmp_adev);
@@ -395,6 +437,12 @@ static struct amdgpu_reset_handler aldebaran_mode2_handler = {
 	.do_reset		= aldebaran_mode2_reset,
 };
 
+static struct amdgpu_reset_handler
+	*aldebaran_rst_handlers[AMDGPU_RESET_MAX_HANDLERS] = {
+		&aldebaran_mode2_handler,
+		&xgmi_reset_on_init_handler,
+	};
+
 int aldebaran_reset_init(struct amdgpu_device *adev)
 {
 	struct amdgpu_reset_control *reset_ctl;
@@ -408,10 +456,9 @@ int aldebaran_reset_init(struct amdgpu_device *adev)
 	reset_ctl->active_reset = AMD_RESET_METHOD_NONE;
 	reset_ctl->get_reset_handler = aldebaran_get_reset_handler;
 
-	INIT_LIST_HEAD(&reset_ctl->reset_handlers);
 	INIT_WORK(&reset_ctl->reset_work, reset_ctl->async_reset);
 	/* Only mode2 is handled through reset control now */
-	amdgpu_reset_add_handler(reset_ctl, &aldebaran_mode2_handler);
+	reset_ctl->reset_handlers = &aldebaran_rst_handlers;
 
 	adev->reset_cntl = reset_ctl;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 8cf2cc50b3de..9f9774f58ce1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -53,7 +53,6 @@
 
 #include <drm/ttm/ttm_bo.h>
 #include <drm/ttm/ttm_placement.h>
-#include <drm/ttm/ttm_execbuf_util.h>
 
 #include <drm/amdgpu_drm.h>
 #include <drm/drm_gem.h>
@@ -64,6 +63,7 @@
 #include "kgd_pp_interface.h"
 
 #include "amd_shared.h"
+#include "amdgpu_utils.h"
 #include "amdgpu_mode.h"
 #include "amdgpu_ih.h"
 #include "amdgpu_irq.h"
@@ -80,6 +80,8 @@
 #include "amdgpu_vce.h"
 #include "amdgpu_vcn.h"
 #include "amdgpu_jpeg.h"
+#include "amdgpu_vpe.h"
+#include "amdgpu_umsch_mm.h"
 #include "amdgpu_gmc.h"
 #include "amdgpu_gfx.h"
 #include "amdgpu_sdma.h"
@@ -106,27 +108,33 @@
 #include "amdgpu_smuio.h"
 #include "amdgpu_fdinfo.h"
 #include "amdgpu_mca.h"
+#include "amdgpu_aca.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_cper.h"
+#include "amdgpu_xcp.h"
+#include "amdgpu_seq64.h"
+#include "amdgpu_reg_state.h"
+#include "amdgpu_userq.h"
+#include "amdgpu_eviction_fence.h"
+#if defined(CONFIG_DRM_AMD_ISP)
+#include "amdgpu_isp.h"
+#endif
 
-#define MAX_GPU_INSTANCE		16
+#define MAX_GPU_INSTANCE		64
 
-struct amdgpu_gpu_instance
-{
+#define GFX_SLICE_PERIOD_MS		250
+
+struct amdgpu_gpu_instance {
 	struct amdgpu_device		*adev;
 	int				mgpu_fan_enabled;
 };
 
-struct amdgpu_mgpu_info
-{
+struct amdgpu_mgpu_info {
 	struct amdgpu_gpu_instance	gpu_ins[MAX_GPU_INSTANCE];
 	struct mutex			mutex;
 	uint32_t			num_gpu;
 	uint32_t			num_dgpu;
 	uint32_t			num_apu;
-
-	/* delayed reset_func for XGMI configuration if necessary */
-	struct delayed_work		delayed_reset_work;
-	bool				pending_reset;
 };
 
 enum amdgpu_ss {
@@ -136,8 +144,15 @@ enum amdgpu_ss {
 	AMDGPU_SS_DRV_UNLOAD
 };
 
-struct amdgpu_watchdog_timer
-{
+struct amdgpu_hwip_reg_entry {
+	u32		hwip;
+	u32		inst;
+	u32		seg;
+	u32		reg_offset;
+	const char	*reg_name;
+};
+
+struct amdgpu_watchdog_timer {
 	bool timeout_fatal_disable;
 	uint32_t period; /* maxCycles = (1 << period), the number of cycles before a timeout */
 };
@@ -185,7 +200,6 @@ extern char *amdgpu_disable_cu;
 extern char *amdgpu_virtual_display;
 extern uint amdgpu_pp_feature_mask;
 extern uint amdgpu_force_long_training;
-extern int amdgpu_job_hang_limit;
 extern int amdgpu_lbpw;
 extern int amdgpu_compute_multipipe;
 extern int amdgpu_gpu_recovery;
@@ -196,8 +210,9 @@ extern uint amdgpu_dc_feature_mask;
 extern uint amdgpu_freesync_vid_mode;
 extern uint amdgpu_dc_debug_mask;
 extern uint amdgpu_dc_visual_confirm;
-extern uint amdgpu_dm_abm_level;
+extern int amdgpu_dm_abm_level;
 extern int amdgpu_backlight;
+extern int amdgpu_damage_clips;
 extern struct amdgpu_mgpu_info mgpu_info;
 extern int amdgpu_ras_enable;
 extern uint amdgpu_ras_mask;
@@ -208,16 +223,21 @@ extern int amdgpu_async_gfx_ring;
 extern int amdgpu_mcbp;
 extern int amdgpu_discovery;
 extern int amdgpu_mes;
+extern int amdgpu_mes_log_enable;
 extern int amdgpu_mes_kiq;
+extern int amdgpu_uni_mes;
 extern int amdgpu_noretry;
 extern int amdgpu_force_asic_type;
 extern int amdgpu_smartshift_bias;
 extern int amdgpu_use_xgmi_p2p;
+extern int amdgpu_mtype_local;
+extern int amdgpu_enforce_isolation;
 #ifdef CONFIG_HSA_AMD
 extern int sched_policy;
 extern bool debug_evictions;
 extern bool no_system_mem_limit;
 extern int halt_if_hws_hang;
+extern uint amdgpu_svm_default_granularity;
 #else
 static const int __maybe_unused sched_policy = KFD_SCHED_POLICY_HWS;
 static const bool __maybe_unused debug_evictions; /* = false */
@@ -240,12 +260,22 @@ extern int amdgpu_cik_support;
 extern int amdgpu_num_kcq;
 
 #define AMDGPU_VCNFW_LOG_SIZE (32 * 1024)
+#define AMDGPU_UMSCHFW_LOG_SIZE (32 * 1024)
 extern int amdgpu_vcnfw_log;
 extern int amdgpu_sg_display;
+extern int amdgpu_umsch_mm;
+extern int amdgpu_seamless;
+extern int amdgpu_umsch_mm_fwlog;
+
+extern int amdgpu_user_partt_mode;
+extern int amdgpu_agp;
+extern int amdgpu_rebar;
+
+extern int amdgpu_wbrf;
+extern int amdgpu_user_queue;
 
 #define AMDGPU_VM_MAX_NUM_CTX			4096
 #define AMDGPU_SG_THRESHOLD			(256*1024*1024)
-#define AMDGPU_DEFAULT_GTT_SIZE_MB		3072ULL /* 3GB by default */
 #define AMDGPU_WAIT_IDLE_TIMEOUT_IN_MS	        3000
 #define AMDGPU_MAX_USEC_TIMEOUT			100000	/* 100 ms */
 #define AMDGPU_FENCE_JIFFIES_TIMEOUT		(HZ / 2)
@@ -275,6 +305,12 @@ extern int amdgpu_sg_display;
 #define AMDGPU_RESET_VCE			(1 << 13)
 #define AMDGPU_RESET_VCE1			(1 << 14)
 
+/* reset mask */
+#define AMDGPU_RESET_TYPE_FULL (1 << 0) /* full adapter reset, mode1/mode2/BACO/etc. */
+#define AMDGPU_RESET_TYPE_SOFT_RESET (1 << 1) /* IP level soft reset */
+#define AMDGPU_RESET_TYPE_PER_QUEUE (1 << 2) /* per queue */
+#define AMDGPU_RESET_TYPE_PER_PIPE (1 << 3) /* per pipe */
+
 /* max cursor sizes (in pixels) */
 #define CIK_CURSOR_WIDTH 128
 #define CIK_CURSOR_HEIGHT 128
@@ -283,6 +319,10 @@ extern int amdgpu_sg_display;
 #define AMDGPU_SMARTSHIFT_MAX_BIAS (100)
 #define AMDGPU_SMARTSHIFT_MIN_BIAS (-100)
 
+/* Extra time delay(in ms) to eliminate the influence of temperature momentary fluctuation */
+#define AMDGPU_SWCTF_EXTRA_DELAY		50
+
+struct amdgpu_xcp_mgr;
 struct amdgpu_device;
 struct amdgpu_irq_src;
 struct amdgpu_fpriv;
@@ -318,7 +358,6 @@ enum amdgpu_kiq_irq {
 	AMDGPU_CP_KIQ_IRQ_DRIVER0 = 0,
 	AMDGPU_CP_KIQ_IRQ_LAST
 };
-#define SRIOV_USEC_TIMEOUT  1200000 /* wait 12 * 100ms for SRIOV */
 #define MAX_KIQ_REG_WAIT       5000 /* in usecs, 5ms */
 #define MAX_KIQ_REG_BAILOUT_INTERVAL   5 /* in msecs, 5ms */
 #define MAX_KIQ_REG_TRY 1000
@@ -333,10 +372,15 @@ void amdgpu_device_ip_get_clockgating_state(struct amdgpu_device *adev,
 					    u64 *flags);
 int amdgpu_device_ip_wait_for_idle(struct amdgpu_device *adev,
 				   enum amd_ip_block_type block_type);
-bool amdgpu_device_ip_is_idle(struct amdgpu_device *adev,
+bool amdgpu_device_ip_is_hw(struct amdgpu_device *adev,
+			    enum amd_ip_block_type block_type);
+bool amdgpu_device_ip_is_valid(struct amdgpu_device *adev,
 			      enum amd_ip_block_type block_type);
+int amdgpu_ip_block_suspend(struct amdgpu_ip_block *ip_block);
 
-#define AMDGPU_MAX_IP_NUM 16
+int amdgpu_ip_block_resume(struct amdgpu_ip_block *ip_block);
+
+#define AMDGPU_MAX_IP_NUM AMD_IP_BLOCK_TYPE_NUM
 
 struct amdgpu_ip_block_status {
 	bool valid;
@@ -354,12 +398,10 @@ struct amdgpu_ip_block_version {
 	const struct amd_ip_funcs *funcs;
 };
 
-#define HW_REV(_Major, _Minor, _Rev) \
-	((((uint32_t) (_Major)) << 16) | ((uint32_t) (_Minor) << 8) | ((uint32_t) (_Rev)))
-
 struct amdgpu_ip_block {
 	struct amdgpu_ip_block_status status;
 	const struct amdgpu_ip_block_version *version;
+	struct amdgpu_device *adev;
 };
 
 int amdgpu_device_ip_block_version_cmp(struct amdgpu_device *adev,
@@ -380,6 +422,7 @@ bool amdgpu_get_bios(struct amdgpu_device *adev);
 bool amdgpu_read_bios(struct amdgpu_device *adev);
 bool amdgpu_soc15_read_bios_from_rom(struct amdgpu_device *adev,
 				     u8 *bios, u32 length_bytes);
+void amdgpu_bios_release(struct amdgpu_device *adev);
 /*
  * Clocks
  */
@@ -394,7 +437,6 @@ struct amdgpu_clock {
 	uint32_t default_mclk;
 	uint32_t default_sclk;
 	uint32_t default_dispclk;
-	uint32_t current_dispclk;
 	uint32_t dp_extclk;
 	uint32_t max_pixel_clock;
 };
@@ -430,9 +472,6 @@ struct amdgpu_sa_manager {
 	void				*cpu_ptr;
 };
 
-int amdgpu_fence_slab_init(void);
-void amdgpu_fence_slab_fini(void);
-
 /*
  * IRQS.
  */
@@ -452,7 +491,6 @@ struct amdgpu_flip_work {
 	bool				async;
 };
 
-
 /*
  * file private structure
  */
@@ -461,9 +499,17 @@ struct amdgpu_fpriv {
 	struct amdgpu_vm	vm;
 	struct amdgpu_bo_va	*prt_va;
 	struct amdgpu_bo_va	*csa_va;
+	struct amdgpu_bo_va	*seq64_va;
 	struct mutex		bo_list_lock;
 	struct idr		bo_list_handles;
 	struct amdgpu_ctx_mgr	ctx_mgr;
+	struct amdgpu_userq_mgr	userq_mgr;
+
+	/* Eviction fence infra */
+	struct amdgpu_eviction_fence_mgr evf_mgr;
+
+	/** GPU partition selection */
+	uint32_t		xcp_id;
 };
 
 int amdgpu_file_to_fpriv(struct file *filp, struct amdgpu_fpriv **fpriv);
@@ -471,14 +517,65 @@ int amdgpu_file_to_fpriv(struct file *filp, struct amdgpu_fpriv **fpriv);
 /*
  * Writeback
  */
-#define AMDGPU_MAX_WB 256	/* Reserve at most 256 WB slots for amdgpu-owned rings. */
+#define AMDGPU_MAX_WB 1024	/* Reserve at most 1024 WB slots for amdgpu-owned rings. */
 
+/**
+ * amdgpu_wb - This struct is used for small GPU memory allocation.
+ *
+ * This struct is used to allocate a small amount of GPU memory that can be
+ * used to shadow certain states into the memory. This is especially useful for
+ * providing easy CPU access to some states without requiring register access
+ * (e.g., if some block is power gated, reading register may be problematic).
+ *
+ * Note: the term writeback was initially used because many of the amdgpu
+ * components had some level of writeback memory, and this struct initially
+ * described those components.
+ */
 struct amdgpu_wb {
+
+	/**
+	 * @wb_obj:
+	 *
+	 * Buffer Object used for the writeback memory.
+	 */
 	struct amdgpu_bo	*wb_obj;
-	volatile uint32_t	*wb;
+
+	/**
+	 * @wb:
+	 *
+	 * Pointer to the first writeback slot. In terms of CPU address
+	 * this value can be accessed directly by using the offset as an index.
+	 * For the GPU address, it is necessary to use gpu_addr and the offset.
+	 */
+	uint32_t		*wb;
+
+	/**
+	 * @gpu_addr:
+	 *
+	 * Writeback base address in the GPU.
+	 */
 	uint64_t		gpu_addr;
-	u32			num_wb;	/* Number of wb slots actually reserved for amdgpu. */
+
+	/**
+	 * @num_wb:
+	 *
+	 * Number of writeback slots reserved for amdgpu.
+	 */
+	u32			num_wb;
+
+	/**
+	 * @used:
+	 *
+	 * Track the writeback slot already used.
+	 */
 	unsigned long		used[DIV_ROUND_UP(AMDGPU_MAX_WB, BITS_PER_LONG)];
+
+	/**
+	 * @lock:
+	 *
+	 * Protects read and write of the used field array.
+	 */
+	spinlock_t		lock;
 };
 
 int amdgpu_device_wb_get(struct amdgpu_device *adev, u32 *wb);
@@ -497,14 +594,42 @@ struct amdgpu_allowed_register_entry {
 	bool grbm_indexed;
 };
 
+/**
+ * enum amd_reset_method - Methods for resetting AMD GPU devices
+ *
+ * @AMD_RESET_METHOD_NONE: The device will not be reset.
+ * @AMD_RESET_LEGACY: Method reserved for SI, CIK and VI ASICs.
+ * @AMD_RESET_MODE0: Reset the entire ASIC. Not currently available for the
+ *                   any device.
+ * @AMD_RESET_MODE1: Resets all IP blocks on the ASIC (SDMA, GFX, VCN, etc.)
+ *                   individually. Suitable only for some discrete GPU, not
+ *                   available for all ASICs.
+ * @AMD_RESET_MODE2: Resets a lesser level of IPs compared to MODE1. Which IPs
+ *                   are reset depends on the ASIC. Notably doesn't reset IPs
+ *                   shared with the CPU on APUs or the memory controllers (so
+ *                   VRAM is not lost). Not available on all ASICs.
+ * @AMD_RESET_LINK: Triggers SW-UP link reset on other GPUs
+ * @AMD_RESET_BACO: BACO (Bus Alive, Chip Off) method powers off and on the card
+ *                  but without powering off the PCI bus. Suitable only for
+ *                  discrete GPUs.
+ * @AMD_RESET_PCI: Does a full bus reset using core Linux subsystem PCI reset
+ *                 and does a secondary bus reset or FLR, depending on what the
+ *                 underlying hardware supports.
+ *
+ * Methods available for AMD GPU driver for resetting the device. Not all
+ * methods are suitable for every device. User can override the method using
+ * module parameter `reset_method`.
+ */
 enum amd_reset_method {
 	AMD_RESET_METHOD_NONE = -1,
 	AMD_RESET_METHOD_LEGACY = 0,
 	AMD_RESET_METHOD_MODE0,
 	AMD_RESET_METHOD_MODE1,
 	AMD_RESET_METHOD_MODE2,
+	AMD_RESET_METHOD_LINK,
 	AMD_RESET_METHOD_BACO,
 	AMD_RESET_METHOD_PCI,
+	AMD_RESET_METHOD_ON_INIT,
 };
 
 struct amdgpu_video_codec_info {
@@ -566,7 +691,7 @@ struct amdgpu_asic_funcs {
 	/* PCIe replay counter */
 	uint64_t (*get_pcie_replay_count)(struct amdgpu_device *adev);
 	/* device supports BACO */
-	bool (*supports_baco)(struct amdgpu_device *adev);
+	int (*supports_baco)(struct amdgpu_device *adev);
 	/* pre asic_init quirks */
 	void (*pre_asic_init)(struct amdgpu_device *adev);
 	/* enter/exit umd stable pstate */
@@ -574,6 +699,12 @@ struct amdgpu_asic_funcs {
 	/* query video codecs */
 	int (*query_video_codecs)(struct amdgpu_device *adev, bool encode,
 				  const struct amdgpu_video_codecs **codecs);
+	/* encode "> 32bits" smn addressing */
+	u64 (*encode_ext_smn_addressing)(int ext_id);
+
+	ssize_t (*get_reg_state)(struct amdgpu_device *adev,
+				 enum amdgpu_reg_state reg_state, void *buf,
+				 size_t max_size);
 };
 
 /*
@@ -592,7 +723,7 @@ int amdgpu_cs_wait_fences_ioctl(struct drm_device *dev, void *data,
 /* VRAM scratch page for HDP bug, default vram page */
 struct amdgpu_mem_scratch {
 	struct amdgpu_bo		*robj;
-	volatile uint32_t		*ptr;
+	uint32_t			*ptr;
 	u64				gpu_addr;
 };
 
@@ -608,15 +739,22 @@ void amdgpu_cgs_destroy_device(struct cgs_device *cgs_device);
 typedef uint32_t (*amdgpu_rreg_t)(struct amdgpu_device*, uint32_t);
 typedef void (*amdgpu_wreg_t)(struct amdgpu_device*, uint32_t, uint32_t);
 
+typedef uint32_t (*amdgpu_rreg_ext_t)(struct amdgpu_device*, uint64_t);
+typedef void (*amdgpu_wreg_ext_t)(struct amdgpu_device*, uint64_t, uint32_t);
+
 typedef uint64_t (*amdgpu_rreg64_t)(struct amdgpu_device*, uint32_t);
 typedef void (*amdgpu_wreg64_t)(struct amdgpu_device*, uint32_t, uint64_t);
 
+typedef uint64_t (*amdgpu_rreg64_ext_t)(struct amdgpu_device*, uint64_t);
+typedef void (*amdgpu_wreg64_ext_t)(struct amdgpu_device*, uint64_t, uint64_t);
+
 typedef uint32_t (*amdgpu_block_rreg_t)(struct amdgpu_device*, uint32_t, uint32_t);
 typedef void (*amdgpu_block_wreg_t)(struct amdgpu_device*, uint32_t, uint32_t, uint32_t);
 
 struct amdgpu_mmio_remap {
 	u32 reg_offset;
 	resource_size_t bus_addr;
+	struct amdgpu_bo *bo;
 };
 
 /* Define the HW IP blocks will be used in driver , add more if necessary */
@@ -642,6 +780,7 @@ enum amd_hw_ip_block_type {
 	JPEG_HWIP = VCN_HWIP,
 	VCN1_HWIP,
 	VCE_HWIP,
+	VPE_HWIP,
 	DF_HWIP,
 	DCE_HWIP,
 	OSSSYS_HWIP,
@@ -655,24 +794,53 @@ enum amd_hw_ip_block_type {
 	XGMI_HWIP,
 	DCI_HWIP,
 	PCIE_HWIP,
+	ISP_HWIP,
 	MAX_HWIP
 };
 
-#define HWIP_MAX_INSTANCE	28
+#define HWIP_MAX_INSTANCE	44
 
 #define HW_ID_MAX		300
-#define IP_VERSION(mj, mn, rv) (((mj) << 16) | ((mn) << 8) | (rv))
-#define IP_VERSION_MAJ(ver) ((ver) >> 16)
-#define IP_VERSION_MIN(ver) (((ver) >> 8) & 0xFF)
-#define IP_VERSION_REV(ver) ((ver) & 0xFF)
+#define IP_VERSION_FULL(mj, mn, rv, var, srev) \
+	(((mj) << 24) | ((mn) << 16) | ((rv) << 8) | ((var) << 4) | (srev))
+#define IP_VERSION(mj, mn, rv)		IP_VERSION_FULL(mj, mn, rv, 0, 0)
+#define IP_VERSION_MAJ(ver)		((ver) >> 24)
+#define IP_VERSION_MIN(ver)		(((ver) >> 16) & 0xFF)
+#define IP_VERSION_REV(ver)		(((ver) >> 8) & 0xFF)
+#define IP_VERSION_VARIANT(ver)		(((ver) >> 4) & 0xF)
+#define IP_VERSION_SUBREV(ver)		((ver) & 0xF)
+#define IP_VERSION_MAJ_MIN_REV(ver)	((ver) >> 8)
+
+struct amdgpu_ip_map_info {
+	/* Map of logical to actual dev instances/mask */
+	uint32_t 		dev_inst[MAX_HWIP][HWIP_MAX_INSTANCE];
+	int8_t (*logical_to_dev_inst)(struct amdgpu_device *adev,
+				      enum amd_hw_ip_block_type block,
+				      int8_t inst);
+	uint32_t (*logical_to_dev_mask)(struct amdgpu_device *adev,
+					enum amd_hw_ip_block_type block,
+					uint32_t mask);
+};
+
+enum amdgpu_uid_type {
+	AMDGPU_UID_TYPE_XCD,
+	AMDGPU_UID_TYPE_AID,
+	AMDGPU_UID_TYPE_SOC,
+	AMDGPU_UID_TYPE_MAX
+};
+
+#define AMDGPU_UID_INST_MAX 8 /* max number of instances for each UID type */
+
+struct amdgpu_uid {
+	uint64_t uid[AMDGPU_UID_TYPE_MAX][AMDGPU_UID_INST_MAX];
+	struct amdgpu_device *adev;
+};
 
 struct amd_powerplay {
 	void *pp_handle;
 	const struct amd_pm_funcs *pp_funcs;
 };
 
-struct ip_discovery_top;
-
 /* polaris10 kickers */
 #define ASICID_IS_P20(did, rid)		(((did == 0x67DF) && \
 					 ((rid == 0xE3) || \
@@ -723,7 +891,14 @@ struct amdgpu_mqd_prop {
 	uint64_t eop_gpu_addr;
 	uint32_t hqd_pipe_priority;
 	uint32_t hqd_queue_priority;
+	bool allow_tunneling;
 	bool hqd_active;
+	uint64_t shadow_addr;
+	uint64_t gds_bkup_addr;
+	uint64_t csa_addr;
+	uint64_t fence_address;
+	bool tmz_queue;
+	bool kernel_queue;
 };
 
 struct amdgpu_mqd {
@@ -732,15 +907,45 @@ struct amdgpu_mqd {
 			struct amdgpu_mqd_prop *p);
 };
 
+struct amdgpu_pcie_reset_ctx {
+	bool in_link_reset;
+	bool occurs_dpc;
+	bool audio_suspended;
+	struct pci_dev *swus;
+	struct pci_saved_state *swus_pcistate;
+	struct pci_saved_state *swds_pcistate;
+};
+
+/*
+ * Custom Init levels could be defined for different situations where a full
+ * initialization of all hardware blocks are not expected. Sample cases are
+ * custom init sequences after resume after S0i3/S3, reset on initialization,
+ * partial reset of blocks etc. Presently, this defines only two levels. Levels
+ * are described in corresponding struct definitions - amdgpu_init_default,
+ * amdgpu_init_minimal_xgmi.
+ */
+enum amdgpu_init_lvl_id {
+	AMDGPU_INIT_LEVEL_DEFAULT,
+	AMDGPU_INIT_LEVEL_MINIMAL_XGMI,
+	AMDGPU_INIT_LEVEL_RESET_RECOVERY,
+};
+
+struct amdgpu_init_level {
+	enum amdgpu_init_lvl_id level;
+	uint32_t hwini_ip_block_mask;
+};
+
 #define AMDGPU_RESET_MAGIC_NUM 64
 #define AMDGPU_MAX_DF_PERFMONS 4
-#define AMDGPU_PRODUCT_NAME_LEN 64
 struct amdgpu_reset_domain;
+struct amdgpu_fru_info;
 
-/*
- * Non-zero (true) if the GPU has VRAM. Zero (false) otherwise.
- */
-#define AMDGPU_HAS_VRAM(_adev) ((_adev)->gmc.real_vram_size)
+enum amdgpu_enforce_isolation_mode {
+	AMDGPU_ENFORCE_ISOLATION_DISABLE = 0,
+	AMDGPU_ENFORCE_ISOLATION_ENABLE = 1,
+	AMDGPU_ENFORCE_ISOLATION_ENABLE_LEGACY = 2,
+	AMDGPU_ENFORCE_ISOLATION_NO_CLEANER_SHADER = 3,
+};
 
 struct amdgpu_device {
 	struct device			*dev;
@@ -751,6 +956,7 @@ struct amdgpu_device {
 	struct amdgpu_acp		acp;
 #endif
 	struct amdgpu_hive_info *hive;
+	struct amdgpu_xcp_mgr *xcp_mgr;
 	/* ASIC */
 	enum amd_asic_type		asic_type;
 	uint32_t			family;
@@ -764,9 +970,9 @@ struct amdgpu_device {
 	bool				need_swiotlb;
 	bool				accel_working;
 	struct notifier_block		acpi_nb;
+	struct notifier_block		pm_nb;
 	struct amdgpu_i2c_chan		*i2c_bus[AMDGPU_MAX_I2C_BUS];
-	struct debugfs_blob_wrapper     debugfs_vbios_blob;
-	struct debugfs_blob_wrapper     debugfs_discovery_blob;
+	struct debugfs_blob_wrapper debugfs_vbios_blob;
 	struct mutex			srbm_mutex;
 	/* GRBM index mutex. Protects concurrent access to GRBM index */
 	struct mutex                    grbm_idx_mutex;
@@ -798,8 +1004,12 @@ struct amdgpu_device {
 	amdgpu_wreg_t			pcie_wreg;
 	amdgpu_rreg_t			pciep_rreg;
 	amdgpu_wreg_t			pciep_wreg;
+	amdgpu_rreg_ext_t		pcie_rreg_ext;
+	amdgpu_wreg_ext_t		pcie_wreg_ext;
 	amdgpu_rreg64_t			pcie_rreg64;
 	amdgpu_wreg64_t			pcie_wreg64;
+	amdgpu_rreg64_ext_t			pcie_rreg64_ext;
+	amdgpu_wreg64_ext_t			pcie_wreg64_ext;
 	/* protects concurrent UVD register access */
 	spinlock_t uvd_ctx_idx_lock;
 	amdgpu_rreg_t			uvd_ctx_rreg;
@@ -831,7 +1041,7 @@ struct amdgpu_device {
 	dma_addr_t			dummy_page_addr;
 	struct amdgpu_vm_manager	vm_manager;
 	struct amdgpu_vmhub             vmhub[AMDGPU_MAX_VMHUBS];
-	unsigned			num_vmhubs;
+	DECLARE_BITMAP(vmhubs_mask, AMDGPU_MAX_VMHUBS);
 
 	/* memory management */
 	struct amdgpu_mman		mman;
@@ -852,6 +1062,9 @@ struct amdgpu_device {
 		u32			log2_max_MBps;
 	} mm_stats;
 
+	/* discovery*/
+	struct amdgpu_discovery_info discovery;
+
 	/* display */
 	bool				enable_virtual_display;
 	struct amdgpu_vkms_output       *amdgpu_vkms_output;
@@ -920,6 +1133,13 @@ struct amdgpu_device {
 	/* jpeg */
 	struct amdgpu_jpeg		jpeg;
 
+	/* vpe */
+	struct amdgpu_vpe		vpe;
+
+	/* umsch */
+	struct amdgpu_umsch_mm		umsch_mm;
+	bool				enable_umsch_mm;
+
 	/* firmwares */
 	struct amdgpu_firmware		firmware;
 
@@ -929,8 +1149,8 @@ struct amdgpu_device {
 	/* GDS */
 	struct amdgpu_gds		gds;
 
-	/* KFD */
-	struct amdgpu_kfd_dev		kfd;
+	/* for userq and VM fences */
+	struct amdgpu_seq64		seq64;
 
 	/* UMC */
 	struct amdgpu_umc		umc;
@@ -938,11 +1158,30 @@ struct amdgpu_device {
 	/* display related functionality */
 	struct amdgpu_display_manager dm;
 
+#if defined(CONFIG_DRM_AMD_ISP)
+	/* isp */
+	struct amdgpu_isp		isp;
+#endif
+
 	/* mes */
 	bool                            enable_mes;
 	bool                            enable_mes_kiq;
+	bool                            enable_uni_mes;
 	struct amdgpu_mes               mes;
 	struct amdgpu_mqd               mqds[AMDGPU_HW_IP_NUM];
+	const struct amdgpu_userq_funcs *userq_funcs[AMDGPU_HW_IP_NUM];
+
+	/* xarray used to retrieve the user queue fence driver reference
+	 * in the EOP interrupt handler to signal the particular user
+	 * queue fence.
+	 */
+	struct xarray			userq_xa;
+	/**
+	 * @userq_doorbell_xa: Global user queue map (doorbell index → queue)
+	 * Key: doorbell_index (unique global identifier for the queue)
+	 * Value: struct amdgpu_usermode_queue
+	 */
+	struct xarray userq_doorbell_xa;
 
 	/* df */
 	struct amdgpu_df                df;
@@ -950,6 +1189,12 @@ struct amdgpu_device {
 	/* MCA */
 	struct amdgpu_mca               mca;
 
+	/* ACA */
+	struct amdgpu_aca		aca;
+
+	/* CPER */
+	struct amdgpu_cper		cper;
+
 	struct amdgpu_ip_block          ip_blocks[AMDGPU_MAX_IP_NUM];
 	uint32_t		        harvest_ip_mask;
 	int				num_ip_blocks;
@@ -963,16 +1208,13 @@ struct amdgpu_device {
 
 	/* soc15 register offset based on ip, instance and  segment */
 	uint32_t		*reg_offset[MAX_HWIP][HWIP_MAX_INSTANCE];
+	struct amdgpu_ip_map_info	ip_map;
 
 	/* delayed work_func for deferring clockgating during resume */
 	struct delayed_work     delayed_init_work;
 
 	struct amdgpu_virt	virt;
 
-	/* link all shadow bo */
-	struct list_head                shadow_list;
-	struct mutex                    shadow_list_lock;
-
 	/* record hw reset is performed */
 	bool has_hw_reset;
 	u8				reset_magic[AMDGPU_RESET_MAGIC_NUM];
@@ -982,6 +1224,7 @@ struct amdgpu_device {
 	bool				in_s3;
 	bool				in_s4;
 	bool				in_s0ix;
+	suspend_state_t			last_suspend_state;
 
 	enum pp_mp1_state               mp1_state;
 	struct amdgpu_doorbell_index doorbell_index;
@@ -996,6 +1239,7 @@ struct amdgpu_device {
 	long				sdma_timeout;
 	long				video_timeout;
 	long				compute_timeout;
+	long				psp_timeout;
 
 	uint64_t			unique_id;
 	uint64_t	df_perfmon_config_assign_mask[AMDGPU_MAX_DF_PERFMONS];
@@ -1005,22 +1249,23 @@ struct amdgpu_device {
 	bool                            has_pr3;
 
 	bool                            ucode_sysfs_en;
-	bool                            psp_sysfs_en;
-
-	/* Chip product information */
-	char				product_number[20];
-	char				product_name[AMDGPU_PRODUCT_NAME_LEN];
-	char				serial[20];
 
+	struct amdgpu_fru_info		*fru_info;
 	atomic_t			throttling_logging_enabled;
 	struct ratelimit_state		throttling_logging_rs;
 	uint32_t                        ras_hw_enabled;
 	uint32_t                        ras_enabled;
+	bool                            ras_default_ecc_enabled;
 
 	bool                            no_hw_access;
 	struct pci_saved_state          *pci_state;
 	pci_channel_state_t		pci_channel_state;
 
+	struct amdgpu_pcie_reset_ctx	pcie_reset_ctx;
+
+	/* Track auto wait count on s_barrier settings */
+	bool				barrier_has_auto_waitcnt;
+
 	struct amdgpu_reset_control     *reset_cntl;
 	uint32_t                        ip_versions[MAX_HWIP][HWIP_MAX_INSTANCE];
 
@@ -1028,31 +1273,74 @@ struct amdgpu_device {
 
 	struct list_head                ras_list;
 
-	struct ip_discovery_top         *ip_top;
-
 	struct amdgpu_reset_domain	*reset_domain;
 
 	struct mutex			benchmark_mutex;
 
-	/* reset dump register */
-	uint32_t                        *reset_dump_reg_list;
-	uint32_t			*reset_dump_reg_value;
-	int                             num_regs;
-#ifdef CONFIG_DEV_COREDUMP
-	struct amdgpu_task_info         reset_task_info;
-	bool                            reset_vram_lost;
-	struct timespec64               reset_time;
-#endif
-
 	bool                            scpm_enabled;
 	uint32_t                        scpm_status;
 
 	struct work_struct		reset_work;
 
-	bool                            job_hang;
 	bool                            dc_enabled;
+	/* Mask of active clusters */
+	uint32_t			aid_mask;
+
+	/* Debug */
+	bool                            debug_vm;
+	bool                            debug_largebar;
+	bool                            debug_disable_soft_recovery;
+	bool                            debug_use_vram_fw_buf;
+	bool                            debug_enable_ras_aca;
+	bool                            debug_exp_resets;
+	bool                            debug_disable_gpu_ring_reset;
+	bool                            debug_vm_userptr;
+	bool                            debug_disable_ce_logs;
+	bool                            debug_enable_ce_cs;
+
+	/* Protection for the following isolation structure */
+	struct mutex                    enforce_isolation_mutex;
+	enum amdgpu_enforce_isolation_mode	enforce_isolation[MAX_XCP];
+	struct amdgpu_isolation {
+		void			*owner;
+		struct dma_fence	*spearhead;
+		struct amdgpu_sync	active;
+		struct amdgpu_sync	prev;
+	} isolation[MAX_XCP];
+
+	struct amdgpu_init_level *init_lvl;
+
+	/* This flag is used to determine how VRAM allocations are handled for APUs
+	 * in KFD: VRAM or GTT.
+	 */
+	bool                            apu_prefer_gtt;
+
+	bool                            userq_halt_for_enforce_isolation;
+	struct work_struct              userq_reset_work;
+	struct amdgpu_uid *uid_info;
+
+	/* KFD
+	 * Must be last --ends in a flexible-array member.
+	 */
+	struct amdgpu_kfd_dev		kfd;
 };
 
+static inline uint32_t amdgpu_ip_version(const struct amdgpu_device *adev,
+					 uint8_t ip, uint8_t inst)
+{
+	/* This considers only major/minor/rev and ignores
+	 * subrevision/variant fields.
+	 */
+	return adev->ip_versions[ip][inst] & ~0xFFU;
+}
+
+static inline uint32_t amdgpu_ip_version_full(const struct amdgpu_device *adev,
+					      uint8_t ip, uint8_t inst)
+{
+	/* This returns full version - major/minor/rev/variant/subrevision */
+	return adev->ip_versions[ip][inst];
+}
+
 static inline struct amdgpu_device *drm_to_adev(struct drm_device *ddev)
 {
 	return container_of(ddev, struct amdgpu_device, ddev);
@@ -1068,6 +1356,11 @@ static inline struct amdgpu_device *amdgpu_ttm_adev(struct ttm_device *bdev)
 	return container_of(bdev, struct amdgpu_device, mman.bdev);
 }
 
+static inline bool amdgpu_is_multi_aid(struct amdgpu_device *adev)
+{
+	return !!adev->aid_mask;
+}
+
 int amdgpu_device_init(struct amdgpu_device *adev,
 		       uint32_t flags);
 void amdgpu_device_fini_hw(struct amdgpu_device *adev);
@@ -1082,13 +1375,27 @@ size_t amdgpu_device_aper_access(struct amdgpu_device *adev, loff_t pos,
 
 void amdgpu_device_vram_access(struct amdgpu_device *adev, loff_t pos,
 			       void *buf, size_t size, bool write);
+uint32_t amdgpu_device_wait_on_rreg(struct amdgpu_device *adev,
+			    uint32_t inst, uint32_t reg_addr, char reg_name[],
+			    uint32_t expected_value, uint32_t mask);
 uint32_t amdgpu_device_rreg(struct amdgpu_device *adev,
 			    uint32_t reg, uint32_t acc_flags);
+u32 amdgpu_device_indirect_rreg_ext(struct amdgpu_device *adev,
+				    u64 reg_addr);
+uint32_t amdgpu_device_xcc_rreg(struct amdgpu_device *adev,
+				uint32_t reg, uint32_t acc_flags,
+				uint32_t xcc_id);
 void amdgpu_device_wreg(struct amdgpu_device *adev,
 			uint32_t reg, uint32_t v,
 			uint32_t acc_flags);
+void amdgpu_device_indirect_wreg_ext(struct amdgpu_device *adev,
+				     u64 reg_addr, u32 reg_data);
+void amdgpu_device_xcc_wreg(struct amdgpu_device *adev,
+			    uint32_t reg, uint32_t v,
+			    uint32_t acc_flags,
+			    uint32_t xcc_id);
 void amdgpu_mm_wreg_mmio_rlc(struct amdgpu_device *adev,
-			     uint32_t reg, uint32_t v);
+			     uint32_t reg, uint32_t v, uint32_t xcc_id);
 void amdgpu_mm_wreg8(struct amdgpu_device *adev, uint32_t offset, uint8_t value);
 uint8_t amdgpu_mm_rreg8(struct amdgpu_device *adev, uint32_t offset);
 
@@ -1096,12 +1403,17 @@ u32 amdgpu_device_indirect_rreg(struct amdgpu_device *adev,
 				u32 reg_addr);
 u64 amdgpu_device_indirect_rreg64(struct amdgpu_device *adev,
 				  u32 reg_addr);
+u64 amdgpu_device_indirect_rreg64_ext(struct amdgpu_device *adev,
+				  u64 reg_addr);
 void amdgpu_device_indirect_wreg(struct amdgpu_device *adev,
 				 u32 reg_addr, u32 reg_data);
 void amdgpu_device_indirect_wreg64(struct amdgpu_device *adev,
 				   u32 reg_addr, u64 reg_data);
+void amdgpu_device_indirect_wreg64_ext(struct amdgpu_device *adev,
+				   u64 reg_addr, u64 reg_data);
 u32 amdgpu_device_get_rev_id(struct amdgpu_device *adev);
-bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type);
+bool amdgpu_device_asic_has_dc_support(struct pci_dev *pdev,
+				       enum amd_asic_type asic_type);
 bool amdgpu_device_has_dc_support(struct amdgpu_device *adev);
 
 void amdgpu_device_set_sriov_virtual_display(struct amdgpu_device *adev);
@@ -1112,6 +1424,8 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 			 struct amdgpu_reset_context *reset_context);
 
+int amdgpu_device_reinit_after_reset(struct amdgpu_reset_context *reset_context);
+
 int emu_soc_asic_init(struct amdgpu_device *adev);
 
 /*
@@ -1123,8 +1437,8 @@ int emu_soc_asic_init(struct amdgpu_device *adev);
 #define RREG32_NO_KIQ(reg) amdgpu_device_rreg(adev, (reg), AMDGPU_REGS_NO_KIQ)
 #define WREG32_NO_KIQ(reg, v) amdgpu_device_wreg(adev, (reg), (v), AMDGPU_REGS_NO_KIQ)
 
-#define RREG32_KIQ(reg) amdgpu_kiq_rreg(adev, (reg))
-#define WREG32_KIQ(reg, v) amdgpu_kiq_wreg(adev, (reg), (v))
+#define RREG32_KIQ(reg) amdgpu_kiq_rreg(adev, (reg), 0)
+#define WREG32_KIQ(reg, v) amdgpu_kiq_wreg(adev, (reg), (v), 0)
 
 #define RREG8(reg) amdgpu_mm_rreg8(adev, (reg))
 #define WREG8(reg, v) amdgpu_mm_wreg8(adev, (reg), (v))
@@ -1134,12 +1448,18 @@ int emu_soc_asic_init(struct amdgpu_device *adev);
 #define WREG32(reg, v) amdgpu_device_wreg(adev, (reg), (v), 0)
 #define REG_SET(FIELD, v) (((v) << FIELD##_SHIFT) & FIELD##_MASK)
 #define REG_GET(FIELD, v) (((v) << FIELD##_SHIFT) & FIELD##_MASK)
+#define RREG32_XCC(reg, inst) amdgpu_device_xcc_rreg(adev, (reg), 0, inst)
+#define WREG32_XCC(reg, v, inst) amdgpu_device_xcc_wreg(adev, (reg), (v), 0, inst)
 #define RREG32_PCIE(reg) adev->pcie_rreg(adev, (reg))
 #define WREG32_PCIE(reg, v) adev->pcie_wreg(adev, (reg), (v))
 #define RREG32_PCIE_PORT(reg) adev->pciep_rreg(adev, (reg))
 #define WREG32_PCIE_PORT(reg, v) adev->pciep_wreg(adev, (reg), (v))
+#define RREG32_PCIE_EXT(reg) adev->pcie_rreg_ext(adev, (reg))
+#define WREG32_PCIE_EXT(reg, v) adev->pcie_wreg_ext(adev, (reg), (v))
 #define RREG64_PCIE(reg) adev->pcie_rreg64(adev, (reg))
 #define WREG64_PCIE(reg, v) adev->pcie_wreg64(adev, (reg), (v))
+#define RREG64_PCIE_EXT(reg) adev->pcie_rreg64_ext(adev, (reg))
+#define WREG64_PCIE_EXT(reg, v) adev->pcie_wreg64_ext(adev, (reg), (v))
 #define RREG32_SMC(reg) adev->smc_rreg(adev, (reg))
 #define WREG32_SMC(reg, v) adev->smc_wreg(adev, (reg), (v))
 #define RREG32_UVD_CTX(reg) adev->uvd_ctx_rreg(adev, (reg))
@@ -1195,6 +1515,7 @@ int emu_soc_asic_init(struct amdgpu_device *adev);
 #define WREG32_FIELD_OFFSET(reg, offset, field, val)	\
 	WREG32(mm##reg + offset, (RREG32(mm##reg + offset) & ~REG_FIELD_MASK(reg, field)) | (val) << REG_FIELD_SHIFT(reg, field))
 
+#define AMDGPU_GET_REG_FIELD(x, h, l) (((x) & GENMASK_ULL(h, l)) >> (l))
 /*
  * BIOS helpers.
  */
@@ -1205,7 +1526,8 @@ int emu_soc_asic_init(struct amdgpu_device *adev);
 /*
  * ASICs macro.
  */
-#define amdgpu_asic_set_vga_state(adev, state) (adev)->asic_funcs->set_vga_state((adev), (state))
+#define amdgpu_asic_set_vga_state(adev, state) \
+    ((adev)->asic_funcs->set_vga_state ? (adev)->asic_funcs->set_vga_state((adev), (state)) : 0)
 #define amdgpu_asic_reset(adev) (adev)->asic_funcs->reset((adev))
 #define amdgpu_asic_reset_method(adev) (adev)->asic_funcs->reset_method((adev))
 #define amdgpu_asic_get_xclk(adev) (adev)->asic_funcs->get_xclk((adev))
@@ -1218,11 +1540,6 @@ int emu_soc_asic_init(struct amdgpu_device *adev);
 #define amdgpu_asic_read_bios_from_rom(adev, b, l) (adev)->asic_funcs->read_bios_from_rom((adev), (b), (l))
 #define amdgpu_asic_read_register(adev, se, sh, offset, v)((adev)->asic_funcs->read_register((adev), (se), (sh), (offset), (v)))
 #define amdgpu_asic_get_config_memsize(adev) (adev)->asic_funcs->get_config_memsize((adev))
-#define amdgpu_asic_flush_hdp(adev, r) \
-	((adev)->asic_funcs->flush_hdp ? (adev)->asic_funcs->flush_hdp((adev), (r)) : (adev)->hdp.funcs->flush_hdp((adev), (r)))
-#define amdgpu_asic_invalidate_hdp(adev, r) \
-	((adev)->asic_funcs->invalidate_hdp ? (adev)->asic_funcs->invalidate_hdp((adev), (r)) : \
-	 ((adev)->hdp.funcs->invalidate_hdp ? (adev)->hdp.funcs->invalidate_hdp((adev), (r)) : 0))
 #define amdgpu_asic_need_full_reset(adev) (adev)->asic_funcs->need_full_reset((adev))
 #define amdgpu_asic_init_doorbell_index(adev) (adev)->asic_funcs->init_doorbell_index((adev))
 #define amdgpu_asic_get_pcie_usage(adev, cnt0, cnt1) ((adev)->asic_funcs->get_pcie_usage((adev), (cnt0), (cnt1)))
@@ -1234,9 +1551,12 @@ int emu_soc_asic_init(struct amdgpu_device *adev);
 	((adev)->asic_funcs->update_umd_stable_pstate ? (adev)->asic_funcs->update_umd_stable_pstate((adev), (enter)) : 0)
 #define amdgpu_asic_query_video_codecs(adev, e, c) (adev)->asic_funcs->query_video_codecs((adev), (e), (c))
 
-#define amdgpu_inc_vram_lost(adev) atomic_inc(&((adev)->vram_lost_counter));
+#define amdgpu_inc_vram_lost(adev) atomic_inc(&((adev)->vram_lost_counter))
 
-#define MIN(X, Y) ((X) < (Y) ? (X) : (Y))
+#define BIT_MASK_UPPER(i) ((i) >= BITS_PER_LONG ? 0 : ~0UL << (i))
+#define for_each_inst(i, inst_mask)        \
+	for (i = ffs(inst_mask); i-- != 0; \
+	     i = ffs(inst_mask & BIT_MASK_UPPER(i + 1)))
 
 /* Common functions */
 bool amdgpu_device_has_job_running(struct amdgpu_device *adev);
@@ -1247,8 +1567,8 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 void amdgpu_device_pci_config_reset(struct amdgpu_device *adev);
 int amdgpu_device_pci_reset(struct amdgpu_device *adev);
 bool amdgpu_device_need_post(struct amdgpu_device *adev);
+bool amdgpu_device_seamless_boot_supported(struct amdgpu_device *adev);
 bool amdgpu_device_should_use_aspm(struct amdgpu_device *adev);
-bool amdgpu_device_aspm_support_quirk(void);
 
 void amdgpu_cs_report_moved_bytes(struct amdgpu_device *adev, u64 num_bytes,
 				  u64 num_vis_bytes);
@@ -1258,15 +1578,17 @@ void amdgpu_device_program_register_sequence(struct amdgpu_device *adev,
 					     const u32 array_size);
 
 int amdgpu_device_mode1_reset(struct amdgpu_device *adev);
-bool amdgpu_device_supports_atpx(struct drm_device *dev);
-bool amdgpu_device_supports_px(struct drm_device *dev);
-bool amdgpu_device_supports_boco(struct drm_device *dev);
-bool amdgpu_device_supports_smart_shift(struct drm_device *dev);
-bool amdgpu_device_supports_baco(struct drm_device *dev);
+int amdgpu_device_link_reset(struct amdgpu_device *adev);
+bool amdgpu_device_supports_atpx(struct amdgpu_device *adev);
+bool amdgpu_device_supports_px(struct amdgpu_device *adev);
+bool amdgpu_device_supports_boco(struct amdgpu_device *adev);
+bool amdgpu_device_supports_smart_shift(struct amdgpu_device *adev);
+int amdgpu_device_supports_baco(struct amdgpu_device *adev);
+void amdgpu_device_detect_runtime_pm_mode(struct amdgpu_device *adev);
 bool amdgpu_device_is_peer_accessible(struct amdgpu_device *adev,
 				      struct amdgpu_device *peer_adev);
-int amdgpu_device_baco_enter(struct drm_device *dev);
-int amdgpu_device_baco_exit(struct drm_device *dev);
+int amdgpu_device_baco_enter(struct amdgpu_device *adev);
+int amdgpu_device_baco_exit(struct amdgpu_device *adev);
 
 void amdgpu_device_flush_hdp(struct amdgpu_device *adev,
 		struct amdgpu_ring *ring);
@@ -1278,9 +1600,15 @@ u32 amdgpu_device_pcie_port_rreg(struct amdgpu_device *adev,
 				u32 reg);
 void amdgpu_device_pcie_port_wreg(struct amdgpu_device *adev,
 				u32 reg, u32 v);
+struct dma_fence *amdgpu_device_get_gang(struct amdgpu_device *adev);
 struct dma_fence *amdgpu_device_switch_gang(struct amdgpu_device *adev,
 					    struct dma_fence *gang);
+struct dma_fence *amdgpu_device_enforce_isolation(struct amdgpu_device *adev,
+						  struct amdgpu_ring *ring,
+						  struct amdgpu_job *job);
 bool amdgpu_device_has_display_hardware(struct amdgpu_device *adev);
+ssize_t amdgpu_get_soft_full_reset_mask(struct amdgpu_ring *ring);
+ssize_t amdgpu_show_reset_mask(char *buf, uint32_t supported_reset);
 
 /* atpx handler */
 #if defined(CONFIG_VGA_SWITCHEROO)
@@ -1288,23 +1616,15 @@ void amdgpu_register_atpx_handler(void);
 void amdgpu_unregister_atpx_handler(void);
 bool amdgpu_has_atpx_dgpu_power_cntl(void);
 bool amdgpu_is_atpx_hybrid(void);
-bool amdgpu_atpx_dgpu_req_power_for_displays(void);
 bool amdgpu_has_atpx(void);
 #else
 static inline void amdgpu_register_atpx_handler(void) {}
 static inline void amdgpu_unregister_atpx_handler(void) {}
 static inline bool amdgpu_has_atpx_dgpu_power_cntl(void) { return false; }
 static inline bool amdgpu_is_atpx_hybrid(void) { return false; }
-static inline bool amdgpu_atpx_dgpu_req_power_for_displays(void) { return false; }
 static inline bool amdgpu_has_atpx(void) { return false; }
 #endif
 
-#if defined(CONFIG_VGA_SWITCHEROO) && defined(CONFIG_ACPI)
-void *amdgpu_atpx_get_dhandle(void);
-#else
-static inline void *amdgpu_atpx_get_dhandle(void) { return NULL; }
-#endif
-
 /*
  * KMS
  */
@@ -1313,13 +1633,13 @@ extern const int amdgpu_max_kms_ioctl;
 
 int amdgpu_driver_load_kms(struct amdgpu_device *adev, unsigned long flags);
 void amdgpu_driver_unload_kms(struct drm_device *dev);
-void amdgpu_driver_lastclose_kms(struct drm_device *dev);
 int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv);
 void amdgpu_driver_postclose_kms(struct drm_device *dev,
 				 struct drm_file *file_priv);
 void amdgpu_driver_release_kms(struct drm_device *dev);
 
-int amdgpu_device_ip_suspend(struct amdgpu_device *adev);
+int amdgpu_device_prepare(struct drm_device *dev);
+void amdgpu_device_complete(struct drm_device *dev);
 int amdgpu_device_suspend(struct drm_device *dev, bool fbcon);
 int amdgpu_device_resume(struct drm_device *dev, bool fbcon);
 u32 amdgpu_get_vblank_counter_kms(struct drm_crtc *crtc);
@@ -1349,6 +1669,12 @@ struct amdgpu_afmt_acr amdgpu_afmt_acr(uint32_t clock);
 
 /* amdgpu_acpi.c */
 
+struct amdgpu_numa_info {
+	uint64_t size;
+	int pxm;
+	int nid;
+};
+
 /* ATCS Device/Driver State */
 #define AMDGPU_ATCS_PSC_DEV_STATE_D0		0
 #define AMDGPU_ATCS_PSC_DEV_STATE_D3_HOT	3
@@ -1364,22 +1690,44 @@ int amdgpu_acpi_pcie_performance_request(struct amdgpu_device *adev,
 						u8 perf_req, bool advertise);
 int amdgpu_acpi_power_shift_control(struct amdgpu_device *adev,
 				    u8 dev_state, bool drv_state);
-int amdgpu_acpi_smart_shift_update(struct drm_device *dev, enum amdgpu_ss ss_state);
+int amdgpu_acpi_smart_shift_update(struct amdgpu_device *adev,
+				   enum amdgpu_ss ss_state);
 int amdgpu_acpi_pcie_notify_device_ready(struct amdgpu_device *adev);
+int amdgpu_acpi_get_tmr_info(struct amdgpu_device *adev, u64 *tmr_offset,
+			     u64 *tmr_size);
+int amdgpu_acpi_get_mem_info(struct amdgpu_device *adev, int xcc_id,
+			     struct amdgpu_numa_info *numa_info);
 
 void amdgpu_acpi_get_backlight_caps(struct amdgpu_dm_backlight_caps *caps);
 bool amdgpu_acpi_should_gpu_reset(struct amdgpu_device *adev);
 void amdgpu_acpi_detect(void);
+void amdgpu_acpi_release(void);
 #else
 static inline int amdgpu_acpi_init(struct amdgpu_device *adev) { return 0; }
+static inline int amdgpu_acpi_get_tmr_info(struct amdgpu_device *adev,
+					   u64 *tmr_offset, u64 *tmr_size)
+{
+	return -EINVAL;
+}
+static inline int amdgpu_acpi_get_mem_info(struct amdgpu_device *adev,
+					   int xcc_id,
+					   struct amdgpu_numa_info *numa_info)
+{
+	return -EINVAL;
+}
 static inline void amdgpu_acpi_fini(struct amdgpu_device *adev) { }
 static inline bool amdgpu_acpi_should_gpu_reset(struct amdgpu_device *adev) { return false; }
 static inline void amdgpu_acpi_detect(void) { }
+static inline void amdgpu_acpi_release(void) { }
 static inline bool amdgpu_acpi_is_power_shift_control_supported(void) { return false; }
 static inline int amdgpu_acpi_power_shift_control(struct amdgpu_device *adev,
 						  u8 dev_state, bool drv_state) { return 0; }
-static inline int amdgpu_acpi_smart_shift_update(struct drm_device *dev,
-						 enum amdgpu_ss ss_state) { return 0; }
+static inline int amdgpu_acpi_smart_shift_update(struct amdgpu_device *adev,
+						 enum amdgpu_ss ss_state)
+{
+	return 0;
+}
+static inline void amdgpu_acpi_get_backlight_caps(struct amdgpu_dm_backlight_caps *caps) { }
 #endif
 
 #if defined(CONFIG_ACPI) && defined(CONFIG_SUSPEND)
@@ -1390,13 +1738,10 @@ static inline bool amdgpu_acpi_is_s0ix_active(struct amdgpu_device *adev) { retu
 static inline bool amdgpu_acpi_is_s3_active(struct amdgpu_device *adev) { return false; }
 #endif
 
-#if defined(CONFIG_DRM_AMD_DC)
-int amdgpu_dm_display_resume(struct amdgpu_device *adev );
-#else
-static inline int amdgpu_dm_display_resume(struct amdgpu_device *adev) { return 0; }
+#if defined(CONFIG_DRM_AMD_ISP)
+int amdgpu_acpi_get_isp4_dev(struct acpi_device **dev);
 #endif
 
-
 void amdgpu_register_gpu_instance(struct amdgpu_device *adev);
 void amdgpu_unregister_gpu_instance(struct amdgpu_device *adev);
 
@@ -1434,4 +1779,30 @@ static inline bool amdgpu_is_tmz(struct amdgpu_device *adev)
 
 int amdgpu_in_reset(struct amdgpu_device *adev);
 
+extern const struct attribute_group amdgpu_vram_mgr_attr_group;
+extern const struct attribute_group amdgpu_gtt_mgr_attr_group;
+extern const struct attribute_group amdgpu_flash_attr_group;
+
+void amdgpu_set_init_level(struct amdgpu_device *adev,
+			   enum amdgpu_init_lvl_id lvl);
+
+static inline int amdgpu_device_bus_status_check(struct amdgpu_device *adev)
+{
+       u32 status;
+       int r;
+
+       r = pci_read_config_dword(adev->pdev, PCI_COMMAND, &status);
+       if (r || PCI_POSSIBLE_ERROR(status)) {
+		dev_err(adev->dev, "device lost from bus!");
+		return -ENODEV;
+       }
+
+       return 0;
+}
+
+void amdgpu_device_set_uid(struct amdgpu_uid *uid_info,
+			   enum amdgpu_uid_type type, uint8_t inst,
+			   uint64_t uid);
+uint64_t amdgpu_device_get_uid(struct amdgpu_uid *uid_info,
+			       enum amdgpu_uid_type type, uint8_t inst);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c
new file mode 100644
index 000000000000..9b3180449150
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.c
@@ -0,0 +1,984 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/list.h>
+#include "amdgpu.h"
+#include "amdgpu_aca.h"
+#include "amdgpu_ras.h"
+
+#define ACA_BANK_HWID(type, hwid, mcatype) [ACA_HWIP_TYPE_##type] = {hwid, mcatype}
+
+typedef int bank_handler_t(struct aca_handle *handle, struct aca_bank *bank, enum aca_smu_type type, void *data);
+
+static struct aca_hwip aca_hwid_mcatypes[ACA_HWIP_TYPE_COUNT] = {
+	ACA_BANK_HWID(SMU,	0x01,	0x01),
+	ACA_BANK_HWID(PCS_XGMI, 0x50,	0x00),
+	ACA_BANK_HWID(UMC,	0x96,	0x00),
+};
+
+static void aca_banks_init(struct aca_banks *banks)
+{
+	if (!banks)
+		return;
+
+	memset(banks, 0, sizeof(*banks));
+	INIT_LIST_HEAD(&banks->list);
+}
+
+static int aca_banks_add_bank(struct aca_banks *banks, struct aca_bank *bank)
+{
+	struct aca_bank_node *node;
+
+	if (!bank)
+		return -EINVAL;
+
+	node = kvzalloc(sizeof(*node), GFP_KERNEL);
+	if (!node)
+		return -ENOMEM;
+
+	memcpy(&node->bank, bank, sizeof(*bank));
+
+	INIT_LIST_HEAD(&node->node);
+	list_add_tail(&node->node, &banks->list);
+
+	banks->nr_banks++;
+
+	return 0;
+}
+
+static void aca_banks_release(struct aca_banks *banks)
+{
+	struct aca_bank_node *node, *tmp;
+
+	if (list_empty(&banks->list))
+		return;
+
+	list_for_each_entry_safe(node, tmp, &banks->list, node) {
+		list_del(&node->node);
+		kvfree(node);
+		banks->nr_banks--;
+	}
+}
+
+static int aca_smu_get_valid_aca_count(struct amdgpu_device *adev, enum aca_smu_type type, u32 *count)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+	const struct aca_smu_funcs *smu_funcs = aca->smu_funcs;
+
+	if (!count)
+		return -EINVAL;
+
+	if (!smu_funcs || !smu_funcs->get_valid_aca_count)
+		return -EOPNOTSUPP;
+
+	return smu_funcs->get_valid_aca_count(adev, type, count);
+}
+
+static struct aca_regs_dump {
+	const char *name;
+	int reg_idx;
+} aca_regs[] = {
+	{"CONTROL",		ACA_REG_IDX_CTL},
+	{"STATUS",		ACA_REG_IDX_STATUS},
+	{"ADDR",		ACA_REG_IDX_ADDR},
+	{"MISC",		ACA_REG_IDX_MISC0},
+	{"CONFIG",		ACA_REG_IDX_CONFIG},
+	{"IPID",		ACA_REG_IDX_IPID},
+	{"SYND",		ACA_REG_IDX_SYND},
+	{"DESTAT",		ACA_REG_IDX_DESTAT},
+	{"DEADDR",		ACA_REG_IDX_DEADDR},
+	{"CONTROL_MASK",	ACA_REG_IDX_CTL_MASK},
+};
+
+static void aca_smu_bank_dump(struct amdgpu_device *adev, int idx, int total, struct aca_bank *bank,
+			      struct ras_query_context *qctx)
+{
+	u64 event_id = qctx ? qctx->evid.event_id : RAS_EVENT_INVALID_ID;
+	int i;
+
+	if (adev->debug_disable_ce_logs &&
+	    bank->smu_err_type == ACA_SMU_TYPE_CE &&
+	    !ACA_BANK_ERR_IS_DEFFERED(bank))
+		return;
+
+	RAS_EVENT_LOG(adev, event_id, HW_ERR "Accelerator Check Architecture events logged\n");
+	/* plus 1 for output format, e.g: ACA[08/08]: xxxx */
+	for (i = 0; i < ARRAY_SIZE(aca_regs); i++)
+		RAS_EVENT_LOG(adev, event_id, HW_ERR "ACA[%02d/%02d].%s=0x%016llx\n",
+			      idx + 1, total, aca_regs[i].name, bank->regs[aca_regs[i].reg_idx]);
+
+	if (ACA_REG__STATUS__SCRUB(bank->regs[ACA_REG_IDX_STATUS]))
+		RAS_EVENT_LOG(adev, event_id, HW_ERR "hardware error logged by the scrubber\n");
+}
+
+static bool aca_bank_hwip_is_matched(struct aca_bank *bank, enum aca_hwip_type type)
+{
+
+	struct aca_hwip *hwip;
+	int hwid, mcatype;
+	u64 ipid;
+
+	if (!bank || type == ACA_HWIP_TYPE_UNKNOW)
+		return false;
+
+	hwip = &aca_hwid_mcatypes[type];
+	if (!hwip->hwid)
+		return false;
+
+	ipid = bank->regs[ACA_REG_IDX_IPID];
+	hwid = ACA_REG__IPID__HARDWAREID(ipid);
+	mcatype = ACA_REG__IPID__MCATYPE(ipid);
+
+	return hwip->hwid == hwid && hwip->mcatype == mcatype;
+}
+
+static int aca_smu_get_valid_aca_banks(struct amdgpu_device *adev, enum aca_smu_type type,
+				       int start, int count,
+				       struct aca_banks *banks, struct ras_query_context *qctx)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+	const struct aca_smu_funcs *smu_funcs = aca->smu_funcs;
+	struct aca_bank bank;
+	int i, max_count, ret;
+
+	if (!count)
+		return 0;
+
+	if (!smu_funcs || !smu_funcs->get_valid_aca_bank)
+		return -EOPNOTSUPP;
+
+	switch (type) {
+	case ACA_SMU_TYPE_UE:
+		max_count = smu_funcs->max_ue_bank_count;
+		break;
+	case ACA_SMU_TYPE_CE:
+		max_count = smu_funcs->max_ce_bank_count;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (start + count > max_count)
+		return -EINVAL;
+
+	count = min_t(int, count, max_count);
+	for (i = 0; i < count; i++) {
+		memset(&bank, 0, sizeof(bank));
+		ret = smu_funcs->get_valid_aca_bank(adev, type, start + i, &bank);
+		if (ret)
+			return ret;
+
+		bank.smu_err_type = type;
+
+		/*
+		 * Poison being consumed when injecting a UE while running background workloads,
+		 * which are unexpected.
+		 */
+		if (type == ACA_SMU_TYPE_UE &&
+		    ACA_REG__STATUS__POISON(bank.regs[ACA_REG_IDX_STATUS]) &&
+		    !aca_bank_hwip_is_matched(&bank, ACA_HWIP_TYPE_UMC))
+			continue;
+
+		aca_smu_bank_dump(adev, i, count, &bank, qctx);
+
+		ret = aca_banks_add_bank(banks, &bank);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+static bool aca_bank_is_valid(struct aca_handle *handle, struct aca_bank *bank, enum aca_smu_type type)
+{
+	const struct aca_bank_ops *bank_ops = handle->bank_ops;
+
+	/* Parse all deferred errors with UMC aca handle */
+	if (ACA_BANK_ERR_IS_DEFFERED(bank))
+		return handle->hwip == ACA_HWIP_TYPE_UMC;
+
+	if (!aca_bank_hwip_is_matched(bank, handle->hwip))
+		return false;
+
+	if (!bank_ops->aca_bank_is_valid)
+		return true;
+
+	return bank_ops->aca_bank_is_valid(handle, bank, type, handle->data);
+}
+
+static struct aca_bank_error *new_bank_error(struct aca_error *aerr, struct aca_bank_info *info)
+{
+	struct aca_bank_error *bank_error;
+
+	bank_error = kvzalloc(sizeof(*bank_error), GFP_KERNEL);
+	if (!bank_error)
+		return NULL;
+
+	INIT_LIST_HEAD(&bank_error->node);
+	memcpy(&bank_error->info, info, sizeof(*info));
+
+	mutex_lock(&aerr->lock);
+	list_add_tail(&bank_error->node, &aerr->list);
+	aerr->nr_errors++;
+	mutex_unlock(&aerr->lock);
+
+	return bank_error;
+}
+
+static struct aca_bank_error *find_bank_error(struct aca_error *aerr, struct aca_bank_info *info)
+{
+	struct aca_bank_error *bank_error = NULL;
+	struct aca_bank_info *tmp_info;
+	bool found = false;
+
+	mutex_lock(&aerr->lock);
+	list_for_each_entry(bank_error, &aerr->list, node) {
+		tmp_info = &bank_error->info;
+		if (tmp_info->socket_id == info->socket_id &&
+		    tmp_info->die_id == info->die_id) {
+			found = true;
+			goto out_unlock;
+		}
+	}
+
+out_unlock:
+	mutex_unlock(&aerr->lock);
+
+	return found ? bank_error : NULL;
+}
+
+static void aca_bank_error_remove(struct aca_error *aerr, struct aca_bank_error *bank_error)
+{
+	if (!aerr || !bank_error)
+		return;
+
+	list_del(&bank_error->node);
+	aerr->nr_errors--;
+
+	kvfree(bank_error);
+}
+
+static struct aca_bank_error *get_bank_error(struct aca_error *aerr, struct aca_bank_info *info)
+{
+	struct aca_bank_error *bank_error;
+
+	if (!aerr || !info)
+		return NULL;
+
+	bank_error = find_bank_error(aerr, info);
+	if (bank_error)
+		return bank_error;
+
+	return new_bank_error(aerr, info);
+}
+
+int aca_error_cache_log_bank_error(struct aca_handle *handle, struct aca_bank_info *info,
+				   enum aca_error_type type, u64 count)
+{
+	struct aca_error_cache *error_cache = &handle->error_cache;
+	struct aca_bank_error *bank_error;
+	struct aca_error *aerr;
+
+	if (!handle || !info || type >= ACA_ERROR_TYPE_COUNT)
+		return -EINVAL;
+
+	if (!count)
+		return 0;
+
+	aerr = &error_cache->errors[type];
+	bank_error = get_bank_error(aerr, info);
+	if (!bank_error)
+		return -ENOMEM;
+
+	bank_error->count += count;
+
+	return 0;
+}
+
+static int aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank, enum aca_smu_type type)
+{
+	const struct aca_bank_ops *bank_ops = handle->bank_ops;
+
+	if (!bank)
+		return -EINVAL;
+
+	if (!bank_ops->aca_bank_parser)
+		return -EOPNOTSUPP;
+
+	return bank_ops->aca_bank_parser(handle, bank, type,
+					 handle->data);
+}
+
+static int handler_aca_log_bank_error(struct aca_handle *handle, struct aca_bank *bank,
+				      enum aca_smu_type type, void *data)
+{
+	int ret;
+
+	ret = aca_bank_parser(handle, bank, type);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static int aca_dispatch_bank(struct aca_handle_manager *mgr, struct aca_bank *bank,
+			     enum aca_smu_type type, bank_handler_t handler, void *data)
+{
+	struct aca_handle *handle;
+	int ret;
+
+	if (list_empty(&mgr->list))
+		return 0;
+
+	list_for_each_entry(handle, &mgr->list, node) {
+		if (!aca_bank_is_valid(handle, bank, type))
+			continue;
+
+		ret = handler(handle, bank, type, data);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+static int aca_dispatch_banks(struct aca_handle_manager *mgr, struct aca_banks *banks,
+			      enum aca_smu_type type, bank_handler_t handler, void *data)
+{
+	struct aca_bank_node *node;
+	struct aca_bank *bank;
+	int ret;
+
+	if (!mgr || !banks)
+		return -EINVAL;
+
+	/* pre check to avoid unnecessary operations */
+	if (list_empty(&mgr->list) || list_empty(&banks->list))
+		return 0;
+
+	list_for_each_entry(node, &banks->list, node) {
+		bank = &node->bank;
+
+		ret = aca_dispatch_bank(mgr, bank, type, handler, data);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+static bool aca_bank_should_update(struct amdgpu_device *adev, enum aca_smu_type type)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+	bool ret = true;
+
+	/*
+	 * Because the UE Valid MCA count will only be cleared after reset,
+	 * in order to avoid repeated counting of the error count,
+	 * the aca bank is only updated once during the gpu recovery stage.
+	 */
+	if (type == ACA_SMU_TYPE_UE) {
+		if (amdgpu_ras_intr_triggered())
+			ret = atomic_cmpxchg(&aca->ue_update_flag, 0, 1) == 0;
+		else
+			atomic_set(&aca->ue_update_flag, 0);
+	}
+
+	return ret;
+}
+
+static void aca_banks_generate_cper(struct amdgpu_device *adev,
+				    enum aca_smu_type type,
+				    struct aca_banks *banks,
+				    int count)
+{
+	struct aca_bank_node *node;
+	struct aca_bank *bank;
+	int r;
+
+	if (!adev->cper.enabled)
+		return;
+
+	if (!banks || !count) {
+		dev_warn(adev->dev, "fail to generate cper records\n");
+		return;
+	}
+
+	/* UEs must be encoded into separate CPER entries */
+	if (type == ACA_SMU_TYPE_UE) {
+		struct aca_banks de_banks;
+
+		aca_banks_init(&de_banks);
+		list_for_each_entry(node, &banks->list, node) {
+			bank = &node->bank;
+			if (bank->aca_err_type == ACA_ERROR_TYPE_DEFERRED) {
+				r = aca_banks_add_bank(&de_banks, bank);
+				if (r)
+					dev_warn(adev->dev, "fail to add de banks, ret = %d\n", r);
+			} else {
+				if (amdgpu_cper_generate_ue_record(adev, bank))
+					dev_warn(adev->dev, "fail to generate ue cper records\n");
+			}
+		}
+
+		if (!list_empty(&de_banks.list)) {
+			if (amdgpu_cper_generate_ce_records(adev, &de_banks, de_banks.nr_banks))
+				dev_warn(adev->dev, "fail to generate de cper records\n");
+		}
+
+		aca_banks_release(&de_banks);
+	} else {
+		/*
+		 * SMU_TYPE_CE banks are combined into 1 CPER entries,
+		 * they could be CEs or DEs or both
+		 */
+		if (amdgpu_cper_generate_ce_records(adev, banks, count))
+			dev_warn(adev->dev, "fail to generate ce cper records\n");
+	}
+}
+
+static int aca_banks_update(struct amdgpu_device *adev, enum aca_smu_type type,
+			    bank_handler_t handler, struct ras_query_context *qctx, void *data)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+	struct aca_banks banks;
+	u32 count = 0;
+	int ret;
+
+	if (list_empty(&aca->mgr.list))
+		return 0;
+
+	if (!aca_bank_should_update(adev, type))
+		return 0;
+
+	ret = aca_smu_get_valid_aca_count(adev, type, &count);
+	if (ret)
+		return ret;
+
+	if (!count)
+		return 0;
+
+	aca_banks_init(&banks);
+
+	ret = aca_smu_get_valid_aca_banks(adev, type, 0, count, &banks, qctx);
+	if (ret)
+		goto err_release_banks;
+
+	if (list_empty(&banks.list)) {
+		ret = 0;
+		goto err_release_banks;
+	}
+
+	ret = aca_dispatch_banks(&aca->mgr, &banks, type,
+				 handler, data);
+	if (ret)
+		goto err_release_banks;
+
+	aca_banks_generate_cper(adev, type, &banks, count);
+
+err_release_banks:
+	aca_banks_release(&banks);
+
+	return ret;
+}
+
+static int aca_log_aca_error_data(struct aca_bank_error *bank_error, enum aca_error_type type, struct ras_err_data *err_data)
+{
+	struct aca_bank_info *info;
+	struct amdgpu_smuio_mcm_config_info mcm_info;
+	u64 count;
+
+	if (type >= ACA_ERROR_TYPE_COUNT)
+		return -EINVAL;
+
+	count = bank_error->count;
+	if (!count)
+		return 0;
+
+	info = &bank_error->info;
+	mcm_info.die_id = info->die_id;
+	mcm_info.socket_id = info->socket_id;
+
+	switch (type) {
+	case ACA_ERROR_TYPE_UE:
+		amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, count);
+		break;
+	case ACA_ERROR_TYPE_CE:
+		amdgpu_ras_error_statistic_ce_count(err_data, &mcm_info, count);
+		break;
+	case ACA_ERROR_TYPE_DEFERRED:
+		amdgpu_ras_error_statistic_de_count(err_data, &mcm_info, count);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int aca_log_aca_error(struct aca_handle *handle, enum aca_error_type type, struct ras_err_data *err_data)
+{
+	struct aca_error_cache *error_cache = &handle->error_cache;
+	struct aca_error *aerr = &error_cache->errors[type];
+	struct aca_bank_error *bank_error, *tmp;
+
+	mutex_lock(&aerr->lock);
+
+	if (list_empty(&aerr->list))
+		goto out_unlock;
+
+	list_for_each_entry_safe(bank_error, tmp, &aerr->list, node) {
+		aca_log_aca_error_data(bank_error, type, err_data);
+		aca_bank_error_remove(aerr, bank_error);
+	}
+
+out_unlock:
+	mutex_unlock(&aerr->lock);
+
+	return 0;
+}
+
+static int __aca_get_error_data(struct amdgpu_device *adev, struct aca_handle *handle, enum aca_error_type type,
+				struct ras_err_data *err_data, struct ras_query_context *qctx)
+{
+	enum aca_smu_type smu_type;
+	int ret;
+
+	switch (type) {
+	case ACA_ERROR_TYPE_UE:
+		smu_type = ACA_SMU_TYPE_UE;
+		break;
+	case ACA_ERROR_TYPE_CE:
+	case ACA_ERROR_TYPE_DEFERRED:
+		smu_type = ACA_SMU_TYPE_CE;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	/* update aca bank to aca source error_cache first */
+	ret = aca_banks_update(adev, smu_type, handler_aca_log_bank_error, qctx, NULL);
+	if (ret)
+		return ret;
+
+	/* DEs may contain in CEs or UEs */
+	if (type != ACA_ERROR_TYPE_DEFERRED)
+		aca_log_aca_error(handle, ACA_ERROR_TYPE_DEFERRED, err_data);
+
+	return aca_log_aca_error(handle, type, err_data);
+}
+
+static bool aca_handle_is_valid(struct aca_handle *handle)
+{
+	if (!handle->mask || !list_empty(&handle->node))
+		return false;
+
+	return true;
+}
+
+int amdgpu_aca_get_error_data(struct amdgpu_device *adev, struct aca_handle *handle,
+			      enum aca_error_type type, struct ras_err_data *err_data,
+			      struct ras_query_context *qctx)
+{
+	if (!handle || !err_data)
+		return -EINVAL;
+
+	if (aca_handle_is_valid(handle))
+		return -EOPNOTSUPP;
+
+	if ((type < 0) || (!(BIT(type) & handle->mask)))
+		return  0;
+
+	return __aca_get_error_data(adev, handle, type, err_data, qctx);
+}
+
+static void aca_error_init(struct aca_error *aerr, enum aca_error_type type)
+{
+	mutex_init(&aerr->lock);
+	INIT_LIST_HEAD(&aerr->list);
+	aerr->type = type;
+	aerr->nr_errors = 0;
+}
+
+static void aca_init_error_cache(struct aca_handle *handle)
+{
+	struct aca_error_cache *error_cache = &handle->error_cache;
+	int type;
+
+	for (type = ACA_ERROR_TYPE_UE; type < ACA_ERROR_TYPE_COUNT; type++)
+		aca_error_init(&error_cache->errors[type], type);
+}
+
+static void aca_error_fini(struct aca_error *aerr)
+{
+	struct aca_bank_error *bank_error, *tmp;
+
+	mutex_lock(&aerr->lock);
+	if (list_empty(&aerr->list))
+		goto out_unlock;
+
+	list_for_each_entry_safe(bank_error, tmp, &aerr->list, node)
+		aca_bank_error_remove(aerr, bank_error);
+
+out_unlock:
+	mutex_destroy(&aerr->lock);
+}
+
+static void aca_fini_error_cache(struct aca_handle *handle)
+{
+	struct aca_error_cache *error_cache = &handle->error_cache;
+	int type;
+
+	for (type = ACA_ERROR_TYPE_UE; type < ACA_ERROR_TYPE_COUNT; type++)
+		aca_error_fini(&error_cache->errors[type]);
+}
+
+static int add_aca_handle(struct amdgpu_device *adev, struct aca_handle_manager *mgr, struct aca_handle *handle,
+			  const char *name, const struct aca_info *ras_info, void *data)
+{
+	memset(handle, 0, sizeof(*handle));
+
+	handle->adev = adev;
+	handle->mgr = mgr;
+	handle->name = name;
+	handle->hwip = ras_info->hwip;
+	handle->mask = ras_info->mask;
+	handle->bank_ops = ras_info->bank_ops;
+	handle->data = data;
+	aca_init_error_cache(handle);
+
+	INIT_LIST_HEAD(&handle->node);
+	list_add_tail(&handle->node, &mgr->list);
+	mgr->nr_handles++;
+
+	return 0;
+}
+
+static ssize_t aca_sysfs_read(struct device *dev,
+			      struct device_attribute *attr, char *buf)
+{
+	struct aca_handle *handle = container_of(attr, struct aca_handle, aca_attr);
+
+	/* NOTE: the aca cache will be auto cleared once read,
+	 * So the driver should unify the query entry point, forward request to ras query interface directly */
+	return amdgpu_ras_aca_sysfs_read(dev, attr, handle, buf, handle->data);
+}
+
+static int add_aca_sysfs(struct amdgpu_device *adev, struct aca_handle *handle)
+{
+	struct device_attribute *aca_attr = &handle->aca_attr;
+
+	snprintf(handle->attr_name, sizeof(handle->attr_name) - 1, "aca_%s", handle->name);
+	aca_attr->show = aca_sysfs_read;
+	aca_attr->attr.name = handle->attr_name;
+	aca_attr->attr.mode = S_IRUGO;
+	sysfs_attr_init(&aca_attr->attr);
+
+	return sysfs_add_file_to_group(&adev->dev->kobj,
+				       &aca_attr->attr,
+				       "ras");
+}
+
+int amdgpu_aca_add_handle(struct amdgpu_device *adev, struct aca_handle *handle,
+			  const char *name, const struct aca_info *ras_info, void *data)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+	int ret;
+
+	if (!amdgpu_aca_is_enabled(adev))
+		return 0;
+
+	ret = add_aca_handle(adev, &aca->mgr, handle, name, ras_info, data);
+	if (ret)
+		return ret;
+
+	return add_aca_sysfs(adev, handle);
+}
+
+static void remove_aca_handle(struct aca_handle *handle)
+{
+	struct aca_handle_manager *mgr = handle->mgr;
+
+	aca_fini_error_cache(handle);
+	list_del(&handle->node);
+	mgr->nr_handles--;
+}
+
+static void remove_aca_sysfs(struct aca_handle *handle)
+{
+	struct amdgpu_device *adev = handle->adev;
+	struct device_attribute *aca_attr = &handle->aca_attr;
+
+	if (adev->dev->kobj.sd)
+		sysfs_remove_file_from_group(&adev->dev->kobj,
+					     &aca_attr->attr,
+					     "ras");
+}
+
+void amdgpu_aca_remove_handle(struct aca_handle *handle)
+{
+	if (!handle || list_empty(&handle->node))
+		return;
+
+	remove_aca_sysfs(handle);
+	remove_aca_handle(handle);
+}
+
+static int aca_manager_init(struct aca_handle_manager *mgr)
+{
+	INIT_LIST_HEAD(&mgr->list);
+	mgr->nr_handles = 0;
+
+	return 0;
+}
+
+static void aca_manager_fini(struct aca_handle_manager *mgr)
+{
+	struct aca_handle *handle, *tmp;
+
+	if (list_empty(&mgr->list))
+		return;
+
+	list_for_each_entry_safe(handle, tmp, &mgr->list, node)
+		amdgpu_aca_remove_handle(handle);
+}
+
+bool amdgpu_aca_is_enabled(struct amdgpu_device *adev)
+{
+	return (adev->aca.is_enabled ||
+		adev->debug_enable_ras_aca);
+}
+
+int amdgpu_aca_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+	int ret;
+
+	atomic_set(&aca->ue_update_flag, 0);
+
+	ret = aca_manager_init(&aca->mgr);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+void amdgpu_aca_fini(struct amdgpu_device *adev)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+
+	aca_manager_fini(&aca->mgr);
+
+	atomic_set(&aca->ue_update_flag, 0);
+}
+
+int amdgpu_aca_reset(struct amdgpu_device *adev)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+
+	atomic_set(&aca->ue_update_flag, 0);
+
+	return 0;
+}
+
+void amdgpu_aca_set_smu_funcs(struct amdgpu_device *adev, const struct aca_smu_funcs *smu_funcs)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+
+	WARN_ON(aca->smu_funcs);
+	aca->smu_funcs = smu_funcs;
+}
+
+int aca_bank_info_decode(struct aca_bank *bank, struct aca_bank_info *info)
+{
+	u64 ipid;
+	u32 instidhi, instidlo;
+
+	if (!bank || !info)
+		return -EINVAL;
+
+	ipid = bank->regs[ACA_REG_IDX_IPID];
+	info->hwid = ACA_REG__IPID__HARDWAREID(ipid);
+	info->mcatype = ACA_REG__IPID__MCATYPE(ipid);
+	/*
+	 * Unfied DieID Format: SAASS. A:AID, S:Socket.
+	 * Unfied DieID[4:4] = InstanceId[0:0]
+	 * Unfied DieID[0:3] = InstanceIdHi[0:3]
+	 */
+	instidhi = ACA_REG__IPID__INSTANCEIDHI(ipid);
+	instidlo = ACA_REG__IPID__INSTANCEIDLO(ipid);
+	info->die_id = ((instidhi >> 2) & 0x03);
+	info->socket_id = ((instidlo & 0x1) << 2) | (instidhi & 0x03);
+
+	return 0;
+}
+
+static int aca_bank_get_error_code(struct amdgpu_device *adev, struct aca_bank *bank)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+	const struct aca_smu_funcs *smu_funcs = aca->smu_funcs;
+
+	if (!smu_funcs || !smu_funcs->parse_error_code)
+		return -EOPNOTSUPP;
+
+	return smu_funcs->parse_error_code(adev, bank);
+}
+
+int aca_bank_check_error_codes(struct amdgpu_device *adev, struct aca_bank *bank, int *err_codes, int size)
+{
+	int i, error_code;
+
+	if (!bank || !err_codes)
+		return -EINVAL;
+
+	error_code = aca_bank_get_error_code(adev, bank);
+	if (error_code < 0)
+		return error_code;
+
+	for (i = 0; i < size; i++) {
+		if (err_codes[i] == error_code)
+			return 0;
+	}
+
+	return -EINVAL;
+}
+
+int amdgpu_aca_smu_set_debug_mode(struct amdgpu_device *adev, bool en)
+{
+	struct amdgpu_aca *aca = &adev->aca;
+	const struct aca_smu_funcs *smu_funcs = aca->smu_funcs;
+
+	if (!smu_funcs || !smu_funcs->set_debug_mode)
+		return -EOPNOTSUPP;
+
+	return smu_funcs->set_debug_mode(adev, en);
+}
+
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_aca_smu_debug_mode_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	int ret;
+
+	ret = amdgpu_ras_set_aca_debug_mode(adev, val ? true : false);
+	if (ret)
+		return ret;
+
+	dev_info(adev->dev, "amdgpu set smu aca debug mode %s success\n", val ? "on" : "off");
+
+	return 0;
+}
+
+static void aca_dump_entry(struct seq_file *m, struct aca_bank *bank, enum aca_smu_type type, int idx)
+{
+	struct aca_bank_info info;
+	int i, ret;
+
+	ret = aca_bank_info_decode(bank, &info);
+	if (ret)
+		return;
+
+	seq_printf(m, "aca entry[%d].type: %s\n", idx, type ==  ACA_SMU_TYPE_UE ? "UE" : "CE");
+	seq_printf(m, "aca entry[%d].info: socketid:%d aid:%d hwid:0x%03x mcatype:0x%04x\n",
+		   idx, info.socket_id, info.die_id, info.hwid, info.mcatype);
+
+	for (i = 0; i < ARRAY_SIZE(aca_regs); i++)
+		seq_printf(m, "aca entry[%d].regs[%d]: 0x%016llx\n", idx, aca_regs[i].reg_idx, bank->regs[aca_regs[i].reg_idx]);
+}
+
+struct aca_dump_context {
+	struct seq_file *m;
+	int idx;
+};
+
+static int handler_aca_bank_dump(struct aca_handle *handle, struct aca_bank *bank,
+				 enum aca_smu_type type, void *data)
+{
+	struct aca_dump_context *ctx = (struct aca_dump_context *)data;
+
+	aca_dump_entry(ctx->m, bank, type, ctx->idx++);
+
+	return handler_aca_log_bank_error(handle, bank, type, NULL);
+}
+
+static int aca_dump_show(struct seq_file *m, enum aca_smu_type type)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct aca_dump_context context = {
+		.m = m,
+		.idx = 0,
+	};
+
+	return aca_banks_update(adev, type, handler_aca_bank_dump, NULL, (void *)&context);
+}
+
+static int aca_dump_ce_show(struct seq_file *m, void *unused)
+{
+	return aca_dump_show(m, ACA_SMU_TYPE_CE);
+}
+
+static int aca_dump_ce_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, aca_dump_ce_show, inode->i_private);
+}
+
+static const struct file_operations aca_ce_dump_debug_fops = {
+	.owner = THIS_MODULE,
+	.open = aca_dump_ce_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = single_release,
+};
+
+static int aca_dump_ue_show(struct seq_file *m, void *unused)
+{
+	return aca_dump_show(m, ACA_SMU_TYPE_UE);
+}
+
+static int aca_dump_ue_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, aca_dump_ue_show, inode->i_private);
+}
+
+static const struct file_operations aca_ue_dump_debug_fops = {
+	.owner = THIS_MODULE,
+	.open = aca_dump_ue_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = single_release,
+};
+
+DEFINE_DEBUGFS_ATTRIBUTE(aca_debug_mode_fops, NULL, amdgpu_aca_smu_debug_mode_set, "%llu\n");
+#endif
+
+void amdgpu_aca_smu_debugfs_init(struct amdgpu_device *adev, struct dentry *root)
+{
+#if defined(CONFIG_DEBUG_FS)
+	if (!root)
+		return;
+
+	debugfs_create_file("aca_debug_mode", 0200, root, adev, &aca_debug_mode_fops);
+	debugfs_create_file("aca_ue_dump", 0400, root, adev, &aca_ue_dump_debug_fops);
+	debugfs_create_file("aca_ce_dump", 0400, root, adev, &aca_ce_dump_debug_fops);
+#endif
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h
new file mode 100644
index 000000000000..38c88897e1ec
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_aca.h
@@ -0,0 +1,232 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __AMDGPU_ACA_H__
+#define __AMDGPU_ACA_H__
+
+#include <linux/list.h>
+
+struct ras_err_data;
+struct ras_query_context;
+
+#define ACA_MAX_REGS_COUNT	(16)
+
+#define ACA_REG_FIELD(x, h, l)			(((x) & GENMASK_ULL(h, l)) >> l)
+#define ACA_REG__STATUS__VAL(x)			ACA_REG_FIELD(x, 63, 63)
+#define ACA_REG__STATUS__OVERFLOW(x)		ACA_REG_FIELD(x, 62, 62)
+#define ACA_REG__STATUS__UC(x)			ACA_REG_FIELD(x, 61, 61)
+#define ACA_REG__STATUS__EN(x)			ACA_REG_FIELD(x, 60, 60)
+#define ACA_REG__STATUS__MISCV(x)		ACA_REG_FIELD(x, 59, 59)
+#define ACA_REG__STATUS__ADDRV(x)		ACA_REG_FIELD(x, 58, 58)
+#define ACA_REG__STATUS__PCC(x)			ACA_REG_FIELD(x, 57, 57)
+#define ACA_REG__STATUS__ERRCOREIDVAL(x)	ACA_REG_FIELD(x, 56, 56)
+#define ACA_REG__STATUS__TCC(x)			ACA_REG_FIELD(x, 55, 55)
+#define ACA_REG__STATUS__SYNDV(x)		ACA_REG_FIELD(x, 53, 53)
+#define ACA_REG__STATUS__CECC(x)		ACA_REG_FIELD(x, 46, 46)
+#define ACA_REG__STATUS__UECC(x)		ACA_REG_FIELD(x, 45, 45)
+#define ACA_REG__STATUS__DEFERRED(x)		ACA_REG_FIELD(x, 44, 44)
+#define ACA_REG__STATUS__POISON(x)		ACA_REG_FIELD(x, 43, 43)
+#define ACA_REG__STATUS__SCRUB(x)		ACA_REG_FIELD(x, 40, 40)
+#define ACA_REG__STATUS__ERRCOREID(x)		ACA_REG_FIELD(x, 37, 32)
+#define ACA_REG__STATUS__ADDRLSB(x)		ACA_REG_FIELD(x, 29, 24)
+#define ACA_REG__STATUS__ERRORCODEEXT(x)	ACA_REG_FIELD(x, 21, 16)
+#define ACA_REG__STATUS__ERRORCODE(x)		ACA_REG_FIELD(x, 15, 0)
+
+#define ACA_REG__IPID__MCATYPE(x)		ACA_REG_FIELD(x, 63, 48)
+#define ACA_REG__IPID__INSTANCEIDHI(x)		ACA_REG_FIELD(x, 47, 44)
+#define ACA_REG__IPID__HARDWAREID(x)		ACA_REG_FIELD(x, 43, 32)
+#define ACA_REG__IPID__INSTANCEIDLO(x)		ACA_REG_FIELD(x, 31, 0)
+
+#define ACA_REG__MISC0__VALID(x)		ACA_REG_FIELD(x, 63, 63)
+#define ACA_REG__MISC0__OVRFLW(x)		ACA_REG_FIELD(x, 48, 48)
+#define ACA_REG__MISC0__ERRCNT(x)		ACA_REG_FIELD(x, 43, 32)
+
+#define ACA_REG__SYND__ERRORINFORMATION(x)	ACA_REG_FIELD(x, 17, 0)
+
+/* NOTE: The following codes refers to the smu header file */
+#define ACA_EXTERROR_CODE_CE			0x3a
+#define ACA_EXTERROR_CODE_FAULT			0x3b
+
+#define ACA_ERROR_UE_MASK		BIT_MASK(ACA_ERROR_TYPE_UE)
+#define ACA_ERROR_CE_MASK		BIT_MASK(ACA_ERROR_TYPE_CE)
+#define ACA_ERROR_DEFERRED_MASK		BIT_MASK(ACA_ERROR_TYPE_DEFERRED)
+
+#define mmSMNAID_AID0_MCA_SMU		0x03b30400	/* SMN AID AID0 */
+#define mmSMNAID_XCD0_MCA_SMU		0x36430400	/* SMN AID XCD0 */
+#define mmSMNAID_XCD1_MCA_SMU		0x38430400	/* SMN AID XCD1 */
+#define mmSMNXCD_XCD0_MCA_SMU		0x40430400	/* SMN XCD XCD0 */
+
+#define ACA_BANK_ERR_IS_DEFFERED(bank)                                \
+	(ACA_REG__STATUS__POISON((bank)->regs[ACA_REG_IDX_STATUS]) || \
+	 ACA_REG__STATUS__DEFERRED((bank)->regs[ACA_REG_IDX_STATUS]))
+
+enum aca_reg_idx {
+	ACA_REG_IDX_CTL			= 0,
+	ACA_REG_IDX_STATUS		= 1,
+	ACA_REG_IDX_ADDR		= 2,
+	ACA_REG_IDX_MISC0		= 3,
+	ACA_REG_IDX_CONFIG		= 4,
+	ACA_REG_IDX_IPID		= 5,
+	ACA_REG_IDX_SYND		= 6,
+	ACA_REG_IDX_DESTAT		= 8,
+	ACA_REG_IDX_DEADDR		= 9,
+	ACA_REG_IDX_CTL_MASK		= 10,
+	ACA_REG_IDX_COUNT		= 16,
+};
+
+enum aca_hwip_type {
+	ACA_HWIP_TYPE_UNKNOW = -1,
+	ACA_HWIP_TYPE_PSP = 0,
+	ACA_HWIP_TYPE_UMC,
+	ACA_HWIP_TYPE_SMU,
+	ACA_HWIP_TYPE_PCS_XGMI,
+	ACA_HWIP_TYPE_COUNT,
+};
+
+enum aca_error_type {
+	ACA_ERROR_TYPE_INVALID = -1,
+	ACA_ERROR_TYPE_UE = 0,
+	ACA_ERROR_TYPE_CE,
+	ACA_ERROR_TYPE_DEFERRED,
+	ACA_ERROR_TYPE_COUNT
+};
+
+enum aca_smu_type {
+	ACA_SMU_TYPE_INVALID = -1,
+	ACA_SMU_TYPE_UE = 0,
+	ACA_SMU_TYPE_CE,
+	ACA_SMU_TYPE_COUNT,
+};
+
+struct aca_hwip {
+	int hwid;
+	int mcatype;
+};
+
+struct aca_bank {
+	enum aca_error_type aca_err_type;
+	enum aca_smu_type smu_err_type;
+	u64 regs[ACA_MAX_REGS_COUNT];
+};
+
+struct aca_bank_node {
+	struct aca_bank bank;
+	struct list_head node;
+};
+
+struct aca_banks {
+	int nr_banks;
+	struct list_head list;
+};
+
+struct aca_bank_info {
+	int die_id;
+	int socket_id;
+	int hwid;
+	int mcatype;
+};
+
+struct aca_bank_error {
+	struct list_head node;
+	struct aca_bank_info info;
+	u64 count;
+};
+
+struct aca_error {
+	struct list_head list;
+	struct mutex lock;
+	enum aca_error_type type;
+	int nr_errors;
+};
+
+struct aca_handle_manager {
+	struct list_head list;
+	int nr_handles;
+};
+
+struct aca_error_cache {
+	struct aca_error errors[ACA_ERROR_TYPE_COUNT];
+};
+
+struct aca_handle {
+	struct list_head node;
+	enum aca_hwip_type hwip;
+	struct amdgpu_device *adev;
+	struct aca_handle_manager *mgr;
+	struct aca_error_cache error_cache;
+	const struct aca_bank_ops *bank_ops;
+	struct device_attribute aca_attr;
+	char attr_name[64];
+	const char *name;
+	u32 mask;
+	void *data;
+};
+
+struct aca_bank_ops {
+	int (*aca_bank_parser)(struct aca_handle *handle, struct aca_bank *bank, enum aca_smu_type type, void *data);
+	bool (*aca_bank_is_valid)(struct aca_handle *handle, struct aca_bank *bank, enum aca_smu_type type,
+				  void *data);
+};
+
+struct aca_smu_funcs {
+	int max_ue_bank_count;
+	int max_ce_bank_count;
+	int (*set_debug_mode)(struct amdgpu_device *adev, bool enable);
+	int (*get_valid_aca_count)(struct amdgpu_device *adev, enum aca_smu_type type, u32 *count);
+	int (*get_valid_aca_bank)(struct amdgpu_device *adev, enum aca_smu_type type, int idx, struct aca_bank *bank);
+	int (*parse_error_code)(struct amdgpu_device *adev, struct aca_bank *bank);
+};
+
+struct amdgpu_aca {
+	struct aca_handle_manager mgr;
+	const struct aca_smu_funcs *smu_funcs;
+	atomic_t ue_update_flag;
+	bool is_enabled;
+};
+
+struct aca_info {
+	enum aca_hwip_type hwip;
+	const struct aca_bank_ops *bank_ops;
+	u32 mask;
+};
+
+int amdgpu_aca_init(struct amdgpu_device *adev);
+void amdgpu_aca_fini(struct amdgpu_device *adev);
+int amdgpu_aca_reset(struct amdgpu_device *adev);
+void amdgpu_aca_set_smu_funcs(struct amdgpu_device *adev, const struct aca_smu_funcs *smu_funcs);
+bool amdgpu_aca_is_enabled(struct amdgpu_device *adev);
+
+int aca_bank_info_decode(struct aca_bank *bank, struct aca_bank_info *info);
+int aca_bank_check_error_codes(struct amdgpu_device *adev, struct aca_bank *bank, int *err_codes, int size);
+
+int amdgpu_aca_add_handle(struct amdgpu_device *adev, struct aca_handle *handle,
+			  const char *name, const struct aca_info *aca_info, void *data);
+void amdgpu_aca_remove_handle(struct aca_handle *handle);
+int amdgpu_aca_get_error_data(struct amdgpu_device *adev, struct aca_handle *handle,
+			      enum aca_error_type type, struct ras_err_data *err_data,
+			      struct ras_query_context *qctx);
+int amdgpu_aca_smu_set_debug_mode(struct amdgpu_device *adev, bool en);
+void amdgpu_aca_smu_debugfs_init(struct amdgpu_device *adev, struct dentry *root);
+int aca_error_cache_log_bank_error(struct aca_handle *handle, struct aca_bank_info *info,
+				   enum aca_error_type type, u64 count);
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c
index 6d72355ac492..381ef205b0df 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c
@@ -98,9 +98,9 @@ enum {
 	ACP_TILE_DSP2,
 };
 
-static int acp_sw_init(void *handle)
+static int acp_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->acp.parent = adev->dev;
 
@@ -112,9 +112,9 @@ static int acp_sw_init(void *handle)
 	return 0;
 }
 
-static int acp_sw_fini(void *handle)
+static int acp_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (adev->acp.cgs_device)
 		amdgpu_cgs_destroy_device(adev->acp.cgs_device);
@@ -140,7 +140,7 @@ static int acp_poweroff(struct generic_pm_domain *genpd)
 	 * 2. power off the acp tiles
 	 * 3. check and enter ulv state
 	 */
-	amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true);
+	amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true, 0);
 	return 0;
 }
 
@@ -157,7 +157,7 @@ static int acp_poweron(struct generic_pm_domain *genpd)
 	 * 2. turn on acp clock
 	 * 3. power on acp tiles
 	 */
-	amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false);
+	amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false, 0);
 	return 0;
 }
 
@@ -219,10 +219,10 @@ static const struct dmi_system_id acp_quirk_table[] = {
 /**
  * acp_hw_init - start and test ACP block
  *
- * @handle: handle used to pass amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  */
-static int acp_hw_init(void *handle)
+static int acp_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
 	u64 acp_base;
@@ -230,19 +230,13 @@ static int acp_hw_init(void *handle)
 	u32 count = 0;
 	struct i2s_platform_data *i2s_pdata = NULL;
 
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	const struct amdgpu_ip_block *ip_block =
-		amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_ACP);
-
-	if (!ip_block)
-		return -EINVAL;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amd_acp_hw_init(adev->acp.cgs_device,
 			    ip_block->version->major, ip_block->version->minor);
 	/* -ENODEV means board uses AZ rather than ACP */
 	if (r == -ENODEV) {
-		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true);
+		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true, 0);
 		return 0;
 	} else if (r) {
 		return r;
@@ -308,17 +302,19 @@ static int acp_hw_init(void *handle)
 		adev->acp.acp_res[2].end = adev->acp.acp_res[2].start;
 
 		adev->acp.acp_cell[0].name = "acp_audio_dma";
+		adev->acp.acp_cell[0].id = 0;
 		adev->acp.acp_cell[0].num_resources = 3;
 		adev->acp.acp_cell[0].resources = &adev->acp.acp_res[0];
 		adev->acp.acp_cell[0].platform_data = &adev->asic_type;
 		adev->acp.acp_cell[0].pdata_size = sizeof(adev->asic_type);
 
 		adev->acp.acp_cell[1].name = "designware-i2s";
+		adev->acp.acp_cell[1].id = 1;
 		adev->acp.acp_cell[1].num_resources = 1;
 		adev->acp.acp_cell[1].resources = &adev->acp.acp_res[1];
 		adev->acp.acp_cell[1].platform_data = &i2s_pdata[0];
 		adev->acp.acp_cell[1].pdata_size = sizeof(struct i2s_platform_data);
-		r = mfd_add_hotplug_devices(adev->acp.parent, adev->acp.acp_cell, 2);
+		r = mfd_add_devices(adev->acp.parent, 0, adev->acp.acp_cell, 2, NULL, 0, NULL);
 		if (r)
 			goto failure;
 		r = device_for_each_child(adev->acp.parent, &adev->acp.acp_genpd->gpd,
@@ -416,30 +412,34 @@ static int acp_hw_init(void *handle)
 		adev->acp.acp_res[4].end = adev->acp.acp_res[4].start;
 
 		adev->acp.acp_cell[0].name = "acp_audio_dma";
+		adev->acp.acp_cell[0].id = 0;
 		adev->acp.acp_cell[0].num_resources = 5;
 		adev->acp.acp_cell[0].resources = &adev->acp.acp_res[0];
 		adev->acp.acp_cell[0].platform_data = &adev->asic_type;
 		adev->acp.acp_cell[0].pdata_size = sizeof(adev->asic_type);
 
 		adev->acp.acp_cell[1].name = "designware-i2s";
+		adev->acp.acp_cell[1].id = 1;
 		adev->acp.acp_cell[1].num_resources = 1;
 		adev->acp.acp_cell[1].resources = &adev->acp.acp_res[1];
 		adev->acp.acp_cell[1].platform_data = &i2s_pdata[0];
 		adev->acp.acp_cell[1].pdata_size = sizeof(struct i2s_platform_data);
 
 		adev->acp.acp_cell[2].name = "designware-i2s";
+		adev->acp.acp_cell[2].id = 2;
 		adev->acp.acp_cell[2].num_resources = 1;
 		adev->acp.acp_cell[2].resources = &adev->acp.acp_res[2];
 		adev->acp.acp_cell[2].platform_data = &i2s_pdata[1];
 		adev->acp.acp_cell[2].pdata_size = sizeof(struct i2s_platform_data);
 
 		adev->acp.acp_cell[3].name = "designware-i2s";
+		adev->acp.acp_cell[3].id = 3;
 		adev->acp.acp_cell[3].num_resources = 1;
 		adev->acp.acp_cell[3].resources = &adev->acp.acp_res[3];
 		adev->acp.acp_cell[3].platform_data = &i2s_pdata[2];
 		adev->acp.acp_cell[3].pdata_size = sizeof(struct i2s_platform_data);
 
-		r = mfd_add_hotplug_devices(adev->acp.parent, adev->acp.acp_cell, ACP_DEVS);
+		r = mfd_add_devices(adev->acp.parent, 0, adev->acp.acp_cell, ACP_DEVS, NULL, 0, NULL);
 		if (r)
 			goto failure;
 
@@ -503,18 +503,18 @@ failure:
 /**
  * acp_hw_fini - stop the hardware block
  *
- * @handle: handle used to pass amdgpu_device pointer
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
  *
  */
-static int acp_hw_fini(void *handle)
+static int acp_hw_fini(struct amdgpu_ip_block *ip_block)
 {
 	u32 val = 0;
 	u32 count = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* return early if no ACP */
 	if (!adev->acp.acp_genpd) {
-		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false);
+		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false, 0);
 		return 0;
 	}
 
@@ -565,67 +565,50 @@ static int acp_hw_fini(void *handle)
 	return 0;
 }
 
-static int acp_suspend(void *handle)
+static int acp_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* power up on suspend */
 	if (!adev->acp.acp_cell)
-		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false);
+		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false, 0);
 	return 0;
 }
 
-static int acp_resume(void *handle)
+static int acp_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* power down again on resume */
 	if (!adev->acp.acp_cell)
-		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true);
-	return 0;
-}
-
-static int acp_early_init(void *handle)
-{
+		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true, 0);
 	return 0;
 }
 
-static bool acp_is_idle(void *handle)
+static bool acp_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	return true;
 }
 
-static int acp_wait_for_idle(void *handle)
-{
-	return 0;
-}
-
-static int acp_soft_reset(void *handle)
-{
-	return 0;
-}
-
-static int acp_set_clockgating_state(void *handle,
+static int acp_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 				     enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int acp_set_powergating_state(void *handle,
+static int acp_set_powergating_state(struct amdgpu_ip_block *ip_block,
 				     enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_PG_STATE_GATE);
 
-	amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, enable);
+	amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, enable, 0);
 
 	return 0;
 }
 
 static const struct amd_ip_funcs acp_ip_funcs = {
 	.name = "acp_ip",
-	.early_init = acp_early_init,
-	.late_init = NULL,
 	.sw_init = acp_sw_init,
 	.sw_fini = acp_sw_fini,
 	.hw_init = acp_hw_init,
@@ -633,8 +616,6 @@ static const struct amd_ip_funcs acp_ip_funcs = {
 	.suspend = acp_suspend,
 	.resume = acp_resume,
 	.is_idle = acp_is_idle,
-	.wait_for_idle = acp_wait_for_idle,
-	.soft_reset = acp_soft_reset,
 	.set_clockgating_state = acp_set_clockgating_state,
 	.set_powergating_state = acp_set_powergating_state,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c
index 60b1857f469e..d31460a9e958 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c
@@ -26,6 +26,7 @@
 #include <linux/acpi.h>
 #include <linux/backlight.h>
 #include <linux/slab.h>
+#include <linux/xarray.h>
 #include <linux/power_supply.h>
 #include <linux/pm_runtime.h>
 #include <linux/suspend.h>
@@ -38,6 +39,45 @@
 #include "amd_acpi.h"
 #include "atom.h"
 
+/* Declare GUID for AMD _DSM method for XCCs */
+static const guid_t amd_xcc_dsm_guid = GUID_INIT(0x8267f5d5, 0xa556, 0x44f2,
+						 0xb8, 0xb4, 0x45, 0x56, 0x2e,
+						 0x8c, 0x5b, 0xec);
+
+#define AMD_XCC_HID_START 3000
+#define AMD_XCC_DSM_GET_NUM_FUNCS 0
+#define AMD_XCC_DSM_GET_SUPP_MODE 1
+#define AMD_XCC_DSM_GET_XCP_MODE 2
+#define AMD_XCC_DSM_GET_VF_XCC_MAPPING 4
+#define AMD_XCC_DSM_GET_TMR_INFO 5
+#define AMD_XCC_DSM_NUM_FUNCS 5
+
+#define AMD_XCC_MAX_HID 24
+
+struct xarray numa_info_xa;
+
+/* Encapsulates the XCD acpi object information */
+struct amdgpu_acpi_xcc_info {
+	struct list_head list;
+	struct amdgpu_numa_info *numa_info;
+	uint8_t xcp_node;
+	uint8_t phy_id;
+	acpi_handle handle;
+};
+
+struct amdgpu_acpi_dev_info {
+	struct list_head list;
+	struct list_head xcc_list;
+	uint32_t sbdf;
+	uint16_t supp_xcp_mode;
+	uint16_t xcp_mode;
+	uint16_t mem_mode;
+	uint64_t tmr_base;
+	uint64_t tmr_size;
+};
+
+struct list_head amdgpu_acpi_dev_list;
+
 struct amdgpu_atif_notification_cfg {
 	bool enabled;
 	int command_code;
@@ -107,6 +147,7 @@ static union acpi_object *amdgpu_atif_call(struct amdgpu_atif *atif,
 					   struct acpi_buffer *params)
 {
 	acpi_status status;
+	union acpi_object *obj;
 	union acpi_object atif_arg_elements[2];
 	struct acpi_object_list atif_arg;
 	struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL };
@@ -129,16 +170,24 @@ static union acpi_object *amdgpu_atif_call(struct amdgpu_atif *atif,
 
 	status = acpi_evaluate_object(atif->handle, NULL, &atif_arg,
 				      &buffer);
+	obj = (union acpi_object *)buffer.pointer;
 
-	/* Fail only if calling the method fails and ATIF is supported */
-	if (ACPI_FAILURE(status) && status != AE_NOT_FOUND) {
+	/* Fail if calling the method fails */
+	if (ACPI_FAILURE(status)) {
 		DRM_DEBUG_DRIVER("failed to evaluate ATIF got %s\n",
 				 acpi_format_exception(status));
-		kfree(buffer.pointer);
+		kfree(obj);
 		return NULL;
 	}
 
-	return buffer.pointer;
+	if (obj->type != ACPI_TYPE_BUFFER) {
+		DRM_DEBUG_DRIVER("bad object returned from ATIF: %d\n",
+				 obj->type);
+		kfree(obj);
+		return NULL;
+	}
+
+	return obj;
 }
 
 /**
@@ -343,6 +392,12 @@ static int amdgpu_atif_query_backlight_caps(struct amdgpu_atif *atif)
 			characteristics.min_input_signal;
 	atif->backlight_caps.max_input_signal =
 			characteristics.max_input_signal;
+	atif->backlight_caps.ac_level = characteristics.ac_level;
+	atif->backlight_caps.dc_level = characteristics.dc_level;
+	atif->backlight_caps.data_points = characteristics.number_of_points;
+	memcpy(atif->backlight_caps.luminance_data,
+	       characteristics.data_points,
+	       sizeof(atif->backlight_caps.luminance_data));
 out:
 	kfree(info);
 	return err;
@@ -452,7 +507,6 @@ static int amdgpu_atif_handler(struct amdgpu_device *adev,
 				pm_runtime_get_sync(adev_to_drm(adev)->dev);
 				/* Just fire off a uevent and let userspace tell us what to do */
 				drm_helper_hpd_irq_event(adev_to_drm(adev));
-				pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 				pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 			}
 		}
@@ -666,7 +720,7 @@ int amdgpu_acpi_pcie_performance_request(struct amdgpu_device *adev,
 
 	atcs_input.size = sizeof(struct atcs_pref_req_input);
 	/* client id (bit 2-0: func num, 7-3: dev num, 15-8: bus num) */
-	atcs_input.client_id = adev->pdev->devfn | (adev->pdev->bus->number << 8);
+	atcs_input.client_id = pci_dev_id(adev->pdev);
 	atcs_input.valid_flags_mask = ATCS_VALID_FLAGS_MASK;
 	atcs_input.flags = ATCS_WAIT_FOR_COMPLETION;
 	if (advertise)
@@ -736,7 +790,7 @@ int amdgpu_acpi_power_shift_control(struct amdgpu_device *adev,
 
 	atcs_input.size = sizeof(struct atcs_pwr_shift_input);
 	/* dGPU id (bit 2-0: func num, 7-3: dev num, 15-8: bus num) */
-	atcs_input.dgpu_id = adev->pdev->devfn | (adev->pdev->bus->number << 8);
+	atcs_input.dgpu_id = pci_dev_id(adev->pdev);
 	atcs_input.dev_acpi_state = dev_state;
 	atcs_input.drv_state = drv_state;
 
@@ -749,24 +803,25 @@ int amdgpu_acpi_power_shift_control(struct amdgpu_device *adev,
 		return -EIO;
 	}
 
+	kfree(info);
 	return 0;
 }
 
 /**
  * amdgpu_acpi_smart_shift_update - update dGPU device state to SBIOS
  *
- * @dev: drm_device pointer
+ * @adev: amdgpu device pointer
  * @ss_state: current smart shift event
  *
  * returns 0 on success,
  * otherwise return error number.
  */
-int amdgpu_acpi_smart_shift_update(struct drm_device *dev, enum amdgpu_ss ss_state)
+int amdgpu_acpi_smart_shift_update(struct amdgpu_device *adev,
+				   enum amdgpu_ss ss_state)
 {
-	struct amdgpu_device *adev = drm_to_adev(dev);
 	int r;
 
-	if (!amdgpu_device_supports_smart_shift(dev))
+	if (!amdgpu_device_supports_smart_shift(adev))
 		return 0;
 
 	switch (ss_state) {
@@ -801,6 +856,347 @@ int amdgpu_acpi_smart_shift_update(struct drm_device *dev, enum amdgpu_ss ss_sta
 	return r;
 }
 
+#ifdef CONFIG_ACPI_NUMA
+static inline uint64_t amdgpu_acpi_get_numa_size(int nid)
+{
+	/* This is directly using si_meminfo_node implementation as the
+	 * function is not exported.
+	 */
+	int zone_type;
+	uint64_t managed_pages = 0;
+
+	pg_data_t *pgdat = NODE_DATA(nid);
+
+	for (zone_type = 0; zone_type < MAX_NR_ZONES; zone_type++)
+		managed_pages +=
+			zone_managed_pages(&pgdat->node_zones[zone_type]);
+	return managed_pages * PAGE_SIZE;
+}
+
+static struct amdgpu_numa_info *amdgpu_acpi_get_numa_info(uint32_t pxm)
+{
+	struct amdgpu_numa_info *numa_info;
+	int nid;
+
+	numa_info = xa_load(&numa_info_xa, pxm);
+
+	if (!numa_info) {
+		struct sysinfo info;
+
+		numa_info = kzalloc(sizeof(*numa_info), GFP_KERNEL);
+		if (!numa_info)
+			return NULL;
+
+		nid = pxm_to_node(pxm);
+		numa_info->pxm = pxm;
+		numa_info->nid = nid;
+
+		if (numa_info->nid == NUMA_NO_NODE) {
+			si_meminfo(&info);
+			numa_info->size = info.totalram * info.mem_unit;
+		} else {
+			numa_info->size = amdgpu_acpi_get_numa_size(nid);
+		}
+		xa_store(&numa_info_xa, numa_info->pxm, numa_info, GFP_KERNEL);
+	}
+
+	return numa_info;
+}
+#endif
+
+/**
+ * amdgpu_acpi_get_node_id - obtain the NUMA node id for corresponding amdgpu
+ * acpi device handle
+ *
+ * @handle: acpi handle
+ * @numa_info: amdgpu_numa_info structure holding numa information
+ *
+ * Queries the ACPI interface to fetch the corresponding NUMA Node ID for a
+ * given amdgpu acpi device.
+ *
+ * Returns ACPI STATUS OK with Node ID on success or the corresponding failure reason
+ */
+static acpi_status amdgpu_acpi_get_node_id(acpi_handle handle,
+				    struct amdgpu_numa_info **numa_info)
+{
+#ifdef CONFIG_ACPI_NUMA
+	u64 pxm;
+	acpi_status status;
+
+	if (!numa_info)
+		return_ACPI_STATUS(AE_ERROR);
+
+	status = acpi_evaluate_integer(handle, "_PXM", NULL, &pxm);
+
+	if (ACPI_FAILURE(status))
+		return status;
+
+	*numa_info = amdgpu_acpi_get_numa_info(pxm);
+
+	if (!*numa_info)
+		return_ACPI_STATUS(AE_ERROR);
+
+	return_ACPI_STATUS(AE_OK);
+#else
+	return_ACPI_STATUS(AE_NOT_EXIST);
+#endif
+}
+
+static struct amdgpu_acpi_dev_info *amdgpu_acpi_get_dev(u32 sbdf)
+{
+	struct amdgpu_acpi_dev_info *acpi_dev;
+
+	if (list_empty(&amdgpu_acpi_dev_list))
+		return NULL;
+
+	list_for_each_entry(acpi_dev, &amdgpu_acpi_dev_list, list)
+		if (acpi_dev->sbdf == sbdf)
+			return acpi_dev;
+
+	return NULL;
+}
+
+static int amdgpu_acpi_dev_init(struct amdgpu_acpi_dev_info **dev_info,
+				struct amdgpu_acpi_xcc_info *xcc_info, u32 sbdf)
+{
+	struct amdgpu_acpi_dev_info *tmp;
+	union acpi_object *obj;
+	int ret = -ENOENT;
+
+	*dev_info = NULL;
+	tmp = kzalloc(sizeof(struct amdgpu_acpi_dev_info), GFP_KERNEL);
+	if (!tmp)
+		return -ENOMEM;
+
+	INIT_LIST_HEAD(&tmp->xcc_list);
+	INIT_LIST_HEAD(&tmp->list);
+	tmp->sbdf = sbdf;
+
+	obj = acpi_evaluate_dsm_typed(xcc_info->handle, &amd_xcc_dsm_guid, 0,
+				      AMD_XCC_DSM_GET_SUPP_MODE, NULL,
+				      ACPI_TYPE_INTEGER);
+
+	if (!obj) {
+		acpi_handle_debug(xcc_info->handle,
+				  "_DSM function %d evaluation failed",
+				  AMD_XCC_DSM_GET_SUPP_MODE);
+		ret = -ENOENT;
+		goto out;
+	}
+
+	tmp->supp_xcp_mode = obj->integer.value & 0xFFFF;
+	ACPI_FREE(obj);
+
+	obj = acpi_evaluate_dsm_typed(xcc_info->handle, &amd_xcc_dsm_guid, 0,
+				      AMD_XCC_DSM_GET_XCP_MODE, NULL,
+				      ACPI_TYPE_INTEGER);
+
+	if (!obj) {
+		acpi_handle_debug(xcc_info->handle,
+				  "_DSM function %d evaluation failed",
+				  AMD_XCC_DSM_GET_XCP_MODE);
+		ret = -ENOENT;
+		goto out;
+	}
+
+	tmp->xcp_mode = obj->integer.value & 0xFFFF;
+	tmp->mem_mode = (obj->integer.value >> 32) & 0xFFFF;
+	ACPI_FREE(obj);
+
+	/* Evaluate DSMs and fill XCC information */
+	obj = acpi_evaluate_dsm_typed(xcc_info->handle, &amd_xcc_dsm_guid, 0,
+				      AMD_XCC_DSM_GET_TMR_INFO, NULL,
+				      ACPI_TYPE_PACKAGE);
+
+	if (!obj || obj->package.count < 2) {
+		acpi_handle_debug(xcc_info->handle,
+				  "_DSM function %d evaluation failed",
+				  AMD_XCC_DSM_GET_TMR_INFO);
+		ret = -ENOENT;
+		goto out;
+	}
+
+	tmp->tmr_base = obj->package.elements[0].integer.value;
+	tmp->tmr_size = obj->package.elements[1].integer.value;
+	ACPI_FREE(obj);
+
+	DRM_DEBUG_DRIVER(
+		"New dev(%x): Supported xcp mode: %x curr xcp_mode : %x mem mode : %x, tmr base: %llx tmr size: %llx  ",
+		tmp->sbdf, tmp->supp_xcp_mode, tmp->xcp_mode, tmp->mem_mode,
+		tmp->tmr_base, tmp->tmr_size);
+	list_add_tail(&tmp->list, &amdgpu_acpi_dev_list);
+	*dev_info = tmp;
+
+	return 0;
+
+out:
+	if (obj)
+		ACPI_FREE(obj);
+	kfree(tmp);
+
+	return ret;
+}
+
+static int amdgpu_acpi_get_xcc_info(struct amdgpu_acpi_xcc_info *xcc_info,
+				    u32 *sbdf)
+{
+	union acpi_object *obj;
+	acpi_status status;
+	int ret = -ENOENT;
+
+	obj = acpi_evaluate_dsm_typed(xcc_info->handle, &amd_xcc_dsm_guid, 0,
+				      AMD_XCC_DSM_GET_NUM_FUNCS, NULL,
+				      ACPI_TYPE_INTEGER);
+
+	if (!obj || obj->integer.value != AMD_XCC_DSM_NUM_FUNCS)
+		goto out;
+	ACPI_FREE(obj);
+
+	/* Evaluate DSMs and fill XCC information */
+	obj = acpi_evaluate_dsm_typed(xcc_info->handle, &amd_xcc_dsm_guid, 0,
+				      AMD_XCC_DSM_GET_VF_XCC_MAPPING, NULL,
+				      ACPI_TYPE_INTEGER);
+
+	if (!obj) {
+		acpi_handle_debug(xcc_info->handle,
+				  "_DSM function %d evaluation failed",
+				  AMD_XCC_DSM_GET_VF_XCC_MAPPING);
+		ret = -EINVAL;
+		goto out;
+	}
+
+	/* PF xcc id [39:32] */
+	xcc_info->phy_id = (obj->integer.value >> 32) & 0xFF;
+	/* xcp node of this xcc [47:40] */
+	xcc_info->xcp_node = (obj->integer.value >> 40) & 0xFF;
+	/* PF domain of this xcc [31:16] */
+	*sbdf = (obj->integer.value) & 0xFFFF0000;
+	/* PF bus/dev/fn of this xcc [63:48] */
+	*sbdf |= (obj->integer.value >> 48) & 0xFFFF;
+	ACPI_FREE(obj);
+	obj = NULL;
+
+	status =
+		amdgpu_acpi_get_node_id(xcc_info->handle, &xcc_info->numa_info);
+
+	/* TODO: check if this check is required */
+	if (ACPI_SUCCESS(status))
+		ret = 0;
+out:
+	if (obj)
+		ACPI_FREE(obj);
+
+	return ret;
+}
+
+static int amdgpu_acpi_enumerate_xcc(void)
+{
+	struct amdgpu_acpi_dev_info *dev_info = NULL;
+	struct amdgpu_acpi_xcc_info *xcc_info;
+	struct acpi_device *acpi_dev;
+	char hid[ACPI_ID_LEN];
+	int ret, id;
+	u32 sbdf;
+
+	INIT_LIST_HEAD(&amdgpu_acpi_dev_list);
+	xa_init(&numa_info_xa);
+
+	for (id = 0; id < AMD_XCC_MAX_HID; id++) {
+		sprintf(hid, "%s%d", "AMD", AMD_XCC_HID_START + id);
+		acpi_dev = acpi_dev_get_first_match_dev(hid, NULL, -1);
+		/* These ACPI objects are expected to be in sequential order. If
+		 * one is not found, no need to check the rest.
+		 */
+		if (!acpi_dev) {
+			DRM_DEBUG_DRIVER("No matching acpi device found for %s",
+					 hid);
+			break;
+		}
+
+		xcc_info = kzalloc(sizeof(struct amdgpu_acpi_xcc_info),
+				   GFP_KERNEL);
+		if (!xcc_info) {
+			DRM_ERROR("Failed to allocate memory for xcc info\n");
+			return -ENOMEM;
+		}
+
+		INIT_LIST_HEAD(&xcc_info->list);
+		xcc_info->handle = acpi_device_handle(acpi_dev);
+		acpi_dev_put(acpi_dev);
+
+		ret = amdgpu_acpi_get_xcc_info(xcc_info, &sbdf);
+		if (ret) {
+			kfree(xcc_info);
+			continue;
+		}
+
+		dev_info = amdgpu_acpi_get_dev(sbdf);
+
+		if (!dev_info)
+			ret = amdgpu_acpi_dev_init(&dev_info, xcc_info, sbdf);
+
+		if (ret == -ENOMEM)
+			return ret;
+
+		if (!dev_info) {
+			kfree(xcc_info);
+			continue;
+		}
+
+		list_add_tail(&xcc_info->list, &dev_info->xcc_list);
+	}
+
+	return 0;
+}
+
+int amdgpu_acpi_get_tmr_info(struct amdgpu_device *adev, u64 *tmr_offset,
+			     u64 *tmr_size)
+{
+	struct amdgpu_acpi_dev_info *dev_info;
+	u32 sbdf;
+
+	if (!tmr_offset || !tmr_size)
+		return -EINVAL;
+
+	sbdf = (pci_domain_nr(adev->pdev->bus) << 16);
+	sbdf |= pci_dev_id(adev->pdev);
+	dev_info = amdgpu_acpi_get_dev(sbdf);
+	if (!dev_info)
+		return -ENOENT;
+
+	*tmr_offset = dev_info->tmr_base;
+	*tmr_size = dev_info->tmr_size;
+
+	return 0;
+}
+
+int amdgpu_acpi_get_mem_info(struct amdgpu_device *adev, int xcc_id,
+			     struct amdgpu_numa_info *numa_info)
+{
+	struct amdgpu_acpi_dev_info *dev_info;
+	struct amdgpu_acpi_xcc_info *xcc_info;
+	u32 sbdf;
+
+	if (!numa_info)
+		return -EINVAL;
+
+	sbdf = (pci_domain_nr(adev->pdev->bus) << 16);
+	sbdf |= pci_dev_id(adev->pdev);
+	dev_info = amdgpu_acpi_get_dev(sbdf);
+	if (!dev_info)
+		return -ENOENT;
+
+	list_for_each_entry(xcc_info, &dev_info->xcc_list, list) {
+		if (xcc_info->phy_id == xcc_id) {
+			memcpy(numa_info, xcc_info->numa_info,
+			       sizeof(*numa_info));
+			return 0;
+		}
+	}
+
+	return -ENOENT;
+}
+
 /**
  * amdgpu_acpi_event - handle notify events
  *
@@ -884,9 +1280,7 @@ void amdgpu_acpi_get_backlight_caps(struct amdgpu_dm_backlight_caps *caps)
 {
 	struct amdgpu_atif *atif = &amdgpu_acpi_priv.atif;
 
-	caps->caps_valid = atif->backlight_caps.caps_valid;
-	caps->min_input_signal = atif->backlight_caps.min_input_signal;
-	caps->max_input_signal = atif->backlight_caps.max_input_signal;
+	memcpy(caps, &atif->backlight_caps, sizeof(*caps));
 }
 
 /**
@@ -981,7 +1375,12 @@ static bool amdgpu_atcs_pci_probe_handle(struct pci_dev *pdev)
  */
 bool amdgpu_acpi_should_gpu_reset(struct amdgpu_device *adev)
 {
-	if (adev->flags & AMD_IS_APU)
+	if ((adev->flags & AMD_IS_APU) &&
+	    adev->gfx.imu.funcs) /* Not need to do mode2 reset for IMU enabled APUs */
+		return false;
+
+	if ((adev->flags & AMD_IS_APU) &&
+	    amdgpu_acpi_is_s3_active(adev))
 		return false;
 
 	if (amdgpu_sriov_vf(adev))
@@ -1007,14 +1406,11 @@ void amdgpu_acpi_detect(void)
 	struct pci_dev *pdev = NULL;
 	int ret;
 
-	while ((pdev = pci_get_class(PCI_CLASS_DISPLAY_VGA << 8, pdev)) != NULL) {
-		if (!atif->handle)
-			amdgpu_atif_pci_probe_handle(pdev);
-		if (!atcs->handle)
-			amdgpu_atcs_pci_probe_handle(pdev);
-	}
+	while ((pdev = pci_get_base_class(PCI_BASE_CLASS_DISPLAY, pdev))) {
+		if ((pdev->class != PCI_CLASS_DISPLAY_VGA << 8) &&
+		    (pdev->class != PCI_CLASS_DISPLAY_OTHER << 8))
+			continue;
 
-	while ((pdev = pci_get_class(PCI_CLASS_DISPLAY_OTHER << 8, pdev)) != NULL) {
 		if (!atif->handle)
 			amdgpu_atif_pci_probe_handle(pdev);
 		if (!atcs->handle)
@@ -1049,6 +1445,36 @@ void amdgpu_acpi_detect(void)
 	} else {
 		atif->backlight_caps.caps_valid = false;
 	}
+
+	amdgpu_acpi_enumerate_xcc();
+}
+
+void amdgpu_acpi_release(void)
+{
+	struct amdgpu_acpi_dev_info *dev_info, *dev_tmp;
+	struct amdgpu_acpi_xcc_info *xcc_info, *xcc_tmp;
+	struct amdgpu_numa_info *numa_info;
+	unsigned long index;
+
+	xa_for_each(&numa_info_xa, index, numa_info) {
+		kfree(numa_info);
+		xa_erase(&numa_info_xa, index);
+	}
+
+	if (list_empty(&amdgpu_acpi_dev_list))
+		return;
+
+	list_for_each_entry_safe(dev_info, dev_tmp, &amdgpu_acpi_dev_list,
+				 list) {
+		list_for_each_entry_safe(xcc_info, xcc_tmp, &dev_info->xcc_list,
+					 list) {
+			list_del(&xcc_info->list);
+			kfree(xcc_info);
+		}
+
+		list_del(&dev_info->list);
+		kfree(dev_info);
+	}
 }
 
 #if IS_ENABLED(CONFIG_SUSPEND)
@@ -1081,22 +1507,59 @@ bool amdgpu_acpi_is_s0ix_active(struct amdgpu_device *adev)
 	if (adev->asic_type < CHIP_RAVEN)
 		return false;
 
+	if (!(adev->pm.pp_feature & PP_GFXOFF_MASK))
+		return false;
+
 	/*
 	 * If ACPI_FADT_LOW_POWER_S0 is not set in the FADT, it is generally
 	 * risky to do any special firmware-related preparations for entering
 	 * S0ix even though the system is suspending to idle, so return false
 	 * in that case.
 	 */
-	if (!(acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0))
-		dev_warn_once(adev->dev,
+	if (!(acpi_gbl_FADT.flags & ACPI_FADT_LOW_POWER_S0)) {
+		dev_err_once(adev->dev,
 			      "Power consumption will be higher as BIOS has not been configured for suspend-to-idle.\n"
 			      "To use suspend-to-idle change the sleep mode in BIOS setup.\n");
+		return false;
+	}
 
 #if !IS_ENABLED(CONFIG_AMD_PMC)
-	dev_warn_once(adev->dev,
+	dev_err_once(adev->dev,
 		      "Power consumption will be higher as the kernel has not been compiled with CONFIG_AMD_PMC.\n");
-#endif /* CONFIG_AMD_PMC */
+	return false;
+#else
 	return true;
+#endif /* CONFIG_AMD_PMC */
 }
-
 #endif /* CONFIG_SUSPEND */
+
+#if IS_ENABLED(CONFIG_DRM_AMD_ISP)
+static const struct acpi_device_id isp_sensor_ids[] = {
+	{ "OMNI5C10" },
+	{ }
+};
+
+static int isp_match_acpi_device_ids(struct device *dev, const void *data)
+{
+	return acpi_match_device(data, dev) ? 1 : 0;
+}
+
+int amdgpu_acpi_get_isp4_dev(struct acpi_device **dev)
+{
+	struct device *pdev __free(put_device) = NULL;
+	struct acpi_device *acpi_pdev;
+
+	pdev = bus_find_device(&platform_bus_type, NULL, isp_sensor_ids,
+			       isp_match_acpi_device_ids);
+	if (!pdev)
+		return -EINVAL;
+
+	acpi_pdev = ACPI_COMPANION(pdev);
+	if (!acpi_pdev)
+		return -ENODEV;
+
+	*dev = acpi_pdev;
+
+	return 0;
+}
+#endif /* CONFIG_DRM_AMD_ISP */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_afmt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_afmt.c
index a4d65973bf7c..80771b1480ff 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_afmt.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_afmt.c
@@ -100,6 +100,7 @@ struct amdgpu_afmt_acr amdgpu_afmt_acr(uint32_t clock)
 	amdgpu_afmt_calc_cts(clock, &res.cts_32khz, &res.n_32khz, 32000);
 	amdgpu_afmt_calc_cts(clock, &res.cts_44_1khz, &res.n_44_1khz, 44100);
 	amdgpu_afmt_calc_cts(clock, &res.cts_48khz, &res.n_48khz, 48000);
+	res.clock = clock;
 
 	return res;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index f99d4873bf22..a2879d2b7c8e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -28,6 +28,7 @@
 #include "amdgpu.h"
 #include "amdgpu_gfx.h"
 #include "amdgpu_dma_buf.h"
+#include <drm/ttm/ttm_tt.h>
 #include <linux/module.h>
 #include <linux/dma-buf.h>
 #include "amdgpu_xgmi.h"
@@ -53,7 +54,6 @@ int amdgpu_amdkfd_init(void)
 	amdgpu_amdkfd_total_mem_size *= si.mem_unit;
 
 	ret = kgd2kfd_init();
-	amdgpu_amdkfd_gpuvm_init_mem_limits();
 	kfd_initialized = !ret;
 
 	return ret;
@@ -96,7 +96,7 @@ static void amdgpu_doorbell_get_kfd_info(struct amdgpu_device *adev,
 					 size_t *start_offset)
 {
 	/*
-	 * The first num_doorbells are used by amdgpu.
+	 * The first num_kernel_doorbells are used by amdgpu.
 	 * amdkfd takes whatever's left in the aperture.
 	 */
 	if (adev->enable_mes) {
@@ -109,11 +109,11 @@ static void amdgpu_doorbell_get_kfd_info(struct amdgpu_device *adev,
 		*aperture_base = adev->doorbell.base;
 		*aperture_size = 0;
 		*start_offset = 0;
-	} else if (adev->doorbell.size > adev->doorbell.num_doorbells *
+	} else if (adev->doorbell.size > adev->doorbell.num_kernel_doorbells *
 						sizeof(u32)) {
 		*aperture_base = adev->doorbell.base;
 		*aperture_size = adev->doorbell.size;
-		*start_offset = adev->doorbell.num_doorbells * sizeof(u32);
+		*start_offset = adev->doorbell.num_kernel_doorbells * sizeof(u32);
 	} else {
 		*aperture_base = 0;
 		*aperture_size = 0;
@@ -133,16 +133,45 @@ static void amdgpu_amdkfd_reset_work(struct work_struct *work)
 
 	reset_context.method = AMD_RESET_METHOD_NONE;
 	reset_context.reset_req_dev = adev;
+	reset_context.src = adev->enable_mes ?
+			    AMDGPU_RESET_SRC_MES :
+			    AMDGPU_RESET_SRC_HWS;
 	clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
 
 	amdgpu_device_gpu_recover(adev, NULL, &reset_context);
 }
 
+static const struct drm_client_funcs kfd_client_funcs = {
+	.unregister	= drm_client_release,
+};
+
+int amdgpu_amdkfd_drm_client_create(struct amdgpu_device *adev)
+{
+	int ret;
+
+	if (!adev->kfd.init_complete || adev->kfd.client.dev)
+		return 0;
+
+	ret = drm_client_init(&adev->ddev, &adev->kfd.client, "kfd",
+			      &kfd_client_funcs);
+	if (ret) {
+		dev_err(adev->dev, "Failed to init DRM client: %d\n",
+			ret);
+		return ret;
+	}
+
+	drm_client_register(&adev->kfd.client);
+
+	return 0;
+}
+
 void amdgpu_amdkfd_device_init(struct amdgpu_device *adev)
 {
 	int i;
 	int last_valid_bit;
 
+	amdgpu_amdkfd_gpuvm_init_mem_limits();
+
 	if (adev->kfd.dev) {
 		struct kgd2kfd_shared_resources gpu_resources = {
 			.compute_vmid_bitmap =
@@ -162,8 +191,8 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev)
 		 * clear
 		 */
 		bitmap_complement(gpu_resources.cp_queue_bitmap,
-				  adev->gfx.mec.queue_bitmap,
-				  KGD_MAX_QUEUES);
+				  adev->gfx.mec_bitmap[0].queue_bitmap,
+				  AMDGPU_MAX_QUEUES);
 
 		/* According to linux/bitmap.h we shouldn't use bitmap_clear if
 		 * nbits is not compile time constant
@@ -171,7 +200,7 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev)
 		last_valid_bit = 1 /* only first MEC can have compute queues */
 				* adev->gfx.mec.num_pipe_per_mec
 				* adev->gfx.mec.num_queue_per_pipe;
-		for (i = last_valid_bit; i < KGD_MAX_QUEUES; ++i)
+		for (i = last_valid_bit; i < AMDGPU_MAX_QUEUES; ++i)
 			clear_bit(i, gpu_resources.cp_queue_bitmap);
 
 		amdgpu_doorbell_get_kfd_info(adev,
@@ -219,38 +248,53 @@ void amdgpu_amdkfd_interrupt(struct amdgpu_device *adev,
 		kgd2kfd_interrupt(adev->kfd.dev, ih_ring_entry);
 }
 
-void amdgpu_amdkfd_suspend(struct amdgpu_device *adev, bool run_pm)
+void amdgpu_amdkfd_suspend(struct amdgpu_device *adev, bool suspend_proc)
 {
-	if (adev->kfd.dev)
-		kgd2kfd_suspend(adev->kfd.dev, run_pm);
+	if (adev->kfd.dev) {
+		if (adev->in_s0ix)
+			kgd2kfd_stop_sched_all_nodes(adev->kfd.dev);
+		else
+			kgd2kfd_suspend(adev->kfd.dev, suspend_proc);
+	}
 }
 
-int amdgpu_amdkfd_resume_iommu(struct amdgpu_device *adev)
+int amdgpu_amdkfd_resume(struct amdgpu_device *adev, bool resume_proc)
 {
 	int r = 0;
 
-	if (adev->kfd.dev)
-		r = kgd2kfd_resume_iommu(adev->kfd.dev);
+	if (adev->kfd.dev) {
+		if (adev->in_s0ix)
+			r = kgd2kfd_start_sched_all_nodes(adev->kfd.dev);
+		else
+			r = kgd2kfd_resume(adev->kfd.dev, resume_proc);
+	}
 
 	return r;
 }
 
-int amdgpu_amdkfd_resume(struct amdgpu_device *adev, bool run_pm)
+void amdgpu_amdkfd_suspend_process(struct amdgpu_device *adev)
+{
+	if (adev->kfd.dev)
+		kgd2kfd_suspend_process(adev->kfd.dev);
+}
+
+int amdgpu_amdkfd_resume_process(struct amdgpu_device *adev)
 {
 	int r = 0;
 
 	if (adev->kfd.dev)
-		r = kgd2kfd_resume(adev->kfd.dev, run_pm);
+		r = kgd2kfd_resume_process(adev->kfd.dev);
 
 	return r;
 }
 
-int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev)
+int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev,
+			    struct amdgpu_reset_context *reset_context)
 {
 	int r = 0;
 
 	if (adev->kfd.dev)
-		r = kgd2kfd_pre_reset(adev->kfd.dev);
+		r = kgd2kfd_pre_reset(adev->kfd.dev, reset_context);
 
 	return r;
 }
@@ -344,15 +388,18 @@ allocate_mem_reserve_bo_failed:
 	return r;
 }
 
-void amdgpu_amdkfd_free_gtt_mem(struct amdgpu_device *adev, void *mem_obj)
+void amdgpu_amdkfd_free_gtt_mem(struct amdgpu_device *adev, void **mem_obj)
 {
-	struct amdgpu_bo *bo = (struct amdgpu_bo *) mem_obj;
+	struct amdgpu_bo **bo = (struct amdgpu_bo **) mem_obj;
 
-	amdgpu_bo_reserve(bo, true);
-	amdgpu_bo_kunmap(bo);
-	amdgpu_bo_unpin(bo);
-	amdgpu_bo_unreserve(bo);
-	amdgpu_bo_unref(&(bo));
+	if (!bo || !*bo)
+		return;
+
+	(void)amdgpu_bo_reserve(*bo, true);
+	amdgpu_bo_kunmap(*bo);
+	amdgpu_bo_unpin(*bo);
+	amdgpu_bo_unreserve(*bo);
+	amdgpu_bo_unref(bo);
 }
 
 int amdgpu_amdkfd_alloc_gws(struct amdgpu_device *adev, size_t size,
@@ -427,14 +474,26 @@ uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
 }
 
 void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
-				      struct kfd_local_mem_info *mem_info)
+				      struct kfd_local_mem_info *mem_info,
+				      struct amdgpu_xcp *xcp)
 {
 	memset(mem_info, 0, sizeof(*mem_info));
 
-	mem_info->local_mem_size_public = adev->gmc.visible_vram_size;
-	mem_info->local_mem_size_private = adev->gmc.real_vram_size -
+	if (xcp) {
+		if (adev->gmc.real_vram_size == adev->gmc.visible_vram_size)
+			mem_info->local_mem_size_public =
+					KFD_XCP_MEMORY_SIZE(adev, xcp->id);
+		else
+			mem_info->local_mem_size_private =
+					KFD_XCP_MEMORY_SIZE(adev, xcp->id);
+	} else if (adev->apu_prefer_gtt) {
+		mem_info->local_mem_size_public = (ttm_tt_pages_limit() << PAGE_SHIFT);
+		mem_info->local_mem_size_private = 0;
+	} else {
+		mem_info->local_mem_size_public = adev->gmc.visible_vram_size;
+		mem_info->local_mem_size_private = adev->gmc.real_vram_size -
 						adev->gmc.visible_vram_size;
-
+	}
 	mem_info->vram_width = adev->gmc.vram_width;
 
 	pr_debug("Address base: %pap public 0x%llx private 0x%llx\n",
@@ -442,9 +501,7 @@ void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
 			mem_info->local_mem_size_public,
 			mem_info->local_mem_size_private);
 
-	if (amdgpu_sriov_vf(adev))
-		mem_info->mem_clk_max = adev->clock.default_mclk / 100;
-	else if (adev->pm.dpm_enabled) {
+	if (adev->pm.dpm_enabled) {
 		if (amdgpu_emu_mode == 1)
 			mem_info->mem_clk_max = 0;
 		else
@@ -463,41 +520,17 @@ uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct amdgpu_device *adev)
 uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct amdgpu_device *adev)
 {
 	/* the sclk is in quantas of 10kHz */
-	if (amdgpu_sriov_vf(adev))
-		return adev->clock.default_sclk / 100;
-	else if (adev->pm.dpm_enabled)
+	if (adev->pm.dpm_enabled)
 		return amdgpu_dpm_get_sclk(adev, false) / 100;
 	else
 		return 100;
 }
 
-void amdgpu_amdkfd_get_cu_info(struct amdgpu_device *adev, struct kfd_cu_info *cu_info)
-{
-	struct amdgpu_cu_info acu_info = adev->gfx.cu_info;
-
-	memset(cu_info, 0, sizeof(*cu_info));
-	if (sizeof(cu_info->cu_bitmap) != sizeof(acu_info.bitmap))
-		return;
-
-	cu_info->cu_active_number = acu_info.number;
-	cu_info->cu_ao_mask = acu_info.ao_cu_mask;
-	memcpy(&cu_info->cu_bitmap[0], &acu_info.bitmap[0],
-	       sizeof(acu_info.bitmap));
-	cu_info->num_shader_engines = adev->gfx.config.max_shader_engines;
-	cu_info->num_shader_arrays_per_engine = adev->gfx.config.max_sh_per_se;
-	cu_info->num_cu_per_sh = adev->gfx.config.max_cu_per_sh;
-	cu_info->simd_per_cu = acu_info.simd_per_cu;
-	cu_info->max_waves_per_simd = acu_info.max_waves_per_simd;
-	cu_info->wave_front_size = acu_info.wave_front_size;
-	cu_info->max_scratch_slots_per_cu = acu_info.max_scratch_slots_per_cu;
-	cu_info->lds_size = acu_info.lds_size;
-}
-
 int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd,
 				  struct amdgpu_device **dmabuf_adev,
 				  uint64_t *bo_size, void *metadata_buffer,
 				  size_t buffer_size, uint32_t *metadata_size,
-				  uint32_t *flags)
+				  uint32_t *flags, int8_t *xcp_id)
 {
 	struct dma_buf *dma_buf;
 	struct drm_gem_object *obj;
@@ -541,54 +574,14 @@ int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd,
 		if (bo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)
 			*flags |= KFD_IOC_ALLOC_MEM_FLAGS_PUBLIC;
 	}
+	if (xcp_id)
+		*xcp_id = bo->xcp_id;
 
 out_put:
 	dma_buf_put(dma_buf);
 	return r;
 }
 
-uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device *dst,
-					  struct amdgpu_device *src)
-{
-	struct amdgpu_device *peer_adev = src;
-	struct amdgpu_device *adev = dst;
-	int ret = amdgpu_xgmi_get_hops_count(adev, peer_adev);
-
-	if (ret < 0) {
-		DRM_ERROR("amdgpu: failed to get  xgmi hops count between node %d and %d. ret = %d\n",
-			adev->gmc.xgmi.physical_node_id,
-			peer_adev->gmc.xgmi.physical_node_id, ret);
-		ret = 0;
-	}
-	return  (uint8_t)ret;
-}
-
-int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct amdgpu_device *dst,
-					    struct amdgpu_device *src,
-					    bool is_min)
-{
-	struct amdgpu_device *adev = dst, *peer_adev;
-	int num_links;
-
-	if (adev->asic_type != CHIP_ALDEBARAN)
-		return 0;
-
-	if (src)
-		peer_adev = src;
-
-	/* num links returns 0 for indirect peers since indirect route is unknown. */
-	num_links = is_min ? 1 : amdgpu_xgmi_get_num_links(adev, peer_adev);
-	if (num_links < 0) {
-		DRM_ERROR("amdgpu: failed to get xgmi num links between node %d and %d. ret = %d\n",
-			adev->gmc.xgmi.physical_node_id,
-			peer_adev->gmc.xgmi.physical_node_id, num_links);
-		num_links = 0;
-	}
-
-	/* Aldebaran xGMI DPM is defeatured so assume x16 x 25Gbps for bandwidth. */
-	return (num_links * 16 * 25000)/BITS_PER_BYTE;
-}
-
 int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_min)
 {
 	int num_lanes_shift = (is_min ? ffs(adev->pm.pcie_mlw_mask) :
@@ -673,7 +666,7 @@ int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev,
 		goto err;
 	}
 
-	ret = amdgpu_job_alloc(adev, NULL, NULL, NULL, 1, &job);
+	ret = amdgpu_job_alloc(adev, NULL, NULL, NULL, 1, &job, 0);
 	if (ret)
 		goto err;
 
@@ -706,12 +699,20 @@ err:
 
 void amdgpu_amdkfd_set_compute_idle(struct amdgpu_device *adev, bool idle)
 {
-	/* Temporary workaround to fix issues observed in some
-	 * compute applications when GFXOFF is enabled on GFX11.
-	 */
-	if (IP_VERSION_MAJ(adev->ip_versions[GC_HWIP][0]) == 11) {
+	enum amd_powergating_state state = idle ? AMD_PG_STATE_GATE : AMD_PG_STATE_UNGATE;
+	if ((IP_VERSION_MAJ(amdgpu_ip_version(adev, GC_HWIP, 0)) == 11 &&
+	    ((adev->mes.kiq_version & AMDGPU_MES_VERSION_MASK) <= 64)) ||
+		(IP_VERSION_MAJ(amdgpu_ip_version(adev, GC_HWIP, 0)) == 12)) {
 		pr_debug("GFXOFF is %s\n", idle ? "enabled" : "disabled");
 		amdgpu_gfx_off_ctrl(adev, idle);
+	} else if ((IP_VERSION_MAJ(amdgpu_ip_version(adev, GC_HWIP, 0)) == 9) &&
+		(adev->flags & AMD_IS_APU)) {
+		/* Disable GFXOFF and PG. Temporary workaround
+		 * to fix some compute applications issue on GFX9.
+		 */
+		struct amdgpu_ip_block *gfx_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GFX);
+		if (gfx_block != NULL)
+			gfx_block->version->funcs->set_powergating_state((void *)gfx_block, state);
 	}
 	amdgpu_dpm_switch_power_profile(adev,
 					PP_SMC_POWER_PROFILE_COMPUTE,
@@ -726,47 +727,186 @@ bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid)
 	return false;
 }
 
-int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct amdgpu_device *adev,
-				     uint16_t vmid)
+bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev)
 {
-	if (adev->family == AMDGPU_FAMILY_AI) {
-		int i;
+	return adev->have_atomics_support;
+}
 
-		for (i = 0; i < adev->num_vmhubs; i++)
-			amdgpu_gmc_flush_gpu_tlb(adev, vmid, i, 0);
-	} else {
-		amdgpu_gmc_flush_gpu_tlb(adev, vmid, AMDGPU_GFXHUB_0, 0);
-	}
+void amdgpu_amdkfd_debug_mem_fence(struct amdgpu_device *adev)
+{
+	amdgpu_device_flush_hdp(adev, NULL);
+}
+
+bool amdgpu_amdkfd_is_fed(struct amdgpu_device *adev)
+{
+	return amdgpu_ras_get_fed_status(adev);
+}
+
+void amdgpu_amdkfd_ras_pasid_poison_consumption_handler(struct amdgpu_device *adev,
+				enum amdgpu_ras_block block, uint16_t pasid,
+				pasid_notify pasid_fn, void *data, uint32_t reset)
+{
+	amdgpu_umc_pasid_poison_handler(adev, block, pasid, pasid_fn, data, reset);
+}
+
+void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev,
+	enum amdgpu_ras_block block, uint32_t reset)
+{
+	amdgpu_umc_pasid_poison_handler(adev, block, 0, NULL, NULL, reset);
+}
+
+int amdgpu_amdkfd_send_close_event_drain_irq(struct amdgpu_device *adev,
+					uint32_t *payload)
+{
+	int ret;
+
+	/* Device or IH ring is not ready so bail. */
+	ret = amdgpu_ih_wait_on_checkpoint_process_ts(adev, &adev->irq.ih);
+	if (ret)
+		return ret;
+
+	/* Send payload to fence KFD interrupts */
+	amdgpu_amdkfd_interrupt(adev, payload);
 
 	return 0;
 }
 
-int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
-				      uint16_t pasid, enum TLB_FLUSH_TYPE flush_type)
+int amdgpu_amdkfd_check_and_lock_kfd(struct amdgpu_device *adev)
 {
-	bool all_hub = false;
+	return kgd2kfd_check_and_lock_kfd(adev->kfd.dev);
+}
 
-	if (adev->family == AMDGPU_FAMILY_AI ||
-	    adev->family == AMDGPU_FAMILY_RV)
-		all_hub = true;
+void amdgpu_amdkfd_unlock_kfd(struct amdgpu_device *adev)
+{
+	kgd2kfd_unlock_kfd(adev->kfd.dev);
+}
 
-	return amdgpu_gmc_flush_gpu_tlb_pasid(adev, pasid, flush_type, all_hub);
+
+u64 amdgpu_amdkfd_xcp_memory_size(struct amdgpu_device *adev, int xcp_id)
+{
+	s8 mem_id = KFD_XCP_MEM_ID(adev, xcp_id);
+	u64 tmp;
+
+	if (adev->gmc.num_mem_partitions && xcp_id >= 0 && mem_id >= 0) {
+		if (adev->gmc.is_app_apu && adev->gmc.num_mem_partitions == 1) {
+			/* In NPS1 mode, we should restrict the vram reporting
+			 * tied to the ttm_pages_limit which is 1/2 of the system
+			 * memory. For other partition modes, the HBM is uniformly
+			 * divided already per numa node reported. If user wants to
+			 * go beyond the default ttm limit and maximize the ROCm
+			 * allocations, they can go up to max ttm and sysmem limits.
+			 */
+
+			tmp = (ttm_tt_pages_limit() << PAGE_SHIFT) / num_online_nodes();
+		} else {
+			tmp = adev->gmc.mem_partitions[mem_id].size;
+		}
+		do_div(tmp, adev->xcp_mgr->num_xcp_per_mem_partition);
+		return ALIGN_DOWN(tmp, PAGE_SIZE);
+	} else if (adev->apu_prefer_gtt) {
+		return (ttm_tt_pages_limit() << PAGE_SHIFT);
+	} else {
+		return adev->gmc.real_vram_size;
+	}
 }
 
-bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev)
+int amdgpu_amdkfd_unmap_hiq(struct amdgpu_device *adev, u32 doorbell_off,
+			    u32 inst)
 {
-	return adev->have_atomics_support;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[inst];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	struct amdgpu_ring_funcs *ring_funcs;
+	struct amdgpu_ring *ring;
+	int r = 0;
+
+	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
+		return -EINVAL;
+
+	if (!kiq_ring->sched.ready || amdgpu_in_reset(adev))
+		return 0;
+
+	ring_funcs = kzalloc(sizeof(*ring_funcs), GFP_KERNEL);
+	if (!ring_funcs)
+		return -ENOMEM;
+
+	ring = kzalloc(sizeof(*ring), GFP_KERNEL);
+	if (!ring) {
+		r = -ENOMEM;
+		goto free_ring_funcs;
+	}
+
+	ring_funcs->type = AMDGPU_RING_TYPE_COMPUTE;
+	ring->doorbell_index = doorbell_off;
+	ring->funcs = ring_funcs;
+
+	spin_lock(&kiq->ring_lock);
+
+	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size)) {
+		spin_unlock(&kiq->ring_lock);
+		r = -ENOMEM;
+		goto free_ring;
+	}
+
+	kiq->pmf->kiq_unmap_queues(kiq_ring, ring, RESET_QUEUES, 0, 0);
+
+	/* Submit unmap queue packet */
+	amdgpu_ring_commit(kiq_ring);
+	/*
+	 * Ring test will do a basic scratch register change check. Just run
+	 * this to ensure that unmap queues that is submitted before got
+	 * processed successfully before returning.
+	 */
+	r = amdgpu_ring_test_helper(kiq_ring);
+
+	spin_unlock(&kiq->ring_lock);
+
+free_ring:
+	kfree(ring);
+
+free_ring_funcs:
+	kfree(ring_funcs);
+
+	return r;
+}
+
+/* Stop scheduling on KFD */
+int amdgpu_amdkfd_stop_sched(struct amdgpu_device *adev, uint32_t node_id)
+{
+	if (!adev->kfd.init_complete)
+		return 0;
+
+	return kgd2kfd_stop_sched(adev->kfd.dev, node_id);
 }
 
-void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev, bool reset)
+/* Start scheduling on KFD */
+int amdgpu_amdkfd_start_sched(struct amdgpu_device *adev, uint32_t node_id)
 {
-	amdgpu_umc_poison_handler(adev, reset);
+	if (!adev->kfd.init_complete)
+		return 0;
+
+	return kgd2kfd_start_sched(adev->kfd.dev, node_id);
 }
 
-bool amdgpu_amdkfd_ras_query_utcl2_poison_status(struct amdgpu_device *adev)
+/* check if there are KFD queues active */
+bool amdgpu_amdkfd_compute_active(struct amdgpu_device *adev, uint32_t node_id)
 {
-	if (adev->gfx.ras && adev->gfx.ras->query_utcl2_poison_status)
-		return adev->gfx.ras->query_utcl2_poison_status(adev);
-	else
+	if (!adev->kfd.init_complete)
 		return false;
+
+	return kgd2kfd_compute_active(adev->kfd.dev, node_id);
+}
+
+/* Config CGTT_SQ_CLK_CTRL */
+int amdgpu_amdkfd_config_sq_perfmon(struct amdgpu_device *adev, uint32_t xcp_id,
+	bool core_override_enable, bool reg_override_enable, bool perfmon_override_enable)
+{
+	int r;
+
+	if (!adev->kfd.init_complete)
+		return 0;
+
+	r = psp_config_sq_perfmon(&adev->psp, xcp_id, core_override_enable,
+					reg_override_enable, perfmon_override_enable);
+
+	return r;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
index 01ba3589b60a..8bdfcde2029b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
@@ -25,15 +25,18 @@
 #ifndef AMDGPU_AMDKFD_H_INCLUDED
 #define AMDGPU_AMDKFD_H_INCLUDED
 
+#include <linux/list.h>
 #include <linux/types.h>
 #include <linux/mm.h>
 #include <linux/kthread.h>
 #include <linux/workqueue.h>
 #include <linux/mmu_notifier.h>
+#include <linux/memremap.h>
 #include <kgd_kfd_interface.h>
-#include <drm/ttm/ttm_execbuf_util.h>
+#include <drm/drm_client.h>
 #include "amdgpu_sync.h"
 #include "amdgpu_vm.h"
+#include "amdgpu_xcp.h"
 
 extern uint64_t amdgpu_amdkfd_total_mem_size;
 
@@ -44,6 +47,8 @@ enum TLB_FLUSH_TYPE {
 };
 
 struct amdgpu_device;
+struct kfd_process_device;
+struct amdgpu_reset_context;
 
 enum kfd_mem_attachment_type {
 	KFD_MEM_ATT_SHARED,	/* Share kgd_mem->bo or another attachment's */
@@ -66,11 +71,10 @@ struct kgd_mem {
 	struct mutex lock;
 	struct amdgpu_bo *bo;
 	struct dma_buf *dmabuf;
-	struct hmm_range *range;
+	struct amdgpu_hmm_range *range;
 	struct list_head attachments;
 	/* protected by amdkfd_process_info.lock */
-	struct ttm_validate_buffer validate_list;
-	struct ttm_validate_buffer resv_list;
+	struct list_head validate_list;
 	uint32_t domain;
 	unsigned int mapped_to_gpu_memory;
 	uint64_t va;
@@ -82,6 +86,7 @@ struct kgd_mem {
 
 	struct amdgpu_sync sync;
 
+	uint32_t gem_handle;
 	bool aql_queue;
 	bool is_imported;
 };
@@ -97,10 +102,18 @@ struct amdgpu_amdkfd_fence {
 
 struct amdgpu_kfd_dev {
 	struct kfd_dev *dev;
-	int64_t vram_used;
-	uint64_t vram_used_aligned;
+	int64_t vram_used[MAX_XCP];
+	uint64_t vram_used_aligned[MAX_XCP];
 	bool init_complete;
 	struct work_struct reset_work;
+
+	/* Client for KFD BO GEM handle allocations */
+	struct drm_client_dev client;
+
+	/* HMM page migration MEMORY_DEVICE_PRIVATE mapping
+	 * Must be last --ends in a flexible-array member.
+	 */
+	struct dev_pagemap pgmap;
 };
 
 enum kgd_engine_type {
@@ -143,28 +156,28 @@ struct amdkfd_process_info {
 int amdgpu_amdkfd_init(void);
 void amdgpu_amdkfd_fini(void);
 
-void amdgpu_amdkfd_suspend(struct amdgpu_device *adev, bool run_pm);
-int amdgpu_amdkfd_resume_iommu(struct amdgpu_device *adev);
-int amdgpu_amdkfd_resume(struct amdgpu_device *adev, bool run_pm);
+void amdgpu_amdkfd_suspend(struct amdgpu_device *adev, bool suspend_proc);
+int amdgpu_amdkfd_resume(struct amdgpu_device *adev, bool resume_proc);
+void amdgpu_amdkfd_suspend_process(struct amdgpu_device *adev);
+int amdgpu_amdkfd_resume_process(struct amdgpu_device *adev);
 void amdgpu_amdkfd_interrupt(struct amdgpu_device *adev,
 			const void *ih_ring_entry);
 void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev);
 void amdgpu_amdkfd_device_init(struct amdgpu_device *adev);
 void amdgpu_amdkfd_device_fini_sw(struct amdgpu_device *adev);
+int amdgpu_amdkfd_check_and_lock_kfd(struct amdgpu_device *adev);
+void amdgpu_amdkfd_unlock_kfd(struct amdgpu_device *adev);
 int amdgpu_amdkfd_submit_ib(struct amdgpu_device *adev,
 				enum kgd_engine_type engine,
 				uint32_t vmid, uint64_t gpu_addr,
 				uint32_t *ib_cmd, uint32_t ib_len);
 void amdgpu_amdkfd_set_compute_idle(struct amdgpu_device *adev, bool idle);
 bool amdgpu_amdkfd_have_atomics_support(struct amdgpu_device *adev);
-int amdgpu_amdkfd_flush_gpu_tlb_vmid(struct amdgpu_device *adev,
-				uint16_t vmid);
-int amdgpu_amdkfd_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
-				uint16_t pasid, enum TLB_FLUSH_TYPE flush_type);
 
 bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid);
 
-int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev);
+int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev,
+			    struct amdgpu_reset_context *reset_context);
 
 int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev);
 
@@ -176,15 +189,20 @@ int amdgpu_queue_mask_bit_to_set_resource_bit(struct amdgpu_device *adev,
 struct amdgpu_amdkfd_fence *amdgpu_amdkfd_fence_create(u64 context,
 				struct mm_struct *mm,
 				struct svm_range_bo *svm_bo);
+
+int amdgpu_amdkfd_drm_client_create(struct amdgpu_device *adev);
 #if defined(CONFIG_DEBUG_FS)
 int kfd_debugfs_kfd_mem_limits(struct seq_file *m, void *data);
 #endif
 #if IS_ENABLED(CONFIG_HSA_AMD)
 bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm);
 struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f);
-int amdgpu_amdkfd_remove_fence_on_pt_pd_bos(struct amdgpu_bo *bo);
+void amdgpu_amdkfd_remove_all_eviction_fences(struct amdgpu_bo *bo);
 int amdgpu_amdkfd_evict_userptr(struct mmu_interval_notifier *mni,
 				unsigned long cur_seq, struct kgd_mem *mem);
+int amdgpu_amdkfd_bo_validate_and_fence(struct amdgpu_bo *bo,
+					uint32_t domain,
+					struct dma_fence *fence);
 #else
 static inline
 bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm)
@@ -199,9 +217,8 @@ struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f)
 }
 
 static inline
-int amdgpu_amdkfd_remove_fence_on_pt_pd_bos(struct amdgpu_bo *bo)
+void amdgpu_amdkfd_remove_all_eviction_fences(struct amdgpu_bo *bo)
 {
-	return 0;
 }
 
 static inline
@@ -210,12 +227,19 @@ int amdgpu_amdkfd_evict_userptr(struct mmu_interval_notifier *mni,
 {
 	return 0;
 }
+static inline
+int amdgpu_amdkfd_bo_validate_and_fence(struct amdgpu_bo *bo,
+					uint32_t domain,
+					struct dma_fence *fence)
+{
+	return 0;
+}
 #endif
 /* Shared API */
 int amdgpu_amdkfd_alloc_gtt_mem(struct amdgpu_device *adev, size_t size,
 				void **mem_obj, uint64_t *gpu_addr,
 				void **cpu_ptr, bool mqd_gfx9);
-void amdgpu_amdkfd_free_gtt_mem(struct amdgpu_device *adev, void *mem_obj);
+void amdgpu_amdkfd_free_gtt_mem(struct amdgpu_device *adev, void **mem_obj);
 int amdgpu_amdkfd_alloc_gws(struct amdgpu_device *adev, size_t size,
 				void **mem_obj);
 void amdgpu_amdkfd_free_gws(struct amdgpu_device *adev, void *mem_obj);
@@ -224,23 +248,27 @@ int amdgpu_amdkfd_remove_gws_from_process(void *info, void *mem);
 uint32_t amdgpu_amdkfd_get_fw_version(struct amdgpu_device *adev,
 				      enum kgd_engine_type type);
 void amdgpu_amdkfd_get_local_mem_info(struct amdgpu_device *adev,
-				      struct kfd_local_mem_info *mem_info);
+				      struct kfd_local_mem_info *mem_info,
+				      struct amdgpu_xcp *xcp);
 uint64_t amdgpu_amdkfd_get_gpu_clock_counter(struct amdgpu_device *adev);
 
 uint32_t amdgpu_amdkfd_get_max_engine_clock_in_mhz(struct amdgpu_device *adev);
-void amdgpu_amdkfd_get_cu_info(struct amdgpu_device *adev,
-			       struct kfd_cu_info *cu_info);
 int amdgpu_amdkfd_get_dmabuf_info(struct amdgpu_device *adev, int dma_buf_fd,
 				  struct amdgpu_device **dmabuf_adev,
 				  uint64_t *bo_size, void *metadata_buffer,
 				  size_t buffer_size, uint32_t *metadata_size,
-				  uint32_t *flags);
-uint8_t amdgpu_amdkfd_get_xgmi_hops_count(struct amdgpu_device *dst,
-					  struct amdgpu_device *src);
-int amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(struct amdgpu_device *dst,
-					    struct amdgpu_device *src,
-					    bool is_min);
+				  uint32_t *flags, int8_t *xcp_id);
 int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_min);
+int amdgpu_amdkfd_send_close_event_drain_irq(struct amdgpu_device *adev,
+					uint32_t *payload);
+int amdgpu_amdkfd_unmap_hiq(struct amdgpu_device *adev, u32 doorbell_off,
+				u32 inst);
+int amdgpu_amdkfd_start_sched(struct amdgpu_device *adev, uint32_t node_id);
+int amdgpu_amdkfd_stop_sched(struct amdgpu_device *adev, uint32_t node_id);
+int amdgpu_amdkfd_config_sq_perfmon(struct amdgpu_device *adev, uint32_t xcp_id,
+	bool core_override_enable, bool reg_override_enable, bool perfmon_override_enable);
+bool amdgpu_amdkfd_compute_active(struct amdgpu_device *adev, uint32_t node_id);
+
 
 /* Read user wptr from a specified user address space with page fault
  * disabled. The memory must be pinned and mapped to the hardware when
@@ -270,16 +298,13 @@ int amdgpu_amdkfd_get_pcie_bandwidth_mbytes(struct amdgpu_device *adev, bool is_
 	(&((struct amdgpu_fpriv *)					\
 		((struct drm_file *)(drm_priv))->driver_priv)->vm)
 
-int amdgpu_amdkfd_gpuvm_set_vm_pasid(struct amdgpu_device *adev,
-				     struct amdgpu_vm *avm, u32 pasid);
 int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev,
 					struct amdgpu_vm *avm,
 					void **process_info,
 					struct dma_fence **ef);
-void amdgpu_amdkfd_gpuvm_release_process_vm(struct amdgpu_device *adev,
-					void *drm_priv);
 uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv);
-size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev);
+size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev,
+					uint8_t xcp_id);
 int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 		struct amdgpu_device *adev, uint64_t va, uint64_t size,
 		void *drm_priv, struct kgd_mem **mem,
@@ -291,37 +316,52 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(struct amdgpu_device *adev,
 					  struct kgd_mem *mem, void *drm_priv);
 int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
 		struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv);
+int amdgpu_amdkfd_gpuvm_dmaunmap_mem(struct kgd_mem *mem, void *drm_priv);
 int amdgpu_amdkfd_gpuvm_sync_memory(
 		struct amdgpu_device *adev, struct kgd_mem *mem, bool intr);
 int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_mem *mem,
 					     void **kptr, uint64_t *size);
 void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct kgd_mem *mem);
 
-int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_device *adev, struct amdgpu_bo *bo);
+int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_bo *bo, struct amdgpu_bo **bo_gart);
 
 int amdgpu_amdkfd_gpuvm_restore_process_bos(void *process_info,
-					    struct dma_fence **ef);
+					    struct dma_fence __rcu **ef);
 int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct amdgpu_device *adev,
 					      struct kfd_vm_fault_info *info);
-int amdgpu_amdkfd_gpuvm_import_dmabuf(struct amdgpu_device *adev,
-				      struct dma_buf *dmabuf,
-				      uint64_t va, void *drm_priv,
-				      struct kgd_mem **mem, uint64_t *size,
-				      uint64_t *mmap_offset);
+int amdgpu_amdkfd_gpuvm_import_dmabuf_fd(struct amdgpu_device *adev, int fd,
+					 uint64_t va, void *drm_priv,
+					 struct kgd_mem **mem, uint64_t *size,
+					 uint64_t *mmap_offset);
 int amdgpu_amdkfd_gpuvm_export_dmabuf(struct kgd_mem *mem,
 				      struct dma_buf **dmabuf);
+void amdgpu_amdkfd_debug_mem_fence(struct amdgpu_device *adev);
 int amdgpu_amdkfd_get_tile_config(struct amdgpu_device *adev,
 				struct tile_config *config);
 void amdgpu_amdkfd_ras_poison_consumption_handler(struct amdgpu_device *adev,
-				bool reset);
-bool amdgpu_amdkfd_bo_mapped_to_dev(struct amdgpu_device *adev, struct kgd_mem *mem);
+			enum amdgpu_ras_block block, uint32_t reset);
+
+void amdgpu_amdkfd_ras_pasid_poison_consumption_handler(struct amdgpu_device *adev,
+			enum amdgpu_ras_block block, uint16_t pasid,
+			pasid_notify pasid_fn, void *data, uint32_t reset);
+
+bool amdgpu_amdkfd_is_fed(struct amdgpu_device *adev);
+bool amdgpu_amdkfd_bo_mapped_to_dev(void *drm_priv, struct kgd_mem *mem);
 void amdgpu_amdkfd_block_mmu_notifications(void *p);
 int amdgpu_amdkfd_criu_resume(void *p);
-bool amdgpu_amdkfd_ras_query_utcl2_poison_status(struct amdgpu_device *adev);
 int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
-		uint64_t size, u32 alloc_flag);
+		uint64_t size, u32 alloc_flag, int8_t xcp_id);
 void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
-		uint64_t size, u32 alloc_flag);
+		uint64_t size, u32 alloc_flag, int8_t xcp_id);
+
+u64 amdgpu_amdkfd_xcp_memory_size(struct amdgpu_device *adev, int xcp_id);
+
+#define KFD_XCP_MEM_ID(adev, xcp_id) \
+		((adev)->xcp_mgr && (xcp_id) >= 0 ?\
+		(adev)->xcp_mgr->xcp[(xcp_id)].mem_id : -1)
+
+#define KFD_XCP_MEMORY_SIZE(adev, xcp_id) amdgpu_amdkfd_xcp_memory_size((adev), (xcp_id))
+
 
 #if IS_ENABLED(CONFIG_HSA_AMD)
 void amdgpu_amdkfd_gpuvm_init_mem_limits(void);
@@ -352,6 +392,17 @@ void amdgpu_amdkfd_release_notify(struct amdgpu_bo *bo)
 {
 }
 #endif
+
+#if IS_ENABLED(CONFIG_HSA_AMD_SVM)
+int kgd2kfd_init_zone_device(struct amdgpu_device *adev);
+#else
+static inline
+int kgd2kfd_init_zone_device(struct amdgpu_device *adev)
+{
+	return 0;
+}
+#endif
+
 /* KGD2KFD callbacks */
 int kgd2kfd_quiesce_mm(struct mm_struct *mm, uint32_t trigger);
 int kgd2kfd_resume_mm(struct mm_struct *mm);
@@ -364,14 +415,26 @@ struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf);
 bool kgd2kfd_device_init(struct kfd_dev *kfd,
 			 const struct kgd2kfd_shared_resources *gpu_resources);
 void kgd2kfd_device_exit(struct kfd_dev *kfd);
-void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm);
-int kgd2kfd_resume_iommu(struct kfd_dev *kfd);
-int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm);
-int kgd2kfd_pre_reset(struct kfd_dev *kfd);
+void kgd2kfd_suspend(struct kfd_dev *kfd, bool suspend_proc);
+int kgd2kfd_resume(struct kfd_dev *kfd, bool resume_proc);
+void kgd2kfd_suspend_process(struct kfd_dev *kfd);
+int kgd2kfd_resume_process(struct kfd_dev *kfd);
+int kgd2kfd_pre_reset(struct kfd_dev *kfd,
+		      struct amdgpu_reset_context *reset_context);
 int kgd2kfd_post_reset(struct kfd_dev *kfd);
 void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry);
 void kgd2kfd_set_sram_ecc_flag(struct kfd_dev *kfd);
 void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask);
+int kgd2kfd_check_and_lock_kfd(struct kfd_dev *kfd);
+void kgd2kfd_unlock_kfd(struct kfd_dev *kfd);
+int kgd2kfd_start_sched(struct kfd_dev *kfd, uint32_t node_id);
+int kgd2kfd_start_sched_all_nodes(struct kfd_dev *kfd);
+int kgd2kfd_stop_sched(struct kfd_dev *kfd, uint32_t node_id);
+int kgd2kfd_stop_sched_all_nodes(struct kfd_dev *kfd);
+bool kgd2kfd_compute_active(struct kfd_dev *kfd, uint32_t node_id);
+bool kgd2kfd_vmfault_fast_path(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry,
+			       bool retry_fault);
+
 #else
 static inline int kgd2kfd_init(void)
 {
@@ -399,21 +462,26 @@ static inline void kgd2kfd_device_exit(struct kfd_dev *kfd)
 {
 }
 
-static inline void kgd2kfd_suspend(struct kfd_dev *kfd, bool run_pm)
+static inline void kgd2kfd_suspend(struct kfd_dev *kfd, bool suspend_proc)
 {
 }
 
-static int __maybe_unused kgd2kfd_resume_iommu(struct kfd_dev *kfd)
+static inline int kgd2kfd_resume(struct kfd_dev *kfd, bool resume_proc)
 {
 	return 0;
 }
 
-static inline int kgd2kfd_resume(struct kfd_dev *kfd, bool run_pm)
+static inline void kgd2kfd_suspend_process(struct kfd_dev *kfd)
+{
+}
+
+static inline int kgd2kfd_resume_process(struct kfd_dev *kfd)
 {
 	return 0;
 }
 
-static inline int kgd2kfd_pre_reset(struct kfd_dev *kfd)
+static inline int kgd2kfd_pre_reset(struct kfd_dev *kfd,
+				    struct amdgpu_reset_context *reset_context)
 {
 	return 0;
 }
@@ -437,5 +505,46 @@ static inline
 void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask)
 {
 }
+
+static inline int kgd2kfd_check_and_lock_kfd(struct kfd_dev *kfd)
+{
+	return 0;
+}
+
+static inline void kgd2kfd_unlock_kfd(struct kfd_dev *kfd)
+{
+}
+
+static inline int kgd2kfd_start_sched(struct kfd_dev *kfd, uint32_t node_id)
+{
+	return 0;
+}
+
+static inline int kgd2kfd_start_sched_all_nodes(struct kfd_dev *kfd)
+{
+	return 0;
+}
+
+static inline int kgd2kfd_stop_sched(struct kfd_dev *kfd, uint32_t node_id)
+{
+	return 0;
+}
+
+static inline int kgd2kfd_stop_sched_all_nodes(struct kfd_dev *kfd)
+{
+	return 0;
+}
+
+static inline bool kgd2kfd_compute_active(struct kfd_dev *kfd, uint32_t node_id)
+{
+	return false;
+}
+
+static inline bool kgd2kfd_vmfault_fast_path(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry,
+				      bool retry_fault)
+{
+	return false;
+}
+
 #endif
 #endif /* AMDGPU_AMDKFD_H_INCLUDED */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
index 4485bb29bec9..7e9f7a280c1b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.c
@@ -23,6 +23,145 @@
 #include "amdgpu_amdkfd.h"
 #include "amdgpu_amdkfd_arcturus.h"
 #include "amdgpu_amdkfd_gfx_v9.h"
+#include "amdgpu_amdkfd_aldebaran.h"
+#include "gc/gc_9_4_2_offset.h"
+#include "gc/gc_9_4_2_sh_mask.h"
+#include <uapi/linux/kfd_ioctl.h>
+
+/*
+ * Returns TRAP_EN, EXCP_EN and EXCP_REPLACE.
+ *
+ * restore_dbg_registers is ignored here but is a general interface requirement
+ * for devices that support GFXOFF and where the RLC save/restore list
+ * does not support hw registers for debugging i.e. the driver has to manually
+ * initialize the debug mode registers after it has disabled GFX off during the
+ * debug session.
+ */
+uint32_t kgd_aldebaran_enable_debug_trap(struct amdgpu_device *adev,
+					    bool restore_dbg_registers,
+					    uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, 0);
+
+	return data;
+}
+
+/* returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_aldebaran_disable_debug_trap(struct amdgpu_device *adev,
+						bool keep_trap_enabled,
+						uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, keep_trap_enabled);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, 0);
+
+	return data;
+}
+
+static int kgd_aldebaran_validate_trap_override_request(struct amdgpu_device *adev,
+							uint32_t trap_override,
+							uint32_t *trap_mask_supported)
+{
+	*trap_mask_supported &= KFD_DBG_TRAP_MASK_FP_INVALID |
+				KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+				KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+				KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+				KFD_DBG_TRAP_MASK_FP_INEXACT |
+				KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+				KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION;
+
+	if (trap_override != KFD_DBG_TRAP_OVERRIDE_OR &&
+			trap_override != KFD_DBG_TRAP_OVERRIDE_REPLACE)
+		return -EPERM;
+
+	return 0;
+}
+
+/* returns TRAP_EN, EXCP_EN and EXCP_RPLACE. */
+static uint32_t kgd_aldebaran_set_wave_launch_trap_override(struct amdgpu_device *adev,
+					uint32_t vmid,
+					uint32_t trap_override,
+					uint32_t trap_mask_bits,
+					uint32_t trap_mask_request,
+					uint32_t *trap_mask_prev,
+					uint32_t kfd_dbg_trap_cntl_prev)
+
+{
+	uint32_t data = 0;
+
+	*trap_mask_prev = REG_GET_FIELD(kfd_dbg_trap_cntl_prev, SPI_GDBG_PER_VMID_CNTL, EXCP_EN);
+	trap_mask_bits = (trap_mask_bits & trap_mask_request) |
+		(*trap_mask_prev & ~trap_mask_request);
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, trap_mask_bits);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, trap_override);
+
+	return data;
+}
+
+uint32_t kgd_aldebaran_set_wave_launch_mode(struct amdgpu_device *adev,
+					uint8_t wave_launch_mode,
+					uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, LAUNCH_MODE, wave_launch_mode);
+
+	return data;
+}
+
+#define TCP_WATCH_STRIDE (regTCP_WATCH1_ADDR_H - regTCP_WATCH0_ADDR_H)
+static uint32_t kgd_gfx_aldebaran_set_address_watch(
+					struct amdgpu_device *adev,
+					uint64_t watch_address,
+					uint32_t watch_address_mask,
+					uint32_t watch_id,
+					uint32_t watch_mode,
+					uint32_t debug_vmid,
+					uint32_t inst)
+{
+	uint32_t watch_address_high;
+	uint32_t watch_address_low;
+	uint32_t watch_address_cntl;
+
+	watch_address_cntl = 0;
+	watch_address_low = lower_32_bits(watch_address);
+	watch_address_high = upper_32_bits(watch_address) & 0xffff;
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MODE,
+			watch_mode);
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MASK,
+			watch_address_mask >> 6);
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VALID,
+			1);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_H) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_high);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_L) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_low);
+
+	return watch_address_cntl;
+}
 
 const struct kfd2kgd_calls aldebaran_kfd2kgd = {
 	.program_sh_mem_settings = kgd_gfx_v9_program_sh_mem_settings,
@@ -42,5 +181,17 @@ const struct kfd2kgd_calls aldebaran_kfd2kgd = {
 				kgd_gfx_v9_get_atc_vmid_pasid_mapping_info,
 	.set_vm_context_page_table_base = kgd_gfx_v9_set_vm_context_page_table_base,
 	.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
-	.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings
+	.enable_debug_trap = kgd_aldebaran_enable_debug_trap,
+	.disable_debug_trap = kgd_aldebaran_disable_debug_trap,
+	.validate_trap_override_request = kgd_aldebaran_validate_trap_override_request,
+	.set_wave_launch_trap_override = kgd_aldebaran_set_wave_launch_trap_override,
+	.set_wave_launch_mode = kgd_aldebaran_set_wave_launch_mode,
+	.set_address_watch = kgd_gfx_aldebaran_set_address_watch,
+	.clear_address_watch = kgd_gfx_v9_clear_address_watch,
+	.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
+	.build_dequeue_wait_counts_packet_info = kgd_gfx_v9_build_dequeue_wait_counts_packet_info,
+	.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings,
+	.hqd_get_pq_addr = kgd_gfx_v9_hqd_get_pq_addr,
+	.hqd_reset = kgd_gfx_v9_hqd_reset,
+	.hqd_sdma_get_doorbell = kgd_gfx_v9_hqd_sdma_get_doorbell
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.h
new file mode 100644
index 000000000000..a7bdaf8d82dd
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_aldebaran.h
@@ -0,0 +1,27 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+uint32_t kgd_aldebaran_enable_debug_trap(struct amdgpu_device *adev,
+					bool restore_dbg_registers,
+					uint32_t vmid);
+uint32_t kgd_aldebaran_set_wave_launch_mode(struct amdgpu_device *adev,
+					uint8_t wave_launch_mode,
+					uint32_t vmid);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
index 4191af5a3f13..1105a09e55dc 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_arcturus.c
@@ -20,12 +20,12 @@
  * OTHER DEALINGS IN THE SOFTWARE.
  */
 #include <linux/module.h>
-#include <linux/fdtable.h>
 #include <linux/uaccess.h>
 #include <linux/firmware.h>
 #include "amdgpu.h"
 #include "amdgpu_amdkfd.h"
 #include "amdgpu_amdkfd_arcturus.h"
+#include "amdgpu_reset.h"
 #include "sdma0/sdma0_4_2_2_offset.h"
 #include "sdma0/sdma0_4_2_2_sh_mask.h"
 #include "sdma1/sdma1_4_2_2_offset.h"
@@ -48,6 +48,8 @@
 #include "amdgpu_amdkfd_gfx_v9.h"
 #include "gfxhub_v1_0.h"
 #include "mmhub_v9_4.h"
+#include "gc/gc_9_0_offset.h"
+#include "gc/gc_9_0_sh_mask.h"
 
 #define HQD_N_REGS 56
 #define DUMP_REG(addr) do {				\
@@ -197,7 +199,7 @@ int kgd_arcturus_hqd_sdma_dump(struct amdgpu_device *adev,
 #undef HQD_N_REGS
 #define HQD_N_REGS (19+6+7+10)
 
-	*dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -276,6 +278,117 @@ int kgd_arcturus_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
 	return 0;
 }
 
+/*
+ * Helper used to suspend/resume gfx pipe for image post process work to set
+ * barrier behaviour.
+ */
+static int suspend_resume_compute_scheduler(struct amdgpu_device *adev, bool suspend)
+{
+	int i, r = 0;
+
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		struct amdgpu_ring *ring = &adev->gfx.compute_ring[i];
+
+		if (!amdgpu_ring_sched_ready(ring))
+			continue;
+
+		/* stop secheduler and drain ring. */
+		if (suspend) {
+			drm_sched_stop(&ring->sched, NULL);
+			r = amdgpu_fence_wait_empty(ring);
+			if (r)
+				goto out;
+		} else {
+			drm_sched_start(&ring->sched, 0);
+		}
+	}
+
+out:
+	/* return on resume or failure to drain rings. */
+	if (!suspend || r)
+		return r;
+
+	return amdgpu_device_ip_wait_for_idle(adev, AMD_IP_BLOCK_TYPE_GFX);
+}
+
+static void set_barrier_auto_waitcnt(struct amdgpu_device *adev, bool enable_waitcnt)
+{
+	uint32_t data;
+
+	WRITE_ONCE(adev->barrier_has_auto_waitcnt, enable_waitcnt);
+
+	if (!down_read_trylock(&adev->reset_domain->sem))
+		return;
+
+	amdgpu_amdkfd_suspend(adev, true);
+
+	if (suspend_resume_compute_scheduler(adev, true))
+		goto out;
+
+	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_CONFIG));
+	data = REG_SET_FIELD(data, SQ_CONFIG, DISABLE_BARRIER_WAITCNT,
+						!enable_waitcnt);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_CONFIG), data);
+
+out:
+	suspend_resume_compute_scheduler(adev, false);
+
+	amdgpu_amdkfd_resume(adev, true);
+
+	up_read(&adev->reset_domain->sem);
+}
+
+/*
+ * restore_dbg_registers is ignored here but is a general interface requirement
+ * for devices that support GFXOFF and where the RLC save/restore list
+ * does not support hw registers for debugging i.e. the driver has to manually
+ * initialize the debug mode registers after it has disabled GFX off during the
+ * debug session.
+ */
+static uint32_t kgd_arcturus_enable_debug_trap(struct amdgpu_device *adev,
+				bool restore_dbg_registers,
+				uint32_t vmid)
+{
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+	set_barrier_auto_waitcnt(adev, true);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+/*
+ * keep_trap_enabled is ignored here but is a general interface requirement
+ * for devices that support multi-process debugging where the performance
+ * overhead from trap temporary setup needs to be bypassed when the debug
+ * session has ended.
+ */
+static uint32_t kgd_arcturus_disable_debug_trap(struct amdgpu_device *adev,
+					bool keep_trap_enabled,
+					uint32_t vmid)
+{
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+	set_barrier_auto_waitcnt(adev, false);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
 const struct kfd2kgd_calls arcturus_kfd2kgd = {
 	.program_sh_mem_settings = kgd_gfx_v9_program_sh_mem_settings,
 	.set_pasid_vmid_mapping = kgd_gfx_v9_set_pasid_vmid_mapping,
@@ -294,6 +407,18 @@ const struct kfd2kgd_calls arcturus_kfd2kgd = {
 				kgd_gfx_v9_get_atc_vmid_pasid_mapping_info,
 	.set_vm_context_page_table_base =
 				kgd_gfx_v9_set_vm_context_page_table_base,
+	.enable_debug_trap = kgd_arcturus_enable_debug_trap,
+	.disable_debug_trap = kgd_arcturus_disable_debug_trap,
+	.validate_trap_override_request = kgd_gfx_v9_validate_trap_override_request,
+	.set_wave_launch_trap_override = kgd_gfx_v9_set_wave_launch_trap_override,
+	.set_wave_launch_mode = kgd_gfx_v9_set_wave_launch_mode,
+	.set_address_watch = kgd_gfx_v9_set_address_watch,
+	.clear_address_watch = kgd_gfx_v9_clear_address_watch,
+	.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
+	.build_dequeue_wait_counts_packet_info = kgd_gfx_v9_build_dequeue_wait_counts_packet_info,
 	.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
-	.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings
+	.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings,
+	.hqd_get_pq_addr = kgd_gfx_v9_hqd_get_pq_addr,
+	.hqd_reset = kgd_gfx_v9_hqd_reset,
+	.hqd_sdma_get_doorbell = kgd_gfx_v9_hqd_sdma_get_doorbell
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c
index 469785d33791..1ef758ac5076 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c
@@ -90,7 +90,7 @@ struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f)
 		return NULL;
 
 	fence = container_of(f, struct amdgpu_amdkfd_fence, base);
-	if (fence && f->ops == &amdkfd_fence_ops)
+	if (f->ops == &amdkfd_fence_ops)
 		return fence;
 
 	return NULL;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c
new file mode 100644
index 000000000000..89a45a9218f3
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c
@@ -0,0 +1,559 @@
+/*
+ * Copyright 2021 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+#include "amdgpu.h"
+#include "amdgpu_amdkfd.h"
+#include "amdgpu_amdkfd_gfx_v9.h"
+#include "amdgpu_amdkfd_aldebaran.h"
+#include "gc/gc_9_4_3_offset.h"
+#include "gc/gc_9_4_3_sh_mask.h"
+#include "athub/athub_1_8_0_offset.h"
+#include "athub/athub_1_8_0_sh_mask.h"
+#include "oss/osssys_4_4_2_offset.h"
+#include "oss/osssys_4_4_2_sh_mask.h"
+#include "v9_structs.h"
+#include "soc15.h"
+#include "sdma/sdma_4_4_2_offset.h"
+#include "sdma/sdma_4_4_2_sh_mask.h"
+#include <uapi/linux/kfd_ioctl.h>
+
+static inline struct v9_sdma_mqd *get_sdma_mqd(void *mqd)
+{
+	return (struct v9_sdma_mqd *)mqd;
+}
+
+static uint32_t get_sdma_rlc_reg_offset(struct amdgpu_device *adev,
+					unsigned int engine_id,
+					unsigned int queue_id)
+{
+	uint32_t sdma_engine_reg_base =
+		SOC15_REG_OFFSET(SDMA0, GET_INST(SDMA0, engine_id),
+				 regSDMA_RLC0_RB_CNTL) -
+		regSDMA_RLC0_RB_CNTL;
+	uint32_t retval = sdma_engine_reg_base +
+		  queue_id * (regSDMA_RLC1_RB_CNTL - regSDMA_RLC0_RB_CNTL);
+
+	pr_debug("RLC register offset for SDMA%d RLC%d: 0x%x\n", engine_id,
+							queue_id, retval);
+	return retval;
+}
+
+static int kgd_gfx_v9_4_3_hqd_sdma_load(struct amdgpu_device *adev, void *mqd,
+				 uint32_t __user *wptr, struct mm_struct *mm)
+{
+	struct v9_sdma_mqd *m;
+	uint32_t sdma_rlc_reg_offset;
+	unsigned long end_jiffies;
+	uint32_t data;
+	uint64_t data64;
+	uint64_t __user *wptr64 = (uint64_t __user *)wptr;
+
+	m = get_sdma_mqd(mqd);
+	sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, m->sdma_engine_id,
+							m->sdma_queue_id);
+
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_CNTL,
+		m->sdmax_rlcx_rb_cntl & (~SDMA_RLC0_RB_CNTL__RB_ENABLE_MASK));
+
+	end_jiffies = msecs_to_jiffies(2000) + jiffies;
+	while (true) {
+		data = RREG32(sdma_rlc_reg_offset + regSDMA_RLC0_CONTEXT_STATUS);
+		if (data & SDMA_RLC0_CONTEXT_STATUS__IDLE_MASK)
+			break;
+		if (time_after(jiffies, end_jiffies)) {
+			pr_err("SDMA RLC not idle in %s\n", __func__);
+			return -ETIME;
+		}
+		usleep_range(500, 1000);
+	}
+
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_DOORBELL_OFFSET,
+		m->sdmax_rlcx_doorbell_offset);
+
+	data = REG_SET_FIELD(m->sdmax_rlcx_doorbell, SDMA_RLC0_DOORBELL,
+				ENABLE, 1);
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_DOORBELL, data);
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_RPTR,
+					m->sdmax_rlcx_rb_rptr);
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_RPTR_HI,
+					m->sdmax_rlcx_rb_rptr_hi);
+
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_MINOR_PTR_UPDATE, 1);
+	if (read_user_wptr(mm, wptr64, data64)) {
+		WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_WPTR,
+			lower_32_bits(data64));
+		WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_WPTR_HI,
+			upper_32_bits(data64));
+	} else {
+		WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_WPTR,
+			m->sdmax_rlcx_rb_rptr);
+		WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_WPTR_HI,
+			m->sdmax_rlcx_rb_rptr_hi);
+	}
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_MINOR_PTR_UPDATE, 0);
+
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_BASE, m->sdmax_rlcx_rb_base);
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_BASE_HI,
+			m->sdmax_rlcx_rb_base_hi);
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_RPTR_ADDR_LO,
+			m->sdmax_rlcx_rb_rptr_addr_lo);
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_RPTR_ADDR_HI,
+			m->sdmax_rlcx_rb_rptr_addr_hi);
+
+	data = REG_SET_FIELD(m->sdmax_rlcx_rb_cntl, SDMA_RLC0_RB_CNTL,
+				RB_ENABLE, 1);
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_CNTL, data);
+
+	return 0;
+}
+
+static int kgd_gfx_v9_4_3_hqd_sdma_dump(struct amdgpu_device *adev,
+				 uint32_t engine_id, uint32_t queue_id,
+				 uint32_t (**dump)[2], uint32_t *n_regs)
+{
+	uint32_t sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev,
+							engine_id, queue_id);
+	uint32_t i = 0, reg;
+#undef HQD_N_REGS
+#define HQD_N_REGS (19+6+7+12)
+#define DUMP_REG(addr) do {				\
+		if (WARN_ON_ONCE(i >= HQD_N_REGS))      \
+			break;				\
+		(*dump)[i][0] = (addr) << 2;            \
+		(*dump)[i++][1] = RREG32(addr);         \
+	} while (0)
+
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
+	if (*dump == NULL)
+		return -ENOMEM;
+
+	for (reg = regSDMA_RLC0_RB_CNTL; reg <= regSDMA_RLC0_DOORBELL; reg++)
+		DUMP_REG(sdma_rlc_reg_offset + reg);
+	for (reg = regSDMA_RLC0_STATUS; reg <= regSDMA_RLC0_CSA_ADDR_HI; reg++)
+		DUMP_REG(sdma_rlc_reg_offset + reg);
+	for (reg = regSDMA_RLC0_IB_SUB_REMAIN;
+	     reg <= regSDMA_RLC0_MINOR_PTR_UPDATE; reg++)
+		DUMP_REG(sdma_rlc_reg_offset + reg);
+	for (reg = regSDMA_RLC0_MIDCMD_DATA0;
+	     reg <= regSDMA_RLC0_MIDCMD_CNTL; reg++)
+		DUMP_REG(sdma_rlc_reg_offset + reg);
+
+	WARN_ON_ONCE(i != HQD_N_REGS);
+	*n_regs = i;
+
+	return 0;
+}
+
+static bool kgd_gfx_v9_4_3_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd)
+{
+	struct v9_sdma_mqd *m;
+	uint32_t sdma_rlc_reg_offset;
+	uint32_t sdma_rlc_rb_cntl;
+
+	m = get_sdma_mqd(mqd);
+	sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, m->sdma_engine_id,
+							m->sdma_queue_id);
+
+	sdma_rlc_rb_cntl = RREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_CNTL);
+
+	if (sdma_rlc_rb_cntl & SDMA_RLC0_RB_CNTL__RB_ENABLE_MASK)
+		return true;
+
+	return false;
+}
+
+static int kgd_gfx_v9_4_3_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
+				    unsigned int utimeout)
+{
+	struct v9_sdma_mqd *m;
+	uint32_t sdma_rlc_reg_offset;
+	uint32_t temp;
+	unsigned long end_jiffies = (utimeout * HZ / 1000) + jiffies;
+
+	m = get_sdma_mqd(mqd);
+	sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev, m->sdma_engine_id,
+							m->sdma_queue_id);
+
+	temp = RREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_CNTL);
+	temp = temp & ~SDMA_RLC0_RB_CNTL__RB_ENABLE_MASK;
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_CNTL, temp);
+
+	while (true) {
+		temp = RREG32(sdma_rlc_reg_offset + regSDMA_RLC0_CONTEXT_STATUS);
+		if (temp & SDMA_RLC0_CONTEXT_STATUS__IDLE_MASK)
+			break;
+		if (time_after(jiffies, end_jiffies)) {
+			pr_err("SDMA RLC not idle in %s\n", __func__);
+			return -ETIME;
+		}
+		usleep_range(500, 1000);
+	}
+
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_DOORBELL, 0);
+	WREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_CNTL,
+		RREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_CNTL) |
+		SDMA_RLC0_RB_CNTL__RB_ENABLE_MASK);
+
+	m->sdmax_rlcx_rb_rptr =
+			RREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_RPTR);
+	m->sdmax_rlcx_rb_rptr_hi =
+			RREG32(sdma_rlc_reg_offset + regSDMA_RLC0_RB_RPTR_HI);
+
+	return 0;
+}
+
+static int kgd_gfx_v9_4_3_set_pasid_vmid_mapping(struct amdgpu_device *adev,
+			u32 pasid, unsigned int vmid, uint32_t xcc_inst)
+{
+	unsigned long timeout;
+	unsigned int reg;
+	unsigned int phy_inst = GET_INST(GC, xcc_inst);
+	/* Every two XCCs share one AID */
+	unsigned int aid = phy_inst / 2;
+
+	/*
+	 * We have to assume that there is no outstanding mapping.
+	 * The ATC_VMID_PASID_MAPPING_UPDATE_STATUS bit could be 0 because
+	 * a mapping is in progress or because a mapping finished
+	 * and the SW cleared it.
+	 * So the protocol is to always wait & clear.
+	 */
+	uint32_t pasid_mapping = (pasid == 0) ? 0 : (uint32_t)pasid |
+			ATC_VMID0_PASID_MAPPING__VALID_MASK;
+
+	WREG32(SOC15_REG_OFFSET(ATHUB, 0,
+		regATC_VMID0_PASID_MAPPING) + vmid, pasid_mapping);
+
+	timeout = jiffies + msecs_to_jiffies(10);
+	while (!(RREG32(SOC15_REG_OFFSET(ATHUB, 0,
+			regATC_VMID_PASID_MAPPING_UPDATE_STATUS)) &
+			(1U << vmid))) {
+		if (time_after(jiffies, timeout)) {
+			pr_err("Fail to program VMID-PASID mapping\n");
+			return -ETIME;
+		}
+		cpu_relax();
+	}
+
+	WREG32(SOC15_REG_OFFSET(ATHUB, 0,
+		regATC_VMID_PASID_MAPPING_UPDATE_STATUS),
+		1U << vmid);
+
+	reg = RREG32(SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_LUT_INDEX));
+	/* Every 4 numbers is a cycle. 1st is AID, 2nd and 3rd are XCDs,
+	 * and the 4th is reserved. Therefore "aid * 4 + (xcc_inst % 2) + 1"
+	 * programs _LUT for XCC and "aid * 4" for AID where the XCC connects
+	 * to.
+	 */
+	WREG32(SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_LUT_INDEX),
+		aid * 4 + (phy_inst % 2) + 1);
+	WREG32(SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT) + vmid,
+		pasid_mapping);
+	WREG32(SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_LUT_INDEX),
+		aid * 4);
+	WREG32(SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT_MM) + vmid,
+		pasid_mapping);
+	WREG32(SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_LUT_INDEX), reg);
+
+	return 0;
+}
+
+static inline struct v9_mqd *get_mqd(void *mqd)
+{
+	return (struct v9_mqd *)mqd;
+}
+
+static int kgd_gfx_v9_4_3_hqd_load(struct amdgpu_device *adev, void *mqd,
+			uint32_t pipe_id, uint32_t queue_id,
+			uint32_t __user *wptr, uint32_t wptr_shift,
+			uint32_t wptr_mask, struct mm_struct *mm, uint32_t inst)
+{
+	struct v9_mqd *m;
+	uint32_t *mqd_hqd;
+	uint32_t reg, hqd_base, hqd_end, data;
+
+	m = get_mqd(mqd);
+
+	kgd_gfx_v9_acquire_queue(adev, pipe_id, queue_id, inst);
+
+	/* HQD registers extend to CP_HQD_AQL_DISPATCH_ID_HI */
+	mqd_hqd = &m->cp_mqd_base_addr_lo;
+	hqd_base = SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_MQD_BASE_ADDR);
+	hqd_end = SOC15_REG_OFFSET(GC, GET_INST(GC, inst), regCP_HQD_AQL_DISPATCH_ID_HI);
+
+	for (reg = hqd_base; reg <= hqd_end; reg++)
+		WREG32_XCC(reg, mqd_hqd[reg - hqd_base], inst);
+
+
+	/* Activate doorbell logic before triggering WPTR poll. */
+	data = REG_SET_FIELD(m->cp_hqd_pq_doorbell_control,
+			     CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), regCP_HQD_PQ_DOORBELL_CONTROL, data);
+
+	if (wptr) {
+		/* Don't read wptr with get_user because the user
+		 * context may not be accessible (if this function
+		 * runs in a work queue). Instead trigger a one-shot
+		 * polling read from memory in the CP. This assumes
+		 * that wptr is GPU-accessible in the queue's VMID via
+		 * ATC or SVM. WPTR==RPTR before starting the poll so
+		 * the CP starts fetching new commands from the right
+		 * place.
+		 *
+		 * Guessing a 64-bit WPTR from a 32-bit RPTR is a bit
+		 * tricky. Assume that the queue didn't overflow. The
+		 * number of valid bits in the 32-bit RPTR depends on
+		 * the queue size. The remaining bits are taken from
+		 * the saved 64-bit WPTR. If the WPTR wrapped, add the
+		 * queue size.
+		 */
+		uint32_t queue_size =
+			2 << REG_GET_FIELD(m->cp_hqd_pq_control,
+					   CP_HQD_PQ_CONTROL, QUEUE_SIZE);
+		uint64_t guessed_wptr = m->cp_hqd_pq_rptr & (queue_size - 1);
+
+		if ((m->cp_hqd_pq_wptr_lo & (queue_size - 1)) < guessed_wptr)
+			guessed_wptr += queue_size;
+		guessed_wptr += m->cp_hqd_pq_wptr_lo & ~(queue_size - 1);
+		guessed_wptr += (uint64_t)m->cp_hqd_pq_wptr_hi << 32;
+
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), regCP_HQD_PQ_WPTR_LO,
+			lower_32_bits(guessed_wptr));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), regCP_HQD_PQ_WPTR_HI,
+			upper_32_bits(guessed_wptr));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), regCP_HQD_PQ_WPTR_POLL_ADDR,
+			lower_32_bits((uintptr_t)wptr));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), regCP_HQD_PQ_WPTR_POLL_ADDR_HI,
+			upper_32_bits((uintptr_t)wptr));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), regCP_PQ_WPTR_POLL_CNTL1,
+			(uint32_t)kgd_gfx_v9_get_queue_mask(adev, pipe_id, queue_id));
+	}
+
+	/* Start the EOP fetcher */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), regCP_HQD_EOP_RPTR,
+	       REG_SET_FIELD(m->cp_hqd_eop_rptr, CP_HQD_EOP_RPTR, INIT_FETCHER, 1));
+
+	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), regCP_HQD_ACTIVE, data);
+
+	kgd_gfx_v9_release_queue(adev, inst);
+
+	return 0;
+}
+
+/* returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_gfx_v9_4_3_disable_debug_trap(struct amdgpu_device *adev,
+						bool keep_trap_enabled,
+						uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, 0);
+
+	return data;
+}
+
+static int kgd_gfx_v9_4_3_validate_trap_override_request(
+				struct amdgpu_device *adev,
+				uint32_t trap_override,
+				uint32_t *trap_mask_supported)
+{
+	*trap_mask_supported &= KFD_DBG_TRAP_MASK_FP_INVALID |
+				KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+				KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+				KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+				KFD_DBG_TRAP_MASK_FP_INEXACT |
+				KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+				KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION |
+				KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START |
+				KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END;
+
+	if (trap_override != KFD_DBG_TRAP_OVERRIDE_OR &&
+			trap_override != KFD_DBG_TRAP_OVERRIDE_REPLACE)
+		return -EPERM;
+
+	return 0;
+}
+
+static uint32_t trap_mask_map_sw_to_hw(uint32_t mask)
+{
+	uint32_t trap_on_start = (mask & KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START) ? 1 : 0;
+	uint32_t trap_on_end = (mask & KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END) ? 1 : 0;
+	uint32_t excp_en = mask & (KFD_DBG_TRAP_MASK_FP_INVALID |
+				KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+				KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+				KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+				KFD_DBG_TRAP_MASK_FP_INEXACT |
+				KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+				KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION);
+	uint32_t ret;
+
+	ret = REG_SET_FIELD(0, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, excp_en);
+	ret = REG_SET_FIELD(ret, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_START, trap_on_start);
+	ret = REG_SET_FIELD(ret, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_END, trap_on_end);
+
+	return ret;
+}
+
+static uint32_t trap_mask_map_hw_to_sw(uint32_t mask)
+{
+	uint32_t ret = REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, EXCP_EN);
+
+	if (REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_START))
+		ret |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START;
+
+	if (REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_END))
+		ret |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END;
+
+	return ret;
+}
+
+/* returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_gfx_v9_4_3_set_wave_launch_trap_override(
+				struct amdgpu_device *adev,
+				uint32_t vmid,
+				uint32_t trap_override,
+				uint32_t trap_mask_bits,
+				uint32_t trap_mask_request,
+				uint32_t *trap_mask_prev,
+				uint32_t kfd_dbg_trap_cntl_prev)
+
+{
+	uint32_t data = 0;
+
+	*trap_mask_prev = trap_mask_map_hw_to_sw(kfd_dbg_trap_cntl_prev);
+
+	data = (trap_mask_bits & trap_mask_request) |
+	       (*trap_mask_prev & ~trap_mask_request);
+	data = trap_mask_map_sw_to_hw(data);
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, trap_override);
+
+	return data;
+}
+
+#define TCP_WATCH_STRIDE (regTCP_WATCH1_ADDR_H - regTCP_WATCH0_ADDR_H)
+static uint32_t kgd_gfx_v9_4_3_set_address_watch(
+				struct amdgpu_device *adev,
+				uint64_t watch_address,
+				uint32_t watch_address_mask,
+				uint32_t watch_id,
+				uint32_t watch_mode,
+				uint32_t debug_vmid,
+				uint32_t inst)
+{
+	uint32_t watch_address_high;
+	uint32_t watch_address_low;
+	uint32_t watch_address_cntl;
+
+	watch_address_cntl = 0;
+	watch_address_low = lower_32_bits(watch_address);
+	watch_address_high = upper_32_bits(watch_address) & 0xffff;
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MODE,
+			watch_mode);
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MASK,
+			watch_address_mask >> 7);
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VALID,
+			1);
+
+	WREG32_XCC((SOC15_REG_OFFSET(GC, GET_INST(GC, inst),
+			regTCP_WATCH0_ADDR_H) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_high, inst);
+
+	WREG32_XCC((SOC15_REG_OFFSET(GC, GET_INST(GC, inst),
+			regTCP_WATCH0_ADDR_L) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_low, inst);
+
+	return watch_address_cntl;
+}
+
+static uint32_t kgd_gfx_v9_4_3_clear_address_watch(struct amdgpu_device *adev,
+				uint32_t watch_id)
+{
+	return 0;
+}
+
+static uint32_t kgd_gfx_v9_4_3_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+						     int engine, int queue)
+{
+	uint32_t reg_offset = get_sdma_rlc_reg_offset(adev, engine, queue);
+	uint32_t status = RREG32(regSDMA_RLC0_CONTEXT_STATUS + reg_offset);
+	uint32_t doorbell_off = RREG32(regSDMA_RLC0_DOORBELL_OFFSET + reg_offset);
+	bool is_active = !!REG_GET_FIELD(status, SDMA_RLC0_CONTEXT_STATUS, SELECTED);
+
+	return is_active ? doorbell_off >> 2 : 0;
+}
+
+const struct kfd2kgd_calls gc_9_4_3_kfd2kgd = {
+	.program_sh_mem_settings = kgd_gfx_v9_program_sh_mem_settings,
+	.set_pasid_vmid_mapping = kgd_gfx_v9_4_3_set_pasid_vmid_mapping,
+	.init_interrupts = kgd_gfx_v9_init_interrupts,
+	.hqd_load = kgd_gfx_v9_4_3_hqd_load,
+	.hiq_mqd_load = kgd_gfx_v9_hiq_mqd_load,
+	.hqd_sdma_load = kgd_gfx_v9_4_3_hqd_sdma_load,
+	.hqd_dump = kgd_gfx_v9_hqd_dump,
+	.hqd_sdma_dump = kgd_gfx_v9_4_3_hqd_sdma_dump,
+	.hqd_is_occupied = kgd_gfx_v9_hqd_is_occupied,
+	.hqd_sdma_is_occupied = kgd_gfx_v9_4_3_hqd_sdma_is_occupied,
+	.hqd_destroy = kgd_gfx_v9_hqd_destroy,
+	.hqd_sdma_destroy = kgd_gfx_v9_4_3_hqd_sdma_destroy,
+	.wave_control_execute = kgd_gfx_v9_wave_control_execute,
+	.get_atc_vmid_pasid_mapping_info =
+				kgd_gfx_v9_get_atc_vmid_pasid_mapping_info,
+	.set_vm_context_page_table_base =
+				kgd_gfx_v9_set_vm_context_page_table_base,
+	.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
+	.program_trap_handler_settings =
+				kgd_gfx_v9_program_trap_handler_settings,
+	.build_dequeue_wait_counts_packet_info =
+				kgd_gfx_v9_build_dequeue_wait_counts_packet_info,
+	.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
+	.enable_debug_trap = kgd_aldebaran_enable_debug_trap,
+	.disable_debug_trap = kgd_gfx_v9_4_3_disable_debug_trap,
+	.validate_trap_override_request =
+			kgd_gfx_v9_4_3_validate_trap_override_request,
+	.set_wave_launch_trap_override =
+			kgd_gfx_v9_4_3_set_wave_launch_trap_override,
+	.set_wave_launch_mode = kgd_aldebaran_set_wave_launch_mode,
+	.set_address_watch = kgd_gfx_v9_4_3_set_address_watch,
+	.clear_address_watch = kgd_gfx_v9_4_3_clear_address_watch,
+	.hqd_get_pq_addr = kgd_gfx_v9_hqd_get_pq_addr,
+	.hqd_reset = kgd_gfx_v9_hqd_reset,
+	.hqd_sdma_get_doorbell = kgd_gfx_v9_4_3_hqd_sdma_get_doorbell
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
index 9378fc79e9ea..0239114fb6c4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.c
@@ -21,6 +21,7 @@
  */
 #include "amdgpu.h"
 #include "amdgpu_amdkfd.h"
+#include "amdgpu_amdkfd_gfx_v10.h"
 #include "gc/gc_10_1_0_offset.h"
 #include "gc/gc_10_1_0_sh_mask.h"
 #include "athub/athub_2_0_0_offset.h"
@@ -31,6 +32,7 @@
 #include "v10_structs.h"
 #include "nv.h"
 #include "nvd.h"
+#include <uapi/linux/kfd_ioctl.h>
 
 enum hqd_dequeue_request_type {
 	NO_ACTION = 0,
@@ -79,7 +81,7 @@ static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmi
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
-					uint32_t sh_mem_bases)
+					uint32_t sh_mem_bases, uint32_t inst)
 {
 	lock_srbm(adev, 0, 0, 0, vmid);
 
@@ -91,7 +93,7 @@ static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmi
 }
 
 static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
-					unsigned int vmid)
+					unsigned int vmid, uint32_t inst)
 {
 	/*
 	 * We have to assume that there is no outstanding mapping.
@@ -135,7 +137,8 @@ static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
  * but still works
  */
 
-static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id)
+static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t inst)
 {
 	uint32_t mec;
 	uint32_t pipe;
@@ -205,7 +208,7 @@ static inline struct v10_sdma_mqd *get_sdma_mqd(void *mqd)
 static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t __user *wptr, uint32_t wptr_shift,
-			uint32_t wptr_mask, struct mm_struct *mm)
+			uint32_t wptr_mask, struct mm_struct *mm, uint32_t inst)
 {
 	struct v10_compute_mqd *m;
 	uint32_t *mqd_hqd;
@@ -286,9 +289,9 @@ static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd,
 
 static int kgd_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 			    uint32_t pipe_id, uint32_t queue_id,
-			    uint32_t doorbell_off)
+			    uint32_t doorbell_off, uint32_t inst)
 {
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[0].ring;
 	struct v10_compute_mqd *m;
 	uint32_t mec, pipe;
 	int r;
@@ -303,7 +306,7 @@ static int kgd_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 	pr_debug("kfd: set HIQ, mec:%d, pipe:%d, queue:%d.\n",
 		 mec, pipe, queue_id);
 
-	spin_lock(&adev->gfx.kiq.ring_lock);
+	spin_lock(&adev->gfx.kiq[0].ring_lock);
 	r = amdgpu_ring_alloc(kiq_ring, 7);
 	if (r) {
 		pr_err("Failed to alloc KIQ (%d).\n", r);
@@ -330,7 +333,7 @@ static int kgd_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 	amdgpu_ring_commit(kiq_ring);
 
 out_unlock:
-	spin_unlock(&adev->gfx.kiq.ring_lock);
+	spin_unlock(&adev->gfx.kiq[0].ring_lock);
 	release_queue(adev);
 
 	return r;
@@ -338,7 +341,7 @@ out_unlock:
 
 static int kgd_hqd_dump(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
-			uint32_t (**dump)[2], uint32_t *n_regs)
+			uint32_t (**dump)[2], uint32_t *n_regs, uint32_t inst)
 {
 	uint32_t i = 0, reg;
 #define HQD_N_REGS 56
@@ -349,7 +352,7 @@ static int kgd_hqd_dump(struct amdgpu_device *adev,
 		(*dump)[i++][1] = RREG32_SOC15_IP(GC, addr);		\
 	} while (0)
 
-	*dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -446,7 +449,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 #undef HQD_N_REGS
 #define HQD_N_REGS (19+6+7+10)
 
-	*dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -469,7 +472,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 
 static bool kgd_hqd_is_occupied(struct amdgpu_device *adev,
 				uint64_t queue_address, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	uint32_t act;
 	bool retval = false;
@@ -510,7 +513,7 @@ static bool kgd_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd)
 static int kgd_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	enum hqd_dequeue_request_type type;
 	unsigned long end_jiffies;
@@ -673,7 +676,7 @@ static bool get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev,
 
 static int kgd_wave_control_execute(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
-					uint32_t sq_cmd)
+					uint32_t sq_cmd, uint32_t inst)
 {
 	uint32_t data = 0;
 
@@ -708,8 +711,342 @@ static void set_vm_context_page_table_base(struct amdgpu_device *adev,
 	adev->gfxhub.funcs->setup_vm_pt_regs(adev, vmid, page_table_base);
 }
 
+/*
+ * GFX10 helper for wave launch stall requirements on debug trap setting.
+ *
+ * vmid:
+ *   Target VMID to stall/unstall.
+ *
+ * stall:
+ *   0-unstall wave launch (enable), 1-stall wave launch (disable).
+ *   After wavefront launch has been stalled, allocated waves must drain from
+ *   SPI in order for debug trap settings to take effect on those waves.
+ *   This is roughly a ~3500 clock cycle wait on SPI where a read on
+ *   SPI_GDBG_WAVE_CNTL translates to ~32 clock cycles.
+ *   KGD_GFX_V10_WAVE_LAUNCH_SPI_DRAIN_LATENCY indicates the number of reads required.
+ *
+ *   NOTE: We can afford to clear the entire STALL_VMID field on unstall
+ *   because current GFX10 chips cannot support multi-process debugging due to
+ *   trap configuration and masking being limited to global scope.  Always
+ *   assume single process conditions.
+ *
+ */
+
+#define KGD_GFX_V10_WAVE_LAUNCH_SPI_DRAIN_LATENCY	110
+static void kgd_gfx_v10_set_wave_launch_stall(struct amdgpu_device *adev, uint32_t vmid, bool stall)
+{
+	uint32_t data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+	int i;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_VMID,
+							stall ? 1 << vmid : 0);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);
+
+	if (!stall)
+		return;
+
+	for (i = 0; i < KGD_GFX_V10_WAVE_LAUNCH_SPI_DRAIN_LATENCY; i++)
+		RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+}
+
+uint32_t kgd_gfx_v10_enable_debug_trap(struct amdgpu_device *adev,
+				bool restore_dbg_registers,
+				uint32_t vmid)
+{
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v10_set_wave_launch_stall(adev, vmid, true);
+
+	/* assume gfx off is disabled for the debug session if rlc restore not supported. */
+	if (restore_dbg_registers) {
+		uint32_t data = 0;
+
+		data = REG_SET_FIELD(data, SPI_GDBG_TRAP_CONFIG,
+				VMID_SEL, 1 << vmid);
+		data = REG_SET_FIELD(data, SPI_GDBG_TRAP_CONFIG,
+				TRAP_EN, 1);
+		WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_CONFIG), data);
+		WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA0), 0);
+		WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA1), 0);
+
+		kgd_gfx_v10_set_wave_launch_stall(adev, vmid, false);
+
+		mutex_unlock(&adev->grbm_idx_mutex);
+
+		return 0;
+	}
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	kgd_gfx_v10_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+uint32_t kgd_gfx_v10_disable_debug_trap(struct amdgpu_device *adev,
+					bool keep_trap_enabled,
+					uint32_t vmid)
+{
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v10_set_wave_launch_stall(adev, vmid, true);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	kgd_gfx_v10_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+int kgd_gfx_v10_validate_trap_override_request(struct amdgpu_device *adev,
+					      uint32_t trap_override,
+					      uint32_t *trap_mask_supported)
+{
+	*trap_mask_supported &= KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH;
+
+	/* The SPI_GDBG_TRAP_MASK register is global and affects all
+	 * processes. Only allow OR-ing the address-watch bit, since
+	 * this only affects processes under the debugger. Other bits
+	 * should stay 0 to avoid the debugger interfering with other
+	 * processes.
+	 */
+	if (trap_override != KFD_DBG_TRAP_OVERRIDE_OR)
+		return -EINVAL;
+
+	return 0;
+}
+
+uint32_t kgd_gfx_v10_set_wave_launch_trap_override(struct amdgpu_device *adev,
+					      uint32_t vmid,
+					      uint32_t trap_override,
+					      uint32_t trap_mask_bits,
+					      uint32_t trap_mask_request,
+					      uint32_t *trap_mask_prev,
+					      uint32_t kfd_dbg_trap_cntl_prev)
+{
+	uint32_t data, wave_cntl_prev;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	wave_cntl_prev = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+
+	kgd_gfx_v10_set_wave_launch_stall(adev, vmid, true);
+
+	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK));
+	*trap_mask_prev = REG_GET_FIELD(data, SPI_GDBG_TRAP_MASK, EXCP_EN);
+
+	trap_mask_bits = (trap_mask_bits & trap_mask_request) |
+		(*trap_mask_prev & ~trap_mask_request);
+
+	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK, EXCP_EN, trap_mask_bits);
+	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK, REPLACE, trap_override);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), data);
+
+	/* We need to preserve wave launch mode stall settings. */
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), wave_cntl_prev);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+uint32_t kgd_gfx_v10_set_wave_launch_mode(struct amdgpu_device *adev,
+					uint8_t wave_launch_mode,
+					uint32_t vmid)
+{
+	uint32_t data = 0;
+	bool is_mode_set = !!wave_launch_mode;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v10_set_wave_launch_stall(adev, vmid, true);
+
+	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
+			VMID_MASK, is_mode_set ? 1 << vmid : 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
+			MODE, is_mode_set ? wave_launch_mode : 0);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL2), data);
+
+	kgd_gfx_v10_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+#define TCP_WATCH_STRIDE (mmTCP_WATCH1_ADDR_H - mmTCP_WATCH0_ADDR_H)
+#define SQ_WATCH_STRIDE (mmSQ_WATCH1_ADDR_H - mmSQ_WATCH0_ADDR_H)
+uint32_t kgd_gfx_v10_set_address_watch(struct amdgpu_device *adev,
+					uint64_t watch_address,
+					uint32_t watch_address_mask,
+					uint32_t watch_id,
+					uint32_t watch_mode,
+					uint32_t debug_vmid,
+					uint32_t inst)
+{
+	/* SQ_WATCH?_ADDR_* and TCP_WATCH?_ADDR_* are programmed with the
+	 * same values.
+	 */
+	uint32_t watch_address_high;
+	uint32_t watch_address_low;
+	uint32_t tcp_watch_address_cntl;
+	uint32_t sq_watch_address_cntl;
+
+	watch_address_low = lower_32_bits(watch_address);
+	watch_address_high = upper_32_bits(watch_address) & 0xffff;
+
+	tcp_watch_address_cntl = 0;
+	tcp_watch_address_cntl = REG_SET_FIELD(tcp_watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VMID,
+			debug_vmid);
+	tcp_watch_address_cntl = REG_SET_FIELD(tcp_watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MODE,
+			watch_mode);
+	tcp_watch_address_cntl = REG_SET_FIELD(tcp_watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MASK,
+			watch_address_mask >> 7);
+
+	sq_watch_address_cntl = 0;
+	sq_watch_address_cntl = REG_SET_FIELD(sq_watch_address_cntl,
+			SQ_WATCH0_CNTL,
+			VMID,
+			debug_vmid);
+	sq_watch_address_cntl = REG_SET_FIELD(sq_watch_address_cntl,
+			SQ_WATCH0_CNTL,
+			MODE,
+			watch_mode);
+	sq_watch_address_cntl = REG_SET_FIELD(sq_watch_address_cntl,
+			SQ_WATCH0_CNTL,
+			MASK,
+			watch_address_mask >> 6);
+
+	/* Turning off this watch point until we set all the registers */
+	tcp_watch_address_cntl = REG_SET_FIELD(tcp_watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VALID,
+			0);
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			tcp_watch_address_cntl);
+
+	sq_watch_address_cntl = REG_SET_FIELD(sq_watch_address_cntl,
+			SQ_WATCH0_CNTL,
+			VALID,
+			0);
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmSQ_WATCH0_CNTL) +
+			(watch_id * SQ_WATCH_STRIDE)),
+			sq_watch_address_cntl);
+
+	/* Program {TCP,SQ}_WATCH?_ADDR* */
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_ADDR_H) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_high);
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_ADDR_L) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_low);
+
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmSQ_WATCH0_ADDR_H) +
+			(watch_id * SQ_WATCH_STRIDE)),
+			watch_address_high);
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmSQ_WATCH0_ADDR_L) +
+			(watch_id * SQ_WATCH_STRIDE)),
+			watch_address_low);
+
+	/* Enable the watch point */
+	tcp_watch_address_cntl = REG_SET_FIELD(tcp_watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VALID,
+			1);
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			tcp_watch_address_cntl);
+
+	sq_watch_address_cntl = REG_SET_FIELD(sq_watch_address_cntl,
+			SQ_WATCH0_CNTL,
+			VALID,
+			1);
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmSQ_WATCH0_CNTL) +
+			(watch_id * SQ_WATCH_STRIDE)),
+			sq_watch_address_cntl);
+
+	return 0;
+}
+
+uint32_t kgd_gfx_v10_clear_address_watch(struct amdgpu_device *adev,
+					uint32_t watch_id)
+{
+	uint32_t watch_address_cntl;
+
+	watch_address_cntl = 0;
+
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_cntl);
+
+	WREG32((SOC15_REG_OFFSET(GC, 0, mmSQ_WATCH0_CNTL) +
+			(watch_id * SQ_WATCH_STRIDE)),
+			watch_address_cntl);
+
+	return 0;
+}
+#undef TCP_WATCH_STRIDE
+#undef SQ_WATCH_STRIDE
+
+
+/* kgd_gfx_v10_get_iq_wait_times: Returns the mmCP_IQ_WAIT_TIME1/2 values
+ * The values read are:
+ *     ib_offload_wait_time     -- Wait Count for Indirect Buffer Offloads.
+ *     atomic_offload_wait_time -- Wait Count for L2 and GDS Atomics Offloads.
+ *     wrm_offload_wait_time    -- Wait Count for WAIT_REG_MEM Offloads.
+ *     gws_wait_time            -- Wait Count for Global Wave Syncs.
+ *     que_sleep_wait_time      -- Wait Count for Dequeue Retry.
+ *     sch_wave_wait_time       -- Wait Count for Scheduling Wave Message.
+ *     sem_rearm_wait_time      -- Wait Count for Semaphore re-arm.
+ *     deq_retry_wait_time      -- Wait Count for Global Wave Syncs.
+ */
+void kgd_gfx_v10_get_iq_wait_times(struct amdgpu_device *adev,
+					uint32_t *wait_times,
+					uint32_t inst)
+
+{
+	*wait_times = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2));
+}
+
+void kgd_gfx_v10_build_dequeue_wait_counts_packet_info(struct amdgpu_device *adev,
+						uint32_t wait_times,
+						uint32_t sch_wave,
+						uint32_t que_sleep,
+						uint32_t *reg_offset,
+						uint32_t *reg_data)
+{
+	*reg_data = wait_times;
+
+	if (sch_wave)
+		*reg_data = REG_SET_FIELD(*reg_data,
+				CP_IQ_WAIT_TIME2,
+				SCH_WAVE,
+				sch_wave);
+	if (que_sleep)
+		*reg_data = REG_SET_FIELD(*reg_data,
+				CP_IQ_WAIT_TIME2,
+				QUE_SLEEP,
+				que_sleep);
+
+	*reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2);
+}
+
 static void program_trap_handler_settings(struct amdgpu_device *adev,
-		uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr)
+		uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr,
+		uint32_t inst)
 {
 	lock_srbm(adev, 0, 0, 0, vmid);
 
@@ -733,6 +1070,26 @@ static void program_trap_handler_settings(struct amdgpu_device *adev,
 	unlock_srbm(adev);
 }
 
+uint64_t kgd_gfx_v10_hqd_get_pq_addr(struct amdgpu_device *adev,
+				     uint32_t pipe_id, uint32_t queue_id,
+				     uint32_t inst)
+{
+	return 0;
+}
+
+uint64_t kgd_gfx_v10_hqd_reset(struct amdgpu_device *adev,
+			       uint32_t pipe_id, uint32_t queue_id,
+			       uint32_t inst, unsigned int utimeout)
+{
+	return 0;
+}
+
+uint32_t kgd_gfx_v10_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+					   int engine, int queue)
+{
+	return 0;
+}
+
 const struct kfd2kgd_calls gfx_v10_kfd2kgd = {
 	.program_sh_mem_settings = kgd_program_sh_mem_settings,
 	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
@@ -750,5 +1107,17 @@ const struct kfd2kgd_calls gfx_v10_kfd2kgd = {
 	.get_atc_vmid_pasid_mapping_info =
 			get_atc_vmid_pasid_mapping_info,
 	.set_vm_context_page_table_base = set_vm_context_page_table_base,
+	.enable_debug_trap = kgd_gfx_v10_enable_debug_trap,
+	.disable_debug_trap = kgd_gfx_v10_disable_debug_trap,
+	.validate_trap_override_request = kgd_gfx_v10_validate_trap_override_request,
+	.set_wave_launch_trap_override = kgd_gfx_v10_set_wave_launch_trap_override,
+	.set_wave_launch_mode = kgd_gfx_v10_set_wave_launch_mode,
+	.set_address_watch = kgd_gfx_v10_set_address_watch,
+	.clear_address_watch = kgd_gfx_v10_clear_address_watch,
+	.get_iq_wait_times = kgd_gfx_v10_get_iq_wait_times,
+	.build_dequeue_wait_counts_packet_info = kgd_gfx_v10_build_dequeue_wait_counts_packet_info,
 	.program_trap_handler_settings = program_trap_handler_settings,
+	.hqd_get_pq_addr = kgd_gfx_v10_hqd_get_pq_addr,
+	.hqd_reset = kgd_gfx_v10_hqd_reset,
+	.hqd_sdma_get_doorbell = kgd_gfx_v10_hqd_sdma_get_doorbell
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
new file mode 100644
index 000000000000..a4c607c88178
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10.h
@@ -0,0 +1,70 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+uint32_t kgd_gfx_v10_enable_debug_trap(struct amdgpu_device *adev,
+				      bool restore_dbg_registers,
+				      uint32_t vmid);
+uint32_t kgd_gfx_v10_disable_debug_trap(struct amdgpu_device *adev,
+					bool keep_trap_enabled,
+					uint32_t vmid);
+int kgd_gfx_v10_validate_trap_override_request(struct amdgpu_device *adev,
+					     uint32_t trap_override,
+					     uint32_t *trap_mask_supported);
+uint32_t kgd_gfx_v10_set_wave_launch_trap_override(struct amdgpu_device *adev,
+					     uint32_t vmid,
+					     uint32_t trap_override,
+					     uint32_t trap_mask_bits,
+					     uint32_t trap_mask_request,
+					     uint32_t *trap_mask_prev,
+					     uint32_t kfd_dbg_trap_cntl_prev);
+uint32_t kgd_gfx_v10_set_wave_launch_mode(struct amdgpu_device *adev,
+					 uint8_t wave_launch_mode,
+					 uint32_t vmid);
+uint32_t kgd_gfx_v10_set_address_watch(struct amdgpu_device *adev,
+					uint64_t watch_address,
+					uint32_t watch_address_mask,
+					uint32_t watch_id,
+					uint32_t watch_mode,
+					uint32_t debug_vmid,
+					uint32_t inst);
+uint32_t kgd_gfx_v10_clear_address_watch(struct amdgpu_device *adev,
+					uint32_t watch_id);
+void kgd_gfx_v10_get_iq_wait_times(struct amdgpu_device *adev,
+				uint32_t *wait_times,
+				uint32_t inst);
+void kgd_gfx_v10_build_dequeue_wait_counts_packet_info(struct amdgpu_device *adev,
+					       uint32_t wait_times,
+					       uint32_t sch_wave,
+					       uint32_t que_sleep,
+					       uint32_t *reg_offset,
+					       uint32_t *reg_data);
+uint64_t kgd_gfx_v10_hqd_get_pq_addr(struct amdgpu_device *adev,
+				    uint32_t pipe_id,
+				    uint32_t queue_id,
+				    uint32_t inst);
+uint64_t kgd_gfx_v10_hqd_reset(struct amdgpu_device *adev,
+			      uint32_t pipe_id,
+			      uint32_t queue_id,
+			      uint32_t inst,
+			      unsigned int utimeout);
+uint32_t kgd_gfx_v10_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+					   int engine, int queue);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
index ba21ec6b35e0..f2278a0937ff 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v10_3.c
@@ -22,6 +22,7 @@
 #include <linux/mmu_context.h>
 #include "amdgpu.h"
 #include "amdgpu_amdkfd.h"
+#include "amdgpu_amdkfd_gfx_v10.h"
 #include "gc/gc_10_3_0_offset.h"
 #include "gc/gc_10_3_0_sh_mask.h"
 #include "oss/osssys_5_0_0_offset.h"
@@ -80,7 +81,7 @@ static void program_sh_mem_settings_v10_3(struct amdgpu_device *adev, uint32_t v
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
-					uint32_t sh_mem_bases)
+					uint32_t sh_mem_bases, uint32_t inst)
 {
 	lock_srbm(adev, 0, 0, 0, vmid);
 
@@ -93,7 +94,7 @@ static void program_sh_mem_settings_v10_3(struct amdgpu_device *adev, uint32_t v
 
 /* ATC is defeatured on Sienna_Cichlid */
 static int set_pasid_vmid_mapping_v10_3(struct amdgpu_device *adev, unsigned int pasid,
-					unsigned int vmid)
+					unsigned int vmid, uint32_t inst)
 {
 	uint32_t value = pasid << IH_VMID_0_LUT__PASID__SHIFT;
 
@@ -105,7 +106,8 @@ static int set_pasid_vmid_mapping_v10_3(struct amdgpu_device *adev, unsigned int
 	return 0;
 }
 
-static int init_interrupts_v10_3(struct amdgpu_device *adev, uint32_t pipe_id)
+static int init_interrupts_v10_3(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t inst)
 {
 	uint32_t mec;
 	uint32_t pipe;
@@ -177,7 +179,7 @@ static inline struct v10_sdma_mqd *get_sdma_mqd(void *mqd)
 static int hqd_load_v10_3(struct amdgpu_device *adev, void *mqd,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t __user *wptr, uint32_t wptr_shift,
-			uint32_t wptr_mask, struct mm_struct *mm)
+			uint32_t wptr_mask, struct mm_struct *mm, uint32_t inst)
 {
 	struct v10_compute_mqd *m;
 	uint32_t *mqd_hqd;
@@ -273,9 +275,9 @@ static int hqd_load_v10_3(struct amdgpu_device *adev, void *mqd,
 
 static int hiq_mqd_load_v10_3(struct amdgpu_device *adev, void *mqd,
 			    uint32_t pipe_id, uint32_t queue_id,
-			    uint32_t doorbell_off)
+			    uint32_t doorbell_off, uint32_t inst)
 {
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[0].ring;
 	struct v10_compute_mqd *m;
 	uint32_t mec, pipe;
 	int r;
@@ -290,7 +292,7 @@ static int hiq_mqd_load_v10_3(struct amdgpu_device *adev, void *mqd,
 	pr_debug("kfd: set HIQ, mec:%d, pipe:%d, queue:%d.\n",
 		 mec, pipe, queue_id);
 
-	spin_lock(&adev->gfx.kiq.ring_lock);
+	spin_lock(&adev->gfx.kiq[0].ring_lock);
 	r = amdgpu_ring_alloc(kiq_ring, 7);
 	if (r) {
 		pr_err("Failed to alloc KIQ (%d).\n", r);
@@ -317,7 +319,7 @@ static int hiq_mqd_load_v10_3(struct amdgpu_device *adev, void *mqd,
 	amdgpu_ring_commit(kiq_ring);
 
 out_unlock:
-	spin_unlock(&adev->gfx.kiq.ring_lock);
+	spin_unlock(&adev->gfx.kiq[0].ring_lock);
 	release_queue(adev);
 
 	return r;
@@ -325,7 +327,7 @@ out_unlock:
 
 static int hqd_dump_v10_3(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
-			uint32_t (**dump)[2], uint32_t *n_regs)
+			uint32_t (**dump)[2], uint32_t *n_regs, uint32_t inst)
 {
 	uint32_t i = 0, reg;
 #define HQD_N_REGS 56
@@ -336,7 +338,7 @@ static int hqd_dump_v10_3(struct amdgpu_device *adev,
 		(*dump)[i++][1] = RREG32_SOC15_IP(GC, addr);		\
 	} while (0)
 
-	*dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -433,7 +435,7 @@ static int hqd_sdma_dump_v10_3(struct amdgpu_device *adev,
 #undef HQD_N_REGS
 #define HQD_N_REGS (19+6+7+12)
 
-	*dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -456,7 +458,7 @@ static int hqd_sdma_dump_v10_3(struct amdgpu_device *adev,
 
 static bool hqd_is_occupied_v10_3(struct amdgpu_device *adev,
 				uint64_t queue_address, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	uint32_t act;
 	bool retval = false;
@@ -498,7 +500,7 @@ static bool hqd_sdma_is_occupied_v10_3(struct amdgpu_device *adev,
 static int hqd_destroy_v10_3(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	enum hqd_dequeue_request_type type;
 	unsigned long end_jiffies;
@@ -586,7 +588,7 @@ static int hqd_sdma_destroy_v10_3(struct amdgpu_device *adev, void *mqd,
 
 static int wave_control_execute_v10_3(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
-					uint32_t sq_cmd)
+					uint32_t sq_cmd, uint32_t inst)
 {
 	uint32_t data = 0;
 
@@ -628,7 +630,8 @@ static void set_vm_context_page_table_base_v10_3(struct amdgpu_device *adev,
 }
 
 static void program_trap_handler_settings_v10_3(struct amdgpu_device *adev,
-			uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr)
+			uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr,
+			uint32_t inst)
 {
 	lock_srbm(adev, 0, 0, 0, vmid);
 
@@ -652,142 +655,6 @@ static void program_trap_handler_settings_v10_3(struct amdgpu_device *adev,
 	unlock_srbm(adev);
 }
 
-#if 0
-uint32_t enable_debug_trap_v10_3(struct amdgpu_device *adev,
-				uint32_t trap_debug_wave_launch_mode,
-				uint32_t vmid)
-{
-	uint32_t data = 0;
-	uint32_t orig_wave_cntl_value;
-	uint32_t orig_stall_vmid;
-
-	mutex_lock(&adev->grbm_idx_mutex);
-
-	orig_wave_cntl_value = RREG32(SOC15_REG_OFFSET(GC,
-				0,
-				mmSPI_GDBG_WAVE_CNTL));
-	orig_stall_vmid = REG_GET_FIELD(orig_wave_cntl_value,
-			SPI_GDBG_WAVE_CNTL,
-			STALL_VMID);
-
-	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_RA, 1);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);
-
-	data = 0;
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), data);
-
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), orig_stall_vmid);
-
-	mutex_unlock(&adev->grbm_idx_mutex);
-
-	return 0;
-}
-
-uint32_t disable_debug_trap_v10_3(struct amdgpu_device *adev)
-{
-	mutex_lock(&adev->grbm_idx_mutex);
-
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
-
-	mutex_unlock(&adev->grbm_idx_mutex);
-
-	return 0;
-}
-
-uint32_t set_wave_launch_trap_override_v10_3(struct amdgpu_device *adev,
-						uint32_t trap_override,
-						uint32_t trap_mask)
-{
-	uint32_t data = 0;
-
-	mutex_lock(&adev->grbm_idx_mutex);
-
-	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
-	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_RA, 1);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);
-
-	data = 0;
-	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK,
-			EXCP_EN, trap_mask);
-	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK,
-			REPLACE, trap_override);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), data);
-
-	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
-	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_RA, 0);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);
-
-	mutex_unlock(&adev->grbm_idx_mutex);
-
-	return 0;
-}
-
-uint32_t set_wave_launch_mode_v10_3(struct amdgpu_device *adev,
-					uint8_t wave_launch_mode,
-					uint32_t vmid)
-{
-	uint32_t data = 0;
-	bool is_stall_mode;
-	bool is_mode_set;
-
-	is_stall_mode = (wave_launch_mode == 4);
-	is_mode_set = (wave_launch_mode != 0 && wave_launch_mode != 4);
-
-	mutex_lock(&adev->grbm_idx_mutex);
-
-	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
-			VMID_MASK, is_mode_set ? 1 << vmid : 0);
-	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
-			MODE, is_mode_set ? wave_launch_mode : 0);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL2), data);
-
-	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
-	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL,
-			STALL_VMID, is_stall_mode ? 1 << vmid : 0);
-	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL,
-			STALL_RA, is_stall_mode ? 1 : 0);
-	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);
-
-	mutex_unlock(&adev->grbm_idx_mutex);
-
-	return 0;
-}
-
-/* kgd_get_iq_wait_times: Returns the mmCP_IQ_WAIT_TIME1/2 values
- * The values read are:
- *	ib_offload_wait_time     -- Wait Count for Indirect Buffer Offloads.
- *	atomic_offload_wait_time -- Wait Count for L2 and GDS Atomics Offloads.
- *	wrm_offload_wait_time    -- Wait Count for WAIT_REG_MEM Offloads.
- *	gws_wait_time            -- Wait Count for Global Wave Syncs.
- *	que_sleep_wait_time      -- Wait Count for Dequeue Retry.
- *	sch_wave_wait_time       -- Wait Count for Scheduling Wave Message.
- *	sem_rearm_wait_time      -- Wait Count for Semaphore re-arm.
- *	deq_retry_wait_time      -- Wait Count for Global Wave Syncs.
- */
-void get_iq_wait_times_v10_3(struct amdgpu_device *adev,
-					uint32_t *wait_times)
-
-{
-	*wait_times = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2));
-}
-
-void build_grace_period_packet_info_v10_3(struct amdgpu_device *adev,
-						uint32_t wait_times,
-						uint32_t grace_period,
-						uint32_t *reg_offset,
-						uint32_t *reg_data)
-{
-	*reg_data = wait_times;
-
-	*reg_data = REG_SET_FIELD(*reg_data,
-			CP_IQ_WAIT_TIME2,
-			SCH_WAVE,
-			grace_period);
-
-	*reg_offset = mmCP_IQ_WAIT_TIME2;
-}
-#endif
-
 const struct kfd2kgd_calls gfx_v10_3_kfd2kgd = {
 	.program_sh_mem_settings = program_sh_mem_settings_v10_3,
 	.set_pasid_vmid_mapping = set_pasid_vmid_mapping_v10_3,
@@ -805,12 +672,16 @@ const struct kfd2kgd_calls gfx_v10_3_kfd2kgd = {
 	.get_atc_vmid_pasid_mapping_info = get_atc_vmid_pasid_mapping_info_v10_3,
 	.set_vm_context_page_table_base = set_vm_context_page_table_base_v10_3,
 	.program_trap_handler_settings = program_trap_handler_settings_v10_3,
-#if 0
-	.enable_debug_trap = enable_debug_trap_v10_3,
-	.disable_debug_trap = disable_debug_trap_v10_3,
-	.set_wave_launch_trap_override = set_wave_launch_trap_override_v10_3,
-	.set_wave_launch_mode = set_wave_launch_mode_v10_3,
-	.get_iq_wait_times = get_iq_wait_times_v10_3,
-	.build_grace_period_packet_info = build_grace_period_packet_info_v10_3,
-#endif
+	.get_iq_wait_times = kgd_gfx_v10_get_iq_wait_times,
+	.build_dequeue_wait_counts_packet_info = kgd_gfx_v10_build_dequeue_wait_counts_packet_info,
+	.enable_debug_trap = kgd_gfx_v10_enable_debug_trap,
+	.disable_debug_trap = kgd_gfx_v10_disable_debug_trap,
+	.validate_trap_override_request = kgd_gfx_v10_validate_trap_override_request,
+	.set_wave_launch_trap_override = kgd_gfx_v10_set_wave_launch_trap_override,
+	.set_wave_launch_mode = kgd_gfx_v10_set_wave_launch_mode,
+	.set_address_watch = kgd_gfx_v10_set_address_watch,
+	.clear_address_watch = kgd_gfx_v10_clear_address_watch,
+	.hqd_get_pq_addr = kgd_gfx_v10_hqd_get_pq_addr,
+	.hqd_reset = kgd_gfx_v10_hqd_reset,
+	.hqd_sdma_get_doorbell = kgd_gfx_v10_hqd_sdma_get_doorbell
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
index 7e80caa05060..aaccf0b9947d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v11.c
@@ -30,6 +30,7 @@
 #include "soc15d.h"
 #include "v11_structs.h"
 #include "soc21.h"
+#include <uapi/linux/kfd_ioctl.h>
 
 enum hqd_dequeue_request_type {
 	NO_ACTION = 0,
@@ -78,7 +79,7 @@ static void program_sh_mem_settings_v11(struct amdgpu_device *adev, uint32_t vmi
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
-					uint32_t sh_mem_bases)
+					uint32_t sh_mem_bases, uint32_t inst)
 {
 	lock_srbm(adev, 0, 0, 0, vmid);
 
@@ -89,7 +90,7 @@ static void program_sh_mem_settings_v11(struct amdgpu_device *adev, uint32_t vmi
 }
 
 static int set_pasid_vmid_mapping_v11(struct amdgpu_device *adev, unsigned int pasid,
-					unsigned int vmid)
+					unsigned int vmid, uint32_t inst)
 {
 	uint32_t value = pasid << IH_VMID_0_LUT__PASID__SHIFT;
 
@@ -101,7 +102,8 @@ static int set_pasid_vmid_mapping_v11(struct amdgpu_device *adev, unsigned int p
 	return 0;
 }
 
-static int init_interrupts_v11(struct amdgpu_device *adev, uint32_t pipe_id)
+static int init_interrupts_v11(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t inst)
 {
 	uint32_t mec;
 	uint32_t pipe;
@@ -162,7 +164,7 @@ static inline struct v11_sdma_mqd *get_sdma_mqd(void *mqd)
 static int hqd_load_v11(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id,
 			uint32_t queue_id, uint32_t __user *wptr,
 			uint32_t wptr_shift, uint32_t wptr_mask,
-			struct mm_struct *mm)
+			struct mm_struct *mm, uint32_t inst)
 {
 	struct v11_compute_mqd *m;
 	uint32_t *mqd_hqd;
@@ -258,9 +260,9 @@ static int hqd_load_v11(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id,
 
 static int hiq_mqd_load_v11(struct amdgpu_device *adev, void *mqd,
 			      uint32_t pipe_id, uint32_t queue_id,
-			      uint32_t doorbell_off)
+			      uint32_t doorbell_off, uint32_t inst)
 {
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[0].ring;
 	struct v11_compute_mqd *m;
 	uint32_t mec, pipe;
 	int r;
@@ -275,7 +277,7 @@ static int hiq_mqd_load_v11(struct amdgpu_device *adev, void *mqd,
 	pr_debug("kfd: set HIQ, mec:%d, pipe:%d, queue:%d.\n",
 		 mec, pipe, queue_id);
 
-	spin_lock(&adev->gfx.kiq.ring_lock);
+	spin_lock(&adev->gfx.kiq[0].ring_lock);
 	r = amdgpu_ring_alloc(kiq_ring, 7);
 	if (r) {
 		pr_err("Failed to alloc KIQ (%d).\n", r);
@@ -302,7 +304,7 @@ static int hiq_mqd_load_v11(struct amdgpu_device *adev, void *mqd,
 	amdgpu_ring_commit(kiq_ring);
 
 out_unlock:
-	spin_unlock(&adev->gfx.kiq.ring_lock);
+	spin_unlock(&adev->gfx.kiq[0].ring_lock);
 	release_queue(adev);
 
 	return r;
@@ -310,7 +312,7 @@ out_unlock:
 
 static int hqd_dump_v11(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
-			uint32_t (**dump)[2], uint32_t *n_regs)
+			uint32_t (**dump)[2], uint32_t *n_regs, uint32_t inst)
 {
 	uint32_t i = 0, reg;
 #define HQD_N_REGS 56
@@ -321,7 +323,7 @@ static int hqd_dump_v11(struct amdgpu_device *adev,
 		(*dump)[i++][1] = RREG32(addr);		\
 	} while (0)
 
-	*dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -418,7 +420,7 @@ static int hqd_sdma_dump_v11(struct amdgpu_device *adev,
 #undef HQD_N_REGS
 #define HQD_N_REGS (7+11+1+12+12)
 
-	*dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -445,7 +447,7 @@ static int hqd_sdma_dump_v11(struct amdgpu_device *adev,
 }
 
 static bool hqd_is_occupied_v11(struct amdgpu_device *adev, uint64_t queue_address,
-				uint32_t pipe_id, uint32_t queue_id)
+				uint32_t pipe_id, uint32_t queue_id, uint32_t inst)
 {
 	uint32_t act;
 	bool retval = false;
@@ -486,7 +488,7 @@ static bool hqd_sdma_is_occupied_v11(struct amdgpu_device *adev, void *mqd)
 static int hqd_destroy_v11(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	enum hqd_dequeue_request_type type;
 	unsigned long end_jiffies;
@@ -571,7 +573,7 @@ static int hqd_sdma_destroy_v11(struct amdgpu_device *adev, void *mqd,
 
 static int wave_control_execute_v11(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
-					uint32_t sq_cmd)
+					uint32_t sq_cmd, uint32_t inst)
 {
 	uint32_t data = 0;
 
@@ -606,6 +608,204 @@ static void set_vm_context_page_table_base_v11(struct amdgpu_device *adev,
 	adev->gfxhub.funcs->setup_vm_pt_regs(adev, vmid, page_table_base);
 }
 
+/*
+ * Returns TRAP_EN, EXCP_EN and EXCP_REPLACE.
+ *
+ * restore_dbg_registers is ignored here but is a general interface requirement
+ * for devices that support GFXOFF and where the RLC save/restore list
+ * does not support hw registers for debugging i.e. the driver has to manually
+ * initialize the debug mode registers after it has disabled GFX off during the
+ * debug session.
+ */
+static uint32_t kgd_gfx_v11_enable_debug_trap(struct amdgpu_device *adev,
+					    bool restore_dbg_registers,
+					    uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, 0);
+
+	return data;
+}
+
+/* Returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_gfx_v11_disable_debug_trap(struct amdgpu_device *adev,
+						bool keep_trap_enabled,
+						uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, 0);
+
+	return data;
+}
+
+static int kgd_gfx_v11_validate_trap_override_request(struct amdgpu_device *adev,
+							uint32_t trap_override,
+							uint32_t *trap_mask_supported)
+{
+	*trap_mask_supported &= KFD_DBG_TRAP_MASK_FP_INVALID |
+				KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+				KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+				KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+				KFD_DBG_TRAP_MASK_FP_INEXACT |
+				KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+				KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION;
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(11, 0, 4))
+		*trap_mask_supported |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START |
+					KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END;
+
+	if (trap_override != KFD_DBG_TRAP_OVERRIDE_OR &&
+			trap_override != KFD_DBG_TRAP_OVERRIDE_REPLACE)
+		return -EPERM;
+
+	return 0;
+}
+
+static uint32_t trap_mask_map_sw_to_hw(uint32_t mask)
+{
+	uint32_t trap_on_start = (mask & KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START) ? 1 : 0;
+	uint32_t trap_on_end = (mask & KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END) ? 1 : 0;
+	uint32_t excp_en = mask & (KFD_DBG_TRAP_MASK_FP_INVALID |
+			KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+			KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+			KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+			KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+			KFD_DBG_TRAP_MASK_FP_INEXACT |
+			KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+			KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+			KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION);
+	uint32_t ret;
+
+	ret = REG_SET_FIELD(0, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, excp_en);
+	ret = REG_SET_FIELD(ret, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_START, trap_on_start);
+	ret = REG_SET_FIELD(ret, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_END, trap_on_end);
+
+	return ret;
+}
+
+static uint32_t trap_mask_map_hw_to_sw(uint32_t mask)
+{
+	uint32_t ret = REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, EXCP_EN);
+
+	if (REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_START))
+		ret |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START;
+
+	if (REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_END))
+		ret |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END;
+
+	return ret;
+}
+
+/* Returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_gfx_v11_set_wave_launch_trap_override(struct amdgpu_device *adev,
+					uint32_t vmid,
+					uint32_t trap_override,
+					uint32_t trap_mask_bits,
+					uint32_t trap_mask_request,
+					uint32_t *trap_mask_prev,
+					uint32_t kfd_dbg_trap_cntl_prev)
+{
+	uint32_t data = 0;
+
+	*trap_mask_prev = trap_mask_map_hw_to_sw(kfd_dbg_trap_cntl_prev);
+
+	data = (trap_mask_bits & trap_mask_request) | (*trap_mask_prev & ~trap_mask_request);
+	data = trap_mask_map_sw_to_hw(data);
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, trap_override);
+
+	return data;
+}
+
+static uint32_t kgd_gfx_v11_set_wave_launch_mode(struct amdgpu_device *adev,
+					uint8_t wave_launch_mode,
+					uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, LAUNCH_MODE, wave_launch_mode);
+
+	return data;
+}
+
+#define TCP_WATCH_STRIDE (regTCP_WATCH1_ADDR_H - regTCP_WATCH0_ADDR_H)
+static uint32_t kgd_gfx_v11_set_address_watch(struct amdgpu_device *adev,
+					uint64_t watch_address,
+					uint32_t watch_address_mask,
+					uint32_t watch_id,
+					uint32_t watch_mode,
+					uint32_t debug_vmid,
+					uint32_t inst)
+{
+	uint32_t watch_address_high;
+	uint32_t watch_address_low;
+	uint32_t watch_address_cntl;
+
+	watch_address_cntl = 0;
+	watch_address_low = lower_32_bits(watch_address);
+	watch_address_high = upper_32_bits(watch_address) & 0xffff;
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MODE,
+			watch_mode);
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MASK,
+			watch_address_mask >> 7);
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VALID,
+			1);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_H) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_high);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_L) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_low);
+
+	return watch_address_cntl;
+}
+
+static uint32_t kgd_gfx_v11_clear_address_watch(struct amdgpu_device *adev,
+						uint32_t watch_id)
+{
+	return 0;
+}
+
+static uint64_t kgd_gfx_v11_hqd_get_pq_addr(struct amdgpu_device *adev,
+					    uint32_t pipe_id, uint32_t queue_id,
+					    uint32_t inst)
+{
+	return 0;
+}
+
+static uint64_t kgd_gfx_v11_hqd_reset(struct amdgpu_device *adev,
+				      uint32_t pipe_id, uint32_t queue_id,
+				      uint32_t inst, unsigned int utimeout)
+{
+	return 0;
+}
+
+static uint32_t kgd_gfx_v11_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+						  int engine, int queue)
+{
+	return 0;
+}
+
 const struct kfd2kgd_calls gfx_v11_kfd2kgd = {
 	.program_sh_mem_settings = program_sh_mem_settings_v11,
 	.set_pasid_vmid_mapping = set_pasid_vmid_mapping_v11,
@@ -622,4 +822,14 @@ const struct kfd2kgd_calls gfx_v11_kfd2kgd = {
 	.wave_control_execute = wave_control_execute_v11,
 	.get_atc_vmid_pasid_mapping_info = NULL,
 	.set_vm_context_page_table_base = set_vm_context_page_table_base_v11,
+	.enable_debug_trap = kgd_gfx_v11_enable_debug_trap,
+	.disable_debug_trap = kgd_gfx_v11_disable_debug_trap,
+	.validate_trap_override_request = kgd_gfx_v11_validate_trap_override_request,
+	.set_wave_launch_trap_override = kgd_gfx_v11_set_wave_launch_trap_override,
+	.set_wave_launch_mode = kgd_gfx_v11_set_wave_launch_mode,
+	.set_address_watch = kgd_gfx_v11_set_address_watch,
+	.clear_address_watch = kgd_gfx_v11_clear_address_watch,
+	.hqd_get_pq_addr = kgd_gfx_v11_hqd_get_pq_addr,
+	.hqd_reset = kgd_gfx_v11_hqd_reset,
+	.hqd_sdma_get_doorbell = kgd_gfx_v11_hqd_sdma_get_doorbell
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v12.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v12.c
new file mode 100644
index 000000000000..e0ceab400b2d
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v12.c
@@ -0,0 +1,384 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include "amdgpu.h"
+#include "amdgpu_amdkfd.h"
+#include "gc/gc_12_0_0_offset.h"
+#include "gc/gc_12_0_0_sh_mask.h"
+#include "soc24.h"
+#include <uapi/linux/kfd_ioctl.h>
+
+static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
+			uint32_t queue, uint32_t vmid)
+{
+	mutex_lock(&adev->srbm_mutex);
+	soc24_grbm_select(adev, mec, pipe, queue, vmid);
+}
+
+static void unlock_srbm(struct amdgpu_device *adev)
+{
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+
+static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t queue_id)
+{
+	uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
+	uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
+
+	lock_srbm(adev, mec, pipe, queue_id, 0);
+}
+
+static void release_queue(struct amdgpu_device *adev)
+{
+	unlock_srbm(adev);
+}
+
+static int init_interrupts_v12(struct amdgpu_device *adev, uint32_t pipe_id, uint32_t inst)
+{
+	uint32_t mec;
+	uint32_t pipe;
+
+	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
+	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
+
+	lock_srbm(adev, mec, pipe, 0, 0);
+
+	WREG32_SOC15(GC, 0, regCPC_INT_CNTL,
+		CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK |
+		CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK);
+
+	unlock_srbm(adev);
+
+	return 0;
+}
+
+static uint32_t get_sdma_rlc_reg_offset(struct amdgpu_device *adev,
+				unsigned int engine_id,
+				unsigned int queue_id)
+{
+	uint32_t sdma_engine_reg_base = 0;
+	uint32_t sdma_rlc_reg_offset;
+
+	switch (engine_id) {
+	case 0:
+		sdma_engine_reg_base = SOC15_REG_OFFSET(SDMA0, 0,
+				regSDMA0_QUEUE0_RB_CNTL) - regSDMA0_QUEUE0_RB_CNTL;
+		break;
+	case 1:
+		sdma_engine_reg_base = SOC15_REG_OFFSET(SDMA1, 0,
+				regSDMA1_QUEUE0_RB_CNTL) - regSDMA0_QUEUE0_RB_CNTL;
+		break;
+	default:
+		BUG();
+	}
+
+	sdma_rlc_reg_offset = sdma_engine_reg_base
+		+ queue_id * (regSDMA0_QUEUE1_RB_CNTL - regSDMA0_QUEUE0_RB_CNTL);
+
+	pr_debug("RLC register offset for SDMA%d RLC%d: 0x%x\n", engine_id,
+			queue_id, sdma_rlc_reg_offset);
+
+	return sdma_rlc_reg_offset;
+}
+
+static int hqd_dump_v12(struct amdgpu_device *adev,
+			uint32_t pipe_id, uint32_t queue_id,
+			uint32_t (**dump)[2], uint32_t *n_regs, uint32_t inst)
+{
+	uint32_t i = 0, reg;
+#define HQD_N_REGS 56
+#define DUMP_REG(addr) do {				\
+		if (WARN_ON_ONCE(i >= HQD_N_REGS))	\
+			break;				\
+		(*dump)[i][0] = (addr) << 2;		\
+		(*dump)[i++][1] = RREG32(addr);		\
+	} while (0)
+
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
+	if (*dump == NULL)
+		return -ENOMEM;
+
+	acquire_queue(adev, pipe_id, queue_id);
+
+	for (reg = SOC15_REG_OFFSET(GC, 0, regCP_MQD_BASE_ADDR);
+	     reg <= SOC15_REG_OFFSET(GC, 0, regCP_HQD_PQ_WPTR_HI); reg++)
+		DUMP_REG(reg);
+
+	release_queue(adev);
+
+	WARN_ON_ONCE(i != HQD_N_REGS);
+	*n_regs = i;
+
+	return 0;
+}
+
+static int hqd_sdma_dump_v12(struct amdgpu_device *adev,
+			     uint32_t engine_id, uint32_t queue_id,
+			     uint32_t (**dump)[2], uint32_t *n_regs)
+{
+	uint32_t sdma_rlc_reg_offset = get_sdma_rlc_reg_offset(adev,
+			engine_id, queue_id);
+	uint32_t i = 0, reg;
+
+	const uint32_t first_reg = regSDMA0_QUEUE0_RB_CNTL;
+	const uint32_t last_reg = regSDMA0_QUEUE0_CONTEXT_STATUS;
+#undef HQD_N_REGS
+#define HQD_N_REGS (last_reg - first_reg + 1)
+
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
+	if (*dump == NULL)
+		return -ENOMEM;
+
+	for (reg = first_reg;
+	     reg <= last_reg; reg++)
+		DUMP_REG(sdma_rlc_reg_offset + reg);
+
+	WARN_ON_ONCE(i != HQD_N_REGS);
+	*n_regs = i;
+
+	return 0;
+}
+
+static int wave_control_execute_v12(struct amdgpu_device *adev,
+					uint32_t gfx_index_val,
+					uint32_t sq_cmd, uint32_t inst)
+{
+	uint32_t data = 0;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, regGRBM_GFX_INDEX), gfx_index_val);
+	WREG32(SOC15_REG_OFFSET(GC, 0, regSQ_CMD), sq_cmd);
+
+	data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
+		INSTANCE_BROADCAST_WRITES, 1);
+	data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
+		SA_BROADCAST_WRITES, 1);
+	data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
+		SE_BROADCAST_WRITES, 1);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, regGRBM_GFX_INDEX), data);
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+/* returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_gfx_v12_enable_debug_trap(struct amdgpu_device *adev,
+					    bool restore_dbg_registers,
+					    uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, 0);
+
+	return data;
+}
+
+/* returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_gfx_v12_disable_debug_trap(struct amdgpu_device *adev,
+						bool keep_trap_enabled,
+						uint32_t vmid)
+{
+	uint32_t data = 0;
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, 0);
+
+	return data;
+}
+
+static int kgd_gfx_v12_validate_trap_override_request(struct amdgpu_device *adev,
+							uint32_t trap_override,
+							uint32_t *trap_mask_supported)
+{
+	*trap_mask_supported &= KFD_DBG_TRAP_MASK_FP_INVALID |
+				KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+				KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+				KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+				KFD_DBG_TRAP_MASK_FP_INEXACT |
+				KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+				KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+				KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION |
+				KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START |
+				KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END;
+
+
+	if (trap_override != KFD_DBG_TRAP_OVERRIDE_OR &&
+			trap_override != KFD_DBG_TRAP_OVERRIDE_REPLACE)
+		return -EPERM;
+
+	return 0;
+}
+
+static uint32_t trap_mask_map_sw_to_hw(uint32_t mask)
+{
+	uint32_t trap_on_start = (mask & KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START) ? 1 : 0;
+	uint32_t trap_on_end = (mask & KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END) ? 1 : 0;
+	uint32_t excp_en = mask & (KFD_DBG_TRAP_MASK_FP_INVALID |
+			KFD_DBG_TRAP_MASK_FP_INPUT_DENORMAL |
+			KFD_DBG_TRAP_MASK_FP_DIVIDE_BY_ZERO |
+			KFD_DBG_TRAP_MASK_FP_OVERFLOW |
+			KFD_DBG_TRAP_MASK_FP_UNDERFLOW |
+			KFD_DBG_TRAP_MASK_FP_INEXACT |
+			KFD_DBG_TRAP_MASK_INT_DIVIDE_BY_ZERO |
+			KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH |
+			KFD_DBG_TRAP_MASK_DBG_MEMORY_VIOLATION);
+	uint32_t ret;
+
+	ret = REG_SET_FIELD(0, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, excp_en);
+	ret = REG_SET_FIELD(ret, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_START, trap_on_start);
+	ret = REG_SET_FIELD(ret, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_END, trap_on_end);
+
+	return ret;
+}
+
+static uint32_t trap_mask_map_hw_to_sw(uint32_t mask)
+{
+	uint32_t ret = REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, EXCP_EN);
+
+	if (REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_START))
+		ret |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_START;
+
+	if (REG_GET_FIELD(mask, SPI_GDBG_PER_VMID_CNTL, TRAP_ON_END))
+		ret |= KFD_DBG_TRAP_MASK_TRAP_ON_WAVE_END;
+
+	return ret;
+}
+
+/* returns TRAP_EN, EXCP_EN and EXCP_REPLACE. */
+static uint32_t kgd_gfx_v12_set_wave_launch_trap_override(struct amdgpu_device *adev,
+					uint32_t vmid,
+					uint32_t trap_override,
+					uint32_t trap_mask_bits,
+					uint32_t trap_mask_request,
+					uint32_t *trap_mask_prev,
+					uint32_t kfd_dbg_trap_cntl_prev)
+
+{
+	uint32_t data = 0;
+
+	*trap_mask_prev = trap_mask_map_hw_to_sw(kfd_dbg_trap_cntl_prev);
+
+	data = (trap_mask_bits & trap_mask_request) | (*trap_mask_prev & ~trap_mask_request);
+	data = trap_mask_map_sw_to_hw(data);
+
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+	data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE, trap_override);
+
+	return data;
+}
+
+/* returns STALL_VMID or LAUNCH_MODE. */
+static uint32_t kgd_gfx_v12_set_wave_launch_mode(struct amdgpu_device *adev,
+					uint8_t wave_launch_mode,
+					uint32_t vmid)
+{
+	uint32_t data = 0;
+	bool is_stall_mode = wave_launch_mode == 4;
+
+	if (is_stall_mode)
+		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, STALL_VMID,
+									1);
+	else
+		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, LAUNCH_MODE,
+							wave_launch_mode);
+
+	return data;
+}
+
+#define TCP_WATCH_STRIDE (regTCP_WATCH1_ADDR_H - regTCP_WATCH0_ADDR_H)
+static uint32_t kgd_gfx_v12_set_address_watch(struct amdgpu_device *adev,
+					uint64_t watch_address,
+					uint32_t watch_address_mask,
+					uint32_t watch_id,
+					uint32_t watch_mode,
+					uint32_t debug_vmid,
+					uint32_t inst)
+{
+	uint32_t watch_address_high;
+	uint32_t watch_address_low;
+	uint32_t watch_address_cntl;
+
+	watch_address_cntl = 0;
+	watch_address_low = lower_32_bits(watch_address);
+	watch_address_high = upper_32_bits(watch_address) & 0xffff;
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MODE,
+			watch_mode);
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MASK,
+			watch_address_mask >> 7);
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VALID,
+			1);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_H) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_high);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, regTCP_WATCH0_ADDR_L) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_low);
+
+	return watch_address_cntl;
+}
+
+static uint32_t kgd_gfx_v12_clear_address_watch(struct amdgpu_device *adev,
+					uint32_t watch_id)
+{
+	return 0;
+}
+
+static uint32_t kgd_gfx_v12_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+						 int engine, int queue)
+{
+	return 0;
+}
+
+const struct kfd2kgd_calls gfx_v12_kfd2kgd = {
+	.init_interrupts = init_interrupts_v12,
+	.hqd_dump = hqd_dump_v12,
+	.hqd_sdma_dump = hqd_sdma_dump_v12,
+	.wave_control_execute = wave_control_execute_v12,
+	.get_atc_vmid_pasid_mapping_info = NULL,
+	.enable_debug_trap = kgd_gfx_v12_enable_debug_trap,
+	.disable_debug_trap = kgd_gfx_v12_disable_debug_trap,
+	.validate_trap_override_request = kgd_gfx_v12_validate_trap_override_request,
+	.set_wave_launch_trap_override = kgd_gfx_v12_set_wave_launch_trap_override,
+	.set_wave_launch_mode = kgd_gfx_v12_set_wave_launch_mode,
+	.set_address_watch = kgd_gfx_v12_set_address_watch,
+	.clear_address_watch = kgd_gfx_v12_clear_address_watch,
+	.hqd_sdma_get_doorbell = kgd_gfx_v12_hqd_sdma_get_doorbell
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
index e83cb1c09610..df77558e03ef 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
@@ -78,7 +78,7 @@ static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmi
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
-					uint32_t sh_mem_bases)
+					uint32_t sh_mem_bases, uint32_t inst)
 {
 	lock_srbm(adev, 0, 0, 0, vmid);
 
@@ -91,7 +91,7 @@ static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmi
 }
 
 static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
-					unsigned int vmid)
+					unsigned int vmid, uint32_t inst)
 {
 	/*
 	 * We have to assume that there is no outstanding mapping.
@@ -114,7 +114,8 @@ static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
 	return 0;
 }
 
-static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id)
+static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t inst)
 {
 	uint32_t mec;
 	uint32_t pipe;
@@ -158,7 +159,7 @@ static inline struct cik_sdma_rlc_registers *get_sdma_mqd(void *mqd)
 static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t __user *wptr, uint32_t wptr_shift,
-			uint32_t wptr_mask, struct mm_struct *mm)
+			uint32_t wptr_mask, struct mm_struct *mm, uint32_t inst)
 {
 	struct cik_mqd *m;
 	uint32_t *mqd_hqd;
@@ -202,7 +203,7 @@ static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd,
 
 static int kgd_hqd_dump(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
-			uint32_t (**dump)[2], uint32_t *n_regs)
+			uint32_t (**dump)[2], uint32_t *n_regs, uint32_t inst)
 {
 	uint32_t i = 0, reg;
 #define HQD_N_REGS (35+4)
@@ -213,7 +214,7 @@ static int kgd_hqd_dump(struct amdgpu_device *adev,
 		(*dump)[i++][1] = RREG32(addr);		\
 	} while (0)
 
-	*dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -300,7 +301,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 #undef HQD_N_REGS
 #define HQD_N_REGS (19+4)
 
-	*dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -318,7 +319,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 
 static bool kgd_hqd_is_occupied(struct amdgpu_device *adev,
 				uint64_t queue_address, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	uint32_t act;
 	bool retval = false;
@@ -358,7 +359,7 @@ static bool kgd_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd)
 static int kgd_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	uint32_t temp;
 	enum hqd_dequeue_request_type type;
@@ -494,7 +495,7 @@ static int kgd_hqd_sdma_destroy(struct amdgpu_device *adev, void *mqd,
 
 static int kgd_wave_control_execute(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
-					uint32_t sq_cmd)
+					uint32_t sq_cmd, uint32_t inst)
 {
 	uint32_t data;
 
@@ -560,6 +561,13 @@ static uint32_t read_vmid_from_vmfault_reg(struct amdgpu_device *adev)
 	return REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, VMID);
 }
 
+static uint32_t kgd_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+					  int engine, int queue)
+
+{
+	return 0;
+}
+
 const struct kfd2kgd_calls gfx_v7_kfd2kgd = {
 	.program_sh_mem_settings = kgd_program_sh_mem_settings,
 	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
@@ -577,4 +585,5 @@ const struct kfd2kgd_calls gfx_v7_kfd2kgd = {
 	.set_scratch_backing_va = set_scratch_backing_va,
 	.set_vm_context_page_table_base = set_vm_context_page_table_base,
 	.read_vmid_from_vmfault_reg = read_vmid_from_vmfault_reg,
+	.hqd_sdma_get_doorbell = kgd_hqd_sdma_get_doorbell,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
index 870f352837fc..e68c0fa8d751 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
@@ -72,7 +72,7 @@ static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmi
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
-					uint32_t sh_mem_bases)
+					uint32_t sh_mem_bases, uint32_t inst)
 {
 	lock_srbm(adev, 0, 0, 0, vmid);
 
@@ -85,7 +85,7 @@ static void kgd_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmi
 }
 
 static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
-					unsigned int vmid)
+					unsigned int vmid, uint32_t inst)
 {
 	/*
 	 * We have to assume that there is no outstanding mapping.
@@ -109,7 +109,8 @@ static int kgd_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
 	return 0;
 }
 
-static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id)
+static int kgd_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t inst)
 {
 	uint32_t mec;
 	uint32_t pipe;
@@ -153,7 +154,7 @@ static inline struct vi_sdma_mqd *get_sdma_mqd(void *mqd)
 static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t __user *wptr, uint32_t wptr_shift,
-			uint32_t wptr_mask, struct mm_struct *mm)
+			uint32_t wptr_mask, struct mm_struct *mm, uint32_t inst)
 {
 	struct vi_mqd *m;
 	uint32_t *mqd_hqd;
@@ -226,7 +227,7 @@ static int kgd_hqd_load(struct amdgpu_device *adev, void *mqd,
 
 static int kgd_hqd_dump(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
-			uint32_t (**dump)[2], uint32_t *n_regs)
+			uint32_t (**dump)[2], uint32_t *n_regs, uint32_t inst)
 {
 	uint32_t i = 0, reg;
 #define HQD_N_REGS (54+4)
@@ -237,7 +238,7 @@ static int kgd_hqd_dump(struct amdgpu_device *adev,
 		(*dump)[i++][1] = RREG32(addr);		\
 	} while (0)
 
-	*dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -323,7 +324,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 #undef HQD_N_REGS
 #define HQD_N_REGS (19+4+2+3+7)
 
-	*dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -350,7 +351,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 
 static bool kgd_hqd_is_occupied(struct amdgpu_device *adev,
 				uint64_t queue_address, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	uint32_t act;
 	bool retval = false;
@@ -390,7 +391,7 @@ static bool kgd_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd)
 static int kgd_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	uint32_t temp;
 	enum hqd_dequeue_request_type type;
@@ -540,7 +541,7 @@ static bool get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev,
 
 static int kgd_wave_control_execute(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
-					uint32_t sq_cmd)
+					uint32_t sq_cmd, uint32_t inst)
 {
 	uint32_t data = 0;
 
@@ -581,6 +582,13 @@ static void set_vm_context_page_table_base(struct amdgpu_device *adev,
 			lower_32_bits(page_table_base));
 }
 
+static uint32_t kgd_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+					  int engine, int queue)
+
+{
+	return 0;
+}
+
 const struct kfd2kgd_calls gfx_v8_kfd2kgd = {
 	.program_sh_mem_settings = kgd_program_sh_mem_settings,
 	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
@@ -598,4 +606,5 @@ const struct kfd2kgd_calls gfx_v8_kfd2kgd = {
 			get_atc_vmid_pasid_mapping_info,
 	.set_scratch_backing_va = set_scratch_backing_va,
 	.set_vm_context_page_table_base = set_vm_context_page_table_base,
+	.hqd_sdma_get_doorbell = kgd_hqd_sdma_get_doorbell,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
index e92b93557c13..088d09cc7a72 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
@@ -38,6 +38,7 @@
 #include "soc15d.h"
 #include "gfx_v9_0.h"
 #include "amdgpu_amdkfd_gfx_v9.h"
+#include <uapi/linux/kfd_ioctl.h>
 
 enum hqd_dequeue_request_type {
 	NO_ACTION = 0,
@@ -46,29 +47,29 @@ enum hqd_dequeue_request_type {
 	SAVE_WAVES
 };
 
-static void lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
-			uint32_t queue, uint32_t vmid)
+static void kgd_gfx_v9_lock_srbm(struct amdgpu_device *adev, uint32_t mec, uint32_t pipe,
+			uint32_t queue, uint32_t vmid, uint32_t inst)
 {
 	mutex_lock(&adev->srbm_mutex);
-	soc15_grbm_select(adev, mec, pipe, queue, vmid);
+	soc15_grbm_select(adev, mec, pipe, queue, vmid, GET_INST(GC, inst));
 }
 
-static void unlock_srbm(struct amdgpu_device *adev)
+static void kgd_gfx_v9_unlock_srbm(struct amdgpu_device *adev, uint32_t inst)
 {
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, inst));
 	mutex_unlock(&adev->srbm_mutex);
 }
 
-static void acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id,
-				uint32_t queue_id)
+void kgd_gfx_v9_acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t queue_id, uint32_t inst)
 {
 	uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
 
-	lock_srbm(adev, mec, pipe, queue_id, 0);
+	kgd_gfx_v9_lock_srbm(adev, mec, pipe, queue_id, 0, inst);
 }
 
-static uint64_t get_queue_mask(struct amdgpu_device *adev,
+uint64_t kgd_gfx_v9_get_queue_mask(struct amdgpu_device *adev,
 			       uint32_t pipe_id, uint32_t queue_id)
 {
 	unsigned int bit = pipe_id * adev->gfx.mec.num_queue_per_pipe +
@@ -77,28 +78,28 @@ static uint64_t get_queue_mask(struct amdgpu_device *adev,
 	return 1ull << bit;
 }
 
-static void release_queue(struct amdgpu_device *adev)
+void kgd_gfx_v9_release_queue(struct amdgpu_device *adev, uint32_t inst)
 {
-	unlock_srbm(adev);
+	kgd_gfx_v9_unlock_srbm(adev, inst);
 }
 
 void kgd_gfx_v9_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmid,
 					uint32_t sh_mem_config,
 					uint32_t sh_mem_ape1_base,
 					uint32_t sh_mem_ape1_limit,
-					uint32_t sh_mem_bases)
+					uint32_t sh_mem_bases, uint32_t inst)
 {
-	lock_srbm(adev, 0, 0, 0, vmid);
+	kgd_gfx_v9_lock_srbm(adev, 0, 0, 0, vmid, inst);
 
-	WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_CONFIG), sh_mem_config);
-	WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_BASES), sh_mem_bases);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmSH_MEM_CONFIG, sh_mem_config);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmSH_MEM_BASES, sh_mem_bases);
 	/* APE1 no longer exists on GFX9 */
 
-	unlock_srbm(adev);
+	kgd_gfx_v9_unlock_srbm(adev, inst);
 }
 
 int kgd_gfx_v9_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
-					unsigned int vmid)
+					unsigned int vmid, uint32_t inst)
 {
 	/*
 	 * We have to assume that there is no outstanding mapping.
@@ -156,7 +157,8 @@ int kgd_gfx_v9_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
  * but still works
  */
 
-int kgd_gfx_v9_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id)
+int kgd_gfx_v9_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t inst)
 {
 	uint32_t mec;
 	uint32_t pipe;
@@ -164,13 +166,13 @@ int kgd_gfx_v9_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id)
 	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
 
-	lock_srbm(adev, mec, pipe, 0, 0);
+	kgd_gfx_v9_lock_srbm(adev, mec, pipe, 0, 0, inst);
 
-	WREG32_SOC15(GC, 0, mmCPC_INT_CNTL,
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmCPC_INT_CNTL,
 		CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK |
 		CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK);
 
-	unlock_srbm(adev);
+	kgd_gfx_v9_unlock_srbm(adev, inst);
 
 	return 0;
 }
@@ -220,7 +222,8 @@ static inline struct v9_sdma_mqd *get_sdma_mqd(void *mqd)
 int kgd_gfx_v9_hqd_load(struct amdgpu_device *adev, void *mqd,
 			uint32_t pipe_id, uint32_t queue_id,
 			uint32_t __user *wptr, uint32_t wptr_shift,
-			uint32_t wptr_mask, struct mm_struct *mm)
+			uint32_t wptr_mask, struct mm_struct *mm,
+			uint32_t inst)
 {
 	struct v9_mqd *m;
 	uint32_t *mqd_hqd;
@@ -228,21 +231,21 @@ int kgd_gfx_v9_hqd_load(struct amdgpu_device *adev, void *mqd,
 
 	m = get_mqd(mqd);
 
-	acquire_queue(adev, pipe_id, queue_id);
+	kgd_gfx_v9_acquire_queue(adev, pipe_id, queue_id, inst);
 
 	/* HQD registers extend from CP_MQD_BASE_ADDR to CP_HQD_EOP_WPTR_MEM. */
 	mqd_hqd = &m->cp_mqd_base_addr_lo;
-	hqd_base = SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR);
+	hqd_base = SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_MQD_BASE_ADDR);
 
 	for (reg = hqd_base;
-	     reg <= SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI); reg++)
-		WREG32_RLC(reg, mqd_hqd[reg - hqd_base]);
+	     reg <= SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_HI); reg++)
+		WREG32_XCC(reg, mqd_hqd[reg - hqd_base], inst);
 
 
 	/* Activate doorbell logic before triggering WPTR poll. */
 	data = REG_SET_FIELD(m->cp_hqd_pq_doorbell_control,
 			     CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1);
-	WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL), data);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_HQD_PQ_DOORBELL_CONTROL, data);
 
 	if (wptr) {
 		/* Don't read wptr with get_user because the user
@@ -271,43 +274,42 @@ int kgd_gfx_v9_hqd_load(struct amdgpu_device *adev, void *mqd,
 		guessed_wptr += m->cp_hqd_pq_wptr_lo & ~(queue_size - 1);
 		guessed_wptr += (uint64_t)m->cp_hqd_pq_wptr_hi << 32;
 
-		WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_LO),
-		       lower_32_bits(guessed_wptr));
-		WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI),
-		       upper_32_bits(guessed_wptr));
-		WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR),
-		       lower_32_bits((uintptr_t)wptr));
-		WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI),
-		       upper_32_bits((uintptr_t)wptr));
-		WREG32_SOC15(GC, 0, mmCP_PQ_WPTR_POLL_CNTL1,
-		       (uint32_t)get_queue_mask(adev, pipe_id, queue_id));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_LO,
+			lower_32_bits(guessed_wptr));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_HI,
+			upper_32_bits(guessed_wptr));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_POLL_ADDR,
+			lower_32_bits((uintptr_t)wptr));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_POLL_ADDR_HI,
+			upper_32_bits((uintptr_t)wptr));
+		WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_PQ_WPTR_POLL_CNTL1,
+			(uint32_t)kgd_gfx_v9_get_queue_mask(adev, pipe_id, queue_id));
 	}
 
 	/* Start the EOP fetcher */
-	WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_EOP_RPTR),
-	       REG_SET_FIELD(m->cp_hqd_eop_rptr,
-			     CP_HQD_EOP_RPTR, INIT_FETCHER, 1));
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_HQD_EOP_RPTR,
+	       REG_SET_FIELD(m->cp_hqd_eop_rptr, CP_HQD_EOP_RPTR, INIT_FETCHER, 1));
 
 	data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1);
-	WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE), data);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_HQD_ACTIVE, data);
 
-	release_queue(adev);
+	kgd_gfx_v9_release_queue(adev, inst);
 
 	return 0;
 }
 
 int kgd_gfx_v9_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 			    uint32_t pipe_id, uint32_t queue_id,
-			    uint32_t doorbell_off)
+			    uint32_t doorbell_off, uint32_t inst)
 {
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[inst].ring;
 	struct v9_mqd *m;
 	uint32_t mec, pipe;
 	int r;
 
 	m = get_mqd(mqd);
 
-	acquire_queue(adev, pipe_id, queue_id);
+	kgd_gfx_v9_acquire_queue(adev, pipe_id, queue_id, inst);
 
 	mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1;
 	pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec);
@@ -315,7 +317,7 @@ int kgd_gfx_v9_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 	pr_debug("kfd: set HIQ, mec:%d, pipe:%d, queue:%d.\n",
 		 mec, pipe, queue_id);
 
-	spin_lock(&adev->gfx.kiq.ring_lock);
+	spin_lock(&adev->gfx.kiq[inst].ring_lock);
 	r = amdgpu_ring_alloc(kiq_ring, 7);
 	if (r) {
 		pr_err("Failed to alloc KIQ (%d).\n", r);
@@ -342,15 +344,15 @@ int kgd_gfx_v9_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 	amdgpu_ring_commit(kiq_ring);
 
 out_unlock:
-	spin_unlock(&adev->gfx.kiq.ring_lock);
-	release_queue(adev);
+	spin_unlock(&adev->gfx.kiq[inst].ring_lock);
+	kgd_gfx_v9_release_queue(adev, inst);
 
 	return r;
 }
 
 int kgd_gfx_v9_hqd_dump(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
-			uint32_t (**dump)[2], uint32_t *n_regs)
+			uint32_t (**dump)[2], uint32_t *n_regs, uint32_t inst)
 {
 	uint32_t i = 0, reg;
 #define HQD_N_REGS 56
@@ -361,17 +363,17 @@ int kgd_gfx_v9_hqd_dump(struct amdgpu_device *adev,
 		(*dump)[i++][1] = RREG32(addr);		\
 	} while (0)
 
-	*dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
-	acquire_queue(adev, pipe_id, queue_id);
+	kgd_gfx_v9_acquire_queue(adev, pipe_id, queue_id, inst);
 
-	for (reg = SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR);
-	     reg <= SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI); reg++)
+	for (reg = SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_MQD_BASE_ADDR);
+	     reg <= SOC15_REG_OFFSET(GC, GET_INST(GC, inst), mmCP_HQD_PQ_WPTR_HI); reg++)
 		DUMP_REG(reg);
 
-	release_queue(adev);
+	kgd_gfx_v9_release_queue(adev, inst);
 
 	WARN_ON_ONCE(i != HQD_N_REGS);
 	*n_regs = i;
@@ -458,7 +460,7 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 #undef HQD_N_REGS
 #define HQD_N_REGS (19+6+7+10)
 
-	*dump = kmalloc_array(HQD_N_REGS * 2, sizeof(uint32_t), GFP_KERNEL);
+	*dump = kmalloc_array(HQD_N_REGS, sizeof(**dump), GFP_KERNEL);
 	if (*dump == NULL)
 		return -ENOMEM;
 
@@ -481,23 +483,23 @@ static int kgd_hqd_sdma_dump(struct amdgpu_device *adev,
 
 bool kgd_gfx_v9_hqd_is_occupied(struct amdgpu_device *adev,
 				uint64_t queue_address, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	uint32_t act;
 	bool retval = false;
 	uint32_t low, high;
 
-	acquire_queue(adev, pipe_id, queue_id);
-	act = RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE);
+	kgd_gfx_v9_acquire_queue(adev, pipe_id, queue_id, inst);
+	act = RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_ACTIVE);
 	if (act) {
 		low = lower_32_bits(queue_address >> 8);
 		high = upper_32_bits(queue_address >> 8);
 
-		if (low == RREG32_SOC15(GC, 0, mmCP_HQD_PQ_BASE) &&
-		   high == RREG32_SOC15(GC, 0, mmCP_HQD_PQ_BASE_HI))
+		if (low == RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_PQ_BASE) &&
+		   high == RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_PQ_BASE_HI))
 			retval = true;
 	}
-	release_queue(adev);
+	kgd_gfx_v9_release_queue(adev, inst);
 	return retval;
 }
 
@@ -522,7 +524,7 @@ static bool kgd_hqd_sdma_is_occupied(struct amdgpu_device *adev, void *mqd)
 int kgd_gfx_v9_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
-				uint32_t queue_id)
+				uint32_t queue_id, uint32_t inst)
 {
 	enum hqd_dequeue_request_type type;
 	unsigned long end_jiffies;
@@ -532,10 +534,10 @@ int kgd_gfx_v9_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 	if (amdgpu_in_reset(adev))
 		return -EIO;
 
-	acquire_queue(adev, pipe_id, queue_id);
+	kgd_gfx_v9_acquire_queue(adev, pipe_id, queue_id, inst);
 
 	if (m->cp_hqd_vmid == 0)
-		WREG32_FIELD15_RLC(GC, 0, RLC_CP_SCHEDULERS, scheduler1, 0);
+		WREG32_FIELD15_RLC(GC, GET_INST(GC, inst), RLC_CP_SCHEDULERS, scheduler1, 0);
 
 	switch (reset_type) {
 	case KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN:
@@ -552,22 +554,22 @@ int kgd_gfx_v9_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 		break;
 	}
 
-	WREG32_RLC(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_DEQUEUE_REQUEST), type);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, inst), mmCP_HQD_DEQUEUE_REQUEST, type);
 
 	end_jiffies = (utimeout * HZ / 1000) + jiffies;
 	while (true) {
-		temp = RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE);
+		temp = RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_ACTIVE);
 		if (!(temp & CP_HQD_ACTIVE__ACTIVE_MASK))
 			break;
 		if (time_after(jiffies, end_jiffies)) {
 			pr_err("cp queue preemption time out.\n");
-			release_queue(adev);
+			kgd_gfx_v9_release_queue(adev, inst);
 			return -ETIME;
 		}
 		usleep_range(500, 1000);
 	}
 
-	release_queue(adev);
+	kgd_gfx_v9_release_queue(adev, inst);
 	return 0;
 }
 
@@ -624,14 +626,14 @@ bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev,
 
 int kgd_gfx_v9_wave_control_execute(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
-					uint32_t sq_cmd)
+					uint32_t sq_cmd, uint32_t inst)
 {
 	uint32_t data = 0;
 
 	mutex_lock(&adev->grbm_idx_mutex);
 
-	WREG32_SOC15_RLC_SHADOW(GC, 0, mmGRBM_GFX_INDEX, gfx_index_val);
-	WREG32_SOC15(GC, 0, mmSQ_CMD, sq_cmd);
+	WREG32_SOC15_RLC_SHADOW(GC, GET_INST(GC, inst), mmGRBM_GFX_INDEX, gfx_index_val);
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmSQ_CMD, sq_cmd);
 
 	data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
 		INSTANCE_BROADCAST_WRITES, 1);
@@ -640,12 +642,274 @@ int kgd_gfx_v9_wave_control_execute(struct amdgpu_device *adev,
 	data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
 		SE_BROADCAST_WRITES, 1);
 
-	WREG32_SOC15_RLC_SHADOW(GC, 0, mmGRBM_GFX_INDEX, data);
+	WREG32_SOC15_RLC_SHADOW(GC, GET_INST(GC, inst), mmGRBM_GFX_INDEX, data);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	return 0;
 }
 
+/*
+ * GFX9 helper for wave launch stall requirements on debug trap setting.
+ *
+ * vmid:
+ *   Target VMID to stall/unstall.
+ *
+ * stall:
+ *   0-unstall wave launch (enable), 1-stall wave launch (disable).
+ *   After wavefront launch has been stalled, allocated waves must drain from
+ *   SPI in order for debug trap settings to take effect on those waves.
+ *   This is roughly a ~96 clock cycle wait on SPI where a read on
+ *   SPI_GDBG_WAVE_CNTL translates to ~32 clock cycles.
+ *   KGD_GFX_V9_WAVE_LAUNCH_SPI_DRAIN_LATENCY indicates the number of reads required.
+ *
+ *   NOTE: We can afford to clear the entire STALL_VMID field on unstall
+ *   because GFX9.4.1 cannot support multi-process debugging due to trap
+ *   configuration and masking being limited to global scope.  Always assume
+ *   single process conditions.
+ */
+#define KGD_GFX_V9_WAVE_LAUNCH_SPI_DRAIN_LATENCY	3
+void kgd_gfx_v9_set_wave_launch_stall(struct amdgpu_device *adev,
+					uint32_t vmid,
+					bool stall)
+{
+	int i;
+	uint32_t data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1))
+		data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_VMID,
+							stall ? 1 << vmid : 0);
+	else
+		data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL, STALL_RA,
+							stall ? 1 : 0);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), data);
+
+	if (!stall)
+		return;
+
+	for (i = 0; i < KGD_GFX_V9_WAVE_LAUNCH_SPI_DRAIN_LATENCY; i++)
+		RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+}
+
+/*
+ * restore_dbg_registers is ignored here but is a general interface requirement
+ * for devices that support GFXOFF and where the RLC save/restore list
+ * does not support hw registers for debugging i.e. the driver has to manually
+ * initialize the debug mode registers after it has disabled GFX off during the
+ * debug session.
+ */
+uint32_t kgd_gfx_v9_enable_debug_trap(struct amdgpu_device *adev,
+				bool restore_dbg_registers,
+				uint32_t vmid)
+{
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+/*
+ * keep_trap_enabled is ignored here but is a general interface requirement
+ * for devices that support multi-process debugging where the performance
+ * overhead from trap temporary setup needs to be bypassed when the debug
+ * session has ended.
+ */
+uint32_t kgd_gfx_v9_disable_debug_trap(struct amdgpu_device *adev,
+					bool keep_trap_enabled,
+					uint32_t vmid)
+{
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+int kgd_gfx_v9_validate_trap_override_request(struct amdgpu_device *adev,
+					uint32_t trap_override,
+					uint32_t *trap_mask_supported)
+{
+	*trap_mask_supported &= KFD_DBG_TRAP_MASK_DBG_ADDRESS_WATCH;
+
+	/* The SPI_GDBG_TRAP_MASK register is global and affects all
+	 * processes. Only allow OR-ing the address-watch bit, since
+	 * this only affects processes under the debugger. Other bits
+	 * should stay 0 to avoid the debugger interfering with other
+	 * processes.
+	 */
+	if (trap_override != KFD_DBG_TRAP_OVERRIDE_OR)
+		return -EINVAL;
+
+	return 0;
+}
+
+uint32_t kgd_gfx_v9_set_wave_launch_trap_override(struct amdgpu_device *adev,
+					     uint32_t vmid,
+					     uint32_t trap_override,
+					     uint32_t trap_mask_bits,
+					     uint32_t trap_mask_request,
+					     uint32_t *trap_mask_prev,
+					     uint32_t kfd_dbg_cntl_prev)
+{
+	uint32_t data, wave_cntl_prev;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	wave_cntl_prev = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL));
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+	data = RREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK));
+	*trap_mask_prev = REG_GET_FIELD(data, SPI_GDBG_TRAP_MASK, EXCP_EN);
+
+	trap_mask_bits = (trap_mask_bits & trap_mask_request) |
+		(*trap_mask_prev & ~trap_mask_request);
+
+	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK, EXCP_EN, trap_mask_bits);
+	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_MASK, REPLACE, trap_override);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), data);
+
+	/* We need to preserve wave launch mode stall settings. */
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL), wave_cntl_prev);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+uint32_t kgd_gfx_v9_set_wave_launch_mode(struct amdgpu_device *adev,
+					uint8_t wave_launch_mode,
+					uint32_t vmid)
+{
+	uint32_t data = 0;
+	bool is_mode_set = !!wave_launch_mode;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, true);
+
+	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
+		VMID_MASK, is_mode_set ? 1 << vmid : 0);
+	data = REG_SET_FIELD(data, SPI_GDBG_WAVE_CNTL2,
+		MODE, is_mode_set ? wave_launch_mode : 0);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_WAVE_CNTL2), data);
+
+	kgd_gfx_v9_set_wave_launch_stall(adev, vmid, false);
+
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	return 0;
+}
+
+#define TCP_WATCH_STRIDE (mmTCP_WATCH1_ADDR_H - mmTCP_WATCH0_ADDR_H)
+uint32_t kgd_gfx_v9_set_address_watch(struct amdgpu_device *adev,
+					uint64_t watch_address,
+					uint32_t watch_address_mask,
+					uint32_t watch_id,
+					uint32_t watch_mode,
+					uint32_t debug_vmid,
+					uint32_t inst)
+{
+	uint32_t watch_address_high;
+	uint32_t watch_address_low;
+	uint32_t watch_address_cntl;
+
+	watch_address_cntl = 0;
+
+	watch_address_low = lower_32_bits(watch_address);
+	watch_address_high = upper_32_bits(watch_address) & 0xffff;
+
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VMID,
+			debug_vmid);
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MODE,
+			watch_mode);
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			MASK,
+			watch_address_mask >> 6);
+
+	/* Turning off this watch point until we set all the registers */
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VALID,
+			0);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_cntl);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_ADDR_H) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_high);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_ADDR_L) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_low);
+
+	/* Enable the watch point */
+	watch_address_cntl = REG_SET_FIELD(watch_address_cntl,
+			TCP_WATCH0_CNTL,
+			VALID,
+			1);
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_cntl);
+
+	return 0;
+}
+
+uint32_t kgd_gfx_v9_clear_address_watch(struct amdgpu_device *adev,
+					uint32_t watch_id)
+{
+	uint32_t watch_address_cntl;
+
+	watch_address_cntl = 0;
+
+	WREG32_RLC((SOC15_REG_OFFSET(GC, 0, mmTCP_WATCH0_CNTL) +
+			(watch_id * TCP_WATCH_STRIDE)),
+			watch_address_cntl);
+
+	return 0;
+}
+
+/* kgd_gfx_v9_get_iq_wait_times: Returns the mmCP_IQ_WAIT_TIME1/2 values
+ * The values read are:
+ *     ib_offload_wait_time     -- Wait Count for Indirect Buffer Offloads.
+ *     atomic_offload_wait_time -- Wait Count for L2 and GDS Atomics Offloads.
+ *     wrm_offload_wait_time    -- Wait Count for WAIT_REG_MEM Offloads.
+ *     gws_wait_time            -- Wait Count for Global Wave Syncs.
+ *     que_sleep_wait_time      -- Wait Count for Dequeue Retry.
+ *     sch_wave_wait_time       -- Wait Count for Scheduling Wave Message.
+ *     sem_rearm_wait_time      -- Wait Count for Semaphore re-arm.
+ *     deq_retry_wait_time      -- Wait Count for Global Wave Syncs.
+ */
+void kgd_gfx_v9_get_iq_wait_times(struct amdgpu_device *adev,
+					uint32_t *wait_times,
+					uint32_t inst)
+
+{
+	*wait_times = RREG32_SOC15_RLC(GC, GET_INST(GC, inst),
+			mmCP_IQ_WAIT_TIME2);
+}
+
 void kgd_gfx_v9_set_vm_context_page_table_base(struct amdgpu_device *adev,
 			uint32_t vmid, uint64_t page_table_base)
 {
@@ -680,33 +944,34 @@ static void unlock_spi_csq_mutexes(struct amdgpu_device *adev)
  *
  * @adev: Handle of device whose registers are to be read
  * @queue_idx: Index of queue in the queue-map bit-field
- * @wave_cnt: Output parameter updated with number of waves in flight
- * @vmid: Output parameter updated with VMID of queue whose wave count
- * is being collected
+ * @queue_cnt: Stores the wave count and doorbell offset for an active queue
+ * @inst: xcc's instance number on a multi-XCC setup
  */
 static void get_wave_count(struct amdgpu_device *adev, int queue_idx,
-		int *wave_cnt, int *vmid)
+		struct kfd_cu_occupancy *queue_cnt, uint32_t inst)
 {
 	int pipe_idx;
 	int queue_slot;
 	unsigned int reg_val;
-
+	unsigned int wave_cnt;
 	/*
 	 * Program GRBM with appropriate MEID, PIPEID, QUEUEID and VMID
 	 * parameters to read out waves in flight. Get VMID if there are
 	 * non-zero waves in flight.
 	 */
-	*vmid = 0xFF;
-	*wave_cnt = 0;
 	pipe_idx = queue_idx / adev->gfx.mec.num_queue_per_pipe;
 	queue_slot = queue_idx % adev->gfx.mec.num_queue_per_pipe;
-	soc15_grbm_select(adev, 1, pipe_idx, queue_slot, 0);
-	reg_val = RREG32_SOC15_IP(GC, SOC15_REG_OFFSET(GC, 0, mmSPI_CSQ_WF_ACTIVE_COUNT_0) +
-			 queue_slot);
-	*wave_cnt = reg_val & SPI_CSQ_WF_ACTIVE_COUNT_0__COUNT_MASK;
-	if (*wave_cnt != 0)
-		*vmid = (RREG32_SOC15(GC, 0, mmCP_HQD_VMID) &
-			 CP_HQD_VMID__VMID_MASK) >> CP_HQD_VMID__VMID__SHIFT;
+	soc15_grbm_select(adev, 1, pipe_idx, queue_slot, 0, GET_INST(GC, inst));
+	reg_val = RREG32_SOC15_IP(GC, SOC15_REG_OFFSET(GC, GET_INST(GC, inst),
+				  mmSPI_CSQ_WF_ACTIVE_COUNT_0) + queue_slot);
+	wave_cnt = reg_val & SPI_CSQ_WF_ACTIVE_COUNT_0__COUNT_MASK;
+	if (wave_cnt != 0) {
+		queue_cnt->wave_cnt += wave_cnt;
+		queue_cnt->doorbell_off =
+			(RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_PQ_DOORBELL_CONTROL) &
+			 CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET_MASK) >>
+			 CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET__SHIFT;
+	}
 }
 
 /**
@@ -716,11 +981,11 @@ static void get_wave_count(struct amdgpu_device *adev, int queue_idx,
  * or more queues running and submitting waves to compute units.
  *
  * @adev: Handle of device from which to get number of waves in flight
- * @pasid: Identifies the process for which this query call is invoked
- * @pasid_wave_cnt: Output parameter updated with number of waves in flight that
- * belong to process with given pasid
+ * @cu_occupancy: Array that gets filled with wave_cnt and doorbell offset
+ *		  for comparison later.
  * @max_waves_per_cu: Output parameter updated with maximum number of waves
- * possible per Compute Unit
+ *                    possible per Compute Unit
+ * @inst: xcc's instance number on a multi-XCC setup
  *
  * Note: It's possible that the device has too many queues (oversubscription)
  * in which case a VMID could be remapped to a different PASID. This could lead
@@ -745,113 +1010,224 @@ static void get_wave_count(struct amdgpu_device *adev, int queue_idx,
  *    number of waves that are in flight for the queue at specified index. The
  *    index ranges from 0 to 7.
  *
- *    If non-zero waves are in flight, read CP_HQD_VMID register to obtain VMID
- *    of the wave(s).
+ *    If non-zero waves are in flight, store the corresponding doorbell offset
+ *    of the queue, along with the wave count.
  *
- *    Determine if VMID from above step maps to pasid provided as parameter. If
- *    it matches agrregate the wave count. That the VMID will not match pasid is
- *    a normal condition i.e. a device is expected to support multiple queues
- *    from multiple proceses.
+ *    Determine if the queue belongs to the process by comparing the doorbell
+ *    offset against the process's queues. If it matches, aggregate the wave
+ *    count for the process.
  *
  *  Reading registers referenced above involves programming GRBM appropriately
  */
-void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev, int pasid,
-		int *pasid_wave_cnt, int *max_waves_per_cu)
+void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev,
+				 struct kfd_cu_occupancy *cu_occupancy,
+				 int *max_waves_per_cu, uint32_t inst)
 {
 	int qidx;
-	int vmid;
 	int se_idx;
-	int sh_idx;
 	int se_cnt;
-	int sh_cnt;
-	int wave_cnt;
 	int queue_map;
-	int pasid_tmp;
 	int max_queue_cnt;
-	int vmid_wave_cnt = 0;
-	DECLARE_BITMAP(cp_queue_bitmap, KGD_MAX_QUEUES);
+	DECLARE_BITMAP(cp_queue_bitmap, AMDGPU_MAX_QUEUES);
 
 	lock_spi_csq_mutexes(adev);
-	soc15_grbm_select(adev, 1, 0, 0, 0);
+	soc15_grbm_select(adev, 1, 0, 0, 0, GET_INST(GC, inst));
 
 	/*
 	 * Iterate through the shader engines and arrays of the device
 	 * to get number of waves in flight
 	 */
-	bitmap_complement(cp_queue_bitmap, adev->gfx.mec.queue_bitmap,
-			  KGD_MAX_QUEUES);
+	bitmap_complement(cp_queue_bitmap, adev->gfx.mec_bitmap[0].queue_bitmap,
+			  AMDGPU_MAX_QUEUES);
 	max_queue_cnt = adev->gfx.mec.num_pipe_per_mec *
 			adev->gfx.mec.num_queue_per_pipe;
-	sh_cnt = adev->gfx.config.max_sh_per_se;
 	se_cnt = adev->gfx.config.max_shader_engines;
 	for (se_idx = 0; se_idx < se_cnt; se_idx++) {
-		for (sh_idx = 0; sh_idx < sh_cnt; sh_idx++) {
+		amdgpu_gfx_select_se_sh(adev, se_idx, 0, 0xffffffff, inst);
+		queue_map = RREG32_SOC15(GC, GET_INST(GC, inst), mmSPI_CSQ_WF_ACTIVE_STATUS);
+
+		/*
+		 * Assumption: queue map encodes following schema: four
+		 * pipes per each micro-engine, with each pipe mapping
+		 * eight queues. This schema is true for GFX9 devices
+		 * and must be verified for newer device families
+		 */
+		for (qidx = 0; qidx < max_queue_cnt; qidx++) {
+			/* Skip qeueus that are not associated with
+			 * compute functions
+			 */
+			if (!test_bit(qidx, cp_queue_bitmap))
+				continue;
 
-			amdgpu_gfx_select_se_sh(adev, se_idx, sh_idx, 0xffffffff);
-			queue_map = RREG32_SOC15(GC, 0, mmSPI_CSQ_WF_ACTIVE_STATUS);
+			if (!(queue_map & (1 << qidx)))
+				continue;
 
-			/*
-			 * Assumption: queue map encodes following schema: four
-			 * pipes per each micro-engine, with each pipe mapping
-			 * eight queues. This schema is true for GFX9 devices
-			 * and must be verified for newer device families
-			 */
-			for (qidx = 0; qidx < max_queue_cnt; qidx++) {
-
-				/* Skip qeueus that are not associated with
-				 * compute functions
-				 */
-				if (!test_bit(qidx, cp_queue_bitmap))
-					continue;
-
-				if (!(queue_map & (1 << qidx)))
-					continue;
-
-				/* Get number of waves in flight and aggregate them */
-				get_wave_count(adev, qidx, &wave_cnt, &vmid);
-				if (wave_cnt != 0) {
-					pasid_tmp =
-					  RREG32(SOC15_REG_OFFSET(OSSSYS, 0,
-						 mmIH_VMID_0_LUT) + vmid);
-					if (pasid_tmp == pasid)
-						vmid_wave_cnt += wave_cnt;
-				}
-			}
+			/* Get number of waves in flight and aggregate them */
+			get_wave_count(adev, qidx, &cu_occupancy[qidx],
+					inst);
 		}
 	}
 
-	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, inst);
+	soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, inst));
 	unlock_spi_csq_mutexes(adev);
 
 	/* Update the output parameters and return */
-	*pasid_wave_cnt = vmid_wave_cnt;
 	*max_waves_per_cu = adev->gfx.cu_info.simd_per_cu *
 				adev->gfx.cu_info.max_waves_per_simd;
 }
 
+void kgd_gfx_v9_build_dequeue_wait_counts_packet_info(struct amdgpu_device *adev,
+		uint32_t wait_times,
+		uint32_t sch_wave,
+		uint32_t que_sleep,
+		uint32_t *reg_offset,
+		uint32_t *reg_data)
+{
+	*reg_data = wait_times;
+
+	if (sch_wave)
+		*reg_data = REG_SET_FIELD(*reg_data,
+				CP_IQ_WAIT_TIME2,
+				SCH_WAVE,
+				sch_wave);
+	if (que_sleep)
+		*reg_data = REG_SET_FIELD(*reg_data,
+				CP_IQ_WAIT_TIME2,
+				QUE_SLEEP,
+				que_sleep);
+
+	*reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_IQ_WAIT_TIME2);
+}
+
 void kgd_gfx_v9_program_trap_handler_settings(struct amdgpu_device *adev,
-                        uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr)
+		uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr, uint32_t inst)
 {
-	lock_srbm(adev, 0, 0, 0, vmid);
+	kgd_gfx_v9_lock_srbm(adev, 0, 0, 0, vmid, inst);
 
 	/*
 	 * Program TBA registers
 	 */
-	WREG32_SOC15(GC, 0, mmSQ_SHADER_TBA_LO,
-                        lower_32_bits(tba_addr >> 8));
-	WREG32_SOC15(GC, 0, mmSQ_SHADER_TBA_HI,
-                        upper_32_bits(tba_addr >> 8));
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmSQ_SHADER_TBA_LO,
+			lower_32_bits(tba_addr >> 8));
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmSQ_SHADER_TBA_HI,
+			upper_32_bits(tba_addr >> 8));
 
 	/*
 	 * Program TMA registers
 	 */
-	WREG32_SOC15(GC, 0, mmSQ_SHADER_TMA_LO,
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmSQ_SHADER_TMA_LO,
 			lower_32_bits(tma_addr >> 8));
-	WREG32_SOC15(GC, 0, mmSQ_SHADER_TMA_HI,
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmSQ_SHADER_TMA_HI,
 			upper_32_bits(tma_addr >> 8));
 
-	unlock_srbm(adev);
+	kgd_gfx_v9_unlock_srbm(adev, inst);
+}
+
+uint64_t kgd_gfx_v9_hqd_get_pq_addr(struct amdgpu_device *adev,
+				    uint32_t pipe_id, uint32_t queue_id,
+				    uint32_t inst)
+{
+	uint32_t low, high;
+	uint64_t queue_addr = 0;
+
+	kgd_gfx_v9_acquire_queue(adev, pipe_id, queue_id, inst);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, inst);
+
+	if (!RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_ACTIVE))
+		goto unlock_out;
+
+	low = RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_PQ_BASE);
+	high = RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_PQ_BASE_HI);
+
+	/* only concerned with user queues. */
+	if (!high)
+		goto unlock_out;
+
+	queue_addr = (((queue_addr | high) << 32) | low) << 8;
+
+unlock_out:
+	amdgpu_gfx_rlc_exit_safe_mode(adev, inst);
+	kgd_gfx_v9_release_queue(adev, inst);
+
+	return queue_addr;
+}
+
+/* assume queue acquired  */
+static int kgd_gfx_v9_hqd_dequeue_wait(struct amdgpu_device *adev, uint32_t inst,
+				       unsigned int utimeout)
+{
+	unsigned long end_jiffies = (utimeout * HZ / 1000) + jiffies;
+
+	while (true) {
+		uint32_t temp = RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_ACTIVE);
+
+		if (!(temp & CP_HQD_ACTIVE__ACTIVE_MASK))
+			return 0;
+
+		if (time_after(jiffies, end_jiffies))
+			return -ETIME;
+
+		usleep_range(500, 1000);
+	}
+}
+
+uint64_t kgd_gfx_v9_hqd_reset(struct amdgpu_device *adev,
+			      uint32_t pipe_id, uint32_t queue_id,
+			      uint32_t inst, unsigned int utimeout)
+{
+	uint32_t low, high, pipe_reset_data = 0;
+	uint64_t queue_addr = 0;
+
+	kgd_gfx_v9_acquire_queue(adev, pipe_id, queue_id, inst);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, inst);
+
+	if (!RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_ACTIVE))
+		goto unlock_out;
+
+	low = RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_PQ_BASE);
+	high = RREG32_SOC15(GC, GET_INST(GC, inst), mmCP_HQD_PQ_BASE_HI);
+
+	/* only concerned with user queues. */
+	if (!high)
+		goto unlock_out;
+
+	queue_addr = (((queue_addr | high) << 32) | low) << 8;
+
+	pr_debug("Attempting queue reset on XCC %i pipe id %i queue id %i\n",
+		 inst, pipe_id, queue_id);
+
+	/* assume previous dequeue request issued will take affect after reset */
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmSPI_COMPUTE_QUEUE_RESET, 0x1);
+
+	if (!kgd_gfx_v9_hqd_dequeue_wait(adev, inst, utimeout))
+		goto unlock_out;
+
+	pr_debug("Attempting pipe reset on XCC %i pipe id %i\n", inst, pipe_id);
+
+	pipe_reset_data = REG_SET_FIELD(pipe_reset_data, CP_MEC_CNTL, MEC_ME1_PIPE0_RESET, 1);
+	pipe_reset_data = pipe_reset_data << pipe_id;
+
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmCP_MEC_CNTL, pipe_reset_data);
+	WREG32_SOC15(GC, GET_INST(GC, inst), mmCP_MEC_CNTL, 0);
+
+	if (kgd_gfx_v9_hqd_dequeue_wait(adev, inst, utimeout))
+		queue_addr = 0;
+
+unlock_out:
+	pr_debug("queue reset on XCC %i pipe id %i queue id %i %s\n",
+		 inst, pipe_id, queue_id, !!queue_addr ? "succeeded!" : "failed!");
+	amdgpu_gfx_rlc_exit_safe_mode(adev, inst);
+	kgd_gfx_v9_release_queue(adev, inst);
+
+	return queue_addr;
+}
+
+uint32_t kgd_gfx_v9_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+					  int engine, int queue)
+
+{
+	return 0;
 }
 
 const struct kfd2kgd_calls gfx_v9_kfd2kgd = {
@@ -871,6 +1247,18 @@ const struct kfd2kgd_calls gfx_v9_kfd2kgd = {
 	.get_atc_vmid_pasid_mapping_info =
 			kgd_gfx_v9_get_atc_vmid_pasid_mapping_info,
 	.set_vm_context_page_table_base = kgd_gfx_v9_set_vm_context_page_table_base,
+	.enable_debug_trap = kgd_gfx_v9_enable_debug_trap,
+	.disable_debug_trap = kgd_gfx_v9_disable_debug_trap,
+	.validate_trap_override_request = kgd_gfx_v9_validate_trap_override_request,
+	.set_wave_launch_trap_override = kgd_gfx_v9_set_wave_launch_trap_override,
+	.set_wave_launch_mode = kgd_gfx_v9_set_wave_launch_mode,
+	.set_address_watch = kgd_gfx_v9_set_address_watch,
+	.clear_address_watch = kgd_gfx_v9_clear_address_watch,
+	.get_iq_wait_times = kgd_gfx_v9_get_iq_wait_times,
+	.build_dequeue_wait_counts_packet_info = kgd_gfx_v9_build_dequeue_wait_counts_packet_info,
 	.get_cu_occupancy = kgd_gfx_v9_get_cu_occupancy,
 	.program_trap_handler_settings = kgd_gfx_v9_program_trap_handler_settings,
+	.hqd_get_pq_addr = kgd_gfx_v9_hqd_get_pq_addr,
+	.hqd_reset = kgd_gfx_v9_hqd_reset,
+	.hqd_sdma_get_doorbell = kgd_gfx_v9_hqd_sdma_get_doorbell
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
index c7ed3bc9053c..704452ca62f8 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.h
@@ -20,41 +20,97 @@
  * OTHER DEALINGS IN THE SOFTWARE.
  */
 
-
-
 void kgd_gfx_v9_program_sh_mem_settings(struct amdgpu_device *adev, uint32_t vmid,
 		uint32_t sh_mem_config,
 		uint32_t sh_mem_ape1_base, uint32_t sh_mem_ape1_limit,
-		uint32_t sh_mem_bases);
+		uint32_t sh_mem_bases, uint32_t inst);
 int kgd_gfx_v9_set_pasid_vmid_mapping(struct amdgpu_device *adev, u32 pasid,
-		unsigned int vmid);
-int kgd_gfx_v9_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id);
+		unsigned int vmid, uint32_t inst);
+int kgd_gfx_v9_init_interrupts(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t inst);
 int kgd_gfx_v9_hqd_load(struct amdgpu_device *adev, void *mqd, uint32_t pipe_id,
 			uint32_t queue_id, uint32_t __user *wptr,
 			uint32_t wptr_shift, uint32_t wptr_mask,
-			struct mm_struct *mm);
+			struct mm_struct *mm, uint32_t inst);
 int kgd_gfx_v9_hiq_mqd_load(struct amdgpu_device *adev, void *mqd,
 			    uint32_t pipe_id, uint32_t queue_id,
-			    uint32_t doorbell_off);
+			    uint32_t doorbell_off, uint32_t inst);
 int kgd_gfx_v9_hqd_dump(struct amdgpu_device *adev,
 			uint32_t pipe_id, uint32_t queue_id,
-			uint32_t (**dump)[2], uint32_t *n_regs);
+			uint32_t (**dump)[2], uint32_t *n_regs, uint32_t inst);
 bool kgd_gfx_v9_hqd_is_occupied(struct amdgpu_device *adev,
 			uint64_t queue_address, uint32_t pipe_id,
-			uint32_t queue_id);
+			uint32_t queue_id, uint32_t inst);
 int kgd_gfx_v9_hqd_destroy(struct amdgpu_device *adev, void *mqd,
 				enum kfd_preempt_type reset_type,
 				unsigned int utimeout, uint32_t pipe_id,
-				uint32_t queue_id);
+				uint32_t queue_id, uint32_t inst);
 int kgd_gfx_v9_wave_control_execute(struct amdgpu_device *adev,
 					uint32_t gfx_index_val,
-					uint32_t sq_cmd);
+					uint32_t sq_cmd, uint32_t inst);
 bool kgd_gfx_v9_get_atc_vmid_pasid_mapping_info(struct amdgpu_device *adev,
 					uint8_t vmid, uint16_t *p_pasid);
-
 void kgd_gfx_v9_set_vm_context_page_table_base(struct amdgpu_device *adev,
 			uint32_t vmid, uint64_t page_table_base);
-void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev, int pasid,
-		int *pasid_wave_cnt, int *max_waves_per_cu);
+void kgd_gfx_v9_get_cu_occupancy(struct amdgpu_device *adev,
+				 struct kfd_cu_occupancy *cu_occupancy,
+				 int *max_waves_per_cu, uint32_t inst);
 void kgd_gfx_v9_program_trap_handler_settings(struct amdgpu_device *adev,
-		uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr);
+		uint32_t vmid, uint64_t tba_addr, uint64_t tma_addr,
+		uint32_t inst);
+void kgd_gfx_v9_acquire_queue(struct amdgpu_device *adev, uint32_t pipe_id,
+				uint32_t queue_id, uint32_t inst);
+uint64_t kgd_gfx_v9_get_queue_mask(struct amdgpu_device *adev,
+				uint32_t pipe_id, uint32_t queue_id);
+void kgd_gfx_v9_release_queue(struct amdgpu_device *adev, uint32_t inst);
+void kgd_gfx_v9_set_wave_launch_stall(struct amdgpu_device *adev,
+					uint32_t vmid,
+					bool stall);
+uint32_t kgd_gfx_v9_enable_debug_trap(struct amdgpu_device *adev,
+				      bool restore_dbg_registers,
+				      uint32_t vmid);
+uint32_t kgd_gfx_v9_disable_debug_trap(struct amdgpu_device *adev,
+					bool keep_trap_enabled,
+					uint32_t vmid);
+int kgd_gfx_v9_validate_trap_override_request(struct amdgpu_device *adev,
+					     uint32_t trap_override,
+					     uint32_t *trap_mask_supported);
+uint32_t kgd_gfx_v9_set_wave_launch_mode(struct amdgpu_device *adev,
+					uint8_t wave_launch_mode,
+					uint32_t vmid);
+uint32_t kgd_gfx_v9_set_wave_launch_trap_override(struct amdgpu_device *adev,
+					     uint32_t vmid,
+					     uint32_t trap_override,
+					     uint32_t trap_mask_bits,
+					     uint32_t trap_mask_request,
+					     uint32_t *trap_mask_prev,
+					     uint32_t kfd_dbg_trap_cntl_prev);
+uint32_t kgd_gfx_v9_set_address_watch(struct amdgpu_device *adev,
+					uint64_t watch_address,
+					uint32_t watch_address_mask,
+					uint32_t watch_id,
+					uint32_t watch_mode,
+					uint32_t debug_vmid,
+					uint32_t inst);
+uint32_t kgd_gfx_v9_clear_address_watch(struct amdgpu_device *adev,
+					uint32_t watch_id);
+void kgd_gfx_v9_get_iq_wait_times(struct amdgpu_device *adev,
+				uint32_t *wait_times,
+				uint32_t inst);
+void kgd_gfx_v9_build_dequeue_wait_counts_packet_info(struct amdgpu_device *adev,
+					       uint32_t wait_times,
+					       uint32_t sch_wave,
+					       uint32_t que_sleep,
+					       uint32_t *reg_offset,
+					       uint32_t *reg_data);
+uint64_t kgd_gfx_v9_hqd_get_pq_addr(struct amdgpu_device *adev,
+				    uint32_t pipe_id,
+				    uint32_t queue_id,
+				    uint32_t inst);
+uint64_t kgd_gfx_v9_hqd_reset(struct amdgpu_device *adev,
+			      uint32_t pipe_id,
+			      uint32_t queue_id,
+			      uint32_t inst,
+			      unsigned int utimeout);
+uint32_t kgd_gfx_v9_hqd_sdma_get_doorbell(struct amdgpu_device *adev,
+					  int engine, int queue);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
index c87515210c4f..b1c24c8fa686 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
@@ -27,6 +27,8 @@
 #include <linux/sched/task.h>
 #include <drm/ttm/ttm_tt.h>
 
+#include <drm/drm_exec.h>
+
 #include "amdgpu_object.h"
 #include "amdgpu_gem.h"
 #include "amdgpu_vm.h"
@@ -35,12 +37,14 @@
 #include "amdgpu_dma_buf.h"
 #include <uapi/linux/kfd_ioctl.h>
 #include "amdgpu_xgmi.h"
+#include "kfd_priv.h"
 #include "kfd_smi_events.h"
 
 /* Userptr restore delay, just long enough to allow consecutive VM
  * changes to accumulate
  */
 #define AMDGPU_USERPTR_RESTORE_DELAY_MS 1
+#define AMDGPU_RESERVE_MEM_LIMIT			(3UL << 29)
 
 /*
  * Align VRAM availability to 2MB to avoid fragmentation caused by 4K allocations in the tail 2MB
@@ -82,6 +86,25 @@ static bool kfd_mem_is_attached(struct amdgpu_vm *avm,
 	return false;
 }
 
+/**
+ * reuse_dmamap() - Check whether adev can share the original
+ * userptr BO
+ *
+ * If both adev and bo_adev are in direct mapping or
+ * in the same iommu group, they can share the original BO.
+ *
+ * @adev: Device to which can or cannot share the original BO
+ * @bo_adev: Device to which allocated BO belongs to
+ *
+ * Return: returns true if adev can share original userptr BO,
+ * false otherwise.
+ */
+static bool reuse_dmamap(struct amdgpu_device *adev, struct amdgpu_device *bo_adev)
+{
+	return (adev->ram_is_direct_mapped && bo_adev->ram_is_direct_mapped) ||
+			(adev->dev->iommu_group == bo_adev->dev->iommu_group);
+}
+
 /* Set memory usage limits. Current, limits are
  *  System (TTM + userptr) memory - 15/16th System RAM
  *  TTM memory - 3/8th System RAM
@@ -91,13 +114,21 @@ void amdgpu_amdkfd_gpuvm_init_mem_limits(void)
 	struct sysinfo si;
 	uint64_t mem;
 
+	if (kfd_mem_limit.max_system_mem_limit)
+		return;
+
 	si_meminfo(&si);
-	mem = si.freeram - si.freehigh;
+	mem = si.totalram - si.totalhigh;
 	mem *= si.mem_unit;
 
 	spin_lock_init(&kfd_mem_limit.mem_limit_lock);
-	kfd_mem_limit.max_system_mem_limit = mem - (mem >> 4);
-	kfd_mem_limit.max_ttm_mem_limit = (mem >> 1) - (mem >> 3);
+	kfd_mem_limit.max_system_mem_limit = mem - (mem >> 6);
+	if (kfd_mem_limit.max_system_mem_limit < 2 * AMDGPU_RESERVE_MEM_LIMIT)
+		kfd_mem_limit.max_system_mem_limit >>= 1;
+	else
+		kfd_mem_limit.max_system_mem_limit -= AMDGPU_RESERVE_MEM_LIMIT;
+
+	kfd_mem_limit.max_ttm_mem_limit = ttm_tt_pages_limit() << PAGE_SHIFT;
 	pr_debug("Kernel memory limit %lluM, TTM limit %lluM\n",
 		(kfd_mem_limit.max_system_mem_limit >> 20),
 		(kfd_mem_limit.max_ttm_mem_limit >> 20));
@@ -129,16 +160,22 @@ void amdgpu_amdkfd_reserve_system_mem(uint64_t size)
  * @size: Size of buffer, in bytes, encapsulated by B0. This should be
  * equivalent to amdgpu_bo_size(BO)
  * @alloc_flag: Flag used in allocating a BO as noted above
+ * @xcp_id: xcp_id is used to get xcp from xcp manager, one xcp is
+ * managed as one compute node in driver for app
  *
- * Return: returns -ENOMEM in case of error, ZERO otherwise
+ * Return:
+ *	returns -ENOMEM in case of error, ZERO otherwise
  */
 int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
-		uint64_t size, u32 alloc_flag)
+		uint64_t size, u32 alloc_flag, int8_t xcp_id)
 {
 	uint64_t reserved_for_pt =
 		ESTIMATE_PT_SIZE(amdgpu_amdkfd_total_mem_size);
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	uint64_t reserved_for_ras = (con ? con->reserved_pages_in_bytes : 0);
 	size_t system_mem_needed, ttm_mem_needed, vram_needed;
 	int ret = 0;
+	uint64_t vram_size = 0;
 
 	system_mem_needed = 0;
 	ttm_mem_needed = 0;
@@ -153,6 +190,17 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
 		 * 2M BO chunk.
 		 */
 		vram_needed = size;
+		/*
+		 * For GFX 9.4.3, get the VRAM size from XCP structs
+		 */
+		if (WARN_ONCE(xcp_id < 0, "invalid XCP ID %d", xcp_id))
+			return -EINVAL;
+
+		vram_size = KFD_XCP_MEMORY_SIZE(adev, xcp_id);
+		if (adev->apu_prefer_gtt) {
+			system_mem_needed = size;
+			ttm_mem_needed = size;
+		}
 	} else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) {
 		system_mem_needed = size;
 	} else if (!(alloc_flag &
@@ -165,27 +213,46 @@ int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev,
 	spin_lock(&kfd_mem_limit.mem_limit_lock);
 
 	if (kfd_mem_limit.system_mem_used + system_mem_needed >
-	    kfd_mem_limit.max_system_mem_limit)
+	    kfd_mem_limit.max_system_mem_limit) {
 		pr_debug("Set no_system_mem_limit=1 if using shared memory\n");
+		if (!no_system_mem_limit) {
+			ret = -ENOMEM;
+			goto release;
+		}
+	}
 
-	if ((kfd_mem_limit.system_mem_used + system_mem_needed >
-	     kfd_mem_limit.max_system_mem_limit && !no_system_mem_limit) ||
-	    (kfd_mem_limit.ttm_mem_used + ttm_mem_needed >
-	     kfd_mem_limit.max_ttm_mem_limit) ||
-	    (adev && adev->kfd.vram_used + vram_needed >
-	     adev->gmc.real_vram_size - reserved_for_pt)) {
+	if (kfd_mem_limit.ttm_mem_used + ttm_mem_needed >
+		kfd_mem_limit.max_ttm_mem_limit) {
 		ret = -ENOMEM;
 		goto release;
 	}
 
+	/*if is_app_apu is false and apu_prefer_gtt is true, it is an APU with
+	 * carve out < gtt. In that case, VRAM allocation will go to gtt domain, skip
+	 * VRAM check since ttm_mem_limit check already cover this allocation
+	 */
+
+	if (adev && xcp_id >= 0 && (!adev->apu_prefer_gtt || adev->gmc.is_app_apu)) {
+		uint64_t vram_available =
+			vram_size - reserved_for_pt - reserved_for_ras -
+			atomic64_read(&adev->vram_pin_size);
+		if (adev->kfd.vram_used[xcp_id] + vram_needed > vram_available) {
+			ret = -ENOMEM;
+			goto release;
+		}
+	}
+
 	/* Update memory accounting by decreasing available system
 	 * memory, TTM memory and GPU memory as computed above
 	 */
 	WARN_ONCE(vram_needed && !adev,
 		  "adev reference can't be null when vram is used");
-	if (adev) {
-		adev->kfd.vram_used += vram_needed;
-		adev->kfd.vram_used_aligned += ALIGN(vram_needed, VRAM_AVAILABLITY_ALIGN);
+	if (adev && xcp_id >= 0) {
+		adev->kfd.vram_used[xcp_id] += vram_needed;
+		adev->kfd.vram_used_aligned[xcp_id] +=
+				adev->apu_prefer_gtt ?
+				vram_needed :
+				ALIGN(vram_needed, VRAM_AVAILABLITY_ALIGN);
 	}
 	kfd_mem_limit.system_mem_used += system_mem_needed;
 	kfd_mem_limit.ttm_mem_used += ttm_mem_needed;
@@ -196,7 +263,7 @@ release:
 }
 
 void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
-		uint64_t size, u32 alloc_flag)
+		uint64_t size, u32 alloc_flag, int8_t xcp_id)
 {
 	spin_lock(&kfd_mem_limit.mem_limit_lock);
 
@@ -206,9 +273,19 @@ void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
 	} else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) {
 		WARN_ONCE(!adev,
 			  "adev reference can't be null when alloc mem flags vram is set");
+		if (WARN_ONCE(xcp_id < 0, "invalid XCP ID %d", xcp_id))
+			goto release;
+
 		if (adev) {
-			adev->kfd.vram_used -= size;
-			adev->kfd.vram_used_aligned -= ALIGN(size, VRAM_AVAILABLITY_ALIGN);
+			adev->kfd.vram_used[xcp_id] -= size;
+			if (adev->apu_prefer_gtt) {
+				adev->kfd.vram_used_aligned[xcp_id] -= size;
+				kfd_mem_limit.system_mem_used -= size;
+				kfd_mem_limit.ttm_mem_used -= size;
+			} else {
+				adev->kfd.vram_used_aligned[xcp_id] -=
+					ALIGN(size, VRAM_AVAILABLITY_ALIGN);
+			}
 		}
 	} else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) {
 		kfd_mem_limit.system_mem_used -= size;
@@ -218,8 +295,8 @@ void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev,
 		pr_err("%s: Invalid BO type %#x\n", __func__, alloc_flag);
 		goto release;
 	}
-	WARN_ONCE(adev && adev->kfd.vram_used < 0,
-		  "KFD VRAM memory accounting unbalanced");
+	WARN_ONCE(adev && xcp_id >= 0 && adev->kfd.vram_used[xcp_id] < 0,
+		  "KFD VRAM memory accounting unbalanced for xcp: %d", xcp_id);
 	WARN_ONCE(kfd_mem_limit.ttm_mem_used < 0,
 		  "KFD TTM memory accounting unbalanced");
 	WARN_ONCE(kfd_mem_limit.system_mem_used < 0,
@@ -235,14 +312,16 @@ void amdgpu_amdkfd_release_notify(struct amdgpu_bo *bo)
 	u32 alloc_flags = bo->kfd_bo->alloc_flags;
 	u64 size = amdgpu_bo_size(bo);
 
-	amdgpu_amdkfd_unreserve_mem_limit(adev, size, alloc_flags);
+	amdgpu_amdkfd_unreserve_mem_limit(adev, size, alloc_flags,
+					  bo->xcp_id);
 
 	kfree(bo->kfd_bo);
 }
 
 /**
- * @create_dmamap_sg_bo: Creates a amdgpu_bo object to reflect information
+ * create_dmamap_sg_bo() - Creates a amdgpu_bo object to reflect information
  * about USERPTR or DOOREBELL or MMIO BO.
+ *
  * @adev: Device for which dmamap BO is being created
  * @mem: BO of peer device that is being DMA mapped. Provides parameters
  *	 in building the dmamap BO
@@ -253,16 +332,20 @@ create_dmamap_sg_bo(struct amdgpu_device *adev,
 		 struct kgd_mem *mem, struct amdgpu_bo **bo_out)
 {
 	struct drm_gem_object *gem_obj;
-	int ret, align;
+	int ret;
+	uint64_t flags = 0;
 
 	ret = amdgpu_bo_reserve(mem->bo, false);
 	if (ret)
 		return ret;
 
-	align = 1;
-	ret = amdgpu_gem_object_create(adev, mem->bo->tbo.base.size, align,
-			AMDGPU_GEM_DOMAIN_CPU, AMDGPU_GEM_CREATE_PREEMPTIBLE,
-			ttm_bo_type_sg, mem->bo->tbo.base.resv, &gem_obj);
+	if (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR)
+		flags |= mem->bo->flags & (AMDGPU_GEM_CREATE_COHERENT |
+					AMDGPU_GEM_CREATE_UNCACHED);
+
+	ret = amdgpu_gem_object_create(adev, mem->bo->tbo.base.size, 1,
+			AMDGPU_GEM_DOMAIN_CPU, AMDGPU_GEM_CREATE_PREEMPTIBLE | flags,
+			ttm_bo_type_sg, mem->bo->tbo.base.resv, &gem_obj, 0);
 
 	amdgpu_bo_unreserve(mem->bo);
 
@@ -303,40 +386,32 @@ static int amdgpu_amdkfd_remove_eviction_fence(struct amdgpu_bo *bo,
 	return 0;
 }
 
-int amdgpu_amdkfd_remove_fence_on_pt_pd_bos(struct amdgpu_bo *bo)
+/**
+ * amdgpu_amdkfd_remove_all_eviction_fences - Remove all eviction fences
+ * @bo: the BO where to remove the evictions fences from.
+ *
+ * This functions should only be used on release when all references to the BO
+ * are already dropped. We remove the eviction fence from the private copy of
+ * the dma_resv object here since that is what is used during release to
+ * determine of the BO is idle or not.
+ */
+void amdgpu_amdkfd_remove_all_eviction_fences(struct amdgpu_bo *bo)
 {
-	struct amdgpu_bo *root = bo;
-	struct amdgpu_vm_bo_base *vm_bo;
-	struct amdgpu_vm *vm;
-	struct amdkfd_process_info *info;
-	struct amdgpu_amdkfd_fence *ef;
-	int ret;
+	struct dma_resv *resv = &bo->tbo.base._resv;
+	struct dma_fence *fence, *stub;
+	struct dma_resv_iter cursor;
 
-	/* we can always get vm_bo from root PD bo.*/
-	while (root->parent)
-		root = root->parent;
+	dma_resv_assert_held(resv);
 
-	vm_bo = root->vm_bo;
-	if (!vm_bo)
-		return 0;
-
-	vm = vm_bo->vm;
-	if (!vm)
-		return 0;
-
-	info = vm->process_info;
-	if (!info || !info->eviction_fence)
-		return 0;
-
-	ef = container_of(dma_fence_get(&info->eviction_fence->base),
-			struct amdgpu_amdkfd_fence, base);
-
-	BUG_ON(!dma_resv_trylock(bo->tbo.base.resv));
-	ret = amdgpu_amdkfd_remove_eviction_fence(bo, ef);
-	dma_resv_unlock(bo->tbo.base.resv);
+	stub = dma_fence_get_stub();
+	dma_resv_for_each_fence(&cursor, resv, DMA_RESV_USAGE_BOOKKEEP, fence) {
+		if (!to_amdgpu_amdkfd_fence(fence))
+			continue;
 
-	dma_fence_put(&ef->base);
-	return ret;
+		dma_resv_replace_fences(resv, fence->context, stub,
+					DMA_RESV_USAGE_BOOKKEEP);
+	}
+	dma_fence_put(stub);
 }
 
 static int amdgpu_amdkfd_bo_validate(struct amdgpu_bo *bo, uint32_t domain,
@@ -349,6 +424,10 @@ static int amdgpu_amdkfd_bo_validate(struct amdgpu_bo *bo, uint32_t domain,
 		 "Called with userptr BO"))
 		return -EINVAL;
 
+	/* bo has been pinned, not need validate it */
+	if (bo->tbo.pin_count)
+		return 0;
+
 	amdgpu_bo_placement_from_domain(bo, domain);
 
 	ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
@@ -361,6 +440,32 @@ validate_fail:
 	return ret;
 }
 
+int amdgpu_amdkfd_bo_validate_and_fence(struct amdgpu_bo *bo,
+					uint32_t domain,
+					struct dma_fence *fence)
+{
+	int ret = amdgpu_bo_reserve(bo, false);
+
+	if (ret)
+		return ret;
+
+	ret = amdgpu_amdkfd_bo_validate(bo, domain, true);
+	if (ret)
+		goto unreserve_out;
+
+	ret = dma_resv_reserve_fences(bo->tbo.base.resv, 1);
+	if (ret)
+		goto unreserve_out;
+
+	dma_resv_add_fence(bo->tbo.base.resv, fence,
+			   DMA_RESV_USAGE_BOOKKEEP);
+
+unreserve_out:
+	amdgpu_bo_unreserve(bo);
+
+	return ret;
+}
+
 static int amdgpu_amdkfd_validate_vm_bo(void *_unused, struct amdgpu_bo *bo)
 {
 	return amdgpu_amdkfd_bo_validate(bo, bo->allowed_domains, false);
@@ -373,13 +478,15 @@ static int amdgpu_amdkfd_validate_vm_bo(void *_unused, struct amdgpu_bo *bo)
  * again. Page directories are only updated after updating page
  * tables.
  */
-static int vm_validate_pt_pd_bos(struct amdgpu_vm *vm)
+static int vm_validate_pt_pd_bos(struct amdgpu_vm *vm,
+				 struct ww_acquire_ctx *ticket)
 {
 	struct amdgpu_bo *pd = vm->root.bo;
 	struct amdgpu_device *adev = amdgpu_ttm_adev(pd->tbo.bdev);
 	int ret;
 
-	ret = amdgpu_vm_validate_pt_bos(adev, vm, amdgpu_amdkfd_validate_vm_bo, NULL);
+	ret = amdgpu_vm_validate(adev, vm, ticket,
+				 amdgpu_amdkfd_validate_vm_bo, NULL);
 	if (ret) {
 		pr_err("failed to validate PT BOs\n");
 		return ret;
@@ -400,10 +507,11 @@ static int vm_update_pds(struct amdgpu_vm *vm, struct amdgpu_sync *sync)
 	if (ret)
 		return ret;
 
-	return amdgpu_sync_fence(sync, vm->last_update);
+	return amdgpu_sync_fence(sync, vm->last_update, GFP_KERNEL);
 }
 
-static uint64_t get_pte_flags(struct amdgpu_device *adev, struct kgd_mem *mem)
+static uint64_t get_pte_flags(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+			      struct kgd_mem *mem)
 {
 	uint32_t mapping_flags = AMDGPU_VM_PAGE_READABLE |
 				 AMDGPU_VM_MTYPE_DEFAULT;
@@ -413,7 +521,7 @@ static uint64_t get_pte_flags(struct amdgpu_device *adev, struct kgd_mem *mem)
 	if (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_EXECUTABLE)
 		mapping_flags |= AMDGPU_VM_PAGE_EXECUTABLE;
 
-	return amdgpu_gem_va_map_flags(adev, mapping_flags);
+	return mapping_flags;
 }
 
 /**
@@ -481,9 +589,6 @@ kfd_mem_dmamap_userptr(struct kgd_mem *mem,
 	if (unlikely(ret))
 		goto release_sg;
 
-	drm_prime_sg_to_dma_addr_array(ttm->sg, ttm->dma_address,
-				       ttm->num_pages);
-
 	amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT);
 	ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 	if (ret)
@@ -628,7 +733,7 @@ kfd_mem_dmaunmap_userptr(struct kgd_mem *mem,
 		return;
 
 	amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU);
-	ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+	(void)ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 
 	dma_unmap_sgtable(adev->dev, ttm->sg, direction, 0);
 	sg_free_table(ttm->sg);
@@ -639,11 +744,10 @@ kfd_mem_dmaunmap_userptr(struct kgd_mem *mem,
 static void
 kfd_mem_dmaunmap_dmabuf(struct kfd_mem_attachment *attachment)
 {
-	struct ttm_operation_ctx ctx = {.interruptible = true};
-	struct amdgpu_bo *bo = attachment->bo_va->base.bo;
-
-	amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU);
-	ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+	/* This is a no-op. We don't want to trigger eviction fences when
+	 * unmapping DMABufs. Therefore the invalidation (moving to system
+	 * domain) is done in kfd_mem_dmamap_dmabuf.
+	 */
 }
 
 /**
@@ -673,12 +777,12 @@ kfd_mem_dmaunmap_sg_bo(struct kgd_mem *mem,
 	enum dma_data_direction dir;
 
 	if (unlikely(!ttm->sg)) {
-		pr_err("SG Table of BO is UNEXPECTEDLY NULL");
+		pr_debug("SG Table of BO is NULL");
 		return;
 	}
 
 	amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU);
-	ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+	(void)ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 
 	dir = mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_WRITABLE ?
 				DMA_BIDIRECTIONAL : DMA_TO_DEVICE;
@@ -714,13 +818,17 @@ kfd_mem_dmaunmap_attachment(struct kgd_mem *mem,
 static int kfd_mem_export_dmabuf(struct kgd_mem *mem)
 {
 	if (!mem->dmabuf) {
-		struct dma_buf *ret = amdgpu_gem_prime_export(
-			&mem->bo->tbo.base,
+		struct amdgpu_device *bo_adev;
+		struct dma_buf *dmabuf;
+
+		bo_adev = amdgpu_ttm_adev(mem->bo->tbo.bdev);
+		dmabuf = drm_gem_prime_handle_to_dmabuf(&bo_adev->ddev, bo_adev->kfd.client.file,
+					       mem->gem_handle,
 			mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_WRITABLE ?
-				DRM_RDWR : 0);
-		if (IS_ERR(ret))
-			return PTR_ERR(ret);
-		mem->dmabuf = ret;
+					       DRM_RDWR : 0);
+		if (IS_ERR(dmabuf))
+			return PTR_ERR(dmabuf);
+		mem->dmabuf = dmabuf;
 	}
 
 	return 0;
@@ -768,6 +876,7 @@ static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem,
 	uint64_t va = mem->va;
 	struct kfd_mem_attachment *attachment[2] = {NULL, NULL};
 	struct amdgpu_bo *bo[2] = {NULL, NULL};
+	struct amdgpu_bo_va *bo_va;
 	bool same_hive = false;
 	int i, ret;
 
@@ -784,7 +893,7 @@ static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem,
 	 * if peer device has large BAR. In contrast, access over xGMI is
 	 * allowed for both small and large BAR configurations of peer device
 	 */
-	if ((adev != bo_adev) &&
+	if ((adev != bo_adev && !adev->apu_prefer_gtt) &&
 	    ((mem->domain == AMDGPU_GEM_DOMAIN_VRAM) ||
 	     (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL) ||
 	     (mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP))) {
@@ -805,11 +914,12 @@ static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem,
 			 va + bo_size, vm);
 
 		if ((adev == bo_adev && !(mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP)) ||
-		    (amdgpu_ttm_tt_get_usermm(mem->bo->tbo.ttm) && adev->ram_is_direct_mapped) ||
+		    (amdgpu_ttm_tt_get_usermm(mem->bo->tbo.ttm) && reuse_dmamap(adev, bo_adev)) ||
+		    (mem->domain == AMDGPU_GEM_DOMAIN_GTT && reuse_dmamap(adev, bo_adev)) ||
 		    same_hive) {
 			/* Mappings on the local GPU, or VRAM mappings in the
-			 * local hive, or userptr mapping IOMMU direct map mode
-			 * share the original BO
+			 * local hive, or userptr, or GTT mapping can reuse dma map
+			 * address space share the original BO
 			 */
 			attachment[i]->type = KFD_MEM_ATT_SHARED;
 			bo[i] = mem->bo;
@@ -854,7 +964,12 @@ static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem,
 			pr_debug("Unable to reserve BO during memory attach");
 			goto unwind;
 		}
-		attachment[i]->bo_va = amdgpu_vm_bo_add(adev, vm, bo[i]);
+		bo_va = amdgpu_vm_bo_find(vm, bo[i]);
+		if (!bo_va)
+			bo_va = amdgpu_vm_bo_add(adev, vm, bo[i]);
+		else
+			++bo_va->ref_count;
+		attachment[i]->bo_va = bo_va;
 		amdgpu_bo_unreserve(bo[i]);
 		if (unlikely(!attachment[i]->bo_va)) {
 			ret = -ENOMEM;
@@ -863,7 +978,7 @@ static int kfd_mem_attach(struct amdgpu_device *adev, struct kgd_mem *mem,
 			goto unwind;
 		}
 		attachment[i]->va = va;
-		attachment[i]->pte_flags = get_pte_flags(adev, mem);
+		attachment[i]->pte_flags = get_pte_flags(adev, vm, mem);
 		attachment[i]->adev = adev;
 		list_add(&attachment[i]->list, &mem->attachments);
 
@@ -877,8 +992,9 @@ unwind:
 		if (!attachment[i])
 			continue;
 		if (attachment[i]->bo_va) {
-			amdgpu_bo_reserve(bo[i], true);
-			amdgpu_vm_bo_del(adev, attachment[i]->bo_va);
+			(void)amdgpu_bo_reserve(bo[i], true);
+			if (--attachment[i]->bo_va->ref_count == 0)
+				amdgpu_vm_bo_del(adev, attachment[i]->bo_va);
 			amdgpu_bo_unreserve(bo[i]);
 			list_del(&attachment[i]->list);
 		}
@@ -895,7 +1011,8 @@ static void kfd_mem_detach(struct kfd_mem_attachment *attachment)
 
 	pr_debug("\t remove VA 0x%llx in entry %p\n",
 			attachment->va, attachment);
-	amdgpu_vm_bo_del(attachment->adev, attachment->bo_va);
+	if (--attachment->bo_va->ref_count == 0)
+		amdgpu_vm_bo_del(attachment->adev, attachment->bo_va);
 	drm_gem_object_put(&bo->tbo.base);
 	list_del(&attachment->list);
 	kfree(attachment);
@@ -905,28 +1022,20 @@ static void add_kgd_mem_to_kfd_bo_list(struct kgd_mem *mem,
 				struct amdkfd_process_info *process_info,
 				bool userptr)
 {
-	struct ttm_validate_buffer *entry = &mem->validate_list;
-	struct amdgpu_bo *bo = mem->bo;
-
-	INIT_LIST_HEAD(&entry->head);
-	entry->num_shared = 1;
-	entry->bo = &bo->tbo;
 	mutex_lock(&process_info->lock);
 	if (userptr)
-		list_add_tail(&entry->head, &process_info->userptr_valid_list);
+		list_add_tail(&mem->validate_list,
+			      &process_info->userptr_valid_list);
 	else
-		list_add_tail(&entry->head, &process_info->kfd_bo_list);
+		list_add_tail(&mem->validate_list, &process_info->kfd_bo_list);
 	mutex_unlock(&process_info->lock);
 }
 
 static void remove_kgd_mem_from_kfd_bo_list(struct kgd_mem *mem,
 		struct amdkfd_process_info *process_info)
 {
-	struct ttm_validate_buffer *bo_list_entry;
-
-	bo_list_entry = &mem->validate_list;
 	mutex_lock(&process_info->lock);
-	list_del(&bo_list_entry->head);
+	list_del(&mem->validate_list);
 	mutex_unlock(&process_info->lock);
 }
 
@@ -948,7 +1057,7 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr,
 	struct amdkfd_process_info *process_info = mem->process_info;
 	struct amdgpu_bo *bo = mem->bo;
 	struct ttm_operation_ctx ctx = { true, false };
-	struct hmm_range *range;
+	struct amdgpu_hmm_range *range;
 	int ret = 0;
 
 	mutex_lock(&process_info->lock);
@@ -980,9 +1089,19 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr,
 		return 0;
 	}
 
-	ret = amdgpu_ttm_tt_get_user_pages(bo, bo->tbo.ttm->pages, &range);
+	range = amdgpu_hmm_range_alloc(NULL);
+	if (unlikely(!range)) {
+		ret = -ENOMEM;
+		goto unregister_out;
+	}
+
+	ret = amdgpu_ttm_tt_get_user_pages(bo, range);
 	if (ret) {
-		pr_err("%s: Failed to get user pages: %d\n", __func__, ret);
+		amdgpu_hmm_range_free(range);
+		if (ret == -EAGAIN)
+			pr_debug("Failed to get user pages, try again\n");
+		else
+			pr_err("%s: Failed to get user pages: %d\n", __func__, ret);
 		goto unregister_out;
 	}
 
@@ -991,6 +1110,9 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr,
 		pr_err("%s: Failed to reserve BO\n", __func__);
 		goto release_out;
 	}
+
+	amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, range);
+
 	amdgpu_bo_placement_from_domain(bo, mem->domain);
 	ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 	if (ret)
@@ -998,7 +1120,7 @@ static int init_user_pages(struct kgd_mem *mem, uint64_t user_addr,
 	amdgpu_bo_unreserve(bo);
 
 release_out:
-	amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm, range);
+	amdgpu_hmm_range_free(range);
 unregister_out:
 	if (ret)
 		amdgpu_hmm_unregister(bo);
@@ -1013,13 +1135,12 @@ out:
  * object can track VM updates.
  */
 struct bo_vm_reservation_context {
-	struct amdgpu_bo_list_entry kfd_bo; /* BO list entry for the KFD BO */
-	unsigned int n_vms;		    /* Number of VMs reserved	    */
-	struct amdgpu_bo_list_entry *vm_pd; /* Array of VM BO list entries  */
-	struct ww_acquire_ctx ticket;	    /* Reservation ticket	    */
-	struct list_head list, duplicates;  /* BO lists			    */
-	struct amdgpu_sync *sync;	    /* Pointer to sync object	    */
-	bool reserved;			    /* Whether BOs are reserved	    */
+	/* DRM execution context for the reservation */
+	struct drm_exec exec;
+	/* Number of VMs reserved */
+	unsigned int n_vms;
+	/* Pointer to sync object */
+	struct amdgpu_sync *sync;
 };
 
 enum bo_vm_match {
@@ -1043,35 +1164,26 @@ static int reserve_bo_and_vm(struct kgd_mem *mem,
 
 	WARN_ON(!vm);
 
-	ctx->reserved = false;
 	ctx->n_vms = 1;
 	ctx->sync = &mem->sync;
-
-	INIT_LIST_HEAD(&ctx->list);
-	INIT_LIST_HEAD(&ctx->duplicates);
-
-	ctx->vm_pd = kcalloc(ctx->n_vms, sizeof(*ctx->vm_pd), GFP_KERNEL);
-	if (!ctx->vm_pd)
-		return -ENOMEM;
-
-	ctx->kfd_bo.priority = 0;
-	ctx->kfd_bo.tv.bo = &bo->tbo;
-	ctx->kfd_bo.tv.num_shared = 1;
-	list_add(&ctx->kfd_bo.tv.head, &ctx->list);
-
-	amdgpu_vm_get_pd_bo(vm, &ctx->list, &ctx->vm_pd[0]);
-
-	ret = ttm_eu_reserve_buffers(&ctx->ticket, &ctx->list,
-				     false, &ctx->duplicates);
-	if (ret) {
-		pr_err("Failed to reserve buffers in ttm.\n");
-		kfree(ctx->vm_pd);
-		ctx->vm_pd = NULL;
-		return ret;
+	drm_exec_init(&ctx->exec, DRM_EXEC_INTERRUPTIBLE_WAIT, 0);
+	drm_exec_until_all_locked(&ctx->exec) {
+		ret = amdgpu_vm_lock_pd(vm, &ctx->exec, 2);
+		drm_exec_retry_on_contention(&ctx->exec);
+		if (unlikely(ret))
+			goto error;
+
+		ret = drm_exec_prepare_obj(&ctx->exec, &bo->tbo.base, 1);
+		drm_exec_retry_on_contention(&ctx->exec);
+		if (unlikely(ret))
+			goto error;
 	}
-
-	ctx->reserved = true;
 	return 0;
+
+error:
+	pr_err("Failed to reserve buffers in ttm.\n");
+	drm_exec_fini(&ctx->exec);
+	return ret;
 }
 
 /**
@@ -1088,63 +1200,40 @@ static int reserve_bo_and_cond_vms(struct kgd_mem *mem,
 				struct amdgpu_vm *vm, enum bo_vm_match map_type,
 				struct bo_vm_reservation_context *ctx)
 {
-	struct amdgpu_bo *bo = mem->bo;
 	struct kfd_mem_attachment *entry;
-	unsigned int i;
+	struct amdgpu_bo *bo = mem->bo;
 	int ret;
 
-	ctx->reserved = false;
-	ctx->n_vms = 0;
-	ctx->vm_pd = NULL;
 	ctx->sync = &mem->sync;
+	drm_exec_init(&ctx->exec, DRM_EXEC_INTERRUPTIBLE_WAIT |
+		      DRM_EXEC_IGNORE_DUPLICATES, 0);
+	drm_exec_until_all_locked(&ctx->exec) {
+		ctx->n_vms = 0;
+		list_for_each_entry(entry, &mem->attachments, list) {
+			if ((vm && vm != entry->bo_va->base.vm) ||
+				(entry->is_mapped != map_type
+				&& map_type != BO_VM_ALL))
+				continue;
 
-	INIT_LIST_HEAD(&ctx->list);
-	INIT_LIST_HEAD(&ctx->duplicates);
-
-	list_for_each_entry(entry, &mem->attachments, list) {
-		if ((vm && vm != entry->bo_va->base.vm) ||
-			(entry->is_mapped != map_type
-			&& map_type != BO_VM_ALL))
-			continue;
-
-		ctx->n_vms++;
-	}
-
-	if (ctx->n_vms != 0) {
-		ctx->vm_pd = kcalloc(ctx->n_vms, sizeof(*ctx->vm_pd),
-				     GFP_KERNEL);
-		if (!ctx->vm_pd)
-			return -ENOMEM;
-	}
-
-	ctx->kfd_bo.priority = 0;
-	ctx->kfd_bo.tv.bo = &bo->tbo;
-	ctx->kfd_bo.tv.num_shared = 1;
-	list_add(&ctx->kfd_bo.tv.head, &ctx->list);
-
-	i = 0;
-	list_for_each_entry(entry, &mem->attachments, list) {
-		if ((vm && vm != entry->bo_va->base.vm) ||
-			(entry->is_mapped != map_type
-			&& map_type != BO_VM_ALL))
-			continue;
-
-		amdgpu_vm_get_pd_bo(entry->bo_va->base.vm, &ctx->list,
-				&ctx->vm_pd[i]);
-		i++;
-	}
+			ret = amdgpu_vm_lock_pd(entry->bo_va->base.vm,
+						&ctx->exec, 2);
+			drm_exec_retry_on_contention(&ctx->exec);
+			if (unlikely(ret))
+				goto error;
+			++ctx->n_vms;
+		}
 
-	ret = ttm_eu_reserve_buffers(&ctx->ticket, &ctx->list,
-				     false, &ctx->duplicates);
-	if (ret) {
-		pr_err("Failed to reserve buffers in ttm.\n");
-		kfree(ctx->vm_pd);
-		ctx->vm_pd = NULL;
-		return ret;
+		ret = drm_exec_prepare_obj(&ctx->exec, &bo->tbo.base, 1);
+		drm_exec_retry_on_contention(&ctx->exec);
+		if (unlikely(ret))
+			goto error;
 	}
-
-	ctx->reserved = true;
 	return 0;
+
+error:
+	pr_err("Failed to reserve buffers in ttm.\n");
+	drm_exec_fini(&ctx->exec);
+	return ret;
 }
 
 /**
@@ -1165,19 +1254,12 @@ static int unreserve_bo_and_vms(struct bo_vm_reservation_context *ctx,
 	if (wait)
 		ret = amdgpu_sync_wait(ctx->sync, intr);
 
-	if (ctx->reserved)
-		ttm_eu_backoff_reservation(&ctx->ticket, &ctx->list);
-	kfree(ctx->vm_pd);
-
+	drm_exec_fini(&ctx->exec);
 	ctx->sync = NULL;
-
-	ctx->reserved = false;
-	ctx->vm_pd = NULL;
-
 	return ret;
 }
 
-static void unmap_bo_from_gpuvm(struct kgd_mem *mem,
+static int unmap_bo_from_gpuvm(struct kgd_mem *mem,
 				struct kfd_mem_attachment *entry,
 				struct amdgpu_sync *sync)
 {
@@ -1185,13 +1267,22 @@ static void unmap_bo_from_gpuvm(struct kgd_mem *mem,
 	struct amdgpu_device *adev = entry->adev;
 	struct amdgpu_vm *vm = bo_va->base.vm;
 
-	amdgpu_vm_bo_unmap(adev, bo_va, entry->va);
+	if (bo_va->queue_refcount) {
+		pr_debug("bo_va->queue_refcount %d\n", bo_va->queue_refcount);
+		return -EBUSY;
+	}
+
+	(void)amdgpu_vm_bo_unmap(adev, bo_va, entry->va);
 
-	amdgpu_vm_clear_freed(adev, vm, &bo_va->last_pt_update);
+	/* VM entity stopped if process killed, don't clear freed pt bo */
+	if (!amdgpu_vm_ready(vm))
+		return 0;
 
-	amdgpu_sync_fence(sync, bo_va->last_pt_update);
+	(void)amdgpu_vm_clear_freed(adev, vm, &bo_va->last_pt_update);
 
-	kfd_mem_dmaunmap_attachment(mem, entry);
+	(void)amdgpu_sync_fence(sync, bo_va->last_pt_update, GFP_KERNEL);
+
+	return 0;
 }
 
 static int update_gpuvm_pte(struct kgd_mem *mem,
@@ -1213,7 +1304,7 @@ static int update_gpuvm_pte(struct kgd_mem *mem,
 		return ret;
 	}
 
-	return amdgpu_sync_fence(sync, bo_va->last_pt_update);
+	return amdgpu_sync_fence(sync, bo_va->last_pt_update, GFP_KERNEL);
 }
 
 static int map_bo_to_gpuvm(struct kgd_mem *mem,
@@ -1246,17 +1337,19 @@ static int map_bo_to_gpuvm(struct kgd_mem *mem,
 
 update_gpuvm_pte_failed:
 	unmap_bo_from_gpuvm(mem, entry, sync);
+	kfd_mem_dmaunmap_attachment(mem, entry);
 	return ret;
 }
 
-static int process_validate_vms(struct amdkfd_process_info *process_info)
+static int process_validate_vms(struct amdkfd_process_info *process_info,
+				struct ww_acquire_ctx *ticket)
 {
 	struct amdgpu_vm *peer_vm;
 	int ret;
 
 	list_for_each_entry(peer_vm, &process_info->vm_list_head,
 			    vm_list_node) {
-		ret = vm_validate_pt_pd_bos(peer_vm);
+		ret = vm_validate_pt_pd_bos(peer_vm, ticket);
 		if (ret)
 			return ret;
 	}
@@ -1333,7 +1426,6 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info,
 				  amdgpu_amdkfd_restore_userptr_worker);
 
 		*process_info = info;
-		*ef = dma_fence_get(&info->eviction_fence->base);
 	}
 
 	vm->process_info = *process_info;
@@ -1342,7 +1434,7 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info,
 	ret = amdgpu_bo_reserve(vm->root.bo, true);
 	if (ret)
 		goto reserve_pd_fail;
-	ret = vm_validate_pt_pd_bos(vm);
+	ret = vm_validate_pt_pd_bos(vm, NULL);
 	if (ret) {
 		pr_err("validate_pt_pd_bos() failed\n");
 		goto validate_pd_fail;
@@ -1364,6 +1456,8 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info,
 	list_add_tail(&vm->vm_list_node,
 			&(vm->process_info->vm_list_head));
 	vm->process_info->n_vms++;
+	if (ef)
+		*ef = dma_fence_get(&vm->process_info->eviction_fence->base);
 	mutex_unlock(&vm->process_info->lock);
 
 	return 0;
@@ -1375,10 +1469,7 @@ validate_pd_fail:
 reserve_pd_fail:
 	vm->process_info = NULL;
 	if (info) {
-		/* Two fence references: one in info and one in *ef */
 		dma_fence_put(&info->eviction_fence->base);
-		dma_fence_put(*ef);
-		*ef = NULL;
 		*process_info = NULL;
 		put_pid(info->pid);
 create_evict_fence_fail:
@@ -1408,13 +1499,30 @@ static int amdgpu_amdkfd_gpuvm_pin_bo(struct amdgpu_bo *bo, u32 domain)
 	if (unlikely(ret))
 		return ret;
 
-	ret = amdgpu_bo_pin_restricted(bo, domain, 0, 0);
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS) {
+		/*
+		 * If bo is not contiguous on VRAM, move to system memory first to ensure
+		 * we can get contiguous VRAM space after evicting other BOs.
+		 */
+		if (!(bo->tbo.resource->placement & TTM_PL_FLAG_CONTIGUOUS)) {
+			struct ttm_operation_ctx ctx = { true, false };
+
+			amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT);
+			ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+			if (unlikely(ret)) {
+				pr_debug("validate bo 0x%p to GTT failed %d\n", &bo->tbo, ret);
+				goto out;
+			}
+		}
+	}
+
+	ret = amdgpu_bo_pin(bo, domain);
 	if (ret)
 		pr_err("Error in Pinning BO to domain: %d\n", domain);
 
 	amdgpu_bo_sync_wait(bo, AMDGPU_FENCE_OWNER_KFD, false);
+out:
 	amdgpu_bo_unreserve(bo);
-
 	return ret;
 }
 
@@ -1438,27 +1546,6 @@ static void amdgpu_amdkfd_gpuvm_unpin_bo(struct amdgpu_bo *bo)
 	amdgpu_bo_unreserve(bo);
 }
 
-int amdgpu_amdkfd_gpuvm_set_vm_pasid(struct amdgpu_device *adev,
-				     struct amdgpu_vm *avm, u32 pasid)
-
-{
-	int ret;
-
-	/* Free the original amdgpu allocated pasid,
-	 * will be replaced with kfd allocated pasid.
-	 */
-	if (avm->pasid) {
-		amdgpu_pasid_free(avm->pasid);
-		amdgpu_vm_set_pasid(adev, avm, 0);
-	}
-
-	ret = amdgpu_vm_set_pasid(adev, avm, pasid);
-	if (ret)
-		return ret;
-
-	return 0;
-}
-
 int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct amdgpu_device *adev,
 					   struct amdgpu_vm *avm,
 					   void **process_info,
@@ -1516,27 +1603,6 @@ void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev,
 	}
 }
 
-void amdgpu_amdkfd_gpuvm_release_process_vm(struct amdgpu_device *adev,
-					    void *drm_priv)
-{
-	struct amdgpu_vm *avm;
-
-	if (WARN_ON(!adev || !drm_priv))
-		return;
-
-	avm = drm_priv_to_vm(drm_priv);
-
-	pr_debug("Releasing process vm %p\n", avm);
-
-	/* The original pasid of amdgpu vm has already been
-	 * released during making a amdgpu vm to a compute vm
-	 * The current pasid is managed by kfd and will be
-	 * released on kfd process destroy. Set amdgpu pasid
-	 * to 0 to avoid duplicate release.
-	 */
-	amdgpu_vm_release_compute(adev, avm);
-}
-
 uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *drm_priv)
 {
 	struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv);
@@ -1572,30 +1638,57 @@ int amdgpu_amdkfd_criu_resume(void *p)
 		goto out_unlock;
 	}
 	WRITE_ONCE(pinfo->block_mmu_notifications, false);
-	schedule_delayed_work(&pinfo->restore_userptr_work, 0);
+	queue_delayed_work(system_freezable_wq,
+			   &pinfo->restore_userptr_work, 0);
 
 out_unlock:
 	mutex_unlock(&pinfo->lock);
 	return ret;
 }
 
-size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev)
+size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev,
+					  uint8_t xcp_id)
 {
 	uint64_t reserved_for_pt =
 		ESTIMATE_PT_SIZE(amdgpu_amdkfd_total_mem_size);
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	uint64_t reserved_for_ras = (con ? con->reserved_pages_in_bytes : 0);
 	ssize_t available;
+	uint64_t vram_available, system_mem_available, ttm_mem_available;
 
 	spin_lock(&kfd_mem_limit.mem_limit_lock);
-	available = adev->gmc.real_vram_size
-		- adev->kfd.vram_used_aligned
-		- atomic64_read(&adev->vram_pin_size)
-		- reserved_for_pt;
+	if (adev->apu_prefer_gtt && !adev->gmc.is_app_apu)
+		vram_available = KFD_XCP_MEMORY_SIZE(adev, xcp_id)
+			- adev->kfd.vram_used_aligned[xcp_id];
+	else
+		vram_available = KFD_XCP_MEMORY_SIZE(adev, xcp_id)
+			- adev->kfd.vram_used_aligned[xcp_id]
+			- atomic64_read(&adev->vram_pin_size)
+			- reserved_for_pt
+			- reserved_for_ras;
+
+	if (adev->apu_prefer_gtt) {
+		system_mem_available = no_system_mem_limit ?
+					kfd_mem_limit.max_system_mem_limit :
+					kfd_mem_limit.max_system_mem_limit -
+					kfd_mem_limit.system_mem_used;
+
+		ttm_mem_available = kfd_mem_limit.max_ttm_mem_limit -
+				kfd_mem_limit.ttm_mem_used;
+
+		available = min3(system_mem_available, ttm_mem_available,
+				 vram_available);
+		available = ALIGN_DOWN(available, PAGE_SIZE);
+	} else {
+		available = ALIGN_DOWN(vram_available, VRAM_AVAILABLITY_ALIGN);
+	}
+
 	spin_unlock(&kfd_mem_limit.mem_limit_lock);
 
 	if (available < 0)
 		available = 0;
 
-	return ALIGN_DOWN(available, VRAM_AVAILABLITY_ALIGN);
+	return available;
 }
 
 int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
@@ -1604,6 +1697,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 		uint64_t *offset, uint32_t flags, bool criu_resume)
 {
 	struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv);
+	struct amdgpu_fpriv *fpriv = container_of(avm, struct amdgpu_fpriv, vm);
 	enum ttm_bo_type bo_type = ttm_bo_type_device;
 	struct sg_table *sg = NULL;
 	uint64_t user_addr = 0;
@@ -1611,6 +1705,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 	struct drm_gem_object *gobj = NULL;
 	u32 domain, alloc_domain;
 	uint64_t aligned_size;
+	int8_t xcp_id = -1;
 	u64 alloc_flags;
 	int ret;
 
@@ -1619,9 +1714,22 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 	 */
 	if (flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) {
 		domain = alloc_domain = AMDGPU_GEM_DOMAIN_VRAM;
-		alloc_flags = AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE;
-		alloc_flags |= (flags & KFD_IOC_ALLOC_MEM_FLAGS_PUBLIC) ?
+
+		if (adev->apu_prefer_gtt) {
+			domain = AMDGPU_GEM_DOMAIN_GTT;
+			alloc_domain = AMDGPU_GEM_DOMAIN_GTT;
+			alloc_flags = 0;
+		} else {
+			alloc_flags = AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE;
+			alloc_flags |= (flags & KFD_IOC_ALLOC_MEM_FLAGS_PUBLIC) ?
 			AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED : 0;
+
+			/* For contiguous VRAM allocation */
+			if (flags & KFD_IOC_ALLOC_MEM_FLAGS_CONTIGUOUS)
+				alloc_flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
+		}
+		xcp_id = fpriv->xcp_id == AMDGPU_XCP_NO_PARTITION ?
+					0 : fpriv->xcp_id;
 	} else if (flags & KFD_IOC_ALLOC_MEM_FLAGS_GTT) {
 		domain = alloc_domain = AMDGPU_GEM_DOMAIN_GTT;
 		alloc_flags = 0;
@@ -1649,6 +1757,8 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 
 	if (flags & KFD_IOC_ALLOC_MEM_FLAGS_COHERENT)
 		alloc_flags |= AMDGPU_GEM_CREATE_COHERENT;
+	if (flags & KFD_IOC_ALLOC_MEM_FLAGS_EXT_COHERENT)
+		alloc_flags |= AMDGPU_GEM_CREATE_EXT_COHERENT;
 	if (flags & KFD_IOC_ALLOC_MEM_FLAGS_UNCACHED)
 		alloc_flags |= AMDGPU_GEM_CREATE_UNCACHED;
 
@@ -1673,17 +1783,19 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 
 	amdgpu_sync_create(&(*mem)->sync);
 
-	ret = amdgpu_amdkfd_reserve_mem_limit(adev, aligned_size, flags);
+	ret = amdgpu_amdkfd_reserve_mem_limit(adev, aligned_size, flags,
+					      xcp_id);
 	if (ret) {
 		pr_debug("Insufficient memory\n");
 		goto err_reserve_limit;
 	}
 
-	pr_debug("\tcreate BO VA 0x%llx size 0x%llx domain %s\n",
-			va, (*mem)->aql_queue ? size << 1 : size, domain_string(alloc_domain));
+	pr_debug("\tcreate BO VA 0x%llx size 0x%llx domain %s xcp_id %d\n",
+		 va, (*mem)->aql_queue ? size << 1 : size,
+		 domain_string(alloc_domain), xcp_id);
 
 	ret = amdgpu_gem_object_create(adev, aligned_size, 1, alloc_domain, alloc_flags,
-				       bo_type, NULL, &gobj);
+				       bo_type, NULL, &gobj, xcp_id + 1);
 	if (ret) {
 		pr_debug("Failed to create BO on domain %s. ret %d\n",
 			 domain_string(alloc_domain), ret);
@@ -1694,6 +1806,9 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 		pr_debug("Failed to allow vma node access. ret %d\n", ret);
 		goto err_node_allow;
 	}
+	ret = drm_gem_handle_create(adev->kfd.client.file, gobj, &(*mem)->gem_handle);
+	if (ret)
+		goto err_gem_handle_create;
 	bo = gem_to_amdgpu_bo(gobj);
 	if (bo_type == ttm_bo_type_sg) {
 		bo->tbo.sg = sg;
@@ -1708,6 +1823,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 	(*mem)->domain = domain;
 	(*mem)->mapped_to_gpu_memory = 0;
 	(*mem)->process_info = avm->process_info;
+
 	add_kgd_mem_to_kfd_bo_list(*mem, avm->process_info, user_addr);
 
 	if (user_addr) {
@@ -1724,6 +1840,15 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 		}
 		bo->allowed_domains = AMDGPU_GEM_DOMAIN_GTT;
 		bo->preferred_domains = AMDGPU_GEM_DOMAIN_GTT;
+	} else {
+		mutex_lock(&avm->process_info->lock);
+		if (avm->process_info->eviction_fence &&
+		    !dma_fence_is_signaled(&avm->process_info->eviction_fence->base))
+			ret = amdgpu_amdkfd_bo_validate_and_fence(bo, domain,
+				&avm->process_info->eviction_fence->base);
+		mutex_unlock(&avm->process_info->lock);
+		if (ret)
+			goto err_validate_bo;
 	}
 
 	if (offset)
@@ -1733,14 +1858,18 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
 
 allocate_init_user_pages_failed:
 err_pin_bo:
+err_validate_bo:
 	remove_kgd_mem_from_kfd_bo_list(*mem, avm->process_info);
+	drm_gem_handle_delete(adev->kfd.client.file, (*mem)->gem_handle);
+err_gem_handle_create:
 	drm_vma_node_revoke(&gobj->vma_node, drm_priv);
 err_node_allow:
 	/* Don't unreserve system mem limit twice */
 	goto err_reserve_limit;
 err_bo_create:
-	amdgpu_amdkfd_unreserve_mem_limit(adev, aligned_size, flags);
+	amdgpu_amdkfd_unreserve_mem_limit(adev, aligned_size, flags, xcp_id);
 err_reserve_limit:
+	amdgpu_sync_free(&(*mem)->sync);
 	mutex_destroy(&(*mem)->lock);
 	if (gobj)
 		drm_gem_object_put(gobj);
@@ -1763,7 +1892,6 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
 	bool use_release_notifier = (mem->bo->kfd_bo == mem);
 	struct kfd_mem_attachment *entry, *tmp;
 	struct bo_vm_reservation_context ctx;
-	struct ttm_validate_buffer *bo_list_entry;
 	unsigned int mapped_to_gpu_memory;
 	int ret;
 	bool is_imported = false;
@@ -1791,16 +1919,15 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
 	}
 
 	/* Make sure restore workers don't access the BO any more */
-	bo_list_entry = &mem->validate_list;
 	mutex_lock(&process_info->lock);
-	list_del(&bo_list_entry->head);
+	list_del(&mem->validate_list);
 	mutex_unlock(&process_info->lock);
 
 	/* Cleanup user pages and MMU notifiers */
 	if (amdgpu_ttm_tt_get_usermm(mem->bo->tbo.ttm)) {
 		amdgpu_hmm_unregister(mem->bo);
 		mutex_lock(&process_info->notifier_lock);
-		amdgpu_ttm_tt_discard_user_pages(mem->bo->tbo.ttm, mem->range);
+		amdgpu_hmm_range_free(mem->range);
 		mutex_unlock(&process_info->notifier_lock);
 	}
 
@@ -1808,18 +1935,16 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
 	if (unlikely(ret))
 		return ret;
 
-	/* The eviction fence should be removed by the last unmap.
-	 * TODO: Log an error condition if the bo still has the eviction fence
-	 * attached
-	 */
 	amdgpu_amdkfd_remove_eviction_fence(mem->bo,
 					process_info->eviction_fence);
 	pr_debug("Release VA 0x%llx - 0x%llx\n", mem->va,
 		mem->va + bo_size * (1 + mem->aql_queue));
 
 	/* Remove from VM internal data structures */
-	list_for_each_entry_safe(entry, tmp, &mem->attachments, list)
+	list_for_each_entry_safe(entry, tmp, &mem->attachments, list) {
+		kfd_mem_dmaunmap_attachment(mem, entry);
 		kfd_mem_detach(entry);
+	}
 
 	ret = unreserve_bo_and_vms(&ctx, false, false);
 
@@ -1835,11 +1960,12 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
 	}
 
 	/* Update the size of the BO being freed if it was allocated from
-	 * VRAM and is not imported.
+	 * VRAM and is not imported. For APP APU VRAM allocations are done
+	 * in GTT domain
 	 */
 	if (size) {
-		if ((mem->bo->preferred_domains == AMDGPU_GEM_DOMAIN_VRAM) &&
-		    (!is_imported))
+		if (!is_imported &&
+		   mem->alloc_flags & KFD_IOC_ALLOC_MEM_FLAGS_VRAM)
 			*size = bo_size;
 		else
 			*size = 0;
@@ -1847,8 +1973,11 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
 
 	/* Free the BO*/
 	drm_vma_node_revoke(&mem->bo->tbo.base.vma_node, drm_priv);
-	if (mem->dmabuf)
+	drm_gem_handle_delete(adev->kfd.client.file, mem->gem_handle);
+	if (mem->dmabuf) {
 		dma_buf_put(mem->dmabuf);
+		mem->dmabuf = NULL;
+	}
 	mutex_destroy(&mem->lock);
 
 	/* If this releases the last reference, it will end up calling
@@ -1931,23 +2060,10 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(
 	    bo->tbo.resource->mem_type == TTM_PL_SYSTEM)
 		is_invalid_userptr = true;
 
-	ret = vm_validate_pt_pd_bos(avm);
+	ret = vm_validate_pt_pd_bos(avm, NULL);
 	if (unlikely(ret))
 		goto out_unreserve;
 
-	if (mem->mapped_to_gpu_memory == 0 &&
-	    !amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) {
-		/* Validate BO only once. The eviction fence gets added to BO
-		 * the first time it is mapped. Validate will wait for all
-		 * background evictions to complete.
-		 */
-		ret = amdgpu_amdkfd_bo_validate(bo, domain, true);
-		if (ret) {
-			pr_debug("Validate failed\n");
-			goto out_unreserve;
-		}
-	}
-
 	list_for_each_entry(entry, &mem->attachments, list) {
 		if (entry->bo_va->base.vm != avm || entry->is_mapped)
 			continue;
@@ -1974,10 +2090,6 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu(
 			 mem->mapped_to_gpu_memory);
 	}
 
-	if (!amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) && !bo->tbo.pin_count)
-		dma_resv_add_fence(bo->tbo.base.resv,
-				   &avm->process_info->eviction_fence->base,
-				   DMA_RESV_USAGE_BOOKKEEP);
 	ret = unreserve_bo_and_vms(&ctx, false, false);
 
 	goto out;
@@ -1990,11 +2102,41 @@ out:
 	return ret;
 }
 
+int amdgpu_amdkfd_gpuvm_dmaunmap_mem(struct kgd_mem *mem, void *drm_priv)
+{
+	struct kfd_mem_attachment *entry;
+	struct amdgpu_vm *vm;
+	int ret;
+
+	vm = drm_priv_to_vm(drm_priv);
+
+	mutex_lock(&mem->lock);
+
+	ret = amdgpu_bo_reserve(mem->bo, true);
+	if (ret)
+		goto out;
+
+	list_for_each_entry(entry, &mem->attachments, list) {
+		if (entry->bo_va->base.vm != vm)
+			continue;
+		if (entry->bo_va->base.bo->tbo.ttm &&
+		    !entry->bo_va->base.bo->tbo.ttm->sg)
+			continue;
+
+		kfd_mem_dmaunmap_attachment(mem, entry);
+	}
+
+	amdgpu_bo_unreserve(mem->bo);
+out:
+	mutex_unlock(&mem->lock);
+
+	return ret;
+}
+
 int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
 		struct amdgpu_device *adev, struct kgd_mem *mem, void *drm_priv)
 {
 	struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv);
-	struct amdkfd_process_info *process_info = avm->process_info;
 	unsigned long bo_size = mem->bo->tbo.base.size;
 	struct kfd_mem_attachment *entry;
 	struct bo_vm_reservation_context ctx;
@@ -2011,7 +2153,7 @@ int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
 		goto unreserve_out;
 	}
 
-	ret = vm_validate_pt_pd_bos(avm);
+	ret = vm_validate_pt_pd_bos(avm, NULL);
 	if (unlikely(ret))
 		goto unreserve_out;
 
@@ -2027,7 +2169,10 @@ int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
 		pr_debug("\t unmap VA 0x%llx - 0x%llx from entry %p\n",
 			 entry->va, entry->va + bo_size, entry);
 
-		unmap_bo_from_gpuvm(mem, entry, ctx.sync);
+		ret = unmap_bo_from_gpuvm(mem, entry, ctx.sync);
+		if (ret)
+			goto unreserve_out;
+
 		entry->is_mapped = false;
 
 		mem->mapped_to_gpu_memory--;
@@ -2035,15 +2180,6 @@ int amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu(
 			 mem->mapped_to_gpu_memory);
 	}
 
-	/* If BO is unmapped from all VMs, unfence it. It can be evicted if
-	 * required.
-	 */
-	if (mem->mapped_to_gpu_memory == 0 &&
-	    !amdgpu_ttm_tt_get_usermm(mem->bo->tbo.ttm) &&
-	    !mem->bo->tbo.pin_count)
-		amdgpu_amdkfd_remove_eviction_fence(mem->bo,
-						process_info->eviction_fence);
-
 unreserve_out:
 	unreserve_bo_and_vms(&ctx, false, false);
 out:
@@ -2070,13 +2206,13 @@ int amdgpu_amdkfd_gpuvm_sync_memory(
 
 /**
  * amdgpu_amdkfd_map_gtt_bo_to_gart - Map BO to GART and increment reference count
- * @adev: Device to which allocated BO belongs
  * @bo: Buffer object to be mapped
+ * @bo_gart: Return bo reference
  *
  * Before return, bo reference count is incremented. To release the reference and unpin/
  * unmap the BO, call amdgpu_amdkfd_free_gtt_mem.
  */
-int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_device *adev, struct amdgpu_bo *bo)
+int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_bo *bo, struct amdgpu_bo **bo_gart)
 {
 	int ret;
 
@@ -2103,7 +2239,7 @@ int amdgpu_amdkfd_map_gtt_bo_to_gart(struct amdgpu_device *adev, struct amdgpu_b
 
 	amdgpu_bo_unreserve(bo);
 
-	bo = amdgpu_bo_ref(bo);
+	*bo_gart = amdgpu_bo_ref(bo);
 
 	return 0;
 
@@ -2193,7 +2329,7 @@ void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct kgd_mem *mem)
 {
 	struct amdgpu_bo *bo = mem->bo;
 
-	amdgpu_bo_reserve(bo, true);
+	(void)amdgpu_bo_reserve(bo, true);
 	amdgpu_bo_kunmap(bo);
 	amdgpu_bo_unpin(bo);
 	amdgpu_bo_unreserve(bo);
@@ -2202,42 +2338,33 @@ void amdgpu_amdkfd_gpuvm_unmap_gtt_bo_from_kernel(struct kgd_mem *mem)
 int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct amdgpu_device *adev,
 					  struct kfd_vm_fault_info *mem)
 {
-	if (atomic_read(&adev->gmc.vm_fault_info_updated) == 1) {
+	if (atomic_read_acquire(&adev->gmc.vm_fault_info_updated) == 1) {
 		*mem = *adev->gmc.vm_fault_info;
-		mb(); /* make sure read happened */
-		atomic_set(&adev->gmc.vm_fault_info_updated, 0);
+		atomic_set_release(&adev->gmc.vm_fault_info_updated, 0);
 	}
 	return 0;
 }
 
-int amdgpu_amdkfd_gpuvm_import_dmabuf(struct amdgpu_device *adev,
-				      struct dma_buf *dma_buf,
-				      uint64_t va, void *drm_priv,
-				      struct kgd_mem **mem, uint64_t *size,
-				      uint64_t *mmap_offset)
+static int import_obj_create(struct amdgpu_device *adev,
+			     struct dma_buf *dma_buf,
+			     struct drm_gem_object *obj,
+			     uint64_t va, void *drm_priv,
+			     struct kgd_mem **mem, uint64_t *size,
+			     uint64_t *mmap_offset)
 {
 	struct amdgpu_vm *avm = drm_priv_to_vm(drm_priv);
-	struct drm_gem_object *obj;
 	struct amdgpu_bo *bo;
 	int ret;
 
-	obj = amdgpu_gem_prime_import(adev_to_drm(adev), dma_buf);
-	if (IS_ERR(obj))
-		return PTR_ERR(obj);
-
 	bo = gem_to_amdgpu_bo(obj);
 	if (!(bo->preferred_domains & (AMDGPU_GEM_DOMAIN_VRAM |
-				    AMDGPU_GEM_DOMAIN_GTT))) {
+				    AMDGPU_GEM_DOMAIN_GTT)))
 		/* Only VRAM and GTT BOs are supported */
-		ret = -EINVAL;
-		goto err_put_obj;
-	}
+		return -EINVAL;
 
 	*mem = kzalloc(sizeof(struct kgd_mem), GFP_KERNEL);
-	if (!*mem) {
-		ret = -ENOMEM;
-		goto err_put_obj;
-	}
+	if (!*mem)
+		return -ENOMEM;
 
 	ret = drm_vma_node_allow(&obj->vma_node, drm_priv);
 	if (ret)
@@ -2262,20 +2389,67 @@ int amdgpu_amdkfd_gpuvm_import_dmabuf(struct amdgpu_device *adev,
 	(*mem)->dmabuf = dma_buf;
 	(*mem)->bo = bo;
 	(*mem)->va = va;
-	(*mem)->domain = (bo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM) ?
-		AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT;
+	(*mem)->domain = (bo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM) &&
+			 !adev->apu_prefer_gtt ?
+			 AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT;
+
 	(*mem)->mapped_to_gpu_memory = 0;
 	(*mem)->process_info = avm->process_info;
 	add_kgd_mem_to_kfd_bo_list(*mem, avm->process_info, false);
 	amdgpu_sync_create(&(*mem)->sync);
 	(*mem)->is_imported = true;
 
+	mutex_lock(&avm->process_info->lock);
+	if (avm->process_info->eviction_fence &&
+	    !dma_fence_is_signaled(&avm->process_info->eviction_fence->base))
+		ret = amdgpu_amdkfd_bo_validate_and_fence(bo, (*mem)->domain,
+				&avm->process_info->eviction_fence->base);
+	mutex_unlock(&avm->process_info->lock);
+	if (ret)
+		goto err_remove_mem;
+
 	return 0;
 
+err_remove_mem:
+	remove_kgd_mem_from_kfd_bo_list(*mem, avm->process_info);
+	drm_vma_node_revoke(&obj->vma_node, drm_priv);
 err_free_mem:
 	kfree(*mem);
+	return ret;
+}
+
+int amdgpu_amdkfd_gpuvm_import_dmabuf_fd(struct amdgpu_device *adev, int fd,
+					 uint64_t va, void *drm_priv,
+					 struct kgd_mem **mem, uint64_t *size,
+					 uint64_t *mmap_offset)
+{
+	struct drm_gem_object *obj;
+	uint32_t handle;
+	int ret;
+
+	ret = drm_gem_prime_fd_to_handle(&adev->ddev, adev->kfd.client.file, fd,
+					 &handle);
+	if (ret)
+		return ret;
+	obj = drm_gem_object_lookup(adev->kfd.client.file, handle);
+	if (!obj) {
+		ret = -EINVAL;
+		goto err_release_handle;
+	}
+
+	ret = import_obj_create(adev, obj->dma_buf, obj, va, drm_priv, mem, size,
+				mmap_offset);
+	if (ret)
+		goto err_put_obj;
+
+	(*mem)->gem_handle = handle;
+
+	return 0;
+
 err_put_obj:
 	drm_gem_object_put(obj);
+err_release_handle:
+	drm_gem_handle_delete(adev->kfd.client.file, handle);
 	return ret;
 }
 
@@ -2326,10 +2500,14 @@ int amdgpu_amdkfd_evict_userptr(struct mmu_interval_notifier *mni,
 		/* First eviction, stop the queues */
 		r = kgd2kfd_quiesce_mm(mni->mm,
 				       KFD_QUEUE_EVICTION_TRIGGER_USERPTR);
-		if (r)
+
+		if (r && r != -ESRCH)
 			pr_err("Failed to quiesce KFD\n");
-		schedule_delayed_work(&process_info->restore_userptr_work,
-			msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS));
+
+		if (r != -ESRCH)
+			queue_delayed_work(system_freezable_wq,
+				&process_info->restore_userptr_work,
+				msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS));
 	}
 	mutex_unlock(&process_info->notifier_lock);
 
@@ -2356,14 +2534,14 @@ static int update_invalid_user_pages(struct amdkfd_process_info *process_info,
 	/* Move all invalidated BOs to the userptr_inval_list */
 	list_for_each_entry_safe(mem, tmp_mem,
 				 &process_info->userptr_valid_list,
-				 validate_list.head)
+				 validate_list)
 		if (mem->invalid)
-			list_move_tail(&mem->validate_list.head,
+			list_move_tail(&mem->validate_list,
 				       &process_info->userptr_inval_list);
 
 	/* Go through userptr_inval_list and update any invalid user_pages */
 	list_for_each_entry(mem, &process_info->userptr_inval_list,
-			    validate_list.head) {
+			    validate_list) {
 		invalid = mem->invalid;
 		if (!invalid)
 			/* BO hasn't been invalidated since the last
@@ -2373,7 +2551,7 @@ static int update_invalid_user_pages(struct amdkfd_process_info *process_info,
 
 		bo = mem->bo;
 
-		amdgpu_ttm_tt_discard_user_pages(bo->tbo.ttm, mem->range);
+		amdgpu_hmm_range_free(mem->range);
 		mem->range = NULL;
 
 		/* BO reservations and getting user pages (hmm_range_fault)
@@ -2397,10 +2575,14 @@ static int update_invalid_user_pages(struct amdkfd_process_info *process_info,
 			}
 		}
 
+		mem->range = amdgpu_hmm_range_alloc(NULL);
+		if (unlikely(!mem->range))
+			return -ENOMEM;
 		/* Get updated user pages */
-		ret = amdgpu_ttm_tt_get_user_pages(bo, bo->tbo.ttm->pages,
-						   &mem->range);
+		ret = amdgpu_ttm_tt_get_user_pages(bo, mem->range);
 		if (ret) {
+			amdgpu_hmm_range_free(mem->range);
+			mem->range = NULL;
 			pr_debug("Failed %d to get user pages\n", ret);
 
 			/* Return -EFAULT bad address error as success. It will
@@ -2413,9 +2595,28 @@ static int update_invalid_user_pages(struct amdkfd_process_info *process_info,
 			if (ret != -EFAULT)
 				return ret;
 
+			/* If applications unmap memory before destroying the userptr
+			 * from the KFD, trigger a segmentation fault in VM debug mode.
+			 */
+			if (amdgpu_ttm_adev(bo->tbo.bdev)->debug_vm_userptr) {
+				struct kfd_process *p;
+
+				pr_err("Pid %d unmapped memory before destroying userptr at GPU addr 0x%llx\n",
+								pid_nr(process_info->pid), mem->va);
+
+				// Send GPU VM fault to user space
+				p = kfd_lookup_process_by_pid(process_info->pid);
+				if (p) {
+					kfd_signal_vm_fault_event_with_userptr(p, mem->va);
+					kfd_unref_process(p);
+				}
+			}
+
 			ret = 0;
 		}
 
+		amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, mem->range);
+
 		mutex_lock(&process_info->notifier_lock);
 
 		/* Mark the BO as valid unless it was invalidated
@@ -2425,7 +2626,9 @@ static int update_invalid_user_pages(struct amdkfd_process_info *process_info,
 			ret = -EAGAIN;
 			goto unlock_out;
 		}
-		mem->invalid = 0;
+		 /* set mem valid if mem has hmm range associated */
+		if (mem->range)
+			mem->invalid = 0;
 	}
 
 unlock_out:
@@ -2441,59 +2644,50 @@ unlock_out:
  */
 static int validate_invalid_user_pages(struct amdkfd_process_info *process_info)
 {
-	struct amdgpu_bo_list_entry *pd_bo_list_entries;
-	struct list_head resv_list, duplicates;
-	struct ww_acquire_ctx ticket;
+	struct ttm_operation_ctx ctx = { false, false };
 	struct amdgpu_sync sync;
+	struct drm_exec exec;
 
 	struct amdgpu_vm *peer_vm;
 	struct kgd_mem *mem, *tmp_mem;
 	struct amdgpu_bo *bo;
-	struct ttm_operation_ctx ctx = { false, false };
-	int i, ret;
-
-	pd_bo_list_entries = kcalloc(process_info->n_vms,
-				     sizeof(struct amdgpu_bo_list_entry),
-				     GFP_KERNEL);
-	if (!pd_bo_list_entries) {
-		pr_err("%s: Failed to allocate PD BO list entries\n", __func__);
-		ret = -ENOMEM;
-		goto out_no_mem;
-	}
-
-	INIT_LIST_HEAD(&resv_list);
-	INIT_LIST_HEAD(&duplicates);
+	int ret;
 
-	/* Get all the page directory BOs that need to be reserved */
-	i = 0;
-	list_for_each_entry(peer_vm, &process_info->vm_list_head,
-			    vm_list_node)
-		amdgpu_vm_get_pd_bo(peer_vm, &resv_list,
-				    &pd_bo_list_entries[i++]);
-	/* Add the userptr_inval_list entries to resv_list */
-	list_for_each_entry(mem, &process_info->userptr_inval_list,
-			    validate_list.head) {
-		list_add_tail(&mem->resv_list.head, &resv_list);
-		mem->resv_list.bo = mem->validate_list.bo;
-		mem->resv_list.num_shared = mem->validate_list.num_shared;
-	}
+	amdgpu_sync_create(&sync);
 
+	drm_exec_init(&exec, 0, 0);
 	/* Reserve all BOs and page tables for validation */
-	ret = ttm_eu_reserve_buffers(&ticket, &resv_list, false, &duplicates);
-	WARN(!list_empty(&duplicates), "Duplicates should be empty");
-	if (ret)
-		goto out_free;
+	drm_exec_until_all_locked(&exec) {
+		/* Reserve all the page directories */
+		list_for_each_entry(peer_vm, &process_info->vm_list_head,
+				    vm_list_node) {
+			ret = amdgpu_vm_lock_pd(peer_vm, &exec, 2);
+			drm_exec_retry_on_contention(&exec);
+			if (unlikely(ret))
+				goto unreserve_out;
+		}
 
-	amdgpu_sync_create(&sync);
+		/* Reserve the userptr_inval_list entries to resv_list */
+		list_for_each_entry(mem, &process_info->userptr_inval_list,
+				    validate_list) {
+			struct drm_gem_object *gobj;
 
-	ret = process_validate_vms(process_info);
+			gobj = &mem->bo->tbo.base;
+			ret = drm_exec_prepare_obj(&exec, gobj, 1);
+			drm_exec_retry_on_contention(&exec);
+			if (unlikely(ret))
+				goto unreserve_out;
+		}
+	}
+
+	ret = process_validate_vms(process_info, NULL);
 	if (ret)
 		goto unreserve_out;
 
 	/* Validate BOs and update GPUVM page tables */
 	list_for_each_entry_safe(mem, tmp_mem,
 				 &process_info->userptr_inval_list,
-				 validate_list.head) {
+				 validate_list) {
 		struct kfd_mem_attachment *attachment;
 
 		bo = mem->bo;
@@ -2535,12 +2729,9 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info)
 	ret = process_update_pds(process_info, &sync);
 
 unreserve_out:
-	ttm_eu_backoff_reservation(&ticket, &resv_list);
+	drm_exec_fini(&exec);
 	amdgpu_sync_wait(&sync, false);
 	amdgpu_sync_free(&sync);
-out_free:
-	kfree(pd_bo_list_entries);
-out_no_mem:
 
 	return ret;
 }
@@ -2556,9 +2747,16 @@ static int confirm_valid_user_pages_locked(struct amdkfd_process_info *process_i
 
 	list_for_each_entry_safe(mem, tmp_mem,
 				 &process_info->userptr_inval_list,
-				 validate_list.head) {
-		bool valid = amdgpu_ttm_tt_get_user_pages_done(
-				mem->bo->tbo.ttm, mem->range);
+				 validate_list) {
+		bool valid;
+
+		/* keep mem without hmm range at userptr_inval_list */
+		if (!mem->range)
+			continue;
+
+		/* Only check mem with hmm range associated */
+		valid = amdgpu_hmm_range_valid(mem->range);
+		amdgpu_hmm_range_free(mem->range);
 
 		mem->range = NULL;
 		if (!valid) {
@@ -2566,9 +2764,14 @@ static int confirm_valid_user_pages_locked(struct amdkfd_process_info *process_i
 			ret = -EAGAIN;
 			continue;
 		}
-		WARN(mem->invalid, "Valid BO is marked invalid");
 
-		list_move_tail(&mem->validate_list.head,
+		if (mem->invalid) {
+			WARN(1, "Valid BO is marked invalid");
+			ret = -EAGAIN;
+			continue;
+		}
+
+		list_move_tail(&mem->validate_list,
 			       &process_info->userptr_valid_list);
 	}
 
@@ -2649,7 +2852,8 @@ unlock_out:
 
 	/* If validation failed, reschedule another attempt */
 	if (evicted_bos) {
-		schedule_delayed_work(&process_info->restore_userptr_work,
+		queue_delayed_work(system_freezable_wq,
+			&process_info->restore_userptr_work,
 			msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS));
 
 		kfd_smi_event_queue_restore_rescheduled(mm);
@@ -2658,6 +2862,23 @@ unlock_out:
 	put_task_struct(usertask);
 }
 
+static void replace_eviction_fence(struct dma_fence __rcu **ef,
+				   struct dma_fence *new_ef)
+{
+	struct dma_fence *old_ef = rcu_replace_pointer(*ef, new_ef, true
+		/* protected by process_info->lock */);
+
+	/* If we're replacing an unsignaled eviction fence, that fence will
+	 * never be signaled, and if anyone is still waiting on that fence,
+	 * they will hang forever. This should never happen. We should only
+	 * replace the fence in restore_work that only gets scheduled after
+	 * eviction work signaled the fence.
+	 */
+	WARN_ONCE(!dma_fence_is_signaled(old_ef),
+		  "Replacing unsignaled eviction fence");
+	dma_fence_put(old_ef);
+}
+
 /** amdgpu_amdkfd_gpuvm_restore_process_bos - Restore all BOs for the given
  *   KFD process identified by process_info
  *
@@ -2676,74 +2897,59 @@ unlock_out:
  * 7.  Add fence to all PD and PT BOs.
  * 8.  Unreserve all BOs
  */
-int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
+int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence __rcu **ef)
 {
-	struct amdgpu_bo_list_entry *pd_bo_list;
 	struct amdkfd_process_info *process_info = info;
 	struct amdgpu_vm *peer_vm;
 	struct kgd_mem *mem;
-	struct bo_vm_reservation_context ctx;
-	struct amdgpu_amdkfd_fence *new_fence;
-	int ret = 0, i;
 	struct list_head duplicate_save;
 	struct amdgpu_sync sync_obj;
 	unsigned long failed_size = 0;
 	unsigned long total_size = 0;
+	struct drm_exec exec;
+	int ret;
 
 	INIT_LIST_HEAD(&duplicate_save);
-	INIT_LIST_HEAD(&ctx.list);
-	INIT_LIST_HEAD(&ctx.duplicates);
-
-	pd_bo_list = kcalloc(process_info->n_vms,
-			     sizeof(struct amdgpu_bo_list_entry),
-			     GFP_KERNEL);
-	if (!pd_bo_list)
-		return -ENOMEM;
 
-	i = 0;
 	mutex_lock(&process_info->lock);
-	list_for_each_entry(peer_vm, &process_info->vm_list_head,
-			vm_list_node)
-		amdgpu_vm_get_pd_bo(peer_vm, &ctx.list, &pd_bo_list[i++]);
 
-	/* Reserve all BOs and page tables/directory. Add all BOs from
-	 * kfd_bo_list to ctx.list
-	 */
-	list_for_each_entry(mem, &process_info->kfd_bo_list,
-			    validate_list.head) {
-
-		list_add_tail(&mem->resv_list.head, &ctx.list);
-		mem->resv_list.bo = mem->validate_list.bo;
-		mem->resv_list.num_shared = mem->validate_list.num_shared;
-	}
+	drm_exec_init(&exec, DRM_EXEC_IGNORE_DUPLICATES, 0);
+	drm_exec_until_all_locked(&exec) {
+		list_for_each_entry(peer_vm, &process_info->vm_list_head,
+				    vm_list_node) {
+			ret = amdgpu_vm_lock_pd(peer_vm, &exec, 2);
+			drm_exec_retry_on_contention(&exec);
+			if (unlikely(ret)) {
+				pr_err("Locking VM PD failed, ret: %d\n", ret);
+				goto ttm_reserve_fail;
+			}
+		}
 
-	ret = ttm_eu_reserve_buffers(&ctx.ticket, &ctx.list,
-				     false, &duplicate_save);
-	if (ret) {
-		pr_debug("Memory eviction: TTM Reserve Failed. Try again\n");
-		goto ttm_reserve_fail;
+		/* Reserve all BOs and page tables/directory. Add all BOs from
+		 * kfd_bo_list to ctx.list
+		 */
+		list_for_each_entry(mem, &process_info->kfd_bo_list,
+				    validate_list) {
+			struct drm_gem_object *gobj;
+
+			gobj = &mem->bo->tbo.base;
+			ret = drm_exec_prepare_obj(&exec, gobj, 1);
+			drm_exec_retry_on_contention(&exec);
+			if (unlikely(ret)) {
+				pr_err("drm_exec_prepare_obj failed, ret: %d\n", ret);
+				goto ttm_reserve_fail;
+			}
+		}
 	}
 
 	amdgpu_sync_create(&sync_obj);
 
-	/* Validate PDs and PTs */
-	ret = process_validate_vms(process_info);
-	if (ret)
-		goto validate_map_fail;
-
-	ret = process_sync_pds_resv(process_info, &sync_obj);
-	if (ret) {
-		pr_debug("Memory eviction: Failed to sync to PD BO moving fence. Try again\n");
-		goto validate_map_fail;
-	}
-
-	/* Validate BOs and map them to GPUVM (update VM page tables). */
+	/* Validate BOs managed by KFD */
 	list_for_each_entry(mem, &process_info->kfd_bo_list,
-			    validate_list.head) {
+			    validate_list) {
 
 		struct amdgpu_bo *bo = mem->bo;
 		uint32_t domain = mem->domain;
-		struct kfd_mem_attachment *attachment;
 		struct dma_resv_iter cursor;
 		struct dma_fence *fence;
 
@@ -2762,12 +2968,31 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
 		}
 		dma_resv_for_each_fence(&cursor, bo->tbo.base.resv,
 					DMA_RESV_USAGE_KERNEL, fence) {
-			ret = amdgpu_sync_fence(&sync_obj, fence);
+			ret = amdgpu_sync_fence(&sync_obj, fence, GFP_KERNEL);
 			if (ret) {
 				pr_debug("Memory eviction: Sync BO fence failed. Try again\n");
 				goto validate_map_fail;
 			}
 		}
+	}
+
+	if (failed_size)
+		pr_debug("0x%lx/0x%lx in system\n", failed_size, total_size);
+
+	/* Validate PDs, PTs and evicted DMABuf imports last. Otherwise BO
+	 * validations above would invalidate DMABuf imports again.
+	 */
+	ret = process_validate_vms(process_info, &exec.ticket);
+	if (ret) {
+		pr_debug("Validating VMs failed, ret: %d\n", ret);
+		goto validate_map_fail;
+	}
+
+	/* Update mappings managed by KFD. */
+	list_for_each_entry(mem, &process_info->kfd_bo_list,
+			    validate_list) {
+		struct kfd_mem_attachment *attachment;
+
 		list_for_each_entry(attachment, &mem->attachments, list) {
 			if (!attachment->is_mapped)
 				continue;
@@ -2781,8 +3006,31 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
 		}
 	}
 
-	if (failed_size)
-		pr_debug("0x%lx/0x%lx in system\n", failed_size, total_size);
+	/* Update mappings not managed by KFD */
+	list_for_each_entry(peer_vm, &process_info->vm_list_head,
+			vm_list_node) {
+		struct amdgpu_device *adev = amdgpu_ttm_adev(
+			peer_vm->root.bo->tbo.bdev);
+
+		struct amdgpu_fpriv *fpriv =
+			container_of(peer_vm, struct amdgpu_fpriv, vm);
+
+		ret = amdgpu_vm_bo_update(adev, fpriv->prt_va, false);
+		if (ret) {
+			dev_dbg(adev->dev,
+				"Memory eviction: handle PRT moved failed, pid %8d. Try again.\n",
+				pid_nr(process_info->pid));
+			goto validate_map_fail;
+		}
+
+		ret = amdgpu_vm_handle_moved(adev, peer_vm, &exec.ticket);
+		if (ret) {
+			dev_dbg(adev->dev,
+				"Memory eviction: handle moved failed, pid %8d. Try again.\n",
+				pid_nr(process_info->pid));
+			goto validate_map_fail;
+		}
+	}
 
 	/* Update page directories */
 	ret = process_update_pds(process_info, &sync_obj);
@@ -2791,29 +3039,50 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
 		goto validate_map_fail;
 	}
 
+	/* Sync with fences on all the page tables. They implicitly depend on any
+	 * move fences from amdgpu_vm_handle_moved above.
+	 */
+	ret = process_sync_pds_resv(process_info, &sync_obj);
+	if (ret) {
+		pr_debug("Memory eviction: Failed to sync to PD BO moving fence. Try again\n");
+		goto validate_map_fail;
+	}
+
 	/* Wait for validate and PT updates to finish */
 	amdgpu_sync_wait(&sync_obj, false);
 
-	/* Release old eviction fence and create new one, because fence only
-	 * goes from unsignaled to signaled, fence cannot be reused.
-	 * Use context and mm from the old fence.
+	/* The old eviction fence may be unsignaled if restore happens
+	 * after a GPU reset or suspend/resume. Keep the old fence in that
+	 * case. Otherwise release the old eviction fence and create new
+	 * one, because fence only goes from unsignaled to signaled once
+	 * and cannot be reused. Use context and mm from the old fence.
+	 *
+	 * If an old eviction fence signals after this check, that's OK.
+	 * Anyone signaling an eviction fence must stop the queues first
+	 * and schedule another restore worker.
 	 */
-	new_fence = amdgpu_amdkfd_fence_create(
+	if (dma_fence_is_signaled(&process_info->eviction_fence->base)) {
+		struct amdgpu_amdkfd_fence *new_fence =
+			amdgpu_amdkfd_fence_create(
 				process_info->eviction_fence->base.context,
 				process_info->eviction_fence->mm,
 				NULL);
-	if (!new_fence) {
-		pr_err("Failed to create eviction fence\n");
-		ret = -ENOMEM;
-		goto validate_map_fail;
+
+		if (!new_fence) {
+			pr_err("Failed to create eviction fence\n");
+			ret = -ENOMEM;
+			goto validate_map_fail;
+		}
+		dma_fence_put(&process_info->eviction_fence->base);
+		process_info->eviction_fence = new_fence;
+		replace_eviction_fence(ef, dma_fence_get(&new_fence->base));
+	} else {
+		WARN_ONCE(*ef != &process_info->eviction_fence->base,
+			  "KFD eviction fence doesn't match KGD process_info");
 	}
-	dma_fence_put(&process_info->eviction_fence->base);
-	process_info->eviction_fence = new_fence;
-	*ef = dma_fence_get(&new_fence->base);
 
 	/* Attach new eviction fence to all BOs except pinned ones */
-	list_for_each_entry(mem, &process_info->kfd_bo_list,
-		validate_list.head) {
+	list_for_each_entry(mem, &process_info->kfd_bo_list, validate_list) {
 		if (mem->bo->tbo.pin_count)
 			continue;
 
@@ -2821,7 +3090,7 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
 				   &process_info->eviction_fence->base,
 				   DMA_RESV_USAGE_BOOKKEEP);
 	}
-	/* Attach eviction fence to PD / PT BOs */
+	/* Attach eviction fence to PD / PT BOs and DMABuf imports */
 	list_for_each_entry(peer_vm, &process_info->vm_list_head,
 			    vm_list_node) {
 		struct amdgpu_bo *bo = peer_vm->root.bo;
@@ -2832,11 +3101,10 @@ int amdgpu_amdkfd_gpuvm_restore_process_bos(void *info, struct dma_fence **ef)
 	}
 
 validate_map_fail:
-	ttm_eu_backoff_reservation(&ctx.ticket, &ctx.list);
 	amdgpu_sync_free(&sync_obj);
 ttm_reserve_fail:
+	drm_exec_fini(&exec);
 	mutex_unlock(&process_info->lock);
-	kfree(pd_bo_list);
 	return ret;
 }
 
@@ -2952,12 +3220,13 @@ int amdgpu_amdkfd_get_tile_config(struct amdgpu_device *adev,
 	return 0;
 }
 
-bool amdgpu_amdkfd_bo_mapped_to_dev(struct amdgpu_device *adev, struct kgd_mem *mem)
+bool amdgpu_amdkfd_bo_mapped_to_dev(void *drm_priv, struct kgd_mem *mem)
 {
+	struct amdgpu_vm *vm = drm_priv_to_vm(drm_priv);
 	struct kfd_mem_attachment *entry;
 
 	list_for_each_entry(entry, &mem->attachments, list) {
-		if (entry->is_mapped && entry->adev == adev)
+		if (entry->is_mapped && entry->bo_va->base.vm == vm)
 			return true;
 	}
 	return false;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c
index 9ba4817a9148..763f2b8dcf13 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c
@@ -36,13 +36,6 @@
 #include "atombios_encoders.h"
 #include "bif/bif_4_1_d.h"
 
-static void amdgpu_atombios_lookup_i2c_gpio_quirks(struct amdgpu_device *adev,
-					  ATOM_GPIO_I2C_ASSIGMENT *gpio,
-					  u8 index)
-{
-
-}
-
 static struct amdgpu_i2c_bus_rec amdgpu_atombios_get_bus_rec_for_i2c_gpio(ATOM_GPIO_I2C_ASSIGMENT *gpio)
 {
 	struct amdgpu_i2c_bus_rec i2c;
@@ -108,9 +101,6 @@ struct amdgpu_i2c_bus_rec amdgpu_atombios_lookup_i2c_gpio(struct amdgpu_device *
 
 		gpio = &i2c_info->asGPIO_Info[0];
 		for (i = 0; i < num_indices; i++) {
-
-			amdgpu_atombios_lookup_i2c_gpio_quirks(adev, gpio, i);
-
 			if (gpio->sucI2cId.ucAccess == id) {
 				i2c = amdgpu_atombios_get_bus_rec_for_i2c_gpio(gpio);
 				break;
@@ -142,8 +132,6 @@ void amdgpu_atombios_i2c_init(struct amdgpu_device *adev)
 
 		gpio = &i2c_info->asGPIO_Info[0];
 		for (i = 0; i < num_indices; i++) {
-			amdgpu_atombios_lookup_i2c_gpio_quirks(adev, gpio, i);
-
 			i2c = amdgpu_atombios_get_bus_rec_for_i2c_gpio(gpio);
 
 			if (i2c.valid) {
@@ -156,6 +144,38 @@ void amdgpu_atombios_i2c_init(struct amdgpu_device *adev)
 	}
 }
 
+void amdgpu_atombios_oem_i2c_init(struct amdgpu_device *adev, u8 i2c_id)
+{
+	struct atom_context *ctx = adev->mode_info.atom_context;
+	ATOM_GPIO_I2C_ASSIGMENT *gpio;
+	struct amdgpu_i2c_bus_rec i2c;
+	int index = GetIndexIntoMasterTable(DATA, GPIO_I2C_Info);
+	struct _ATOM_GPIO_I2C_INFO *i2c_info;
+	uint16_t data_offset, size;
+	int i, num_indices;
+	char stmp[32];
+
+	if (amdgpu_atom_parse_data_header(ctx, index, &size, NULL, NULL, &data_offset)) {
+		i2c_info = (struct _ATOM_GPIO_I2C_INFO *)(ctx->bios + data_offset);
+
+		num_indices = (size - sizeof(ATOM_COMMON_TABLE_HEADER)) /
+			sizeof(ATOM_GPIO_I2C_ASSIGMENT);
+
+		gpio = &i2c_info->asGPIO_Info[0];
+		for (i = 0; i < num_indices; i++) {
+			i2c = amdgpu_atombios_get_bus_rec_for_i2c_gpio(gpio);
+
+			if (i2c.valid && i2c.i2c_id == i2c_id) {
+				sprintf(stmp, "OEM 0x%x", i2c.i2c_id);
+				adev->i2c_bus[i] = amdgpu_i2c_create(adev_to_drm(adev), &i2c, stmp);
+				break;
+			}
+			gpio = (ATOM_GPIO_I2C_ASSIGMENT *)
+				((u8 *)gpio + sizeof(ATOM_GPIO_I2C_ASSIGMENT));
+		}
+	}
+}
+
 struct amdgpu_gpio_rec
 amdgpu_atombios_lookup_gpio(struct amdgpu_device *adev,
 			    u8 id)
@@ -686,7 +706,6 @@ int amdgpu_atombios_get_clock_info(struct amdgpu_device *adev)
 		}
 		adev->clock.dp_extclk =
 			le16_to_cpu(firmware_info->info_21.usUniphyDPModeExtClkFreq);
-		adev->clock.current_dispclk = adev->clock.default_dispclk;
 
 		adev->clock.max_pixel_clock = le16_to_cpu(firmware_info->info.usMaxPixelClock);
 		if (adev->clock.max_pixel_clock == 0)
@@ -1018,7 +1037,9 @@ int amdgpu_atombios_get_clock_dividers(struct amdgpu_device *adev,
 		if (clock_type == COMPUTE_ENGINE_PLL_PARAM) {
 			args.v3.ulClockParams = cpu_to_le32((clock_type << 24) | clock);
 
-			amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+			if (amdgpu_atom_execute_table(adev->mode_info.atom_context,
+			    index, (uint32_t *)&args, sizeof(args)))
+				return -EINVAL;
 
 			dividers->post_div = args.v3.ucPostDiv;
 			dividers->enable_post_div = (args.v3.ucCntlFlag &
@@ -1038,7 +1059,9 @@ int amdgpu_atombios_get_clock_dividers(struct amdgpu_device *adev,
 			if (strobe_mode)
 				args.v5.ucInputFlag = ATOM_PLL_INPUT_FLAG_PLL_STROBE_MODE_EN;
 
-			amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+			if (amdgpu_atom_execute_table(adev->mode_info.atom_context,
+			    index, (uint32_t *)&args, sizeof(args)))
+				return -EINVAL;
 
 			dividers->post_div = args.v5.ucPostDiv;
 			dividers->enable_post_div = (args.v5.ucCntlFlag &
@@ -1056,7 +1079,9 @@ int amdgpu_atombios_get_clock_dividers(struct amdgpu_device *adev,
 		/* fusion */
 		args.v4.ulClock = cpu_to_le32(clock);	/* 10 khz */
 
-		amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+		if (amdgpu_atom_execute_table(adev->mode_info.atom_context,
+		    index, (uint32_t *)&args, sizeof(args)))
+			return -EINVAL;
 
 		dividers->post_divider = dividers->post_div = args.v4.ucPostDiv;
 		dividers->real_clock = le32_to_cpu(args.v4.ulClock);
@@ -1067,7 +1092,9 @@ int amdgpu_atombios_get_clock_dividers(struct amdgpu_device *adev,
 		args.v6_in.ulClock.ulComputeClockFlag = clock_type;
 		args.v6_in.ulClock.ulClockFreq = cpu_to_le32(clock);	/* 10 khz */
 
-		amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+		if (amdgpu_atom_execute_table(adev->mode_info.atom_context,
+		    index, (uint32_t *)&args, sizeof(args)))
+			return -EINVAL;
 
 		dividers->whole_fb_div = le16_to_cpu(args.v6_out.ulFbDiv.usFbDiv);
 		dividers->frac_fb_div = le16_to_cpu(args.v6_out.ulFbDiv.usFbDivFrac);
@@ -1109,7 +1136,9 @@ int amdgpu_atombios_get_memory_pll_dividers(struct amdgpu_device *adev,
 			if (strobe_mode)
 				args.ucInputFlag |= MPLL_INPUT_FLAG_STROBE_MODE_EN;
 
-			amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+			if (amdgpu_atom_execute_table(adev->mode_info.atom_context,
+			    index, (uint32_t *)&args, sizeof(args)))
+				return -EINVAL;
 
 			mpll_param->clkfrac = le16_to_cpu(args.ulFbDiv.usFbDivFrac);
 			mpll_param->clkf = le16_to_cpu(args.ulFbDiv.usFbDiv);
@@ -1135,8 +1164,8 @@ int amdgpu_atombios_get_memory_pll_dividers(struct amdgpu_device *adev,
 	return 0;
 }
 
-void amdgpu_atombios_set_engine_dram_timings(struct amdgpu_device *adev,
-					     u32 eng_clock, u32 mem_clock)
+int amdgpu_atombios_set_engine_dram_timings(struct amdgpu_device *adev,
+					    u32 eng_clock, u32 mem_clock)
 {
 	SET_ENGINE_CLOCK_PS_ALLOCATION args;
 	int index = GetIndexIntoMasterTable(COMMAND, DynamicMemorySettings);
@@ -1151,7 +1180,8 @@ void amdgpu_atombios_set_engine_dram_timings(struct amdgpu_device *adev,
 	if (mem_clock)
 		args.sReserved.ulClock = cpu_to_le32(mem_clock & SET_CLOCK_FREQ_MASK);
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	return amdgpu_atom_execute_table(adev->mode_info.atom_context, index,
+					 (uint32_t *)&args, sizeof(args));
 }
 
 void amdgpu_atombios_get_default_voltages(struct amdgpu_device *adev,
@@ -1205,7 +1235,9 @@ int amdgpu_atombios_get_max_vddc(struct amdgpu_device *adev, u8 voltage_type,
 		args.v2.ucVoltageMode = 0;
 		args.v2.usVoltageLevel = 0;
 
-		amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+		if (amdgpu_atom_execute_table(adev->mode_info.atom_context,
+		    index, (uint32_t *)&args, sizeof(args)))
+			return -EINVAL;
 
 		*voltage = le16_to_cpu(args.v2.usVoltageLevel);
 		break;
@@ -1214,7 +1246,9 @@ int amdgpu_atombios_get_max_vddc(struct amdgpu_device *adev, u8 voltage_type,
 		args.v3.ucVoltageMode = ATOM_GET_VOLTAGE_LEVEL;
 		args.v3.usVoltageLevel = cpu_to_le16(voltage_id);
 
-		amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+		if (amdgpu_atom_execute_table(adev->mode_info.atom_context,
+		    index, (uint32_t *)&args, sizeof(args)))
+			return -EINVAL;
 
 		*voltage = le16_to_cpu(args.v3.usVoltageLevel);
 		break;
@@ -1476,6 +1510,8 @@ int amdgpu_atombios_init_mc_reg_table(struct amdgpu_device *adev,
 										(u32)le32_to_cpu(*((u32 *)reg_data + j));
 									j++;
 								} else if ((reg_table->mc_reg_address[i].pre_reg_data & LOW_NIBBLE_MASK) == DATA_EQU_PREV) {
+									if (i == 0)
+										continue;
 									reg_table->mc_reg_table_entry[num_ranges].mc_data[i] =
 										reg_table->mc_reg_table_entry[num_ranges].mc_data[i - 1];
 								}
@@ -1776,21 +1812,57 @@ static ssize_t amdgpu_atombios_get_vbios_version(struct device *dev,
 	struct amdgpu_device *adev = drm_to_adev(ddev);
 	struct atom_context *ctx = adev->mode_info.atom_context;
 
-	return sysfs_emit(buf, "%s\n", ctx->vbios_version);
+	return sysfs_emit(buf, "%s\n", ctx->vbios_pn);
+}
+
+static ssize_t amdgpu_atombios_get_vbios_build(struct device *dev,
+					       struct device_attribute *attr,
+					       char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	struct atom_context *ctx = adev->mode_info.atom_context;
+
+	return sysfs_emit(buf, "%s\n", ctx->build_num);
 }
 
 static DEVICE_ATTR(vbios_version, 0444, amdgpu_atombios_get_vbios_version,
 		   NULL);
+static DEVICE_ATTR(vbios_build, 0444, amdgpu_atombios_get_vbios_build, NULL);
 
 static struct attribute *amdgpu_vbios_version_attrs[] = {
-	&dev_attr_vbios_version.attr,
-	NULL
+	&dev_attr_vbios_version.attr, &dev_attr_vbios_build.attr, NULL
 };
 
+static umode_t amdgpu_vbios_version_attrs_is_visible(struct kobject *kobj,
+						     struct attribute *attr,
+						     int index)
+{
+	struct device *dev = kobj_to_dev(kobj);
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	struct atom_context *ctx = adev->mode_info.atom_context;
+
+	if (attr == &dev_attr_vbios_build.attr && !strlen(ctx->build_num))
+		return 0;
+
+	return attr->mode;
+}
+
 const struct attribute_group amdgpu_vbios_version_attr_group = {
-	.attrs = amdgpu_vbios_version_attrs
+	.attrs = amdgpu_vbios_version_attrs,
+	.is_visible = amdgpu_vbios_version_attrs_is_visible,
 };
 
+int amdgpu_atombios_sysfs_init(struct amdgpu_device *adev)
+{
+	if (adev->mode_info.atom_context)
+		return devm_device_add_group(adev->dev,
+					     &amdgpu_vbios_version_attr_group);
+
+	return 0;
+}
+
 /**
  * amdgpu_atombios_fini - free the driver info and callbacks for atombios
  *
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h
index 4153d520e2a3..867bc5c5ce67 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.h
@@ -89,8 +89,7 @@ struct atom_memory_info {
 
 #define MAX_AC_TIMING_ENTRIES 16
 
-struct atom_memory_clock_range_table
-{
+struct atom_memory_clock_range_table {
 	u8 num_entries;
 	u8 rsv[3];
 	u32 mclk[MAX_AC_TIMING_ENTRIES];
@@ -118,14 +117,12 @@ struct atom_mc_reg_table {
 
 #define MAX_VOLTAGE_ENTRIES 32
 
-struct atom_voltage_table_entry
-{
+struct atom_voltage_table_entry {
 	u16 value;
 	u32 smio_low;
 };
 
-struct atom_voltage_table
-{
+struct atom_voltage_table {
 	u32 count;
 	u32 mask_low;
 	u32 phase_delay;
@@ -139,6 +136,7 @@ amdgpu_atombios_lookup_gpio(struct amdgpu_device *adev,
 struct amdgpu_i2c_bus_rec amdgpu_atombios_lookup_i2c_gpio(struct amdgpu_device *adev,
 							  uint8_t id);
 void amdgpu_atombios_i2c_init(struct amdgpu_device *adev);
+void amdgpu_atombios_oem_i2c_init(struct amdgpu_device *adev, u8 i2c_id);
 
 bool amdgpu_atombios_has_dce_engine_info(struct amdgpu_device *adev);
 
@@ -166,8 +164,8 @@ int amdgpu_atombios_get_memory_pll_dividers(struct amdgpu_device *adev,
 					    bool strobe_mode,
 					    struct atom_mpll_param *mpll_param);
 
-void amdgpu_atombios_set_engine_dram_timings(struct amdgpu_device *adev,
-					     u32 eng_clock, u32 mem_clock);
+int amdgpu_atombios_set_engine_dram_timings(struct amdgpu_device *adev,
+					    u32 eng_clock, u32 mem_clock);
 
 bool
 amdgpu_atombios_is_voltage_gpio(struct amdgpu_device *adev,
@@ -217,5 +215,6 @@ int amdgpu_atombios_get_data_table(struct amdgpu_device *adev,
 
 void amdgpu_atombios_fini(struct amdgpu_device *adev);
 int amdgpu_atombios_init(struct amdgpu_device *adev);
+int amdgpu_atombios_sysfs_init(struct amdgpu_device *adev);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c
index ac6fe0ae4609..636385c80f64 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c
@@ -34,6 +34,7 @@ union firmware_info {
 	struct atom_firmware_info_v3_2 v32;
 	struct atom_firmware_info_v3_3 v33;
 	struct atom_firmware_info_v3_4 v34;
+	struct atom_firmware_info_v3_5 v35;
 };
 
 /*
@@ -58,7 +59,7 @@ uint32_t amdgpu_atomfirmware_query_firmware_capability(struct amdgpu_device *ade
 	if (amdgpu_atom_parse_data_header(adev->mode_info.atom_context,
 				index, &size, &frev, &crev, &data_offset)) {
 		/* support firmware_info 3.1 + */
-		if ((frev == 3 && crev >=1) || (frev > 3)) {
+		if ((frev == 3 && crev >= 1) || (frev > 3)) {
 			firmware_info = (union firmware_info *)
 				(mode_info->atom_context->bios + data_offset);
 			fw_cap = le32_to_cpu(firmware_info->v31.firmware_capability);
@@ -180,19 +181,22 @@ int amdgpu_atomfirmware_allocate_fb_scratch(struct amdgpu_device *adev)
 	u8 frev, crev;
 	int usage_bytes = 0;
 
-	if (amdgpu_atom_parse_data_header(ctx, index, NULL, &frev, &crev, &data_offset)) {
-		if (frev == 2 && crev == 1) {
-			fw_usage_v2_1 =
-				(struct vram_usagebyfirmware_v2_1 *)(ctx->bios + data_offset);
-			amdgpu_atomfirmware_allocate_fb_v2_1(adev,
-					fw_usage_v2_1,
-					&usage_bytes);
-		} else if (frev >= 2 && crev >= 2) {
-			fw_usage_v2_2 =
-				(struct vram_usagebyfirmware_v2_2 *)(ctx->bios + data_offset);
-			amdgpu_atomfirmware_allocate_fb_v2_2(adev,
-					fw_usage_v2_2,
-					&usage_bytes);
+	/* Skip atomfirmware allocation for SRIOV VFs when dynamic crit regn is enabled */
+	if (!(amdgpu_sriov_vf(adev) && adev->virt.is_dynamic_crit_regn_enabled)) {
+		if (amdgpu_atom_parse_data_header(ctx, index, NULL, &frev, &crev, &data_offset)) {
+			if (frev == 2 && crev == 1) {
+				fw_usage_v2_1 =
+					(struct vram_usagebyfirmware_v2_1 *)(ctx->bios + data_offset);
+				amdgpu_atomfirmware_allocate_fb_v2_1(adev,
+						fw_usage_v2_1,
+						&usage_bytes);
+			} else if (frev >= 2 && crev >= 2) {
+				fw_usage_v2_2 =
+					(struct vram_usagebyfirmware_v2_2 *)(ctx->bios + data_offset);
+				amdgpu_atomfirmware_allocate_fb_v2_2(adev,
+						fw_usage_v2_2,
+						&usage_bytes);
+			}
 		}
 	}
 
@@ -211,12 +215,14 @@ union igp_info {
 	struct atom_integrated_system_info_v1_11 v11;
 	struct atom_integrated_system_info_v1_12 v12;
 	struct atom_integrated_system_info_v2_1 v21;
+	struct atom_integrated_system_info_v2_3 v23;
 };
 
 union umc_info {
 	struct atom_umc_info_v3_1 v31;
 	struct atom_umc_info_v3_2 v32;
 	struct atom_umc_info_v3_3 v33;
+	struct atom_umc_info_v4_0 v40;
 };
 
 union vram_info {
@@ -272,11 +278,15 @@ static int convert_atom_mem_type_to_vram_type(struct amdgpu_device *adev,
 			break;
 		case ATOM_DGPU_VRAM_TYPE_HBM2:
 		case ATOM_DGPU_VRAM_TYPE_HBM2E:
+		case ATOM_DGPU_VRAM_TYPE_HBM3:
 			vram_type = AMDGPU_VRAM_TYPE_HBM;
 			break;
 		case ATOM_DGPU_VRAM_TYPE_GDDR6:
 			vram_type = AMDGPU_VRAM_TYPE_GDDR6;
 			break;
+		case ATOM_DGPU_VRAM_TYPE_HBM3E:
+			vram_type = AMDGPU_VRAM_TYPE_HBM3E;
+			break;
 		default:
 			vram_type = AMDGPU_VRAM_TYPE_UNKNOWN;
 			break;
@@ -286,7 +296,6 @@ static int convert_atom_mem_type_to_vram_type(struct amdgpu_device *adev,
 	return vram_type;
 }
 
-
 int
 amdgpu_atomfirmware_get_vram_info(struct amdgpu_device *adev,
 				  int *vram_width, int *vram_type,
@@ -297,6 +306,7 @@ amdgpu_atomfirmware_get_vram_info(struct amdgpu_device *adev,
 	u16 data_offset, size;
 	union igp_info *igp_info;
 	union vram_info *vram_info;
+	union umc_info *umc_info;
 	union vram_module *vram_module;
 	u8 frev, crev;
 	u8 mem_type;
@@ -308,10 +318,16 @@ amdgpu_atomfirmware_get_vram_info(struct amdgpu_device *adev,
 	if (adev->flags & AMD_IS_APU)
 		index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1,
 						    integratedsysteminfo);
-	else
-		index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1,
-						    vram_info);
-
+	else {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+		case IP_VERSION(12, 0, 0):
+		case IP_VERSION(12, 0, 1):
+			index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, umc_info);
+			break;
+		default:
+			index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, vram_info);
+		}
+	}
 	if (amdgpu_atom_parse_data_header(mode_info->atom_context,
 					  index, &size,
 					  &frev, &crev, &data_offset)) {
@@ -326,10 +342,13 @@ amdgpu_atomfirmware_get_vram_info(struct amdgpu_device *adev,
 					mem_channel_number = igp_info->v11.umachannelnumber;
 					if (!mem_channel_number)
 						mem_channel_number = 1;
-					/* channel width is 64 */
-					if (vram_width)
-						*vram_width = mem_channel_number * 64;
 					mem_type = igp_info->v11.memorytype;
+					if (mem_type == LpDdr5MemType)
+						mem_channel_width = 32;
+					else
+						mem_channel_width = 64;
+					if (vram_width)
+						*vram_width = mem_channel_number * mem_channel_width;
 					if (vram_type)
 						*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
 					break;
@@ -344,10 +363,27 @@ amdgpu_atomfirmware_get_vram_info(struct amdgpu_device *adev,
 					mem_channel_number = igp_info->v21.umachannelnumber;
 					if (!mem_channel_number)
 						mem_channel_number = 1;
-					/* channel width is 64 */
-					if (vram_width)
-						*vram_width = mem_channel_number * 64;
 					mem_type = igp_info->v21.memorytype;
+					if (mem_type == LpDdr5MemType)
+						mem_channel_width = 32;
+					else
+						mem_channel_width = 64;
+					if (vram_width)
+						*vram_width = mem_channel_number * mem_channel_width;
+					if (vram_type)
+						*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
+					break;
+				case 3:
+					mem_channel_number = igp_info->v23.umachannelnumber;
+					if (!mem_channel_number)
+						mem_channel_number = 1;
+					mem_type = igp_info->v23.memorytype;
+					if (mem_type == LpDdr5MemType)
+						mem_channel_width = 32;
+					else
+						mem_channel_width = 64;
+					if (vram_width)
+						*vram_width = mem_channel_number * mem_channel_width;
 					if (vram_type)
 						*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
 					break;
@@ -359,123 +395,150 @@ amdgpu_atomfirmware_get_vram_info(struct amdgpu_device *adev,
 				return -EINVAL;
 			}
 		} else {
-			vram_info = (union vram_info *)
-				(mode_info->atom_context->bios + data_offset);
-			module_id = (RREG32(adev->bios_scratch_reg_offset + 4) & 0x00ff0000) >> 16;
-			if (frev == 3) {
-				switch (crev) {
-				/* v30 */
-				case 0:
-					vram_module = (union vram_module *)vram_info->v30.vram_module;
-					mem_vendor = (vram_module->v30.dram_vendor_id) & 0xF;
-					if (vram_vendor)
-						*vram_vendor = mem_vendor;
-					mem_type = vram_info->v30.memory_type;
-					if (vram_type)
-						*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
-					mem_channel_number = vram_info->v30.channel_num;
-					mem_channel_width = vram_info->v30.channel_width;
-					if (vram_width)
-						*vram_width = mem_channel_number * (1 << mem_channel_width);
-					break;
-				default:
-					return -EINVAL;
-				}
-			} else if (frev == 2) {
-				switch (crev) {
-				/* v23 */
-				case 3:
-					if (module_id > vram_info->v23.vram_module_num)
-						module_id = 0;
-					vram_module = (union vram_module *)vram_info->v23.vram_module;
-					while (i < module_id) {
-						vram_module = (union vram_module *)
-							((u8 *)vram_module + vram_module->v9.vram_module_size);
-						i++;
+			switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+			case IP_VERSION(12, 0, 0):
+			case IP_VERSION(12, 0, 1):
+				umc_info = (union umc_info *)(mode_info->atom_context->bios + data_offset);
+
+				if (frev == 4) {
+					switch (crev) {
+					case 0:
+						mem_channel_number = le32_to_cpu(umc_info->v40.channel_num);
+						mem_type = le32_to_cpu(umc_info->v40.vram_type);
+						mem_channel_width = le32_to_cpu(umc_info->v40.channel_width);
+						mem_vendor = RREG32(adev->bios_scratch_reg_offset + 4) & 0xF;
+						if (vram_vendor)
+							*vram_vendor = mem_vendor;
+						if (vram_type)
+							*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
+						if (vram_width)
+							*vram_width = mem_channel_number * (1 << mem_channel_width);
+						break;
+					default:
+						return -EINVAL;
 					}
-					mem_type = vram_module->v9.memory_type;
-					if (vram_type)
-						*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
-					mem_channel_number = vram_module->v9.channel_num;
-					mem_channel_width = vram_module->v9.channel_width;
-					if (vram_width)
-						*vram_width = mem_channel_number * (1 << mem_channel_width);
-					mem_vendor = (vram_module->v9.vender_rev_id) & 0xF;
-					if (vram_vendor)
-						*vram_vendor = mem_vendor;
-					break;
-				/* v24 */
-				case 4:
-					if (module_id > vram_info->v24.vram_module_num)
-						module_id = 0;
-					vram_module = (union vram_module *)vram_info->v24.vram_module;
-					while (i < module_id) {
-						vram_module = (union vram_module *)
-							((u8 *)vram_module + vram_module->v10.vram_module_size);
-						i++;
+				} else
+					return -EINVAL;
+				break;
+			default:
+				vram_info = (union vram_info *)
+					(mode_info->atom_context->bios + data_offset);
+
+				module_id = (RREG32(adev->bios_scratch_reg_offset + 4) & 0x00ff0000) >> 16;
+				if (frev == 3) {
+					switch (crev) {
+					/* v30 */
+					case 0:
+						vram_module = (union vram_module *)vram_info->v30.vram_module;
+						mem_vendor = (vram_module->v30.dram_vendor_id) & 0xF;
+						if (vram_vendor)
+							*vram_vendor = mem_vendor;
+						mem_type = vram_info->v30.memory_type;
+						if (vram_type)
+							*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
+						mem_channel_number = vram_info->v30.channel_num;
+						mem_channel_width = vram_info->v30.channel_width;
+						if (vram_width)
+							*vram_width = mem_channel_number * 16;
+						break;
+					default:
+						return -EINVAL;
 					}
-					mem_type = vram_module->v10.memory_type;
-					if (vram_type)
-						*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
-					mem_channel_number = vram_module->v10.channel_num;
-					mem_channel_width = vram_module->v10.channel_width;
-					if (vram_width)
-						*vram_width = mem_channel_number * (1 << mem_channel_width);
-					mem_vendor = (vram_module->v10.vender_rev_id) & 0xF;
-					if (vram_vendor)
-						*vram_vendor = mem_vendor;
-					break;
-				/* v25 */
-				case 5:
-					if (module_id > vram_info->v25.vram_module_num)
-						module_id = 0;
-					vram_module = (union vram_module *)vram_info->v25.vram_module;
-					while (i < module_id) {
-						vram_module = (union vram_module *)
-							((u8 *)vram_module + vram_module->v11.vram_module_size);
-						i++;
+				} else if (frev == 2) {
+					switch (crev) {
+					/* v23 */
+					case 3:
+						if (module_id > vram_info->v23.vram_module_num)
+							module_id = 0;
+						vram_module = (union vram_module *)vram_info->v23.vram_module;
+						while (i < module_id) {
+							vram_module = (union vram_module *)
+								((u8 *)vram_module + vram_module->v9.vram_module_size);
+							i++;
+						}
+						mem_type = vram_module->v9.memory_type;
+						if (vram_type)
+							*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
+						mem_channel_number = vram_module->v9.channel_num;
+						mem_channel_width = vram_module->v9.channel_width;
+						if (vram_width)
+							*vram_width = mem_channel_number * (1 << mem_channel_width);
+						mem_vendor = (vram_module->v9.vender_rev_id) & 0xF;
+						if (vram_vendor)
+							*vram_vendor = mem_vendor;
+						break;
+					/* v24 */
+					case 4:
+						if (module_id > vram_info->v24.vram_module_num)
+							module_id = 0;
+						vram_module = (union vram_module *)vram_info->v24.vram_module;
+						while (i < module_id) {
+							vram_module = (union vram_module *)
+								((u8 *)vram_module + vram_module->v10.vram_module_size);
+							i++;
+						}
+						mem_type = vram_module->v10.memory_type;
+						if (vram_type)
+							*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
+						mem_channel_number = vram_module->v10.channel_num;
+						mem_channel_width = vram_module->v10.channel_width;
+						if (vram_width)
+							*vram_width = mem_channel_number * (1 << mem_channel_width);
+						mem_vendor = (vram_module->v10.vender_rev_id) & 0xF;
+						if (vram_vendor)
+							*vram_vendor = mem_vendor;
+						break;
+					/* v25 */
+					case 5:
+						if (module_id > vram_info->v25.vram_module_num)
+							module_id = 0;
+						vram_module = (union vram_module *)vram_info->v25.vram_module;
+						while (i < module_id) {
+							vram_module = (union vram_module *)
+								((u8 *)vram_module + vram_module->v11.vram_module_size);
+							i++;
+						}
+						mem_type = vram_module->v11.memory_type;
+						if (vram_type)
+							*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
+						mem_channel_number = vram_module->v11.channel_num;
+						mem_channel_width = vram_module->v11.channel_width;
+						if (vram_width)
+							*vram_width = mem_channel_number * (1 << mem_channel_width);
+						mem_vendor = (vram_module->v11.vender_rev_id) & 0xF;
+						if (vram_vendor)
+							*vram_vendor = mem_vendor;
+						break;
+					/* v26 */
+					case 6:
+						if (module_id > vram_info->v26.vram_module_num)
+							module_id = 0;
+						vram_module = (union vram_module *)vram_info->v26.vram_module;
+						while (i < module_id) {
+							vram_module = (union vram_module *)
+								((u8 *)vram_module + vram_module->v9.vram_module_size);
+							i++;
+						}
+						mem_type = vram_module->v9.memory_type;
+						if (vram_type)
+							*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
+						mem_channel_number = vram_module->v9.channel_num;
+						mem_channel_width = vram_module->v9.channel_width;
+						if (vram_width)
+							*vram_width = mem_channel_number * (1 << mem_channel_width);
+						mem_vendor = (vram_module->v9.vender_rev_id) & 0xF;
+						if (vram_vendor)
+							*vram_vendor = mem_vendor;
+						break;
+					default:
+						return -EINVAL;
 					}
-					mem_type = vram_module->v11.memory_type;
-					if (vram_type)
-						*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
-					mem_channel_number = vram_module->v11.channel_num;
-					mem_channel_width = vram_module->v11.channel_width;
-					if (vram_width)
-						*vram_width = mem_channel_number * (1 << mem_channel_width);
-					mem_vendor = (vram_module->v11.vender_rev_id) & 0xF;
-					if (vram_vendor)
-						*vram_vendor = mem_vendor;
-					break;
-				/* v26 */
-				case 6:
-					if (module_id > vram_info->v26.vram_module_num)
-						module_id = 0;
-					vram_module = (union vram_module *)vram_info->v26.vram_module;
-					while (i < module_id) {
-						vram_module = (union vram_module *)
-							((u8 *)vram_module + vram_module->v9.vram_module_size);
-						i++;
-					}
-					mem_type = vram_module->v9.memory_type;
-					if (vram_type)
-						*vram_type = convert_atom_mem_type_to_vram_type(adev, mem_type);
-					mem_channel_number = vram_module->v9.channel_num;
-					mem_channel_width = vram_module->v9.channel_width;
-					if (vram_width)
-						*vram_width = mem_channel_number * (1 << mem_channel_width);
-					mem_vendor = (vram_module->v9.vender_rev_id) & 0xF;
-					if (vram_vendor)
-						*vram_vendor = mem_vendor;
-					break;
-				default:
+				} else {
+					/* invalid frev */
 					return -EINVAL;
 				}
-			} else {
-				/* invalid frev */
-				return -EINVAL;
 			}
 		}
-
 	}
 
 	return 0;
@@ -492,44 +555,63 @@ bool amdgpu_atomfirmware_mem_ecc_supported(struct amdgpu_device *adev)
 	u16 data_offset, size;
 	union umc_info *umc_info;
 	u8 frev, crev;
-	bool ecc_default_enabled = false;
+	bool mem_ecc_enabled = false;
 	u8 umc_config;
 	u32 umc_config1;
+	adev->ras_default_ecc_enabled = false;
 
 	index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1,
 			umc_info);
 
 	if (amdgpu_atom_parse_data_header(mode_info->atom_context,
 				index, &size, &frev, &crev, &data_offset)) {
+		umc_info = (union umc_info *)(mode_info->atom_context->bios + data_offset);
 		if (frev == 3) {
-			umc_info = (union umc_info *)
-				(mode_info->atom_context->bios + data_offset);
 			switch (crev) {
 			case 1:
 				umc_config = le32_to_cpu(umc_info->v31.umc_config);
-				ecc_default_enabled =
+				mem_ecc_enabled =
 					(umc_config & UMC_CONFIG__DEFAULT_MEM_ECC_ENABLE) ? true : false;
 				break;
 			case 2:
 				umc_config = le32_to_cpu(umc_info->v32.umc_config);
-				ecc_default_enabled =
+				mem_ecc_enabled =
 					(umc_config & UMC_CONFIG__DEFAULT_MEM_ECC_ENABLE) ? true : false;
 				break;
 			case 3:
 				umc_config = le32_to_cpu(umc_info->v33.umc_config);
 				umc_config1 = le32_to_cpu(umc_info->v33.umc_config1);
-				ecc_default_enabled =
+				mem_ecc_enabled =
 					((umc_config & UMC_CONFIG__DEFAULT_MEM_ECC_ENABLE) ||
 					 (umc_config1 & UMC_CONFIG1__ENABLE_ECC_CAPABLE)) ? true : false;
+				adev->ras_default_ecc_enabled =
+					(umc_config & UMC_CONFIG__DEFAULT_MEM_ECC_ENABLE) ? true : false;
 				break;
 			default:
 				/* unsupported crev */
 				return false;
 			}
+		} else if (frev == 4) {
+			switch (crev) {
+			case 0:
+				umc_config = le32_to_cpu(umc_info->v40.umc_config);
+				umc_config1 = le32_to_cpu(umc_info->v40.umc_config1);
+				mem_ecc_enabled =
+					(umc_config1 & UMC_CONFIG1__ENABLE_ECC_CAPABLE) ? true : false;
+				adev->ras_default_ecc_enabled =
+					(umc_config & UMC_CONFIG__DEFAULT_MEM_ECC_ENABLE) ? true : false;
+				break;
+			default:
+				/* unsupported crev */
+				return false;
+			}
+		} else {
+			/* unsupported frev */
+			return false;
 		}
 	}
 
-	return ecc_default_enabled;
+	return mem_ecc_enabled;
 }
 
 /*
@@ -590,7 +672,7 @@ bool amdgpu_atomfirmware_ras_rom_addr(struct amdgpu_device *adev,
 					  index, &size, &frev, &crev,
 					  &data_offset)) {
 		/* support firmware_info 3.4 + */
-		if ((frev == 3 && crev >=4) || (frev > 3)) {
+		if ((frev == 3 && crev >= 4) || (frev > 3)) {
 			firmware_info = (union firmware_info *)
 				(mode_info->atom_context->bios + data_offset);
 			/* The ras_rom_i2c_slave_addr should ideally
@@ -843,7 +925,7 @@ int amdgpu_atomfirmware_get_fw_reserved_fb_size(struct amdgpu_device *adev)
 
 	firmware_info = (union firmware_info *)(ctx->bios + data_offset);
 
-	if (frev !=3)
+	if (frev != 3)
 		return -EINVAL;
 
 	switch (crev) {
@@ -851,6 +933,10 @@ int amdgpu_atomfirmware_get_fw_reserved_fb_size(struct amdgpu_device *adev)
 		fw_reserved_fb_size =
 			(firmware_info->v34.fw_reserved_size_in_kb << 10);
 		break;
+	case 5:
+		fw_reserved_fb_size =
+			(firmware_info->v35.fw_reserved_size_in_kb << 10);
+		break;
 	default:
 		fw_reserved_fb_size = 0;
 		break;
@@ -902,7 +988,7 @@ int amdgpu_atomfirmware_asic_init(struct amdgpu_device *adev, bool fb_reset)
 	}
 
 	index = get_index_into_master_table(atom_master_list_of_command_functions_v2_1,
-                                            asic_init);
+					asic_init);
 	if (amdgpu_atom_parse_cmd_header(mode_info->atom_context, index, &frev, &crev)) {
 		if (frev == 2 && crev >= 1) {
 			memset(&asic_init_ps_v2_1, 0, sizeof(asic_init_ps_v2_1));
@@ -920,5 +1006,6 @@ int amdgpu_atomfirmware_asic_init(struct amdgpu_device *adev, bool fb_reset)
 		return -EINVAL;
 	}
 
-	return amdgpu_atom_execute_table(ctx, ATOM_CMD_INIT, (uint32_t *)&asic_init_ps_v2_1);
+	return amdgpu_atom_execute_table(ctx, ATOM_CMD_INIT, (uint32_t *)&asic_init_ps_v2_1,
+		sizeof(asic_init_ps_v2_1));
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h
index c7eb2caec65a..649b5530d8ae 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h
@@ -36,7 +36,7 @@ int amdgpu_atomfirmware_get_clock_info(struct amdgpu_device *adev);
 int amdgpu_atomfirmware_get_gfx_info(struct amdgpu_device *adev);
 bool amdgpu_atomfirmware_mem_ecc_supported(struct amdgpu_device *adev);
 bool amdgpu_atomfirmware_sram_ecc_supported(struct amdgpu_device *adev);
-bool amdgpu_atomfirmware_ras_rom_addr(struct amdgpu_device *adev, uint8_t* i2c_address);
+bool amdgpu_atomfirmware_ras_rom_addr(struct amdgpu_device *adev, uint8_t *i2c_address);
 bool amdgpu_atomfirmware_mem_training_supported(struct amdgpu_device *adev);
 bool amdgpu_atomfirmware_dynamic_boot_config_supported(struct amdgpu_device *adev);
 int amdgpu_atomfirmware_get_fw_reserved_fb_size(struct amdgpu_device *adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c
index d6d986be906a..3893e6fc2f03 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c
@@ -74,28 +74,21 @@ struct atpx_mux {
 	u16 mux;
 } __packed;
 
-bool amdgpu_has_atpx(void) {
+bool amdgpu_has_atpx(void)
+{
 	return amdgpu_atpx_priv.atpx_detected;
 }
 
-bool amdgpu_has_atpx_dgpu_power_cntl(void) {
+bool amdgpu_has_atpx_dgpu_power_cntl(void)
+{
 	return amdgpu_atpx_priv.atpx.functions.power_cntl;
 }
 
-bool amdgpu_is_atpx_hybrid(void) {
+bool amdgpu_is_atpx_hybrid(void)
+{
 	return amdgpu_atpx_priv.atpx.is_hybrid;
 }
 
-bool amdgpu_atpx_dgpu_req_power_for_displays(void) {
-	return amdgpu_atpx_priv.atpx.dgpu_req_power_for_displays;
-}
-
-#if defined(CONFIG_ACPI)
-void *amdgpu_atpx_get_dhandle(void) {
-	return amdgpu_atpx_priv.dhandle;
-}
-#endif
-
 /**
  * amdgpu_atpx_call - call an ATPX method
  *
@@ -134,7 +127,7 @@ static union acpi_object *amdgpu_atpx_call(acpi_handle handle, int function,
 
 	/* Fail only if calling the method fails and ATPX is supported */
 	if (ACPI_FAILURE(status) && status != AE_NOT_FOUND) {
-		printk("failed to evaluate ATPX got %s\n",
+		pr_err("failed to evaluate ATPX got %s\n",
 		       acpi_format_exception(status));
 		kfree(buffer.pointer);
 		return NULL;
@@ -190,7 +183,7 @@ static int amdgpu_atpx_validate(struct amdgpu_atpx *atpx)
 
 		size = *(u16 *) info->buffer.pointer;
 		if (size < 10) {
-			printk("ATPX buffer is too small: %zu\n", size);
+			pr_err("ATPX buffer is too small: %zu\n", size);
 			kfree(info);
 			return -EINVAL;
 		}
@@ -223,11 +216,11 @@ static int amdgpu_atpx_validate(struct amdgpu_atpx *atpx)
 	atpx->is_hybrid = false;
 	if (valid_bits & ATPX_MS_HYBRID_GFX_SUPPORTED) {
 		if (amdgpu_atpx_priv.quirks & AMDGPU_PX_QUIRK_FORCE_ATPX) {
-			printk("ATPX Hybrid Graphics, forcing to ATPX\n");
+			pr_warn("ATPX Hybrid Graphics, forcing to ATPX\n");
 			atpx->functions.power_cntl = true;
 			atpx->is_hybrid = false;
 		} else {
-			printk("ATPX Hybrid Graphics\n");
+			pr_notice("ATPX Hybrid Graphics\n");
 			/*
 			 * Disable legacy PM methods only when pcie port PM is usable,
 			 * otherwise the device might fail to power off or power on.
@@ -269,7 +262,7 @@ static int amdgpu_atpx_verify_interface(struct amdgpu_atpx *atpx)
 
 	size = *(u16 *) info->buffer.pointer;
 	if (size < 8) {
-		printk("ATPX buffer is too small: %zu\n", size);
+		pr_err("ATPX buffer is too small: %zu\n", size);
 		err = -EINVAL;
 		goto out;
 	}
@@ -278,8 +271,8 @@ static int amdgpu_atpx_verify_interface(struct amdgpu_atpx *atpx)
 	memcpy(&output, info->buffer.pointer, size);
 
 	/* TODO: check version? */
-	printk("ATPX version %u, functions 0x%08x\n",
-	       output.version, output.function_bits);
+	pr_notice("ATPX version %u, functions 0x%08x\n",
+		  output.version, output.function_bits);
 
 	amdgpu_atpx_parse_functions(&atpx->functions, output.function_bits);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
index edc6377ec5ff..199693369c7c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
@@ -39,7 +39,7 @@ static int amdgpu_benchmark_do_move(struct amdgpu_device *adev, unsigned size,
 	for (i = 0; i < n; i++) {
 		struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
 		r = amdgpu_copy_buffer(ring, saddr, daddr, size, NULL, &fence,
-				       false, false, false);
+				       false, false, 0);
 		if (r)
 			goto exit_do_move;
 		r = dma_fence_wait(fence, false);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c
index 30c28a69e847..35d04e69aec0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bios.c
@@ -29,6 +29,7 @@
 #include "amdgpu.h"
 #include "atom.h"
 
+#include <linux/device.h>
 #include <linux/pci.h>
 #include <linux/slab.h>
 #include <linux/acpi.h>
@@ -46,52 +47,63 @@
 /* Check if current bios is an ATOM BIOS.
  * Return true if it is ATOM BIOS. Otherwise, return false.
  */
-static bool check_atom_bios(uint8_t *bios, size_t size)
+static bool check_atom_bios(struct amdgpu_device *adev, size_t size)
 {
 	uint16_t tmp, bios_header_start;
+	uint8_t *bios = adev->bios;
 
 	if (!bios || size < 0x49) {
-		DRM_INFO("vbios mem is null or mem size is wrong\n");
+		dev_dbg(adev->dev, "VBIOS mem is null or mem size is wrong\n");
 		return false;
 	}
 
 	if (!AMD_IS_VALID_VBIOS(bios)) {
-		DRM_INFO("BIOS signature incorrect %x %x\n", bios[0], bios[1]);
+		dev_dbg(adev->dev, "VBIOS signature incorrect %x %x\n", bios[0],
+			bios[1]);
 		return false;
 	}
 
 	bios_header_start = bios[0x48] | (bios[0x49] << 8);
 	if (!bios_header_start) {
-		DRM_INFO("Can't locate bios header\n");
+		dev_dbg(adev->dev, "Can't locate VBIOS header\n");
 		return false;
 	}
 
 	tmp = bios_header_start + 4;
 	if (size < tmp) {
-		DRM_INFO("BIOS header is broken\n");
+		dev_dbg(adev->dev, "VBIOS header is broken\n");
 		return false;
 	}
 
 	if (!memcmp(bios + tmp, "ATOM", 4) ||
 	    !memcmp(bios + tmp, "MOTA", 4)) {
-		DRM_DEBUG("ATOMBIOS detected\n");
+		dev_dbg(adev->dev, "ATOMBIOS detected\n");
 		return true;
 	}
 
 	return false;
 }
 
+void amdgpu_bios_release(struct amdgpu_device *adev)
+{
+	kfree(adev->bios);
+	adev->bios = NULL;
+	adev->bios_size = 0;
+}
+
 /* If you boot an IGP board with a discrete card as the primary,
  * the IGP rom is not accessible via the rom bar as the IGP rom is
  * part of the system bios.  On boot, the system bios puts a
  * copy of the igp rom at the start of vram if a discrete card is
  * present.
+ * For SR-IOV, if dynamic critical region is not enabled,
+ * the vbios image is also put at the start of VRAM in the VF.
  */
-static bool igp_read_bios_from_vram(struct amdgpu_device *adev)
+static bool amdgpu_read_bios_from_vram(struct amdgpu_device *adev)
 {
-	uint8_t __iomem *bios;
+	uint8_t __iomem *bios = NULL;
 	resource_size_t vram_base;
-	resource_size_t size = 256 * 1024; /* ??? */
+	u32 size = 256U * 1024U; /* ??? */
 
 	if (!(adev->flags & AMD_IS_APU))
 		if (amdgpu_device_need_post(adev))
@@ -103,22 +115,36 @@ static bool igp_read_bios_from_vram(struct amdgpu_device *adev)
 
 	adev->bios = NULL;
 	vram_base = pci_resource_start(adev->pdev, 0);
-	bios = ioremap_wc(vram_base, size);
-	if (!bios) {
-		return false;
-	}
 
 	adev->bios = kmalloc(size, GFP_KERNEL);
-	if (!adev->bios) {
-		iounmap(bios);
+	if (!adev->bios)
 		return false;
+
+	/* For SRIOV with dynamic critical region is enabled,
+	 * the vbios image is put at a dynamic offset of VRAM in the VF.
+	 * If dynamic critical region is disabled, follow the existing logic as on baremetal.
+	 */
+	if (amdgpu_sriov_vf(adev) && adev->virt.is_dynamic_crit_regn_enabled) {
+		if (amdgpu_virt_get_dynamic_data_info(adev,
+				AMD_SRIOV_MSG_VBIOS_IMG_TABLE_ID, adev->bios, &size)) {
+			amdgpu_bios_release(adev);
+			return false;
+		}
+	} else {
+		bios = ioremap_wc(vram_base, size);
+		if (!bios) {
+			amdgpu_bios_release(adev);
+			return false;
+		}
+
+		memcpy_fromio(adev->bios, bios, size);
+		iounmap(bios);
 	}
+
 	adev->bios_size = size;
-	memcpy_fromio(adev->bios, bios, size);
-	iounmap(bios);
 
-	if (!check_atom_bios(adev->bios, size)) {
-		kfree(adev->bios);
+	if (!check_atom_bios(adev, size)) {
+		amdgpu_bios_release(adev);
 		return false;
 	}
 
@@ -133,9 +159,8 @@ bool amdgpu_read_bios(struct amdgpu_device *adev)
 	adev->bios = NULL;
 	/* XXX: some cards may return 0 for rom size? ddx has a workaround */
 	bios = pci_map_rom(adev->pdev, &size);
-	if (!bios) {
+	if (!bios)
 		return false;
-	}
 
 	adev->bios = kzalloc(size, GFP_KERNEL);
 	if (adev->bios == NULL) {
@@ -146,8 +171,8 @@ bool amdgpu_read_bios(struct amdgpu_device *adev)
 	memcpy_fromio(adev->bios, bios, size);
 	pci_unmap_rom(adev->pdev, bios);
 
-	if (!check_atom_bios(adev->bios, size)) {
-		kfree(adev->bios);
+	if (!check_atom_bios(adev, size)) {
+		amdgpu_bios_release(adev);
 		return false;
 	}
 
@@ -168,9 +193,9 @@ static bool amdgpu_read_bios_from_rom(struct amdgpu_device *adev)
 	header[AMD_VBIOS_SIGNATURE_END] = 0;
 
 	if ((!AMD_IS_VALID_VBIOS(header)) ||
-	    0 != memcmp((char *)&header[AMD_VBIOS_SIGNATURE_OFFSET],
-			AMD_VBIOS_SIGNATURE,
-			strlen(AMD_VBIOS_SIGNATURE)))
+		memcmp((char *)&header[AMD_VBIOS_SIGNATURE_OFFSET],
+		       AMD_VBIOS_SIGNATURE,
+		       strlen(AMD_VBIOS_SIGNATURE)) != 0)
 		return false;
 
 	/* valid vbios, go on */
@@ -186,8 +211,8 @@ static bool amdgpu_read_bios_from_rom(struct amdgpu_device *adev)
 	/* read complete BIOS */
 	amdgpu_asic_read_bios_from_rom(adev, adev->bios, len);
 
-	if (!check_atom_bios(adev->bios, len)) {
-		kfree(adev->bios);
+	if (!check_atom_bios(adev, len)) {
+		amdgpu_bios_release(adev);
 		return false;
 	}
 
@@ -216,14 +241,15 @@ static bool amdgpu_read_platform_bios(struct amdgpu_device *adev)
 	memcpy_fromio(adev->bios, bios, romlen);
 	iounmap(bios);
 
-	if (!check_atom_bios(adev->bios, romlen))
+	if (!check_atom_bios(adev, romlen))
 		goto free_bios;
 
 	adev->bios_size = romlen;
 
 	return true;
 free_bios:
-	kfree(adev->bios);
+	amdgpu_bios_release(adev);
+
 	return false;
 }
 
@@ -264,7 +290,7 @@ static int amdgpu_atrm_call(acpi_handle atrm_handle, uint8_t *bios,
 
 	status = acpi_evaluate_object(atrm_handle, NULL, &atrm_arg, &buffer);
 	if (ACPI_FAILURE(status)) {
-		printk("failed to evaluate ATRM got %s\n", acpi_format_exception(status));
+		DRM_ERROR("failed to evaluate ATRM got %s\n", acpi_format_exception(status));
 		return -ENODEV;
 	}
 
@@ -285,11 +311,15 @@ static bool amdgpu_atrm_get_bios(struct amdgpu_device *adev)
 	acpi_status status;
 	bool found = false;
 
-	/* ATRM is for the discrete card only */
-	if (adev->flags & AMD_IS_APU)
+	/* ATRM is for on-platform devices only */
+	if (dev_is_removable(&adev->pdev->dev))
 		return false;
 
-	while ((pdev = pci_get_class(PCI_CLASS_DISPLAY_VGA << 8, pdev)) != NULL) {
+	while ((pdev = pci_get_base_class(PCI_BASE_CLASS_DISPLAY, pdev))) {
+		if ((pdev->class != PCI_CLASS_DISPLAY_VGA << 8) &&
+		    (pdev->class != PCI_CLASS_DISPLAY_OTHER << 8))
+			continue;
+
 		dhandle = ACPI_HANDLE(&pdev->dev);
 		if (!dhandle)
 			continue;
@@ -301,20 +331,6 @@ static bool amdgpu_atrm_get_bios(struct amdgpu_device *adev)
 		}
 	}
 
-	if (!found) {
-		while ((pdev = pci_get_class(PCI_CLASS_DISPLAY_OTHER << 8, pdev)) != NULL) {
-			dhandle = ACPI_HANDLE(&pdev->dev);
-			if (!dhandle)
-				continue;
-
-			status = acpi_get_handle(dhandle, "ATRM", &atrm_handle);
-			if (ACPI_SUCCESS(status)) {
-				found = true;
-				break;
-			}
-		}
-	}
-
 	if (!found)
 		return false;
 	pci_dev_put(pdev);
@@ -334,8 +350,8 @@ static bool amdgpu_atrm_get_bios(struct amdgpu_device *adev)
 			break;
 	}
 
-	if (!check_atom_bios(adev->bios, size)) {
-		kfree(adev->bios);
+	if (!check_atom_bios(adev, size)) {
+		amdgpu_bios_release(adev);
 		return false;
 	}
 	adev->bios_size = size;
@@ -350,11 +366,8 @@ static inline bool amdgpu_atrm_get_bios(struct amdgpu_device *adev)
 
 static bool amdgpu_read_disabled_bios(struct amdgpu_device *adev)
 {
-	if (adev->flags & AMD_IS_APU)
-		return igp_read_bios_from_vram(adev);
-	else
-		return (!adev->asic_funcs || !adev->asic_funcs->read_disabled_bios) ?
-			false : amdgpu_asic_read_disabled_bios(adev);
+	return (!adev->asic_funcs || !adev->asic_funcs->read_disabled_bios) ?
+		false : amdgpu_asic_read_disabled_bios(adev);
 }
 
 #ifdef CONFIG_ACPI
@@ -363,7 +376,7 @@ static bool amdgpu_acpi_vfct_bios(struct amdgpu_device *adev)
 	struct acpi_table_header *hdr;
 	acpi_size tbl_size;
 	UEFI_ACPI_VFCT *vfct;
-	unsigned offset;
+	unsigned int offset;
 
 	if (!ACPI_SUCCESS(acpi_get_table("VFCT", 1, &hdr)))
 		return false;
@@ -402,8 +415,8 @@ static bool amdgpu_acpi_vfct_bios(struct amdgpu_device *adev)
 					     vhdr->ImageLength,
 					     GFP_KERNEL);
 
-			if (!check_atom_bios(adev->bios, vhdr->ImageLength)) {
-				kfree(adev->bios);
+			if (!check_atom_bios(adev, vhdr->ImageLength)) {
+				amdgpu_bios_release(adev);
 				return false;
 			}
 			adev->bios_size = vhdr->ImageLength;
@@ -421,7 +434,43 @@ static inline bool amdgpu_acpi_vfct_bios(struct amdgpu_device *adev)
 }
 #endif
 
-bool amdgpu_get_bios(struct amdgpu_device *adev)
+static bool amdgpu_get_bios_apu(struct amdgpu_device *adev)
+{
+	if (amdgpu_acpi_vfct_bios(adev)) {
+		dev_info(adev->dev, "Fetched VBIOS from VFCT\n");
+		goto success;
+	}
+
+	if (amdgpu_read_bios_from_vram(adev)) {
+		dev_info(adev->dev, "Fetched VBIOS from VRAM BAR\n");
+		goto success;
+	}
+
+	if (amdgpu_read_bios(adev)) {
+		dev_info(adev->dev, "Fetched VBIOS from ROM BAR\n");
+		goto success;
+	}
+
+	if (amdgpu_read_platform_bios(adev)) {
+		dev_info(adev->dev, "Fetched VBIOS from platform\n");
+		goto success;
+	}
+
+	dev_err(adev->dev, "Unable to locate a BIOS ROM\n");
+	return false;
+
+success:
+	return true;
+}
+
+static bool amdgpu_prefer_rom_resource(struct amdgpu_device *adev)
+{
+	struct resource *res = &adev->pdev->resource[PCI_ROM_RESOURCE];
+
+	return (res->flags & IORESOURCE_ROM_SHADOW);
+}
+
+static bool amdgpu_get_bios_dgpu(struct amdgpu_device *adev)
 {
 	if (amdgpu_atrm_get_bios(adev)) {
 		dev_info(adev->dev, "Fetched VBIOS from ATRM\n");
@@ -433,14 +482,33 @@ bool amdgpu_get_bios(struct amdgpu_device *adev)
 		goto success;
 	}
 
-	if (igp_read_bios_from_vram(adev)) {
+	/* this is required for SR-IOV */
+	if (amdgpu_read_bios_from_vram(adev)) {
 		dev_info(adev->dev, "Fetched VBIOS from VRAM BAR\n");
 		goto success;
 	}
 
-	if (amdgpu_read_bios(adev)) {
-		dev_info(adev->dev, "Fetched VBIOS from ROM BAR\n");
-		goto success;
+	if (amdgpu_prefer_rom_resource(adev)) {
+		if (amdgpu_read_bios(adev)) {
+			dev_info(adev->dev, "Fetched VBIOS from ROM BAR\n");
+			goto success;
+		}
+
+		if (amdgpu_read_platform_bios(adev)) {
+			dev_info(adev->dev, "Fetched VBIOS from platform\n");
+			goto success;
+		}
+
+	} else {
+		if (amdgpu_read_platform_bios(adev)) {
+			dev_info(adev->dev, "Fetched VBIOS from platform\n");
+			goto success;
+		}
+
+		if (amdgpu_read_bios(adev)) {
+			dev_info(adev->dev, "Fetched VBIOS from ROM BAR\n");
+			goto success;
+		}
 	}
 
 	if (amdgpu_read_bios_from_rom(adev)) {
@@ -453,19 +521,28 @@ bool amdgpu_get_bios(struct amdgpu_device *adev)
 		goto success;
 	}
 
-	if (amdgpu_read_platform_bios(adev)) {
-		dev_info(adev->dev, "Fetched VBIOS from platform\n");
-		goto success;
-	}
-
 	dev_err(adev->dev, "Unable to locate a BIOS ROM\n");
 	return false;
 
 success:
-	adev->is_atom_fw = (adev->asic_type >= CHIP_VEGA10) ? true : false;
 	return true;
 }
 
+bool amdgpu_get_bios(struct amdgpu_device *adev)
+{
+	bool found;
+
+	if (adev->flags & AMD_IS_APU)
+		found = amdgpu_get_bios_apu(adev);
+	else
+		found = amdgpu_get_bios_dgpu(adev);
+
+	if (found)
+		adev->is_atom_fw = adev->asic_type >= CHIP_VEGA10;
+
+	return found;
+}
+
 /* helper function for soc15 and onwards to read bios from rom */
 bool amdgpu_soc15_read_bios_from_rom(struct amdgpu_device *adev,
 				     u8 *bios, u32 length_bytes)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c
index 252a876b0725..66fb37b64388 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c
@@ -28,6 +28,7 @@
  *    Christian König <deathsimple@vodafone.de>
  */
 
+#include <linux/sort.h>
 #include <linux/uaccess.h>
 
 #include "amdgpu.h"
@@ -50,13 +51,20 @@ static void amdgpu_bo_list_free(struct kref *ref)
 						   refcount);
 	struct amdgpu_bo_list_entry *e;
 
-	amdgpu_bo_list_for_each_entry(e, list) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
+	amdgpu_bo_list_for_each_entry(e, list)
+		amdgpu_bo_unref(&e->bo);
+	call_rcu(&list->rhead, amdgpu_bo_list_free_rcu);
+}
 
-		amdgpu_bo_unref(&bo);
-	}
+static int amdgpu_bo_list_entry_cmp(const void *_a, const void *_b)
+{
+	const struct amdgpu_bo_list_entry *a = _a, *b = _b;
 
-	call_rcu(&list->rhead, amdgpu_bo_list_free_rcu);
+	if (a->priority > b->priority)
+		return 1;
+	if (a->priority < b->priority)
+		return -1;
+	return 0;
 }
 
 int amdgpu_bo_list_create(struct amdgpu_device *adev, struct drm_file *filp,
@@ -67,27 +75,17 @@ int amdgpu_bo_list_create(struct amdgpu_device *adev, struct drm_file *filp,
 	struct amdgpu_bo_list_entry *array;
 	struct amdgpu_bo_list *list;
 	uint64_t total_size = 0;
-	size_t size;
 	unsigned i;
 	int r;
 
-	if (num_entries > (SIZE_MAX - sizeof(struct amdgpu_bo_list))
-				/ sizeof(struct amdgpu_bo_list_entry))
-		return -EINVAL;
-
-	size = sizeof(struct amdgpu_bo_list);
-	size += num_entries * sizeof(struct amdgpu_bo_list_entry);
-	list = kvmalloc(size, GFP_KERNEL);
+	list = kvzalloc(struct_size(list, entries, num_entries), GFP_KERNEL);
 	if (!list)
 		return -ENOMEM;
 
 	kref_init(&list->refcount);
-	list->gds_obj = NULL;
-	list->gws_obj = NULL;
-	list->oa_obj = NULL;
 
-	array = amdgpu_bo_list_array_entry(list, 0);
-	memset(array, 0, num_entries * sizeof(struct amdgpu_bo_list_entry));
+	list->num_entries = num_entries;
+	array = list->entries;
 
 	for (i = 0; i < num_entries; ++i) {
 		struct amdgpu_bo_list_entry *entry;
@@ -118,7 +116,7 @@ int amdgpu_bo_list_create(struct amdgpu_device *adev, struct drm_file *filp,
 
 		entry->priority = min(info[i].bo_priority,
 				      AMDGPU_BO_LIST_MAX_PRIORITY);
-		entry->tv.bo = &bo->tbo;
+		entry->bo = bo;
 
 		if (bo->preferred_domains == AMDGPU_GEM_DOMAIN_GDS)
 			list->gds_obj = bo;
@@ -132,7 +130,8 @@ int amdgpu_bo_list_create(struct amdgpu_device *adev, struct drm_file *filp,
 	}
 
 	list->first_userptr = first_userptr;
-	list->num_entries = num_entries;
+	sort(array, last_entry, sizeof(struct amdgpu_bo_list_entry),
+	     amdgpu_bo_list_entry_cmp, NULL);
 
 	trace_amdgpu_cs_bo_status(list->num_entries, total_size);
 
@@ -141,16 +140,10 @@ int amdgpu_bo_list_create(struct amdgpu_device *adev, struct drm_file *filp,
 	return 0;
 
 error_free:
-	for (i = 0; i < last_entry; ++i) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(array[i].tv.bo);
-
-		amdgpu_bo_unref(&bo);
-	}
-	for (i = first_userptr; i < num_entries; ++i) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(array[i].tv.bo);
-
-		amdgpu_bo_unref(&bo);
-	}
+	for (i = 0; i < last_entry; ++i)
+		amdgpu_bo_unref(&array[i].bo);
+	for (i = first_userptr; i < num_entries; ++i)
+		amdgpu_bo_unref(&array[i].bo);
 	kvfree(list);
 	return r;
 
@@ -179,44 +172,10 @@ int amdgpu_bo_list_get(struct amdgpu_fpriv *fpriv, int id,
 	}
 
 	rcu_read_unlock();
+	*result = NULL;
 	return -ENOENT;
 }
 
-void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list,
-			     struct list_head *validated)
-{
-	/* This is based on the bucket sort with O(n) time complexity.
-	 * An item with priority "i" is added to bucket[i]. The lists are then
-	 * concatenated in descending order.
-	 */
-	struct list_head bucket[AMDGPU_BO_LIST_NUM_BUCKETS];
-	struct amdgpu_bo_list_entry *e;
-	unsigned i;
-
-	for (i = 0; i < AMDGPU_BO_LIST_NUM_BUCKETS; i++)
-		INIT_LIST_HEAD(&bucket[i]);
-
-	/* Since buffers which appear sooner in the relocation list are
-	 * likely to be used more often than buffers which appear later
-	 * in the list, the sort mustn't change the ordering of buffers
-	 * with the same priority, i.e. it must be stable.
-	 */
-	amdgpu_bo_list_for_each_entry(e, list) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
-		unsigned priority = e->priority;
-
-		if (!bo->parent)
-			list_add_tail(&e->tv.head, &bucket[priority]);
-
-		e->user_pages = NULL;
-		e->range = NULL;
-	}
-
-	/* Connect the sorted buckets in the output list. */
-	for (i = 0; i < AMDGPU_BO_LIST_NUM_BUCKETS; i++)
-		list_splice(&bucket[i], validated);
-}
-
 void amdgpu_bo_list_put(struct amdgpu_bo_list *list)
 {
 	kref_put(&list->refcount, amdgpu_bo_list_free);
@@ -225,43 +184,36 @@ void amdgpu_bo_list_put(struct amdgpu_bo_list *list)
 int amdgpu_bo_create_list_entry_array(struct drm_amdgpu_bo_list_in *in,
 				      struct drm_amdgpu_bo_list_entry **info_param)
 {
-	const void __user *uptr = u64_to_user_ptr(in->bo_info_ptr);
 	const uint32_t info_size = sizeof(struct drm_amdgpu_bo_list_entry);
+	const void __user *uptr = u64_to_user_ptr(in->bo_info_ptr);
+	const uint32_t bo_info_size = in->bo_info_size;
+	const uint32_t bo_number = in->bo_number;
 	struct drm_amdgpu_bo_list_entry *info;
-	int r;
-
-	info = kvmalloc_array(in->bo_number, info_size, GFP_KERNEL);
-	if (!info)
-		return -ENOMEM;
 
 	/* copy the handle array from userspace to a kernel buffer */
-	r = -EFAULT;
-	if (likely(info_size == in->bo_info_size)) {
-		unsigned long bytes = in->bo_number *
-			in->bo_info_size;
-
-		if (copy_from_user(info, uptr, bytes))
-			goto error_free;
-
+	if (likely(info_size == bo_info_size)) {
+		info = vmemdup_array_user(uptr, bo_number, info_size);
+		if (IS_ERR(info))
+			return PTR_ERR(info);
 	} else {
-		unsigned long bytes = min(in->bo_info_size, info_size);
+		const uint32_t bytes = min(bo_info_size, info_size);
 		unsigned i;
 
-		memset(info, 0, in->bo_number * info_size);
-		for (i = 0; i < in->bo_number; ++i) {
-			if (copy_from_user(&info[i], uptr, bytes))
-				goto error_free;
+		info = kvmalloc_array(bo_number, info_size, GFP_KERNEL);
+		if (!info)
+			return -ENOMEM;
 
-			uptr += in->bo_info_size;
+		memset(info, 0, bo_number * info_size);
+		for (i = 0; i < bo_number; ++i, uptr += bo_info_size) {
+			if (copy_from_user(&info[i], uptr, bytes)) {
+				kvfree(info);
+				return -EFAULT;
+			}
 		}
 	}
 
 	*info_param = info;
 	return 0;
-
-error_free:
-	kvfree(info);
-	return r;
 }
 
 int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h
index ededdc01ca28..2b5e7c46a39d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h
@@ -23,7 +23,6 @@
 #ifndef __AMDGPU_BO_LIST_H__
 #define __AMDGPU_BO_LIST_H__
 
-#include <drm/ttm/ttm_execbuf_util.h>
 #include <drm/amdgpu_drm.h>
 
 struct hmm_range;
@@ -36,11 +35,10 @@ struct amdgpu_bo_va;
 struct amdgpu_fpriv;
 
 struct amdgpu_bo_list_entry {
-	struct ttm_validate_buffer	tv;
+	struct amdgpu_bo		*bo;
 	struct amdgpu_bo_va		*bo_va;
 	uint32_t			priority;
-	struct page			**user_pages;
-	struct hmm_range		*range;
+	struct amdgpu_hmm_range		*range;
 	bool				user_invalidated;
 };
 
@@ -56,12 +54,12 @@ struct amdgpu_bo_list {
 	/* Protect access during command submission.
 	 */
 	struct mutex bo_list_mutex;
+
+	struct amdgpu_bo_list_entry entries[] __counted_by(num_entries);
 };
 
 int amdgpu_bo_list_get(struct amdgpu_fpriv *fpriv, int id,
 		       struct amdgpu_bo_list **result);
-void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list,
-			     struct list_head *validated);
 void amdgpu_bo_list_put(struct amdgpu_bo_list *list);
 int amdgpu_bo_create_list_entry_array(struct drm_amdgpu_bo_list_in *in,
 				      struct drm_amdgpu_bo_list_entry **info_param);
@@ -72,22 +70,14 @@ int amdgpu_bo_list_create(struct amdgpu_device *adev,
 				 size_t num_entries,
 				 struct amdgpu_bo_list **list);
 
-static inline struct amdgpu_bo_list_entry *
-amdgpu_bo_list_array_entry(struct amdgpu_bo_list *list, unsigned index)
-{
-	struct amdgpu_bo_list_entry *array = (void *)&list[1];
-
-	return &array[index];
-}
-
 #define amdgpu_bo_list_for_each_entry(e, list) \
-	for (e = amdgpu_bo_list_array_entry(list, 0); \
-	     e != amdgpu_bo_list_array_entry(list, (list)->num_entries); \
+	for (e = list->entries; \
+	     e != &list->entries[list->num_entries]; \
 	     ++e)
 
 #define amdgpu_bo_list_for_each_userptr_entry(e, list) \
-	for (e = amdgpu_bo_list_array_entry(list, (list)->first_userptr); \
-	     e != amdgpu_bo_list_array_entry(list, (list)->num_entries); \
+	for (e = &list->entries[list->first_userptr]; \
+	     e != &list->entries[list->num_entries]; \
 	     ++e)
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c
index 456e385333b6..004a6a9d6b9f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c
@@ -41,13 +41,13 @@ struct amdgpu_cgs_device {
 		((struct amdgpu_cgs_device *)cgs_device)->adev
 
 
-static uint32_t amdgpu_cgs_read_register(struct cgs_device *cgs_device, unsigned offset)
+static uint32_t amdgpu_cgs_read_register(struct cgs_device *cgs_device, unsigned int offset)
 {
 	CGS_FUNC_ADEV;
 	return RREG32(offset);
 }
 
-static void amdgpu_cgs_write_register(struct cgs_device *cgs_device, unsigned offset,
+static void amdgpu_cgs_write_register(struct cgs_device *cgs_device, unsigned int offset,
 				      uint32_t value)
 {
 	CGS_FUNC_ADEV;
@@ -56,7 +56,7 @@ static void amdgpu_cgs_write_register(struct cgs_device *cgs_device, unsigned of
 
 static uint32_t amdgpu_cgs_read_ind_register(struct cgs_device *cgs_device,
 					     enum cgs_ind_reg space,
-					     unsigned index)
+					     unsigned int index)
 {
 	CGS_FUNC_ADEV;
 	switch (space) {
@@ -84,7 +84,7 @@ static uint32_t amdgpu_cgs_read_ind_register(struct cgs_device *cgs_device,
 
 static void amdgpu_cgs_write_ind_register(struct cgs_device *cgs_device,
 					  enum cgs_ind_reg space,
-					  unsigned index, uint32_t value)
+					  unsigned int index, uint32_t value)
 {
 	CGS_FUNC_ADEV;
 	switch (space) {
@@ -163,38 +163,38 @@ static uint16_t amdgpu_get_firmware_version(struct cgs_device *cgs_device,
 	uint16_t fw_version = 0;
 
 	switch (type) {
-		case CGS_UCODE_ID_SDMA0:
-			fw_version = adev->sdma.instance[0].fw_version;
-			break;
-		case CGS_UCODE_ID_SDMA1:
-			fw_version = adev->sdma.instance[1].fw_version;
-			break;
-		case CGS_UCODE_ID_CP_CE:
-			fw_version = adev->gfx.ce_fw_version;
-			break;
-		case CGS_UCODE_ID_CP_PFP:
-			fw_version = adev->gfx.pfp_fw_version;
-			break;
-		case CGS_UCODE_ID_CP_ME:
-			fw_version = adev->gfx.me_fw_version;
-			break;
-		case CGS_UCODE_ID_CP_MEC:
-			fw_version = adev->gfx.mec_fw_version;
-			break;
-		case CGS_UCODE_ID_CP_MEC_JT1:
-			fw_version = adev->gfx.mec_fw_version;
-			break;
-		case CGS_UCODE_ID_CP_MEC_JT2:
-			fw_version = adev->gfx.mec_fw_version;
-			break;
-		case CGS_UCODE_ID_RLC_G:
-			fw_version = adev->gfx.rlc_fw_version;
-			break;
-		case CGS_UCODE_ID_STORAGE:
-			break;
-		default:
-			DRM_ERROR("firmware type %d do not have version\n", type);
-			break;
+	case CGS_UCODE_ID_SDMA0:
+		fw_version = adev->sdma.instance[0].fw_version;
+		break;
+	case CGS_UCODE_ID_SDMA1:
+		fw_version = adev->sdma.instance[1].fw_version;
+		break;
+	case CGS_UCODE_ID_CP_CE:
+		fw_version = adev->gfx.ce_fw_version;
+		break;
+	case CGS_UCODE_ID_CP_PFP:
+		fw_version = adev->gfx.pfp_fw_version;
+		break;
+	case CGS_UCODE_ID_CP_ME:
+		fw_version = adev->gfx.me_fw_version;
+		break;
+	case CGS_UCODE_ID_CP_MEC:
+		fw_version = adev->gfx.mec_fw_version;
+		break;
+	case CGS_UCODE_ID_CP_MEC_JT1:
+		fw_version = adev->gfx.mec_fw_version;
+		break;
+	case CGS_UCODE_ID_CP_MEC_JT2:
+		fw_version = adev->gfx.mec_fw_version;
+		break;
+	case CGS_UCODE_ID_RLC_G:
+		fw_version = adev->gfx.rlc_fw_version;
+		break;
+	case CGS_UCODE_ID_STORAGE:
+		break;
+	default:
+		DRM_ERROR("firmware type %d do not have version\n", type);
+		break;
 	}
 	return fw_version;
 }
@@ -205,7 +205,7 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
 {
 	CGS_FUNC_ADEV;
 
-	if ((CGS_UCODE_ID_SMU != type) && (CGS_UCODE_ID_SMU_SK != type)) {
+	if (type != CGS_UCODE_ID_SMU && type != CGS_UCODE_ID_SMU_SK) {
 		uint64_t gpu_addr;
 		uint32_t data_size;
 		const struct gfx_firmware_header_v1_0 *header;
@@ -213,6 +213,9 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
 		struct amdgpu_firmware_info *ucode;
 
 		id = fw_type_convert(cgs_device, type);
+		if (id >= AMDGPU_UCODE_ID_MAXIMUM)
+			return -EINVAL;
+
 		ucode = &adev->firmware.ucode[id];
 		if (ucode->fw == NULL)
 			return -EINVAL;
@@ -232,7 +235,7 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
 		info->mc_addr = gpu_addr;
 		info->version = (uint16_t)le32_to_cpu(header->header.ucode_version);
 
-		if (CGS_UCODE_ID_CP_MEC == type)
+		if (type == CGS_UCODE_ID_CP_MEC)
 			info->image_size = le32_to_cpu(header->jt_offset) << 2;
 
 		info->fw_version = amdgpu_get_firmware_version(cgs_device, type);
@@ -249,83 +252,22 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
 
 		if (!adev->pm.fw) {
 			switch (adev->asic_type) {
-			case CHIP_TAHITI:
-				strcpy(fw_name, "radeon/tahiti_smc.bin");
-				break;
-			case CHIP_PITCAIRN:
-				if ((adev->pdev->revision == 0x81) &&
-				    ((adev->pdev->device == 0x6810) ||
-				    (adev->pdev->device == 0x6811))) {
-					info->is_kicker = true;
-					strcpy(fw_name, "radeon/pitcairn_k_smc.bin");
-				} else {
-					strcpy(fw_name, "radeon/pitcairn_smc.bin");
-				}
-				break;
-			case CHIP_VERDE:
-				if (((adev->pdev->device == 0x6820) &&
-					((adev->pdev->revision == 0x81) ||
-					(adev->pdev->revision == 0x83))) ||
-				    ((adev->pdev->device == 0x6821) &&
-					((adev->pdev->revision == 0x83) ||
-					(adev->pdev->revision == 0x87))) ||
-				    ((adev->pdev->revision == 0x87) &&
-					((adev->pdev->device == 0x6823) ||
-					(adev->pdev->device == 0x682b)))) {
-					info->is_kicker = true;
-					strcpy(fw_name, "radeon/verde_k_smc.bin");
-				} else {
-					strcpy(fw_name, "radeon/verde_smc.bin");
-				}
-				break;
-			case CHIP_OLAND:
-				if (((adev->pdev->revision == 0x81) &&
-					((adev->pdev->device == 0x6600) ||
-					(adev->pdev->device == 0x6604) ||
-					(adev->pdev->device == 0x6605) ||
-					(adev->pdev->device == 0x6610))) ||
-				    ((adev->pdev->revision == 0x83) &&
-					(adev->pdev->device == 0x6610))) {
-					info->is_kicker = true;
-					strcpy(fw_name, "radeon/oland_k_smc.bin");
-				} else {
-					strcpy(fw_name, "radeon/oland_smc.bin");
-				}
-				break;
-			case CHIP_HAINAN:
-				if (((adev->pdev->revision == 0x81) &&
-					(adev->pdev->device == 0x6660)) ||
-				    ((adev->pdev->revision == 0x83) &&
-					((adev->pdev->device == 0x6660) ||
-					(adev->pdev->device == 0x6663) ||
-					(adev->pdev->device == 0x6665) ||
-					 (adev->pdev->device == 0x6667)))) {
-					info->is_kicker = true;
-					strcpy(fw_name, "radeon/hainan_k_smc.bin");
-				} else if ((adev->pdev->revision == 0xc3) &&
-					 (adev->pdev->device == 0x6665)) {
-					info->is_kicker = true;
-					strcpy(fw_name, "radeon/banks_k_2_smc.bin");
-				} else {
-					strcpy(fw_name, "radeon/hainan_smc.bin");
-				}
-				break;
 			case CHIP_BONAIRE:
 				if ((adev->pdev->revision == 0x80) ||
 					(adev->pdev->revision == 0x81) ||
 					(adev->pdev->device == 0x665f)) {
 					info->is_kicker = true;
-					strcpy(fw_name, "amdgpu/bonaire_k_smc.bin");
+					strscpy(fw_name, "amdgpu/bonaire_k_smc.bin");
 				} else {
-					strcpy(fw_name, "amdgpu/bonaire_smc.bin");
+					strscpy(fw_name, "amdgpu/bonaire_smc.bin");
 				}
 				break;
 			case CHIP_HAWAII:
 				if (adev->pdev->revision == 0x80) {
 					info->is_kicker = true;
-					strcpy(fw_name, "amdgpu/hawaii_k_smc.bin");
+					strscpy(fw_name, "amdgpu/hawaii_k_smc.bin");
 				} else {
-					strcpy(fw_name, "amdgpu/hawaii_smc.bin");
+					strscpy(fw_name, "amdgpu/hawaii_smc.bin");
 				}
 				break;
 			case CHIP_TOPAZ:
@@ -335,83 +277,85 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device,
 				    ((adev->pdev->device == 0x6900) && (adev->pdev->revision == 0xD1)) ||
 				    ((adev->pdev->device == 0x6900) && (adev->pdev->revision == 0xD3))) {
 					info->is_kicker = true;
-					strcpy(fw_name, "amdgpu/topaz_k_smc.bin");
+					strscpy(fw_name, "amdgpu/topaz_k_smc.bin");
 				} else
-					strcpy(fw_name, "amdgpu/topaz_smc.bin");
+					strscpy(fw_name, "amdgpu/topaz_smc.bin");
 				break;
 			case CHIP_TONGA:
 				if (((adev->pdev->device == 0x6939) && (adev->pdev->revision == 0xf1)) ||
 				    ((adev->pdev->device == 0x6938) && (adev->pdev->revision == 0xf1))) {
 					info->is_kicker = true;
-					strcpy(fw_name, "amdgpu/tonga_k_smc.bin");
+					strscpy(fw_name, "amdgpu/tonga_k_smc.bin");
 				} else
-					strcpy(fw_name, "amdgpu/tonga_smc.bin");
+					strscpy(fw_name, "amdgpu/tonga_smc.bin");
 				break;
 			case CHIP_FIJI:
-				strcpy(fw_name, "amdgpu/fiji_smc.bin");
+				strscpy(fw_name, "amdgpu/fiji_smc.bin");
 				break;
 			case CHIP_POLARIS11:
 				if (type == CGS_UCODE_ID_SMU) {
 					if (ASICID_IS_P21(adev->pdev->device, adev->pdev->revision)) {
 						info->is_kicker = true;
-						strcpy(fw_name, "amdgpu/polaris11_k_smc.bin");
+						strscpy(fw_name, "amdgpu/polaris11_k_smc.bin");
 					} else if (ASICID_IS_P31(adev->pdev->device, adev->pdev->revision)) {
 						info->is_kicker = true;
-						strcpy(fw_name, "amdgpu/polaris11_k2_smc.bin");
+						strscpy(fw_name, "amdgpu/polaris11_k2_smc.bin");
 					} else {
-						strcpy(fw_name, "amdgpu/polaris11_smc.bin");
+						strscpy(fw_name, "amdgpu/polaris11_smc.bin");
 					}
 				} else if (type == CGS_UCODE_ID_SMU_SK) {
-					strcpy(fw_name, "amdgpu/polaris11_smc_sk.bin");
+					strscpy(fw_name, "amdgpu/polaris11_smc_sk.bin");
 				}
 				break;
 			case CHIP_POLARIS10:
 				if (type == CGS_UCODE_ID_SMU) {
 					if (ASICID_IS_P20(adev->pdev->device, adev->pdev->revision)) {
 						info->is_kicker = true;
-						strcpy(fw_name, "amdgpu/polaris10_k_smc.bin");
+						strscpy(fw_name, "amdgpu/polaris10_k_smc.bin");
 					} else if (ASICID_IS_P30(adev->pdev->device, adev->pdev->revision)) {
 						info->is_kicker = true;
-						strcpy(fw_name, "amdgpu/polaris10_k2_smc.bin");
+						strscpy(fw_name, "amdgpu/polaris10_k2_smc.bin");
 					} else {
-						strcpy(fw_name, "amdgpu/polaris10_smc.bin");
+						strscpy(fw_name, "amdgpu/polaris10_smc.bin");
 					}
 				} else if (type == CGS_UCODE_ID_SMU_SK) {
-					strcpy(fw_name, "amdgpu/polaris10_smc_sk.bin");
+					strscpy(fw_name, "amdgpu/polaris10_smc_sk.bin");
 				}
 				break;
 			case CHIP_POLARIS12:
 				if (ASICID_IS_P23(adev->pdev->device, adev->pdev->revision)) {
 					info->is_kicker = true;
-					strcpy(fw_name, "amdgpu/polaris12_k_smc.bin");
+					strscpy(fw_name, "amdgpu/polaris12_k_smc.bin");
 				} else {
-					strcpy(fw_name, "amdgpu/polaris12_smc.bin");
+					strscpy(fw_name, "amdgpu/polaris12_smc.bin");
 				}
 				break;
 			case CHIP_VEGAM:
-				strcpy(fw_name, "amdgpu/vegam_smc.bin");
+				strscpy(fw_name, "amdgpu/vegam_smc.bin");
 				break;
 			case CHIP_VEGA10:
 				if ((adev->pdev->device == 0x687f) &&
 					((adev->pdev->revision == 0xc0) ||
 					(adev->pdev->revision == 0xc1) ||
 					(adev->pdev->revision == 0xc3)))
-					strcpy(fw_name, "amdgpu/vega10_acg_smc.bin");
+					strscpy(fw_name, "amdgpu/vega10_acg_smc.bin");
 				else
-					strcpy(fw_name, "amdgpu/vega10_smc.bin");
+					strscpy(fw_name, "amdgpu/vega10_smc.bin");
 				break;
 			case CHIP_VEGA12:
-				strcpy(fw_name, "amdgpu/vega12_smc.bin");
+				strscpy(fw_name, "amdgpu/vega12_smc.bin");
 				break;
 			case CHIP_VEGA20:
-				strcpy(fw_name, "amdgpu/vega20_smc.bin");
+				strscpy(fw_name, "amdgpu/vega20_smc.bin");
 				break;
 			default:
 				DRM_ERROR("SMC firmware not supported\n");
 				return -EINVAL;
 			}
 
-			err = amdgpu_ucode_request(adev, &adev->pm.fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->pm.fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "%s", fw_name);
 			if (err) {
 				DRM_ERROR("Failed to load firmware \"%s\"", fw_name);
 				amdgpu_ucode_release(&adev->pm.fw);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c
index 6be30dcb029d..9f96d568acf2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c
@@ -103,7 +103,7 @@ int amdgpu_connector_get_monitor_bpc(struct drm_connector *connector)
 	struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
 	struct amdgpu_connector_atom_dig *dig_connector;
 	int bpc = 8;
-	unsigned mode_clock, max_tmds_clock;
+	unsigned int mode_clock, max_tmds_clock;
 
 	switch (connector->connector_type) {
 	case DRM_MODE_CONNECTOR_DVII:
@@ -246,36 +246,10 @@ amdgpu_connector_find_encoder(struct drm_connector *connector,
 	return NULL;
 }
 
-struct edid *amdgpu_connector_edid(struct drm_connector *connector)
-{
-	struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
-	struct drm_property_blob *edid_blob = connector->edid_blob_ptr;
-
-	if (amdgpu_connector->edid) {
-		return amdgpu_connector->edid;
-	} else if (edid_blob) {
-		struct edid *edid = kmemdup(edid_blob->data, edid_blob->length, GFP_KERNEL);
-		if (edid)
-			amdgpu_connector->edid = edid;
-	}
-	return amdgpu_connector->edid;
-}
-
 static struct edid *
 amdgpu_connector_get_hardcoded_edid(struct amdgpu_device *adev)
 {
-	struct edid *edid;
-
-	if (adev->mode_info.bios_hardcoded_edid) {
-		edid = kmalloc(adev->mode_info.bios_hardcoded_edid_size, GFP_KERNEL);
-		if (edid) {
-			memcpy((unsigned char *)edid,
-			       (unsigned char *)adev->mode_info.bios_hardcoded_edid,
-			       adev->mode_info.bios_hardcoded_edid_size);
-			return edid;
-		}
-	}
-	return NULL;
+	return drm_edid_duplicate(drm_edid_raw(adev->mode_info.bios_hardcoded_edid));
 }
 
 static void amdgpu_connector_get_edid(struct drm_connector *connector)
@@ -424,30 +398,28 @@ static void amdgpu_connector_add_common_modes(struct drm_encoder *encoder,
 	struct drm_display_mode *mode = NULL;
 	struct drm_display_mode *native_mode = &amdgpu_encoder->native_mode;
 	int i;
-	static const struct mode_size {
+	int n;
+	struct mode_size {
+		char name[DRM_DISPLAY_MODE_LEN];
 		int w;
 		int h;
-	} common_modes[17] = {
-		{ 640,  480},
-		{ 720,  480},
-		{ 800,  600},
-		{ 848,  480},
-		{1024,  768},
-		{1152,  768},
-		{1280,  720},
-		{1280,  800},
-		{1280,  854},
-		{1280,  960},
-		{1280, 1024},
-		{1440,  900},
-		{1400, 1050},
-		{1680, 1050},
-		{1600, 1200},
-		{1920, 1080},
-		{1920, 1200}
+	} common_modes[] = {
+		{  "640x480",  640,  480},
+		{  "800x600",  800,  600},
+		{ "1024x768", 1024,  768},
+		{ "1280x720", 1280,  720},
+		{ "1280x800", 1280,  800},
+		{"1280x1024", 1280, 1024},
+		{ "1440x900", 1440,  900},
+		{"1680x1050", 1680, 1050},
+		{"1600x1200", 1600, 1200},
+		{"1920x1080", 1920, 1080},
+		{"1920x1200", 1920, 1200}
 	};
 
-	for (i = 0; i < 17; i++) {
+	n = ARRAY_SIZE(common_modes);
+
+	for (i = 0; i < n; i++) {
 		if (amdgpu_encoder->devices & (ATOM_DEVICE_TV_SUPPORT)) {
 			if (common_modes[i].w > 1024 ||
 			    common_modes[i].h > 768)
@@ -460,10 +432,12 @@ static void amdgpu_connector_add_common_modes(struct drm_encoder *encoder,
 			     common_modes[i].h == native_mode->vdisplay))
 				continue;
 		}
-		if (common_modes[i].w < 320 || common_modes[i].h < 200)
-			continue;
 
 		mode = drm_cvt_mode(dev, common_modes[i].w, common_modes[i].h, 60, false, false, false);
+		if (!mode)
+			return;
+		strscpy(mode->name, common_modes[i].name, DRM_DISPLAY_MODE_LEN);
+
 		drm_mode_probed_add(connector, mode);
 	}
 }
@@ -588,16 +562,26 @@ static int amdgpu_connector_set_property(struct drm_connector *connector,
 			amdgpu_encoder = to_amdgpu_encoder(connector->encoder);
 		} else {
 			const struct drm_connector_helper_funcs *connector_funcs = connector->helper_private;
+
 			amdgpu_encoder = to_amdgpu_encoder(connector_funcs->best_encoder(connector));
 		}
 
 		switch (val) {
 		default:
-		case DRM_MODE_SCALE_NONE: rmx_type = RMX_OFF; break;
-		case DRM_MODE_SCALE_CENTER: rmx_type = RMX_CENTER; break;
-		case DRM_MODE_SCALE_ASPECT: rmx_type = RMX_ASPECT; break;
-		case DRM_MODE_SCALE_FULLSCREEN: rmx_type = RMX_FULL; break;
+		case DRM_MODE_SCALE_NONE:
+			rmx_type = RMX_OFF;
+			break;
+		case DRM_MODE_SCALE_CENTER:
+			rmx_type = RMX_CENTER;
+			break;
+		case DRM_MODE_SCALE_ASPECT:
+			rmx_type = RMX_ASPECT;
+			break;
+		case DRM_MODE_SCALE_FULLSCREEN:
+			rmx_type = RMX_FULL;
+			break;
 		}
+
 		if (amdgpu_encoder->rmx_type == rmx_type)
 			return 0;
 
@@ -687,7 +671,7 @@ static int amdgpu_connector_lvds_get_modes(struct drm_connector *connector)
 }
 
 static enum drm_mode_status amdgpu_connector_lvds_mode_valid(struct drm_connector *connector,
-					     struct drm_display_mode *mode)
+					     const struct drm_display_mode *mode)
 {
 	struct drm_encoder *encoder = amdgpu_connector_best_single_encoder(connector);
 
@@ -750,10 +734,8 @@ amdgpu_connector_lvds_detect(struct drm_connector *connector, bool force)
 
 	amdgpu_connector_update_scratch_regs(connector, ret);
 
-	if (!drm_kms_helper_is_poll_worker()) {
-		pm_runtime_mark_last_busy(connector->dev->dev);
+	if (!drm_kms_helper_is_poll_worker())
 		pm_runtime_put_autosuspend(connector->dev->dev);
-	}
 
 	return ret;
 }
@@ -795,16 +777,26 @@ static int amdgpu_connector_set_lcd_property(struct drm_connector *connector,
 		amdgpu_encoder = to_amdgpu_encoder(connector->encoder);
 	else {
 		const struct drm_connector_helper_funcs *connector_funcs = connector->helper_private;
+
 		amdgpu_encoder = to_amdgpu_encoder(connector_funcs->best_encoder(connector));
 	}
 
 	switch (value) {
-	case DRM_MODE_SCALE_NONE: rmx_type = RMX_OFF; break;
-	case DRM_MODE_SCALE_CENTER: rmx_type = RMX_CENTER; break;
-	case DRM_MODE_SCALE_ASPECT: rmx_type = RMX_ASPECT; break;
+	case DRM_MODE_SCALE_NONE:
+		rmx_type = RMX_OFF;
+		break;
+	case DRM_MODE_SCALE_CENTER:
+		rmx_type = RMX_CENTER;
+		break;
+	case DRM_MODE_SCALE_ASPECT:
+		rmx_type = RMX_ASPECT;
+		break;
 	default:
-	case DRM_MODE_SCALE_FULLSCREEN: rmx_type = RMX_FULL; break;
+	case DRM_MODE_SCALE_FULLSCREEN:
+		rmx_type = RMX_FULL;
+		break;
 	}
+
 	if (amdgpu_encoder->rmx_type == rmx_type)
 		return 0;
 
@@ -842,7 +834,7 @@ static int amdgpu_connector_vga_get_modes(struct drm_connector *connector)
 }
 
 static enum drm_mode_status amdgpu_connector_vga_mode_valid(struct drm_connector *connector,
-					    struct drm_display_mode *mode)
+					    const struct drm_display_mode *mode)
 {
 	struct drm_device *dev = connector->dev;
 	struct amdgpu_device *adev = drm_to_adev(dev);
@@ -925,10 +917,8 @@ amdgpu_connector_vga_detect(struct drm_connector *connector, bool force)
 	amdgpu_connector_update_scratch_regs(connector, ret);
 
 out:
-	if (!drm_kms_helper_is_poll_worker()) {
-		pm_runtime_mark_last_busy(connector->dev->dev);
+	if (!drm_kms_helper_is_poll_worker())
 		pm_runtime_put_autosuspend(connector->dev->dev);
-	}
 
 	return ret;
 }
@@ -968,6 +958,41 @@ amdgpu_connector_check_hpd_status_unchanged(struct drm_connector *connector)
 	return false;
 }
 
+static void amdgpu_connector_shared_ddc(enum drm_connector_status *status,
+					struct drm_connector *connector,
+					struct amdgpu_connector *amdgpu_connector)
+{
+	struct drm_connector *list_connector;
+	struct drm_connector_list_iter iter;
+	struct amdgpu_connector *list_amdgpu_connector;
+	struct drm_device *dev = connector->dev;
+	struct amdgpu_device *adev = drm_to_adev(dev);
+
+	if (amdgpu_connector->shared_ddc && *status == connector_status_connected) {
+		drm_connector_list_iter_begin(dev, &iter);
+		drm_for_each_connector_iter(list_connector,
+					    &iter) {
+			if (connector == list_connector)
+				continue;
+			list_amdgpu_connector = to_amdgpu_connector(list_connector);
+			if (list_amdgpu_connector->shared_ddc &&
+			    list_amdgpu_connector->ddc_bus->rec.i2c_id ==
+			     amdgpu_connector->ddc_bus->rec.i2c_id) {
+				/* cases where both connectors are digital */
+				if (list_connector->connector_type != DRM_MODE_CONNECTOR_VGA) {
+					/* hpd is our only option in this case */
+					if (!amdgpu_display_hpd_sense(adev,
+								      amdgpu_connector->hpd.hpd)) {
+						amdgpu_connector_free_edid(connector);
+						*status = connector_status_disconnected;
+					}
+				}
+			}
+		}
+		drm_connector_list_iter_end(&iter);
+	}
+}
+
 /*
  * DVI is complicated
  * Do a DDC probe, if DDC probe passes, get the full EDID so
@@ -1054,32 +1079,7 @@ amdgpu_connector_dvi_detect(struct drm_connector *connector, bool force)
 			 * DDC line.  The latter is more complex because with DVI<->HDMI adapters
 			 * you don't really know what's connected to which port as both are digital.
 			 */
-			if (amdgpu_connector->shared_ddc && (ret == connector_status_connected)) {
-				struct drm_connector *list_connector;
-				struct drm_connector_list_iter iter;
-				struct amdgpu_connector *list_amdgpu_connector;
-
-				drm_connector_list_iter_begin(dev, &iter);
-				drm_for_each_connector_iter(list_connector,
-							    &iter) {
-					if (connector == list_connector)
-						continue;
-					list_amdgpu_connector = to_amdgpu_connector(list_connector);
-					if (list_amdgpu_connector->shared_ddc &&
-					    (list_amdgpu_connector->ddc_bus->rec.i2c_id ==
-					     amdgpu_connector->ddc_bus->rec.i2c_id)) {
-						/* cases where both connectors are digital */
-						if (list_connector->connector_type != DRM_MODE_CONNECTOR_VGA) {
-							/* hpd is our only option in this case */
-							if (!amdgpu_display_hpd_sense(adev, amdgpu_connector->hpd.hpd)) {
-								amdgpu_connector_free_edid(connector);
-								ret = connector_status_disconnected;
-							}
-						}
-					}
-				}
-				drm_connector_list_iter_end(&iter);
-			}
+			amdgpu_connector_shared_ddc(&ret, connector, amdgpu_connector);
 		}
 	}
 
@@ -1127,7 +1127,8 @@ amdgpu_connector_dvi_detect(struct drm_connector *connector, bool force)
 					/* assume digital unless load detected otherwise */
 					amdgpu_connector->use_digital = true;
 					lret = encoder_funcs->detect(encoder, connector);
-					DRM_DEBUG_KMS("load_detect %x returned: %x\n",encoder->encoder_type,lret);
+					DRM_DEBUG_KMS("load_detect %x returned: %x\n",
+						      encoder->encoder_type, lret);
 					if (lret == connector_status_connected)
 						amdgpu_connector->use_digital = false;
 				}
@@ -1141,10 +1142,8 @@ out:
 	amdgpu_connector_update_scratch_regs(connector, ret);
 
 exit:
-	if (!drm_kms_helper_is_poll_worker()) {
-		pm_runtime_mark_last_busy(connector->dev->dev);
+	if (!drm_kms_helper_is_poll_worker())
 		pm_runtime_put_autosuspend(connector->dev->dev);
-	}
 
 	return ret;
 }
@@ -1180,35 +1179,76 @@ amdgpu_connector_dvi_encoder(struct drm_connector *connector)
 static void amdgpu_connector_dvi_force(struct drm_connector *connector)
 {
 	struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
+
 	if (connector->force == DRM_FORCE_ON)
 		amdgpu_connector->use_digital = false;
 	if (connector->force == DRM_FORCE_ON_DIGITAL)
 		amdgpu_connector->use_digital = true;
 }
 
+/**
+ * amdgpu_max_hdmi_pixel_clock - Return max supported HDMI (TMDS) pixel clock
+ * @adev: pointer to amdgpu_device
+ *
+ * Return: maximum supported HDMI (TMDS) pixel clock in KHz.
+ */
+static int amdgpu_max_hdmi_pixel_clock(const struct amdgpu_device *adev)
+{
+	if (adev->asic_type >= CHIP_POLARIS10)
+		return 600000;
+	else if (adev->asic_type >= CHIP_TONGA)
+		return 300000;
+	else
+		return 297000;
+}
+
+/**
+ * amdgpu_connector_dvi_mode_valid - Validate a mode on DVI/HDMI connectors
+ * @connector: DRM connector to validate the mode on
+ * @mode: display mode to validate
+ *
+ * Validate the given display mode on DVI and HDMI connectors, including
+ * analog signals on DVI-I.
+ *
+ * Return: drm_mode_status indicating whether the mode is valid.
+ */
 static enum drm_mode_status amdgpu_connector_dvi_mode_valid(struct drm_connector *connector,
-					    struct drm_display_mode *mode)
+					    const struct drm_display_mode *mode)
 {
 	struct drm_device *dev = connector->dev;
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
+	const int max_hdmi_pixel_clock = amdgpu_max_hdmi_pixel_clock(adev);
+	const int max_dvi_single_link_pixel_clock = 165000;
+	int max_digital_pixel_clock_khz;
 
 	/* XXX check mode bandwidth */
 
-	if (amdgpu_connector->use_digital && (mode->clock > 165000)) {
-		if ((amdgpu_connector->connector_object_id == CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_I) ||
-		    (amdgpu_connector->connector_object_id == CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D) ||
-		    (amdgpu_connector->connector_object_id == CONNECTOR_OBJECT_ID_HDMI_TYPE_B)) {
-			return MODE_OK;
-		} else if (connector->display_info.is_hdmi) {
-			/* HDMI 1.3+ supports max clock of 340 Mhz */
-			if (mode->clock > 340000)
-				return MODE_CLOCK_HIGH;
-			else
-				return MODE_OK;
-		} else {
-			return MODE_CLOCK_HIGH;
+	if (amdgpu_connector->use_digital) {
+		switch (amdgpu_connector->connector_object_id) {
+		case CONNECTOR_OBJECT_ID_HDMI_TYPE_A:
+			max_digital_pixel_clock_khz = max_hdmi_pixel_clock;
+			break;
+		case CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_I:
+		case CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D:
+			max_digital_pixel_clock_khz = max_dvi_single_link_pixel_clock;
+			break;
+		case CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_I:
+		case CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D:
+		case CONNECTOR_OBJECT_ID_HDMI_TYPE_B:
+			max_digital_pixel_clock_khz = max_dvi_single_link_pixel_clock * 2;
+			break;
 		}
+
+		/* When the display EDID claims that it's an HDMI display,
+		 * we use the HDMI encoder mode of the display HW,
+		 * so we should verify against the max HDMI clock here.
+		 */
+		if (connector->display_info.is_hdmi)
+			max_digital_pixel_clock_khz = max_hdmi_pixel_clock;
+
+		if (mode->clock > max_digital_pixel_clock_khz)
+			return MODE_CLOCK_HIGH;
 	}
 
 	/* check against the max pixel clock */
@@ -1414,6 +1454,7 @@ amdgpu_connector_dp_detect(struct drm_connector *connector, bool force)
 				ret = connector_status_connected;
 			else if (amdgpu_connector->dac_load_detect) { /* try load detection */
 				const struct drm_encoder_helper_funcs *encoder_funcs = encoder->helper_private;
+
 				ret = encoder_funcs->detect(encoder, connector);
 			}
 		}
@@ -1439,10 +1480,8 @@ amdgpu_connector_dp_detect(struct drm_connector *connector, bool force)
 
 	amdgpu_connector_update_scratch_regs(connector, ret);
 out:
-	if (!drm_kms_helper_is_poll_worker()) {
-		pm_runtime_mark_last_busy(connector->dev->dev);
+	if (!drm_kms_helper_is_poll_worker())
 		pm_runtime_put_autosuspend(connector->dev->dev);
-	}
 
 	if (connector->connector_type == DRM_MODE_CONNECTOR_DisplayPort ||
 	    connector->connector_type == DRM_MODE_CONNECTOR_eDP)
@@ -1454,7 +1493,7 @@ out:
 }
 
 static enum drm_mode_status amdgpu_connector_dp_mode_valid(struct drm_connector *connector,
-					   struct drm_display_mode *mode)
+					   const struct drm_display_mode *mode)
 {
 	struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
 	struct amdgpu_connector_atom_dig *amdgpu_dig_connector = amdgpu_connector->con_priv;
@@ -1991,7 +2030,7 @@ amdgpu_connector_add(struct amdgpu_device *adev,
 	if (amdgpu_connector->hpd.hpd == AMDGPU_HPD_NONE) {
 		if (i2c_bus->valid) {
 			connector->polled = DRM_CONNECTOR_POLL_CONNECT |
-			                    DRM_CONNECTOR_POLL_DISCONNECT;
+						DRM_CONNECTOR_POLL_DISCONNECT;
 		}
 	} else
 		connector->polled = DRM_CONNECTOR_POLL_HPD;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.h
index 61fcef15ad72..eff833b6ed31 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.h
@@ -24,7 +24,6 @@
 #ifndef __AMDGPU_CONNECTORS_H__
 #define __AMDGPU_CONNECTORS_H__
 
-struct edid *amdgpu_connector_edid(struct drm_connector *connector);
 void amdgpu_connector_hotplug(struct drm_connector *connector);
 int amdgpu_connector_get_monitor_bpc(struct drm_connector *connector);
 u16 amdgpu_connector_encoder_get_dp_bridge_encoder_id(struct drm_connector *connector);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c
new file mode 100644
index 000000000000..425a3e564360
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.c
@@ -0,0 +1,591 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include <linux/list.h>
+#include "amdgpu.h"
+
+static const guid_t MCE			= CPER_NOTIFY_MCE;
+static const guid_t CMC			= CPER_NOTIFY_CMC;
+static const guid_t BOOT		= BOOT_TYPE;
+
+static const guid_t CRASHDUMP		= AMD_CRASHDUMP;
+static const guid_t RUNTIME		= AMD_GPU_NONSTANDARD_ERROR;
+
+static void __inc_entry_length(struct cper_hdr *hdr, uint32_t size)
+{
+	hdr->record_length += size;
+}
+
+static void amdgpu_cper_get_timestamp(struct cper_timestamp *timestamp)
+{
+	struct tm tm;
+	time64_t now = ktime_get_real_seconds();
+
+	time64_to_tm(now, 0, &tm);
+	timestamp->seconds = tm.tm_sec;
+	timestamp->minutes = tm.tm_min;
+	timestamp->hours = tm.tm_hour;
+	timestamp->flag = 0;
+	timestamp->day = tm.tm_mday;
+	timestamp->month = 1 + tm.tm_mon;
+	timestamp->year = (1900 + tm.tm_year) % 100;
+	timestamp->century = (1900 + tm.tm_year) / 100;
+}
+
+void amdgpu_cper_entry_fill_hdr(struct amdgpu_device *adev,
+				struct cper_hdr *hdr,
+				enum amdgpu_cper_type type,
+				enum cper_error_severity sev)
+{
+	char record_id[16];
+
+	hdr->signature[0]		= 'C';
+	hdr->signature[1]		= 'P';
+	hdr->signature[2]		= 'E';
+	hdr->signature[3]		= 'R';
+	hdr->revision			= CPER_HDR_REV_1;
+	hdr->signature_end		= 0xFFFFFFFF;
+	hdr->error_severity		= sev;
+
+	hdr->valid_bits.platform_id	= 1;
+	hdr->valid_bits.timestamp	= 1;
+
+	amdgpu_cper_get_timestamp(&hdr->timestamp);
+
+	snprintf(record_id, 9, "%d:%X",
+		 (adev->smuio.funcs && adev->smuio.funcs->get_socket_id) ?
+			 adev->smuio.funcs->get_socket_id(adev) :
+			 0,
+		 atomic_inc_return(&adev->cper.unique_id));
+	memcpy(hdr->record_id, record_id, 8);
+
+	snprintf(hdr->platform_id, 16, "0x%04X:0x%04X",
+		 adev->pdev->vendor, adev->pdev->device);
+	/* pmfw version should be part of creator_id according to CPER spec */
+	snprintf(hdr->creator_id, 16, "%s", CPER_CREATOR_ID_AMDGPU);
+
+	switch (type) {
+	case AMDGPU_CPER_TYPE_BOOT:
+		hdr->notify_type = BOOT;
+		break;
+	case AMDGPU_CPER_TYPE_FATAL:
+	case AMDGPU_CPER_TYPE_BP_THRESHOLD:
+		hdr->notify_type = MCE;
+		break;
+	case AMDGPU_CPER_TYPE_RUNTIME:
+		if (sev == CPER_SEV_NON_FATAL_CORRECTED)
+			hdr->notify_type = CMC;
+		else
+			hdr->notify_type = MCE;
+		break;
+	default:
+		dev_err(adev->dev, "Unknown CPER Type\n");
+		break;
+	}
+
+	__inc_entry_length(hdr, HDR_LEN);
+}
+
+static int amdgpu_cper_entry_fill_section_desc(struct amdgpu_device *adev,
+					       struct cper_sec_desc *section_desc,
+					       bool bp_threshold,
+					       bool poison,
+					       enum cper_error_severity sev,
+					       guid_t sec_type,
+					       uint32_t section_length,
+					       uint32_t section_offset)
+{
+	section_desc->revision_minor		= CPER_SEC_MINOR_REV_1;
+	section_desc->revision_major		= CPER_SEC_MAJOR_REV_22;
+	section_desc->sec_offset		= section_offset;
+	section_desc->sec_length		= section_length;
+	section_desc->valid_bits.fru_text	= 1;
+	section_desc->flag_bits.primary		= 1;
+	section_desc->severity			= sev;
+	section_desc->sec_type			= sec_type;
+
+	snprintf(section_desc->fru_text, 20, "OAM%d",
+		 (adev->smuio.funcs && adev->smuio.funcs->get_socket_id) ?
+			 adev->smuio.funcs->get_socket_id(adev) :
+			 0);
+
+	if (bp_threshold)
+		section_desc->flag_bits.exceed_err_threshold = 1;
+	if (poison)
+		section_desc->flag_bits.latent_err = 1;
+
+	return 0;
+}
+
+int amdgpu_cper_entry_fill_fatal_section(struct amdgpu_device *adev,
+					 struct cper_hdr *hdr,
+					 uint32_t idx,
+					 struct cper_sec_crashdump_reg_data reg_data)
+{
+	struct cper_sec_desc *section_desc;
+	struct cper_sec_crashdump_fatal *section;
+
+	section_desc = (struct cper_sec_desc *)((uint8_t *)hdr + SEC_DESC_OFFSET(idx));
+	section = (struct cper_sec_crashdump_fatal *)((uint8_t *)hdr +
+		   FATAL_SEC_OFFSET(hdr->sec_cnt, idx));
+
+	amdgpu_cper_entry_fill_section_desc(adev, section_desc, false, false,
+					    CPER_SEV_FATAL, CRASHDUMP, FATAL_SEC_LEN,
+					    FATAL_SEC_OFFSET(hdr->sec_cnt, idx));
+
+	section->body.reg_ctx_type = CPER_CTX_TYPE_CRASH;
+	section->body.reg_arr_size = sizeof(reg_data);
+	section->body.data = reg_data;
+
+	__inc_entry_length(hdr, SEC_DESC_LEN + FATAL_SEC_LEN);
+
+	return 0;
+}
+
+int amdgpu_cper_entry_fill_runtime_section(struct amdgpu_device *adev,
+					   struct cper_hdr *hdr,
+					   uint32_t idx,
+					   enum cper_error_severity sev,
+					   uint32_t *reg_dump,
+					   uint32_t reg_count)
+{
+	struct cper_sec_desc *section_desc;
+	struct cper_sec_nonstd_err *section;
+	bool poison;
+
+	poison = sev != CPER_SEV_NON_FATAL_CORRECTED;
+	section_desc = (struct cper_sec_desc *)((uint8_t *)hdr + SEC_DESC_OFFSET(idx));
+	section = (struct cper_sec_nonstd_err *)((uint8_t *)hdr +
+		   NONSTD_SEC_OFFSET(hdr->sec_cnt, idx));
+
+	amdgpu_cper_entry_fill_section_desc(adev, section_desc, false, poison,
+					    sev, RUNTIME, NONSTD_SEC_LEN,
+					    NONSTD_SEC_OFFSET(hdr->sec_cnt, idx));
+
+	reg_count = umin(reg_count, CPER_ACA_REG_COUNT);
+
+	section->hdr.valid_bits.err_info_cnt = 1;
+	section->hdr.valid_bits.err_context_cnt = 1;
+
+	section->info.error_type = RUNTIME;
+	section->info.ms_chk_bits.err_type_valid = 1;
+	section->ctx.reg_ctx_type = CPER_CTX_TYPE_CRASH;
+	section->ctx.reg_arr_size = sizeof(section->ctx.reg_dump);
+
+	memcpy(section->ctx.reg_dump, reg_dump, reg_count * sizeof(uint32_t));
+
+	__inc_entry_length(hdr, SEC_DESC_LEN + NONSTD_SEC_LEN);
+
+	return 0;
+}
+
+int amdgpu_cper_entry_fill_bad_page_threshold_section(struct amdgpu_device *adev,
+						      struct cper_hdr *hdr,
+						      uint32_t idx)
+{
+	struct cper_sec_desc *section_desc;
+	struct cper_sec_nonstd_err *section;
+	uint32_t socket_id;
+
+	section_desc = (struct cper_sec_desc *)((uint8_t *)hdr + SEC_DESC_OFFSET(idx));
+	section = (struct cper_sec_nonstd_err *)((uint8_t *)hdr +
+		   NONSTD_SEC_OFFSET(hdr->sec_cnt, idx));
+
+	amdgpu_cper_entry_fill_section_desc(adev, section_desc, true, false,
+					    CPER_SEV_FATAL, RUNTIME, NONSTD_SEC_LEN,
+					    NONSTD_SEC_OFFSET(hdr->sec_cnt, idx));
+
+	section->hdr.valid_bits.err_info_cnt = 1;
+	section->hdr.valid_bits.err_context_cnt = 1;
+
+	section->info.error_type = RUNTIME;
+	section->info.valid_bits.ms_chk = 1;
+	section->info.ms_chk_bits.err_type_valid = 1;
+	section->info.ms_chk_bits.err_type = 1;
+	section->info.ms_chk_bits.pcc = 1;
+	section->ctx.reg_ctx_type = CPER_CTX_TYPE_CRASH;
+	section->ctx.reg_arr_size = sizeof(section->ctx.reg_dump);
+
+	/* Hardcoded Reg dump for bad page threshold CPER */
+	socket_id = (adev->smuio.funcs && adev->smuio.funcs->get_socket_id) ?
+				adev->smuio.funcs->get_socket_id(adev) :
+				0;
+	section->ctx.reg_dump[CPER_ACA_REG_CTL_LO]    = 0x1;
+	section->ctx.reg_dump[CPER_ACA_REG_CTL_HI]    = 0x0;
+	section->ctx.reg_dump[CPER_ACA_REG_STATUS_LO] = 0x137;
+	section->ctx.reg_dump[CPER_ACA_REG_STATUS_HI] = 0xB0000000;
+	section->ctx.reg_dump[CPER_ACA_REG_ADDR_LO]   = 0x0;
+	section->ctx.reg_dump[CPER_ACA_REG_ADDR_HI]   = 0x0;
+	section->ctx.reg_dump[CPER_ACA_REG_MISC0_LO]  = 0x0;
+	section->ctx.reg_dump[CPER_ACA_REG_MISC0_HI]  = 0x0;
+	section->ctx.reg_dump[CPER_ACA_REG_CONFIG_LO] = 0x2;
+	section->ctx.reg_dump[CPER_ACA_REG_CONFIG_HI] = 0x1ff;
+	section->ctx.reg_dump[CPER_ACA_REG_IPID_LO]   = (socket_id / 4) & 0x01;
+	section->ctx.reg_dump[CPER_ACA_REG_IPID_HI]   = 0x096 | (((socket_id % 4) & 0x3) << 12);
+	section->ctx.reg_dump[CPER_ACA_REG_SYND_LO]   = 0x0;
+	section->ctx.reg_dump[CPER_ACA_REG_SYND_HI]   = 0x0;
+
+	__inc_entry_length(hdr, SEC_DESC_LEN + NONSTD_SEC_LEN);
+
+	return 0;
+}
+
+struct cper_hdr *amdgpu_cper_alloc_entry(struct amdgpu_device *adev,
+					 enum amdgpu_cper_type type,
+					 uint16_t section_count)
+{
+	struct cper_hdr *hdr;
+	uint32_t size = 0;
+
+	size += HDR_LEN;
+	size += (SEC_DESC_LEN * section_count);
+
+	switch (type) {
+	case AMDGPU_CPER_TYPE_RUNTIME:
+	case AMDGPU_CPER_TYPE_BP_THRESHOLD:
+		size += (NONSTD_SEC_LEN * section_count);
+		break;
+	case AMDGPU_CPER_TYPE_FATAL:
+		size += (FATAL_SEC_LEN * section_count);
+		break;
+	case AMDGPU_CPER_TYPE_BOOT:
+		size += (BOOT_SEC_LEN * section_count);
+		break;
+	default:
+		dev_err(adev->dev, "Unknown CPER Type!\n");
+		return NULL;
+	}
+
+	hdr = kzalloc(size, GFP_KERNEL);
+	if (!hdr)
+		return NULL;
+
+	/* Save this early */
+	hdr->sec_cnt = section_count;
+
+	return hdr;
+}
+
+int amdgpu_cper_generate_ue_record(struct amdgpu_device *adev,
+				   struct aca_bank *bank)
+{
+	struct cper_hdr *fatal = NULL;
+	struct cper_sec_crashdump_reg_data reg_data = { 0 };
+	struct amdgpu_ring *ring = &adev->cper.ring_buf;
+	int ret;
+
+	fatal = amdgpu_cper_alloc_entry(adev, AMDGPU_CPER_TYPE_FATAL, 1);
+	if (!fatal) {
+		dev_err(adev->dev, "fail to alloc cper entry for ue record\n");
+		return -ENOMEM;
+	}
+
+	reg_data.status_lo = lower_32_bits(bank->regs[ACA_REG_IDX_STATUS]);
+	reg_data.status_hi = upper_32_bits(bank->regs[ACA_REG_IDX_STATUS]);
+	reg_data.addr_lo   = lower_32_bits(bank->regs[ACA_REG_IDX_ADDR]);
+	reg_data.addr_hi   = upper_32_bits(bank->regs[ACA_REG_IDX_ADDR]);
+	reg_data.ipid_lo   = lower_32_bits(bank->regs[ACA_REG_IDX_IPID]);
+	reg_data.ipid_hi   = upper_32_bits(bank->regs[ACA_REG_IDX_IPID]);
+	reg_data.synd_lo   = lower_32_bits(bank->regs[ACA_REG_IDX_SYND]);
+	reg_data.synd_hi   = upper_32_bits(bank->regs[ACA_REG_IDX_SYND]);
+
+	amdgpu_cper_entry_fill_hdr(adev, fatal, AMDGPU_CPER_TYPE_FATAL, CPER_SEV_FATAL);
+	ret = amdgpu_cper_entry_fill_fatal_section(adev, fatal, 0, reg_data);
+	if (ret)
+		return ret;
+
+	amdgpu_cper_ring_write(ring, fatal, fatal->record_length);
+	kfree(fatal);
+
+	return 0;
+}
+
+int amdgpu_cper_generate_bp_threshold_record(struct amdgpu_device *adev)
+{
+	struct cper_hdr *bp_threshold = NULL;
+	struct amdgpu_ring *ring = &adev->cper.ring_buf;
+	int ret;
+
+	bp_threshold = amdgpu_cper_alloc_entry(adev, AMDGPU_CPER_TYPE_BP_THRESHOLD, 1);
+	if (!bp_threshold) {
+		dev_err(adev->dev, "fail to alloc cper entry for bad page threshold record\n");
+		return -ENOMEM;
+	}
+
+	amdgpu_cper_entry_fill_hdr(adev, bp_threshold,
+				   AMDGPU_CPER_TYPE_BP_THRESHOLD,
+				   CPER_SEV_FATAL);
+	ret = amdgpu_cper_entry_fill_bad_page_threshold_section(adev, bp_threshold, 0);
+	if (ret)
+		return ret;
+
+	amdgpu_cper_ring_write(ring, bp_threshold, bp_threshold->record_length);
+	kfree(bp_threshold);
+
+	return 0;
+}
+
+static enum cper_error_severity amdgpu_aca_err_type_to_cper_sev(struct amdgpu_device *adev,
+								enum aca_error_type aca_err_type)
+{
+	switch (aca_err_type) {
+	case ACA_ERROR_TYPE_UE:
+		return CPER_SEV_FATAL;
+	case ACA_ERROR_TYPE_CE:
+		return CPER_SEV_NON_FATAL_CORRECTED;
+	case ACA_ERROR_TYPE_DEFERRED:
+		return CPER_SEV_NON_FATAL_UNCORRECTED;
+	default:
+		dev_err(adev->dev, "Unknown ACA error type!\n");
+		return CPER_SEV_FATAL;
+	}
+}
+
+int amdgpu_cper_generate_ce_records(struct amdgpu_device *adev,
+				    struct aca_banks *banks,
+				    uint16_t bank_count)
+{
+	struct cper_hdr *corrected = NULL;
+	enum cper_error_severity sev = CPER_SEV_NON_FATAL_CORRECTED;
+	struct amdgpu_ring *ring = &adev->cper.ring_buf;
+	uint32_t reg_data[CPER_ACA_REG_COUNT] = { 0 };
+	struct aca_bank_node *node;
+	struct aca_bank *bank;
+	uint32_t i = 0;
+	int ret;
+
+	corrected = amdgpu_cper_alloc_entry(adev, AMDGPU_CPER_TYPE_RUNTIME, bank_count);
+	if (!corrected) {
+		dev_err(adev->dev, "fail to allocate cper entry for ce records\n");
+		return -ENOMEM;
+	}
+
+	/* Raise severity if any DE is detected in the ACA bank list */
+	list_for_each_entry(node, &banks->list, node) {
+		bank = &node->bank;
+		if (bank->aca_err_type == ACA_ERROR_TYPE_DEFERRED) {
+			sev = CPER_SEV_NON_FATAL_UNCORRECTED;
+			break;
+		}
+	}
+
+	amdgpu_cper_entry_fill_hdr(adev, corrected, AMDGPU_CPER_TYPE_RUNTIME, sev);
+
+	/* Combine CE and DE in cper record */
+	list_for_each_entry(node, &banks->list, node) {
+		bank = &node->bank;
+		reg_data[CPER_ACA_REG_CTL_LO]    = lower_32_bits(bank->regs[ACA_REG_IDX_CTL]);
+		reg_data[CPER_ACA_REG_CTL_HI]    = upper_32_bits(bank->regs[ACA_REG_IDX_CTL]);
+		reg_data[CPER_ACA_REG_STATUS_LO] = lower_32_bits(bank->regs[ACA_REG_IDX_STATUS]);
+		reg_data[CPER_ACA_REG_STATUS_HI] = upper_32_bits(bank->regs[ACA_REG_IDX_STATUS]);
+		reg_data[CPER_ACA_REG_ADDR_LO]   = lower_32_bits(bank->regs[ACA_REG_IDX_ADDR]);
+		reg_data[CPER_ACA_REG_ADDR_HI]   = upper_32_bits(bank->regs[ACA_REG_IDX_ADDR]);
+		reg_data[CPER_ACA_REG_MISC0_LO]  = lower_32_bits(bank->regs[ACA_REG_IDX_MISC0]);
+		reg_data[CPER_ACA_REG_MISC0_HI]  = upper_32_bits(bank->regs[ACA_REG_IDX_MISC0]);
+		reg_data[CPER_ACA_REG_CONFIG_LO] = lower_32_bits(bank->regs[ACA_REG_IDX_CONFIG]);
+		reg_data[CPER_ACA_REG_CONFIG_HI] = upper_32_bits(bank->regs[ACA_REG_IDX_CONFIG]);
+		reg_data[CPER_ACA_REG_IPID_LO]   = lower_32_bits(bank->regs[ACA_REG_IDX_IPID]);
+		reg_data[CPER_ACA_REG_IPID_HI]   = upper_32_bits(bank->regs[ACA_REG_IDX_IPID]);
+		reg_data[CPER_ACA_REG_SYND_LO]   = lower_32_bits(bank->regs[ACA_REG_IDX_SYND]);
+		reg_data[CPER_ACA_REG_SYND_HI]   = upper_32_bits(bank->regs[ACA_REG_IDX_SYND]);
+
+		ret = amdgpu_cper_entry_fill_runtime_section(adev, corrected, i++,
+				amdgpu_aca_err_type_to_cper_sev(adev, bank->aca_err_type),
+				reg_data, CPER_ACA_REG_COUNT);
+		if (ret)
+			return ret;
+	}
+
+	amdgpu_cper_ring_write(ring, corrected, corrected->record_length);
+	kfree(corrected);
+
+	return 0;
+}
+
+static bool amdgpu_cper_is_hdr(struct amdgpu_ring *ring, u64 pos)
+{
+	struct cper_hdr *chdr;
+
+	chdr = (struct cper_hdr *)&(ring->ring[pos]);
+	return strcmp(chdr->signature, "CPER") ? false : true;
+}
+
+static u32 amdgpu_cper_ring_get_ent_sz(struct amdgpu_ring *ring, u64 pos)
+{
+	struct cper_hdr *chdr;
+	u64 p;
+	u32 chunk, rec_len = 0;
+
+	chdr = (struct cper_hdr *)&(ring->ring[pos]);
+	chunk = ring->ring_size - (pos << 2);
+
+	if (!strcmp(chdr->signature, "CPER")) {
+		rec_len = chdr->record_length;
+		goto calc;
+	}
+
+	/* ring buffer is not full, no cper data after ring->wptr */
+	if (ring->count_dw)
+		goto calc;
+
+	for (p = pos + 1; p <= ring->buf_mask; p++) {
+		chdr = (struct cper_hdr *)&(ring->ring[p]);
+		if (!strcmp(chdr->signature, "CPER")) {
+			rec_len = (p - pos) << 2;
+			goto calc;
+		}
+	}
+
+calc:
+	if (!rec_len)
+		return chunk;
+	else
+		return umin(rec_len, chunk);
+}
+
+void amdgpu_cper_ring_write(struct amdgpu_ring *ring, void *src, int count)
+{
+	u64 pos, wptr_old, rptr;
+	int rec_cnt_dw = count >> 2;
+	u32 chunk, ent_sz;
+	u8 *s = (u8 *)src;
+
+	if (count >= ring->ring_size - 4) {
+		dev_err(ring->adev->dev,
+			"CPER data size(%d) is larger than ring size(%d)\n",
+			count, ring->ring_size - 4);
+
+		return;
+	}
+
+	mutex_lock(&ring->adev->cper.ring_lock);
+
+	wptr_old = ring->wptr;
+	rptr = *ring->rptr_cpu_addr & ring->ptr_mask;
+
+	while (count) {
+		ent_sz = amdgpu_cper_ring_get_ent_sz(ring, ring->wptr);
+		chunk = umin(ent_sz, count);
+
+		memcpy(&ring->ring[ring->wptr], s, chunk);
+
+		ring->wptr += (chunk >> 2);
+		ring->wptr &= ring->ptr_mask;
+		count -= chunk;
+		s += chunk;
+	}
+
+	if (ring->count_dw < rec_cnt_dw)
+		ring->count_dw = 0;
+
+	/* the buffer is overflow, adjust rptr */
+	if (((wptr_old < rptr) && (rptr <= ring->wptr)) ||
+	    ((ring->wptr < wptr_old) && (wptr_old < rptr)) ||
+	    ((rptr <= ring->wptr) && (ring->wptr < wptr_old))) {
+		pos = (ring->wptr + 1) & ring->ptr_mask;
+
+		do {
+			ent_sz = amdgpu_cper_ring_get_ent_sz(ring, pos);
+
+			rptr += (ent_sz >> 2);
+			rptr &= ring->ptr_mask;
+			*ring->rptr_cpu_addr = rptr;
+
+			pos = rptr;
+		} while (!amdgpu_cper_is_hdr(ring, rptr));
+	}
+
+	if (ring->count_dw >= rec_cnt_dw)
+		ring->count_dw -= rec_cnt_dw;
+	mutex_unlock(&ring->adev->cper.ring_lock);
+}
+
+static u64 amdgpu_cper_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	return *(ring->rptr_cpu_addr);
+}
+
+static u64 amdgpu_cper_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	return ring->wptr;
+}
+
+static const struct amdgpu_ring_funcs cper_ring_funcs = {
+	.type = AMDGPU_RING_TYPE_CPER,
+	.align_mask = 0xff,
+	.support_64bit_ptrs = false,
+	.get_rptr = amdgpu_cper_ring_get_rptr,
+	.get_wptr = amdgpu_cper_ring_get_wptr,
+};
+
+static int amdgpu_cper_ring_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring = &(adev->cper.ring_buf);
+
+	mutex_init(&adev->cper.ring_lock);
+
+	ring->adev = NULL;
+	ring->ring_obj = NULL;
+	ring->use_doorbell = false;
+	ring->no_scheduler = true;
+	ring->funcs = &cper_ring_funcs;
+
+	sprintf(ring->name, "cper");
+	return amdgpu_ring_init(adev, ring, CPER_MAX_RING_SIZE, NULL, 0,
+				AMDGPU_RING_PRIO_DEFAULT, NULL);
+}
+
+int amdgpu_cper_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (!amdgpu_aca_is_enabled(adev) && !amdgpu_sriov_ras_cper_en(adev))
+		return 0;
+
+	r = amdgpu_cper_ring_init(adev);
+	if (r) {
+		dev_err(adev->dev, "failed to initialize cper ring, r = %d\n", r);
+		return r;
+	}
+
+	mutex_init(&adev->cper.cper_lock);
+
+	adev->cper.enabled = true;
+	adev->cper.max_count = CPER_MAX_ALLOWED_COUNT;
+
+	return 0;
+}
+
+int amdgpu_cper_fini(struct amdgpu_device *adev)
+{
+	if (!amdgpu_aca_is_enabled(adev) && !amdgpu_sriov_ras_cper_en(adev))
+		return 0;
+
+	adev->cper.enabled = false;
+
+	amdgpu_ring_fini(&(adev->cper.ring_buf));
+	adev->cper.count = 0;
+	adev->cper.wptr = 0;
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h
new file mode 100644
index 000000000000..353421807387
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cper.h
@@ -0,0 +1,105 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __AMDGPU_CPER_H__
+#define __AMDGPU_CPER_H__
+
+#include "amd_cper.h"
+#include "amdgpu_aca.h"
+
+#define CPER_MAX_ALLOWED_COUNT		0x1000
+#define CPER_MAX_RING_SIZE		0X100000
+#define HDR_LEN				(sizeof(struct cper_hdr))
+#define SEC_DESC_LEN			(sizeof(struct cper_sec_desc))
+
+#define BOOT_SEC_LEN			(sizeof(struct cper_sec_crashdump_boot))
+#define FATAL_SEC_LEN			(sizeof(struct cper_sec_crashdump_fatal))
+#define NONSTD_SEC_LEN			(sizeof(struct cper_sec_nonstd_err))
+
+#define SEC_DESC_OFFSET(idx)		(HDR_LEN + (SEC_DESC_LEN * idx))
+
+#define BOOT_SEC_OFFSET(count, idx)	(HDR_LEN + (SEC_DESC_LEN * count) + (BOOT_SEC_LEN * idx))
+#define FATAL_SEC_OFFSET(count, idx)	(HDR_LEN + (SEC_DESC_LEN * count) + (FATAL_SEC_LEN * idx))
+#define NONSTD_SEC_OFFSET(count, idx)	(HDR_LEN + (SEC_DESC_LEN * count) + (NONSTD_SEC_LEN * idx))
+
+enum amdgpu_cper_type {
+	AMDGPU_CPER_TYPE_RUNTIME,
+	AMDGPU_CPER_TYPE_FATAL,
+	AMDGPU_CPER_TYPE_BOOT,
+	AMDGPU_CPER_TYPE_BP_THRESHOLD,
+};
+
+struct amdgpu_cper {
+	bool enabled;
+
+	atomic_t unique_id;
+	struct mutex cper_lock;
+
+	/* Lifetime CPERs generated */
+	uint32_t count;
+	uint32_t max_count;
+
+	uint32_t wptr;
+
+	void *ring[CPER_MAX_ALLOWED_COUNT];
+	struct amdgpu_ring ring_buf;
+	struct mutex ring_lock;
+};
+
+void amdgpu_cper_entry_fill_hdr(struct amdgpu_device *adev,
+				struct cper_hdr *hdr,
+				enum amdgpu_cper_type type,
+				enum cper_error_severity sev);
+int amdgpu_cper_entry_fill_fatal_section(struct amdgpu_device *adev,
+					 struct cper_hdr *hdr,
+					 uint32_t idx,
+					 struct cper_sec_crashdump_reg_data reg_data);
+int amdgpu_cper_entry_fill_runtime_section(struct amdgpu_device *adev,
+					   struct cper_hdr *hdr,
+					   uint32_t idx,
+					   enum cper_error_severity sev,
+					   uint32_t *reg_dump,
+					   uint32_t reg_count);
+int amdgpu_cper_entry_fill_bad_page_threshold_section(struct amdgpu_device *adev,
+						      struct cper_hdr *hdr,
+						      uint32_t section_idx);
+
+struct cper_hdr *amdgpu_cper_alloc_entry(struct amdgpu_device *adev,
+					 enum amdgpu_cper_type type,
+					 uint16_t section_count);
+/* UE must be encoded into separated cper entries, 1 UE 1 cper */
+int amdgpu_cper_generate_ue_record(struct amdgpu_device *adev,
+				   struct aca_bank *bank);
+/* CEs and DEs are combined into 1 cper entry */
+int amdgpu_cper_generate_ce_records(struct amdgpu_device *adev,
+				    struct aca_banks *banks,
+				    uint16_t bank_count);
+/* Bad page threshold is encoded into separated cper entry */
+int amdgpu_cper_generate_bp_threshold_record(struct amdgpu_device *adev);
+void amdgpu_cper_ring_write(struct amdgpu_ring *ring,
+			void *src, int count);
+int amdgpu_cper_init(struct amdgpu_device *adev);
+int amdgpu_cper_fini(struct amdgpu_device *adev);
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index 08eced097bd8..ecdfe6cb36cc 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -40,6 +40,7 @@
 #include "amdgpu_gmc.h"
 #include "amdgpu_gem.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_hmm.h"
 
 static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p,
 				 struct amdgpu_device *adev,
@@ -65,6 +66,8 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p,
 	}
 
 	amdgpu_sync_create(&p->sync);
+	drm_exec_init(&p->exec, DRM_EXEC_INTERRUPTIBLE_WAIT |
+		      DRM_EXEC_IGNORE_DUPLICATES, 0);
 	return 0;
 }
 
@@ -112,6 +115,9 @@ static int amdgpu_cs_p1_ib(struct amdgpu_cs_parser *p,
 	if (r < 0)
 		return r;
 
+	if (num_ibs[r] >= amdgpu_ring_max_ibs(chunk_ib->ip_type))
+		return -EINVAL;
+
 	++(num_ibs[r]);
 	p->gang_leader_idx = r;
 	return 0;
@@ -122,40 +128,24 @@ static int amdgpu_cs_p1_user_fence(struct amdgpu_cs_parser *p,
 				   uint32_t *offset)
 {
 	struct drm_gem_object *gobj;
-	struct amdgpu_bo *bo;
 	unsigned long size;
-	int r;
 
 	gobj = drm_gem_object_lookup(p->filp, data->handle);
 	if (gobj == NULL)
 		return -EINVAL;
 
-	bo = amdgpu_bo_ref(gem_to_amdgpu_bo(gobj));
-	p->uf_entry.priority = 0;
-	p->uf_entry.tv.bo = &bo->tbo;
-	/* One for TTM and two for the CS job */
-	p->uf_entry.tv.num_shared = 3;
-
+	p->uf_bo = amdgpu_bo_ref(gem_to_amdgpu_bo(gobj));
 	drm_gem_object_put(gobj);
 
-	size = amdgpu_bo_size(bo);
-	if (size != PAGE_SIZE || (data->offset + 8) > size) {
-		r = -EINVAL;
-		goto error_unref;
-	}
+	size = amdgpu_bo_size(p->uf_bo);
+	if (size != PAGE_SIZE || data->offset > (size - 8))
+		return -EINVAL;
 
-	if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) {
-		r = -EINVAL;
-		goto error_unref;
-	}
+	if (amdgpu_ttm_tt_get_usermm(p->uf_bo->tbo.ttm))
+		return -EINVAL;
 
 	*offset = data->offset;
-
 	return 0;
-
-error_unref:
-	amdgpu_bo_unref(&bo);
-	return r;
 }
 
 static int amdgpu_cs_p1_bo_handles(struct amdgpu_cs_parser *p,
@@ -189,25 +179,17 @@ static int amdgpu_cs_pass1(struct amdgpu_cs_parser *p,
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
 	unsigned int num_ibs[AMDGPU_CS_GANG_SIZE] = { };
 	struct amdgpu_vm *vm = &fpriv->vm;
-	uint64_t *chunk_array_user;
 	uint64_t *chunk_array;
 	uint32_t uf_offset = 0;
-	unsigned int size;
+	size_t size;
 	int ret;
 	int i;
 
-	chunk_array = kvmalloc_array(cs->in.num_chunks, sizeof(uint64_t),
-				     GFP_KERNEL);
-	if (!chunk_array)
-		return -ENOMEM;
-
-	/* get chunks */
-	chunk_array_user = u64_to_user_ptr(cs->in.chunks);
-	if (copy_from_user(chunk_array, chunk_array_user,
-			   sizeof(uint64_t)*cs->in.num_chunks)) {
-		ret = -EFAULT;
-		goto free_chunk;
-	}
+	chunk_array = memdup_array_user(u64_to_user_ptr(cs->in.chunks),
+					cs->in.num_chunks,
+					sizeof(uint64_t));
+	if (IS_ERR(chunk_array))
+		return PTR_ERR(chunk_array);
 
 	p->nchunks = cs->in.num_chunks;
 	p->chunks = kvmalloc_array(p->nchunks, sizeof(struct amdgpu_cs_chunk),
@@ -218,9 +200,8 @@ static int amdgpu_cs_pass1(struct amdgpu_cs_parser *p,
 	}
 
 	for (i = 0; i < p->nchunks; i++) {
-		struct drm_amdgpu_cs_chunk __user **chunk_ptr = NULL;
+		struct drm_amdgpu_cs_chunk __user *chunk_ptr = NULL;
 		struct drm_amdgpu_cs_chunk user_chunk;
-		uint32_t __user *cdata;
 
 		chunk_ptr = u64_to_user_ptr(chunk_array[i]);
 		if (copy_from_user(&user_chunk, chunk_ptr,
@@ -233,20 +214,16 @@ static int amdgpu_cs_pass1(struct amdgpu_cs_parser *p,
 		p->chunks[i].length_dw = user_chunk.length_dw;
 
 		size = p->chunks[i].length_dw;
-		cdata = u64_to_user_ptr(user_chunk.chunk_data);
 
-		p->chunks[i].kdata = kvmalloc_array(size, sizeof(uint32_t),
-						    GFP_KERNEL);
-		if (p->chunks[i].kdata == NULL) {
-			ret = -ENOMEM;
+		p->chunks[i].kdata = vmemdup_array_user(u64_to_user_ptr(user_chunk.chunk_data),
+							size,
+							sizeof(uint32_t));
+		if (IS_ERR(p->chunks[i].kdata)) {
+			ret = PTR_ERR(p->chunks[i].kdata);
 			i--;
 			goto free_partial_kdata;
 		}
 		size *= sizeof(uint32_t);
-		if (copy_from_user(p->chunks[i].kdata, cdata, size)) {
-			ret = -EFAULT;
-			goto free_partial_kdata;
-		}
 
 		/* Assume the worst on the following checks */
 		ret = -EINVAL;
@@ -274,6 +251,10 @@ static int amdgpu_cs_pass1(struct amdgpu_cs_parser *p,
 			if (size < sizeof(struct drm_amdgpu_bo_list_in))
 				goto free_partial_kdata;
 
+			/* Only a single BO list is allowed to simplify handling. */
+			if (p->bo_list)
+				goto free_partial_kdata;
+
 			ret = amdgpu_cs_p1_bo_handles(p, p->chunks[i].kdata);
 			if (ret)
 				goto free_partial_kdata;
@@ -285,6 +266,7 @@ static int amdgpu_cs_pass1(struct amdgpu_cs_parser *p,
 		case AMDGPU_CHUNK_ID_SCHEDULED_DEPENDENCIES:
 		case AMDGPU_CHUNK_ID_SYNCOBJ_TIMELINE_WAIT:
 		case AMDGPU_CHUNK_ID_SYNCOBJ_TIMELINE_SIGNAL:
+		case AMDGPU_CHUNK_ID_CP_GFX_SHADOW:
 			break;
 
 		default:
@@ -292,25 +274,45 @@ static int amdgpu_cs_pass1(struct amdgpu_cs_parser *p,
 		}
 	}
 
-	if (!p->gang_size) {
+	if (!p->gang_size || (amdgpu_sriov_vf(p->adev) && p->gang_size > 1)) {
 		ret = -EINVAL;
-		goto free_partial_kdata;
+		goto free_all_kdata;
 	}
 
 	for (i = 0; i < p->gang_size; ++i) {
 		ret = amdgpu_job_alloc(p->adev, vm, p->entities[i], vm,
-				       num_ibs[i], &p->jobs[i]);
+				       num_ibs[i], &p->jobs[i],
+				       p->filp->client_id);
 		if (ret)
 			goto free_all_kdata;
+		switch (p->adev->enforce_isolation[fpriv->xcp_id]) {
+		case AMDGPU_ENFORCE_ISOLATION_DISABLE:
+		default:
+			p->jobs[i]->enforce_isolation = false;
+			p->jobs[i]->run_cleaner_shader = false;
+			break;
+		case AMDGPU_ENFORCE_ISOLATION_ENABLE:
+			p->jobs[i]->enforce_isolation = true;
+			p->jobs[i]->run_cleaner_shader = true;
+			break;
+		case AMDGPU_ENFORCE_ISOLATION_ENABLE_LEGACY:
+			p->jobs[i]->enforce_isolation = true;
+			p->jobs[i]->run_cleaner_shader = false;
+			break;
+		case AMDGPU_ENFORCE_ISOLATION_NO_CLEANER_SHADER:
+			p->jobs[i]->enforce_isolation = true;
+			p->jobs[i]->run_cleaner_shader = false;
+			break;
+		}
 	}
 	p->gang_leader = p->jobs[p->gang_leader_idx];
 
-	if (p->ctx->vram_lost_counter != p->gang_leader->vram_lost_counter) {
+	if (p->ctx->generation != p->gang_leader->generation) {
 		ret = -ECANCELED;
 		goto free_all_kdata;
 	}
 
-	if (p->uf_entry.tv.bo)
+	if (p->uf_bo)
 		p->gang_leader->uf_addr = uf_offset;
 	kvfree(chunk_array);
 
@@ -354,10 +356,20 @@ static int amdgpu_cs_p2_ib(struct amdgpu_cs_parser *p,
 	ring = amdgpu_job_ring(job);
 	ib = &job->ibs[job->num_ibs++];
 
+	/* submissions to kernel queues are disabled */
+	if (ring->no_user_submission)
+		return -EINVAL;
+
 	/* MM engine doesn't support user fences */
-	if (p->uf_entry.tv.bo && ring->funcs->no_user_fence)
+	if (p->uf_bo && ring->funcs->no_user_fence)
 		return -EINVAL;
 
+	if (!p->adev->debug_enable_ce_cs &&
+	    chunk_ib->flags & AMDGPU_IB_FLAG_CE) {
+		dev_err_ratelimited(p->adev->dev, "CE CS is blocked, use debug=0x400 to override\n");
+		return -EINVAL;
+	}
+
 	if (chunk_ib->ip_type == AMDGPU_HW_IP_GFX &&
 	    chunk_ib->flags & AMDGPU_IB_FLAG_PREEMPT) {
 		if (chunk_ib->flags & AMDGPU_IB_FLAG_CE)
@@ -378,7 +390,7 @@ static int amdgpu_cs_p2_ib(struct amdgpu_cs_parser *p,
 			   chunk_ib->ib_bytes : 0,
 			   AMDGPU_IB_POOL_DELAYED, ib);
 	if (r) {
-		DRM_ERROR("Failed to get ib !\n");
+		drm_err(adev_to_drm(p->adev), "Failed to get ib !\n");
 		return r;
 	}
 
@@ -393,7 +405,7 @@ static int amdgpu_cs_p2_dependencies(struct amdgpu_cs_parser *p,
 {
 	struct drm_amdgpu_cs_chunk_dep *deps = chunk->kdata;
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
-	unsigned num_deps;
+	unsigned int num_deps;
 	int i, r;
 
 	num_deps = chunk->length_dw * 4 /
@@ -433,7 +445,7 @@ static int amdgpu_cs_p2_dependencies(struct amdgpu_cs_parser *p,
 			dma_fence_put(old);
 		}
 
-		r = amdgpu_sync_fence(&p->sync, fence);
+		r = amdgpu_sync_fence(&p->sync, fence, GFP_KERNEL);
 		dma_fence_put(fence);
 		if (r)
 			return r;
@@ -450,12 +462,12 @@ static int amdgpu_syncobj_lookup_and_add(struct amdgpu_cs_parser *p,
 
 	r = drm_syncobj_find_fence(p->filp, handle, point, flags, &fence);
 	if (r) {
-		DRM_ERROR("syncobj %u failed to find fence @ %llu (%d)!\n",
+		drm_err(adev_to_drm(p->adev), "syncobj %u failed to find fence @ %llu (%d)!\n",
 			  handle, point, r);
 		return r;
 	}
 
-	r = amdgpu_sync_fence(&p->sync, fence);
+	r = amdgpu_sync_fence(&p->sync, fence, GFP_KERNEL);
 	dma_fence_put(fence);
 	return r;
 }
@@ -464,7 +476,7 @@ static int amdgpu_cs_p2_syncobj_in(struct amdgpu_cs_parser *p,
 				   struct amdgpu_cs_chunk *chunk)
 {
 	struct drm_amdgpu_cs_chunk_sem *deps = chunk->kdata;
-	unsigned num_deps;
+	unsigned int num_deps;
 	int i, r;
 
 	num_deps = chunk->length_dw * 4 /
@@ -482,7 +494,7 @@ static int amdgpu_cs_p2_syncobj_timeline_wait(struct amdgpu_cs_parser *p,
 					      struct amdgpu_cs_chunk *chunk)
 {
 	struct drm_amdgpu_cs_chunk_syncobj *syncobj_deps = chunk->kdata;
-	unsigned num_deps;
+	unsigned int num_deps;
 	int i, r;
 
 	num_deps = chunk->length_dw * 4 /
@@ -502,7 +514,7 @@ static int amdgpu_cs_p2_syncobj_out(struct amdgpu_cs_parser *p,
 				    struct amdgpu_cs_chunk *chunk)
 {
 	struct drm_amdgpu_cs_chunk_sem *deps = chunk->kdata;
-	unsigned num_deps;
+	unsigned int num_deps;
 	int i;
 
 	num_deps = chunk->length_dw * 4 /
@@ -536,7 +548,7 @@ static int amdgpu_cs_p2_syncobj_timeline_signal(struct amdgpu_cs_parser *p,
 						struct amdgpu_cs_chunk *chunk)
 {
 	struct drm_amdgpu_cs_chunk_syncobj *syncobj_deps = chunk->kdata;
-	unsigned num_deps;
+	unsigned int num_deps;
 	int i;
 
 	num_deps = chunk->length_dw * 4 /
@@ -575,6 +587,26 @@ static int amdgpu_cs_p2_syncobj_timeline_signal(struct amdgpu_cs_parser *p,
 	return 0;
 }
 
+static int amdgpu_cs_p2_shadow(struct amdgpu_cs_parser *p,
+			       struct amdgpu_cs_chunk *chunk)
+{
+	struct drm_amdgpu_cs_chunk_cp_gfx_shadow *shadow = chunk->kdata;
+	int i;
+
+	if (shadow->flags & ~AMDGPU_CS_CHUNK_CP_GFX_SHADOW_FLAGS_INIT_SHADOW)
+		return -EINVAL;
+
+	for (i = 0; i < p->gang_size; ++i) {
+		p->jobs[i]->shadow_va = shadow->shadow_va;
+		p->jobs[i]->csa_va = shadow->csa_va;
+		p->jobs[i]->gds_va = shadow->gds_va;
+		p->jobs[i]->init_shadow =
+			shadow->flags & AMDGPU_CS_CHUNK_CP_GFX_SHADOW_FLAGS_INIT_SHADOW;
+	}
+
+	return 0;
+}
+
 static int amdgpu_cs_pass2(struct amdgpu_cs_parser *p)
 {
 	unsigned int ce_preempt = 0, de_preempt = 0;
@@ -617,6 +649,11 @@ static int amdgpu_cs_pass2(struct amdgpu_cs_parser *p)
 			if (r)
 				return r;
 			break;
+		case AMDGPU_CHUNK_ID_CP_GFX_SHADOW:
+			r = amdgpu_cs_p2_shadow(p, chunk);
+			if (r)
+				return r;
+			break;
 		}
 	}
 
@@ -671,7 +708,7 @@ static void amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev,
 	 */
 	const s64 us_upper_bound = 200000;
 
-	if (!adev->mm_stats.log2_max_MBps) {
+	if ((!adev->mm_stats.log2_max_MBps) || !ttm_resource_manager_used(&adev->mman.vram_mgr.manager)) {
 		*max_bytes = 0;
 		*max_vis_bytes = 0;
 		return;
@@ -729,6 +766,7 @@ static void amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev,
 
 		if (used_vis_vram < total_vis_vram) {
 			u64 free_vis_vram = total_vis_vram - used_vis_vram;
+
 			adev->mm_stats.accum_us_vis = min(adev->mm_stats.accum_us_vis +
 							  increment_us, us_upper_bound);
 
@@ -803,7 +841,7 @@ retry:
 
 	p->bytes_moved += ctx.bytes_moved;
 	if (!amdgpu_gmc_vram_full_visible(&adev->gmc) &&
-	    amdgpu_bo_in_cpu_visible_vram(bo))
+	    amdgpu_res_cpu_visible(adev, bo->tbo.resource))
 		p->bytes_moved_vis += ctx.bytes_moved;
 
 	if (unlikely(r == -ENOMEM) && domain != bo->allowed_domains) {
@@ -814,55 +852,18 @@ retry:
 	return r;
 }
 
-static int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p,
-			    struct list_head *validated)
-{
-	struct ttm_operation_ctx ctx = { true, false };
-	struct amdgpu_bo_list_entry *lobj;
-	int r;
-
-	list_for_each_entry(lobj, validated, tv.head) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(lobj->tv.bo);
-		struct mm_struct *usermm;
-
-		usermm = amdgpu_ttm_tt_get_usermm(bo->tbo.ttm);
-		if (usermm && usermm != current->mm)
-			return -EPERM;
-
-		if (amdgpu_ttm_tt_is_userptr(bo->tbo.ttm) &&
-		    lobj->user_invalidated && lobj->user_pages) {
-			amdgpu_bo_placement_from_domain(bo,
-							AMDGPU_GEM_DOMAIN_CPU);
-			r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
-			if (r)
-				return r;
-
-			amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm,
-						     lobj->user_pages);
-		}
-
-		r = amdgpu_cs_bo_validate(p, bo);
-		if (r)
-			return r;
-
-		kvfree(lobj->user_pages);
-		lobj->user_pages = NULL;
-	}
-	return 0;
-}
-
 static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 				union drm_amdgpu_cs *cs)
 {
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
+	struct ttm_operation_ctx ctx = { true, false };
 	struct amdgpu_vm *vm = &fpriv->vm;
 	struct amdgpu_bo_list_entry *e;
-	struct list_head duplicates;
+	struct drm_gem_object *obj;
+	unsigned long index;
 	unsigned int i;
 	int r;
 
-	INIT_LIST_HEAD(&p->validated);
-
 	/* p->bo_list could already be assigned if AMDGPU_CHUNK_ID_BO_HANDLES is present */
 	if (cs->in.bo_list_handle) {
 		if (p->bo_list)
@@ -882,45 +883,25 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 
 	mutex_lock(&p->bo_list->bo_list_mutex);
 
-	/* One for TTM and one for the CS job */
-	amdgpu_bo_list_for_each_entry(e, p->bo_list)
-		e->tv.num_shared = 2;
-
-	amdgpu_bo_list_get_list(p->bo_list, &p->validated);
-
-	INIT_LIST_HEAD(&duplicates);
-	amdgpu_vm_get_pd_bo(&fpriv->vm, &p->validated, &p->vm_pd);
-
-	if (p->uf_entry.tv.bo && !ttm_to_amdgpu_bo(p->uf_entry.tv.bo)->parent)
-		list_add(&p->uf_entry.tv.head, &p->validated);
-
 	/* Get userptr backing pages. If pages are updated after registered
 	 * in amdgpu_gem_userptr_ioctl(), amdgpu_cs_list_validate() will do
 	 * amdgpu_ttm_backend_bind() to flush and invalidate new pages
 	 */
 	amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
 		bool userpage_invalidated = false;
-		int i;
-
-		e->user_pages = kvmalloc_array(bo->tbo.ttm->num_pages,
-					sizeof(struct page *),
-					GFP_KERNEL | __GFP_ZERO);
-		if (!e->user_pages) {
-			DRM_ERROR("kvmalloc_array failure\n");
-			r = -ENOMEM;
-			goto out_free_user_pages;
-		}
+		struct amdgpu_bo *bo = e->bo;
 
-		r = amdgpu_ttm_tt_get_user_pages(bo, e->user_pages, &e->range);
-		if (r) {
-			kvfree(e->user_pages);
-			e->user_pages = NULL;
+		e->range = amdgpu_hmm_range_alloc(NULL);
+		if (unlikely(!e->range))
+			return -ENOMEM;
+
+		r = amdgpu_ttm_tt_get_user_pages(bo, e->range);
+		if (r)
 			goto out_free_user_pages;
-		}
 
 		for (i = 0; i < bo->tbo.ttm->num_pages; i++) {
-			if (bo->tbo.ttm->pages[i] != e->user_pages[i]) {
+			if (bo->tbo.ttm->pages[i] !=
+				hmm_pfn_to_page(e->range->hmm_range.hmm_pfns[i])) {
 				userpage_invalidated = true;
 				break;
 			}
@@ -928,18 +909,53 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 		e->user_invalidated = userpage_invalidated;
 	}
 
-	r = ttm_eu_reserve_buffers(&p->ticket, &p->validated, true,
-				   &duplicates);
-	if (unlikely(r != 0)) {
-		if (r != -ERESTARTSYS)
-			DRM_ERROR("ttm_eu_reserve_buffers failed.\n");
-		goto out_free_user_pages;
+	drm_exec_until_all_locked(&p->exec) {
+		r = amdgpu_vm_lock_pd(&fpriv->vm, &p->exec, 1 + p->gang_size);
+		drm_exec_retry_on_contention(&p->exec);
+		if (unlikely(r))
+			goto out_free_user_pages;
+
+		amdgpu_bo_list_for_each_entry(e, p->bo_list) {
+			/* One fence for TTM and one for each CS job */
+			r = drm_exec_prepare_obj(&p->exec, &e->bo->tbo.base,
+						 1 + p->gang_size);
+			drm_exec_retry_on_contention(&p->exec);
+			if (unlikely(r))
+				goto out_free_user_pages;
+
+			e->bo_va = amdgpu_vm_bo_find(vm, e->bo);
+		}
+
+		if (p->uf_bo) {
+			r = drm_exec_prepare_obj(&p->exec, &p->uf_bo->tbo.base,
+						 1 + p->gang_size);
+			drm_exec_retry_on_contention(&p->exec);
+			if (unlikely(r))
+				goto out_free_user_pages;
+		}
 	}
 
-	amdgpu_bo_list_for_each_entry(e, p->bo_list) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
+	amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) {
+		struct mm_struct *usermm;
 
-		e->bo_va = amdgpu_vm_bo_find(vm, bo);
+		usermm = amdgpu_ttm_tt_get_usermm(e->bo->tbo.ttm);
+		if (usermm && usermm != current->mm) {
+			r = -EPERM;
+			goto out_free_user_pages;
+		}
+
+		if (amdgpu_ttm_tt_is_userptr(e->bo->tbo.ttm) &&
+		    e->user_invalidated) {
+			amdgpu_bo_placement_from_domain(e->bo,
+							AMDGPU_GEM_DOMAIN_CPU);
+			r = ttm_bo_validate(&e->bo->tbo, &e->bo->placement,
+					    &ctx);
+			if (r)
+				goto out_free_user_pages;
+
+			amdgpu_ttm_tt_set_user_pages(e->bo->tbo.ttm,
+						     e->range);
+		}
 	}
 
 	amdgpu_cs_get_threshold_for_moves(p->adev, &p->bytes_moved_threshold,
@@ -947,29 +963,25 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 	p->bytes_moved = 0;
 	p->bytes_moved_vis = 0;
 
-	r = amdgpu_vm_validate_pt_bos(p->adev, &fpriv->vm,
-				      amdgpu_cs_bo_validate, p);
+	r = amdgpu_vm_validate(p->adev, &fpriv->vm, NULL,
+			       amdgpu_cs_bo_validate, p);
 	if (r) {
-		DRM_ERROR("amdgpu_vm_validate_pt_bos() failed.\n");
-		goto error_validate;
+		drm_err(adev_to_drm(p->adev), "amdgpu_vm_validate() failed.\n");
+		goto out_free_user_pages;
 	}
 
-	r = amdgpu_cs_list_validate(p, &duplicates);
-	if (r)
-		goto error_validate;
-
-	r = amdgpu_cs_list_validate(p, &p->validated);
-	if (r)
-		goto error_validate;
-
-	if (p->uf_entry.tv.bo) {
-		struct amdgpu_bo *uf = ttm_to_amdgpu_bo(p->uf_entry.tv.bo);
+	drm_exec_for_each_locked_object(&p->exec, index, obj) {
+		r = amdgpu_cs_bo_validate(p, gem_to_amdgpu_bo(obj));
+		if (unlikely(r))
+			goto out_free_user_pages;
+	}
 
-		r = amdgpu_ttm_alloc_gart(&uf->tbo);
-		if (r)
-			goto error_validate;
+	if (p->uf_bo) {
+		r = amdgpu_ttm_alloc_gart(&p->uf_bo->tbo);
+		if (unlikely(r))
+			goto out_free_user_pages;
 
-		p->gang_leader->uf_addr += amdgpu_bo_gpu_offset(uf);
+		p->gang_leader->uf_addr += amdgpu_bo_gpu_offset(p->uf_bo);
 	}
 
 	amdgpu_cs_report_moved_bytes(p->adev, p->bytes_moved,
@@ -981,18 +993,9 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 					 p->bo_list->oa_obj);
 	return 0;
 
-error_validate:
-	ttm_eu_backoff_reservation(&p->ticket, &p->validated);
-
 out_free_user_pages:
 	amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
-
-		if (!e->user_pages)
-			continue;
-		amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm, e->range);
-		kvfree(e->user_pages);
-		e->user_pages = NULL;
+		amdgpu_hmm_range_free(e->range);
 		e->range = NULL;
 	}
 	mutex_unlock(&p->bo_list->bo_list_mutex);
@@ -1035,21 +1038,20 @@ static int amdgpu_cs_patch_ibs(struct amdgpu_cs_parser *p,
 		va_start = ib->gpu_addr & AMDGPU_GMC_HOLE_MASK;
 		r = amdgpu_cs_find_mapping(p, va_start, &aobj, &m);
 		if (r) {
-			DRM_ERROR("IB va_start is invalid\n");
+			drm_err(adev_to_drm(p->adev), "IB va_start is invalid\n");
 			return r;
 		}
 
 		if ((va_start + ib->length_dw * 4) >
 		    (m->last + 1) * AMDGPU_GPU_PAGE_SIZE) {
-			DRM_ERROR("IB va_start+ib_bytes is invalid\n");
+			drm_err(adev_to_drm(p->adev), "IB va_start+ib_bytes is invalid\n");
 			return -EINVAL;
 		}
 
 		/* the IB should be reserved at this point */
 		r = amdgpu_bo_kmap(aobj, (void **)&kptr);
-		if (r) {
+		if (r)
 			return r;
-		}
 
 		kptr += va_start - (m->start * AMDGPU_GPU_PAGE_SIZE);
 
@@ -1060,6 +1062,9 @@ static int amdgpu_cs_patch_ibs(struct amdgpu_cs_parser *p,
 			r = amdgpu_ring_parse_cs(ring, p, job, ib);
 			if (r)
 				return r;
+
+			if (ib->sa_bo)
+				ib->gpu_addr =  amdgpu_sa_bo_gpu_addr(ib->sa_bo);
 		} else {
 			ib->ptr = (uint32_t *)kptr;
 			r = amdgpu_ring_patch_cs_in_place(ring, p, job, ib);
@@ -1093,10 +1098,27 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p)
 	struct amdgpu_vm *vm = &fpriv->vm;
 	struct amdgpu_bo_list_entry *e;
 	struct amdgpu_bo_va *bo_va;
-	struct amdgpu_bo *bo;
 	unsigned int i;
 	int r;
 
+	/*
+	 * We can't use gang submit on with reserved VMIDs when the VM changes
+	 * can't be invalidated by more than one engine at the same time.
+	 */
+	if (p->gang_size > 1 && !adev->vm_manager.concurrent_flush) {
+		for (i = 0; i < p->gang_size; ++i) {
+			struct drm_sched_entity *entity = p->entities[i];
+			struct drm_gpu_scheduler *sched = entity->rq->sched;
+			struct amdgpu_ring *ring = to_amdgpu_ring(sched);
+
+			if (amdgpu_vmid_uses_reserved(vm, ring->vm_hub))
+				return -EINVAL;
+		}
+	}
+
+	if (!amdgpu_vm_ready(vm))
+		return -EINVAL;
+
 	r = amdgpu_vm_clear_freed(adev, vm, NULL);
 	if (r)
 		return r;
@@ -1105,7 +1127,8 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p)
 	if (r)
 		return r;
 
-	r = amdgpu_sync_fence(&p->sync, fpriv->prt_va->last_pt_update);
+	r = amdgpu_sync_fence(&p->sync, fpriv->prt_va->last_pt_update,
+			      GFP_KERNEL);
 	if (r)
 		return r;
 
@@ -1116,17 +1139,18 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p)
 		if (r)
 			return r;
 
-		r = amdgpu_sync_fence(&p->sync, bo_va->last_pt_update);
+		r = amdgpu_sync_fence(&p->sync, bo_va->last_pt_update,
+				      GFP_KERNEL);
 		if (r)
 			return r;
 	}
 
+	/* FIXME: In theory this loop shouldn't be needed any more when
+	 * amdgpu_vm_handle_moved handles all moved BOs that are reserved
+	 * with p->ticket. But removing it caused test regressions, so I'm
+	 * leaving it here for now.
+	 */
 	amdgpu_bo_list_for_each_entry(e, p->bo_list) {
-		/* ignore duplicates */
-		bo = ttm_to_amdgpu_bo(e->tv.bo);
-		if (!bo)
-			continue;
-
 		bo_va = e->bo_va;
 		if (bo_va == NULL)
 			continue;
@@ -1135,12 +1159,13 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p)
 		if (r)
 			return r;
 
-		r = amdgpu_sync_fence(&p->sync, bo_va->last_pt_update);
+		r = amdgpu_sync_fence(&p->sync, bo_va->last_pt_update,
+				      GFP_KERNEL);
 		if (r)
 			return r;
 	}
 
-	r = amdgpu_vm_handle_moved(adev, vm);
+	r = amdgpu_vm_handle_moved(adev, vm, &p->exec.ticket);
 	if (r)
 		return r;
 
@@ -1148,7 +1173,7 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p)
 	if (r)
 		return r;
 
-	r = amdgpu_sync_fence(&p->sync, vm->last_update);
+	r = amdgpu_sync_fence(&p->sync, vm->last_update, GFP_KERNEL);
 	if (r)
 		return r;
 
@@ -1161,16 +1186,16 @@ static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p)
 		job->vm_pd_addr = amdgpu_gmc_pd_addr(vm->root.bo);
 	}
 
-	if (amdgpu_vm_debug) {
+	if (adev->debug_vm) {
 		/* Invalidate all BOs to test for userspace bugs */
 		amdgpu_bo_list_for_each_entry(e, p->bo_list) {
-			struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
+			struct amdgpu_bo *bo = e->bo;
 
 			/* ignore duplicates */
 			if (!bo)
 				continue;
 
-			amdgpu_vm_bo_invalidate(adev, bo, false);
+			amdgpu_vm_bo_invalidate(bo, false);
 		}
 	}
 
@@ -1181,20 +1206,22 @@ static int amdgpu_cs_sync_rings(struct amdgpu_cs_parser *p)
 {
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
 	struct drm_gpu_scheduler *sched;
-	struct amdgpu_bo_list_entry *e;
+	struct drm_gem_object *obj;
 	struct dma_fence *fence;
+	unsigned long index;
 	unsigned int i;
 	int r;
 
 	r = amdgpu_ctx_wait_prev_fence(p->ctx, p->entities[p->gang_leader_idx]);
 	if (r) {
 		if (r != -ERESTARTSYS)
-			DRM_ERROR("amdgpu_ctx_wait_prev_fence failed.\n");
+			drm_err(adev_to_drm(p->adev), "amdgpu_ctx_wait_prev_fence failed.\n");
 		return r;
 	}
 
-	list_for_each_entry(e, &p->validated, tv.head) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
+	drm_exec_for_each_locked_object(&p->exec, index, obj) {
+		struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+
 		struct dma_resv *resv = bo->tbo.base.resv;
 		enum amdgpu_sync_mode sync_mode;
 
@@ -1227,7 +1254,8 @@ static int amdgpu_cs_sync_rings(struct amdgpu_cs_parser *p)
 			continue;
 		}
 
-		r = amdgpu_sync_fence(&p->gang_leader->explicit_sync, fence);
+		r = amdgpu_sync_fence(&p->gang_leader->explicit_sync, fence,
+				      GFP_KERNEL);
 		dma_fence_put(fence);
 		if (r)
 			return r;
@@ -1258,6 +1286,8 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
 	struct amdgpu_job *leader = p->gang_leader;
 	struct amdgpu_bo_list_entry *e;
+	struct drm_gem_object *gobj;
+	unsigned long index;
 	unsigned int i;
 	uint64_t seq;
 	int r;
@@ -1276,7 +1306,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 		r = drm_sched_job_add_dependency(&leader->base, fence);
 		if (r) {
 			dma_fence_put(fence);
-			goto error_cleanup;
+			return r;
 		}
 	}
 
@@ -1296,31 +1326,33 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	 */
 	r = 0;
 	amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) {
-		struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo);
-
-		r |= !amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm, e->range);
+		r |= !amdgpu_hmm_range_valid(e->range);
+		amdgpu_hmm_range_free(e->range);
 		e->range = NULL;
 	}
 	if (r) {
 		r = -EAGAIN;
-		goto error_unlock;
+		mutex_unlock(&p->adev->notifier_lock);
+		return r;
 	}
 
 	p->fence = dma_fence_get(&leader->base.s_fence->finished);
-	list_for_each_entry(e, &p->validated, tv.head) {
+	drm_exec_for_each_locked_object(&p->exec, index, gobj) {
+
+		ttm_bo_move_to_lru_tail_unlocked(&gem_to_amdgpu_bo(gobj)->tbo);
 
 		/* Everybody except for the gang leader uses READ */
 		for (i = 0; i < p->gang_size; ++i) {
 			if (p->jobs[i] == leader)
 				continue;
 
-			dma_resv_add_fence(e->tv.bo->base.resv,
+			dma_resv_add_fence(gobj->resv,
 					   &p->jobs[i]->base.s_fence->finished,
 					   DMA_RESV_USAGE_READ);
 		}
 
-		/* The gang leader is remembered as writer */
-		e->tv.num_shared = 0;
+		/* The gang leader as remembered as writer */
+		dma_resv_add_fence(gobj->resv, p->fence, DMA_RESV_USAGE_WRITE);
 	}
 
 	seq = amdgpu_ctx_add_fence(p->ctx, p->entities[p->gang_leader_idx],
@@ -1336,7 +1368,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	cs->out.handle = seq;
 	leader->uf_sequence = seq;
 
-	amdgpu_vm_bo_trace_cs(&fpriv->vm, &p->ticket);
+	amdgpu_vm_bo_trace_cs(&fpriv->vm, &p->exec.ticket);
 	for (i = 0; i < p->gang_size; ++i) {
 		amdgpu_job_free_resources(p->jobs[i]);
 		trace_amdgpu_cs_ioctl(p->jobs[i]);
@@ -1345,27 +1377,20 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	}
 
 	amdgpu_vm_move_to_lru_tail(p->adev, &fpriv->vm);
-	ttm_eu_fence_buffer_objects(&p->ticket, &p->validated, p->fence);
 
 	mutex_unlock(&p->adev->notifier_lock);
 	mutex_unlock(&p->bo_list->bo_list_mutex);
 	return 0;
-
-error_unlock:
-	mutex_unlock(&p->adev->notifier_lock);
-
-error_cleanup:
-	for (i = 0; i < p->gang_size; ++i)
-		drm_sched_job_cleanup(&p->jobs[i]->base);
-	return r;
 }
 
 /* Cleanup the parser structure */
 static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser)
 {
-	unsigned i;
+	unsigned int i;
 
 	amdgpu_sync_free(&parser->sync);
+	drm_exec_fini(&parser->exec);
+
 	for (i = 0; i < parser->num_post_deps; i++) {
 		drm_syncobj_put(parser->post_deps[i].syncobj);
 		kfree(parser->post_deps[i].chain);
@@ -1386,11 +1411,7 @@ static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser)
 		if (parser->jobs[i])
 			amdgpu_job_free(parser->jobs[i]);
 	}
-	if (parser->uf_entry.tv.bo) {
-		struct amdgpu_bo *uf = ttm_to_amdgpu_bo(parser->uf_entry.tv.bo);
-
-		amdgpu_bo_unref(&uf);
-	}
+	amdgpu_bo_unref(&parser->uf_bo);
 }
 
 int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
@@ -1407,8 +1428,7 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 
 	r = amdgpu_cs_parser_init(&parser, adev, filp, data);
 	if (r) {
-		if (printk_ratelimit())
-			DRM_ERROR("Failed to initialize parser %d!\n", r);
+		drm_err_ratelimited(dev, "Failed to initialize parser %d!\n", r);
 		return r;
 	}
 
@@ -1423,9 +1443,9 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	r = amdgpu_cs_parser_bos(&parser, data);
 	if (r) {
 		if (r == -ENOMEM)
-			DRM_ERROR("Not enough memory for command submission!\n");
+			drm_err(dev, "Not enough memory for command submission!\n");
 		else if (r != -ERESTARTSYS && r != -EAGAIN)
-			DRM_ERROR("Failed to process the buffer list %d!\n", r);
+			drm_dbg(dev, "Failed to process the buffer list %d!\n", r);
 		goto error_fini;
 	}
 
@@ -1451,7 +1471,6 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	return 0;
 
 error_backoff:
-	ttm_eu_backoff_reservation(&parser.ticket, &parser.validated);
 	mutex_unlock(&parser.bo_list->bo_list_mutex);
 
 error_fini:
@@ -1631,15 +1650,15 @@ static int amdgpu_cs_wait_all_fences(struct amdgpu_device *adev,
 			continue;
 
 		r = dma_fence_wait_timeout(fence, true, timeout);
+		if (r > 0 && fence->error)
+			r = fence->error;
+
 		dma_fence_put(fence);
 		if (r < 0)
 			return r;
 
 		if (r == 0)
 			break;
-
-		if (fence->error)
-			return fence->error;
 	}
 
 	memset(wait, 0, sizeof(*wait));
@@ -1725,30 +1744,21 @@ int amdgpu_cs_wait_fences_ioctl(struct drm_device *dev, void *data,
 {
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	union drm_amdgpu_wait_fences *wait = data;
-	uint32_t fence_count = wait->in.fence_count;
-	struct drm_amdgpu_fence *fences_user;
 	struct drm_amdgpu_fence *fences;
 	int r;
 
 	/* Get the fences from userspace */
-	fences = kmalloc_array(fence_count, sizeof(struct drm_amdgpu_fence),
-			GFP_KERNEL);
-	if (fences == NULL)
-		return -ENOMEM;
-
-	fences_user = u64_to_user_ptr(wait->in.fences);
-	if (copy_from_user(fences, fences_user,
-		sizeof(struct drm_amdgpu_fence) * fence_count)) {
-		r = -EFAULT;
-		goto err_free_fences;
-	}
+	fences = memdup_array_user(u64_to_user_ptr(wait->in.fences),
+				   wait->in.fence_count,
+				   sizeof(struct drm_amdgpu_fence));
+	if (IS_ERR(fences))
+		return PTR_ERR(fences);
 
 	if (wait->in.wait_all)
 		r = amdgpu_cs_wait_all_fences(adev, filp, wait, fences);
 	else
 		r = amdgpu_cs_wait_any_fence(adev, filp, wait, fences);
 
-err_free_fences:
 	kfree(fences);
 
 	return r;
@@ -1774,7 +1784,7 @@ int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
 	struct ttm_operation_ctx ctx = { false, false };
 	struct amdgpu_vm *vm = &fpriv->vm;
 	struct amdgpu_bo_va_mapping *mapping;
-	int r;
+	int i, r;
 
 	addr /= AMDGPU_GPU_PAGE_SIZE;
 
@@ -1786,12 +1796,17 @@ int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
 	*map = mapping;
 
 	/* Double check that the BO is reserved by this CS */
-	if (dma_resv_locking_ctx((*bo)->tbo.base.resv) != &parser->ticket)
+	if (dma_resv_locking_ctx((*bo)->tbo.base.resv) != &parser->exec.ticket)
 		return -EINVAL;
 
-	if (!((*bo)->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)) {
-		(*bo)->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
+	/* Make sure VRAM is allocated contigiously */
+	(*bo)->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
+	if ((*bo)->tbo.resource->mem_type == TTM_PL_VRAM &&
+	    !((*bo)->tbo.resource->placement & TTM_PL_FLAG_CONTIGUOUS)) {
+
 		amdgpu_bo_placement_from_domain(*bo, (*bo)->allowed_domains);
+		for (i = 0; i < (*bo)->placement.num_placement; i++)
+			(*bo)->placements[i].flags |= TTM_PL_FLAG_CONTIGUOUS;
 		r = ttm_bo_validate(&(*bo)->tbo, &(*bo)->placement, &ctx);
 		if (r)
 			return r;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.h
index fb3e3d56d427..39c33ad100cb 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.h
@@ -24,6 +24,7 @@
 #define __AMDGPU_CS_H__
 
 #include <linux/ww_mutex.h>
+#include <drm/drm_exec.h>
 
 #include "amdgpu_job.h"
 #include "amdgpu_bo_list.h"
@@ -62,11 +63,9 @@ struct amdgpu_cs_parser {
 	struct amdgpu_job	*gang_leader;
 
 	/* buffer objects */
-	struct ww_acquire_ctx		ticket;
+	struct drm_exec			exec;
 	struct amdgpu_bo_list		*bo_list;
 	struct amdgpu_mn		*mn;
-	struct amdgpu_bo_list_entry	vm_pd;
-	struct list_head		validated;
 	struct dma_fence		*fence;
 	uint64_t			bytes_moved_threshold;
 	uint64_t			bytes_moved_vis_threshold;
@@ -74,7 +73,7 @@ struct amdgpu_cs_parser {
 	uint64_t			bytes_moved_vis;
 
 	/* user fence */
-	struct amdgpu_bo_list_entry	uf_entry;
+	struct amdgpu_bo		*uf_bo;
 
 	unsigned			num_post_deps;
 	struct amdgpu_cs_post_dep	*post_deps;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c
index c6d4d41c4393..02138aa55793 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.c
@@ -22,13 +22,14 @@
  * * Author: Monk.liu@amd.com
  */
 
+#include <drm/drm_exec.h>
+
 #include "amdgpu.h"
 
 uint64_t amdgpu_csa_vaddr(struct amdgpu_device *adev)
 {
-	uint64_t addr = adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT;
+	uint64_t addr = AMDGPU_VA_RESERVED_CSA_START(adev);
 
-	addr -= AMDGPU_VA_RESERVED_SIZE;
 	addr = amdgpu_gmc_sign_extend(addr);
 
 	return addr;
@@ -65,31 +66,25 @@ int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			  struct amdgpu_bo *bo, struct amdgpu_bo_va **bo_va,
 			  uint64_t csa_addr, uint32_t size)
 {
-	struct ww_acquire_ctx ticket;
-	struct list_head list;
-	struct amdgpu_bo_list_entry pd;
-	struct ttm_validate_buffer csa_tv;
+	struct drm_exec exec;
 	int r;
 
-	INIT_LIST_HEAD(&list);
-	INIT_LIST_HEAD(&csa_tv.head);
-	csa_tv.bo = &bo->tbo;
-	csa_tv.num_shared = 1;
-
-	list_add(&csa_tv.head, &list);
-	amdgpu_vm_get_pd_bo(vm, &list, &pd);
-
-	r = ttm_eu_reserve_buffers(&ticket, &list, true, NULL);
-	if (r) {
-		DRM_ERROR("failed to reserve CSA,PD BOs: err=%d\n", r);
-		return r;
+	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT, 0);
+	drm_exec_until_all_locked(&exec) {
+		r = amdgpu_vm_lock_pd(vm, &exec, 0);
+		if (likely(!r))
+			r = drm_exec_lock_obj(&exec, &bo->tbo.base);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(r)) {
+			DRM_ERROR("failed to reserve CSA,PD BOs: err=%d\n", r);
+			goto error;
+		}
 	}
 
 	*bo_va = amdgpu_vm_bo_add(adev, vm, bo);
 	if (!*bo_va) {
-		ttm_eu_backoff_reservation(&ticket, &list);
-		DRM_ERROR("failed to create bo_va for static CSA\n");
-		return -ENOMEM;
+		r = -ENOMEM;
+		goto error;
 	}
 
 	r = amdgpu_vm_bo_map(adev, *bo_va, csa_addr, 0, size,
@@ -99,10 +94,42 @@ int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	if (r) {
 		DRM_ERROR("failed to do bo_map on static CSA, err=%d\n", r);
 		amdgpu_vm_bo_del(adev, *bo_va);
-		ttm_eu_backoff_reservation(&ticket, &list);
-		return r;
+		goto error;
 	}
 
-	ttm_eu_backoff_reservation(&ticket, &list);
-	return 0;
+error:
+	drm_exec_fini(&exec);
+	return r;
+}
+
+int amdgpu_unmap_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+			    struct amdgpu_bo *bo, struct amdgpu_bo_va *bo_va,
+			    uint64_t csa_addr)
+{
+	struct drm_exec exec;
+	int r;
+
+	drm_exec_init(&exec, 0, 0);
+	drm_exec_until_all_locked(&exec) {
+		r = amdgpu_vm_lock_pd(vm, &exec, 0);
+		if (likely(!r))
+			r = drm_exec_lock_obj(&exec, &bo->tbo.base);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(r)) {
+			DRM_ERROR("failed to reserve CSA,PD BOs: err=%d\n", r);
+			goto error;
+		}
+	}
+
+	r = amdgpu_vm_bo_unmap(adev, bo_va, csa_addr);
+	if (r) {
+		DRM_ERROR("failed to do bo_unmap on static CSA, err=%d\n", r);
+		goto error;
+	}
+
+	amdgpu_vm_bo_del(adev, bo_va);
+
+error:
+	drm_exec_fini(&exec);
+	return r;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.h
index 524b4437a021..7dfc1f2012eb 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_csa.h
@@ -34,6 +34,9 @@ int amdgpu_allocate_static_csa(struct amdgpu_device *adev, struct amdgpu_bo **bo
 int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			  struct amdgpu_bo *bo, struct amdgpu_bo_va **bo_va,
 			  uint64_t csa_addr, uint32_t size);
+int amdgpu_unmap_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+			    struct amdgpu_bo *bo, struct amdgpu_bo_va *bo_va,
+			    uint64_t csa_addr);
 void amdgpu_free_static_csa(struct amdgpu_bo **bo);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
index d2139ac12159..afedea02188d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
@@ -42,12 +42,12 @@ const unsigned int amdgpu_ctx_num_entities[AMDGPU_HW_IP_NUM] = {
 	[AMDGPU_HW_IP_VCN_DEC]	=	1,
 	[AMDGPU_HW_IP_VCN_ENC]	=	1,
 	[AMDGPU_HW_IP_VCN_JPEG]	=	1,
+	[AMDGPU_HW_IP_VPE]	=	1,
 };
 
 bool amdgpu_ctx_priority_is_valid(int32_t ctx_prio)
 {
 	switch (ctx_prio) {
-	case AMDGPU_CTX_PRIORITY_UNSET:
 	case AMDGPU_CTX_PRIORITY_VERY_LOW:
 	case AMDGPU_CTX_PRIORITY_LOW:
 	case AMDGPU_CTX_PRIORITY_NORMAL:
@@ -55,6 +55,11 @@ bool amdgpu_ctx_priority_is_valid(int32_t ctx_prio)
 	case AMDGPU_CTX_PRIORITY_VERY_HIGH:
 		return true;
 	default:
+	case AMDGPU_CTX_PRIORITY_UNSET:
+		/* UNSET priority is not valid and we don't carry that
+		 * around, but set it to NORMAL in the only place this
+		 * function is called, amdgpu_ctx_ioctl().
+		 */
 		return false;
 	}
 }
@@ -64,13 +69,14 @@ amdgpu_ctx_to_drm_sched_prio(int32_t ctx_prio)
 {
 	switch (ctx_prio) {
 	case AMDGPU_CTX_PRIORITY_UNSET:
-		return DRM_SCHED_PRIORITY_UNSET;
+		pr_warn_once("AMD-->DRM context priority value UNSET-->NORMAL");
+		return DRM_SCHED_PRIORITY_NORMAL;
 
 	case AMDGPU_CTX_PRIORITY_VERY_LOW:
-		return DRM_SCHED_PRIORITY_MIN;
+		return DRM_SCHED_PRIORITY_LOW;
 
 	case AMDGPU_CTX_PRIORITY_LOW:
-		return DRM_SCHED_PRIORITY_MIN;
+		return DRM_SCHED_PRIORITY_LOW;
 
 	case AMDGPU_CTX_PRIORITY_NORMAL:
 		return DRM_SCHED_PRIORITY_NORMAL;
@@ -94,9 +100,6 @@ amdgpu_ctx_to_drm_sched_prio(int32_t ctx_prio)
 static int amdgpu_ctx_priority_permit(struct drm_file *filp,
 				      int32_t priority)
 {
-	if (!amdgpu_ctx_priority_is_valid(priority))
-		return -EINVAL;
-
 	/* NORMAL and below are accessible by everyone */
 	if (priority <= AMDGPU_CTX_PRIORITY_NORMAL)
 		return 0;
@@ -222,8 +225,19 @@ static int amdgpu_ctx_init_entity(struct amdgpu_ctx *ctx, u32 hw_ip,
 	drm_prio = amdgpu_ctx_to_drm_sched_prio(ctx_prio);
 
 	hw_ip = array_index_nospec(hw_ip, AMDGPU_HW_IP_NUM);
-	scheds = adev->gpu_sched[hw_ip][hw_prio].sched;
-	num_scheds = adev->gpu_sched[hw_ip][hw_prio].num_scheds;
+
+	if (!(adev)->xcp_mgr) {
+		scheds = adev->gpu_sched[hw_ip][hw_prio].sched;
+		num_scheds = adev->gpu_sched[hw_ip][hw_prio].num_scheds;
+	} else {
+		struct amdgpu_fpriv *fpriv;
+
+		fpriv = container_of(ctx->ctx_mgr, struct amdgpu_fpriv, ctx_mgr);
+		r = amdgpu_xcp_select_scheds(adev, hw_ip, hw_prio, fpriv,
+						&num_scheds, &scheds);
+		if (r)
+			goto error_free_entity;
+	}
 
 	/* disable load balance if the hw engine retains context among dependent jobs */
 	if (hw_ip == AMDGPU_HW_IP_VCN_ENC ||
@@ -255,7 +269,8 @@ error_free_entity:
 	return r;
 }
 
-static ktime_t amdgpu_ctx_fini_entity(struct amdgpu_ctx_entity *entity)
+static ktime_t amdgpu_ctx_fini_entity(struct amdgpu_device *adev,
+				  struct amdgpu_ctx_entity *entity)
 {
 	ktime_t res = ns_to_ktime(0);
 	int i;
@@ -268,6 +283,8 @@ static ktime_t amdgpu_ctx_fini_entity(struct amdgpu_ctx_entity *entity)
 		dma_fence_put(entity->fences[i]);
 	}
 
+	amdgpu_xcp_release_sched(adev, entity);
+
 	kfree(entity);
 	return res;
 }
@@ -303,6 +320,7 @@ static int amdgpu_ctx_get_stable_pstate(struct amdgpu_ctx *ctx,
 static int amdgpu_ctx_init(struct amdgpu_ctx_mgr *mgr, int32_t priority,
 			   struct drm_file *filp, struct amdgpu_ctx *ctx)
 {
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
 	u32 current_stable_pstate;
 	int r;
 
@@ -318,7 +336,7 @@ static int amdgpu_ctx_init(struct amdgpu_ctx_mgr *mgr, int32_t priority,
 
 	ctx->reset_counter = atomic_read(&mgr->adev->gpu_reset_counter);
 	ctx->reset_counter_query = ctx->reset_counter;
-	ctx->vram_lost_counter = atomic_read(&mgr->adev->vram_lost_counter);
+	ctx->generation = amdgpu_vm_generation(mgr->adev, &fpriv->vm);
 	ctx->init_priority = priority;
 	ctx->override_priority = AMDGPU_CTX_PRIORITY_UNSET;
 
@@ -331,6 +349,7 @@ static int amdgpu_ctx_init(struct amdgpu_ctx_mgr *mgr, int32_t priority,
 	else
 		ctx->stable_pstate = current_stable_pstate;
 
+	ctx->ctx_mgr = &(fpriv->ctx_mgr);
 	return 0;
 }
 
@@ -399,7 +418,7 @@ static void amdgpu_ctx_fini(struct kref *ref)
 		for (j = 0; j < AMDGPU_MAX_ENTITY_NUM; ++j) {
 			ktime_t spend;
 
-			spend = amdgpu_ctx_fini_entity(ctx->entities[i][j]);
+			spend = amdgpu_ctx_fini_entity(adev, ctx->entities[i][j]);
 			atomic64_add(ktime_to_ns(spend), &mgr->time_spend[i]);
 		}
 	}
@@ -416,6 +435,7 @@ int amdgpu_ctx_get_entity(struct amdgpu_ctx *ctx, u32 hw_ip, u32 instance,
 			  u32 ring, struct drm_sched_entity **entity)
 {
 	int r;
+	struct drm_sched_entity *ctx_entity;
 
 	if (hw_ip >= AMDGPU_HW_IP_NUM) {
 		DRM_ERROR("unknown HW IP type: %d\n", hw_ip);
@@ -439,7 +459,14 @@ int amdgpu_ctx_get_entity(struct amdgpu_ctx *ctx, u32 hw_ip, u32 instance,
 			return r;
 	}
 
-	*entity = &ctx->entities[hw_ip][ring]->entity;
+	ctx_entity = &ctx->entities[hw_ip][ring]->entity;
+	r = drm_sched_entity_error(ctx_entity);
+	if (r) {
+		DRM_DEBUG("error entity %p\n", ctx_entity);
+		return r;
+	}
+
+	*entity = ctx_entity;
 	return 0;
 }
 
@@ -570,12 +597,15 @@ static int amdgpu_ctx_query2(struct amdgpu_device *adev,
 	if (ctx->reset_counter != atomic_read(&adev->gpu_reset_counter))
 		out->state.flags |= AMDGPU_CTX_QUERY2_FLAGS_RESET;
 
-	if (ctx->vram_lost_counter != atomic_read(&adev->vram_lost_counter))
+	if (ctx->generation != amdgpu_vm_generation(adev, &fpriv->vm))
 		out->state.flags |= AMDGPU_CTX_QUERY2_FLAGS_VRAMLOST;
 
 	if (atomic_read(&ctx->guilty))
 		out->state.flags |= AMDGPU_CTX_QUERY2_FLAGS_GUILTY;
 
+	if (amdgpu_in_reset(adev))
+		out->state.flags |= AMDGPU_CTX_QUERY2_FLAGS_RESET_IN_PROGRESS;
+
 	if (adev->ras_enabled && con) {
 		/* Return the cached values in O(1),
 		 * and schedule delayed work to cache
@@ -604,8 +634,6 @@ static int amdgpu_ctx_query2(struct amdgpu_device *adev,
 	return 0;
 }
 
-
-
 static int amdgpu_ctx_stable_pstate(struct amdgpu_device *adev,
 				    struct amdgpu_fpriv *fpriv, uint32_t id,
 				    bool set, u32 *stable_pstate)
@@ -648,23 +676,33 @@ int amdgpu_ctx_ioctl(struct drm_device *dev, void *data,
 	id = args->in.ctx_id;
 	priority = args->in.priority;
 
-	/* For backwards compatibility reasons, we need to accept
-	 * ioctls with garbage in the priority field */
+	/* For backwards compatibility, we need to accept ioctls with garbage
+	 * in the priority field. Garbage values in the priority field, result
+	 * in the priority being set to NORMAL.
+	 */
 	if (!amdgpu_ctx_priority_is_valid(priority))
 		priority = AMDGPU_CTX_PRIORITY_NORMAL;
 
 	switch (args->in.op) {
 	case AMDGPU_CTX_OP_ALLOC_CTX:
+		if (args->in.flags)
+			return -EINVAL;
 		r = amdgpu_ctx_alloc(adev, fpriv, filp, priority, &id);
 		args->out.alloc.ctx_id = id;
 		break;
 	case AMDGPU_CTX_OP_FREE_CTX:
+		if (args->in.flags)
+			return -EINVAL;
 		r = amdgpu_ctx_free(fpriv, id);
 		break;
 	case AMDGPU_CTX_OP_QUERY_STATE:
+		if (args->in.flags)
+			return -EINVAL;
 		r = amdgpu_ctx_query(adev, fpriv, id, &args->out);
 		break;
 	case AMDGPU_CTX_OP_QUERY_STATE2:
+		if (args->in.flags)
+			return -EINVAL;
 		r = amdgpu_ctx_query2(adev, fpriv, id, &args->out);
 		break;
 	case AMDGPU_CTX_OP_GET_STABLE_PSTATE:
@@ -881,7 +919,7 @@ long amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr, long timeout)
 	return timeout;
 }
 
-void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr)
+static void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr)
 {
 	struct amdgpu_ctx *ctx;
 	struct idr *idp;
@@ -906,24 +944,13 @@ void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr)
 				drm_sched_entity_fini(entity);
 			}
 		}
+		kref_put(&ctx->refcount, amdgpu_ctx_fini);
 	}
 }
 
 void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr)
 {
-	struct amdgpu_ctx *ctx;
-	struct idr *idp;
-	uint32_t id;
-
 	amdgpu_ctx_mgr_entity_fini(mgr);
-
-	idp = &mgr->ctx_handles;
-
-	idr_for_each_entry(idp, ctx, id) {
-		if (kref_put(&ctx->refcount, amdgpu_ctx_fini) != 1)
-			DRM_ERROR("ctx %p is still alive\n", ctx);
-	}
-
 	idr_destroy(&mgr->ctx_handles);
 	mutex_destroy(&mgr->lock);
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
index 0fa0e56daf67..090dfe86f75b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
@@ -47,7 +47,7 @@ struct amdgpu_ctx {
 	struct amdgpu_ctx_mgr		*mgr;
 	unsigned			reset_counter;
 	unsigned			reset_counter_query;
-	uint32_t			vram_lost_counter;
+	uint64_t			generation;
 	spinlock_t			ring_lock;
 	struct amdgpu_ctx_entity	*entities[AMDGPU_HW_IP_NUM][AMDGPU_MAX_ENTITY_NUM];
 	bool				preamble_presented;
@@ -57,6 +57,7 @@ struct amdgpu_ctx {
 	unsigned long			ras_counter_ce;
 	unsigned long			ras_counter_ue;
 	uint32_t			stable_pstate;
+	struct amdgpu_ctx_mgr		*ctx_mgr;
 };
 
 struct amdgpu_ctx_mgr {
@@ -91,7 +92,6 @@ int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx,
 
 void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr,
 			 struct amdgpu_device *adev);
-void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr);
 long amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr, long timeout);
 void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr);
 void amdgpu_ctx_mgr_usage(struct amdgpu_ctx_mgr *mgr,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
index f60753f97ac5..62d43b8cbe58 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
@@ -56,14 +56,14 @@
  *
  * Bit 62:  Indicates a GRBM bank switch is needed
  * Bit 61:  Indicates a SRBM bank switch is needed (implies bit 62 is
- * 	    zero)
+ *	    zero)
  * Bits 24..33: The SE or ME selector if needed
  * Bits 34..43: The SH (or SA) or PIPE selector if needed
  * Bits 44..53: The INSTANCE (or CU/WGP) or QUEUE selector if needed
  *
  * Bit 23:  Indicates that the PM power gating lock should be held
- * 	    This is necessary to read registers that might be
- * 	    unreliable during a power gating transistion.
+ *	    This is necessary to read registers that might be
+ *	    unreliable during a power gating transistion.
  *
  * The lower bits are the BYTE offset of the register to read.  This
  * allows reading multiple registers in a single call and having
@@ -76,7 +76,7 @@ static int  amdgpu_debugfs_process_reg_op(bool read, struct file *f,
 	ssize_t result = 0;
 	int r;
 	bool pm_pg_lock, use_bank, use_ring;
-	unsigned instance_bank, sh_bank, se_bank, me, pipe, queue, vmid;
+	unsigned int instance_bank, sh_bank, se_bank, me, pipe, queue, vmid;
 
 	pm_pg_lock = use_bank = use_ring = false;
 	instance_bank = sh_bank = se_bank = me = pipe = queue = vmid = 0;
@@ -129,17 +129,16 @@ static int  amdgpu_debugfs_process_reg_op(bool read, struct file *f,
 	if (use_bank) {
 		if ((sh_bank != 0xFFFFFFFF && sh_bank >= adev->gfx.config.max_sh_per_se) ||
 		    (se_bank != 0xFFFFFFFF && se_bank >= adev->gfx.config.max_shader_engines)) {
-			pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 			pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 			amdgpu_virt_disable_access_debugfs(adev);
 			return -EINVAL;
 		}
 		mutex_lock(&adev->grbm_idx_mutex);
 		amdgpu_gfx_select_se_sh(adev, se_bank,
-					sh_bank, instance_bank);
+					sh_bank, instance_bank, 0);
 	} else if (use_ring) {
 		mutex_lock(&adev->srbm_mutex);
-		amdgpu_gfx_select_me_pipe_q(adev, me, pipe, queue, vmid);
+		amdgpu_gfx_select_me_pipe_q(adev, me, pipe, queue, vmid, 0);
 	}
 
 	if (pm_pg_lock)
@@ -154,7 +153,7 @@ static int  amdgpu_debugfs_process_reg_op(bool read, struct file *f,
 		} else {
 			r = get_user(value, (uint32_t *)buf);
 			if (!r)
-				amdgpu_mm_wreg_mmio_rlc(adev, *pos >> 2, value);
+				amdgpu_mm_wreg_mmio_rlc(adev, *pos >> 2, value, 0);
 		}
 		if (r) {
 			result = r;
@@ -169,17 +168,16 @@ static int  amdgpu_debugfs_process_reg_op(bool read, struct file *f,
 
 end:
 	if (use_bank) {
-		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 		mutex_unlock(&adev->grbm_idx_mutex);
 	} else if (use_ring) {
-		amdgpu_gfx_select_me_pipe_q(adev, 0, 0, 0, 0);
+		amdgpu_gfx_select_me_pipe_q(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 	}
 
 	if (pm_pg_lock)
 		mutex_unlock(&adev->pm.mutex);
 
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	amdgpu_virt_disable_access_debugfs(adev);
@@ -208,7 +206,7 @@ static int amdgpu_debugfs_regs2_open(struct inode *inode, struct file *file)
 {
 	struct amdgpu_debugfs_regs2_data *rd;
 
-	rd = kzalloc(sizeof *rd, GFP_KERNEL);
+	rd = kzalloc(sizeof(*rd), GFP_KERNEL);
 	if (!rd)
 		return -ENOMEM;
 	rd->adev = file_inode(file)->i_private;
@@ -221,6 +219,7 @@ static int amdgpu_debugfs_regs2_open(struct inode *inode, struct file *file)
 static int amdgpu_debugfs_regs2_release(struct inode *inode, struct file *file)
 {
 	struct amdgpu_debugfs_regs2_data *rd = file->private_data;
+
 	mutex_destroy(&rd->lock);
 	kfree(file->private_data);
 	return 0;
@@ -254,7 +253,6 @@ static ssize_t amdgpu_debugfs_regs2_op(struct file *f, char __user *buf, u32 off
 	if (rd->id.use_grbm) {
 		if ((rd->id.grbm.sh != 0xFFFFFFFF && rd->id.grbm.sh >= adev->gfx.config.max_sh_per_se) ||
 		    (rd->id.grbm.se != 0xFFFFFFFF && rd->id.grbm.se >= adev->gfx.config.max_shader_engines)) {
-			pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 			pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 			amdgpu_virt_disable_access_debugfs(adev);
 			mutex_unlock(&rd->lock);
@@ -262,14 +260,14 @@ static ssize_t amdgpu_debugfs_regs2_op(struct file *f, char __user *buf, u32 off
 		}
 		mutex_lock(&adev->grbm_idx_mutex);
 		amdgpu_gfx_select_se_sh(adev, rd->id.grbm.se,
-								rd->id.grbm.sh,
-								rd->id.grbm.instance);
+						  rd->id.grbm.sh,
+						  rd->id.grbm.instance, rd->id.xcc_id);
 	}
 
 	if (rd->id.use_srbm) {
 		mutex_lock(&adev->srbm_mutex);
 		amdgpu_gfx_select_me_pipe_q(adev, rd->id.srbm.me, rd->id.srbm.pipe,
-									rd->id.srbm.queue, rd->id.srbm.vmid);
+					    rd->id.srbm.queue, rd->id.srbm.vmid, rd->id.xcc_id);
 	}
 
 	if (rd->id.pg_lock)
@@ -282,7 +280,7 @@ static ssize_t amdgpu_debugfs_regs2_op(struct file *f, char __user *buf, u32 off
 		} else {
 			r = get_user(value, (uint32_t *)buf);
 			if (!r)
-				amdgpu_mm_wreg_mmio_rlc(adev, offset >> 2, value);
+				amdgpu_mm_wreg_mmio_rlc(adev, offset >> 2, value, rd->id.xcc_id);
 		}
 		if (r) {
 			result = r;
@@ -295,12 +293,12 @@ static ssize_t amdgpu_debugfs_regs2_op(struct file *f, char __user *buf, u32 off
 	}
 end:
 	if (rd->id.use_grbm) {
-		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, rd->id.xcc_id);
 		mutex_unlock(&adev->grbm_idx_mutex);
 	}
 
 	if (rd->id.use_srbm) {
-		amdgpu_gfx_select_me_pipe_q(adev, 0, 0, 0, 0);
+		amdgpu_gfx_select_me_pipe_q(adev, 0, 0, 0, 0, rd->id.xcc_id);
 		mutex_unlock(&adev->srbm_mutex);
 	}
 
@@ -309,7 +307,6 @@ end:
 
 	mutex_unlock(&rd->lock);
 
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	amdgpu_virt_disable_access_debugfs(adev);
@@ -319,18 +316,45 @@ end:
 static long amdgpu_debugfs_regs2_ioctl(struct file *f, unsigned int cmd, unsigned long data)
 {
 	struct amdgpu_debugfs_regs2_data *rd = f->private_data;
+	struct amdgpu_debugfs_regs2_iocdata v1_data;
 	int r;
 
+	mutex_lock(&rd->lock);
+
 	switch (cmd) {
+	case AMDGPU_DEBUGFS_REGS2_IOC_SET_STATE_V2:
+		r = copy_from_user(&rd->id, (struct amdgpu_debugfs_regs2_iocdata_v2 *)data,
+				   sizeof(rd->id));
+		if (r)
+			r = -EINVAL;
+		goto done;
 	case AMDGPU_DEBUGFS_REGS2_IOC_SET_STATE:
-		mutex_lock(&rd->lock);
-		r = copy_from_user(&rd->id, (struct amdgpu_debugfs_regs2_iocdata *)data, sizeof rd->id);
-		mutex_unlock(&rd->lock);
-		return r ? -EINVAL : 0;
+		r = copy_from_user(&v1_data, (struct amdgpu_debugfs_regs2_iocdata *)data,
+				   sizeof(v1_data));
+		if (r) {
+			r = -EINVAL;
+			goto done;
+		}
+		goto v1_copy;
 	default:
-		return -EINVAL;
-	}
-	return 0;
+		r = -EINVAL;
+		goto done;
+	}
+
+v1_copy:
+	rd->id.use_srbm = v1_data.use_srbm;
+	rd->id.use_grbm = v1_data.use_grbm;
+	rd->id.pg_lock = v1_data.pg_lock;
+	rd->id.grbm.se = v1_data.grbm.se;
+	rd->id.grbm.sh = v1_data.grbm.sh;
+	rd->id.grbm.instance = v1_data.grbm.instance;
+	rd->id.srbm.me = v1_data.srbm.me;
+	rd->id.srbm.pipe = v1_data.srbm.pipe;
+	rd->id.srbm.queue = v1_data.srbm.queue;
+	rd->id.xcc_id = 0;
+done:
+	mutex_unlock(&rd->lock);
+	return r;
 }
 
 static ssize_t amdgpu_debugfs_regs2_read(struct file *f, char __user *buf, size_t size, loff_t *pos)
@@ -343,6 +367,136 @@ static ssize_t amdgpu_debugfs_regs2_write(struct file *f, const char __user *buf
 	return amdgpu_debugfs_regs2_op(f, (char __user *)buf, *pos, size, 1);
 }
 
+static int amdgpu_debugfs_gprwave_open(struct inode *inode, struct file *file)
+{
+	struct amdgpu_debugfs_gprwave_data *rd;
+
+	rd = kzalloc(sizeof(*rd), GFP_KERNEL);
+	if (!rd)
+		return -ENOMEM;
+	rd->adev = file_inode(file)->i_private;
+	file->private_data = rd;
+	mutex_init(&rd->lock);
+
+	return 0;
+}
+
+static int amdgpu_debugfs_gprwave_release(struct inode *inode, struct file *file)
+{
+	struct amdgpu_debugfs_gprwave_data *rd = file->private_data;
+
+	mutex_destroy(&rd->lock);
+	kfree(file->private_data);
+	return 0;
+}
+
+static ssize_t amdgpu_debugfs_gprwave_read(struct file *f, char __user *buf, size_t size, loff_t *pos)
+{
+	struct amdgpu_debugfs_gprwave_data *rd = f->private_data;
+	struct amdgpu_device *adev = rd->adev;
+	ssize_t result = 0;
+	int r;
+	uint32_t *data, x;
+
+	if (size > 4096 || size & 0x3 || *pos & 0x3)
+		return -EINVAL;
+
+	r = pm_runtime_get_sync(adev_to_drm(adev)->dev);
+	if (r < 0) {
+		pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
+		return r;
+	}
+
+	r = amdgpu_virt_enable_access_debugfs(adev);
+	if (r < 0) {
+		pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
+		return r;
+	}
+
+	data = kcalloc(1024, sizeof(*data), GFP_KERNEL);
+	if (!data) {
+		pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
+		amdgpu_virt_disable_access_debugfs(adev);
+		return -ENOMEM;
+	}
+
+	/* switch to the specific se/sh/cu */
+	mutex_lock(&adev->grbm_idx_mutex);
+	amdgpu_gfx_select_se_sh(adev, rd->id.se, rd->id.sh, rd->id.cu, rd->id.xcc_id);
+
+	if (!rd->id.gpr_or_wave) {
+		x = 0;
+		if (adev->gfx.funcs->read_wave_data)
+			adev->gfx.funcs->read_wave_data(adev, rd->id.xcc_id, rd->id.simd, rd->id.wave, data, &x);
+	} else {
+		x = size >> 2;
+		if (rd->id.gpr.vpgr_or_sgpr) {
+			if (adev->gfx.funcs->read_wave_vgprs)
+				adev->gfx.funcs->read_wave_vgprs(adev, rd->id.xcc_id, rd->id.simd, rd->id.wave, rd->id.gpr.thread, *pos, size>>2, data);
+		} else {
+			if (adev->gfx.funcs->read_wave_sgprs)
+				adev->gfx.funcs->read_wave_sgprs(adev, rd->id.xcc_id, rd->id.simd, rd->id.wave, *pos, size>>2, data);
+		}
+	}
+
+	amdgpu_gfx_select_se_sh(adev, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, rd->id.xcc_id);
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
+
+	if (!x) {
+		result = -EINVAL;
+		goto done;
+	}
+
+	while (size && (*pos < x * 4)) {
+		uint32_t value;
+
+		value = data[*pos >> 2];
+		r = put_user(value, (uint32_t *)buf);
+		if (r) {
+			result = r;
+			goto done;
+		}
+
+		result += 4;
+		buf += 4;
+		*pos += 4;
+		size -= 4;
+	}
+
+done:
+	amdgpu_virt_disable_access_debugfs(adev);
+	kfree(data);
+	return result;
+}
+
+static long amdgpu_debugfs_gprwave_ioctl(struct file *f, unsigned int cmd, unsigned long data)
+{
+	struct amdgpu_debugfs_gprwave_data *rd = f->private_data;
+	int r = 0;
+
+	mutex_lock(&rd->lock);
+
+	switch (cmd) {
+	case AMDGPU_DEBUGFS_GPRWAVE_IOC_SET_STATE:
+		if (copy_from_user(&rd->id,
+				   (struct amdgpu_debugfs_gprwave_iocdata *)data,
+				   sizeof(rd->id)))
+			r = -EFAULT;
+		goto done;
+	default:
+		r = -EINVAL;
+		goto done;
+	}
+
+done:
+	mutex_unlock(&rd->lock);
+	return r;
+}
+
+
+
 
 /**
  * amdgpu_debugfs_regs_pcie_read - Read from a PCIE register
@@ -381,7 +535,11 @@ static ssize_t amdgpu_debugfs_regs_pcie_read(struct file *f, char __user *buf,
 	while (size) {
 		uint32_t value;
 
-		value = RREG32_PCIE(*pos);
+		if (upper_32_bits(*pos))
+			value = RREG32_PCIE_EXT(*pos);
+		else
+			value = RREG32_PCIE(*pos);
+
 		r = put_user(value, (uint32_t *)buf);
 		if (r)
 			goto out;
@@ -394,7 +552,6 @@ static ssize_t amdgpu_debugfs_regs_pcie_read(struct file *f, char __user *buf,
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 	amdgpu_virt_disable_access_debugfs(adev);
 	return r;
@@ -441,7 +598,10 @@ static ssize_t amdgpu_debugfs_regs_pcie_write(struct file *f, const char __user
 		if (r)
 			goto out;
 
-		WREG32_PCIE(*pos, value);
+		if (upper_32_bits(*pos))
+			WREG32_PCIE_EXT(*pos, value);
+		else
+			WREG32_PCIE(*pos, value);
 
 		result += 4;
 		buf += 4;
@@ -451,7 +611,6 @@ static ssize_t amdgpu_debugfs_regs_pcie_write(struct file *f, const char __user
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 	amdgpu_virt_disable_access_debugfs(adev);
 	return r;
@@ -479,6 +638,9 @@ static ssize_t amdgpu_debugfs_regs_didt_read(struct file *f, char __user *buf,
 	if (size & 0x3 || *pos & 0x3)
 		return -EINVAL;
 
+	if (!adev->didt_rreg)
+		return -EOPNOTSUPP;
+
 	r = pm_runtime_get_sync(adev_to_drm(adev)->dev);
 	if (r < 0) {
 		pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
@@ -507,7 +669,6 @@ static ssize_t amdgpu_debugfs_regs_didt_read(struct file *f, char __user *buf,
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 	amdgpu_virt_disable_access_debugfs(adev);
 	return r;
@@ -535,6 +696,9 @@ static ssize_t amdgpu_debugfs_regs_didt_write(struct file *f, const char __user
 	if (size & 0x3 || *pos & 0x3)
 		return -EINVAL;
 
+	if (!adev->didt_wreg)
+		return -EOPNOTSUPP;
+
 	r = pm_runtime_get_sync(adev_to_drm(adev)->dev);
 	if (r < 0) {
 		pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
@@ -564,7 +728,6 @@ static ssize_t amdgpu_debugfs_regs_didt_write(struct file *f, const char __user
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 	amdgpu_virt_disable_access_debugfs(adev);
 	return r;
@@ -589,6 +752,9 @@ static ssize_t amdgpu_debugfs_regs_smc_read(struct file *f, char __user *buf,
 	ssize_t result = 0;
 	int r;
 
+	if (!adev->smc_rreg)
+		return -EOPNOTSUPP;
+
 	if (size & 0x3 || *pos & 0x3)
 		return -EINVAL;
 
@@ -620,7 +786,6 @@ static ssize_t amdgpu_debugfs_regs_smc_read(struct file *f, char __user *buf,
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 	amdgpu_virt_disable_access_debugfs(adev);
 	return r;
@@ -645,6 +810,9 @@ static ssize_t amdgpu_debugfs_regs_smc_write(struct file *f, const char __user *
 	ssize_t result = 0;
 	int r;
 
+	if (!adev->smc_wreg)
+		return -EOPNOTSUPP;
+
 	if (size & 0x3 || *pos & 0x3)
 		return -EINVAL;
 
@@ -677,7 +845,6 @@ static ssize_t amdgpu_debugfs_regs_smc_write(struct file *f, const char __user *
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 	amdgpu_virt_disable_access_debugfs(adev);
 	return r;
@@ -740,7 +907,7 @@ static ssize_t amdgpu_debugfs_gca_config_read(struct file *f, char __user *buf,
 
 	/* rev==1 */
 	config[no_regs++] = adev->rev_id;
-	config[no_regs++] = lower_32_bits(adev->pg_flags);
+	config[no_regs++] = adev->pg_flags;
 	config[no_regs++] = lower_32_bits(adev->cg_flags);
 
 	/* rev==2 */
@@ -757,7 +924,7 @@ static ssize_t amdgpu_debugfs_gca_config_read(struct file *f, char __user *buf,
 	config[no_regs++] = adev->flags & AMD_IS_APU ? 1 : 0;
 
 	/* rev==5 PG/CG flag upper 32bit */
-	config[no_regs++] = upper_32_bits(adev->pg_flags);
+	config[no_regs++] = 0;
 	config[no_regs++] = upper_32_bits(adev->cg_flags);
 
 	while (size && (*pos < no_regs * 4)) {
@@ -825,7 +992,6 @@ static ssize_t amdgpu_debugfs_sensor_read(struct file *f, char __user *buf,
 
 	r = amdgpu_dpm_read_sensor(adev, idx, &values[0], &valuesize);
 
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	if (r) {
@@ -863,7 +1029,7 @@ static ssize_t amdgpu_debugfs_sensor_read(struct file *f, char __user *buf,
  * The offset being sought changes which wave that the status data
  * will be returned for.  The bits are used as follows:
  *
- * Bits 0..6: 	Byte offset into data
+ * Bits 0..6:	Byte offset into data
  * Bits 7..14:	SE selector
  * Bits 15..22:	SH/SA selector
  * Bits 23..30: CU/{WGP+SIMD} selector
@@ -907,16 +1073,15 @@ static ssize_t amdgpu_debugfs_wave_read(struct file *f, char __user *buf,
 
 	/* switch to the specific se/sh/cu */
 	mutex_lock(&adev->grbm_idx_mutex);
-	amdgpu_gfx_select_se_sh(adev, se, sh, cu);
+	amdgpu_gfx_select_se_sh(adev, se, sh, cu, 0);
 
 	x = 0;
 	if (adev->gfx.funcs->read_wave_data)
-		adev->gfx.funcs->read_wave_data(adev, simd, wave, data, &x);
+		adev->gfx.funcs->read_wave_data(adev, 0, simd, wave, data, &x);
 
-	amdgpu_gfx_select_se_sh(adev, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF);
+	amdgpu_gfx_select_se_sh(adev, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	if (!x) {
@@ -1001,20 +1166,19 @@ static ssize_t amdgpu_debugfs_gpr_read(struct file *f, char __user *buf,
 
 	/* switch to the specific se/sh/cu */
 	mutex_lock(&adev->grbm_idx_mutex);
-	amdgpu_gfx_select_se_sh(adev, se, sh, cu);
+	amdgpu_gfx_select_se_sh(adev, se, sh, cu, 0);
 
 	if (bank == 0) {
 		if (adev->gfx.funcs->read_wave_vgprs)
-			adev->gfx.funcs->read_wave_vgprs(adev, simd, wave, thread, offset, size>>2, data);
+			adev->gfx.funcs->read_wave_vgprs(adev, 0, simd, wave, thread, offset, size>>2, data);
 	} else {
 		if (adev->gfx.funcs->read_wave_sgprs)
-			adev->gfx.funcs->read_wave_sgprs(adev, simd, wave, offset, size>>2, data);
+			adev->gfx.funcs->read_wave_sgprs(adev, 0, simd, wave, offset, size>>2, data);
 	}
 
-	amdgpu_gfx_select_se_sh(adev, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF);
+	amdgpu_gfx_select_se_sh(adev, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	while (size) {
@@ -1088,7 +1252,6 @@ static ssize_t amdgpu_debugfs_gfxoff_residency_read(struct file *f, char __user
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	return r;
@@ -1137,7 +1300,6 @@ static ssize_t amdgpu_debugfs_gfxoff_residency_write(struct file *f, const char
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	return r;
@@ -1187,7 +1349,6 @@ static ssize_t amdgpu_debugfs_gfxoff_count_read(struct file *f, char __user *buf
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	return r;
@@ -1236,7 +1397,6 @@ static ssize_t amdgpu_debugfs_gfxoff_write(struct file *f, const char __user *bu
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	return r;
@@ -1282,7 +1442,6 @@ static ssize_t amdgpu_debugfs_gfxoff_read(struct file *f, char __user *buf,
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	return r;
@@ -1323,7 +1482,6 @@ static ssize_t amdgpu_debugfs_gfxoff_status_read(struct file *f, char __user *bu
 
 	r = result;
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	return r;
@@ -1339,6 +1497,15 @@ static const struct file_operations amdgpu_debugfs_regs2_fops = {
 	.llseek = default_llseek
 };
 
+static const struct file_operations amdgpu_debugfs_gprwave_fops = {
+	.owner = THIS_MODULE,
+	.unlocked_ioctl = amdgpu_debugfs_gprwave_ioctl,
+	.read = amdgpu_debugfs_gprwave_read,
+	.open = amdgpu_debugfs_gprwave_open,
+	.release = amdgpu_debugfs_gprwave_release,
+	.llseek = default_llseek
+};
+
 static const struct file_operations amdgpu_debugfs_regs_fops = {
 	.owner = THIS_MODULE,
 	.read = amdgpu_debugfs_regs_read,
@@ -1416,6 +1583,7 @@ static const struct file_operations amdgpu_debugfs_gfxoff_residency_fops = {
 static const struct file_operations *debugfs_regs[] = {
 	&amdgpu_debugfs_regs_fops,
 	&amdgpu_debugfs_regs2_fops,
+	&amdgpu_debugfs_gprwave_fops,
 	&amdgpu_debugfs_regs_didt_fops,
 	&amdgpu_debugfs_regs_pcie_fops,
 	&amdgpu_debugfs_regs_smc_fops,
@@ -1429,9 +1597,10 @@ static const struct file_operations *debugfs_regs[] = {
 	&amdgpu_debugfs_gfxoff_residency_fops,
 };
 
-static const char *debugfs_regs_names[] = {
+static const char * const debugfs_regs_names[] = {
 	"amdgpu_regs",
 	"amdgpu_regs2",
+	"amdgpu_gprwave",
 	"amdgpu_regs_didt",
 	"amdgpu_regs_pcie",
 	"amdgpu_regs_smc",
@@ -1447,7 +1616,7 @@ static const char *debugfs_regs_names[] = {
 
 /**
  * amdgpu_debugfs_regs_init -	Initialize debugfs entries that provide
- * 				register access.
+ *				register access.
  *
  * @adev: The device to attach the debugfs entries to
  */
@@ -1459,7 +1628,7 @@ int amdgpu_debugfs_regs_init(struct amdgpu_device *adev)
 
 	for (i = 0; i < ARRAY_SIZE(debugfs_regs); i++) {
 		ent = debugfs_create_file(debugfs_regs_names[i],
-					  S_IFREG | S_IRUGO, root,
+					  S_IFREG | 0400, root,
 					  adev, debugfs_regs[i]);
 		if (!i && !IS_ERR_OR_NULL(ent))
 			i_size_write(ent->d_inode, adev->rmmio_size);
@@ -1470,7 +1639,7 @@ int amdgpu_debugfs_regs_init(struct amdgpu_device *adev)
 
 static int amdgpu_debugfs_test_ib_show(struct seq_file *m, void *unused)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct amdgpu_device *adev = m->private;
 	struct drm_device *dev = adev_to_drm(adev);
 	int r = 0, i;
 
@@ -1489,30 +1658,29 @@ static int amdgpu_debugfs_test_ib_show(struct seq_file *m, void *unused)
 	for (i = 0; i < AMDGPU_MAX_RINGS; i++) {
 		struct amdgpu_ring *ring = adev->rings[i];
 
-		if (!ring || !ring->sched.thread)
+		if (!amdgpu_ring_sched_ready(ring))
 			continue;
-		kthread_park(ring->sched.thread);
+		drm_sched_wqueue_stop(&ring->sched);
 	}
 
-	seq_printf(m, "run ib test:\n");
+	seq_puts(m, "run ib test:\n");
 	r = amdgpu_ib_ring_tests(adev);
 	if (r)
 		seq_printf(m, "ib ring tests failed (%d).\n", r);
 	else
-		seq_printf(m, "ib ring tests passed.\n");
+		seq_puts(m, "ib ring tests passed.\n");
 
 	/* go on the scheduler */
 	for (i = 0; i < AMDGPU_MAX_RINGS; i++) {
 		struct amdgpu_ring *ring = adev->rings[i];
 
-		if (!ring || !ring->sched.thread)
+		if (!amdgpu_ring_sched_ready(ring))
 			continue;
-		kthread_unpark(ring->sched.thread);
+		drm_sched_wqueue_start(&ring->sched);
 	}
 
 	up_write(&adev->reset_domain->sem);
 
-	pm_runtime_mark_last_busy(dev->dev);
 	pm_runtime_put_autosuspend(dev->dev);
 
 	return 0;
@@ -1532,7 +1700,6 @@ static int amdgpu_debugfs_evict_vram(void *data, u64 *val)
 
 	*val = amdgpu_ttm_evict_resources(adev, TTM_PL_VRAM);
 
-	pm_runtime_mark_last_busy(dev->dev);
 	pm_runtime_put_autosuspend(dev->dev);
 
 	return 0;
@@ -1553,7 +1720,6 @@ static int amdgpu_debugfs_evict_gtt(void *data, u64 *val)
 
 	*val = amdgpu_ttm_evict_resources(adev, TTM_PL_TT);
 
-	pm_runtime_mark_last_busy(dev->dev);
 	pm_runtime_put_autosuspend(dev->dev);
 
 	return 0;
@@ -1573,7 +1739,6 @@ static int amdgpu_debugfs_benchmark(void *data, u64 val)
 
 	r = amdgpu_benchmark(adev, val);
 
-	pm_runtime_mark_last_busy(dev->dev);
 	pm_runtime_put_autosuspend(dev->dev);
 
 	return r;
@@ -1581,7 +1746,7 @@ static int amdgpu_debugfs_benchmark(void *data, u64 val)
 
 static int amdgpu_debugfs_vm_info_show(struct seq_file *m, void *unused)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct amdgpu_device *adev = m->private;
 	struct drm_device *dev = adev_to_drm(adev);
 	struct drm_file *file;
 	int r;
@@ -1593,9 +1758,14 @@ static int amdgpu_debugfs_vm_info_show(struct seq_file *m, void *unused)
 	list_for_each_entry(file, &dev->filelist, lhead) {
 		struct amdgpu_fpriv *fpriv = file->driver_priv;
 		struct amdgpu_vm *vm = &fpriv->vm;
+		struct amdgpu_task_info *ti;
+
+		ti = amdgpu_vm_get_task_info_vm(vm);
+		if (ti) {
+			seq_printf(m, "pid:%d\tProcess:%s ----------\n", ti->task.pid, ti->process_name);
+			amdgpu_vm_put_task_info(ti);
+		}
 
-		seq_printf(m, "pid:%d\tProcess:%s ----------\n",
-				vm->task_info.pid, vm->task_info.process_name);
 		r = amdgpu_bo_reserve(vm->root.bo, true);
 		if (r)
 			break;
@@ -1708,7 +1878,7 @@ no_preempt:
 			continue;
 		}
 		job = to_amdgpu_job(s_job);
-		if (preempted && (&job->hw_fence) == fence)
+		if (preempted && (&job->hw_fence->base) == fence)
 			/* mark the job as preempted */
 			job->preemption_status |= AMDGPU_IB_PREEMPTED;
 	}
@@ -1727,7 +1897,8 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val)
 
 	ring = adev->rings[val];
 
-	if (!ring || !ring->funcs->preempt_ib || !ring->sched.thread)
+	if (!amdgpu_ring_sched_ready(ring) ||
+	    !ring->funcs->preempt_ib)
 		return -EINVAL;
 
 	/* the last preemption failed */
@@ -1745,7 +1916,7 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val)
 		goto pro_end;
 
 	/* stop the scheduler */
-	kthread_park(ring->sched.thread);
+	drm_sched_wqueue_stop(&ring->sched);
 
 	/* preempt the IB */
 	r = amdgpu_ring_preempt_ib(ring);
@@ -1779,7 +1950,7 @@ static int amdgpu_debugfs_ib_preempt(void *data, u64 val)
 
 failure:
 	/* restart the scheduler */
-	kthread_unpark(ring->sched.thread);
+	drm_sched_wqueue_start(&ring->sched);
 
 	up_read(&adev->reset_domain->sem);
 
@@ -1795,7 +1966,7 @@ static int amdgpu_debugfs_sclk_set(void *data, u64 val)
 	uint32_t max_freq, min_freq;
 	struct amdgpu_device *adev = (struct amdgpu_device *)data;
 
-	if (amdgpu_sriov_vf(adev) && !amdgpu_sriov_is_pp_one_vf(adev))
+	if (amdgpu_sriov_multi_vf_mode(adev))
 		return -EINVAL;
 
 	ret = pm_runtime_get_sync(adev_to_drm(adev)->dev);
@@ -1819,7 +1990,6 @@ static int amdgpu_debugfs_sclk_set(void *data, u64 val)
 		ret = -EINVAL;
 
 out:
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 
 	return ret;
@@ -1831,99 +2001,6 @@ DEFINE_DEBUGFS_ATTRIBUTE(fops_ib_preempt, NULL,
 DEFINE_DEBUGFS_ATTRIBUTE(fops_sclk_set, NULL,
 			amdgpu_debugfs_sclk_set, "%llu\n");
 
-static ssize_t amdgpu_reset_dump_register_list_read(struct file *f,
-				char __user *buf, size_t size, loff_t *pos)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(f)->i_private;
-	char reg_offset[12];
-	int i, ret, len = 0;
-
-	if (*pos)
-		return 0;
-
-	memset(reg_offset, 0, 12);
-	ret = down_read_killable(&adev->reset_domain->sem);
-	if (ret)
-		return ret;
-
-	for (i = 0; i < adev->num_regs; i++) {
-		sprintf(reg_offset, "0x%x\n", adev->reset_dump_reg_list[i]);
-		up_read(&adev->reset_domain->sem);
-		if (copy_to_user(buf + len, reg_offset, strlen(reg_offset)))
-			return -EFAULT;
-
-		len += strlen(reg_offset);
-		ret = down_read_killable(&adev->reset_domain->sem);
-		if (ret)
-			return ret;
-	}
-
-	up_read(&adev->reset_domain->sem);
-	*pos += len;
-
-	return len;
-}
-
-static ssize_t amdgpu_reset_dump_register_list_write(struct file *f,
-			const char __user *buf, size_t size, loff_t *pos)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(f)->i_private;
-	char reg_offset[11];
-	uint32_t *new = NULL, *tmp = NULL;
-	int ret, i = 0, len = 0;
-
-	do {
-		memset(reg_offset, 0, 11);
-		if (copy_from_user(reg_offset, buf + len,
-					min(10, ((int)size-len)))) {
-			ret = -EFAULT;
-			goto error_free;
-		}
-
-		new = krealloc_array(tmp, i + 1, sizeof(uint32_t), GFP_KERNEL);
-		if (!new) {
-			ret = -ENOMEM;
-			goto error_free;
-		}
-		tmp = new;
-		if (sscanf(reg_offset, "%X %n", &tmp[i], &ret) != 1) {
-			ret = -EINVAL;
-			goto error_free;
-		}
-
-		len += ret;
-		i++;
-	} while (len < size);
-
-	new = kmalloc_array(i, sizeof(uint32_t), GFP_KERNEL);
-	if (!new) {
-		ret = -ENOMEM;
-		goto error_free;
-	}
-	ret = down_write_killable(&adev->reset_domain->sem);
-	if (ret)
-		goto error_free;
-
-	swap(adev->reset_dump_reg_list, tmp);
-	swap(adev->reset_dump_reg_value, new);
-	adev->num_regs = i;
-	up_write(&adev->reset_domain->sem);
-	ret = size;
-
-error_free:
-	if (tmp != new)
-		kfree(tmp);
-	kfree(new);
-	return ret;
-}
-
-static const struct file_operations amdgpu_reset_dump_register_list = {
-	.owner = THIS_MODULE,
-	.read = amdgpu_reset_dump_register_list_read,
-	.write = amdgpu_reset_dump_register_list_write,
-	.llseek = default_llseek
-};
-
 int amdgpu_debugfs_init(struct amdgpu_device *adev)
 {
 	struct dentry *root = adev_to_drm(adev)->primary->debugfs_root;
@@ -1964,6 +2041,8 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev)
 	amdgpu_debugfs_firmware_init(adev);
 	amdgpu_ta_if_debugfs_init(adev);
 
+	amdgpu_debugfs_mes_event_log_init(adev);
+
 #if defined(CONFIG_DRM_AMD_DC)
 	if (adev->dc_enabled)
 		dtn_debugfs_init(adev);
@@ -1978,7 +2057,7 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev)
 		amdgpu_debugfs_ring_init(adev, ring);
 	}
 
-	for ( i = 0; i < adev->vcn.num_vcn_inst; i++) {
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
 		if (!amdgpu_vcnfw_log)
 			break;
 
@@ -1988,37 +2067,99 @@ int amdgpu_debugfs_init(struct amdgpu_device *adev)
 		amdgpu_debugfs_vcn_fwlog_init(adev, i, &adev->vcn.inst[i]);
 	}
 
+	if (amdgpu_umsch_mm & amdgpu_umsch_mm_fwlog)
+		amdgpu_debugfs_umsch_fwlog_init(adev, &adev->umsch_mm);
+
+	amdgpu_debugfs_vcn_sched_mask_init(adev);
+	amdgpu_debugfs_jpeg_sched_mask_init(adev);
+	amdgpu_debugfs_gfx_sched_mask_init(adev);
+	amdgpu_debugfs_compute_sched_mask_init(adev);
+	amdgpu_debugfs_sdma_sched_mask_init(adev);
+
 	amdgpu_ras_debugfs_create_all(adev);
 	amdgpu_rap_debugfs_init(adev);
 	amdgpu_securedisplay_debugfs_init(adev);
 	amdgpu_fw_attestation_debugfs_init(adev);
+	amdgpu_psp_debugfs_init(adev);
 
-	debugfs_create_file("amdgpu_evict_vram", 0444, root, adev,
+	debugfs_create_file("amdgpu_evict_vram", 0400, root, adev,
 			    &amdgpu_evict_vram_fops);
-	debugfs_create_file("amdgpu_evict_gtt", 0444, root, adev,
+	debugfs_create_file("amdgpu_evict_gtt", 0400, root, adev,
 			    &amdgpu_evict_gtt_fops);
-	debugfs_create_file("amdgpu_test_ib", 0444, root, adev,
+	debugfs_create_file("amdgpu_test_ib", 0400, root, adev,
 			    &amdgpu_debugfs_test_ib_fops);
 	debugfs_create_file("amdgpu_vm_info", 0444, root, adev,
 			    &amdgpu_debugfs_vm_info_fops);
 	debugfs_create_file("amdgpu_benchmark", 0200, root, adev,
 			    &amdgpu_benchmark_fops);
-	debugfs_create_file("amdgpu_reset_dump_register_list", 0644, root, adev,
-			    &amdgpu_reset_dump_register_list);
 
 	adev->debugfs_vbios_blob.data = adev->bios;
 	adev->debugfs_vbios_blob.size = adev->bios_size;
 	debugfs_create_blob("amdgpu_vbios", 0444, root,
 			    &adev->debugfs_vbios_blob);
 
-	adev->debugfs_discovery_blob.data = adev->mman.discovery_bin;
-	adev->debugfs_discovery_blob.size = adev->mman.discovery_tmr_size;
-	debugfs_create_blob("amdgpu_discovery", 0444, root,
-			    &adev->debugfs_discovery_blob);
+	if (adev->discovery.debugfs_blob.size)
+		debugfs_create_blob("amdgpu_discovery", 0444, root,
+				    &adev->discovery.debugfs_blob);
+
+	return 0;
+}
+
+static int amdgpu_pt_info_read(struct seq_file *m, void *unused)
+{
+	struct drm_file *file;
+	struct amdgpu_fpriv *fpriv;
+	struct amdgpu_bo *root_bo;
+	struct amdgpu_device *adev;
+	int r;
+
+	file = m->private;
+	if (!file)
+		return -EINVAL;
+
+	adev = drm_to_adev(file->minor->dev);
+	fpriv = file->driver_priv;
+	if (!fpriv || !fpriv->vm.root.bo)
+		return -ENODEV;
+
+	root_bo = amdgpu_bo_ref(fpriv->vm.root.bo);
+	r = amdgpu_bo_reserve(root_bo, true);
+	if (r) {
+		amdgpu_bo_unref(&root_bo);
+		return -EINVAL;
+	}
+
+	seq_printf(m, "pd_address: 0x%llx\n", amdgpu_gmc_pd_addr(fpriv->vm.root.bo));
+	seq_printf(m, "max_pfn: 0x%llx\n", adev->vm_manager.max_pfn);
+	seq_printf(m, "num_level: 0x%x\n", adev->vm_manager.num_level);
+	seq_printf(m, "block_size: 0x%x\n", adev->vm_manager.block_size);
+	seq_printf(m, "fragment_size: 0x%x\n", adev->vm_manager.fragment_size);
+
+	amdgpu_bo_unreserve(root_bo);
+	amdgpu_bo_unref(&root_bo);
 
 	return 0;
 }
 
+static int amdgpu_pt_info_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, amdgpu_pt_info_read, inode->i_private);
+}
+
+static const struct file_operations amdgpu_pt_info_fops = {
+	.owner = THIS_MODULE,
+	.open = amdgpu_pt_info_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = single_release,
+};
+
+void amdgpu_debugfs_vm_init(struct drm_file *file)
+{
+	debugfs_create_file("vm_pagetable_info", 0444, file->debugfs_client, file,
+			    &amdgpu_pt_info_fops);
+}
+
 #else
 int amdgpu_debugfs_init(struct amdgpu_device *adev)
 {
@@ -2028,4 +2169,7 @@ int amdgpu_debugfs_regs_init(struct amdgpu_device *adev)
 {
 	return 0;
 }
+void amdgpu_debugfs_vm_init(struct drm_file *file)
+{
+}
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h
index 371a6f0deb29..e7b3c38e5186 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.h
@@ -32,3 +32,6 @@ void amdgpu_debugfs_fini(struct amdgpu_device *adev);
 void amdgpu_debugfs_fence_init(struct amdgpu_device *adev);
 void amdgpu_debugfs_firmware_init(struct amdgpu_device *adev);
 void amdgpu_debugfs_gem_init(struct amdgpu_device *adev);
+void amdgpu_debugfs_mes_event_log_init(struct amdgpu_device *adev);
+void amdgpu_debugfs_vm_init(struct drm_file *file);
+
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c
new file mode 100644
index 000000000000..4e2fe6674db8
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.c
@@ -0,0 +1,371 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <generated/utsrelease.h>
+#include <linux/devcoredump.h>
+#include "amdgpu_dev_coredump.h"
+#include "atom.h"
+
+#ifndef CONFIG_DEV_COREDUMP
+void amdgpu_coredump(struct amdgpu_device *adev, bool skip_vram_check,
+		     bool vram_lost, struct amdgpu_job *job)
+{
+}
+#else
+
+const char *hw_ip_names[MAX_HWIP] = {
+	[GC_HWIP]		= "GC",
+	[HDP_HWIP]		= "HDP",
+	[SDMA0_HWIP]		= "SDMA0",
+	[SDMA1_HWIP]		= "SDMA1",
+	[SDMA2_HWIP]		= "SDMA2",
+	[SDMA3_HWIP]		= "SDMA3",
+	[SDMA4_HWIP]		= "SDMA4",
+	[SDMA5_HWIP]		= "SDMA5",
+	[SDMA6_HWIP]		= "SDMA6",
+	[SDMA7_HWIP]		= "SDMA7",
+	[LSDMA_HWIP]		= "LSDMA",
+	[MMHUB_HWIP]		= "MMHUB",
+	[ATHUB_HWIP]		= "ATHUB",
+	[NBIO_HWIP]		= "NBIO",
+	[MP0_HWIP]		= "MP0",
+	[MP1_HWIP]		= "MP1",
+	[UVD_HWIP]		= "UVD/JPEG/VCN",
+	[VCN1_HWIP]		= "VCN1",
+	[VCE_HWIP]		= "VCE",
+	[VPE_HWIP]		= "VPE",
+	[DF_HWIP]		= "DF",
+	[DCE_HWIP]		= "DCE",
+	[OSSSYS_HWIP]		= "OSSSYS",
+	[SMUIO_HWIP]		= "SMUIO",
+	[PWR_HWIP]		= "PWR",
+	[NBIF_HWIP]		= "NBIF",
+	[THM_HWIP]		= "THM",
+	[CLK_HWIP]		= "CLK",
+	[UMC_HWIP]		= "UMC",
+	[RSMU_HWIP]		= "RSMU",
+	[XGMI_HWIP]		= "XGMI",
+	[DCI_HWIP]		= "DCI",
+	[PCIE_HWIP]		= "PCIE",
+};
+
+static void amdgpu_devcoredump_fw_info(struct amdgpu_device *adev,
+				       struct drm_printer *p)
+{
+	uint32_t version;
+	uint32_t feature;
+	uint8_t smu_program, smu_major, smu_minor, smu_debug;
+	struct atom_context *ctx = adev->mode_info.atom_context;
+
+	drm_printf(p, "VCE feature version: %u, fw version: 0x%08x\n",
+		   adev->vce.fb_version, adev->vce.fw_version);
+	drm_printf(p, "UVD feature version: %u, fw version: 0x%08x\n", 0,
+		   adev->uvd.fw_version);
+	drm_printf(p, "GMC feature version: %u, fw version: 0x%08x\n", 0,
+		   adev->gmc.fw_version);
+	drm_printf(p, "ME feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.me_feature_version, adev->gfx.me_fw_version);
+	drm_printf(p, "PFP feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.pfp_feature_version, adev->gfx.pfp_fw_version);
+	drm_printf(p, "CE feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.ce_feature_version, adev->gfx.ce_fw_version);
+	drm_printf(p, "RLC feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.rlc_feature_version, adev->gfx.rlc_fw_version);
+
+	drm_printf(p, "RLC SRLC feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.rlc_srlc_feature_version,
+		   adev->gfx.rlc_srlc_fw_version);
+	drm_printf(p, "RLC SRLG feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.rlc_srlg_feature_version,
+		   adev->gfx.rlc_srlg_fw_version);
+	drm_printf(p, "RLC SRLS feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.rlc_srls_feature_version,
+		   adev->gfx.rlc_srls_fw_version);
+	drm_printf(p, "RLCP feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.rlcp_ucode_feature_version,
+		   adev->gfx.rlcp_ucode_version);
+	drm_printf(p, "RLCV feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.rlcv_ucode_feature_version,
+		   adev->gfx.rlcv_ucode_version);
+	drm_printf(p, "MEC feature version: %u, fw version: 0x%08x\n",
+		   adev->gfx.mec_feature_version, adev->gfx.mec_fw_version);
+
+	if (adev->gfx.mec2_fw)
+		drm_printf(p, "MEC2 feature version: %u, fw version: 0x%08x\n",
+			   adev->gfx.mec2_feature_version,
+			   adev->gfx.mec2_fw_version);
+
+	drm_printf(p, "IMU feature version: %u, fw version: 0x%08x\n", 0,
+		   adev->gfx.imu_fw_version);
+	drm_printf(p, "PSP SOS feature version: %u, fw version: 0x%08x\n",
+		   adev->psp.sos.feature_version, adev->psp.sos.fw_version);
+	drm_printf(p, "PSP ASD feature version: %u, fw version: 0x%08x\n",
+		   adev->psp.asd_context.bin_desc.feature_version,
+		   adev->psp.asd_context.bin_desc.fw_version);
+
+	drm_printf(p, "TA XGMI feature version: 0x%08x, fw version: 0x%08x\n",
+		   adev->psp.xgmi_context.context.bin_desc.feature_version,
+		   adev->psp.xgmi_context.context.bin_desc.fw_version);
+	drm_printf(p, "TA RAS feature version: 0x%08x, fw version: 0x%08x\n",
+		   adev->psp.ras_context.context.bin_desc.feature_version,
+		   adev->psp.ras_context.context.bin_desc.fw_version);
+	drm_printf(p, "TA HDCP feature version: 0x%08x, fw version: 0x%08x\n",
+		   adev->psp.hdcp_context.context.bin_desc.feature_version,
+		   adev->psp.hdcp_context.context.bin_desc.fw_version);
+	drm_printf(p, "TA DTM feature version: 0x%08x, fw version: 0x%08x\n",
+		   adev->psp.dtm_context.context.bin_desc.feature_version,
+		   adev->psp.dtm_context.context.bin_desc.fw_version);
+	drm_printf(p, "TA RAP feature version: 0x%08x, fw version: 0x%08x\n",
+		   adev->psp.rap_context.context.bin_desc.feature_version,
+		   adev->psp.rap_context.context.bin_desc.fw_version);
+	drm_printf(p,
+		   "TA SECURE DISPLAY feature version: 0x%08x, fw version: 0x%08x\n",
+		   adev->psp.securedisplay_context.context.bin_desc.feature_version,
+		   adev->psp.securedisplay_context.context.bin_desc.fw_version);
+
+	/* SMC firmware */
+	version = adev->pm.fw_version;
+
+	smu_program = (version >> 24) & 0xff;
+	smu_major = (version >> 16) & 0xff;
+	smu_minor = (version >> 8) & 0xff;
+	smu_debug = (version >> 0) & 0xff;
+	drm_printf(p,
+		   "SMC feature version: %u, program: %d, fw version: 0x%08x (%d.%d.%d)\n",
+		   0, smu_program, version, smu_major, smu_minor, smu_debug);
+
+	/* SDMA firmware */
+	for (int i = 0; i < adev->sdma.num_instances; i++) {
+		drm_printf(p,
+			   "SDMA%d feature version: %u, firmware version: 0x%08x\n",
+			   i, adev->sdma.instance[i].feature_version,
+			   adev->sdma.instance[i].fw_version);
+	}
+
+	drm_printf(p, "VCN feature version: %u, fw version: 0x%08x\n", 0,
+		   adev->vcn.fw_version);
+	drm_printf(p, "DMCU feature version: %u, fw version: 0x%08x\n", 0,
+		   adev->dm.dmcu_fw_version);
+	drm_printf(p, "DMCUB feature version: %u, fw version: 0x%08x\n", 0,
+		   adev->dm.dmcub_fw_version);
+	drm_printf(p, "PSP TOC feature version: %u, fw version: 0x%08x\n",
+		   adev->psp.toc.feature_version, adev->psp.toc.fw_version);
+
+	version = adev->mes.kiq_version & AMDGPU_MES_VERSION_MASK;
+	feature = (adev->mes.kiq_version & AMDGPU_MES_FEAT_VERSION_MASK) >>
+		  AMDGPU_MES_FEAT_VERSION_SHIFT;
+	drm_printf(p, "MES_KIQ feature version: %u, fw version: 0x%08x\n",
+		   feature, version);
+
+	version = adev->mes.sched_version & AMDGPU_MES_VERSION_MASK;
+	feature = (adev->mes.sched_version & AMDGPU_MES_FEAT_VERSION_MASK) >>
+		  AMDGPU_MES_FEAT_VERSION_SHIFT;
+	drm_printf(p, "MES feature version: %u, fw version: 0x%08x\n", feature,
+		   version);
+
+	drm_printf(p, "VPE feature version: %u, fw version: 0x%08x\n",
+		   adev->vpe.feature_version, adev->vpe.fw_version);
+
+	drm_printf(p, "\nVBIOS Information\n");
+	drm_printf(p, "vbios name       : %s\n", ctx->name);
+	drm_printf(p, "vbios pn         : %s\n", ctx->vbios_pn);
+	drm_printf(p, "vbios version    : %d\n", ctx->version);
+	drm_printf(p, "vbios ver_str    : %s\n", ctx->vbios_ver_str);
+	drm_printf(p, "vbios date       : %s\n", ctx->date);
+}
+
+static ssize_t
+amdgpu_devcoredump_read(char *buffer, loff_t offset, size_t count,
+			void *data, size_t datalen)
+{
+	struct drm_printer p;
+	struct amdgpu_coredump_info *coredump = data;
+	struct drm_print_iterator iter;
+	struct amdgpu_vm_fault_info *fault_info;
+	struct amdgpu_ip_block *ip_block;
+	int ver;
+
+	iter.data = buffer;
+	iter.offset = 0;
+	iter.start = offset;
+	iter.remain = count;
+
+	p = drm_coredump_printer(&iter);
+
+	drm_printf(&p, "**** AMDGPU Device Coredump ****\n");
+	drm_printf(&p, "version: " AMDGPU_COREDUMP_VERSION "\n");
+	drm_printf(&p, "kernel: " UTS_RELEASE "\n");
+	drm_printf(&p, "module: " KBUILD_MODNAME "\n");
+	drm_printf(&p, "time: %ptSp\n", &coredump->reset_time);
+
+	if (coredump->reset_task_info.task.pid)
+		drm_printf(&p, "process_name: %s PID: %d\n",
+			   coredump->reset_task_info.process_name,
+			   coredump->reset_task_info.task.pid);
+
+	/* SOC Information */
+	drm_printf(&p, "\nSOC Information\n");
+	drm_printf(&p, "SOC Device id: %d\n", coredump->adev->pdev->device);
+	drm_printf(&p, "SOC PCI Revision id: %d\n", coredump->adev->pdev->revision);
+	drm_printf(&p, "SOC Family: %d\n", coredump->adev->family);
+	drm_printf(&p, "SOC Revision id: %d\n", coredump->adev->rev_id);
+	drm_printf(&p, "SOC External Revision id: %d\n", coredump->adev->external_rev_id);
+
+	/* Memory Information */
+	drm_printf(&p, "\nSOC Memory Information\n");
+	drm_printf(&p, "real vram size: %llu\n", coredump->adev->gmc.real_vram_size);
+	drm_printf(&p, "visible vram size: %llu\n", coredump->adev->gmc.visible_vram_size);
+	drm_printf(&p, "gtt size: %llu\n", coredump->adev->mman.gtt_mgr.manager.size);
+
+	/* GDS Config */
+	drm_printf(&p, "\nGDS Config\n");
+	drm_printf(&p, "gds: total size: %d\n", coredump->adev->gds.gds_size);
+	drm_printf(&p, "gds: compute partition size: %d\n", coredump->adev->gds.gds_size);
+	drm_printf(&p, "gds: gws per compute partition: %d\n", coredump->adev->gds.gws_size);
+	drm_printf(&p, "gds: os per compute partition: %d\n", coredump->adev->gds.oa_size);
+
+	/* HWIP Version Information */
+	drm_printf(&p, "\nHW IP Version Information\n");
+	for (int i = 1; i < MAX_HWIP; i++) {
+		for (int j = 0; j < HWIP_MAX_INSTANCE; j++) {
+			ver = coredump->adev->ip_versions[i][j];
+			if (ver)
+				drm_printf(&p, "HWIP: %s[%d][%d]: v%d.%d.%d.%d.%d\n",
+					   hw_ip_names[i], i, j,
+					   IP_VERSION_MAJ(ver),
+					   IP_VERSION_MIN(ver),
+					   IP_VERSION_REV(ver),
+					   IP_VERSION_VARIANT(ver),
+					   IP_VERSION_SUBREV(ver));
+		}
+	}
+
+	/* IP firmware information */
+	drm_printf(&p, "\nIP Firmwares\n");
+	amdgpu_devcoredump_fw_info(coredump->adev, &p);
+
+	if (coredump->ring) {
+		drm_printf(&p, "\nRing timed out details\n");
+		drm_printf(&p, "IP Type: %d Ring Name: %s\n",
+			   coredump->ring->funcs->type,
+			   coredump->ring->name);
+	}
+
+	/* Add page fault information */
+	fault_info = &coredump->adev->vm_manager.fault_info;
+	drm_printf(&p, "\n[%s] Page fault observed\n",
+		   fault_info->vmhub ? "mmhub" : "gfxhub");
+	drm_printf(&p, "Faulty page starting at address: 0x%016llx\n", fault_info->addr);
+	drm_printf(&p, "Protection fault status register: 0x%x\n\n", fault_info->status);
+
+	/* dump the ip state for each ip */
+	drm_printf(&p, "IP Dump\n");
+	for (int i = 0; i < coredump->adev->num_ip_blocks; i++) {
+		ip_block = &coredump->adev->ip_blocks[i];
+		if (ip_block->version->funcs->print_ip_state) {
+			drm_printf(&p, "IP: %s\n", ip_block->version->funcs->name);
+			ip_block->version->funcs->print_ip_state(ip_block, &p);
+			drm_printf(&p, "\n");
+		}
+	}
+
+	/* Add ring buffer information */
+	drm_printf(&p, "Ring buffer information\n");
+	for (int i = 0; i < coredump->adev->num_rings; i++) {
+		int j = 0;
+		struct amdgpu_ring *ring = coredump->adev->rings[i];
+
+		drm_printf(&p, "ring name: %s\n", ring->name);
+		drm_printf(&p, "Rptr: 0x%llx Wptr: 0x%llx RB mask: %x\n",
+			   amdgpu_ring_get_rptr(ring),
+			   amdgpu_ring_get_wptr(ring),
+			   ring->buf_mask);
+		drm_printf(&p, "Ring size in dwords: %d\n",
+			   ring->ring_size / 4);
+		drm_printf(&p, "Ring contents\n");
+		drm_printf(&p, "Offset \t Value\n");
+
+		while (j < ring->ring_size) {
+			drm_printf(&p, "0x%x \t 0x%x\n", j, ring->ring[j / 4]);
+			j += 4;
+		}
+	}
+
+	if (coredump->skip_vram_check)
+		drm_printf(&p, "VRAM lost check is skipped!\n");
+	else if (coredump->reset_vram_lost)
+		drm_printf(&p, "VRAM is lost due to GPU reset!\n");
+
+	return count - iter.remain;
+}
+
+static void amdgpu_devcoredump_free(void *data)
+{
+	kfree(data);
+}
+
+void amdgpu_coredump(struct amdgpu_device *adev, bool skip_vram_check,
+		     bool vram_lost, struct amdgpu_job *job)
+{
+	struct drm_device *dev = adev_to_drm(adev);
+	struct amdgpu_coredump_info *coredump;
+	struct drm_sched_job *s_job;
+
+	coredump = kzalloc(sizeof(*coredump), GFP_NOWAIT);
+
+	if (!coredump) {
+		DRM_ERROR("%s: failed to allocate memory for coredump\n", __func__);
+		return;
+	}
+
+	coredump->skip_vram_check = skip_vram_check;
+	coredump->reset_vram_lost = vram_lost;
+
+	if (job && job->pasid) {
+		struct amdgpu_task_info *ti;
+
+		ti = amdgpu_vm_get_task_info_pasid(adev, job->pasid);
+		if (ti) {
+			coredump->reset_task_info = *ti;
+			amdgpu_vm_put_task_info(ti);
+		}
+	}
+
+	if (job) {
+		s_job = &job->base;
+		coredump->ring = to_amdgpu_ring(s_job->sched);
+	}
+
+	coredump->adev = adev;
+
+	ktime_get_ts64(&coredump->reset_time);
+
+	dev_coredumpm(dev->dev, THIS_MODULE, coredump, 0, GFP_NOWAIT,
+		      amdgpu_devcoredump_read, amdgpu_devcoredump_free);
+
+	drm_info(dev, "AMDGPU device coredump file has been created\n");
+	drm_info(dev, "Check your /sys/class/drm/card%d/device/devcoredump/data\n",
+		 dev->primary->index);
+}
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.h
new file mode 100644
index 000000000000..ef9772c6bcc9
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dev_coredump.h
@@ -0,0 +1,46 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __AMDGPU_DEV_COREDUMP_H__
+#define __AMDGPU_DEV_COREDUMP_H__
+
+#include "amdgpu.h"
+
+#ifdef CONFIG_DEV_COREDUMP
+
+#define AMDGPU_COREDUMP_VERSION "1"
+
+struct amdgpu_coredump_info {
+	struct amdgpu_device            *adev;
+	struct amdgpu_task_info         reset_task_info;
+	struct timespec64               reset_time;
+	bool                            skip_vram_check;
+	bool                            reset_vram_lost;
+	struct amdgpu_ring              *ring;
+};
+#endif
+
+void amdgpu_coredump(struct amdgpu_device *adev, bool skip_vram_check,
+		     bool vram_lost, struct amdgpu_job *job);
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index fac9312b1695..58c3ffe707d1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -25,6 +25,8 @@
  *          Alex Deucher
  *          Jerome Glisse
  */
+
+#include <linux/aperture.h>
 #include <linux/power_supply.h>
 #include <linux/kthread.h>
 #include <linux/module.h>
@@ -32,17 +34,15 @@
 #include <linux/slab.h>
 #include <linux/iommu.h>
 #include <linux/pci.h>
-#include <linux/devcoredump.h>
-#include <generated/utsrelease.h>
 #include <linux/pci-p2pdma.h>
 #include <linux/apple-gmux.h>
 
-#include <drm/drm_aperture.h>
 #include <drm/drm_atomic_helper.h>
+#include <drm/drm_client_event.h>
 #include <drm/drm_crtc_helper.h>
-#include <drm/drm_fb_helper.h>
 #include <drm/drm_probe_helper.h>
 #include <drm/amdgpu_drm.h>
+#include <linux/device.h>
 #include <linux/vgaarb.h>
 #include <linux/vga_switcheroo.h>
 #include <linux/efi.h>
@@ -71,9 +71,12 @@
 
 #include "amdgpu_xgmi.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_ras_mgr.h"
 #include "amdgpu_pmu.h"
 #include "amdgpu_fru_eeprom.h"
 #include "amdgpu_reset.h"
+#include "amdgpu_virt.h"
+#include "amdgpu_dev_coredump.h"
 
 #include <linux/suspend.h>
 #include <drm/task_barrier.h>
@@ -83,6 +86,7 @@
 
 #if IS_ENABLED(CONFIG_X86)
 #include <asm/intel-family.h>
+#include <asm/cpu_device_id.h>
 #endif
 
 MODULE_FIRMWARE("amdgpu/vega10_gpu_info.bin");
@@ -92,10 +96,17 @@ MODULE_FIRMWARE("amdgpu/picasso_gpu_info.bin");
 MODULE_FIRMWARE("amdgpu/raven2_gpu_info.bin");
 MODULE_FIRMWARE("amdgpu/arcturus_gpu_info.bin");
 MODULE_FIRMWARE("amdgpu/navi12_gpu_info.bin");
+MODULE_FIRMWARE("amdgpu/cyan_skillfish_gpu_info.bin");
 
 #define AMDGPU_RESUME_MS		2000
 #define AMDGPU_MAX_RETRY_LIMIT		2
 #define AMDGPU_RETRY_SRIOV_RESET(r) ((r) == -EBUSY || (r) == -ETIMEDOUT || (r) == -EINVAL)
+#define AMDGPU_PCIE_INDEX_FALLBACK (0x38 >> 2)
+#define AMDGPU_PCIE_INDEX_HI_FALLBACK (0x44 >> 2)
+#define AMDGPU_PCIE_DATA_FALLBACK (0x3C >> 2)
+
+#define AMDGPU_VBIOS_SKIP (1U << 0)
+#define AMDGPU_VBIOS_OPTIONAL (1U << 1)
 
 static const struct drm_driver amdgpu_kms_driver;
 
@@ -140,13 +151,76 @@ const char *amdgpu_asic_name[] = {
 	"LAST",
 };
 
+#define AMDGPU_IP_BLK_MASK_ALL GENMASK(AMD_IP_BLOCK_TYPE_NUM  - 1, 0)
+/*
+ * Default init level where all blocks are expected to be initialized. This is
+ * the level of initialization expected by default and also after a full reset
+ * of the device.
+ */
+struct amdgpu_init_level amdgpu_init_default = {
+	.level = AMDGPU_INIT_LEVEL_DEFAULT,
+	.hwini_ip_block_mask = AMDGPU_IP_BLK_MASK_ALL,
+};
+
+struct amdgpu_init_level amdgpu_init_recovery = {
+	.level = AMDGPU_INIT_LEVEL_RESET_RECOVERY,
+	.hwini_ip_block_mask = AMDGPU_IP_BLK_MASK_ALL,
+};
+
+/*
+ * Minimal blocks needed to be initialized before a XGMI hive can be reset. This
+ * is used for cases like reset on initialization where the entire hive needs to
+ * be reset before first use.
+ */
+struct amdgpu_init_level amdgpu_init_minimal_xgmi = {
+	.level = AMDGPU_INIT_LEVEL_MINIMAL_XGMI,
+	.hwini_ip_block_mask =
+		BIT(AMD_IP_BLOCK_TYPE_GMC) | BIT(AMD_IP_BLOCK_TYPE_SMC) |
+		BIT(AMD_IP_BLOCK_TYPE_COMMON) | BIT(AMD_IP_BLOCK_TYPE_IH) |
+		BIT(AMD_IP_BLOCK_TYPE_PSP)
+};
+
+static int amdgpu_device_ip_resume_phase1(struct amdgpu_device *adev);
+static int amdgpu_device_ip_resume_phase2(struct amdgpu_device *adev);
+static int amdgpu_device_ip_resume_phase3(struct amdgpu_device *adev);
+
+static void amdgpu_device_load_switch_state(struct amdgpu_device *adev);
+
+static inline bool amdgpu_ip_member_of_hwini(struct amdgpu_device *adev,
+					     enum amd_ip_block_type block)
+{
+	return (adev->init_lvl->hwini_ip_block_mask & (1U << block)) != 0;
+}
+
+void amdgpu_set_init_level(struct amdgpu_device *adev,
+			   enum amdgpu_init_lvl_id lvl)
+{
+	switch (lvl) {
+	case AMDGPU_INIT_LEVEL_MINIMAL_XGMI:
+		adev->init_lvl = &amdgpu_init_minimal_xgmi;
+		break;
+	case AMDGPU_INIT_LEVEL_RESET_RECOVERY:
+		adev->init_lvl = &amdgpu_init_recovery;
+		break;
+	case AMDGPU_INIT_LEVEL_DEFAULT:
+		fallthrough;
+	default:
+		adev->init_lvl = &amdgpu_init_default;
+		break;
+	}
+}
+
+static inline void amdgpu_device_stop_pending_resets(struct amdgpu_device *adev);
+static int amdgpu_device_pm_notifier(struct notifier_block *nb, unsigned long mode,
+				     void *data);
+
 /**
  * DOC: pcie_replay_count
  *
  * The amdgpu driver provides a sysfs API for reporting the total number
- * of PCIe replays (NAKs)
+ * of PCIe replays (NAKs).
  * The file pcie_replay_count is used for this and returns the total
- * number of replays as a sum of the NAKs generated and NAKs received
+ * number of replays as a sum of the NAKs generated and NAKs received.
  */
 
 static ssize_t amdgpu_device_get_pcie_replay_count(struct device *dev,
@@ -159,89 +233,202 @@ static ssize_t amdgpu_device_get_pcie_replay_count(struct device *dev,
 	return sysfs_emit(buf, "%llu\n", cnt);
 }
 
-static DEVICE_ATTR(pcie_replay_count, S_IRUGO,
+static DEVICE_ATTR(pcie_replay_count, 0444,
 		amdgpu_device_get_pcie_replay_count, NULL);
 
-static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev);
+static int amdgpu_device_attr_sysfs_init(struct amdgpu_device *adev)
+{
+	int ret = 0;
 
-/**
- * DOC: product_name
- *
- * The amdgpu driver provides a sysfs API for reporting the product name
- * for the device
- * The file product_name is used for this and returns the product name
- * as returned from the FRU.
- * NOTE: This is only available for certain server cards
- */
+	if (amdgpu_nbio_is_replay_cnt_supported(adev))
+		ret = sysfs_create_file(&adev->dev->kobj,
+					&dev_attr_pcie_replay_count.attr);
 
-static ssize_t amdgpu_device_get_product_name(struct device *dev,
-		struct device_attribute *attr, char *buf)
+	return ret;
+}
+
+static void amdgpu_device_attr_sysfs_fini(struct amdgpu_device *adev)
 {
+	if (amdgpu_nbio_is_replay_cnt_supported(adev))
+		sysfs_remove_file(&adev->dev->kobj,
+				  &dev_attr_pcie_replay_count.attr);
+}
+
+static ssize_t amdgpu_sysfs_reg_state_get(struct file *f, struct kobject *kobj,
+					  const struct bin_attribute *attr, char *buf,
+					  loff_t ppos, size_t count)
+{
+	struct device *dev = kobj_to_dev(kobj);
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(ddev);
+	ssize_t bytes_read;
+
+	switch (ppos) {
+	case AMDGPU_SYS_REG_STATE_XGMI:
+		bytes_read = amdgpu_asic_get_reg_state(
+			adev, AMDGPU_REG_STATE_TYPE_XGMI, buf, count);
+		break;
+	case AMDGPU_SYS_REG_STATE_WAFL:
+		bytes_read = amdgpu_asic_get_reg_state(
+			adev, AMDGPU_REG_STATE_TYPE_WAFL, buf, count);
+		break;
+	case AMDGPU_SYS_REG_STATE_PCIE:
+		bytes_read = amdgpu_asic_get_reg_state(
+			adev, AMDGPU_REG_STATE_TYPE_PCIE, buf, count);
+		break;
+	case AMDGPU_SYS_REG_STATE_USR:
+		bytes_read = amdgpu_asic_get_reg_state(
+			adev, AMDGPU_REG_STATE_TYPE_USR, buf, count);
+		break;
+	case AMDGPU_SYS_REG_STATE_USR_1:
+		bytes_read = amdgpu_asic_get_reg_state(
+			adev, AMDGPU_REG_STATE_TYPE_USR_1, buf, count);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return bytes_read;
+}
+
+static const BIN_ATTR(reg_state, 0444, amdgpu_sysfs_reg_state_get, NULL,
+		      AMDGPU_SYS_REG_STATE_END);
+
+int amdgpu_reg_state_sysfs_init(struct amdgpu_device *adev)
+{
+	int ret;
+
+	if (!amdgpu_asic_get_reg_state_supported(adev))
+		return 0;
+
+	ret = sysfs_create_bin_file(&adev->dev->kobj, &bin_attr_reg_state);
 
-	return sysfs_emit(buf, "%s\n", adev->product_name);
+	return ret;
+}
+
+void amdgpu_reg_state_sysfs_fini(struct amdgpu_device *adev)
+{
+	if (!amdgpu_asic_get_reg_state_supported(adev))
+		return;
+	sysfs_remove_bin_file(&adev->dev->kobj, &bin_attr_reg_state);
 }
 
-static DEVICE_ATTR(product_name, S_IRUGO,
-		amdgpu_device_get_product_name, NULL);
+int amdgpu_ip_block_suspend(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+
+	if (ip_block->version->funcs->suspend) {
+		r = ip_block->version->funcs->suspend(ip_block);
+		if (r) {
+			dev_err(ip_block->adev->dev,
+				"suspend of IP block <%s> failed %d\n",
+				ip_block->version->funcs->name, r);
+			return r;
+		}
+	}
+
+	ip_block->status.hw = false;
+	return 0;
+}
+
+int amdgpu_ip_block_resume(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+
+	if (ip_block->version->funcs->resume) {
+		r = ip_block->version->funcs->resume(ip_block);
+		if (r) {
+			dev_err(ip_block->adev->dev,
+				"resume of IP block <%s> failed %d\n",
+				ip_block->version->funcs->name, r);
+			return r;
+		}
+	}
+
+	ip_block->status.hw = true;
+	return 0;
+}
 
 /**
- * DOC: product_number
+ * DOC: board_info
+ *
+ * The amdgpu driver provides a sysfs API for giving board related information.
+ * It provides the form factor information in the format
+ *
+ *   type : form factor
+ *
+ * Possible form factor values
+ *
+ * - "cem"		- PCIE CEM card
+ * - "oam"		- Open Compute Accelerator Module
+ * - "unknown"	- Not known
  *
- * The amdgpu driver provides a sysfs API for reporting the part number
- * for the device
- * The file product_number is used for this and returns the part number
- * as returned from the FRU.
- * NOTE: This is only available for certain server cards
  */
 
-static ssize_t amdgpu_device_get_product_number(struct device *dev,
-		struct device_attribute *attr, char *buf)
+static ssize_t amdgpu_device_get_board_info(struct device *dev,
+					    struct device_attribute *attr,
+					    char *buf)
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(ddev);
+	enum amdgpu_pkg_type pkg_type = AMDGPU_PKG_TYPE_CEM;
+	const char *pkg;
+
+	if (adev->smuio.funcs && adev->smuio.funcs->get_pkg_type)
+		pkg_type = adev->smuio.funcs->get_pkg_type(adev);
+
+	switch (pkg_type) {
+	case AMDGPU_PKG_TYPE_CEM:
+		pkg = "cem";
+		break;
+	case AMDGPU_PKG_TYPE_OAM:
+		pkg = "oam";
+		break;
+	default:
+		pkg = "unknown";
+		break;
+	}
 
-	return sysfs_emit(buf, "%s\n", adev->product_number);
+	return sysfs_emit(buf, "%s : %s\n", "type", pkg);
 }
 
-static DEVICE_ATTR(product_number, S_IRUGO,
-		amdgpu_device_get_product_number, NULL);
+static DEVICE_ATTR(board_info, 0444, amdgpu_device_get_board_info, NULL);
 
-/**
- * DOC: serial_number
- *
- * The amdgpu driver provides a sysfs API for reporting the serial number
- * for the device
- * The file serial_number is used for this and returns the serial number
- * as returned from the FRU.
- * NOTE: This is only available for certain server cards
- */
+static struct attribute *amdgpu_board_attrs[] = {
+	&dev_attr_board_info.attr,
+	NULL,
+};
 
-static ssize_t amdgpu_device_get_serial_number(struct device *dev,
-		struct device_attribute *attr, char *buf)
+static umode_t amdgpu_board_attrs_is_visible(struct kobject *kobj,
+					     struct attribute *attr, int n)
 {
+	struct device *dev = kobj_to_dev(kobj);
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(ddev);
 
-	return sysfs_emit(buf, "%s\n", adev->serial);
+	if (adev->flags & AMD_IS_APU)
+		return 0;
+
+	return attr->mode;
 }
 
-static DEVICE_ATTR(serial_number, S_IRUGO,
-		amdgpu_device_get_serial_number, NULL);
+static const struct attribute_group amdgpu_board_attrs_group = {
+	.attrs = amdgpu_board_attrs,
+	.is_visible = amdgpu_board_attrs_is_visible
+};
+
+static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev);
 
 /**
  * amdgpu_device_supports_px - Is the device a dGPU with ATPX power control
  *
- * @dev: drm_device pointer
+ * @adev: amdgpu device pointer
  *
  * Returns true if the device is a dGPU with ATPX power control,
  * otherwise return false.
  */
-bool amdgpu_device_supports_px(struct drm_device *dev)
+bool amdgpu_device_supports_px(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = drm_to_adev(dev);
-
 	if ((adev->flags & AMD_IS_PX) && !amdgpu_is_atpx_hybrid())
 		return true;
 	return false;
@@ -250,14 +437,15 @@ bool amdgpu_device_supports_px(struct drm_device *dev)
 /**
  * amdgpu_device_supports_boco - Is the device a dGPU with ACPI power resources
  *
- * @dev: drm_device pointer
+ * @adev: amdgpu device pointer
  *
  * Returns true if the device is a dGPU with ACPI power control,
  * otherwise return false.
  */
-bool amdgpu_device_supports_boco(struct drm_device *dev)
+bool amdgpu_device_supports_boco(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = drm_to_adev(dev);
+	if (!IS_ENABLED(CONFIG_HOTPLUG_PCI_PCIE))
+		return false;
 
 	if (adev->has_pr3 ||
 	    ((adev->flags & AMD_IS_PX) && amdgpu_is_atpx_hybrid()))
@@ -268,30 +456,105 @@ bool amdgpu_device_supports_boco(struct drm_device *dev)
 /**
  * amdgpu_device_supports_baco - Does the device support BACO
  *
- * @dev: drm_device pointer
+ * @adev: amdgpu device pointer
  *
- * Returns true if the device supporte BACO,
- * otherwise return false.
+ * Return:
+ * 1 if the device supports BACO;
+ * 3 if the device supports MACO (only works if BACO is supported)
+ * otherwise return 0.
  */
-bool amdgpu_device_supports_baco(struct drm_device *dev)
+int amdgpu_device_supports_baco(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = drm_to_adev(dev);
-
 	return amdgpu_asic_supports_baco(adev);
 }
 
+void amdgpu_device_detect_runtime_pm_mode(struct amdgpu_device *adev)
+{
+	int bamaco_support;
+
+	adev->pm.rpm_mode = AMDGPU_RUNPM_NONE;
+	bamaco_support = amdgpu_device_supports_baco(adev);
+
+	switch (amdgpu_runtime_pm) {
+	case 2:
+		if (bamaco_support & MACO_SUPPORT) {
+			adev->pm.rpm_mode = AMDGPU_RUNPM_BAMACO;
+			dev_info(adev->dev, "Forcing BAMACO for runtime pm\n");
+		} else if (bamaco_support == BACO_SUPPORT) {
+			adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
+			dev_info(adev->dev, "Requested mode BAMACO not available,fallback to use BACO\n");
+		}
+		break;
+	case 1:
+		if (bamaco_support & BACO_SUPPORT) {
+			adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
+			dev_info(adev->dev, "Forcing BACO for runtime pm\n");
+		}
+		break;
+	case -1:
+	case -2:
+		if (amdgpu_device_supports_px(adev)) {
+			/* enable PX as runtime mode */
+			adev->pm.rpm_mode = AMDGPU_RUNPM_PX;
+			dev_info(adev->dev, "Using ATPX for runtime pm\n");
+		} else if (amdgpu_device_supports_boco(adev)) {
+			/* enable boco as runtime mode */
+			adev->pm.rpm_mode = AMDGPU_RUNPM_BOCO;
+			dev_info(adev->dev, "Using BOCO for runtime pm\n");
+		} else {
+			if (!bamaco_support)
+				goto no_runtime_pm;
+
+			switch (adev->asic_type) {
+			case CHIP_VEGA20:
+			case CHIP_ARCTURUS:
+				/* BACO are not supported on vega20 and arctrus */
+				break;
+			case CHIP_VEGA10:
+				/* enable BACO as runpm mode if noretry=0 */
+				if (!adev->gmc.noretry && !amdgpu_passthrough(adev))
+					adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
+				break;
+			default:
+				/* enable BACO as runpm mode on CI+ */
+				if (!amdgpu_passthrough(adev))
+					adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
+				break;
+			}
+
+			if (adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) {
+				if (bamaco_support & MACO_SUPPORT) {
+					adev->pm.rpm_mode = AMDGPU_RUNPM_BAMACO;
+					dev_info(adev->dev, "Using BAMACO for runtime pm\n");
+				} else {
+					dev_info(adev->dev, "Using BACO for runtime pm\n");
+				}
+			}
+		}
+		break;
+	case 0:
+		dev_info(adev->dev, "runtime pm is manually disabled\n");
+		break;
+	default:
+		break;
+	}
+
+no_runtime_pm:
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_NONE)
+		dev_info(adev->dev, "Runtime PM not available\n");
+}
 /**
  * amdgpu_device_supports_smart_shift - Is the device dGPU with
  * smart shift support
  *
- * @dev: drm_device pointer
+ * @adev: amdgpu device pointer
  *
  * Returns true if the device is a dGPU with Smart Shift support,
  * otherwise returns false.
  */
-bool amdgpu_device_supports_smart_shift(struct drm_device *dev)
+bool amdgpu_device_supports_smart_shift(struct amdgpu_device *adev)
 {
-	return (amdgpu_device_supports_boco(dev) &&
+	return (amdgpu_device_supports_boco(adev) &&
 		amdgpu_acpi_is_power_shift_control_supported());
 }
 
@@ -342,7 +605,7 @@ void amdgpu_device_mm_access(struct amdgpu_device *adev, loff_t pos,
 }
 
 /**
- * amdgpu_device_aper_access - access vram by vram aperature
+ * amdgpu_device_aper_access - access vram by vram aperture
  *
  * @adev: amdgpu_device pointer
  * @pos: offset of the buffer in vram
@@ -370,10 +633,16 @@ size_t amdgpu_device_aper_access(struct amdgpu_device *adev, loff_t pos,
 
 		if (write) {
 			memcpy_toio(addr, buf, count);
+			/* Make sure HDP write cache flush happens without any reordering
+			 * after the system memory contents are sent over PCIe device
+			 */
 			mb();
 			amdgpu_device_flush_hdp(adev, NULL);
 		} else {
 			amdgpu_device_invalidate_hdp(adev, NULL);
+			/* Make sure HDP read cache is invalidated before issuing a read
+			 * to the PCIe device
+			 */
 			mb();
 			memcpy_fromio(buf, addr, count);
 		}
@@ -427,7 +696,7 @@ bool amdgpu_device_skip_hw_access(struct amdgpu_device *adev)
 	 * here is that the GPU reset is not running on another thread in parallel.
 	 *
 	 * For this we trylock the read side of the reset semaphore, if that succeeds
-	 * we know that the reset is not running in paralell.
+	 * we know that the reset is not running in parallel.
 	 *
 	 * If the trylock fails we assert that we are either already holding the read
 	 * side of the lock or are the reset thread itself and hold the write side of
@@ -464,7 +733,7 @@ uint32_t amdgpu_device_rreg(struct amdgpu_device *adev,
 		if (!(acc_flags & AMDGPU_REGS_NO_KIQ) &&
 		    amdgpu_sriov_runtime(adev) &&
 		    down_read_trylock(&adev->reset_domain->sem)) {
-			ret = amdgpu_kiq_rreg(adev, reg);
+			ret = amdgpu_kiq_rreg(adev, reg, 0);
 			up_read(&adev->reset_domain->sem);
 		} else {
 			ret = readl(((void __iomem *)adev->rmmio) + (reg * 4));
@@ -481,8 +750,7 @@ uint32_t amdgpu_device_rreg(struct amdgpu_device *adev,
 /*
  * MMIO register read with bytes helper functions
  * @offset:bytes offset from MMIO start
- *
-*/
+ */
 
 /**
  * amdgpu_mm_rreg8 - read a memory mapped IO register
@@ -502,12 +770,55 @@ uint8_t amdgpu_mm_rreg8(struct amdgpu_device *adev, uint32_t offset)
 	BUG();
 }
 
+
+/**
+ * amdgpu_device_xcc_rreg - read a memory mapped IO or indirect register with specific XCC
+ *
+ * @adev: amdgpu_device pointer
+ * @reg: dword aligned register offset
+ * @acc_flags: access flags which require special behavior
+ * @xcc_id: xcc accelerated compute core id
+ *
+ * Returns the 32 bit value from the offset specified.
+ */
+uint32_t amdgpu_device_xcc_rreg(struct amdgpu_device *adev,
+				uint32_t reg, uint32_t acc_flags,
+				uint32_t xcc_id)
+{
+	uint32_t ret, rlcg_flag;
+
+	if (amdgpu_device_skip_hw_access(adev))
+		return 0;
+
+	if ((reg * 4) < adev->rmmio_size) {
+		if (amdgpu_sriov_vf(adev) &&
+		    !amdgpu_sriov_runtime(adev) &&
+		    adev->gfx.rlc.rlcg_reg_access_supported &&
+		    amdgpu_virt_get_rlcg_reg_access_flag(adev, acc_flags,
+							 GC_HWIP, false,
+							 &rlcg_flag)) {
+			ret = amdgpu_virt_rlcg_reg_rw(adev, reg, 0, rlcg_flag, GET_INST(GC, xcc_id));
+		} else if (!(acc_flags & AMDGPU_REGS_NO_KIQ) &&
+		    amdgpu_sriov_runtime(adev) &&
+		    down_read_trylock(&adev->reset_domain->sem)) {
+			ret = amdgpu_kiq_rreg(adev, reg, xcc_id);
+			up_read(&adev->reset_domain->sem);
+		} else {
+			ret = readl(((void __iomem *)adev->rmmio) + (reg * 4));
+		}
+	} else {
+		ret = adev->pcie_rreg(adev, reg * 4);
+	}
+
+	return ret;
+}
+
 /*
  * MMIO register write with bytes helper functions
  * @offset:bytes offset from MMIO start
  * @value: the value want to be written to the register
- *
-*/
+ */
+
 /**
  * amdgpu_mm_wreg8 - read a memory mapped IO register
  *
@@ -549,7 +860,7 @@ void amdgpu_device_wreg(struct amdgpu_device *adev,
 		if (!(acc_flags & AMDGPU_REGS_NO_KIQ) &&
 		    amdgpu_sriov_runtime(adev) &&
 		    down_read_trylock(&adev->reset_domain->sem)) {
-			amdgpu_kiq_wreg(adev, reg, v);
+			amdgpu_kiq_wreg(adev, reg, v, 0);
 			up_read(&adev->reset_domain->sem);
 		} else {
 			writel(v, ((void __iomem *)adev->rmmio) + (reg * 4));
@@ -567,11 +878,13 @@ void amdgpu_device_wreg(struct amdgpu_device *adev,
  * @adev: amdgpu_device pointer
  * @reg: mmio/rlc register
  * @v: value to write
+ * @xcc_id: xcc accelerated compute core id
  *
  * this function is invoked only for the debugfs register access
  */
 void amdgpu_mm_wreg_mmio_rlc(struct amdgpu_device *adev,
-			     uint32_t reg, uint32_t v)
+			     uint32_t reg, uint32_t v,
+			     uint32_t xcc_id)
 {
 	if (amdgpu_device_skip_hw_access(adev))
 		return;
@@ -580,7 +893,7 @@ void amdgpu_mm_wreg_mmio_rlc(struct amdgpu_device *adev,
 	    adev->gfx.rlc.funcs &&
 	    adev->gfx.rlc.funcs->is_rlcg_access_range) {
 		if (adev->gfx.rlc.funcs->is_rlcg_access_range(adev, reg))
-			return amdgpu_sriov_wreg(adev, reg, v, 0, 0);
+			return amdgpu_sriov_wreg(adev, reg, v, 0, 0, xcc_id);
 	} else if ((reg * 4) >= adev->rmmio_size) {
 		adev->pcie_wreg(adev, reg * 4, v);
 	} else {
@@ -589,108 +902,144 @@ void amdgpu_mm_wreg_mmio_rlc(struct amdgpu_device *adev,
 }
 
 /**
- * amdgpu_mm_rdoorbell - read a doorbell dword
+ * amdgpu_device_xcc_wreg - write to a memory mapped IO or indirect register with specific XCC
  *
  * @adev: amdgpu_device pointer
- * @index: doorbell index
+ * @reg: dword aligned register offset
+ * @v: 32 bit value to write to the register
+ * @acc_flags: access flags which require special behavior
+ * @xcc_id: xcc accelerated compute core id
  *
- * Returns the value in the doorbell aperture at the
- * requested doorbell index (CIK).
+ * Writes the value specified to the offset specified.
  */
-u32 amdgpu_mm_rdoorbell(struct amdgpu_device *adev, u32 index)
+void amdgpu_device_xcc_wreg(struct amdgpu_device *adev,
+			uint32_t reg, uint32_t v,
+			uint32_t acc_flags, uint32_t xcc_id)
 {
+	uint32_t rlcg_flag;
+
 	if (amdgpu_device_skip_hw_access(adev))
-		return 0;
+		return;
 
-	if (index < adev->doorbell.num_doorbells) {
-		return readl(adev->doorbell.ptr + index);
+	if ((reg * 4) < adev->rmmio_size) {
+		if (amdgpu_sriov_vf(adev) &&
+		    !amdgpu_sriov_runtime(adev) &&
+		    adev->gfx.rlc.rlcg_reg_access_supported &&
+		    amdgpu_virt_get_rlcg_reg_access_flag(adev, acc_flags,
+							 GC_HWIP, true,
+							 &rlcg_flag)) {
+			amdgpu_virt_rlcg_reg_rw(adev, reg, v, rlcg_flag, GET_INST(GC, xcc_id));
+		} else if (!(acc_flags & AMDGPU_REGS_NO_KIQ) &&
+		    amdgpu_sriov_runtime(adev) &&
+		    down_read_trylock(&adev->reset_domain->sem)) {
+			amdgpu_kiq_wreg(adev, reg, v, xcc_id);
+			up_read(&adev->reset_domain->sem);
+		} else {
+			writel(v, ((void __iomem *)adev->rmmio) + (reg * 4));
+		}
 	} else {
-		DRM_ERROR("reading beyond doorbell aperture: 0x%08x!\n", index);
-		return 0;
+		adev->pcie_wreg(adev, reg * 4, v);
 	}
 }
 
 /**
- * amdgpu_mm_wdoorbell - write a doorbell dword
+ * amdgpu_device_indirect_rreg - read an indirect register
  *
  * @adev: amdgpu_device pointer
- * @index: doorbell index
- * @v: value to write
+ * @reg_addr: indirect register address to read from
  *
- * Writes @v to the doorbell aperture at the
- * requested doorbell index (CIK).
+ * Returns the value of indirect register @reg_addr
  */
-void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v)
+u32 amdgpu_device_indirect_rreg(struct amdgpu_device *adev,
+				u32 reg_addr)
 {
-	if (amdgpu_device_skip_hw_access(adev))
-		return;
+	unsigned long flags, pcie_index, pcie_data;
+	void __iomem *pcie_index_offset;
+	void __iomem *pcie_data_offset;
+	u32 r;
 
-	if (index < adev->doorbell.num_doorbells) {
-		writel(v, adev->doorbell.ptr + index);
-	} else {
-		DRM_ERROR("writing beyond doorbell aperture: 0x%08x!\n", index);
-	}
+	pcie_index = adev->nbio.funcs->get_pcie_index_offset(adev);
+	pcie_data = adev->nbio.funcs->get_pcie_data_offset(adev);
+
+	spin_lock_irqsave(&adev->pcie_idx_lock, flags);
+	pcie_index_offset = (void __iomem *)adev->rmmio + pcie_index * 4;
+	pcie_data_offset = (void __iomem *)adev->rmmio + pcie_data * 4;
+
+	writel(reg_addr, pcie_index_offset);
+	readl(pcie_index_offset);
+	r = readl(pcie_data_offset);
+	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
+
+	return r;
 }
 
-/**
- * amdgpu_mm_rdoorbell64 - read a doorbell Qword
- *
- * @adev: amdgpu_device pointer
- * @index: doorbell index
- *
- * Returns the value in the doorbell aperture at the
- * requested doorbell index (VEGA10+).
- */
-u64 amdgpu_mm_rdoorbell64(struct amdgpu_device *adev, u32 index)
+u32 amdgpu_device_indirect_rreg_ext(struct amdgpu_device *adev,
+				    u64 reg_addr)
 {
-	if (amdgpu_device_skip_hw_access(adev))
-		return 0;
+	unsigned long flags, pcie_index, pcie_index_hi, pcie_data;
+	u32 r;
+	void __iomem *pcie_index_offset;
+	void __iomem *pcie_index_hi_offset;
+	void __iomem *pcie_data_offset;
 
-	if (index < adev->doorbell.num_doorbells) {
-		return atomic64_read((atomic64_t *)(adev->doorbell.ptr + index));
+	if (unlikely(!adev->nbio.funcs)) {
+		pcie_index = AMDGPU_PCIE_INDEX_FALLBACK;
+		pcie_data = AMDGPU_PCIE_DATA_FALLBACK;
 	} else {
-		DRM_ERROR("reading beyond doorbell aperture: 0x%08x!\n", index);
-		return 0;
+		pcie_index = adev->nbio.funcs->get_pcie_index_offset(adev);
+		pcie_data = adev->nbio.funcs->get_pcie_data_offset(adev);
 	}
-}
 
-/**
- * amdgpu_mm_wdoorbell64 - write a doorbell Qword
- *
- * @adev: amdgpu_device pointer
- * @index: doorbell index
- * @v: value to write
- *
- * Writes @v to the doorbell aperture at the
- * requested doorbell index (VEGA10+).
- */
-void amdgpu_mm_wdoorbell64(struct amdgpu_device *adev, u32 index, u64 v)
-{
-	if (amdgpu_device_skip_hw_access(adev))
-		return;
-
-	if (index < adev->doorbell.num_doorbells) {
-		atomic64_set((atomic64_t *)(adev->doorbell.ptr + index), v);
+	if (reg_addr >> 32) {
+		if (unlikely(!adev->nbio.funcs))
+			pcie_index_hi = AMDGPU_PCIE_INDEX_HI_FALLBACK;
+		else
+			pcie_index_hi = adev->nbio.funcs->get_pcie_index_hi_offset(adev);
 	} else {
-		DRM_ERROR("writing beyond doorbell aperture: 0x%08x!\n", index);
+		pcie_index_hi = 0;
+	}
+
+	spin_lock_irqsave(&adev->pcie_idx_lock, flags);
+	pcie_index_offset = (void __iomem *)adev->rmmio + pcie_index * 4;
+	pcie_data_offset = (void __iomem *)adev->rmmio + pcie_data * 4;
+	if (pcie_index_hi != 0)
+		pcie_index_hi_offset = (void __iomem *)adev->rmmio +
+				pcie_index_hi * 4;
+
+	writel(reg_addr, pcie_index_offset);
+	readl(pcie_index_offset);
+	if (pcie_index_hi != 0) {
+		writel((reg_addr >> 32) & 0xff, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
 	}
+	r = readl(pcie_data_offset);
+
+	/* clear the high bits */
+	if (pcie_index_hi != 0) {
+		writel(0, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
+
+	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
+
+	return r;
 }
 
 /**
- * amdgpu_device_indirect_rreg - read an indirect register
+ * amdgpu_device_indirect_rreg64 - read a 64bits indirect register
  *
  * @adev: amdgpu_device pointer
  * @reg_addr: indirect register address to read from
  *
  * Returns the value of indirect register @reg_addr
  */
-u32 amdgpu_device_indirect_rreg(struct amdgpu_device *adev,
-				u32 reg_addr)
+u64 amdgpu_device_indirect_rreg64(struct amdgpu_device *adev,
+				  u32 reg_addr)
 {
 	unsigned long flags, pcie_index, pcie_data;
 	void __iomem *pcie_index_offset;
 	void __iomem *pcie_data_offset;
-	u32 r;
+	u64 r;
 
 	pcie_index = adev->nbio.funcs->get_pcie_index_offset(adev);
 	pcie_data = adev->nbio.funcs->get_pcie_data_offset(adev);
@@ -699,45 +1048,64 @@ u32 amdgpu_device_indirect_rreg(struct amdgpu_device *adev,
 	pcie_index_offset = (void __iomem *)adev->rmmio + pcie_index * 4;
 	pcie_data_offset = (void __iomem *)adev->rmmio + pcie_data * 4;
 
+	/* read low 32 bits */
 	writel(reg_addr, pcie_index_offset);
 	readl(pcie_index_offset);
 	r = readl(pcie_data_offset);
+	/* read high 32 bits */
+	writel(reg_addr + 4, pcie_index_offset);
+	readl(pcie_index_offset);
+	r |= ((u64)readl(pcie_data_offset) << 32);
 	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
 
 	return r;
 }
 
-/**
- * amdgpu_device_indirect_rreg64 - read a 64bits indirect register
- *
- * @adev: amdgpu_device pointer
- * @reg_addr: indirect register address to read from
- *
- * Returns the value of indirect register @reg_addr
- */
-u64 amdgpu_device_indirect_rreg64(struct amdgpu_device *adev,
-				  u32 reg_addr)
+u64 amdgpu_device_indirect_rreg64_ext(struct amdgpu_device *adev,
+				  u64 reg_addr)
 {
 	unsigned long flags, pcie_index, pcie_data;
+	unsigned long pcie_index_hi = 0;
 	void __iomem *pcie_index_offset;
+	void __iomem *pcie_index_hi_offset;
 	void __iomem *pcie_data_offset;
 	u64 r;
 
 	pcie_index = adev->nbio.funcs->get_pcie_index_offset(adev);
 	pcie_data = adev->nbio.funcs->get_pcie_data_offset(adev);
+	if ((reg_addr >> 32) && (adev->nbio.funcs->get_pcie_index_hi_offset))
+		pcie_index_hi = adev->nbio.funcs->get_pcie_index_hi_offset(adev);
 
 	spin_lock_irqsave(&adev->pcie_idx_lock, flags);
 	pcie_index_offset = (void __iomem *)adev->rmmio + pcie_index * 4;
 	pcie_data_offset = (void __iomem *)adev->rmmio + pcie_data * 4;
+	if (pcie_index_hi != 0)
+		pcie_index_hi_offset = (void __iomem *)adev->rmmio +
+			pcie_index_hi * 4;
 
 	/* read low 32 bits */
 	writel(reg_addr, pcie_index_offset);
 	readl(pcie_index_offset);
+	if (pcie_index_hi != 0) {
+		writel((reg_addr >> 32) & 0xff, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
 	r = readl(pcie_data_offset);
 	/* read high 32 bits */
 	writel(reg_addr + 4, pcie_index_offset);
 	readl(pcie_index_offset);
+	if (pcie_index_hi != 0) {
+		writel((reg_addr >> 32) & 0xff, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
 	r |= ((u64)readl(pcie_data_offset) << 32);
+
+	/* clear the high bits */
+	if (pcie_index_hi != 0) {
+		writel(0, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
+
 	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
 
 	return r;
@@ -747,8 +1115,6 @@ u64 amdgpu_device_indirect_rreg64(struct amdgpu_device *adev,
  * amdgpu_device_indirect_wreg - write an indirect register address
  *
  * @adev: amdgpu_device pointer
- * @pcie_index: mmio register offset
- * @pcie_data: mmio register offset
  * @reg_addr: indirect register offset
  * @reg_data: indirect register data
  *
@@ -774,12 +1140,50 @@ void amdgpu_device_indirect_wreg(struct amdgpu_device *adev,
 	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
 }
 
+void amdgpu_device_indirect_wreg_ext(struct amdgpu_device *adev,
+				     u64 reg_addr, u32 reg_data)
+{
+	unsigned long flags, pcie_index, pcie_index_hi, pcie_data;
+	void __iomem *pcie_index_offset;
+	void __iomem *pcie_index_hi_offset;
+	void __iomem *pcie_data_offset;
+
+	pcie_index = adev->nbio.funcs->get_pcie_index_offset(adev);
+	pcie_data = adev->nbio.funcs->get_pcie_data_offset(adev);
+	if ((reg_addr >> 32) && (adev->nbio.funcs->get_pcie_index_hi_offset))
+		pcie_index_hi = adev->nbio.funcs->get_pcie_index_hi_offset(adev);
+	else
+		pcie_index_hi = 0;
+
+	spin_lock_irqsave(&adev->pcie_idx_lock, flags);
+	pcie_index_offset = (void __iomem *)adev->rmmio + pcie_index * 4;
+	pcie_data_offset = (void __iomem *)adev->rmmio + pcie_data * 4;
+	if (pcie_index_hi != 0)
+		pcie_index_hi_offset = (void __iomem *)adev->rmmio +
+				pcie_index_hi * 4;
+
+	writel(reg_addr, pcie_index_offset);
+	readl(pcie_index_offset);
+	if (pcie_index_hi != 0) {
+		writel((reg_addr >> 32) & 0xff, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
+	writel(reg_data, pcie_data_offset);
+	readl(pcie_data_offset);
+
+	/* clear the high bits */
+	if (pcie_index_hi != 0) {
+		writel(0, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
+
+	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
+}
+
 /**
  * amdgpu_device_indirect_wreg64 - write a 64bits indirect register address
  *
  * @adev: amdgpu_device pointer
- * @pcie_index: mmio register offset
- * @pcie_data: mmio register offset
  * @reg_addr: indirect register offset
  * @reg_data: indirect register data
  *
@@ -811,6 +1215,55 @@ void amdgpu_device_indirect_wreg64(struct amdgpu_device *adev,
 	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
 }
 
+void amdgpu_device_indirect_wreg64_ext(struct amdgpu_device *adev,
+				   u64 reg_addr, u64 reg_data)
+{
+	unsigned long flags, pcie_index, pcie_data;
+	unsigned long pcie_index_hi = 0;
+	void __iomem *pcie_index_offset;
+	void __iomem *pcie_index_hi_offset;
+	void __iomem *pcie_data_offset;
+
+	pcie_index = adev->nbio.funcs->get_pcie_index_offset(adev);
+	pcie_data = adev->nbio.funcs->get_pcie_data_offset(adev);
+	if ((reg_addr >> 32) && (adev->nbio.funcs->get_pcie_index_hi_offset))
+		pcie_index_hi = adev->nbio.funcs->get_pcie_index_hi_offset(adev);
+
+	spin_lock_irqsave(&adev->pcie_idx_lock, flags);
+	pcie_index_offset = (void __iomem *)adev->rmmio + pcie_index * 4;
+	pcie_data_offset = (void __iomem *)adev->rmmio + pcie_data * 4;
+	if (pcie_index_hi != 0)
+		pcie_index_hi_offset = (void __iomem *)adev->rmmio +
+				pcie_index_hi * 4;
+
+	/* write low 32 bits */
+	writel(reg_addr, pcie_index_offset);
+	readl(pcie_index_offset);
+	if (pcie_index_hi != 0) {
+		writel((reg_addr >> 32) & 0xff, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
+	writel((u32)(reg_data & 0xffffffffULL), pcie_data_offset);
+	readl(pcie_data_offset);
+	/* write high 32 bits */
+	writel(reg_addr + 4, pcie_index_offset);
+	readl(pcie_index_offset);
+	if (pcie_index_hi != 0) {
+		writel((reg_addr >> 32) & 0xff, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
+	writel((u32)(reg_data >> 32), pcie_data_offset);
+	readl(pcie_data_offset);
+
+	/* clear the high bits */
+	if (pcie_index_hi != 0) {
+		writel(0, pcie_index_hi_offset);
+		readl(pcie_index_hi_offset);
+	}
+
+	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
+}
+
 /**
  * amdgpu_device_get_rev_id - query device rev_id
  *
@@ -835,7 +1288,14 @@ u32 amdgpu_device_get_rev_id(struct amdgpu_device *adev)
  */
 static uint32_t amdgpu_invalid_rreg(struct amdgpu_device *adev, uint32_t reg)
 {
-	DRM_ERROR("Invalid callback to read register 0x%04X\n", reg);
+	dev_err(adev->dev, "Invalid callback to read register 0x%04X\n", reg);
+	BUG();
+	return 0;
+}
+
+static uint32_t amdgpu_invalid_rreg_ext(struct amdgpu_device *adev, uint64_t reg)
+{
+	dev_err(adev->dev, "Invalid callback to read register 0x%llX\n", reg);
 	BUG();
 	return 0;
 }
@@ -852,8 +1312,17 @@ static uint32_t amdgpu_invalid_rreg(struct amdgpu_device *adev, uint32_t reg)
  */
 static void amdgpu_invalid_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v)
 {
-	DRM_ERROR("Invalid callback to write register 0x%04X with 0x%08X\n",
-		  reg, v);
+	dev_err(adev->dev,
+		"Invalid callback to write register 0x%04X with 0x%08X\n", reg,
+		v);
+	BUG();
+}
+
+static void amdgpu_invalid_wreg_ext(struct amdgpu_device *adev, uint64_t reg, uint32_t v)
+{
+	dev_err(adev->dev,
+		"Invalid callback to write register 0x%llX with 0x%08X\n", reg,
+		v);
 	BUG();
 }
 
@@ -869,7 +1338,15 @@ static void amdgpu_invalid_wreg(struct amdgpu_device *adev, uint32_t reg, uint32
  */
 static uint64_t amdgpu_invalid_rreg64(struct amdgpu_device *adev, uint32_t reg)
 {
-	DRM_ERROR("Invalid callback to read 64 bit register 0x%04X\n", reg);
+	dev_err(adev->dev, "Invalid callback to read 64 bit register 0x%04X\n",
+		reg);
+	BUG();
+	return 0;
+}
+
+static uint64_t amdgpu_invalid_rreg64_ext(struct amdgpu_device *adev, uint64_t reg)
+{
+	dev_err(adev->dev, "Invalid callback to read register 0x%llX\n", reg);
 	BUG();
 	return 0;
 }
@@ -886,8 +1363,17 @@ static uint64_t amdgpu_invalid_rreg64(struct amdgpu_device *adev, uint32_t reg)
  */
 static void amdgpu_invalid_wreg64(struct amdgpu_device *adev, uint32_t reg, uint64_t v)
 {
-	DRM_ERROR("Invalid callback to write 64 bit register 0x%04X with 0x%08llX\n",
-		  reg, v);
+	dev_err(adev->dev,
+		"Invalid callback to write 64 bit register 0x%04X with 0x%08llX\n",
+		reg, v);
+	BUG();
+}
+
+static void amdgpu_invalid_wreg64_ext(struct amdgpu_device *adev, uint64_t reg, uint64_t v)
+{
+	dev_err(adev->dev,
+		"Invalid callback to write 64 bit register 0x%llX with 0x%08llX\n",
+		reg, v);
 	BUG();
 }
 
@@ -905,8 +1391,9 @@ static void amdgpu_invalid_wreg64(struct amdgpu_device *adev, uint32_t reg, uint
 static uint32_t amdgpu_block_invalid_rreg(struct amdgpu_device *adev,
 					  uint32_t block, uint32_t reg)
 {
-	DRM_ERROR("Invalid callback to read register 0x%04X in block 0x%04X\n",
-		  reg, block);
+	dev_err(adev->dev,
+		"Invalid callback to read register 0x%04X in block 0x%04X\n",
+		reg, block);
 	BUG();
 	return 0;
 }
@@ -926,11 +1413,23 @@ static void amdgpu_block_invalid_wreg(struct amdgpu_device *adev,
 				      uint32_t block,
 				      uint32_t reg, uint32_t v)
 {
-	DRM_ERROR("Invalid block callback to write register 0x%04X in block 0x%04X with 0x%08X\n",
-		  reg, block, v);
+	dev_err(adev->dev,
+		"Invalid block callback to write register 0x%04X in block 0x%04X with 0x%08X\n",
+		reg, block, v);
 	BUG();
 }
 
+static uint32_t amdgpu_device_get_vbios_flags(struct amdgpu_device *adev)
+{
+	if (hweight32(adev->aid_mask) && (adev->flags & AMD_IS_APU))
+		return AMDGPU_VBIOS_SKIP;
+
+	if (hweight32(adev->aid_mask) && amdgpu_passthrough(adev))
+		return AMDGPU_VBIOS_OPTIONAL;
+
+	return 0;
+}
+
 /**
  * amdgpu_device_asic_init - Wrapper for atom asic_init
  *
@@ -940,12 +1439,32 @@ static void amdgpu_block_invalid_wreg(struct amdgpu_device *adev,
  */
 static int amdgpu_device_asic_init(struct amdgpu_device *adev)
 {
+	uint32_t flags;
+	bool optional;
+	int ret;
+
 	amdgpu_asic_pre_asic_init(adev);
+	flags = amdgpu_device_get_vbios_flags(adev);
+	optional = !!(flags & (AMDGPU_VBIOS_OPTIONAL | AMDGPU_VBIOS_SKIP));
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 4) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 5, 0) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(11, 0, 0)) {
+		amdgpu_psp_wait_for_bootloader(adev);
+		if (optional && !adev->bios)
+			return 0;
+
+		ret = amdgpu_atomfirmware_asic_init(adev, true);
+		return ret;
+	} else {
+		if (optional && !adev->bios)
+			return 0;
 
-	if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(11, 0, 0))
-		return amdgpu_atomfirmware_asic_init(adev, true);
-	else
 		return amdgpu_atom_asic_init(adev->mode_info.atom_context);
+	}
+
+	return 0;
 }
 
 /**
@@ -985,7 +1504,7 @@ static void amdgpu_device_mem_scratch_fini(struct amdgpu_device *adev)
  * @registers: pointer to the register array
  * @array_size: size of the register array
  *
- * Programs an array or registers with and and or masks.
+ * Programs an array or registers with and or masks.
  * This is a helper for setting golden registers.
  */
 void amdgpu_device_program_register_sequence(struct amdgpu_device *adev,
@@ -998,7 +1517,7 @@ void amdgpu_device_program_register_sequence(struct amdgpu_device *adev,
 	if (array_size % 3)
 		return;
 
-	for (i = 0; i < array_size; i +=3) {
+	for (i = 0; i < array_size; i += 3) {
 		reg = registers[i + 0];
 		and_mask = registers[i + 1];
 		or_mask = registers[i + 2];
@@ -1043,82 +1562,6 @@ int amdgpu_device_pci_reset(struct amdgpu_device *adev)
 }
 
 /*
- * GPU doorbell aperture helpers function.
- */
-/**
- * amdgpu_device_doorbell_init - Init doorbell driver information.
- *
- * @adev: amdgpu_device pointer
- *
- * Init doorbell driver information (CIK)
- * Returns 0 on success, error on failure.
- */
-static int amdgpu_device_doorbell_init(struct amdgpu_device *adev)
-{
-
-	/* No doorbell on SI hardware generation */
-	if (adev->asic_type < CHIP_BONAIRE) {
-		adev->doorbell.base = 0;
-		adev->doorbell.size = 0;
-		adev->doorbell.num_doorbells = 0;
-		adev->doorbell.ptr = NULL;
-		return 0;
-	}
-
-	if (pci_resource_flags(adev->pdev, 2) & IORESOURCE_UNSET)
-		return -EINVAL;
-
-	amdgpu_asic_init_doorbell_index(adev);
-
-	/* doorbell bar mapping */
-	adev->doorbell.base = pci_resource_start(adev->pdev, 2);
-	adev->doorbell.size = pci_resource_len(adev->pdev, 2);
-
-	if (adev->enable_mes) {
-		adev->doorbell.num_doorbells =
-			adev->doorbell.size / sizeof(u32);
-	} else {
-		adev->doorbell.num_doorbells =
-			min_t(u32, adev->doorbell.size / sizeof(u32),
-			      adev->doorbell_index.max_assignment+1);
-		if (adev->doorbell.num_doorbells == 0)
-			return -EINVAL;
-
-		/* For Vega, reserve and map two pages on doorbell BAR since SDMA
-		 * paging queue doorbell use the second page. The
-		 * AMDGPU_DOORBELL64_MAX_ASSIGNMENT definition assumes all the
-		 * doorbells are in the first page. So with paging queue enabled,
-		 * the max num_doorbells should + 1 page (0x400 in dword)
-		 */
-		if (adev->asic_type >= CHIP_VEGA10)
-			adev->doorbell.num_doorbells += 0x400;
-	}
-
-	adev->doorbell.ptr = ioremap(adev->doorbell.base,
-				     adev->doorbell.num_doorbells *
-				     sizeof(u32));
-	if (adev->doorbell.ptr == NULL)
-		return -ENOMEM;
-
-	return 0;
-}
-
-/**
- * amdgpu_device_doorbell_fini - Tear down doorbell driver information.
- *
- * @adev: amdgpu_device pointer
- *
- * Tear down doorbell driver information (CIK)
- */
-static void amdgpu_device_doorbell_fini(struct amdgpu_device *adev)
-{
-	iounmap(adev->doorbell.ptr);
-	adev->doorbell.ptr = NULL;
-}
-
-
-
-/*
  * amdgpu_device_wb_*()
  * Writeback is the method by which the GPU updates special pages in memory
  * with the status of certain GPU events (fences, ring pointers,etc.).
@@ -1187,13 +1630,17 @@ static int amdgpu_device_wb_init(struct amdgpu_device *adev)
  */
 int amdgpu_device_wb_get(struct amdgpu_device *adev, u32 *wb)
 {
-	unsigned long offset = find_first_zero_bit(adev->wb.used, adev->wb.num_wb);
+	unsigned long flags, offset;
 
+	spin_lock_irqsave(&adev->wb.lock, flags);
+	offset = find_first_zero_bit(adev->wb.used, adev->wb.num_wb);
 	if (offset < adev->wb.num_wb) {
 		__set_bit(offset, adev->wb.used);
+		spin_unlock_irqrestore(&adev->wb.lock, flags);
 		*wb = offset << 3; /* convert to dw offset */
 		return 0;
 	} else {
+		spin_unlock_irqrestore(&adev->wb.lock, flags);
 		return -EINVAL;
 	}
 }
@@ -1208,9 +1655,13 @@ int amdgpu_device_wb_get(struct amdgpu_device *adev, u32 *wb)
  */
 void amdgpu_device_wb_free(struct amdgpu_device *adev, u32 wb)
 {
+	unsigned long flags;
+
 	wb >>= 3;
+	spin_lock_irqsave(&adev->wb.lock, flags);
 	if (wb < adev->wb.num_wb)
 		__clear_bit(wb, adev->wb.used);
+	spin_unlock_irqrestore(&adev->wb.lock, flags);
 }
 
 /**
@@ -1227,14 +1678,33 @@ int amdgpu_device_resize_fb_bar(struct amdgpu_device *adev)
 	int rbar_size = pci_rebar_bytes_to_size(adev->gmc.real_vram_size);
 	struct pci_bus *root;
 	struct resource *res;
-	unsigned i;
+	int max_size, r;
+	unsigned int i;
 	u16 cmd;
-	int r;
+
+	if (!IS_ENABLED(CONFIG_PHYS_ADDR_T_64BIT))
+		return 0;
 
 	/* Bypass for VF */
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
+	if (!amdgpu_rebar)
+		return 0;
+
+	/* resizing on Dell G5 SE platforms causes problems with runtime pm */
+	if ((amdgpu_runtime_pm != 0) &&
+	    adev->pdev->vendor == PCI_VENDOR_ID_ATI &&
+	    adev->pdev->device == 0x731f &&
+	    adev->pdev->subsystem_vendor == PCI_VENDOR_ID_DELL)
+		return 0;
+
+	/* PCI_EXT_CAP_ID_VNDR extended capability is located at 0x100 */
+	if (!pci_find_ext_capability(adev->pdev, PCI_EXT_CAP_ID_VNDR))
+		dev_warn(
+			adev->dev,
+			"System can't access extended configuration space, please check!!\n");
+
 	/* skip if the bios has already enabled large BAR */
 	if (adev->gmc.real_vram_size &&
 	    (pci_resource_len(adev->pdev, 0) >= adev->gmc.real_vram_size))
@@ -1256,33 +1726,32 @@ int amdgpu_device_resize_fb_bar(struct amdgpu_device *adev)
 		return 0;
 
 	/* Limit the BAR size to what is available */
-	rbar_size = min(fls(pci_rebar_get_possible_sizes(adev->pdev, 0)) - 1,
-			rbar_size);
+	max_size = pci_rebar_get_max_size(adev->pdev, 0);
+	if (max_size < 0)
+		return 0;
+	rbar_size = min(max_size, rbar_size);
 
 	/* Disable memory decoding while we change the BAR addresses and size */
 	pci_read_config_word(adev->pdev, PCI_COMMAND, &cmd);
 	pci_write_config_word(adev->pdev, PCI_COMMAND,
 			      cmd & ~PCI_COMMAND_MEMORY);
 
-	/* Free the VRAM and doorbell BAR, we most likely need to move both. */
-	amdgpu_device_doorbell_fini(adev);
-	if (adev->asic_type >= CHIP_BONAIRE)
-		pci_release_resource(adev->pdev, 2);
-
-	pci_release_resource(adev->pdev, 0);
+	/* Tear down doorbell as resizing will release BARs */
+	amdgpu_doorbell_fini(adev);
 
-	r = pci_resize_resource(adev->pdev, 0, rbar_size);
+	r = pci_resize_resource(adev->pdev, 0, rbar_size,
+				(adev->asic_type >= CHIP_BONAIRE) ? 1 << 5
+								  : 1 << 2);
 	if (r == -ENOSPC)
-		DRM_INFO("Not enough PCI address space for a large BAR.");
+		dev_info(adev->dev,
+			 "Not enough PCI address space for a large BAR.");
 	else if (r && r != -ENOTSUPP)
-		DRM_ERROR("Problem resizing BAR0 (%d).", r);
-
-	pci_assign_unassigned_bus_resources(adev->pdev->bus);
+		dev_err(adev->dev, "Problem resizing BAR0 (%d).", r);
 
 	/* When the doorbell or fb BAR isn't available we have no chance of
 	 * using the device.
 	 */
-	r = amdgpu_device_doorbell_init(adev);
+	r = amdgpu_doorbell_init(adev);
 	if (r || (pci_resource_flags(adev->pdev, 0) & IORESOURCE_UNSET))
 		return -ENODEV;
 
@@ -1305,11 +1774,17 @@ int amdgpu_device_resize_fb_bar(struct amdgpu_device *adev)
  */
 bool amdgpu_device_need_post(struct amdgpu_device *adev)
 {
-	uint32_t reg;
+	uint32_t reg, flags;
 
 	if (amdgpu_sriov_vf(adev))
 		return false;
 
+	flags = amdgpu_device_get_vbios_flags(adev);
+	if (flags & AMDGPU_VBIOS_SKIP)
+		return false;
+	if ((flags & AMDGPU_VBIOS_OPTIONAL) && !adev->bios)
+		return false;
+
 	if (amdgpu_passthrough(adev)) {
 		/* for FIJI: In whole GPU pass-through virtualization case, after VM reboot
 		 * some old smc fw still need driver do vPost otherwise gpu hang, while
@@ -1319,19 +1794,21 @@ bool amdgpu_device_need_post(struct amdgpu_device *adev)
 		if (adev->asic_type == CHIP_FIJI) {
 			int err;
 			uint32_t fw_ver;
+
 			err = request_firmware(&adev->pm.fw, "amdgpu/fiji_smc.bin", adev->dev);
-			/* force vPost if error occured */
+			/* force vPost if error occurred */
 			if (err)
 				return true;
 
 			fw_ver = *((uint32_t *)adev->pm.fw->data + 69);
+			release_firmware(adev->pm.fw);
 			if (fw_ver < 0x00160e00)
 				return true;
 		}
 	}
 
 	/* Don't post if we need to reset whole hive on init */
-	if (adev->gmc.xgmi.pending_reset)
+	if (adev->init_lvl->level == AMDGPU_INIT_LEVEL_MINIMAL_XGMI)
 		return false;
 
 	if (adev->has_hw_reset) {
@@ -1352,6 +1829,96 @@ bool amdgpu_device_need_post(struct amdgpu_device *adev)
 	return true;
 }
 
+/*
+ * Check whether seamless boot is supported.
+ *
+ * So far we only support seamless boot on DCE 3.0 or later.
+ * If users report that it works on older ASICS as well, we may
+ * loosen this.
+ */
+bool amdgpu_device_seamless_boot_supported(struct amdgpu_device *adev)
+{
+	switch (amdgpu_seamless) {
+	case -1:
+		break;
+	case 1:
+		return true;
+	case 0:
+		return false;
+	default:
+		dev_err(adev->dev, "Invalid value for amdgpu.seamless: %d\n",
+			amdgpu_seamless);
+		return false;
+	}
+
+	if (!(adev->flags & AMD_IS_APU))
+		return false;
+
+	if (adev->mman.keep_stolen_vga_memory)
+		return false;
+
+	return amdgpu_ip_version(adev, DCE_HWIP, 0) >= IP_VERSION(3, 0, 0);
+}
+
+/*
+ * Intel hosts such as Rocket Lake, Alder Lake, Raptor Lake and Sapphire Rapids
+ * don't support dynamic speed switching. Until we have confirmation from Intel
+ * that a specific host supports it, it's safer that we keep it disabled for all.
+ *
+ * https://edc.intel.com/content/www/us/en/design/products/platforms/details/raptor-lake-s/13th-generation-core-processors-datasheet-volume-1-of-2/005/pci-express-support/
+ * https://gitlab.freedesktop.org/drm/amd/-/issues/2663
+ */
+static bool amdgpu_device_pcie_dynamic_switching_supported(struct amdgpu_device *adev)
+{
+#if IS_ENABLED(CONFIG_X86)
+	struct cpuinfo_x86 *c = &cpu_data(0);
+
+	/* eGPU change speeds based on USB4 fabric conditions */
+	if (dev_is_removable(adev->dev))
+		return true;
+
+	if (c->x86_vendor == X86_VENDOR_INTEL)
+		return false;
+#endif
+	return true;
+}
+
+static bool amdgpu_device_aspm_support_quirk(struct amdgpu_device *adev)
+{
+	/* Enabling ASPM causes randoms hangs on Tahiti and Oland on Zen4.
+	 * It's unclear if this is a platform-specific or GPU-specific issue.
+	 * Disable ASPM on SI for the time being.
+	 */
+	if (adev->family == AMDGPU_FAMILY_SI)
+		return true;
+
+#if IS_ENABLED(CONFIG_X86)
+	struct cpuinfo_x86 *c = &cpu_data(0);
+
+	if (!(amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(12, 0, 0) ||
+		  amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(12, 0, 1)))
+		return false;
+
+	if (c->x86 == 6 &&
+		adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN5) {
+		switch (c->x86_model) {
+		case VFM_MODEL(INTEL_ALDERLAKE):
+		case VFM_MODEL(INTEL_ALDERLAKE_L):
+		case VFM_MODEL(INTEL_RAPTORLAKE):
+		case VFM_MODEL(INTEL_RAPTORLAKE_P):
+		case VFM_MODEL(INTEL_RAPTORLAKE_S):
+			return true;
+		default:
+			return false;
+		}
+	} else {
+		return false;
+	}
+#else
+	return false;
+#endif
+}
+
 /**
  * amdgpu_device_should_use_aspm - check if the device should program ASPM
  *
@@ -1374,20 +1941,13 @@ bool amdgpu_device_should_use_aspm(struct amdgpu_device *adev)
 	default:
 		return false;
 	}
+	if (adev->flags & AMD_IS_APU)
+		return false;
+	if (amdgpu_device_aspm_support_quirk(adev))
+		return false;
 	return pcie_aspm_enabled(adev->pdev);
 }
 
-bool amdgpu_device_aspm_support_quirk(void)
-{
-#if IS_ENABLED(CONFIG_X86)
-	struct cpuinfo_x86 *c = &cpu_data(0);
-
-	return !(c->x86 == 6 && c->x86_model == INTEL_FAM6_ALDERLAKE);
-#else
-	return true;
-#endif
-}
-
 /* if we get transitioned to only one device, take VGA back */
 /**
  * amdgpu_device_vga_set_decode - enable/disable vga decode
@@ -1402,6 +1962,7 @@ static unsigned int amdgpu_device_vga_set_decode(struct pci_dev *pdev,
 		bool state)
 {
 	struct amdgpu_device *adev = drm_to_adev(pci_get_drvdata(pdev));
+
 	amdgpu_asic_set_vga_state(adev, state);
 	if (state)
 		return VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM |
@@ -1424,7 +1985,8 @@ static void amdgpu_device_check_block_size(struct amdgpu_device *adev)
 {
 	/* defines number of bits in page table versus page directory,
 	 * a page is 4KB so we have 12 bits offset, minimum 9 bits in the
-	 * page table and the remaining bits are in the page directory */
+	 * page table and the remaining bits are in the page directory
+	 */
 	if (amdgpu_vm_block_size == -1)
 		return;
 
@@ -1468,7 +2030,7 @@ static void amdgpu_device_check_smu_prv_buffer_size(struct amdgpu_device *adev)
 		return;
 
 	if (!is_os_64) {
-		DRM_WARN("Not 64-bit OS, feature not supported\n");
+		dev_warn(adev->dev, "Not 64-bit OS, feature not supported\n");
 		goto def_value;
 	}
 	si_meminfo(&si);
@@ -1483,7 +2045,7 @@ static void amdgpu_device_check_smu_prv_buffer_size(struct amdgpu_device *adev)
 		if (total_memory < dram_size_seven_GB)
 			goto def_value1;
 	} else {
-		DRM_WARN("Smu memory pool size not supported\n");
+		dev_warn(adev->dev, "Smu memory pool size not supported\n");
 		goto def_value;
 	}
 	adev->pm.smu_prv_buffer_size = amdgpu_smu_memory_pool_size << 28;
@@ -1491,7 +2053,7 @@ static void amdgpu_device_check_smu_prv_buffer_size(struct amdgpu_device *adev)
 	return;
 
 def_value1:
-	DRM_WARN("No enough system memory\n");
+	dev_warn(adev->dev, "No enough system memory\n");
 def_value:
 	adev->pm.smu_prv_buffer_size = 0;
 }
@@ -1543,11 +2105,13 @@ static int amdgpu_device_init_apu_flags(struct amdgpu_device *adev)
  */
 static int amdgpu_device_check_arguments(struct amdgpu_device *adev)
 {
+	int i;
+
 	if (amdgpu_sched_jobs < 4) {
 		dev_warn(adev->dev, "sched jobs (%d) must be at least 4\n",
 			 amdgpu_sched_jobs);
 		amdgpu_sched_jobs = 4;
-	} else if (!is_power_of_2(amdgpu_sched_jobs)){
+	} else if (!is_power_of_2(amdgpu_sched_jobs)) {
 		dev_warn(adev->dev, "sched jobs (%d) must be a power of 2\n",
 			 amdgpu_sched_jobs);
 		amdgpu_sched_jobs = roundup_pow_of_two(amdgpu_sched_jobs);
@@ -1597,6 +2161,32 @@ static int amdgpu_device_check_arguments(struct amdgpu_device *adev)
 
 	adev->firmware.load_type = amdgpu_ucode_get_load_type(adev, amdgpu_fw_load_type);
 
+	for (i = 0; i < MAX_XCP; i++) {
+		switch (amdgpu_enforce_isolation) {
+		case -1:
+		case 0:
+		default:
+			/* disable */
+			adev->enforce_isolation[i] = AMDGPU_ENFORCE_ISOLATION_DISABLE;
+			break;
+		case 1:
+			/* enable */
+			adev->enforce_isolation[i] =
+				AMDGPU_ENFORCE_ISOLATION_ENABLE;
+			break;
+		case 2:
+			/* enable legacy mode */
+			adev->enforce_isolation[i] =
+				AMDGPU_ENFORCE_ISOLATION_ENABLE_LEGACY;
+			break;
+		case 3:
+			/* enable only process isolation without submitting cleaner shader */
+			adev->enforce_isolation[i] =
+				AMDGPU_ENFORCE_ISOLATION_NO_CLEANER_SHADER;
+			break;
+		}
+	}
+
 	return 0;
 }
 
@@ -1615,7 +2205,8 @@ static void amdgpu_switcheroo_set_state(struct pci_dev *pdev,
 	struct drm_device *dev = pci_get_drvdata(pdev);
 	int r;
 
-	if (amdgpu_device_supports_px(dev) && state == VGA_SWITCHEROO_OFF)
+	if (amdgpu_device_supports_px(drm_to_adev(dev)) &&
+	    state == VGA_SWITCHEROO_OFF)
 		return;
 
 	if (state == VGA_SWITCHEROO_ON) {
@@ -1627,13 +2218,15 @@ static void amdgpu_switcheroo_set_state(struct pci_dev *pdev,
 		amdgpu_device_load_pci_state(pdev);
 		r = pci_enable_device(pdev);
 		if (r)
-			DRM_WARN("pci_enable_device failed (%d)\n", r);
+			dev_warn(&pdev->dev, "pci_enable_device failed (%d)\n",
+				 r);
 		amdgpu_device_resume(dev, true);
 
 		dev->switch_power_state = DRM_SWITCH_POWER_ON;
 	} else {
-		pr_info("switched off\n");
+		dev_info(&pdev->dev, "switched off\n");
 		dev->switch_power_state = DRM_SWITCH_POWER_CHANGING;
+		amdgpu_device_prepare(dev);
 		amdgpu_device_suspend(dev, true);
 		amdgpu_device_cache_pci_state(pdev);
 		/* Shut down the device */
@@ -1656,7 +2249,7 @@ static bool amdgpu_switcheroo_can_switch(struct pci_dev *pdev)
 {
 	struct drm_device *dev = pci_get_drvdata(pdev);
 
-	/*
+       /*
 	* FIXME: open_count is protected by drm_global_mutex but that would lead to
 	* locking inversion with the driver load path. And the access here is
 	* completely racy anyway. So don't bother with locking for now.
@@ -1696,10 +2289,11 @@ int amdgpu_device_ip_set_clockgating_state(void *dev,
 		if (!adev->ip_blocks[i].version->funcs->set_clockgating_state)
 			continue;
 		r = adev->ip_blocks[i].version->funcs->set_clockgating_state(
-			(void *)adev, state);
+			&adev->ip_blocks[i], state);
 		if (r)
-			DRM_ERROR("set_clockgating_state of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
+			dev_err(adev->dev,
+				"set_clockgating_state of IP block <%s> failed %d\n",
+				adev->ip_blocks[i].version->funcs->name, r);
 	}
 	return r;
 }
@@ -1730,10 +2324,11 @@ int amdgpu_device_ip_set_powergating_state(void *dev,
 		if (!adev->ip_blocks[i].version->funcs->set_powergating_state)
 			continue;
 		r = adev->ip_blocks[i].version->funcs->set_powergating_state(
-			(void *)adev, state);
+			&adev->ip_blocks[i], state);
 		if (r)
-			DRM_ERROR("set_powergating_state of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
+			dev_err(adev->dev,
+				"set_powergating_state of IP block <%s> failed %d\n",
+				adev->ip_blocks[i].version->funcs->name, r);
 	}
 	return r;
 }
@@ -1758,7 +2353,8 @@ void amdgpu_device_ip_get_clockgating_state(struct amdgpu_device *adev,
 		if (!adev->ip_blocks[i].status.valid)
 			continue;
 		if (adev->ip_blocks[i].version->funcs->get_clockgating_state)
-			adev->ip_blocks[i].version->funcs->get_clockgating_state((void *)adev, flags);
+			adev->ip_blocks[i].version->funcs->get_clockgating_state(
+				&adev->ip_blocks[i], flags);
 	}
 }
 
@@ -1780,9 +2376,12 @@ int amdgpu_device_ip_wait_for_idle(struct amdgpu_device *adev,
 		if (!adev->ip_blocks[i].status.valid)
 			continue;
 		if (adev->ip_blocks[i].version->type == block_type) {
-			r = adev->ip_blocks[i].version->funcs->wait_for_idle((void *)adev);
-			if (r)
-				return r;
+			if (adev->ip_blocks[i].version->funcs->wait_for_idle) {
+				r = adev->ip_blocks[i].version->funcs->wait_for_idle(
+								&adev->ip_blocks[i]);
+				if (r)
+					return r;
+			}
 			break;
 		}
 	}
@@ -1791,26 +2390,45 @@ int amdgpu_device_ip_wait_for_idle(struct amdgpu_device *adev,
 }
 
 /**
- * amdgpu_device_ip_is_idle - is the hardware IP idle
+ * amdgpu_device_ip_is_hw - is the hardware IP enabled
  *
  * @adev: amdgpu_device pointer
  * @block_type: Type of hardware IP (SMU, GFX, UVD, etc.)
  *
- * Check if the hardware IP is idle or not.
- * Returns true if it the IP is idle, false if not.
+ * Check if the hardware IP is enable or not.
+ * Returns true if it the IP is enable, false if not.
  */
-bool amdgpu_device_ip_is_idle(struct amdgpu_device *adev,
-			      enum amd_ip_block_type block_type)
+bool amdgpu_device_ip_is_hw(struct amdgpu_device *adev,
+			    enum amd_ip_block_type block_type)
 {
 	int i;
 
 	for (i = 0; i < adev->num_ip_blocks; i++) {
-		if (!adev->ip_blocks[i].status.valid)
-			continue;
 		if (adev->ip_blocks[i].version->type == block_type)
-			return adev->ip_blocks[i].version->funcs->is_idle((void *)adev);
+			return adev->ip_blocks[i].status.hw;
 	}
-	return true;
+	return false;
+}
+
+/**
+ * amdgpu_device_ip_is_valid - is the hardware IP valid
+ *
+ * @adev: amdgpu_device pointer
+ * @block_type: Type of hardware IP (SMU, GFX, UVD, etc.)
+ *
+ * Check if the hardware IP is valid or not.
+ * Returns true if it the IP is valid, false if not.
+ */
+bool amdgpu_device_ip_is_valid(struct amdgpu_device *adev,
+			       enum amd_ip_block_type block_type)
+{
+	int i;
+
+	for (i = 0; i < adev->num_ip_blocks; i++) {
+		if (adev->ip_blocks[i].version->type == block_type)
+			return adev->ip_blocks[i].status.valid;
+	}
+	return false;
 
 }
 
@@ -1861,6 +2479,34 @@ int amdgpu_device_ip_block_version_cmp(struct amdgpu_device *adev,
 	return 1;
 }
 
+static const char *ip_block_names[] = {
+	[AMD_IP_BLOCK_TYPE_COMMON] = "common",
+	[AMD_IP_BLOCK_TYPE_GMC] = "gmc",
+	[AMD_IP_BLOCK_TYPE_IH] = "ih",
+	[AMD_IP_BLOCK_TYPE_SMC] = "smu",
+	[AMD_IP_BLOCK_TYPE_PSP] = "psp",
+	[AMD_IP_BLOCK_TYPE_DCE] = "dce",
+	[AMD_IP_BLOCK_TYPE_GFX] = "gfx",
+	[AMD_IP_BLOCK_TYPE_SDMA] = "sdma",
+	[AMD_IP_BLOCK_TYPE_UVD] = "uvd",
+	[AMD_IP_BLOCK_TYPE_VCE] = "vce",
+	[AMD_IP_BLOCK_TYPE_ACP] = "acp",
+	[AMD_IP_BLOCK_TYPE_VCN] = "vcn",
+	[AMD_IP_BLOCK_TYPE_MES] = "mes",
+	[AMD_IP_BLOCK_TYPE_JPEG] = "jpeg",
+	[AMD_IP_BLOCK_TYPE_VPE] = "vpe",
+	[AMD_IP_BLOCK_TYPE_UMSCH_MM] = "umsch_mm",
+	[AMD_IP_BLOCK_TYPE_ISP] = "isp",
+	[AMD_IP_BLOCK_TYPE_RAS] = "ras",
+};
+
+static const char *ip_block_name(struct amdgpu_device *adev, enum amd_ip_block_type type)
+{
+	int idx = (int)type;
+
+	return idx < ARRAY_SIZE(ip_block_names) ? ip_block_names[idx] : "unknown";
+}
+
 /**
  * amdgpu_device_ip_block_add
  *
@@ -1889,8 +2535,15 @@ int amdgpu_device_ip_block_add(struct amdgpu_device *adev,
 		break;
 	}
 
-	DRM_INFO("add ip block number %d <%s>\n", adev->num_ip_blocks,
-		  ip_block_version->funcs->name);
+	dev_info(adev->dev, "detected ip block number %d <%s_v%d_%d_%d> (%s)\n",
+		 adev->num_ip_blocks,
+		 ip_block_name(adev, ip_block_version->type),
+		 ip_block_version->major,
+		 ip_block_version->minor,
+		 ip_block_version->rev,
+		 ip_block_version->funcs->name);
+
+	adev->ip_blocks[adev->num_ip_blocks].adev = adev;
 
 	adev->ip_blocks[adev->num_ip_blocks++].version = ip_block_version;
 
@@ -1906,7 +2559,7 @@ int amdgpu_device_ip_block_add(struct amdgpu_device *adev,
  * the module parameter virtual_display.  This feature provides a virtual
  * display hardware on headless boards or in virtualized environments.
  * This function parses and validates the configuration string specified by
- * the user and configues the virtual display configuration (number of
+ * the user and configures the virtual display configuration (number of
  * virtual connectors, crtcs, etc.) specified.
  */
 static void amdgpu_device_enable_virtual_display(struct amdgpu_device *adev)
@@ -1945,9 +2598,11 @@ static void amdgpu_device_enable_virtual_display(struct amdgpu_device *adev)
 			}
 		}
 
-		DRM_INFO("virtual display string:%s, %s:virtual_display:%d, num_crtc:%d\n",
-			 amdgpu_virtual_display, pci_address_name,
-			 adev->enable_virtual_display, adev->mode_info.num_crtc);
+		dev_info(
+			adev->dev,
+			"virtual display string:%s, %s:virtual_display:%d, num_crtc:%d\n",
+			amdgpu_virtual_display, pci_address_name,
+			adev->enable_virtual_display, adev->mode_info.num_crtc);
 
 		kfree(pciaddstr);
 	}
@@ -1958,8 +2613,9 @@ void amdgpu_device_set_sriov_virtual_display(struct amdgpu_device *adev)
 	if (amdgpu_sriov_vf(adev) && !adev->enable_virtual_display) {
 		adev->mode_info.num_crtc = 1;
 		adev->enable_virtual_display = true;
-		DRM_INFO("virtual_display:%d, num_crtc:%d\n",
-			 adev->enable_virtual_display, adev->mode_info.num_crtc);
+		dev_info(adev->dev, "virtual_display:%d, num_crtc:%d\n",
+			 adev->enable_virtual_display,
+			 adev->mode_info.num_crtc);
 	}
 }
 
@@ -1969,29 +2625,18 @@ void amdgpu_device_set_sriov_virtual_display(struct amdgpu_device *adev)
  * @adev: amdgpu_device pointer
  *
  * Parses the asic configuration parameters specified in the gpu info
- * firmware and makes them availale to the driver for use in configuring
+ * firmware and makes them available to the driver for use in configuring
  * the asic.
  * Returns 0 on success, -EINVAL on failure.
  */
 static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[40];
 	int err;
 	const struct gpu_info_firmware_header_v1_0 *hdr;
 
 	adev->firmware.gpu_info_fw = NULL;
 
-	if (adev->mman.discovery_bin) {
-		/*
-		 * FIXME: The bounding box is still needed by Navi12, so
-		 * temporarily read it from gpu_info firmware. Should be dropped
-		 * when DAL no longer needs it.
-		 */
-		if (adev->asic_type != CHIP_NAVI12)
-			return 0;
-	}
-
 	switch (adev->asic_type) {
 	default:
 		return 0;
@@ -2013,16 +2658,24 @@ static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev)
 		chip_name = "arcturus";
 		break;
 	case CHIP_NAVI12:
+		if (adev->discovery.bin)
+			return 0;
 		chip_name = "navi12";
 		break;
+	case CHIP_CYAN_SKILLFISH:
+		if (adev->discovery.bin)
+			return 0;
+		chip_name = "cyan_skillfish";
+		break;
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_gpu_info.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->firmware.gpu_info_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->firmware.gpu_info_fw,
+				   AMDGPU_UCODE_OPTIONAL,
+				   "amdgpu/%s_gpu_info.bin", chip_name);
 	if (err) {
 		dev_err(adev->dev,
-			"Failed to get gpu_info firmware \"%s\"\n",
-			fw_name);
+			"Failed to get gpu_info firmware \"%s_gpu_info.bin\"\n",
+			chip_name);
 		goto out;
 	}
 
@@ -2037,7 +2690,7 @@ static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev)
 								le32_to_cpu(hdr->header.ucode_array_offset_bytes));
 
 		/*
-		 * Should be droped when DAL no longer needs it.
+		 * Should be dropped when DAL no longer needs it.
 		 */
 		if (adev->asic_type == CHIP_NAVI12)
 			goto parse_soc_bounding_box;
@@ -2093,6 +2746,24 @@ out:
 	return err;
 }
 
+static void amdgpu_uid_init(struct amdgpu_device *adev)
+{
+	/* Initialize the UID for the device */
+	adev->uid_info = kzalloc(sizeof(struct amdgpu_uid), GFP_KERNEL);
+	if (!adev->uid_info) {
+		dev_warn(adev->dev, "Failed to allocate memory for UID\n");
+		return;
+	}
+	adev->uid_info->adev = adev;
+}
+
+static void amdgpu_uid_fini(struct amdgpu_device *adev)
+{
+	/* Free the UID memory */
+	kfree(adev->uid_info);
+	adev->uid_info = NULL;
+}
+
 /**
  * amdgpu_device_ip_early_init - run early init for hardware IPs
  *
@@ -2105,10 +2776,11 @@ out:
  */
 static int amdgpu_device_ip_early_init(struct amdgpu_device *adev)
 {
-	struct drm_device *dev = adev_to_drm(adev);
+	struct amdgpu_ip_block *ip_block;
 	struct pci_dev *parent;
+	bool total, skip_bios;
+	uint32_t bios_flags;
 	int i, r;
-	bool total;
 
 	amdgpu_device_enable_virtual_display(adev);
 
@@ -2116,6 +2788,10 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev)
 		r = amdgpu_virt_request_full_gpu(adev, true);
 		if (r)
 			return r;
+
+		r = amdgpu_virt_init_critical_region(adev);
+		if (r)
+			return r;
 	}
 
 	switch (adev->asic_type) {
@@ -2172,47 +2848,62 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev)
 		break;
 	}
 
+	/* Check for IP version 9.4.3 with A0 hardware */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3) &&
+	    !amdgpu_device_get_rev_id(adev)) {
+		dev_err(adev->dev, "Unsupported A0 hardware\n");
+		return -ENODEV;	/* device unsupported - no device error */
+	}
+
 	if (amdgpu_has_atpx() &&
 	    (amdgpu_is_atpx_hybrid() ||
 	     amdgpu_has_atpx_dgpu_power_cntl()) &&
 	    ((adev->flags & AMD_IS_APU) == 0) &&
-	    !pci_is_thunderbolt_attached(to_pci_dev(dev->dev)))
+	    !dev_is_removable(&adev->pdev->dev))
 		adev->flags |= AMD_IS_PX;
 
 	if (!(adev->flags & AMD_IS_APU)) {
-		parent = pci_upstream_bridge(adev->pdev);
+		parent = pcie_find_root_port(adev->pdev);
 		adev->has_pr3 = parent ? pci_pr3_present(parent) : false;
 	}
 
-	amdgpu_amdkfd_device_probe(adev);
-
 	adev->pm.pp_feature = amdgpu_pp_feature_mask;
 	if (amdgpu_sriov_vf(adev) || sched_policy == KFD_SCHED_POLICY_NO_HWS)
 		adev->pm.pp_feature &= ~PP_GFXOFF_MASK;
 	if (amdgpu_sriov_vf(adev) && adev->asic_type == CHIP_SIENNA_CICHLID)
 		adev->pm.pp_feature &= ~PP_OVERDRIVE_MASK;
+	if (!amdgpu_device_pcie_dynamic_switching_supported(adev))
+		adev->pm.pp_feature &= ~PP_PCIE_DPM_MASK;
+
+	adev->virt.is_xgmi_node_migrate_enabled = false;
+	if (amdgpu_sriov_vf(adev)) {
+		adev->virt.is_xgmi_node_migrate_enabled =
+			amdgpu_ip_version((adev), GC_HWIP, 0) == IP_VERSION(9, 4, 4);
+	}
 
 	total = true;
 	for (i = 0; i < adev->num_ip_blocks; i++) {
+		ip_block = &adev->ip_blocks[i];
+
 		if ((amdgpu_ip_block_mask & (1 << i)) == 0) {
-			DRM_ERROR("disabled ip block: %d <%s>\n",
-				  i, adev->ip_blocks[i].version->funcs->name);
+			dev_warn(adev->dev, "disabled ip block: %d <%s>\n", i,
+				 adev->ip_blocks[i].version->funcs->name);
 			adev->ip_blocks[i].status.valid = false;
-		} else {
-			if (adev->ip_blocks[i].version->funcs->early_init) {
-				r = adev->ip_blocks[i].version->funcs->early_init((void *)adev);
-				if (r == -ENOENT) {
-					adev->ip_blocks[i].status.valid = false;
-				} else if (r) {
-					DRM_ERROR("early_init of IP block <%s> failed %d\n",
-						  adev->ip_blocks[i].version->funcs->name, r);
-					total = false;
-				} else {
-					adev->ip_blocks[i].status.valid = true;
-				}
+		} else if (ip_block->version->funcs->early_init) {
+			r = ip_block->version->funcs->early_init(ip_block);
+			if (r == -ENOENT) {
+				adev->ip_blocks[i].status.valid = false;
+			} else if (r) {
+				dev_err(adev->dev,
+					"early_init of IP block <%s> failed %d\n",
+					adev->ip_blocks[i].version->funcs->name,
+					r);
+				total = false;
 			} else {
 				adev->ip_blocks[i].status.valid = true;
 			}
+		} else {
+			adev->ip_blocks[i].status.valid = true;
 		}
 		/* get the vbios after the asic_funcs are set up */
 		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON) {
@@ -2220,15 +2911,32 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev)
 			if (r)
 				return r;
 
+			bios_flags = amdgpu_device_get_vbios_flags(adev);
+			skip_bios = !!(bios_flags & AMDGPU_VBIOS_SKIP);
 			/* Read BIOS */
-			if (!amdgpu_get_bios(adev))
-				return -EINVAL;
-
-			r = amdgpu_atombios_init(adev);
-			if (r) {
-				dev_err(adev->dev, "amdgpu_atombios_init failed\n");
-				amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL, 0, 0);
-				return r;
+			if (!skip_bios) {
+				bool optional =
+					!!(bios_flags & AMDGPU_VBIOS_OPTIONAL);
+				if (!amdgpu_get_bios(adev) && !optional)
+					return -EINVAL;
+
+				if (optional && !adev->bios)
+					dev_info(
+						adev->dev,
+						"VBIOS image optional, proceeding without VBIOS image");
+
+				if (adev->bios) {
+					r = amdgpu_atombios_init(adev);
+					if (r) {
+						dev_err(adev->dev,
+							"amdgpu_atombios_init failed\n");
+						amdgpu_vf_error_put(
+							adev,
+							AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL,
+							0, 0);
+						return r;
+					}
+				}
 			}
 
 			/*get pf2vf msg info at it's earliest time*/
@@ -2240,6 +2948,15 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev)
 	if (!total)
 		return -ENODEV;
 
+	if (adev->gmc.xgmi.supported)
+		amdgpu_xgmi_early_init(adev);
+
+	if (amdgpu_is_multi_aid(adev))
+		amdgpu_uid_init(adev);
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GFX);
+	if (ip_block->status.valid != false)
+		amdgpu_amdkfd_device_probe(adev);
+
 	adev->cg_flags &= amdgpu_cg_mask;
 	adev->pg_flags &= amdgpu_pg_mask;
 
@@ -2255,13 +2972,18 @@ static int amdgpu_device_ip_hw_init_phase1(struct amdgpu_device *adev)
 			continue;
 		if (adev->ip_blocks[i].status.hw)
 			continue;
+		if (!amdgpu_ip_member_of_hwini(
+			    adev, adev->ip_blocks[i].version->type))
+			continue;
 		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON ||
 		    (amdgpu_sriov_vf(adev) && (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP)) ||
 		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH) {
-			r = adev->ip_blocks[i].version->funcs->hw_init(adev);
+			r = adev->ip_blocks[i].version->funcs->hw_init(&adev->ip_blocks[i]);
 			if (r) {
-				DRM_ERROR("hw_init of IP block <%s> failed %d\n",
-					  adev->ip_blocks[i].version->funcs->name, r);
+				dev_err(adev->dev,
+					"hw_init of IP block <%s> failed %d\n",
+					adev->ip_blocks[i].version->funcs->name,
+					r);
 				return r;
 			}
 			adev->ip_blocks[i].status.hw = true;
@@ -2280,10 +3002,14 @@ static int amdgpu_device_ip_hw_init_phase2(struct amdgpu_device *adev)
 			continue;
 		if (adev->ip_blocks[i].status.hw)
 			continue;
-		r = adev->ip_blocks[i].version->funcs->hw_init(adev);
+		if (!amdgpu_ip_member_of_hwini(
+			    adev, adev->ip_blocks[i].version->type))
+			continue;
+		r = adev->ip_blocks[i].version->funcs->hw_init(&adev->ip_blocks[i]);
 		if (r) {
-			DRM_ERROR("hw_init of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
+			dev_err(adev->dev,
+				"hw_init of IP block <%s> failed %d\n",
+				adev->ip_blocks[i].version->funcs->name, r);
 			return r;
 		}
 		adev->ip_blocks[i].status.hw = true;
@@ -2303,6 +3029,10 @@ static int amdgpu_device_fw_loading(struct amdgpu_device *adev)
 			if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_PSP)
 				continue;
 
+			if (!amdgpu_ip_member_of_hwini(adev,
+						       AMD_IP_BLOCK_TYPE_PSP))
+				break;
+
 			if (!adev->ip_blocks[i].status.sw)
 				continue;
 
@@ -2311,22 +3041,21 @@ static int amdgpu_device_fw_loading(struct amdgpu_device *adev)
 				break;
 
 			if (amdgpu_in_reset(adev) || adev->in_suspend) {
-				r = adev->ip_blocks[i].version->funcs->resume(adev);
-				if (r) {
-					DRM_ERROR("resume of IP block <%s> failed %d\n",
-							  adev->ip_blocks[i].version->funcs->name, r);
+				r = amdgpu_ip_block_resume(&adev->ip_blocks[i]);
+				if (r)
 					return r;
-				}
 			} else {
-				r = adev->ip_blocks[i].version->funcs->hw_init(adev);
+				r = adev->ip_blocks[i].version->funcs->hw_init(&adev->ip_blocks[i]);
 				if (r) {
-					DRM_ERROR("hw_init of IP block <%s> failed %d\n",
-							  adev->ip_blocks[i].version->funcs->name, r);
+					dev_err(adev->dev,
+						"hw_init of IP block <%s> failed %d\n",
+						adev->ip_blocks[i]
+							.version->funcs->name,
+						r);
 					return r;
 				}
+				adev->ip_blocks[i].status.hw = true;
 			}
-
-			adev->ip_blocks[i].status.hw = true;
 			break;
 		}
 	}
@@ -2339,6 +3068,12 @@ static int amdgpu_device_fw_loading(struct amdgpu_device *adev)
 
 static int amdgpu_device_init_schedulers(struct amdgpu_device *adev)
 {
+	struct drm_sched_init_args args = {
+		.ops = &amdgpu_sched_ops,
+		.num_rqs = DRM_SCHED_PRIORITY_COUNT,
+		.timeout_wq = adev->reset_domain->wq,
+		.dev = adev->dev,
+	};
 	long timeout;
 	int r, i;
 
@@ -2364,18 +3099,37 @@ static int amdgpu_device_init_schedulers(struct amdgpu_device *adev)
 			break;
 		}
 
-		r = drm_sched_init(&ring->sched, &amdgpu_sched_ops,
-				   ring->num_hw_submission, amdgpu_job_hang_limit,
-				   timeout, adev->reset_domain->wq,
-				   ring->sched_score, ring->name,
-				   adev->dev);
+		args.timeout = timeout;
+		args.credit_limit = ring->num_hw_submission;
+		args.score = ring->sched_score;
+		args.name = ring->name;
+
+		r = drm_sched_init(&ring->sched, &args);
+		if (r) {
+			dev_err(adev->dev,
+				"Failed to create scheduler on ring %s.\n",
+				ring->name);
+			return r;
+		}
+		r = amdgpu_uvd_entity_init(adev, ring);
+		if (r) {
+			dev_err(adev->dev,
+				"Failed to create UVD scheduling entity on ring %s.\n",
+				ring->name);
+			return r;
+		}
+		r = amdgpu_vce_entity_init(adev, ring);
 		if (r) {
-			DRM_ERROR("Failed to create scheduler on ring %s.\n",
-				  ring->name);
+			dev_err(adev->dev,
+				"Failed to create VCE scheduling entity on ring %s.\n",
+				ring->name);
 			return r;
 		}
 	}
 
+	if (adev->xcp_mgr)
+		amdgpu_xcp_update_partition_sched_list(adev);
+
 	return 0;
 }
 
@@ -2393,6 +3147,7 @@ static int amdgpu_device_init_schedulers(struct amdgpu_device *adev)
  */
 static int amdgpu_device_ip_init(struct amdgpu_device *adev)
 {
+	bool init_badpage;
 	int i, r;
 
 	r = amdgpu_ras_init(adev);
@@ -2402,19 +3157,28 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
 	for (i = 0; i < adev->num_ip_blocks; i++) {
 		if (!adev->ip_blocks[i].status.valid)
 			continue;
-		r = adev->ip_blocks[i].version->funcs->sw_init((void *)adev);
-		if (r) {
-			DRM_ERROR("sw_init of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
-			goto init_failed;
+		if (adev->ip_blocks[i].version->funcs->sw_init) {
+			r = adev->ip_blocks[i].version->funcs->sw_init(&adev->ip_blocks[i]);
+			if (r) {
+				dev_err(adev->dev,
+					"sw_init of IP block <%s> failed %d\n",
+					adev->ip_blocks[i].version->funcs->name,
+					r);
+				goto init_failed;
+			}
 		}
 		adev->ip_blocks[i].status.sw = true;
 
+		if (!amdgpu_ip_member_of_hwini(
+			    adev, adev->ip_blocks[i].version->type))
+			continue;
+
 		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON) {
 			/* need to do common hw init early so everything is set up for gmc */
-			r = adev->ip_blocks[i].version->funcs->hw_init((void *)adev);
+			r = adev->ip_blocks[i].version->funcs->hw_init(&adev->ip_blocks[i]);
 			if (r) {
-				DRM_ERROR("hw_init %d failed %d\n", i, r);
+				dev_err(adev->dev, "hw_init %d failed %d\n", i,
+					r);
 				goto init_failed;
 			}
 			adev->ip_blocks[i].status.hw = true;
@@ -2426,32 +3190,44 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
 
 			r = amdgpu_device_mem_scratch_init(adev);
 			if (r) {
-				DRM_ERROR("amdgpu_mem_scratch_init failed %d\n", r);
+				dev_err(adev->dev,
+					"amdgpu_mem_scratch_init failed %d\n",
+					r);
 				goto init_failed;
 			}
-			r = adev->ip_blocks[i].version->funcs->hw_init((void *)adev);
+			r = adev->ip_blocks[i].version->funcs->hw_init(&adev->ip_blocks[i]);
 			if (r) {
-				DRM_ERROR("hw_init %d failed %d\n", i, r);
+				dev_err(adev->dev, "hw_init %d failed %d\n", i,
+					r);
 				goto init_failed;
 			}
 			r = amdgpu_device_wb_init(adev);
 			if (r) {
-				DRM_ERROR("amdgpu_device_wb_init failed %d\n", r);
+				dev_err(adev->dev,
+					"amdgpu_device_wb_init failed %d\n", r);
 				goto init_failed;
 			}
 			adev->ip_blocks[i].status.hw = true;
 
 			/* right after GMC hw init, we create CSA */
-			if (amdgpu_mcbp) {
+			if (adev->gfx.mcbp) {
 				r = amdgpu_allocate_static_csa(adev, &adev->virt.csa_obj,
 							       AMDGPU_GEM_DOMAIN_VRAM |
 							       AMDGPU_GEM_DOMAIN_GTT,
 							       AMDGPU_CSA_SIZE);
 				if (r) {
-					DRM_ERROR("allocate CSA failed %d\n", r);
+					dev_err(adev->dev,
+						"allocate CSA failed %d\n", r);
 					goto init_failed;
 				}
 			}
+
+			r = amdgpu_seq64_init(adev);
+			if (r) {
+				dev_err(adev->dev, "allocate seq64 failed %d\n",
+					r);
+				goto init_failed;
+			}
 		}
 	}
 
@@ -2496,7 +3272,8 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
 	 * Note: theoretically, this should be called before all vram allocations
 	 * to protect retired page from abusing
 	 */
-	r = amdgpu_ras_recovery_init(adev);
+	init_badpage = (adev->init_lvl->level != AMDGPU_INIT_LEVEL_MINIMAL_XGMI);
+	r = amdgpu_ras_recovery_init(adev, init_badpage);
 	if (r)
 		goto init_failed;
 
@@ -2532,15 +3309,21 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev)
 	if (r)
 		goto init_failed;
 
+	if (adev->mman.buffer_funcs_ring->sched.ready)
+		amdgpu_ttm_set_buffer_funcs_status(adev, true);
+
 	/* Don't init kfd if whole hive need to be reset during init */
-	if (!adev->gmc.xgmi.pending_reset)
+	if (adev->init_lvl->level != AMDGPU_INIT_LEVEL_MINIMAL_XGMI) {
+		kgd2kfd_init_zone_device(adev);
 		amdgpu_amdkfd_device_init(adev);
+	}
 
 	amdgpu_fru_get_product_info(adev);
 
+	if (!amdgpu_sriov_vf(adev) || amdgpu_sriov_ras_cper_en(adev))
+		r = amdgpu_cper_init(adev);
+
 init_failed:
-	if (amdgpu_sriov_vf(adev))
-		amdgpu_virt_release_full_gpu(adev, true);
 
 	return r;
 }
@@ -2552,7 +3335,7 @@ init_failed:
  *
  * Writes a reset magic value to the gart pointer in VRAM.  The driver calls
  * this function before a GPU reset.  If the value is retained after a
- * GPU reset, VRAM has not been lost.  Some GPU resets may destry VRAM contents.
+ * GPU reset, VRAM has not been lost. Some GPU resets may destroy VRAM contents.
  */
 static void amdgpu_device_fill_reset_magic(struct amdgpu_device *adev)
 {
@@ -2583,6 +3366,8 @@ static bool amdgpu_device_check_vram_lost(struct amdgpu_device *adev)
 	 * always assumed to be lost.
 	 */
 	switch (amdgpu_asic_reset_method(adev)) {
+	case AMD_RESET_METHOD_LEGACY:
+	case AMD_RESET_METHOD_LINK:
 	case AMD_RESET_METHOD_BACO:
 	case AMD_RESET_METHOD_MODE1:
 		return true;
@@ -2628,11 +3413,13 @@ int amdgpu_device_set_cg_state(struct amdgpu_device *adev,
 		    adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_JPEG &&
 		    adev->ip_blocks[i].version->funcs->set_clockgating_state) {
 			/* enable clockgating to save power */
-			r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev,
+			r = adev->ip_blocks[i].version->funcs->set_clockgating_state(&adev->ip_blocks[i],
 										     state);
 			if (r) {
-				DRM_ERROR("set_clockgating_state(gate) of IP block <%s> failed %d\n",
-					  adev->ip_blocks[i].version->funcs->name, r);
+				dev_err(adev->dev,
+					"set_clockgating_state(gate) of IP block <%s> failed %d\n",
+					adev->ip_blocks[i].version->funcs->name,
+					r);
 				return r;
 			}
 		}
@@ -2658,18 +3445,21 @@ int amdgpu_device_set_pg_state(struct amdgpu_device *adev,
 		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GFX ||
 		     adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SDMA))
 			continue;
-		/* skip CG for VCE/UVD, it's handled specially */
+		/* skip CG for VCE/UVD/VPE, it's handled specially */
 		if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_UVD &&
 		    adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VCE &&
 		    adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VCN &&
+		    adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VPE &&
 		    adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_JPEG &&
 		    adev->ip_blocks[i].version->funcs->set_powergating_state) {
 			/* enable powergating to save power */
-			r = adev->ip_blocks[i].version->funcs->set_powergating_state((void *)adev,
+			r = adev->ip_blocks[i].version->funcs->set_powergating_state(&adev->ip_blocks[i],
 											state);
 			if (r) {
-				DRM_ERROR("set_powergating_state(gate) of IP block <%s> failed %d\n",
-					  adev->ip_blocks[i].version->funcs->name, r);
+				dev_err(adev->dev,
+					"set_powergating_state(gate) of IP block <%s> failed %d\n",
+					adev->ip_blocks[i].version->funcs->name,
+					r);
 				return r;
 			}
 		}
@@ -2696,7 +3486,7 @@ static int amdgpu_device_enable_mgpu_fan_boost(void)
 	for (i = 0; i < mgpu_info.num_dgpu; i++) {
 		gpu_ins = &(mgpu_info.gpu_ins[i]);
 		adev = gpu_ins->adev;
-		if (!(adev->flags & AMD_IS_APU) &&
+		if (!(adev->flags & AMD_IS_APU || amdgpu_sriov_multi_vf_mode(adev)) &&
 		    !gpu_ins->mgpu_fan_enabled) {
 			ret = amdgpu_dpm_enable_mgpu_fan_boost(adev);
 			if (ret)
@@ -2733,10 +3523,12 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
 		if (!adev->ip_blocks[i].status.hw)
 			continue;
 		if (adev->ip_blocks[i].version->funcs->late_init) {
-			r = adev->ip_blocks[i].version->funcs->late_init((void *)adev);
+			r = adev->ip_blocks[i].version->funcs->late_init(&adev->ip_blocks[i]);
 			if (r) {
-				DRM_ERROR("late_init of IP block <%s> failed %d\n",
-					  adev->ip_blocks[i].version->funcs->name, r);
+				dev_err(adev->dev,
+					"late_init of IP block <%s> failed %d\n",
+					adev->ip_blocks[i].version->funcs->name,
+					r);
 				return r;
 			}
 		}
@@ -2745,11 +3537,12 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
 
 	r = amdgpu_ras_late_init(adev);
 	if (r) {
-		DRM_ERROR("amdgpu_ras_late_init failed %d", r);
+		dev_err(adev->dev, "amdgpu_ras_late_init failed %d", r);
 		return r;
 	}
 
-	amdgpu_ras_set_error_query_ready(adev, true);
+	if (!amdgpu_reset_in_recovery(adev))
+		amdgpu_ras_set_error_query_ready(adev, true);
 
 	amdgpu_device_set_cg_state(adev, AMD_CG_STATE_GATE);
 	amdgpu_device_set_pg_state(adev, AMD_PG_STATE_GATE);
@@ -2758,11 +3551,12 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
 
 	r = amdgpu_device_enable_mgpu_fan_boost();
 	if (r)
-		DRM_ERROR("enable mgpu fan boost failed (%d).\n", r);
+		dev_err(adev->dev, "enable mgpu fan boost failed (%d).\n", r);
 
 	/* For passthrough configuration on arcturus and aldebaran, enable special handling SBR */
-	if (amdgpu_passthrough(adev) && ((adev->asic_type == CHIP_ARCTURUS && adev->gmc.xgmi.num_physical_nodes > 1)||
-			       adev->asic_type == CHIP_ALDEBARAN ))
+	if (amdgpu_passthrough(adev) &&
+	    ((adev->asic_type == CHIP_ARCTURUS && adev->gmc.xgmi.num_physical_nodes > 1) ||
+	     adev->asic_type == CHIP_ALDEBARAN))
 		amdgpu_dpm_handle_passthrough_sbr(adev, true);
 
 	if (adev->gmc.xgmi.num_physical_nodes > 1) {
@@ -2790,7 +3584,9 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
 				r = amdgpu_xgmi_set_pstate(gpu_instance->adev,
 						AMDGPU_XGMI_PSTATE_MIN);
 				if (r) {
-					DRM_ERROR("pstate setting failed (%d).\n", r);
+					dev_err(adev->dev,
+						"pstate setting failed (%d).\n",
+						r);
 					break;
 				}
 			}
@@ -2802,6 +3598,27 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
 	return 0;
 }
 
+static void amdgpu_ip_block_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	if (!ip_block->version->funcs->hw_fini) {
+		dev_err(adev->dev, "hw_fini of IP block <%s> not defined\n",
+			ip_block->version->funcs->name);
+	} else {
+		r = ip_block->version->funcs->hw_fini(ip_block);
+		/* XXX handle errors */
+		if (r) {
+			dev_dbg(adev->dev,
+				"hw_fini of IP block <%s> failed %d\n",
+				ip_block->version->funcs->name, r);
+		}
+	}
+
+	ip_block->status.hw = false;
+}
+
 /**
  * amdgpu_device_smu_fini_early - smu hw_fini wrapper
  *
@@ -2811,22 +3628,16 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
  */
 static void amdgpu_device_smu_fini_early(struct amdgpu_device *adev)
 {
-	int i, r;
+	int i;
 
-	if (adev->ip_versions[GC_HWIP][0] > IP_VERSION(9, 0, 0))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) > IP_VERSION(9, 0, 0))
 		return;
 
 	for (i = 0; i < adev->num_ip_blocks; i++) {
 		if (!adev->ip_blocks[i].status.hw)
 			continue;
 		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC) {
-			r = adev->ip_blocks[i].version->funcs->hw_fini((void *)adev);
-			/* XXX handle errors */
-			if (r) {
-				DRM_DEBUG("hw_fini of IP block <%s> failed %d\n",
-					  adev->ip_blocks[i].version->funcs->name, r);
-			}
-			adev->ip_blocks[i].status.hw = false;
+			amdgpu_ip_block_hw_fini(&adev->ip_blocks[i]);
 			break;
 		}
 	}
@@ -2840,38 +3651,48 @@ static int amdgpu_device_ip_fini_early(struct amdgpu_device *adev)
 		if (!adev->ip_blocks[i].version->funcs->early_fini)
 			continue;
 
-		r = adev->ip_blocks[i].version->funcs->early_fini((void *)adev);
+		r = adev->ip_blocks[i].version->funcs->early_fini(&adev->ip_blocks[i]);
 		if (r) {
-			DRM_DEBUG("early_fini of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
+			dev_dbg(adev->dev,
+				"early_fini of IP block <%s> failed %d\n",
+				adev->ip_blocks[i].version->funcs->name, r);
 		}
 	}
 
 	amdgpu_device_set_pg_state(adev, AMD_PG_STATE_UNGATE);
 	amdgpu_device_set_cg_state(adev, AMD_CG_STATE_UNGATE);
 
-	amdgpu_amdkfd_suspend(adev, false);
+	amdgpu_amdkfd_suspend(adev, true);
+	amdgpu_userq_suspend(adev);
 
-	/* Workaroud for ASICs need to disable SMC first */
+	/* Workaround for ASICs need to disable SMC first */
 	amdgpu_device_smu_fini_early(adev);
 
 	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
 		if (!adev->ip_blocks[i].status.hw)
 			continue;
 
-		r = adev->ip_blocks[i].version->funcs->hw_fini((void *)adev);
-		/* XXX handle errors */
-		if (r) {
-			DRM_DEBUG("hw_fini of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
-		}
-
-		adev->ip_blocks[i].status.hw = false;
+		amdgpu_ip_block_hw_fini(&adev->ip_blocks[i]);
 	}
 
 	if (amdgpu_sriov_vf(adev)) {
 		if (amdgpu_virt_release_full_gpu(adev, false))
-			DRM_ERROR("failed to release exclusive mode on fini\n");
+			dev_err(adev->dev,
+				"failed to release exclusive mode on fini\n");
+	}
+
+	/*
+	 * Driver reload on the APU can fail due to firmware validation because
+	 * the PSP is always running, as it is shared across the whole SoC.
+	 * This same issue does not occur on dGPU because it has a mechanism
+	 * that checks whether the PSP is running. A solution for those issues
+	 * in the APU is to trigger a GPU reset, but this should be done during
+	 * the unload phase to avoid adding boot latency and screen flicker.
+	 */
+	if ((adev->flags & AMD_IS_APU) && !adev->gmc.is_app_apu) {
+		r = amdgpu_asic_reset(adev);
+		if (r)
+			dev_err(adev->dev, "asic reset on %s failed\n", __func__);
 	}
 
 	return 0;
@@ -2892,6 +3713,8 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev)
 {
 	int i, r;
 
+	amdgpu_cper_fini(adev);
+
 	if (amdgpu_sriov_vf(adev) && adev->virt.ras_init_done)
 		amdgpu_virt_release_ras_err_handler_data(adev);
 
@@ -2910,13 +3733,18 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev)
 			amdgpu_device_wb_fini(adev);
 			amdgpu_device_mem_scratch_fini(adev);
 			amdgpu_ib_pool_fini(adev);
+			amdgpu_seq64_fini(adev);
+			amdgpu_doorbell_fini(adev);
 		}
-
-		r = adev->ip_blocks[i].version->funcs->sw_fini((void *)adev);
-		/* XXX handle errors */
-		if (r) {
-			DRM_DEBUG("sw_fini of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
+		if (adev->ip_blocks[i].version->funcs->sw_fini) {
+			r = adev->ip_blocks[i].version->funcs->sw_fini(&adev->ip_blocks[i]);
+			/* XXX handle errors */
+			if (r) {
+				dev_dbg(adev->dev,
+					"sw_fini of IP block <%s> failed %d\n",
+					adev->ip_blocks[i].version->funcs->name,
+					r);
+			}
 		}
 		adev->ip_blocks[i].status.sw = false;
 		adev->ip_blocks[i].status.valid = false;
@@ -2926,11 +3754,12 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev)
 		if (!adev->ip_blocks[i].status.late_initialized)
 			continue;
 		if (adev->ip_blocks[i].version->funcs->late_fini)
-			adev->ip_blocks[i].version->funcs->late_fini((void *)adev);
+			adev->ip_blocks[i].version->funcs->late_fini(&adev->ip_blocks[i]);
 		adev->ip_blocks[i].status.late_initialized = false;
 	}
 
 	amdgpu_ras_fini(adev);
+	amdgpu_uid_fini(adev);
 
 	return 0;
 }
@@ -2948,7 +3777,7 @@ static void amdgpu_device_delayed_init_work_handler(struct work_struct *work)
 
 	r = amdgpu_ib_ring_tests(adev);
 	if (r)
-		DRM_ERROR("ib ring test failed (%d).\n", r);
+		dev_err(adev->dev, "ib ring test failed (%d).\n", r);
 }
 
 static void amdgpu_device_delay_enable_gfx_off(struct work_struct *work)
@@ -2959,7 +3788,7 @@ static void amdgpu_device_delay_enable_gfx_off(struct work_struct *work)
 	WARN_ON_ONCE(adev->gfx.gfx_off_state);
 	WARN_ON_ONCE(adev->gfx.gfx_off_req_count);
 
-	if (!amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, true))
+	if (!amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, true, 0))
 		adev->gfx.gfx_off_state = true;
 }
 
@@ -2976,7 +3805,7 @@ static void amdgpu_device_delay_enable_gfx_off(struct work_struct *work)
  */
 static int amdgpu_device_ip_suspend_phase1(struct amdgpu_device *adev)
 {
-	int i, r;
+	int i, r, rec;
 
 	amdgpu_device_set_pg_state(adev, AMD_PG_STATE_UNGATE);
 	amdgpu_device_set_cg_state(adev, AMD_CG_STATE_UNGATE);
@@ -2997,19 +3826,25 @@ static int amdgpu_device_ip_suspend_phase1(struct amdgpu_device *adev)
 		if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_DCE)
 			continue;
 
-		/* XXX handle errors */
-		r = adev->ip_blocks[i].version->funcs->suspend(adev);
-		/* XXX handle errors */
-		if (r) {
-			DRM_ERROR("suspend of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
-			return r;
-		}
-
-		adev->ip_blocks[i].status.hw = false;
+		r = amdgpu_ip_block_suspend(&adev->ip_blocks[i]);
+		if (r)
+			goto unwind;
 	}
 
 	return 0;
+unwind:
+	rec = amdgpu_device_ip_resume_phase3(adev);
+	if (rec)
+		dev_err(adev->dev,
+			"amdgpu_device_ip_resume_phase3 failed during unwind: %d\n",
+			rec);
+
+	amdgpu_dpm_set_df_cstate(adev, DF_CSTATE_ALLOW);
+
+	amdgpu_device_set_pg_state(adev, AMD_PG_STATE_GATE);
+	amdgpu_device_set_cg_state(adev, AMD_CG_STATE_GATE);
+
+	return r;
 }
 
 /**
@@ -3025,7 +3860,7 @@ static int amdgpu_device_ip_suspend_phase1(struct amdgpu_device *adev)
  */
 static int amdgpu_device_ip_suspend_phase2(struct amdgpu_device *adev)
 {
-	int i, r;
+	int i, r, rec;
 
 	if (adev->in_s0ix)
 		amdgpu_dpm_gfx_state_change(adev, sGpuChangeState_D3Entry);
@@ -3044,15 +3879,17 @@ static int amdgpu_device_ip_suspend_phase2(struct amdgpu_device *adev)
 		}
 
 		/* skip unnecessary suspend if we do not initialize them yet */
-		if (adev->gmc.xgmi.pending_reset &&
-		    !(adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC ||
-		      adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC ||
-		      adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON ||
-		      adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH)) {
-			adev->ip_blocks[i].status.hw = false;
+		if (!amdgpu_ip_member_of_hwini(
+			    adev, adev->ip_blocks[i].version->type))
 			continue;
-		}
 
+		/* Since we skip suspend for S0i3, we need to cancel the delayed
+		 * idle work here as the suspend callback never gets called.
+		 */
+		if (adev->in_s0ix &&
+		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GFX &&
+		    amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 0, 0))
+			cancel_delayed_work_sync(&adev->gfx.idle_work);
 		/* skip suspend of gfx/mes and psp for S0ix
 		 * gfx is in gfxoff state, so on resume it will exit gfxoff just
 		 * like at runtime. PSP is also part of the always on hardware
@@ -3066,8 +3903,10 @@ static int amdgpu_device_ip_suspend_phase2(struct amdgpu_device *adev)
 
 		/* SDMA 5.x+ is part of GFX power domain so it's covered by GFXOFF */
 		if (adev->in_s0ix &&
-		    (adev->ip_versions[SDMA0_HWIP][0] >= IP_VERSION(5, 0, 0)) &&
-		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SDMA))
+		    (amdgpu_ip_version(adev, SDMA0_HWIP, 0) >=
+		     IP_VERSION(5, 0, 0)) &&
+		    (adev->ip_blocks[i].version->type ==
+		     AMD_IP_BLOCK_TYPE_SDMA))
 			continue;
 
 		/* Once swPSP provides the IMU, RLC FW binaries to TOS during cold-boot.
@@ -3082,28 +3921,52 @@ static int amdgpu_device_ip_suspend_phase2(struct amdgpu_device *adev)
 		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP)
 			continue;
 
-		/* XXX handle errors */
-		r = adev->ip_blocks[i].version->funcs->suspend(adev);
-		/* XXX handle errors */
-		if (r) {
-			DRM_ERROR("suspend of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
-		}
-		adev->ip_blocks[i].status.hw = false;
+		r = amdgpu_ip_block_suspend(&adev->ip_blocks[i]);
+		if (r)
+			goto unwind;
+
 		/* handle putting the SMC in the appropriate state */
-		if(!amdgpu_sriov_vf(adev)){
+		if (!amdgpu_sriov_vf(adev)) {
 			if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC) {
 				r = amdgpu_dpm_set_mp1_state(adev, adev->mp1_state);
 				if (r) {
-					DRM_ERROR("SMC failed to set mp1 state %d, %d\n",
-							adev->mp1_state, r);
-					return r;
+					dev_err(adev->dev,
+						"SMC failed to set mp1 state %d, %d\n",
+						adev->mp1_state, r);
+					goto unwind;
 				}
 			}
 		}
 	}
 
 	return 0;
+unwind:
+	/* suspend phase 2 = resume phase 1 + resume phase 2 */
+	rec = amdgpu_device_ip_resume_phase1(adev);
+	if (rec) {
+		dev_err(adev->dev,
+			"amdgpu_device_ip_resume_phase1 failed during unwind: %d\n",
+			rec);
+		return r;
+	}
+
+	rec = amdgpu_device_fw_loading(adev);
+	if (rec) {
+		dev_err(adev->dev,
+			"amdgpu_device_fw_loading failed during unwind: %d\n",
+			rec);
+		return r;
+	}
+
+	rec = amdgpu_device_ip_resume_phase2(adev);
+	if (rec) {
+		dev_err(adev->dev,
+			"amdgpu_device_ip_resume_phase2 failed during unwind: %d\n",
+			rec);
+		return r;
+	}
+
+	return r;
 }
 
 /**
@@ -3117,7 +3980,7 @@ static int amdgpu_device_ip_suspend_phase2(struct amdgpu_device *adev)
  * in each IP into a state suitable for suspend.
  * Returns 0 on success, negative error code on failure.
  */
-int amdgpu_device_ip_suspend(struct amdgpu_device *adev)
+static int amdgpu_device_ip_suspend(struct amdgpu_device *adev)
 {
 	int r;
 
@@ -3126,6 +3989,8 @@ int amdgpu_device_ip_suspend(struct amdgpu_device *adev)
 		amdgpu_virt_request_full_gpu(adev, false);
 	}
 
+	amdgpu_ttm_set_buffer_funcs_status(adev, false);
+
 	r = amdgpu_device_ip_suspend_phase1(adev);
 	if (r)
 		return r;
@@ -3161,10 +4026,12 @@ static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev)
 				!block->status.valid)
 				continue;
 
-			r = block->version->funcs->hw_init(adev);
-			DRM_INFO("RE-INIT-early: %s %s\n", block->version->funcs->name, r?"failed":"succeeded");
-			if (r)
+			r = block->version->funcs->hw_init(&adev->ip_blocks[i]);
+			if (r) {
+				dev_err(adev->dev, "RE-INIT-early: %s failed\n",
+					 block->version->funcs->name);
 				return r;
+			}
 			block->status.hw = true;
 		}
 	}
@@ -3174,7 +4041,8 @@ static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev)
 
 static int amdgpu_device_ip_reinit_late_sriov(struct amdgpu_device *adev)
 {
-	int i, r;
+	struct amdgpu_ip_block *block;
+	int i, r = 0;
 
 	static enum amd_ip_block_type ip_order[] = {
 		AMD_IP_BLOCK_TYPE_SMC,
@@ -3189,30 +4057,28 @@ static int amdgpu_device_ip_reinit_late_sriov(struct amdgpu_device *adev)
 	};
 
 	for (i = 0; i < ARRAY_SIZE(ip_order); i++) {
-		int j;
-		struct amdgpu_ip_block *block;
+		block = amdgpu_device_ip_get_ip_block(adev, ip_order[i]);
 
-		for (j = 0; j < adev->num_ip_blocks; j++) {
-			block = &adev->ip_blocks[j];
-
-			if (block->version->type != ip_order[i] ||
-				!block->status.valid ||
-				block->status.hw)
-				continue;
+		if (!block)
+			continue;
 
-			if (block->version->type == AMD_IP_BLOCK_TYPE_SMC)
-				r = block->version->funcs->resume(adev);
-			else
-				r = block->version->funcs->hw_init(adev);
+		if (block->status.valid && !block->status.hw) {
+			if (block->version->type == AMD_IP_BLOCK_TYPE_SMC) {
+				r = amdgpu_ip_block_resume(block);
+			} else {
+				r = block->version->funcs->hw_init(block);
+			}
 
-			DRM_INFO("RE-INIT-late: %s %s\n", block->version->funcs->name, r?"failed":"succeeded");
-			if (r)
-				return r;
+			if (r) {
+				dev_err(adev->dev, "RE-INIT-late: %s failed\n",
+					 block->version->funcs->name);
+				break;
+			}
 			block->status.hw = true;
 		}
 	}
 
-	return 0;
+	return r;
 }
 
 /**
@@ -3239,13 +4105,9 @@ static int amdgpu_device_ip_resume_phase1(struct amdgpu_device *adev)
 		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH ||
 		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP && amdgpu_sriov_vf(adev))) {
 
-			r = adev->ip_blocks[i].version->funcs->resume(adev);
-			if (r) {
-				DRM_ERROR("resume of IP block <%s> failed %d\n",
-					  adev->ip_blocks[i].version->funcs->name, r);
+			r = amdgpu_ip_block_resume(&adev->ip_blocks[i]);
+			if (r)
 				return r;
-			}
-			adev->ip_blocks[i].status.hw = true;
 		}
 	}
 
@@ -3257,7 +4119,7 @@ static int amdgpu_device_ip_resume_phase1(struct amdgpu_device *adev)
  *
  * @adev: amdgpu_device pointer
  *
- * First resume function for hardware IPs.  The list of all the hardware
+ * Second resume function for hardware IPs.  The list of all the hardware
  * IPs that make up the asic is walked and the resume callbacks are run for
  * all blocks except COMMON, GMC, and IH.  resume puts the hardware into a
  * functional state after a suspend and updates the software state as
@@ -3275,15 +4137,42 @@ static int amdgpu_device_ip_resume_phase2(struct amdgpu_device *adev)
 		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON ||
 		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC ||
 		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH ||
+		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE ||
 		    adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP)
 			continue;
-		r = adev->ip_blocks[i].version->funcs->resume(adev);
-		if (r) {
-			DRM_ERROR("resume of IP block <%s> failed %d\n",
-				  adev->ip_blocks[i].version->funcs->name, r);
+		r = amdgpu_ip_block_resume(&adev->ip_blocks[i]);
+		if (r)
 			return r;
+	}
+
+	return 0;
+}
+
+/**
+ * amdgpu_device_ip_resume_phase3 - run resume for hardware IPs
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Third resume function for hardware IPs.  The list of all the hardware
+ * IPs that make up the asic is walked and the resume callbacks are run for
+ * all DCE.  resume puts the hardware into a functional state after a suspend
+ * and updates the software state as necessary.  This function is also used
+ * for restoring the GPU after a GPU reset.
+ *
+ * Returns 0 on success, negative error code on failure.
+ */
+static int amdgpu_device_ip_resume_phase3(struct amdgpu_device *adev)
+{
+	int i, r;
+
+	for (i = 0; i < adev->num_ip_blocks; i++) {
+		if (!adev->ip_blocks[i].status.valid || adev->ip_blocks[i].status.hw)
+			continue;
+		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE) {
+			r = amdgpu_ip_block_resume(&adev->ip_blocks[i]);
+			if (r)
+				return r;
 		}
-		adev->ip_blocks[i].status.hw = true;
 	}
 
 	return 0;
@@ -3296,7 +4185,7 @@ static int amdgpu_device_ip_resume_phase2(struct amdgpu_device *adev)
  *
  * Main resume function for hardware IPs.  The hardware IPs
  * are split into two resume functions because they are
- * are also used in in recovering from a GPU reset and some additional
+ * also used in recovering from a GPU reset and some additional
  * steps need to be take between them.  In this case (S3/S4) they are
  * run sequentially.
  * Returns 0 on success, negative error code on failure.
@@ -3305,10 +4194,6 @@ static int amdgpu_device_ip_resume(struct amdgpu_device *adev)
 {
 	int r;
 
-	r = amdgpu_amdkfd_resume_iommu(adev);
-	if (r)
-		return r;
-
 	r = amdgpu_device_ip_resume_phase1(adev);
 	if (r)
 		return r;
@@ -3319,6 +4204,16 @@ static int amdgpu_device_ip_resume(struct amdgpu_device *adev)
 
 	r = amdgpu_device_ip_resume_phase2(adev);
 
+	if (adev->mman.buffer_funcs_ring->sched.ready)
+		amdgpu_ttm_set_buffer_funcs_status(adev, true);
+
+	if (r)
+		return r;
+
+	amdgpu_fence_driver_hw_init(adev);
+
+	r = amdgpu_device_ip_resume_phase3(adev);
+
 	return r;
 }
 
@@ -3348,12 +4243,14 @@ static void amdgpu_device_detect_sriov_bios(struct amdgpu_device *adev)
 /**
  * amdgpu_device_asic_has_dc_support - determine if DC supports the asic
  *
+ * @pdev : pci device context
  * @asic_type: AMD asic type
  *
  * Check if there is DC (new modesetting infrastructre) support for an asic.
  * returns true if DC has support, false if not.
  */
-bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type)
+bool amdgpu_device_asic_has_dc_support(struct pci_dev *pdev,
+				       enum amd_asic_type asic_type)
 {
 	switch (asic_type) {
 #ifdef CONFIG_DRM_AMDGPU_SI
@@ -3367,25 +4264,13 @@ bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type)
 	case CHIP_PITCAIRN:
 	case CHIP_VERDE:
 	case CHIP_OLAND:
-		/*
-		 * We have systems in the wild with these ASICs that require
-		 * LVDS and VGA support which is not supported with DC.
-		 *
-		 * Fallback to the non-DC driver here by default so as not to
-		 * cause regressions.
-		 */
-#if defined(CONFIG_DRM_AMD_DC_SI)
-		return amdgpu_dc > 0;
-#else
-		return false;
-#endif
-	case CHIP_BONAIRE:
+		return amdgpu_dc != 0 && IS_ENABLED(CONFIG_DRM_AMD_DC_SI);
 	case CHIP_KAVERI:
 	case CHIP_KABINI:
 	case CHIP_MULLINS:
 		/*
 		 * We have systems in the wild with these ASICs that require
-		 * VGA support which is not supported with DC.
+		 * TRAVIS and NUTMEG support which is not supported with DC.
 		 *
 		 * Fallback to the non-DC driver here by default so as not to
 		 * cause regressions.
@@ -3396,8 +4281,9 @@ bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type)
 #else
 	default:
 		if (amdgpu_dc > 0)
-			DRM_INFO_ONCE("Display Core has been requested via kernel parameter "
-					 "but isn't supported by ASIC, ignoring\n");
+			dev_info_once(
+				&pdev->dev,
+				"Display Core has been requested via kernel parameter but isn't supported by ASIC, ignoring\n");
 		return false;
 #endif
 	}
@@ -3416,7 +4302,7 @@ bool amdgpu_device_has_dc_support(struct amdgpu_device *adev)
 	    (adev->harvest_ip_mask & AMD_HARVEST_IP_DMU_MASK))
 		return false;
 
-	return amdgpu_device_asic_has_dc_support(adev->asic_type);
+	return amdgpu_device_asic_has_dc_support(adev->pdev, adev->asic_type);
 }
 
 static void amdgpu_device_xgmi_reset_func(struct work_struct *__work)
@@ -3438,20 +4324,18 @@ static void amdgpu_device_xgmi_reset_func(struct work_struct *__work)
 	if (amdgpu_asic_reset_method(adev) == AMD_RESET_METHOD_BACO) {
 
 		task_barrier_enter(&hive->tb);
-		adev->asic_reset_res = amdgpu_device_baco_enter(adev_to_drm(adev));
+		adev->asic_reset_res = amdgpu_device_baco_enter(adev);
 
 		if (adev->asic_reset_res)
 			goto fail;
 
 		task_barrier_exit(&hive->tb);
-		adev->asic_reset_res = amdgpu_device_baco_exit(adev_to_drm(adev));
+		adev->asic_reset_res = amdgpu_device_baco_exit(adev);
 
 		if (adev->asic_reset_res)
 			goto fail;
 
-		if (adev->mmhub.ras && adev->mmhub.ras->ras_block.hw_ops &&
-		    adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count)
-			adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count(adev);
+		amdgpu_ras_reset_error_count(adev, AMDGPU_RAS_BLOCK__MMHUB);
 	} else {
 
 		task_barrier_full(&hive->tb);
@@ -3460,7 +4344,8 @@ static void amdgpu_device_xgmi_reset_func(struct work_struct *__work)
 
 fail:
 	if (adev->asic_reset_res)
-		DRM_WARN("ASIC reset failed with error, %d for drm dev, %s",
+		dev_warn(adev->dev,
+			 "ASIC reset failed with error, %d for drm dev, %s",
 			 adev->asic_reset_res, adev_to_drm(adev)->unique);
 	amdgpu_put_xgmi_hive(hive);
 }
@@ -3473,66 +4358,53 @@ static int amdgpu_device_get_job_timeout_settings(struct amdgpu_device *adev)
 	long timeout;
 	int ret = 0;
 
-	/*
-	 * By default timeout for non compute jobs is 10000
-	 * and 60000 for compute jobs.
-	 * In SR-IOV or passthrough mode, timeout for compute
-	 * jobs are 60000 by default.
-	 */
-	adev->gfx_timeout = msecs_to_jiffies(10000);
-	adev->sdma_timeout = adev->video_timeout = adev->gfx_timeout;
-	if (amdgpu_sriov_vf(adev))
-		adev->compute_timeout = amdgpu_sriov_is_pp_one_vf(adev) ?
-					msecs_to_jiffies(60000) : msecs_to_jiffies(10000);
-	else
-		adev->compute_timeout =  msecs_to_jiffies(60000);
+	/* By default timeout for all queues is 2 sec */
+	adev->gfx_timeout = adev->compute_timeout = adev->sdma_timeout =
+		adev->video_timeout = msecs_to_jiffies(2000);
 
-	if (strnlen(input, AMDGPU_MAX_TIMEOUT_PARAM_LENGTH)) {
-		while ((timeout_setting = strsep(&input, ",")) &&
-				strnlen(timeout_setting, AMDGPU_MAX_TIMEOUT_PARAM_LENGTH)) {
-			ret = kstrtol(timeout_setting, 0, &timeout);
-			if (ret)
-				return ret;
+	if (!strnlen(input, AMDGPU_MAX_TIMEOUT_PARAM_LENGTH))
+		return 0;
 
-			if (timeout == 0) {
-				index++;
-				continue;
-			} else if (timeout < 0) {
-				timeout = MAX_SCHEDULE_TIMEOUT;
-				dev_warn(adev->dev, "lockup timeout disabled");
-				add_taint(TAINT_SOFTLOCKUP, LOCKDEP_STILL_OK);
-			} else {
-				timeout = msecs_to_jiffies(timeout);
-			}
+	while ((timeout_setting = strsep(&input, ",")) &&
+	       strnlen(timeout_setting, AMDGPU_MAX_TIMEOUT_PARAM_LENGTH)) {
+		ret = kstrtol(timeout_setting, 0, &timeout);
+		if (ret)
+			return ret;
 
-			switch (index++) {
-			case 0:
-				adev->gfx_timeout = timeout;
-				break;
-			case 1:
-				adev->compute_timeout = timeout;
-				break;
-			case 2:
-				adev->sdma_timeout = timeout;
-				break;
-			case 3:
-				adev->video_timeout = timeout;
-				break;
-			default:
-				break;
-			}
+		if (timeout == 0) {
+			index++;
+			continue;
+		} else if (timeout < 0) {
+			timeout = MAX_SCHEDULE_TIMEOUT;
+			dev_warn(adev->dev, "lockup timeout disabled");
+			add_taint(TAINT_SOFTLOCKUP, LOCKDEP_STILL_OK);
+		} else {
+			timeout = msecs_to_jiffies(timeout);
 		}
-		/*
-		 * There is only one value specified and
-		 * it should apply to all non-compute jobs.
-		 */
-		if (index == 1) {
-			adev->sdma_timeout = adev->video_timeout = adev->gfx_timeout;
-			if (amdgpu_sriov_vf(adev) || amdgpu_passthrough(adev))
-				adev->compute_timeout = adev->gfx_timeout;
+
+		switch (index++) {
+		case 0:
+			adev->gfx_timeout = timeout;
+			break;
+		case 1:
+			adev->compute_timeout = timeout;
+			break;
+		case 2:
+			adev->sdma_timeout = timeout;
+			break;
+		case 3:
+			adev->video_timeout = timeout;
+			break;
+		default:
+			break;
 		}
 	}
 
+	/* When only one value specified apply it to all queues. */
+	if (index == 1)
+		adev->gfx_timeout = adev->compute_timeout = adev->sdma_timeout =
+			adev->video_timeout = timeout;
+
 	return ret;
 }
 
@@ -3552,13 +4424,89 @@ static void amdgpu_device_check_iommu_direct_map(struct amdgpu_device *adev)
 		adev->ram_is_direct_mapped = true;
 }
 
-static const struct attribute *amdgpu_dev_attributes[] = {
-	&dev_attr_product_name.attr,
-	&dev_attr_product_number.attr,
-	&dev_attr_serial_number.attr,
-	&dev_attr_pcie_replay_count.attr,
-	NULL
-};
+#if defined(CONFIG_HSA_AMD_P2P)
+/**
+ * amdgpu_device_check_iommu_remap - Check if DMA remapping is enabled.
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * return if IOMMU remapping bar address
+ */
+static bool amdgpu_device_check_iommu_remap(struct amdgpu_device *adev)
+{
+	struct iommu_domain *domain;
+
+	domain = iommu_get_domain_for_dev(adev->dev);
+	if (domain && (domain->type == IOMMU_DOMAIN_DMA ||
+		domain->type ==	IOMMU_DOMAIN_DMA_FQ))
+		return true;
+
+	return false;
+}
+#endif
+
+static void amdgpu_device_set_mcbp(struct amdgpu_device *adev)
+{
+	if (amdgpu_mcbp == 1)
+		adev->gfx.mcbp = true;
+	else if (amdgpu_mcbp == 0)
+		adev->gfx.mcbp = false;
+
+	if (amdgpu_sriov_vf(adev))
+		adev->gfx.mcbp = true;
+
+	if (adev->gfx.mcbp)
+		dev_info(adev->dev, "MCBP is enabled\n");
+}
+
+static int amdgpu_device_sys_interface_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = amdgpu_atombios_sysfs_init(adev);
+	if (r)
+		drm_err(&adev->ddev,
+			"registering atombios sysfs failed (%d).\n", r);
+
+	r = amdgpu_pm_sysfs_init(adev);
+	if (r)
+		dev_err(adev->dev, "registering pm sysfs failed (%d).\n", r);
+
+	r = amdgpu_ucode_sysfs_init(adev);
+	if (r) {
+		adev->ucode_sysfs_en = false;
+		dev_err(adev->dev, "Creating firmware sysfs failed (%d).\n", r);
+	} else
+		adev->ucode_sysfs_en = true;
+
+	r = amdgpu_device_attr_sysfs_init(adev);
+	if (r)
+		dev_err(adev->dev, "Could not create amdgpu device attr\n");
+
+	r = devm_device_add_group(adev->dev, &amdgpu_board_attrs_group);
+	if (r)
+		dev_err(adev->dev,
+			"Could not create amdgpu board attributes\n");
+
+	amdgpu_fru_sysfs_init(adev);
+	amdgpu_reg_state_sysfs_init(adev);
+	amdgpu_xcp_sysfs_init(adev);
+
+	return r;
+}
+
+static void amdgpu_device_sys_interface_fini(struct amdgpu_device *adev)
+{
+	if (adev->pm.sysfs_initialized)
+		amdgpu_pm_sysfs_fini(adev);
+	if (adev->ucode_sysfs_en)
+		amdgpu_ucode_sysfs_fini(adev);
+	amdgpu_device_attr_sysfs_fini(adev);
+	amdgpu_fru_sysfs_fini(adev);
+
+	amdgpu_reg_state_sysfs_fini(adev);
+	amdgpu_xcp_sysfs_fini(adev);
+}
 
 /**
  * amdgpu_device_init - initialize the driver
@@ -3573,11 +4521,11 @@ static const struct attribute *amdgpu_dev_attributes[] = {
 int amdgpu_device_init(struct amdgpu_device *adev,
 		       uint32_t flags)
 {
-	struct drm_device *ddev = adev_to_drm(adev);
 	struct pci_dev *pdev = adev->pdev;
 	int r, i;
 	bool px = false;
 	u32 max_MBps;
+	int tmp;
 
 	adev->shutdown = false;
 	adev->flags = flags;
@@ -3607,10 +4555,14 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	adev->smc_wreg = &amdgpu_invalid_wreg;
 	adev->pcie_rreg = &amdgpu_invalid_rreg;
 	adev->pcie_wreg = &amdgpu_invalid_wreg;
+	adev->pcie_rreg_ext = &amdgpu_invalid_rreg_ext;
+	adev->pcie_wreg_ext = &amdgpu_invalid_wreg_ext;
 	adev->pciep_rreg = &amdgpu_invalid_rreg;
 	adev->pciep_wreg = &amdgpu_invalid_wreg;
 	adev->pcie_rreg64 = &amdgpu_invalid_rreg64;
 	adev->pcie_wreg64 = &amdgpu_invalid_wreg64;
+	adev->pcie_rreg64_ext = &amdgpu_invalid_rreg64_ext;
+	adev->pcie_wreg64_ext = &amdgpu_invalid_wreg64_ext;
 	adev->uvd_ctx_rreg = &amdgpu_invalid_rreg;
 	adev->uvd_ctx_wreg = &amdgpu_invalid_wreg;
 	adev->didt_rreg = &amdgpu_invalid_rreg;
@@ -3620,18 +4572,22 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	adev->audio_endpt_rreg = &amdgpu_block_invalid_rreg;
 	adev->audio_endpt_wreg = &amdgpu_block_invalid_wreg;
 
-	DRM_INFO("initializing kernel modesetting (%s 0x%04X:0x%04X 0x%04X:0x%04X 0x%02X).\n",
-		 amdgpu_asic_name[adev->asic_type], pdev->vendor, pdev->device,
-		 pdev->subsystem_vendor, pdev->subsystem_device, pdev->revision);
+	dev_info(
+		adev->dev,
+		"initializing kernel modesetting (%s 0x%04X:0x%04X 0x%04X:0x%04X 0x%02X).\n",
+		amdgpu_asic_name[adev->asic_type], pdev->vendor, pdev->device,
+		pdev->subsystem_vendor, pdev->subsystem_device, pdev->revision);
 
 	/* mutex initialization are all done here so we
-	 * can recall function without having locking issues */
+	 * can recall function without having locking issues
+	 */
 	mutex_init(&adev->firmware.mutex);
 	mutex_init(&adev->pm.mutex);
 	mutex_init(&adev->gfx.gpu_clock_mutex);
 	mutex_init(&adev->srbm_mutex);
 	mutex_init(&adev->gfx.pipe_reserve_mutex);
 	mutex_init(&adev->gfx.gfx_off_mutex);
+	mutex_init(&adev->gfx.partition_mutex);
 	mutex_init(&adev->grbm_idx_mutex);
 	mutex_init(&adev->mn_lock);
 	mutex_init(&adev->virt.vf_errors.lock);
@@ -3640,6 +4596,17 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	mutex_init(&adev->notifier_lock);
 	mutex_init(&adev->pm.stable_pstate_ctx_lock);
 	mutex_init(&adev->benchmark_mutex);
+	mutex_init(&adev->gfx.reset_sem_mutex);
+	/* Initialize the mutex for cleaner shader isolation between GFX and compute processes */
+	mutex_init(&adev->enforce_isolation_mutex);
+	for (i = 0; i < MAX_XCP; ++i) {
+		adev->isolation[i].spearhead = dma_fence_get_stub();
+		amdgpu_sync_create(&adev->isolation[i].active);
+		amdgpu_sync_create(&adev->isolation[i].prev);
+	}
+	mutex_init(&adev->gfx.userq_sch_mutex);
+	mutex_init(&adev->gfx.workload_profile_mutex);
+	mutex_init(&adev->vcn.workload_profile_mutex);
 
 	amdgpu_device_init_apu_flags(adev);
 
@@ -3656,20 +4623,41 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	spin_lock_init(&adev->se_cac_idx_lock);
 	spin_lock_init(&adev->audio_endpt_idx_lock);
 	spin_lock_init(&adev->mm_stats.lock);
+	spin_lock_init(&adev->virt.rlcg_reg_lock);
+	spin_lock_init(&adev->wb.lock);
 
-	INIT_LIST_HEAD(&adev->shadow_list);
-	mutex_init(&adev->shadow_list_lock);
+	xa_init_flags(&adev->userq_xa, XA_FLAGS_LOCK_IRQ);
 
 	INIT_LIST_HEAD(&adev->reset_list);
 
 	INIT_LIST_HEAD(&adev->ras_list);
 
+	INIT_LIST_HEAD(&adev->pm.od_kobj_list);
+
+	xa_init(&adev->userq_doorbell_xa);
+
 	INIT_DELAYED_WORK(&adev->delayed_init_work,
 			  amdgpu_device_delayed_init_work_handler);
 	INIT_DELAYED_WORK(&adev->gfx.gfx_off_delay_work,
 			  amdgpu_device_delay_enable_gfx_off);
+	/*
+	 * Initialize the enforce_isolation work structures for each XCP
+	 * partition.  This work handler is responsible for enforcing shader
+	 * isolation on AMD GPUs.  It counts the number of emitted fences for
+	 * each GFX and compute ring.  If there are any fences, it schedules
+	 * the `enforce_isolation_work` to be run after a delay.  If there are
+	 * no fences, it signals the Kernel Fusion Driver (KFD) to resume the
+	 * runqueue.
+	 */
+	for (i = 0; i < MAX_XCP; i++) {
+		INIT_DELAYED_WORK(&adev->gfx.enforce_isolation[i].work,
+				  amdgpu_gfx_enforce_isolation_handler);
+		adev->gfx.enforce_isolation[i].adev = adev;
+		adev->gfx.enforce_isolation[i].xcp_id = i;
+	}
 
 	INIT_WORK(&adev->xgmi_reset_work, amdgpu_device_xgmi_reset_func);
+	INIT_WORK(&adev->userq_reset_work, amdgpu_userq_reset_work);
 
 	adev->gfx.gfx_off_req_count = 1;
 	adev->gfx.gfx_off_residency = 0;
@@ -3685,6 +4673,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	 * for throttling interrupt) = 60 seconds.
 	 */
 	ratelimit_state_init(&adev->throttling_logging_rs, (60 - 1) * HZ, 1);
+
 	ratelimit_set_flags(&adev->throttling_logging_rs, RATELIMIT_MSG_ON_RELEASE);
 
 	/* Registers mapping */
@@ -3701,20 +4690,17 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 		atomic_set(&adev->pm.pwr_state[i], POWER_STATE_UNKNOWN);
 
 	adev->rmmio = ioremap(adev->rmmio_base, adev->rmmio_size);
-	if (adev->rmmio == NULL) {
+	if (!adev->rmmio)
 		return -ENOMEM;
-	}
-	DRM_INFO("register mmio base: 0x%08X\n", (uint32_t)adev->rmmio_base);
-	DRM_INFO("register mmio size: %u\n", (unsigned)adev->rmmio_size);
-
-	amdgpu_device_get_pcie_info(adev);
 
-	if (amdgpu_mcbp)
-		DRM_INFO("MCBP is enabled\n");
+	dev_info(adev->dev, "register mmio base: 0x%08X\n",
+		 (uint32_t)adev->rmmio_base);
+	dev_info(adev->dev, "register mmio size: %u\n",
+		 (unsigned int)adev->rmmio_size);
 
 	/*
 	 * Reset domain needs to be present early, before XGMI hive discovered
-	 * (if any) and intitialized to use reset sem and in_gpu reset flag
+	 * (if any) and initialized to use reset sem and in_gpu reset flag
 	 * early on during init and before calling to RREG32.
 	 */
 	adev->reset_domain = amdgpu_reset_create_reset_domain(SINGLE_DEVICE, "amdgpu-reset-dev");
@@ -3722,7 +4708,9 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 		return -ENOMEM;
 
 	/* detect hw virtualization here */
-	amdgpu_detect_virtualization(adev);
+	amdgpu_virt_init(adev);
+
+	amdgpu_device_get_pcie_info(adev);
 
 	r = amdgpu_device_get_job_timeout_settings(adev);
 	if (r) {
@@ -3730,19 +4718,41 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 		return r;
 	}
 
+	amdgpu_device_set_mcbp(adev);
+
+	/*
+	 * By default, use default mode where all blocks are expected to be
+	 * initialized. At present a 'swinit' of blocks is required to be
+	 * completed before the need for a different level is detected.
+	 */
+	amdgpu_set_init_level(adev, AMDGPU_INIT_LEVEL_DEFAULT);
 	/* early init functions */
 	r = amdgpu_device_ip_early_init(adev);
 	if (r)
 		return r;
 
-	/* Get rid of things like offb */
-	r = drm_aperture_remove_conflicting_pci_framebuffers(adev->pdev, &amdgpu_kms_driver);
-	if (r)
-		return r;
+	/*
+	 * No need to remove conflicting FBs for non-display class devices.
+	 * This prevents the sysfb from being freed accidently.
+	 */
+	if ((pdev->class >> 8) == PCI_CLASS_DISPLAY_VGA ||
+	    (pdev->class >> 8) == PCI_CLASS_DISPLAY_OTHER) {
+		/* Get rid of things like offb */
+		r = aperture_remove_conflicting_pci_devices(adev->pdev, amdgpu_kms_driver.name);
+		if (r)
+			return r;
+	}
 
 	/* Enable TMZ based on IP_VERSION */
 	amdgpu_gmc_tmz_set(adev);
 
+	if (amdgpu_sriov_vf(adev) &&
+	    amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 3, 0))
+		/* VF MMIO access (except mailbox range) from CPU
+		 * will be blocked during sriov runtime
+		 */
+		adev->virt.caps |= AMDGPU_VF_MMIO_ACCESS_PROTECT;
+
 	amdgpu_gmc_noretry_set(adev);
 	/* Need to get xgmi info early to decide the reset behavior*/
 	if (adev->gmc.xgmi.supported) {
@@ -3752,20 +4762,30 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	}
 
 	/* enable PCIE atomic ops */
-	if (amdgpu_sriov_vf(adev))
-		adev->have_atomics_support = ((struct amd_sriov_msg_pf2vf_info *)
-			adev->virt.fw_reserve.p_pf2vf)->pcie_atomic_ops_support_flags ==
-			(PCI_EXP_DEVCAP2_ATOMIC_COMP32 | PCI_EXP_DEVCAP2_ATOMIC_COMP64);
-	else
+	if (amdgpu_sriov_vf(adev)) {
+		if (adev->virt.fw_reserve.p_pf2vf)
+			adev->have_atomics_support = ((struct amd_sriov_msg_pf2vf_info *)
+						      adev->virt.fw_reserve.p_pf2vf)->pcie_atomic_ops_support_flags ==
+				(PCI_EXP_DEVCAP2_ATOMIC_COMP32 | PCI_EXP_DEVCAP2_ATOMIC_COMP64);
+	/* APUs w/ gfx9 onwards doesn't reply on PCIe atomics, rather it is a
+	 * internal path natively support atomics, set have_atomics_support to true.
+	 */
+	} else if ((adev->flags & AMD_IS_APU) &&
+		   (amdgpu_ip_version(adev, GC_HWIP, 0) >
+		    IP_VERSION(9, 0, 0))) {
+		adev->have_atomics_support = true;
+	} else {
 		adev->have_atomics_support =
 			!pci_enable_atomic_ops_to_root(adev->pdev,
 					  PCI_EXP_DEVCAP2_ATOMIC_COMP32 |
 					  PCI_EXP_DEVCAP2_ATOMIC_COMP64);
+	}
+
 	if (!adev->have_atomics_support)
 		dev_info(adev->dev, "PCIE atomic ops is not supported\n");
 
 	/* doorbell bar mapping and doorbell index init*/
-	amdgpu_device_doorbell_init(adev);
+	amdgpu_doorbell_init(adev);
 
 	if (amdgpu_emu_mode == 1) {
 		/* post the asic on emulation mode */
@@ -3776,7 +4796,8 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	amdgpu_reset_init(adev);
 
 	/* detect if we are with an SRIOV vbios */
-	amdgpu_device_detect_sriov_bios(adev);
+	if (adev->bios)
+		amdgpu_device_detect_sriov_bios(adev);
 
 	/* check if we need to reset the asic
 	 *  E.g., driver was not cleanly unloaded previously, etc.
@@ -3784,26 +4805,24 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	if (!amdgpu_sriov_vf(adev) && amdgpu_asic_need_reset_on_init(adev)) {
 		if (adev->gmc.xgmi.num_physical_nodes) {
 			dev_info(adev->dev, "Pending hive reset.\n");
-			adev->gmc.xgmi.pending_reset = true;
-			/* Only need to init necessary block for SMU to handle the reset */
-			for (i = 0; i < adev->num_ip_blocks; i++) {
-				if (!adev->ip_blocks[i].status.valid)
-					continue;
-				if (!(adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC ||
-				      adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON ||
-				      adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH ||
-				      adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC)) {
-					DRM_DEBUG("IP %s disabled for hw_init.\n",
-						adev->ip_blocks[i].version->funcs->name);
-					adev->ip_blocks[i].status.hw = true;
-				}
-			}
+			amdgpu_set_init_level(adev,
+					      AMDGPU_INIT_LEVEL_MINIMAL_XGMI);
+		} else if (amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 10) &&
+				   !amdgpu_device_has_display_hardware(adev)) {
+					r = psp_gpu_reset(adev);
 		} else {
-			r = amdgpu_asic_reset(adev);
-			if (r) {
-				dev_err(adev->dev, "asic reset on init failed\n");
-				goto failed;
-			}
+				tmp = amdgpu_reset_method;
+				/* It should do a default reset when loading or reloading the driver,
+				 * regardless of the module parameter reset_method.
+				 */
+				amdgpu_reset_method = AMD_RESET_METHOD_NONE;
+				r = amdgpu_asic_reset(adev);
+				amdgpu_reset_method = tmp;
+		}
+
+		if (r) {
+		  dev_err(adev->dev, "asic reset on init failed\n");
+		  goto failed;
 		}
 	}
 
@@ -3814,7 +4833,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 			r = -EINVAL;
 			goto failed;
 		}
-		DRM_INFO("GPU posting now...\n");
+		dev_info(adev->dev, "GPU posting now...\n");
 		r = amdgpu_device_asic_init(adev);
 		if (r) {
 			dev_err(adev->dev, "gpu post error!\n");
@@ -3822,25 +4841,26 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 		}
 	}
 
-	if (adev->is_atom_fw) {
-		/* Initialize clocks */
-		r = amdgpu_atomfirmware_get_clock_info(adev);
-		if (r) {
-			dev_err(adev->dev, "amdgpu_atomfirmware_get_clock_info failed\n");
-			amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
-			goto failed;
-		}
-	} else {
-		/* Initialize clocks */
-		r = amdgpu_atombios_get_clock_info(adev);
-		if (r) {
-			dev_err(adev->dev, "amdgpu_atombios_get_clock_info failed\n");
-			amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
-			goto failed;
+	if (adev->bios) {
+		if (adev->is_atom_fw) {
+			/* Initialize clocks */
+			r = amdgpu_atomfirmware_get_clock_info(adev);
+			if (r) {
+				dev_err(adev->dev, "amdgpu_atomfirmware_get_clock_info failed\n");
+				amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
+				goto failed;
+			}
+		} else {
+			/* Initialize clocks */
+			r = amdgpu_atombios_get_clock_info(adev);
+			if (r) {
+				dev_err(adev->dev, "amdgpu_atombios_get_clock_info failed\n");
+				amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
+				goto failed;
+			}
+			/* init i2c buses */
+			amdgpu_i2c_init(adev);
 		}
-		/* init i2c buses */
-		if (!amdgpu_device_has_dc_support(adev))
-			amdgpu_atombios_i2c_init(adev);
 	}
 
 fence_driver_init:
@@ -3857,18 +4877,6 @@ fence_driver_init:
 
 	r = amdgpu_device_ip_init(adev);
 	if (r) {
-		/* failed in exclusive mode due to timeout */
-		if (amdgpu_sriov_vf(adev) &&
-		    !amdgpu_sriov_runtime(adev) &&
-		    amdgpu_virt_mmio_blocked(adev) &&
-		    !amdgpu_virt_wait_reset(adev)) {
-			dev_err(adev->dev, "VF exclusive mode timeout\n");
-			/* Don't send request since VF is inactive. */
-			adev->virt.caps &= ~AMDGPU_SRIOV_CAPS_RUNTIME;
-			adev->virt.ops = NULL;
-			r = -EAGAIN;
-			goto release_ras_con;
-		}
 		dev_err(adev->dev, "amdgpu_device_ip_init failed\n");
 		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_INIT_FAIL, 0, 0);
 		goto release_ras_con;
@@ -3895,25 +4903,6 @@ fence_driver_init:
 	/* Get a log2 for easy divisions. */
 	adev->mm_stats.log2_max_MBps = ilog2(max(1u, max_MBps));
 
-	r = amdgpu_pm_sysfs_init(adev);
-	if (r)
-		DRM_ERROR("registering pm sysfs failed (%d).\n", r);
-
-	r = amdgpu_ucode_sysfs_init(adev);
-	if (r) {
-		adev->ucode_sysfs_en = false;
-		DRM_ERROR("Creating firmware sysfs failed (%d).\n", r);
-	} else
-		adev->ucode_sysfs_en = true;
-
-	r = amdgpu_psp_sysfs_init(adev);
-	if (r) {
-		adev->psp_sysfs_en = false;
-		if (!amdgpu_sriov_vf(adev))
-			DRM_ERROR("Creating psp sysfs failed\n");
-	} else
-		adev->psp_sysfs_en = true;
-
 	/*
 	 * Register gpu instance before amdgpu_device_enable_mgpu_fan_boost.
 	 * Otherwise the mgpu fan boost feature will be skipped due to the
@@ -3924,7 +4913,7 @@ fence_driver_init:
 	/* enable clockgating, etc. after ib tests, etc. since some blocks require
 	 * explicit gating rather than handling it automatically.
 	 */
-	if (!adev->gmc.xgmi.pending_reset) {
+	if (adev->init_lvl->level != AMDGPU_INIT_LEVEL_MINIMAL_XGMI) {
 		r = amdgpu_device_ip_late_init(adev);
 		if (r) {
 			dev_err(adev->dev, "amdgpu_device_ip_late_init failed\n");
@@ -3937,12 +4926,19 @@ fence_driver_init:
 				   msecs_to_jiffies(AMDGPU_RESUME_MS));
 	}
 
-	if (amdgpu_sriov_vf(adev))
+	if (amdgpu_sriov_vf(adev)) {
+		amdgpu_virt_release_full_gpu(adev, true);
 		flush_delayed_work(&adev->delayed_init_work);
+	}
 
-	r = sysfs_create_files(&adev->dev->kobj, amdgpu_dev_attributes);
-	if (r)
-		dev_err(adev->dev, "Could not create amdgpu device attr\n");
+	if (adev->init_lvl->level == AMDGPU_INIT_LEVEL_MINIMAL_XGMI)
+		amdgpu_xgmi_reset_on_init(adev);
+	/*
+	 * Place those sysfs registering after `late_init`. As some of those
+	 * operations performed in `late_init` might affect the sysfs
+	 * interfaces creating.
+	 */
+	r = amdgpu_device_sys_interface_init(adev);
 
 	if (IS_ENABLED(CONFIG_PERF_EVENTS))
 		r = amdgpu_pmu_init(adev);
@@ -3955,13 +4951,14 @@ fence_driver_init:
 
 	/* if we have > 1 VGA cards, then disable the amdgpu VGA resources */
 	/* this will fail for cards that aren't VGA class devices, just
-	 * ignore it */
+	 * ignore it
+	 */
 	if ((adev->pdev->class >> 8) == PCI_CLASS_DISPLAY_VGA)
 		vga_client_register(adev->pdev, amdgpu_device_vga_set_decode);
 
-	px = amdgpu_device_supports_px(ddev);
+	px = amdgpu_device_supports_px(adev);
 
-	if (px || (!pci_is_thunderbolt_attached(adev->pdev) &&
+	if (px || (!dev_is_removable(&adev->pdev->dev) &&
 				apple_gmux_detect(NULL, NULL)))
 		vga_switcheroo_register_client(adev->pdev,
 					       &amdgpu_switcheroo_ops, px);
@@ -3969,15 +4966,30 @@ fence_driver_init:
 	if (px)
 		vga_switcheroo_init_domain_pm_ops(adev->dev, &adev->vga_pm_domain);
 
-	if (adev->gmc.xgmi.pending_reset)
-		queue_delayed_work(system_wq, &mgpu_info.delayed_reset_work,
-				   msecs_to_jiffies(AMDGPU_RESUME_MS));
-
 	amdgpu_device_check_iommu_direct_map(adev);
 
+	adev->pm_nb.notifier_call = amdgpu_device_pm_notifier;
+	r = register_pm_notifier(&adev->pm_nb);
+	if (r)
+		goto failed;
+
 	return 0;
 
 release_ras_con:
+	if (amdgpu_sriov_vf(adev))
+		amdgpu_virt_release_full_gpu(adev, true);
+
+	/* failed in exclusive mode due to timeout */
+	if (amdgpu_sriov_vf(adev) &&
+		!amdgpu_sriov_runtime(adev) &&
+		amdgpu_virt_mmio_blocked(adev) &&
+		!amdgpu_virt_wait_reset(adev)) {
+		dev_err(adev->dev, "VF exclusive mode timeout\n");
+		/* Don't send request since VF is inactive. */
+		adev->virt.caps &= ~AMDGPU_SRIOV_CAPS_RUNTIME;
+		adev->virt.ops = NULL;
+		r = -EAGAIN;
+	}
 	amdgpu_release_ras_context(adev);
 
 failed:
@@ -3993,7 +5005,7 @@ static void amdgpu_device_unmap_mmio(struct amdgpu_device *adev)
 	unmap_mapping_range(adev->ddev.anon_inode->i_mapping, 0, 0, 1);
 
 	/* Unmap all mapped bars - Doorbell, registers and VRAM */
-	amdgpu_device_doorbell_fini(adev);
+	amdgpu_doorbell_fini(adev);
 
 	iounmap(adev->rmmio);
 	adev->rmmio = NULL;
@@ -4002,7 +5014,7 @@ static void amdgpu_device_unmap_mmio(struct amdgpu_device *adev)
 	adev->mman.aper_base_kaddr = NULL;
 
 	/* Memory manager related */
-	if (!adev->gmc.xgmi.connected_to_cpu) {
+	if (!adev->gmc.xgmi.connected_to_cpu && !adev->gmc.is_app_apu) {
 		arch_phys_wc_del(adev->gmc.vram_mtrr);
 		arch_io_free_memtype_wc(adev->gmc.aper_base, adev->gmc.aper_size);
 	}
@@ -4020,11 +5032,16 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev)
 {
 	dev_info(adev->dev, "amdgpu: finishing device.\n");
 	flush_delayed_work(&adev->delayed_init_work);
+
+	if (adev->mman.initialized)
+		drain_workqueue(adev->mman.bdev.wq);
 	adev->shutdown = true;
 
+	unregister_pm_notifier(&adev->pm_nb);
+
 	/* make sure IB test finished before entering exclusive mode
 	 * to avoid preemption on IB test
-	 * */
+	 */
 	if (amdgpu_sriov_vf(adev)) {
 		amdgpu_virt_request_full_gpu(adev, false);
 		amdgpu_virt_fini_data_exchange(adev);
@@ -4032,7 +5049,7 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev)
 
 	/* disable all interrupts */
 	amdgpu_irq_disable_all(adev);
-	if (adev->mode_info.mode_config_initialized){
+	if (adev->mode_info.mode_config_initialized) {
 		if (!drm_drv_uses_atomic_modeset(adev_to_drm(adev)))
 			drm_helper_force_disable_all(adev_to_drm(adev));
 		else
@@ -4040,20 +5057,13 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev)
 	}
 	amdgpu_fence_driver_hw_fini(adev);
 
-	if (adev->mman.initialized)
-		drain_workqueue(adev->mman.bdev.wq);
-
-	if (adev->pm.sysfs_initialized)
-		amdgpu_pm_sysfs_fini(adev);
-	if (adev->ucode_sysfs_en)
-		amdgpu_ucode_sysfs_fini(adev);
-	if (adev->psp_sysfs_en)
-		amdgpu_psp_sysfs_fini(adev);
-	sysfs_remove_files(&adev->dev->kobj, amdgpu_dev_attributes);
+	amdgpu_device_sys_interface_fini(adev);
 
 	/* disable ras feature must before hw fini */
 	amdgpu_ras_pre_fini(adev);
 
+	amdgpu_ttm_set_buffer_funcs_status(adev, false);
+
 	amdgpu_device_ip_fini_early(adev);
 
 	amdgpu_irq_fini_hw(adev);
@@ -4070,30 +5080,40 @@ void amdgpu_device_fini_hw(struct amdgpu_device *adev)
 
 void amdgpu_device_fini_sw(struct amdgpu_device *adev)
 {
-	int idx;
+	int i, idx;
 	bool px;
 
-	amdgpu_fence_driver_sw_fini(adev);
 	amdgpu_device_ip_fini(adev);
+	amdgpu_fence_driver_sw_fini(adev);
 	amdgpu_ucode_release(&adev->firmware.gpu_info_fw);
 	adev->accel_working = false;
 	dma_fence_put(rcu_dereference_protected(adev->gang_submit, true));
+	for (i = 0; i < MAX_XCP; ++i) {
+		dma_fence_put(adev->isolation[i].spearhead);
+		amdgpu_sync_free(&adev->isolation[i].active);
+		amdgpu_sync_free(&adev->isolation[i].prev);
+	}
 
 	amdgpu_reset_fini(adev);
 
 	/* free i2c buses */
-	if (!amdgpu_device_has_dc_support(adev))
-		amdgpu_i2c_fini(adev);
+	amdgpu_i2c_fini(adev);
+
+	if (adev->bios) {
+		if (amdgpu_emu_mode != 1)
+			amdgpu_atombios_fini(adev);
+		amdgpu_bios_release(adev);
+	}
 
-	if (amdgpu_emu_mode != 1)
-		amdgpu_atombios_fini(adev);
+	kfree(adev->fru_info);
+	adev->fru_info = NULL;
 
-	kfree(adev->bios);
-	adev->bios = NULL;
+	kfree(adev->xcp_mgr);
+	adev->xcp_mgr = NULL;
 
-	px = amdgpu_device_supports_px(adev_to_drm(adev));
+	px = amdgpu_device_supports_px(adev);
 
-	if (px || (!pci_is_thunderbolt_attached(adev->pdev) &&
+	if (px || (!dev_is_removable(&adev->pdev->dev) &&
 				apple_gmux_detect(NULL, NULL)))
 		vga_switcheroo_unregister_client(adev->pdev);
 
@@ -4107,20 +5127,20 @@ void amdgpu_device_fini_sw(struct amdgpu_device *adev)
 
 		iounmap(adev->rmmio);
 		adev->rmmio = NULL;
-		amdgpu_device_doorbell_fini(adev);
 		drm_dev_exit(idx);
 	}
 
 	if (IS_ENABLED(CONFIG_PERF_EVENTS))
 		amdgpu_pmu_fini(adev);
-	if (adev->mman.discovery_bin)
+	if (adev->discovery.bin)
 		amdgpu_discovery_fini(adev);
 
 	amdgpu_reset_put_reset_domain(adev->reset_domain);
 	adev->reset_domain = NULL;
 
 	kfree(adev->pci_state);
-
+	kfree(adev->pcie_reset_ctx.swds_pcistate);
+	kfree(adev->pcie_reset_ctx.swus_pcistate);
 }
 
 /**
@@ -4136,13 +5156,25 @@ static int amdgpu_device_evict_resources(struct amdgpu_device *adev)
 {
 	int ret;
 
-	/* No need to evict vram on APUs for suspend to ram or s2idle */
-	if ((adev->in_s3 || adev->in_s0ix) && (adev->flags & AMD_IS_APU))
+	/* No need to evict vram on APUs unless going to S4 */
+	if (!adev->in_s4 && (adev->flags & AMD_IS_APU))
+		return 0;
+
+	/* No need to evict when going to S5 through S4 callbacks */
+	if (system_state == SYSTEM_POWER_OFF)
 		return 0;
 
 	ret = amdgpu_ttm_evict_resources(adev, TTM_PL_VRAM);
-	if (ret)
-		DRM_WARN("evicting device resources failed\n");
+	if (ret) {
+		dev_warn(adev->dev, "evicting device resources failed\n");
+		return ret;
+	}
+
+	if (adev->in_s4) {
+		ret = ttm_device_prepare_hibernation(&adev->mman.bdev);
+		if (ret)
+			dev_err(adev->dev, "prepare hibernation failed, %d\n", ret);
+	}
 	return ret;
 }
 
@@ -4150,77 +5182,245 @@ static int amdgpu_device_evict_resources(struct amdgpu_device *adev)
  * Suspend & resume.
  */
 /**
- * amdgpu_device_suspend - initiate device suspend
+ * amdgpu_device_pm_notifier - Notification block for Suspend/Hibernate events
+ * @nb: notifier block
+ * @mode: suspend mode
+ * @data: data
+ *
+ * This function is called when the system is about to suspend or hibernate.
+ * It is used to set the appropriate flags so that eviction can be optimized
+ * in the pm prepare callback.
+ */
+static int amdgpu_device_pm_notifier(struct notifier_block *nb, unsigned long mode,
+				     void *data)
+{
+	struct amdgpu_device *adev = container_of(nb, struct amdgpu_device, pm_nb);
+
+	switch (mode) {
+	case PM_HIBERNATION_PREPARE:
+		adev->in_s4 = true;
+		break;
+	case PM_POST_HIBERNATION:
+		adev->in_s4 = false;
+		break;
+	}
+
+	return NOTIFY_DONE;
+}
+
+/**
+ * amdgpu_device_prepare - prepare for device suspend
  *
  * @dev: drm dev pointer
- * @fbcon : notify the fbdev of suspend
  *
- * Puts the hw in the suspend state (all asics).
+ * Prepare to put the hw in the suspend state (all asics).
  * Returns 0 for success or an error on failure.
  * Called at driver suspend.
  */
-int amdgpu_device_suspend(struct drm_device *dev, bool fbcon)
+int amdgpu_device_prepare(struct drm_device *dev)
 {
 	struct amdgpu_device *adev = drm_to_adev(dev);
-	int r = 0;
+	int i, r;
 
 	if (dev->switch_power_state == DRM_SWITCH_POWER_OFF)
 		return 0;
 
-	adev->in_suspend = true;
-
-	/* Evict the majority of BOs before grabbing the full access */
+	/* Evict the majority of BOs before starting suspend sequence */
 	r = amdgpu_device_evict_resources(adev);
 	if (r)
 		return r;
 
+	flush_delayed_work(&adev->gfx.gfx_off_delay_work);
+
+	for (i = 0; i < adev->num_ip_blocks; i++) {
+		if (!adev->ip_blocks[i].status.valid)
+			continue;
+		if (!adev->ip_blocks[i].version->funcs->prepare_suspend)
+			continue;
+		r = adev->ip_blocks[i].version->funcs->prepare_suspend(&adev->ip_blocks[i]);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+/**
+ * amdgpu_device_complete - complete power state transition
+ *
+ * @dev: drm dev pointer
+ *
+ * Undo the changes from amdgpu_device_prepare. This will be
+ * called on all resume transitions, including those that failed.
+ */
+void amdgpu_device_complete(struct drm_device *dev)
+{
+	struct amdgpu_device *adev = drm_to_adev(dev);
+	int i;
+
+	for (i = 0; i < adev->num_ip_blocks; i++) {
+		if (!adev->ip_blocks[i].status.valid)
+			continue;
+		if (!adev->ip_blocks[i].version->funcs->complete)
+			continue;
+		adev->ip_blocks[i].version->funcs->complete(&adev->ip_blocks[i]);
+	}
+}
+
+/**
+ * amdgpu_device_suspend - initiate device suspend
+ *
+ * @dev: drm dev pointer
+ * @notify_clients: notify in-kernel DRM clients
+ *
+ * Puts the hw in the suspend state (all asics).
+ * Returns 0 for success or an error on failure.
+ * Called at driver suspend.
+ */
+int amdgpu_device_suspend(struct drm_device *dev, bool notify_clients)
+{
+	struct amdgpu_device *adev = drm_to_adev(dev);
+	int r, rec;
+
+	if (dev->switch_power_state == DRM_SWITCH_POWER_OFF)
+		return 0;
+
+	adev->in_suspend = true;
+
 	if (amdgpu_sriov_vf(adev)) {
+		if (!adev->in_runpm)
+			amdgpu_amdkfd_suspend_process(adev);
 		amdgpu_virt_fini_data_exchange(adev);
 		r = amdgpu_virt_request_full_gpu(adev, false);
 		if (r)
 			return r;
 	}
 
-	if (amdgpu_acpi_smart_shift_update(dev, AMDGPU_SS_DEV_D3))
-		DRM_WARN("smart shift update failed\n");
+	r = amdgpu_acpi_smart_shift_update(adev, AMDGPU_SS_DEV_D3);
+	if (r)
+		goto unwind_sriov;
 
-	if (fbcon)
-		drm_fb_helper_set_suspend_unlocked(adev_to_drm(adev)->fb_helper, true);
+	if (notify_clients)
+		drm_client_dev_suspend(adev_to_drm(adev));
 
 	cancel_delayed_work_sync(&adev->delayed_init_work);
 
 	amdgpu_ras_suspend(adev);
 
-	amdgpu_device_ip_suspend_phase1(adev);
+	r = amdgpu_device_ip_suspend_phase1(adev);
+	if (r)
+		goto unwind_smartshift;
 
-	if (!adev->in_s0ix)
-		amdgpu_amdkfd_suspend(adev, adev->in_runpm);
+	amdgpu_amdkfd_suspend(adev, !amdgpu_sriov_vf(adev) && !adev->in_runpm);
+	r = amdgpu_userq_suspend(adev);
+	if (r)
+		goto unwind_ip_phase1;
 
 	r = amdgpu_device_evict_resources(adev);
 	if (r)
-		return r;
+		goto unwind_userq;
+
+	amdgpu_ttm_set_buffer_funcs_status(adev, false);
 
 	amdgpu_fence_driver_hw_fini(adev);
 
-	amdgpu_device_ip_suspend_phase2(adev);
+	r = amdgpu_device_ip_suspend_phase2(adev);
+	if (r)
+		goto unwind_evict;
 
 	if (amdgpu_sriov_vf(adev))
 		amdgpu_virt_release_full_gpu(adev, false);
 
 	return 0;
+
+unwind_evict:
+	if (adev->mman.buffer_funcs_ring->sched.ready)
+		amdgpu_ttm_set_buffer_funcs_status(adev, true);
+	amdgpu_fence_driver_hw_init(adev);
+
+unwind_userq:
+	rec = amdgpu_userq_resume(adev);
+	if (rec) {
+		dev_warn(adev->dev, "failed to re-initialize user queues: %d\n", rec);
+		return r;
+	}
+	rec = amdgpu_amdkfd_resume(adev, !amdgpu_sriov_vf(adev) && !adev->in_runpm);
+	if (rec) {
+		dev_warn(adev->dev, "failed to re-initialize kfd: %d\n", rec);
+		return r;
+	}
+
+unwind_ip_phase1:
+	/* suspend phase 1 = resume phase 3 */
+	rec = amdgpu_device_ip_resume_phase3(adev);
+	if (rec) {
+		dev_warn(adev->dev, "failed to re-initialize IPs phase1: %d\n", rec);
+		return r;
+	}
+
+unwind_smartshift:
+	rec = amdgpu_acpi_smart_shift_update(adev, AMDGPU_SS_DEV_D0);
+	if (rec) {
+		dev_warn(adev->dev, "failed to re-update smart shift: %d\n", rec);
+		return r;
+	}
+
+	if (notify_clients)
+		drm_client_dev_resume(adev_to_drm(adev));
+
+	amdgpu_ras_resume(adev);
+
+unwind_sriov:
+	if (amdgpu_sriov_vf(adev)) {
+		rec = amdgpu_virt_request_full_gpu(adev, true);
+		if (rec) {
+			dev_warn(adev->dev, "failed to reinitialize sriov: %d\n", rec);
+			return r;
+		}
+	}
+
+	adev->in_suspend = adev->in_s0ix = adev->in_s3 = false;
+
+	return r;
+}
+
+static inline int amdgpu_virt_resume(struct amdgpu_device *adev)
+{
+	int r;
+	unsigned int prev_physical_node_id = adev->gmc.xgmi.physical_node_id;
+
+	/* During VM resume, QEMU programming of VF MSIX table (register GFXMSIX_VECT0_ADDR_LO)
+	 * may not work. The access could be blocked by nBIF protection as VF isn't in
+	 * exclusive access mode. Exclusive access is enabled now, disable/enable MSIX
+	 * so that QEMU reprograms MSIX table.
+	 */
+	amdgpu_restore_msix(adev);
+
+	r = adev->gfxhub.funcs->get_xgmi_info(adev);
+	if (r)
+		return r;
+
+	dev_info(adev->dev, "xgmi node, old id %d, new id %d\n",
+		prev_physical_node_id, adev->gmc.xgmi.physical_node_id);
+
+	adev->vm_manager.vram_base_offset = adev->gfxhub.funcs->get_mc_fb_offset(adev);
+	adev->vm_manager.vram_base_offset +=
+		adev->gmc.xgmi.physical_node_id * adev->gmc.xgmi.node_segment_size;
+
+	return 0;
 }
 
 /**
  * amdgpu_device_resume - initiate device resume
  *
  * @dev: drm dev pointer
- * @fbcon : notify the fbdev of resume
+ * @notify_clients: notify in-kernel DRM clients
  *
  * Bring the hw back to operating state (all asics).
  * Returns 0 for success or an error on failure.
  * Called at driver resume.
  */
-int amdgpu_device_resume(struct drm_device *dev, bool fbcon)
+int amdgpu_device_resume(struct drm_device *dev, bool notify_clients)
 {
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	int r = 0;
@@ -4231,6 +5431,12 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon)
 			return r;
 	}
 
+	if (amdgpu_virt_xgmi_migrate_enabled(adev)) {
+		r = amdgpu_virt_resume(adev);
+		if (r)
+			goto exit;
+	}
+
 	if (dev->switch_power_state == DRM_SWITCH_POWER_OFF)
 		return 0;
 
@@ -4250,7 +5456,14 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon)
 		dev_err(adev->dev, "amdgpu_device_ip_resume failed (%d).\n", r);
 		goto exit;
 	}
-	amdgpu_fence_driver_hw_init(adev);
+
+	r = amdgpu_amdkfd_resume(adev, !amdgpu_sriov_vf(adev) && !adev->in_runpm);
+	if (r)
+		goto exit;
+
+	r = amdgpu_userq_resume(adev);
+	if (r)
+		goto exit;
 
 	r = amdgpu_device_ip_late_init(adev);
 	if (r)
@@ -4258,17 +5471,13 @@ int amdgpu_device_resume(struct drm_device *dev, bool fbcon)
 
 	queue_delayed_work(system_wq, &adev->delayed_init_work,
 			   msecs_to_jiffies(AMDGPU_RESUME_MS));
-
-	if (!adev->in_s0ix) {
-		r = amdgpu_amdkfd_resume(adev, adev->in_runpm);
-		if (r)
-			goto exit;
-	}
-
 exit:
 	if (amdgpu_sriov_vf(adev)) {
 		amdgpu_virt_init_data_exchange(adev);
 		amdgpu_virt_release_full_gpu(adev, true);
+
+		if (!r && !adev->in_runpm)
+			r = amdgpu_amdkfd_resume_process(adev);
 	}
 
 	if (r)
@@ -4277,8 +5486,8 @@ exit:
 	/* Make sure IB tests flushed */
 	flush_delayed_work(&adev->delayed_init_work);
 
-	if (fbcon)
-		drm_fb_helper_set_suspend_unlocked(adev_to_drm(adev)->fb_helper, false);
+	if (notify_clients)
+		drm_client_dev_resume(adev_to_drm(adev));
 
 	amdgpu_ras_resume(adev);
 
@@ -4303,13 +5512,12 @@ exit:
 		dev->dev->power.disable_depth--;
 #endif
 	}
-	adev->in_suspend = false;
 
-	if (adev->enable_mes)
-		amdgpu_mes_self_test(adev);
+	amdgpu_vram_mgr_clear_reset_blocks(adev);
+	adev->in_suspend = false;
 
-	if (amdgpu_acpi_smart_shift_update(dev, AMDGPU_SS_DEV_D0))
-		DRM_WARN("smart shift update failed\n");
+	if (amdgpu_acpi_smart_shift_update(adev, AMDGPU_SS_DEV_D0))
+		dev_warn(adev->dev, "smart shift update failed\n");
 
 	return 0;
 }
@@ -4340,7 +5548,8 @@ static bool amdgpu_device_ip_check_soft_reset(struct amdgpu_device *adev)
 			continue;
 		if (adev->ip_blocks[i].version->funcs->check_soft_reset)
 			adev->ip_blocks[i].status.hang =
-				adev->ip_blocks[i].version->funcs->check_soft_reset(adev);
+				adev->ip_blocks[i].version->funcs->check_soft_reset(
+					&adev->ip_blocks[i]);
 		if (adev->ip_blocks[i].status.hang) {
 			dev_info(adev->dev, "IP block:%s is hung!\n", adev->ip_blocks[i].version->funcs->name);
 			asic_hang = true;
@@ -4369,7 +5578,7 @@ static int amdgpu_device_ip_pre_soft_reset(struct amdgpu_device *adev)
 			continue;
 		if (adev->ip_blocks[i].status.hang &&
 		    adev->ip_blocks[i].version->funcs->pre_soft_reset) {
-			r = adev->ip_blocks[i].version->funcs->pre_soft_reset(adev);
+			r = adev->ip_blocks[i].version->funcs->pre_soft_reset(&adev->ip_blocks[i]);
 			if (r)
 				return r;
 		}
@@ -4431,7 +5640,7 @@ static int amdgpu_device_ip_soft_reset(struct amdgpu_device *adev)
 			continue;
 		if (adev->ip_blocks[i].status.hang &&
 		    adev->ip_blocks[i].version->funcs->soft_reset) {
-			r = adev->ip_blocks[i].version->funcs->soft_reset(adev);
+			r = adev->ip_blocks[i].version->funcs->soft_reset(&adev->ip_blocks[i]);
 			if (r)
 				return r;
 		}
@@ -4460,7 +5669,7 @@ static int amdgpu_device_ip_post_soft_reset(struct amdgpu_device *adev)
 			continue;
 		if (adev->ip_blocks[i].status.hang &&
 		    adev->ip_blocks[i].version->funcs->post_soft_reset)
-			r = adev->ip_blocks[i].version->funcs->post_soft_reset(adev);
+			r = adev->ip_blocks[i].version->funcs->post_soft_reset(&adev->ip_blocks[i]);
 		if (r)
 			return r;
 	}
@@ -4469,104 +5678,42 @@ static int amdgpu_device_ip_post_soft_reset(struct amdgpu_device *adev)
 }
 
 /**
- * amdgpu_device_recover_vram - Recover some VRAM contents
- *
- * @adev: amdgpu_device pointer
- *
- * Restores the contents of VRAM buffers from the shadows in GTT.  Used to
- * restore things like GPUVM page tables after a GPU reset where
- * the contents of VRAM might be lost.
- *
- * Returns:
- * 0 on success, negative error code on failure.
- */
-static int amdgpu_device_recover_vram(struct amdgpu_device *adev)
-{
-	struct dma_fence *fence = NULL, *next = NULL;
-	struct amdgpu_bo *shadow;
-	struct amdgpu_bo_vm *vmbo;
-	long r = 1, tmo;
-
-	if (amdgpu_sriov_runtime(adev))
-		tmo = msecs_to_jiffies(8000);
-	else
-		tmo = msecs_to_jiffies(100);
-
-	dev_info(adev->dev, "recover vram bo from shadow start\n");
-	mutex_lock(&adev->shadow_list_lock);
-	list_for_each_entry(vmbo, &adev->shadow_list, shadow_list) {
-		shadow = &vmbo->bo;
-		/* No need to recover an evicted BO */
-		if (shadow->tbo.resource->mem_type != TTM_PL_TT ||
-		    shadow->tbo.resource->start == AMDGPU_BO_INVALID_OFFSET ||
-		    shadow->parent->tbo.resource->mem_type != TTM_PL_VRAM)
-			continue;
-
-		r = amdgpu_bo_restore_shadow(shadow, &next);
-		if (r)
-			break;
-
-		if (fence) {
-			tmo = dma_fence_wait_timeout(fence, false, tmo);
-			dma_fence_put(fence);
-			fence = next;
-			if (tmo == 0) {
-				r = -ETIMEDOUT;
-				break;
-			} else if (tmo < 0) {
-				r = tmo;
-				break;
-			}
-		} else {
-			fence = next;
-		}
-	}
-	mutex_unlock(&adev->shadow_list_lock);
-
-	if (fence)
-		tmo = dma_fence_wait_timeout(fence, false, tmo);
-	dma_fence_put(fence);
-
-	if (r < 0 || tmo <= 0) {
-		dev_err(adev->dev, "recover vram bo from shadow failed, r is %ld, tmo is %ld\n", r, tmo);
-		return -EIO;
-	}
-
-	dev_info(adev->dev, "recover vram bo from shadow done\n");
-	return 0;
-}
-
-
-/**
  * amdgpu_device_reset_sriov - reset ASIC for SR-IOV vf
  *
  * @adev: amdgpu_device pointer
- * @from_hypervisor: request from hypervisor
+ * @reset_context: amdgpu reset context pointer
  *
  * do VF FLR and reinitialize Asic
  * return 0 means succeeded otherwise failed
  */
 static int amdgpu_device_reset_sriov(struct amdgpu_device *adev,
-				     bool from_hypervisor)
+				     struct amdgpu_reset_context *reset_context)
 {
 	int r;
 	struct amdgpu_hive_info *hive = NULL;
-	int retry_limit = 0;
-
-retry:
-	amdgpu_amdkfd_pre_reset(adev);
 
-	if (from_hypervisor)
+	if (test_bit(AMDGPU_HOST_FLR, &reset_context->flags)) {
+		if (!amdgpu_ras_get_fed_status(adev))
+			amdgpu_virt_ready_to_reset(adev);
+		amdgpu_virt_wait_reset(adev);
+		clear_bit(AMDGPU_HOST_FLR, &reset_context->flags);
 		r = amdgpu_virt_request_full_gpu(adev, true);
-	else
+	} else {
 		r = amdgpu_virt_reset_gpu(adev);
+	}
 	if (r)
 		return r;
 
+	amdgpu_ras_clear_err_state(adev);
+	amdgpu_irq_gpu_reset_resume_helper(adev);
+
+	/* some sw clean up VF needs to do before recover */
+	amdgpu_virt_post_reset(adev);
+
 	/* Resume IP prior to SMC */
 	r = amdgpu_device_ip_reinit_early_sriov(adev);
 	if (r)
-		goto error;
+		return r;
 
 	amdgpu_virt_init_data_exchange(adev);
 
@@ -4577,64 +5724,64 @@ retry:
 	/* now we are okay to resume SMC/CP/SDMA */
 	r = amdgpu_device_ip_reinit_late_sriov(adev);
 	if (r)
-		goto error;
+		return r;
 
 	hive = amdgpu_get_xgmi_hive(adev);
 	/* Update PSP FW topology after reset */
 	if (hive && adev->gmc.xgmi.num_physical_nodes > 1)
 		r = amdgpu_xgmi_update_topology(hive, adev);
-
 	if (hive)
 		amdgpu_put_xgmi_hive(hive);
+	if (r)
+		return r;
 
-	if (!r) {
-		amdgpu_irq_gpu_reset_resume_helper(adev);
-		r = amdgpu_ib_ring_tests(adev);
-
-		amdgpu_amdkfd_post_reset(adev);
-	}
+	r = amdgpu_ib_ring_tests(adev);
+	if (r)
+		return r;
 
-error:
-	if (!r && adev->virt.gim_feature & AMDGIM_FEATURE_GIM_FLR_VRAMLOST) {
+	if (adev->virt.gim_feature & AMDGIM_FEATURE_GIM_FLR_VRAMLOST)
 		amdgpu_inc_vram_lost(adev);
-		r = amdgpu_device_recover_vram(adev);
-	}
+
+	/* need to be called during full access so we can't do it later like
+	 * bare-metal does.
+	 */
+	amdgpu_amdkfd_post_reset(adev);
 	amdgpu_virt_release_full_gpu(adev, true);
 
-	if (AMDGPU_RETRY_SRIOV_RESET(r)) {
-		if (retry_limit < AMDGPU_MAX_RETRY_LIMIT) {
-			retry_limit++;
-			goto retry;
-		} else
-			DRM_ERROR("GPU reset retry is beyond the retry limit\n");
-	}
+	/* Aldebaran and gfx_11_0_3 support ras in SRIOV, so need resume ras during reset */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 4) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 5, 0) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 0, 3))
+		amdgpu_ras_resume(adev);
 
-	return r;
+	amdgpu_virt_ras_telemetry_post_reset(adev);
+
+	return 0;
 }
 
 /**
- * amdgpu_device_has_job_running - check if there is any job in mirror list
+ * amdgpu_device_has_job_running - check if there is any unfinished job
  *
  * @adev: amdgpu_device pointer
  *
- * check if there is any job in mirror list
+ * check if there is any job running on the device when guest driver receives
+ * FLR notification from host driver. If there are still jobs running, then
+ * the guest driver will not respond the FLR reset. Instead, let the job hit
+ * the timeout and guest driver then issue the reset request.
  */
 bool amdgpu_device_has_job_running(struct amdgpu_device *adev)
 {
 	int i;
-	struct drm_sched_job *job;
 
 	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 		struct amdgpu_ring *ring = adev->rings[i];
 
-		if (!ring || !ring->sched.thread)
+		if (!amdgpu_ring_sched_ready(ring))
 			continue;
 
-		spin_lock(&ring->sched.job_list_lock);
-		job = list_first_entry_or_null(&ring->sched.pending_list,
-					       struct drm_sched_job, list);
-		spin_unlock(&ring->sched.job_list_lock);
-		if (job)
+		if (amdgpu_fence_count_emitted(ring))
 			return true;
 	}
 	return false;
@@ -4693,42 +5840,83 @@ disabled:
 
 int amdgpu_device_mode1_reset(struct amdgpu_device *adev)
 {
-        u32 i;
-        int ret = 0;
+	u32 i;
+	int ret = 0;
 
-        amdgpu_atombios_scratch_regs_engine_hung(adev, true);
+	if (adev->bios)
+		amdgpu_atombios_scratch_regs_engine_hung(adev, true);
 
-        dev_info(adev->dev, "GPU mode1 reset\n");
+	dev_info(adev->dev, "GPU mode1 reset\n");
 
-        /* disable BM */
-        pci_clear_master(adev->pdev);
+	/* Cache the state before bus master disable. The saved config space
+	 * values are used in other cases like restore after mode-2 reset.
+	 */
+	amdgpu_device_cache_pci_state(adev->pdev);
 
-        amdgpu_device_cache_pci_state(adev->pdev);
+	/* disable BM */
+	pci_clear_master(adev->pdev);
 
-        if (amdgpu_dpm_is_mode1_reset_supported(adev)) {
-                dev_info(adev->dev, "GPU smu mode1 reset\n");
-                ret = amdgpu_dpm_mode1_reset(adev);
-        } else {
-                dev_info(adev->dev, "GPU psp mode1 reset\n");
-                ret = psp_gpu_reset(adev);
-        }
+	if (amdgpu_dpm_is_mode1_reset_supported(adev)) {
+		dev_info(adev->dev, "GPU smu mode1 reset\n");
+		ret = amdgpu_dpm_mode1_reset(adev);
+	} else {
+		dev_info(adev->dev, "GPU psp mode1 reset\n");
+		ret = psp_gpu_reset(adev);
+	}
 
-        if (ret)
-                dev_err(adev->dev, "GPU mode1 reset failed\n");
+	if (ret)
+		goto mode1_reset_failed;
 
-        amdgpu_device_load_pci_state(adev->pdev);
+	amdgpu_device_load_pci_state(adev->pdev);
+	ret = amdgpu_psp_wait_for_bootloader(adev);
+	if (ret)
+		goto mode1_reset_failed;
 
-        /* wait for asic to come out of reset */
-        for (i = 0; i < adev->usec_timeout; i++) {
-                u32 memsize = adev->nbio.funcs->get_memsize(adev);
+	/* wait for asic to come out of reset */
+	for (i = 0; i < adev->usec_timeout; i++) {
+		u32 memsize = adev->nbio.funcs->get_memsize(adev);
 
-                if (memsize != 0xffffffff)
-                        break;
-                udelay(1);
-        }
+		if (memsize != 0xffffffff)
+			break;
+		udelay(1);
+	}
+
+	if (i >= adev->usec_timeout) {
+		ret = -ETIMEDOUT;
+		goto mode1_reset_failed;
+	}
+
+	if (adev->bios)
+		amdgpu_atombios_scratch_regs_engine_hung(adev, false);
+
+	return 0;
 
-        amdgpu_atombios_scratch_regs_engine_hung(adev, false);
-        return ret;
+mode1_reset_failed:
+	dev_err(adev->dev, "GPU mode1 reset failed\n");
+	return ret;
+}
+
+int amdgpu_device_link_reset(struct amdgpu_device *adev)
+{
+	int ret = 0;
+
+	dev_info(adev->dev, "GPU link reset\n");
+
+	if (!amdgpu_reset_in_dpc(adev))
+		ret = amdgpu_dpm_link_reset(adev);
+
+	if (ret)
+		goto link_reset_failed;
+
+	ret = amdgpu_psp_wait_for_bootloader(adev);
+	if (ret)
+		goto link_reset_failed;
+
+	return 0;
+
+link_reset_failed:
+	dev_err(adev->dev, "GPU link reset failed\n");
+	return ret;
 }
 
 int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
@@ -4736,16 +5924,15 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 {
 	int i, r = 0;
 	struct amdgpu_job *job = NULL;
+	struct amdgpu_device *tmp_adev = reset_context->reset_req_dev;
 	bool need_full_reset =
 		test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
 
 	if (reset_context->reset_req_dev == adev)
 		job = reset_context->job;
 
-	if (amdgpu_sriov_vf(adev)) {
-		/* stop the data exchange thread */
-		amdgpu_virt_fini_data_exchange(adev);
-	}
+	if (amdgpu_sriov_vf(adev))
+		amdgpu_virt_pre_reset(adev);
 
 	amdgpu_fence_driver_isr_toggle(adev, true);
 
@@ -4753,13 +5940,9 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 		struct amdgpu_ring *ring = adev->rings[i];
 
-		if (!ring || !ring->sched.thread)
+		if (!amdgpu_ring_sched_ready(ring))
 			continue;
 
-		/*clear job fence from fence drv to avoid force_completion
-		 *leave NULL and vm flush fence in fence drv */
-		amdgpu_fence_driver_clear_job_fences(ring);
-
 		/* after all hw jobs are reset, hw fence is meaningless, so force_completion */
 		amdgpu_fence_driver_force_completion(ring);
 	}
@@ -4771,7 +5954,7 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 
 	r = amdgpu_reset_prepare_hwcontext(adev, reset_context);
 	/* If reset handler not implemented, continue; otherwise return */
-	if (r == -ENOSYS)
+	if (r == -EOPNOTSUPP)
 		r = 0;
 	else
 		return r;
@@ -4793,6 +5976,16 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 			}
 		}
 
+		if (!test_bit(AMDGPU_SKIP_COREDUMP, &reset_context->flags)) {
+			dev_info(tmp_adev->dev, "Dumping IP State\n");
+			/* Trigger ip dump before we reset the asic */
+			for (i = 0; i < tmp_adev->num_ip_blocks; i++)
+				if (tmp_adev->ip_blocks[i].version->funcs->dump_ip_state)
+					tmp_adev->ip_blocks[i].version->funcs
+						->dump_ip_state((void *)&tmp_adev->ip_blocks[i]);
+			dev_info(tmp_adev->dev, "Dumping IP State Completed\n");
+		}
+
 		if (need_full_reset)
 			r = amdgpu_device_ip_suspend(adev);
 		if (need_full_reset)
@@ -4805,188 +5998,55 @@ int amdgpu_device_pre_asic_reset(struct amdgpu_device *adev,
 	return r;
 }
 
-static int amdgpu_reset_reg_dumps(struct amdgpu_device *adev)
-{
-	int i;
-
-	lockdep_assert_held(&adev->reset_domain->sem);
-
-	for (i = 0; i < adev->num_regs; i++) {
-		adev->reset_dump_reg_value[i] = RREG32(adev->reset_dump_reg_list[i]);
-		trace_amdgpu_reset_reg_dumps(adev->reset_dump_reg_list[i],
-					     adev->reset_dump_reg_value[i]);
-	}
-
-	return 0;
-}
-
-#ifdef CONFIG_DEV_COREDUMP
-static ssize_t amdgpu_devcoredump_read(char *buffer, loff_t offset,
-		size_t count, void *data, size_t datalen)
-{
-	struct drm_printer p;
-	struct amdgpu_device *adev = data;
-	struct drm_print_iterator iter;
-	int i;
-
-	iter.data = buffer;
-	iter.offset = 0;
-	iter.start = offset;
-	iter.remain = count;
-
-	p = drm_coredump_printer(&iter);
-
-	drm_printf(&p, "**** AMDGPU Device Coredump ****\n");
-	drm_printf(&p, "kernel: " UTS_RELEASE "\n");
-	drm_printf(&p, "module: " KBUILD_MODNAME "\n");
-	drm_printf(&p, "time: %lld.%09ld\n", adev->reset_time.tv_sec, adev->reset_time.tv_nsec);
-	if (adev->reset_task_info.pid)
-		drm_printf(&p, "process_name: %s PID: %d\n",
-			   adev->reset_task_info.process_name,
-			   adev->reset_task_info.pid);
-
-	if (adev->reset_vram_lost)
-		drm_printf(&p, "VRAM is lost due to GPU reset!\n");
-	if (adev->num_regs) {
-		drm_printf(&p, "AMDGPU register dumps:\nOffset:     Value:\n");
-
-		for (i = 0; i < adev->num_regs; i++)
-			drm_printf(&p, "0x%08x: 0x%08x\n",
-				   adev->reset_dump_reg_list[i],
-				   adev->reset_dump_reg_value[i]);
-	}
-
-	return count - iter.remain;
-}
-
-static void amdgpu_devcoredump_free(void *data)
+int amdgpu_device_reinit_after_reset(struct amdgpu_reset_context *reset_context)
 {
-}
-
-static void amdgpu_reset_capture_coredumpm(struct amdgpu_device *adev)
-{
-	struct drm_device *dev = adev_to_drm(adev);
-
-	ktime_get_ts64(&adev->reset_time);
-	dev_coredumpm(dev->dev, THIS_MODULE, adev, 0, GFP_KERNEL,
-		      amdgpu_devcoredump_read, amdgpu_devcoredump_free);
-}
-#endif
-
-int amdgpu_do_asic_reset(struct list_head *device_list_handle,
-			 struct amdgpu_reset_context *reset_context)
-{
-	struct amdgpu_device *tmp_adev = NULL;
-	bool need_full_reset, skip_hw_reset, vram_lost = false;
-	int r = 0;
-	bool gpu_reset_for_dev_remove = 0;
-
-	/* Try reset handler method first */
-	tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device,
-				    reset_list);
-	amdgpu_reset_reg_dumps(tmp_adev);
-
-	reset_context->reset_device_list = device_list_handle;
-	r = amdgpu_reset_perform_reset(tmp_adev, reset_context);
-	/* If reset handler not implemented, continue; otherwise return */
-	if (r == -ENOSYS)
-		r = 0;
-	else
-		return r;
-
-	/* Reset handler not implemented, use the default method */
-	need_full_reset =
-		test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
-	skip_hw_reset = test_bit(AMDGPU_SKIP_HW_RESET, &reset_context->flags);
-
-	gpu_reset_for_dev_remove =
-		test_bit(AMDGPU_RESET_FOR_DEVICE_REMOVE, &reset_context->flags) &&
-			test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
+	struct list_head *device_list_handle;
+	bool full_reset, vram_lost = false;
+	struct amdgpu_device *tmp_adev;
+	int r, init_level;
 
-	/*
-	 * ASIC reset has to be done on all XGMI hive nodes ASAP
-	 * to allow proper links negotiation in FW (within 1 sec)
-	 */
-	if (!skip_hw_reset && need_full_reset) {
-		list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
-			/* For XGMI run all resets in parallel to speed up the process */
-			if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) {
-				tmp_adev->gmc.xgmi.pending_reset = false;
-				if (!queue_work(system_unbound_wq, &tmp_adev->xgmi_reset_work))
-					r = -EALREADY;
-			} else
-				r = amdgpu_asic_reset(tmp_adev);
-
-			if (r) {
-				dev_err(tmp_adev->dev, "ASIC reset failed with error, %d for drm dev, %s",
-					 r, adev_to_drm(tmp_adev)->unique);
-				break;
-			}
-		}
-
-		/* For XGMI wait for all resets to complete before proceed */
-		if (!r) {
-			list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
-				if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) {
-					flush_work(&tmp_adev->xgmi_reset_work);
-					r = tmp_adev->asic_reset_res;
-					if (r)
-						break;
-				}
-			}
-		}
-	}
+	device_list_handle = reset_context->reset_device_list;
 
-	if (!r && amdgpu_ras_intr_triggered()) {
-		list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
-			if (tmp_adev->mmhub.ras && tmp_adev->mmhub.ras->ras_block.hw_ops &&
-			    tmp_adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count)
-				tmp_adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count(tmp_adev);
-		}
+	if (!device_list_handle)
+		return -EINVAL;
 
-		amdgpu_ras_intr_cleared();
-	}
+	full_reset = test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
 
-	/* Since the mode1 reset affects base ip blocks, the
-	 * phase1 ip blocks need to be resumed. Otherwise there
-	 * will be a BIOS signature error and the psp bootloader
-	 * can't load kdb on the next amdgpu install.
+	/**
+	 * If it's reset on init, it's default init level, otherwise keep level
+	 * as recovery level.
 	 */
-	if (gpu_reset_for_dev_remove) {
-		list_for_each_entry(tmp_adev, device_list_handle, reset_list)
-			amdgpu_device_ip_resume_phase1(tmp_adev);
-
-		goto end;
-	}
+	if (reset_context->method == AMD_RESET_METHOD_ON_INIT)
+			init_level = AMDGPU_INIT_LEVEL_DEFAULT;
+	else
+			init_level = AMDGPU_INIT_LEVEL_RESET_RECOVERY;
 
+	r = 0;
 	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
-		if (need_full_reset) {
+		amdgpu_set_init_level(tmp_adev, init_level);
+		if (full_reset) {
 			/* post card */
+			amdgpu_reset_set_dpc_status(tmp_adev, false);
+			amdgpu_ras_clear_err_state(tmp_adev);
 			r = amdgpu_device_asic_init(tmp_adev);
 			if (r) {
 				dev_warn(tmp_adev->dev, "asic atom init failed!");
 			} else {
 				dev_info(tmp_adev->dev, "GPU reset succeeded, trying to resume\n");
-				r = amdgpu_amdkfd_resume_iommu(tmp_adev);
-				if (r)
-					goto out;
 
 				r = amdgpu_device_ip_resume_phase1(tmp_adev);
 				if (r)
 					goto out;
 
 				vram_lost = amdgpu_device_check_vram_lost(tmp_adev);
-#ifdef CONFIG_DEV_COREDUMP
-				tmp_adev->reset_vram_lost = vram_lost;
-				memset(&tmp_adev->reset_task_info, 0,
-						sizeof(tmp_adev->reset_task_info));
-				if (reset_context->job && reset_context->job->vm)
-					tmp_adev->reset_task_info =
-						reset_context->job->vm->task_info;
-				amdgpu_reset_capture_coredumpm(tmp_adev);
-#endif
+
+				if (!test_bit(AMDGPU_SKIP_COREDUMP, &reset_context->flags))
+					amdgpu_coredump(tmp_adev, false, vram_lost, reset_context->job);
+
 				if (vram_lost) {
-					DRM_INFO("VRAM is lost due to GPU reset!\n");
+					dev_info(
+						tmp_adev->dev,
+						"VRAM is lost due to GPU reset!\n");
 					amdgpu_inc_vram_lost(tmp_adev);
 				}
 
@@ -4994,10 +6054,22 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 				if (r)
 					return r;
 
+				r = amdgpu_xcp_restore_partition_mode(
+					tmp_adev->xcp_mgr);
+				if (r)
+					goto out;
+
 				r = amdgpu_device_ip_resume_phase2(tmp_adev);
 				if (r)
 					goto out;
 
+				if (tmp_adev->mman.buffer_funcs_ring->sched.ready)
+					amdgpu_ttm_set_buffer_funcs_status(tmp_adev, true);
+
+				r = amdgpu_device_ip_resume_phase3(tmp_adev);
+				if (r)
+					goto out;
+
 				if (vram_lost)
 					amdgpu_device_fill_reset_magic(tmp_adev);
 
@@ -5015,7 +6087,11 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 				if (r)
 					goto out;
 
-				drm_fb_helper_set_suspend_unlocked(adev_to_drm(tmp_adev)->fb_helper, false);
+				r = amdgpu_userq_post_reset(tmp_adev, vram_lost);
+				if (r)
+					goto out;
+
+				drm_client_dev_resume(adev_to_drm(tmp_adev));
 
 				/*
 				 * The GPU enters bad state once faulty pages
@@ -5027,7 +6103,7 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 				 * bad_page_threshold value to fix this once
 				 * probing driver again.
 				 */
-				if (!amdgpu_ras_eeprom_check_err_threshold(tmp_adev)) {
+				if (!amdgpu_ras_is_rma(tmp_adev)) {
 					/* must succeed. */
 					amdgpu_ras_resume(tmp_adev);
 				} else {
@@ -5045,27 +6121,102 @@ int amdgpu_do_asic_reset(struct list_head *device_list_handle,
 
 out:
 		if (!r) {
+			/* IP init is complete now, set level as default */
+			amdgpu_set_init_level(tmp_adev,
+					      AMDGPU_INIT_LEVEL_DEFAULT);
 			amdgpu_irq_gpu_reset_resume_helper(tmp_adev);
 			r = amdgpu_ib_ring_tests(tmp_adev);
 			if (r) {
 				dev_err(tmp_adev->dev, "ib ring test failed (%d).\n", r);
-				need_full_reset = true;
 				r = -EAGAIN;
 				goto end;
 			}
 		}
 
-		if (!r)
-			r = amdgpu_device_recover_vram(tmp_adev);
-		else
+		if (r)
 			tmp_adev->asic_reset_res = r;
 	}
 
 end:
-	if (need_full_reset)
+	return r;
+}
+
+int amdgpu_do_asic_reset(struct list_head *device_list_handle,
+			 struct amdgpu_reset_context *reset_context)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+	bool need_full_reset, skip_hw_reset;
+	int r = 0;
+
+	/* Try reset handler method first */
+	tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device,
+				    reset_list);
+
+	reset_context->reset_device_list = device_list_handle;
+	r = amdgpu_reset_perform_reset(tmp_adev, reset_context);
+	/* If reset handler not implemented, continue; otherwise return */
+	if (r == -EOPNOTSUPP)
+		r = 0;
+	else
+		return r;
+
+	/* Reset handler not implemented, use the default method */
+	need_full_reset =
+		test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
+	skip_hw_reset = test_bit(AMDGPU_SKIP_HW_RESET, &reset_context->flags);
+
+	/*
+	 * ASIC reset has to be done on all XGMI hive nodes ASAP
+	 * to allow proper links negotiation in FW (within 1 sec)
+	 */
+	if (!skip_hw_reset && need_full_reset) {
+		list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
+			/* For XGMI run all resets in parallel to speed up the process */
+			if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) {
+				if (!queue_work(system_unbound_wq,
+						&tmp_adev->xgmi_reset_work))
+					r = -EALREADY;
+			} else
+				r = amdgpu_asic_reset(tmp_adev);
+
+			if (r) {
+				dev_err(tmp_adev->dev,
+					"ASIC reset failed with error, %d for drm dev, %s",
+					r, adev_to_drm(tmp_adev)->unique);
+				goto out;
+			}
+		}
+
+		/* For XGMI wait for all resets to complete before proceed */
+		if (!r) {
+			list_for_each_entry(tmp_adev, device_list_handle,
+					    reset_list) {
+				if (tmp_adev->gmc.xgmi.num_physical_nodes > 1) {
+					flush_work(&tmp_adev->xgmi_reset_work);
+					r = tmp_adev->asic_reset_res;
+					if (r)
+						break;
+				}
+			}
+		}
+	}
+
+	if (!r && amdgpu_ras_intr_triggered()) {
+		list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
+			amdgpu_ras_reset_error_count(tmp_adev,
+						     AMDGPU_RAS_BLOCK__MMHUB);
+		}
+
+		amdgpu_ras_intr_cleared();
+	}
+
+	r = amdgpu_device_reinit_after_reset(reset_context);
+	if (r == -EAGAIN)
 		set_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
 	else
 		clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
+
+out:
 	return r;
 }
 
@@ -5074,6 +6225,7 @@ static void amdgpu_device_set_mp1_state(struct amdgpu_device *adev)
 
 	switch (amdgpu_asic_reset_method(adev)) {
 	case AMD_RESET_METHOD_MODE1:
+	case AMD_RESET_METHOD_LINK:
 		adev->mp1_state = PP_MP1_STATE_SHUTDOWN;
 		break;
 	case AMD_RESET_METHOD_MODE2:
@@ -5161,6 +6313,7 @@ static inline void amdgpu_device_stop_pending_resets(struct amdgpu_device *adev)
 	if (!amdgpu_sriov_vf(adev))
 		cancel_work(&adev->reset_work);
 #endif
+	cancel_work(&adev->userq_reset_work);
 
 	if (adev->kfd.dev)
 		cancel_work(&adev->kfd.reset_work);
@@ -5173,89 +6326,80 @@ static inline void amdgpu_device_stop_pending_resets(struct amdgpu_device *adev)
 
 }
 
-/**
- * amdgpu_device_gpu_recover - reset the asic and recover scheduler
- *
- * @adev: amdgpu_device pointer
- * @job: which job trigger hang
- * @reset_context: amdgpu reset context pointer
- *
- * Attempt to reset the GPU if it has hung (all asics).
- * Attempt to do soft-reset or full-reset and reinitialize Asic
- * Returns 0 for success or an error on failure.
- */
-
-int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
-			      struct amdgpu_job *job,
-			      struct amdgpu_reset_context *reset_context)
+static int amdgpu_device_health_check(struct list_head *device_list_handle)
 {
-	struct list_head device_list, *device_list_handle =  NULL;
-	bool job_signaled = false;
-	struct amdgpu_hive_info *hive = NULL;
-	struct amdgpu_device *tmp_adev = NULL;
-	int i, r = 0;
-	bool need_emergency_restart = false;
-	bool audio_suspended = false;
-	bool gpu_reset_for_dev_remove = false;
-
-	gpu_reset_for_dev_remove =
-			test_bit(AMDGPU_RESET_FOR_DEVICE_REMOVE, &reset_context->flags) &&
-				test_bit(AMDGPU_NEED_FULL_RESET, &reset_context->flags);
-
-	/*
-	 * Special case: RAS triggered and full reset isn't supported
-	 */
-	need_emergency_restart = amdgpu_ras_need_emergency_restart(adev);
-
-	/*
-	 * Flush RAM to disk so that after reboot
-	 * the user can read log and see why the system rebooted.
-	 */
-	if (need_emergency_restart && amdgpu_ras_get_context(adev)->reboot) {
-		DRM_WARN("Emergency reboot.");
+	struct amdgpu_device *tmp_adev;
+	int ret = 0;
 
-		ksys_sync_helper();
-		emergency_restart();
+	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
+		ret |= amdgpu_device_bus_status_check(tmp_adev);
 	}
 
-	dev_info(adev->dev, "GPU %s begin!\n",
-		need_emergency_restart ? "jobs stop":"reset");
+	return ret;
+}
 
-	if (!amdgpu_sriov_vf(adev))
-		hive = amdgpu_get_xgmi_hive(adev);
-	if (hive)
-		mutex_lock(&hive->hive_lock);
+static void amdgpu_device_recovery_prepare(struct amdgpu_device *adev,
+					  struct list_head *device_list,
+					  struct amdgpu_hive_info *hive)
+{
+	struct amdgpu_device *tmp_adev = NULL;
 
-	reset_context->job = job;
-	reset_context->hive = hive;
 	/*
 	 * Build list of devices to reset.
 	 * In case we are in XGMI hive mode, resort the device list
 	 * to put adev in the 1st position.
 	 */
-	INIT_LIST_HEAD(&device_list);
-	if (!amdgpu_sriov_vf(adev) && (adev->gmc.xgmi.num_physical_nodes > 1)) {
+	if (!amdgpu_sriov_vf(adev) && (adev->gmc.xgmi.num_physical_nodes > 1) && hive) {
 		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
-			list_add_tail(&tmp_adev->reset_list, &device_list);
-			if (gpu_reset_for_dev_remove && adev->shutdown)
+			list_add_tail(&tmp_adev->reset_list, device_list);
+			if (adev->shutdown)
 				tmp_adev->shutdown = true;
+			if (amdgpu_reset_in_dpc(adev))
+				tmp_adev->pcie_reset_ctx.in_link_reset = true;
 		}
-		if (!list_is_first(&adev->reset_list, &device_list))
-			list_rotate_to_front(&adev->reset_list, &device_list);
-		device_list_handle = &device_list;
+		if (!list_is_first(&adev->reset_list, device_list))
+			list_rotate_to_front(&adev->reset_list, device_list);
 	} else {
-		list_add_tail(&adev->reset_list, &device_list);
-		device_list_handle = &device_list;
+		list_add_tail(&adev->reset_list, device_list);
 	}
+}
 
-	/* We need to lock reset domain only once both for XGMI and single device */
-	tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device,
-				    reset_list);
+static void amdgpu_device_recovery_get_reset_lock(struct amdgpu_device *adev,
+						  struct list_head *device_list)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+
+	if (list_empty(device_list))
+		return;
+	tmp_adev =
+		list_first_entry(device_list, struct amdgpu_device, reset_list);
 	amdgpu_device_lock_reset_domain(tmp_adev->reset_domain);
+}
 
-	/* block all schedulers and reset given job's ring */
-	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
+static void amdgpu_device_recovery_put_reset_lock(struct amdgpu_device *adev,
+						  struct list_head *device_list)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+
+	if (list_empty(device_list))
+		return;
+	tmp_adev =
+		list_first_entry(device_list, struct amdgpu_device, reset_list);
+	amdgpu_device_unlock_reset_domain(tmp_adev->reset_domain);
+}
+
+static void amdgpu_device_halt_activities(struct amdgpu_device *adev,
+					  struct amdgpu_job *job,
+					  struct amdgpu_reset_context *reset_context,
+					  struct list_head *device_list,
+					  struct amdgpu_hive_info *hive,
+					  bool need_emergency_restart)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+	int i;
 
+	/* block all schedulers and reset given job's ring */
+	list_for_each_entry(tmp_adev, device_list, reset_list) {
 		amdgpu_device_set_mp1_state(tmp_adev);
 
 		/*
@@ -5269,32 +6413,33 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 		 * some audio codec errors.
 		 */
 		if (!amdgpu_device_suspend_display_audio(tmp_adev))
-			audio_suspended = true;
+			tmp_adev->pcie_reset_ctx.audio_suspended = true;
 
 		amdgpu_ras_set_error_query_ready(tmp_adev, false);
 
 		cancel_delayed_work_sync(&tmp_adev->delayed_init_work);
 
-		if (!amdgpu_sriov_vf(tmp_adev))
-			amdgpu_amdkfd_pre_reset(tmp_adev);
+		amdgpu_amdkfd_pre_reset(tmp_adev, reset_context);
 
 		/*
-		 * Mark these ASICs to be reseted as untracked first
+		 * Mark these ASICs to be reset as untracked first
 		 * And add them back after reset completed
 		 */
 		amdgpu_unregister_gpu_instance(tmp_adev);
 
-		drm_fb_helper_set_suspend_unlocked(adev_to_drm(tmp_adev)->fb_helper, true);
+		drm_client_dev_suspend(adev_to_drm(tmp_adev));
 
 		/* disable ras on ALL IPs */
-		if (!need_emergency_restart &&
-		      amdgpu_device_ip_need_full_reset(tmp_adev))
+		if (!need_emergency_restart && !amdgpu_reset_in_dpc(adev) &&
+		    amdgpu_device_ip_need_full_reset(tmp_adev))
 			amdgpu_ras_suspend(tmp_adev);
 
+		amdgpu_userq_pre_reset(tmp_adev);
+
 		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 			struct amdgpu_ring *ring = tmp_adev->rings[i];
 
-			if (!ring || !ring->sched.thread)
+			if (!amdgpu_ring_sched_ready(ring))
 				continue;
 
 			drm_sched_stop(&ring->sched, job ? &job->base : NULL);
@@ -5304,28 +6449,18 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 		}
 		atomic_inc(&tmp_adev->gpu_reset_counter);
 	}
+}
 
-	if (need_emergency_restart)
-		goto skip_sched_resume;
-
-	/*
-	 * Must check guilty signal here since after this point all old
-	 * HW fences are force signaled.
-	 *
-	 * job->base holds a reference to parent fence
-	 */
-	if (job && dma_fence_is_signaled(&job->hw_fence)) {
-		job_signaled = true;
-		dev_info(adev->dev, "Guilty job already signaled, skipping HW reset");
-		goto skip_hw_reset;
-	}
+static int amdgpu_device_asic_reset(struct amdgpu_device *adev,
+			      struct list_head *device_list,
+			      struct amdgpu_reset_context *reset_context)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+	int retry_limit = AMDGPU_MAX_RETRY_LIMIT;
+	int r = 0;
 
 retry:	/* Rest of adevs pre asic reset from XGMI hive. */
-	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
-		if (gpu_reset_for_dev_remove) {
-			/* Workaroud for ASICs need to disable SMC first */
-			amdgpu_device_smu_fini_early(tmp_adev);
-		}
+	list_for_each_entry(tmp_adev, device_list, reset_list) {
 		r = amdgpu_device_pre_asic_reset(tmp_adev, reset_context);
 		/*TODO Should we stop ?*/
 		if (r) {
@@ -5333,73 +6468,109 @@ retry:	/* Rest of adevs pre asic reset from XGMI hive. */
 				  r, adev_to_drm(tmp_adev)->unique);
 			tmp_adev->asic_reset_res = r;
 		}
-
-		/*
-		 * Drop all pending non scheduler resets. Scheduler resets
-		 * were already dropped during drm_sched_stop
-		 */
-		amdgpu_device_stop_pending_resets(tmp_adev);
 	}
 
 	/* Actual ASIC resets if needed.*/
 	/* Host driver will handle XGMI hive reset for SRIOV */
 	if (amdgpu_sriov_vf(adev)) {
-		r = amdgpu_device_reset_sriov(adev, job ? false : true);
+
+		/* Bail out of reset early */
+		if (amdgpu_ras_is_rma(adev))
+			return -ENODEV;
+
+		if (amdgpu_ras_get_fed_status(adev) || amdgpu_virt_rcvd_ras_interrupt(adev)) {
+			dev_dbg(adev->dev, "Detected RAS error, wait for FLR completion\n");
+			amdgpu_ras_set_fed(adev, true);
+			set_bit(AMDGPU_HOST_FLR, &reset_context->flags);
+		}
+
+		r = amdgpu_device_reset_sriov(adev, reset_context);
+		if (AMDGPU_RETRY_SRIOV_RESET(r) && (retry_limit--) > 0) {
+			amdgpu_virt_release_full_gpu(adev, true);
+			goto retry;
+		}
 		if (r)
 			adev->asic_reset_res = r;
-
-		/* Aldebaran and gfx_11_0_3 support ras in SRIOV, so need resume ras during reset */
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) ||
-		    adev->ip_versions[GC_HWIP][0] == IP_VERSION(11, 0, 3))
-			amdgpu_ras_resume(adev);
 	} else {
-		r = amdgpu_do_asic_reset(device_list_handle, reset_context);
+		r = amdgpu_do_asic_reset(device_list, reset_context);
 		if (r && r == -EAGAIN)
 			goto retry;
+	}
 
-		if (!r && gpu_reset_for_dev_remove)
-			goto recover_end;
+	list_for_each_entry(tmp_adev, device_list, reset_list) {
+		/*
+		 * Drop any pending non scheduler resets queued before reset is done.
+		 * Any reset scheduled after this point would be valid. Scheduler resets
+		 * were already dropped during drm_sched_stop and no new ones can come
+		 * in before drm_sched_start.
+		 */
+		amdgpu_device_stop_pending_resets(tmp_adev);
 	}
 
-skip_hw_reset:
+	return r;
+}
+
+static int amdgpu_device_sched_resume(struct list_head *device_list,
+			      struct amdgpu_reset_context *reset_context,
+			      bool   job_signaled)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+	int i, r = 0;
 
 	/* Post ASIC reset for all devs .*/
-	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
+	list_for_each_entry(tmp_adev, device_list, reset_list) {
 
 		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 			struct amdgpu_ring *ring = tmp_adev->rings[i];
 
-			if (!ring || !ring->sched.thread)
+			if (!amdgpu_ring_sched_ready(ring))
 				continue;
 
-			drm_sched_start(&ring->sched, true);
+			drm_sched_start(&ring->sched, 0);
 		}
 
-		if (adev->enable_mes && adev->ip_versions[GC_HWIP][0] != IP_VERSION(11, 0, 3))
-			amdgpu_mes_self_test(tmp_adev);
-
-		if (!drm_drv_uses_atomic_modeset(adev_to_drm(tmp_adev)) && !job_signaled) {
+		if (!drm_drv_uses_atomic_modeset(adev_to_drm(tmp_adev)) && !job_signaled)
 			drm_helper_resume_force_mode(adev_to_drm(tmp_adev));
-		}
 
-		if (tmp_adev->asic_reset_res)
-			r = tmp_adev->asic_reset_res;
-
-		tmp_adev->asic_reset_res = 0;
-
-		if (r) {
-			/* bad news, how to tell it to userspace ? */
-			dev_info(tmp_adev->dev, "GPU reset(%d) failed\n", atomic_read(&tmp_adev->gpu_reset_counter));
-			amdgpu_vf_error_put(tmp_adev, AMDGIM_ERROR_VF_GPU_RESET_FAIL, 0, r);
+		if (tmp_adev->asic_reset_res) {
+			/* bad news, how to tell it to userspace ?
+			 * for ras error, we should report GPU bad status instead of
+			 * reset failure
+			 */
+			if (reset_context->src != AMDGPU_RESET_SRC_RAS ||
+			    !amdgpu_ras_eeprom_check_err_threshold(tmp_adev))
+				dev_info(
+					tmp_adev->dev,
+					"GPU reset(%d) failed with error %d \n",
+					atomic_read(
+						&tmp_adev->gpu_reset_counter),
+					tmp_adev->asic_reset_res);
+			amdgpu_vf_error_put(tmp_adev,
+					    AMDGIM_ERROR_VF_GPU_RESET_FAIL, 0,
+					    tmp_adev->asic_reset_res);
+			if (!r)
+				r = tmp_adev->asic_reset_res;
+			tmp_adev->asic_reset_res = 0;
 		} else {
-			dev_info(tmp_adev->dev, "GPU reset(%d) succeeded!\n", atomic_read(&tmp_adev->gpu_reset_counter));
-			if (amdgpu_acpi_smart_shift_update(adev_to_drm(tmp_adev), AMDGPU_SS_DEV_D0))
-				DRM_WARN("smart shift update failed\n");
+			dev_info(tmp_adev->dev, "GPU reset(%d) succeeded!\n",
+				 atomic_read(&tmp_adev->gpu_reset_counter));
+			if (amdgpu_acpi_smart_shift_update(tmp_adev,
+							   AMDGPU_SS_DEV_D0))
+				dev_warn(tmp_adev->dev,
+					 "smart shift update failed\n");
 		}
 	}
 
-skip_sched_resume:
-	list_for_each_entry(tmp_adev, device_list_handle, reset_list) {
+	return r;
+}
+
+static void amdgpu_device_gpu_resume(struct amdgpu_device *adev,
+			      struct list_head *device_list,
+			      bool   need_emergency_restart)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+
+	list_for_each_entry(tmp_adev, device_list, reset_list) {
 		/* unlock kfd: SRIOV would do it separately */
 		if (!need_emergency_restart && !amdgpu_sriov_vf(tmp_adev))
 			amdgpu_amdkfd_post_reset(tmp_adev);
@@ -5410,19 +6581,125 @@ skip_sched_resume:
 		if (!adev->kfd.init_complete)
 			amdgpu_amdkfd_device_init(adev);
 
-		if (audio_suspended)
+		if (tmp_adev->pcie_reset_ctx.audio_suspended)
 			amdgpu_device_resume_display_audio(tmp_adev);
 
 		amdgpu_device_unset_mp1_state(tmp_adev);
 
 		amdgpu_ras_set_error_query_ready(tmp_adev, true);
+
 	}
+}
 
-recover_end:
-	tmp_adev = list_first_entry(device_list_handle, struct amdgpu_device,
-					    reset_list);
-	amdgpu_device_unlock_reset_domain(tmp_adev->reset_domain);
 
+/**
+ * amdgpu_device_gpu_recover - reset the asic and recover scheduler
+ *
+ * @adev: amdgpu_device pointer
+ * @job: which job trigger hang
+ * @reset_context: amdgpu reset context pointer
+ *
+ * Attempt to reset the GPU if it has hung (all asics).
+ * Attempt to do soft-reset or full-reset and reinitialize Asic
+ * Returns 0 for success or an error on failure.
+ */
+
+int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
+			      struct amdgpu_job *job,
+			      struct amdgpu_reset_context *reset_context)
+{
+	struct list_head device_list;
+	bool job_signaled = false;
+	struct amdgpu_hive_info *hive = NULL;
+	int r = 0;
+	bool need_emergency_restart = false;
+
+	/*
+	 * If it reaches here because of hang/timeout and a RAS error is
+	 * detected at the same time, let RAS recovery take care of it.
+	 */
+	if (amdgpu_ras_is_err_state(adev, AMDGPU_RAS_BLOCK__ANY) &&
+	    !amdgpu_sriov_vf(adev) &&
+	    reset_context->src != AMDGPU_RESET_SRC_RAS) {
+		dev_dbg(adev->dev,
+			"Gpu recovery from source: %d yielding to RAS error recovery handling",
+			reset_context->src);
+		return 0;
+	}
+
+	/*
+	 * Special case: RAS triggered and full reset isn't supported
+	 */
+	need_emergency_restart = amdgpu_ras_need_emergency_restart(adev);
+
+	/*
+	 * Flush RAM to disk so that after reboot
+	 * the user can read log and see why the system rebooted.
+	 */
+	if (need_emergency_restart && amdgpu_ras_get_context(adev) &&
+		amdgpu_ras_get_context(adev)->reboot) {
+		dev_warn(adev->dev, "Emergency reboot.");
+
+		ksys_sync_helper();
+		emergency_restart();
+	}
+
+	dev_info(adev->dev, "GPU %s begin!. Source:  %d\n",
+		 need_emergency_restart ? "jobs stop" : "reset",
+		 reset_context->src);
+
+	if (!amdgpu_sriov_vf(adev))
+		hive = amdgpu_get_xgmi_hive(adev);
+	if (hive)
+		mutex_lock(&hive->hive_lock);
+
+	reset_context->job = job;
+	reset_context->hive = hive;
+	INIT_LIST_HEAD(&device_list);
+
+	amdgpu_device_recovery_prepare(adev, &device_list, hive);
+
+	if (!amdgpu_sriov_vf(adev)) {
+		r = amdgpu_device_health_check(&device_list);
+		if (r)
+			goto end_reset;
+	}
+
+	/* Cannot be called after locking reset domain */
+	amdgpu_ras_pre_reset(adev, &device_list);
+
+	/* We need to lock reset domain only once both for XGMI and single device */
+	amdgpu_device_recovery_get_reset_lock(adev, &device_list);
+
+	amdgpu_device_halt_activities(adev, job, reset_context, &device_list,
+				      hive, need_emergency_restart);
+	if (need_emergency_restart)
+		goto skip_sched_resume;
+	/*
+	 * Must check guilty signal here since after this point all old
+	 * HW fences are force signaled.
+	 *
+	 * job->base holds a reference to parent fence
+	 */
+	if (job && dma_fence_is_signaled(&job->hw_fence->base)) {
+		job_signaled = true;
+		dev_info(adev->dev, "Guilty job already signaled, skipping HW reset");
+		goto skip_hw_reset;
+	}
+
+	r = amdgpu_device_asic_reset(adev, &device_list, reset_context);
+	if (r)
+		goto reset_unlock;
+skip_hw_reset:
+	r = amdgpu_device_sched_resume(&device_list, reset_context, job_signaled);
+	if (r)
+		goto reset_unlock;
+skip_sched_resume:
+	amdgpu_device_gpu_resume(adev, &device_list, need_emergency_restart);
+reset_unlock:
+	amdgpu_device_recovery_put_reset_lock(adev, &device_list);
+	amdgpu_ras_post_reset(adev, &device_list);
+end_reset:
 	if (hive) {
 		mutex_unlock(&hive->hive_lock);
 		amdgpu_put_xgmi_hive(hive);
@@ -5432,23 +6709,111 @@ recover_end:
 		dev_info(adev->dev, "GPU reset end with ret = %d\n", r);
 
 	atomic_set(&adev->reset_domain->reset_res, r);
+
+	if (!r) {
+		struct amdgpu_task_info *ti = NULL;
+
+		if (job)
+			ti = amdgpu_vm_get_task_info_pasid(adev, job->pasid);
+
+		drm_dev_wedged_event(adev_to_drm(adev), DRM_WEDGE_RECOVERY_NONE,
+				     ti ? &ti->task : NULL);
+
+		amdgpu_vm_put_task_info(ti);
+	}
+
 	return r;
 }
 
 /**
+ * amdgpu_device_partner_bandwidth - find the bandwidth of appropriate partner
+ *
+ * @adev: amdgpu_device pointer
+ * @speed: pointer to the speed of the link
+ * @width: pointer to the width of the link
+ *
+ * Evaluate the hierarchy to find the speed and bandwidth capabilities of the
+ * first physical partner to an AMD dGPU.
+ * This will exclude any virtual switches and links.
+ */
+static void amdgpu_device_partner_bandwidth(struct amdgpu_device *adev,
+					    enum pci_bus_speed *speed,
+					    enum pcie_link_width *width)
+{
+	struct pci_dev *parent = adev->pdev;
+
+	if (!speed || !width)
+		return;
+
+	*speed = PCI_SPEED_UNKNOWN;
+	*width = PCIE_LNK_WIDTH_UNKNOWN;
+
+	if (amdgpu_device_pcie_dynamic_switching_supported(adev)) {
+		while ((parent = pci_upstream_bridge(parent))) {
+			/* skip upstream/downstream switches internal to dGPU*/
+			if (parent->vendor == PCI_VENDOR_ID_ATI)
+				continue;
+			*speed = pcie_get_speed_cap(parent);
+			*width = pcie_get_width_cap(parent);
+			break;
+		}
+	} else {
+		/* use the current speeds rather than max if switching is not supported */
+		pcie_bandwidth_available(adev->pdev, NULL, speed, width);
+	}
+}
+
+/**
+ * amdgpu_device_gpu_bandwidth - find the bandwidth of the GPU
+ *
+ * @adev: amdgpu_device pointer
+ * @speed: pointer to the speed of the link
+ * @width: pointer to the width of the link
+ *
+ * Evaluate the hierarchy to find the speed and bandwidth capabilities of the
+ * AMD dGPU which may be a virtual upstream bridge.
+ */
+static void amdgpu_device_gpu_bandwidth(struct amdgpu_device *adev,
+					enum pci_bus_speed *speed,
+					enum pcie_link_width *width)
+{
+	struct pci_dev *parent = adev->pdev;
+
+	if (!speed || !width)
+		return;
+
+	parent = pci_upstream_bridge(parent);
+	if (parent && parent->vendor == PCI_VENDOR_ID_ATI) {
+		/* use the upstream/downstream switches internal to dGPU */
+		*speed = pcie_get_speed_cap(parent);
+		*width = pcie_get_width_cap(parent);
+		while ((parent = pci_upstream_bridge(parent))) {
+			if (parent->vendor == PCI_VENDOR_ID_ATI) {
+				/* use the upstream/downstream switches internal to dGPU */
+				*speed = pcie_get_speed_cap(parent);
+				*width = pcie_get_width_cap(parent);
+			}
+		}
+	} else {
+		/* use the device itself */
+		*speed = pcie_get_speed_cap(adev->pdev);
+		*width = pcie_get_width_cap(adev->pdev);
+	}
+}
+
+/**
  * amdgpu_device_get_pcie_info - fence pcie info about the PCIE slot
  *
  * @adev: amdgpu_device pointer
  *
- * Fetchs and stores in the driver the PCIE capabilities (gen speed
+ * Fetches and stores in the driver the PCIE capabilities (gen speed
  * and lanes) of the slot the device is in. Handles APUs and
  * virtualized environments where PCIE config space may not be available.
  */
 static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev)
 {
-	struct pci_dev *pdev;
 	enum pci_bus_speed speed_cap, platform_speed_cap;
-	enum pcie_link_width platform_link_width;
+	enum pcie_link_width platform_link_width, link_width;
 
 	if (amdgpu_pcie_gen_cap)
 		adev->pm.pcie_gen_mask = amdgpu_pcie_gen_cap;
@@ -5457,7 +6822,7 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev)
 		adev->pm.pcie_mlw_mask = amdgpu_pcie_lane_cap;
 
 	/* covers APUs as well */
-	if (pci_is_root_bus(adev->pdev->bus)) {
+	if (pci_is_root_bus(adev->pdev->bus) && !amdgpu_passthrough(adev)) {
 		if (adev->pm.pcie_gen_mask == 0)
 			adev->pm.pcie_gen_mask = AMDGPU_DEFAULT_PCIE_GEN_MASK;
 		if (adev->pm.pcie_mlw_mask == 0)
@@ -5468,13 +6833,12 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev)
 	if (adev->pm.pcie_gen_mask && adev->pm.pcie_mlw_mask)
 		return;
 
-	pcie_bandwidth_available(adev->pdev, NULL,
-				 &platform_speed_cap, &platform_link_width);
+	amdgpu_device_partner_bandwidth(adev, &platform_speed_cap,
+					&platform_link_width);
+	amdgpu_device_gpu_bandwidth(adev, &speed_cap, &link_width);
 
 	if (adev->pm.pcie_gen_mask == 0) {
 		/* asic caps */
-		pdev = adev->pdev;
-		speed_cap = pcie_get_speed_cap(pdev);
 		if (speed_cap == PCI_SPEED_UNKNOWN) {
 			adev->pm.pcie_gen_mask |= (CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN1 |
 						  CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN2 |
@@ -5530,51 +6894,103 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev)
 		}
 	}
 	if (adev->pm.pcie_mlw_mask == 0) {
+		/* asic caps */
+		if (link_width == PCIE_LNK_WIDTH_UNKNOWN) {
+			adev->pm.pcie_mlw_mask |= AMDGPU_DEFAULT_ASIC_PCIE_MLW_MASK;
+		} else {
+			switch (link_width) {
+			case PCIE_LNK_X32:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X32 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X16 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X16:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X16 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X12:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X8:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X4:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X2:
+				adev->pm.pcie_mlw_mask |= (CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1);
+				break;
+			case PCIE_LNK_X1:
+				adev->pm.pcie_mlw_mask |= CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X1;
+				break;
+			default:
+				break;
+			}
+		}
+		/* platform caps */
 		if (platform_link_width == PCIE_LNK_WIDTH_UNKNOWN) {
 			adev->pm.pcie_mlw_mask |= AMDGPU_DEFAULT_PCIE_MLW_MASK;
 		} else {
 			switch (platform_link_width) {
 			case PCIE_LNK_X32:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X32 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X32 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X16:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X12:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X8:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X4:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X2:
-				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
-							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
+				adev->pm.pcie_mlw_mask |= (CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
+							   CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
 				break;
 			case PCIE_LNK_X1:
-				adev->pm.pcie_mlw_mask = CAIL_PCIE_LINK_WIDTH_SUPPORT_X1;
+				adev->pm.pcie_mlw_mask |= CAIL_PCIE_LINK_WIDTH_SUPPORT_X1;
 				break;
 			default:
 				break;
@@ -5597,29 +7013,37 @@ bool amdgpu_device_is_peer_accessible(struct amdgpu_device *adev,
 				      struct amdgpu_device *peer_adev)
 {
 #ifdef CONFIG_HSA_AMD_P2P
-	uint64_t address_mask = peer_adev->dev->dma_mask ?
-		~*peer_adev->dev->dma_mask : ~((1ULL << 32) - 1);
-	resource_size_t aper_limit =
-		adev->gmc.aper_base + adev->gmc.aper_size - 1;
 	bool p2p_access =
 		!adev->gmc.xgmi.connected_to_cpu &&
 		!(pci_p2pdma_distance(adev->pdev, peer_adev->dev, false) < 0);
+	if (!p2p_access)
+		dev_info(adev->dev, "PCIe P2P access from peer device %s is not supported by the chipset\n",
+			pci_name(peer_adev->pdev));
 
-	return pcie_p2p && p2p_access && (adev->gmc.visible_vram_size &&
-		adev->gmc.real_vram_size == adev->gmc.visible_vram_size &&
-		!(adev->gmc.aper_base & address_mask ||
-		  aper_limit & address_mask));
+	bool is_large_bar = adev->gmc.visible_vram_size &&
+		adev->gmc.real_vram_size == adev->gmc.visible_vram_size;
+	bool p2p_addressable = amdgpu_device_check_iommu_remap(peer_adev);
+
+	if (!p2p_addressable) {
+		uint64_t address_mask = peer_adev->dev->dma_mask ?
+			~*peer_adev->dev->dma_mask : ~((1ULL << 32) - 1);
+		resource_size_t aper_limit =
+			adev->gmc.aper_base + adev->gmc.aper_size - 1;
+
+		p2p_addressable = !(adev->gmc.aper_base & address_mask ||
+				     aper_limit & address_mask);
+	}
+	return pcie_p2p && is_large_bar && p2p_access && p2p_addressable;
 #else
 	return false;
 #endif
 }
 
-int amdgpu_device_baco_enter(struct drm_device *dev)
+int amdgpu_device_baco_enter(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 
-	if (!amdgpu_device_supports_baco(dev))
+	if (!amdgpu_device_supports_baco(adev))
 		return -ENOTSUPP;
 
 	if (ras && adev->ras_enabled &&
@@ -5629,13 +7053,12 @@ int amdgpu_device_baco_enter(struct drm_device *dev)
 	return amdgpu_dpm_baco_enter(adev);
 }
 
-int amdgpu_device_baco_exit(struct drm_device *dev)
+int amdgpu_device_baco_exit(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 	int ret = 0;
 
-	if (!amdgpu_device_supports_baco(dev))
+	if (!amdgpu_device_supports_baco(adev))
 		return -ENOTSUPP;
 
 	ret = amdgpu_dpm_baco_exit(adev);
@@ -5646,7 +7069,7 @@ int amdgpu_device_baco_exit(struct drm_device *dev)
 	    adev->nbio.funcs->enable_doorbell_interrupt)
 		adev->nbio.funcs->enable_doorbell_interrupt(adev, true);
 
-	if (amdgpu_passthrough(adev) &&
+	if (amdgpu_passthrough(adev) && adev->nbio.funcs &&
 	    adev->nbio.funcs->clear_doorbell_interrupt)
 		adev->nbio.funcs->clear_doorbell_interrupt(adev);
 
@@ -5666,45 +7089,52 @@ pci_ers_result_t amdgpu_pci_error_detected(struct pci_dev *pdev, pci_channel_sta
 {
 	struct drm_device *dev = pci_get_drvdata(pdev);
 	struct amdgpu_device *adev = drm_to_adev(dev);
-	int i;
+	struct amdgpu_hive_info *hive __free(xgmi_put_hive) =
+		amdgpu_get_xgmi_hive(adev);
+	struct amdgpu_reset_context reset_context;
+	struct list_head device_list;
 
-	DRM_INFO("PCI error: detected callback, state(%d)!!\n", state);
-
-	if (adev->gmc.xgmi.num_physical_nodes > 1) {
-		DRM_WARN("No support for XGMI hive yet...");
-		return PCI_ERS_RESULT_DISCONNECT;
-	}
+	dev_info(adev->dev, "PCI error: detected callback!!\n");
 
 	adev->pci_channel_state = state;
 
 	switch (state) {
 	case pci_channel_io_normal:
+		dev_info(adev->dev, "pci_channel_io_normal: state(%d)!!\n", state);
 		return PCI_ERS_RESULT_CAN_RECOVER;
-	/* Fatal error, prepare for slot reset */
 	case pci_channel_io_frozen:
-		/*
-		 * Locking adev->reset_domain->sem will prevent any external access
-		 * to GPU during PCI error recovery
-		 */
-		amdgpu_device_lock_reset_domain(adev->reset_domain);
-		amdgpu_device_set_mp1_state(adev);
-
-		/*
-		 * Block any work scheduling as we do for regular GPU reset
-		 * for the duration of the recovery
-		 */
-		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
-			struct amdgpu_ring *ring = adev->rings[i];
-
-			if (!ring || !ring->sched.thread)
-				continue;
+		/* Fatal error, prepare for slot reset */
+		dev_info(adev->dev, "pci_channel_io_frozen: state(%d)!!\n", state);
+		if (hive) {
+			/* Hive devices should be able to support FW based
+			 * link reset on other devices, if not return.
+			 */
+			if (!amdgpu_dpm_is_link_reset_supported(adev)) {
+				dev_warn(adev->dev,
+					 "No support for XGMI hive yet...\n");
+				return PCI_ERS_RESULT_DISCONNECT;
+			}
+			/* Set dpc status only if device is part of hive
+			 * Non-hive devices should be able to recover after
+			 * link reset.
+			 */
+			amdgpu_reset_set_dpc_status(adev, true);
 
-			drm_sched_stop(&ring->sched, NULL);
+			mutex_lock(&hive->hive_lock);
 		}
-		atomic_inc(&adev->gpu_reset_counter);
+		memset(&reset_context, 0, sizeof(reset_context));
+		INIT_LIST_HEAD(&device_list);
+
+		amdgpu_device_recovery_prepare(adev, &device_list, hive);
+		amdgpu_device_recovery_get_reset_lock(adev, &device_list);
+		amdgpu_device_halt_activities(adev, NULL, &reset_context, &device_list,
+					      hive, false);
+		if (hive)
+			mutex_unlock(&hive->hive_lock);
 		return PCI_ERS_RESULT_NEED_RESET;
 	case pci_channel_io_perm_failure:
 		/* Permanent error, prepare for device removal */
+		dev_info(adev->dev, "pci_channel_io_perm_failure: state(%d)!!\n", state);
 		return PCI_ERS_RESULT_DISCONNECT;
 	}
 
@@ -5717,8 +7147,10 @@ pci_ers_result_t amdgpu_pci_error_detected(struct pci_dev *pdev, pci_channel_sta
  */
 pci_ers_result_t amdgpu_pci_mmio_enabled(struct pci_dev *pdev)
 {
+	struct drm_device *dev = pci_get_drvdata(pdev);
+	struct amdgpu_device *adev = drm_to_adev(dev);
 
-	DRM_INFO("PCI error: mmio enabled callback!!\n");
+	dev_info(adev->dev, "PCI error: mmio enabled callback!!\n");
 
 	/* TODO - dump whatever for debugging purposes */
 
@@ -5742,21 +7174,38 @@ pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev)
 {
 	struct drm_device *dev = pci_get_drvdata(pdev);
 	struct amdgpu_device *adev = drm_to_adev(dev);
-	int r, i;
 	struct amdgpu_reset_context reset_context;
-	u32 memsize;
+	struct amdgpu_device *tmp_adev;
+	struct amdgpu_hive_info *hive;
 	struct list_head device_list;
+	struct pci_dev *link_dev;
+	int r = 0, i, timeout;
+	u32 memsize;
+	u16 status;
 
-	DRM_INFO("PCI error: slot reset callback!!\n");
+	dev_info(adev->dev, "PCI error: slot reset callback!!\n");
 
 	memset(&reset_context, 0, sizeof(reset_context));
 
-	INIT_LIST_HEAD(&device_list);
-	list_add_tail(&adev->reset_list, &device_list);
+	if (adev->pcie_reset_ctx.swus)
+		link_dev = adev->pcie_reset_ctx.swus;
+	else
+		link_dev = adev->pdev;
+	/* wait for asic to come out of reset, timeout = 10s */
+	timeout = 10000;
+	do {
+		usleep_range(10000, 10500);
+		r = pci_read_config_word(link_dev, PCI_VENDOR_ID, &status);
+		timeout -= 10;
+	} while (timeout > 0 && (status != PCI_VENDOR_ID_ATI) &&
+		 (status != PCI_VENDOR_ID_AMD));
 
-	/* wait for asic to come out of reset */
-	msleep(500);
+	if ((status != PCI_VENDOR_ID_ATI) && (status != PCI_VENDOR_ID_AMD)) {
+		r = -ETIME;
+		goto out;
+	}
 
+	amdgpu_device_load_switch_state(adev);
 	/* Restore PCI confspace */
 	amdgpu_device_load_pci_state(pdev);
 
@@ -5776,26 +7225,40 @@ pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev)
 	reset_context.method = AMD_RESET_METHOD_NONE;
 	reset_context.reset_req_dev = adev;
 	set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
-	set_bit(AMDGPU_SKIP_HW_RESET, &reset_context.flags);
-
-	adev->no_hw_access = true;
-	r = amdgpu_device_pre_asic_reset(adev, &reset_context);
-	adev->no_hw_access = false;
-	if (r)
-		goto out;
+	set_bit(AMDGPU_SKIP_COREDUMP, &reset_context.flags);
+	INIT_LIST_HEAD(&device_list);
 
-	r = amdgpu_do_asic_reset(&device_list, &reset_context);
+	hive = amdgpu_get_xgmi_hive(adev);
+	if (hive) {
+		mutex_lock(&hive->hive_lock);
+		reset_context.hive = hive;
+		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
+			tmp_adev->pcie_reset_ctx.in_link_reset = true;
+			list_add_tail(&tmp_adev->reset_list, &device_list);
+		}
+	} else {
+		set_bit(AMDGPU_SKIP_HW_RESET, &reset_context.flags);
+		list_add_tail(&adev->reset_list, &device_list);
+	}
 
+	r = amdgpu_device_asic_reset(adev, &device_list, &reset_context);
 out:
 	if (!r) {
 		if (amdgpu_device_cache_pci_state(adev->pdev))
 			pci_restore_state(adev->pdev);
-
-		DRM_INFO("PCIe error recovery succeeded\n");
+		dev_info(adev->dev, "PCIe error recovery succeeded\n");
 	} else {
-		DRM_ERROR("PCIe error recovery failed, err:%d", r);
-		amdgpu_device_unset_mp1_state(adev);
-		amdgpu_device_unlock_reset_domain(adev->reset_domain);
+		dev_err(adev->dev, "PCIe error recovery failed, err:%d\n", r);
+		if (hive) {
+			list_for_each_entry(tmp_adev, &device_list, reset_list)
+				amdgpu_device_unset_mp1_state(tmp_adev);
+		}
+		amdgpu_device_recovery_put_reset_lock(adev, &device_list);
+	}
+
+	if (hive) {
+		mutex_unlock(&hive->hive_lock);
+		amdgpu_put_xgmi_hive(hive);
 	}
 
 	return r ? PCI_ERS_RESULT_DISCONNECT : PCI_ERS_RESULT_RECOVERED;
@@ -5812,26 +7275,95 @@ void amdgpu_pci_resume(struct pci_dev *pdev)
 {
 	struct drm_device *dev = pci_get_drvdata(pdev);
 	struct amdgpu_device *adev = drm_to_adev(dev);
-	int i;
-
+	struct list_head device_list;
+	struct amdgpu_hive_info *hive = NULL;
+	struct amdgpu_device *tmp_adev = NULL;
 
-	DRM_INFO("PCI error: resume callback!!\n");
+	dev_info(adev->dev, "PCI error: resume callback!!\n");
 
 	/* Only continue execution for the case of pci_channel_io_frozen */
 	if (adev->pci_channel_state != pci_channel_io_frozen)
 		return;
 
-	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
-		struct amdgpu_ring *ring = adev->rings[i];
+	INIT_LIST_HEAD(&device_list);
 
-		if (!ring || !ring->sched.thread)
-			continue;
+	hive = amdgpu_get_xgmi_hive(adev);
+	if (hive) {
+		mutex_lock(&hive->hive_lock);
+		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
+			tmp_adev->pcie_reset_ctx.in_link_reset = false;
+			list_add_tail(&tmp_adev->reset_list, &device_list);
+		}
+	} else
+		list_add_tail(&adev->reset_list, &device_list);
+
+	amdgpu_device_sched_resume(&device_list, NULL, NULL);
+	amdgpu_device_gpu_resume(adev, &device_list, false);
+	amdgpu_device_recovery_put_reset_lock(adev, &device_list);
 
-		drm_sched_start(&ring->sched, true);
+	if (hive) {
+		mutex_unlock(&hive->hive_lock);
+		amdgpu_put_xgmi_hive(hive);
 	}
+}
+
+static void amdgpu_device_cache_switch_state(struct amdgpu_device *adev)
+{
+	struct pci_dev *swus, *swds;
+	int r;
+
+	swds = pci_upstream_bridge(adev->pdev);
+	if (!swds || swds->vendor != PCI_VENDOR_ID_ATI ||
+	    pci_pcie_type(swds) != PCI_EXP_TYPE_DOWNSTREAM)
+		return;
+	swus = pci_upstream_bridge(swds);
+	if (!swus ||
+	    (swus->vendor != PCI_VENDOR_ID_ATI &&
+	     swus->vendor != PCI_VENDOR_ID_AMD) ||
+	    pci_pcie_type(swus) != PCI_EXP_TYPE_UPSTREAM)
+		return;
+
+	/* If already saved, return */
+	if (adev->pcie_reset_ctx.swus)
+		return;
+	/* Upstream bridge is ATI, assume it's SWUS/DS architecture */
+	r = pci_save_state(swds);
+	if (r)
+		return;
+	adev->pcie_reset_ctx.swds_pcistate = pci_store_saved_state(swds);
+
+	r = pci_save_state(swus);
+	if (r)
+		return;
+	adev->pcie_reset_ctx.swus_pcistate = pci_store_saved_state(swus);
 
-	amdgpu_device_unset_mp1_state(adev);
-	amdgpu_device_unlock_reset_domain(adev->reset_domain);
+	adev->pcie_reset_ctx.swus = swus;
+}
+
+static void amdgpu_device_load_switch_state(struct amdgpu_device *adev)
+{
+	struct pci_dev *pdev;
+	int r;
+
+	if (!adev->pcie_reset_ctx.swds_pcistate ||
+	    !adev->pcie_reset_ctx.swus_pcistate)
+		return;
+
+	pdev = adev->pcie_reset_ctx.swus;
+	r = pci_load_saved_state(pdev, adev->pcie_reset_ctx.swus_pcistate);
+	if (!r) {
+		pci_restore_state(pdev);
+	} else {
+		dev_warn(adev->dev, "Failed to load SWUS state, err:%d\n", r);
+		return;
+	}
+
+	pdev = pci_upstream_bridge(adev->pdev);
+	r = pci_load_saved_state(pdev, adev->pcie_reset_ctx.swds_pcistate);
+	if (!r)
+		pci_restore_state(pdev);
+	else
+		dev_warn(adev->dev, "Failed to load SWDS state, err:%d\n", r);
 }
 
 bool amdgpu_device_cache_pci_state(struct pci_dev *pdev)
@@ -5840,6 +7372,9 @@ bool amdgpu_device_cache_pci_state(struct pci_dev *pdev)
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	int r;
 
+	if (amdgpu_sriov_vf(adev))
+		return false;
+
 	r = pci_save_state(pdev);
 	if (!r) {
 		kfree(adev->pci_state);
@@ -5847,14 +7382,16 @@ bool amdgpu_device_cache_pci_state(struct pci_dev *pdev)
 		adev->pci_state = pci_store_saved_state(pdev);
 
 		if (!adev->pci_state) {
-			DRM_ERROR("Failed to store PCI saved state");
+			dev_err(adev->dev, "Failed to store PCI saved state");
 			return false;
 		}
 	} else {
-		DRM_WARN("Failed to save PCI state, err:%d\n", r);
+		dev_warn(adev->dev, "Failed to save PCI state, err:%d\n", r);
 		return false;
 	}
 
+	amdgpu_device_cache_switch_state(adev);
+
 	return true;
 }
 
@@ -5872,7 +7409,7 @@ bool amdgpu_device_load_pci_state(struct pci_dev *pdev)
 	if (!r) {
 		pci_restore_state(pdev);
 	} else {
-		DRM_WARN("Failed to load PCI state, err:%d\n", r);
+		dev_warn(adev->dev, "Failed to load PCI state, err:%d\n", r);
 		return false;
 	}
 
@@ -5889,10 +7426,17 @@ void amdgpu_device_flush_hdp(struct amdgpu_device *adev,
 	if (adev->gmc.xgmi.connected_to_cpu)
 		return;
 
-	if (ring && ring->funcs->emit_hdp_flush)
+	if (ring && ring->funcs->emit_hdp_flush) {
 		amdgpu_ring_emit_hdp_flush(ring);
-	else
-		amdgpu_asic_flush_hdp(adev, ring);
+		return;
+	}
+
+	if (!ring && amdgpu_sriov_runtime(adev)) {
+		if (!amdgpu_kiq_hdp_flush(adev))
+			return;
+	}
+
+	amdgpu_hdp_flush(adev, ring);
 }
 
 void amdgpu_device_invalidate_hdp(struct amdgpu_device *adev,
@@ -5905,7 +7449,7 @@ void amdgpu_device_invalidate_hdp(struct amdgpu_device *adev,
 	if (adev->gmc.xgmi.connected_to_cpu)
 		return;
 
-	amdgpu_asic_invalidate_hdp(adev, ring);
+	amdgpu_hdp_invalidate(adev, ring);
 }
 
 int amdgpu_in_reset(struct amdgpu_device *adev)
@@ -5938,6 +7482,7 @@ void amdgpu_device_halt(struct amdgpu_device *adev)
 	struct pci_dev *pdev = adev->pdev;
 	struct drm_device *ddev = adev_to_drm(adev);
 
+	amdgpu_xcp_dev_unplug(adev);
 	drm_dev_unplug(ddev);
 
 	amdgpu_irq_disable_all(adev);
@@ -5986,6 +7531,22 @@ void amdgpu_device_pcie_port_wreg(struct amdgpu_device *adev,
 }
 
 /**
+ * amdgpu_device_get_gang - return a reference to the current gang
+ * @adev: amdgpu_device pointer
+ *
+ * Returns: A new reference to the current gang leader.
+ */
+struct dma_fence *amdgpu_device_get_gang(struct amdgpu_device *adev)
+{
+	struct dma_fence *fence;
+
+	rcu_read_lock();
+	fence = dma_fence_get_rcu_safe(&adev->gang_submit);
+	rcu_read_unlock();
+	return fence;
+}
+
+/**
  * amdgpu_device_switch_gang - switch to a new gang
  * @adev: amdgpu_device pointer
  * @gang: the gang to switch to
@@ -5999,25 +7560,117 @@ struct dma_fence *amdgpu_device_switch_gang(struct amdgpu_device *adev,
 {
 	struct dma_fence *old = NULL;
 
+	dma_fence_get(gang);
 	do {
 		dma_fence_put(old);
-		rcu_read_lock();
-		old = dma_fence_get_rcu_safe(&adev->gang_submit);
-		rcu_read_unlock();
-
+		old = amdgpu_device_get_gang(adev);
 		if (old == gang)
 			break;
 
-		if (!dma_fence_is_signaled(old))
+		if (!dma_fence_is_signaled(old)) {
+			dma_fence_put(gang);
 			return old;
+		}
 
 	} while (cmpxchg((struct dma_fence __force **)&adev->gang_submit,
 			 old, gang) != old);
 
+	/*
+	 * Drop it once for the exchanged reference in adev and once for the
+	 * thread local reference acquired in amdgpu_device_get_gang().
+	 */
+	dma_fence_put(old);
 	dma_fence_put(old);
 	return NULL;
 }
 
+/**
+ * amdgpu_device_enforce_isolation - enforce HW isolation
+ * @adev: the amdgpu device pointer
+ * @ring: the HW ring the job is supposed to run on
+ * @job: the job which is about to be pushed to the HW ring
+ *
+ * Makes sure that only one client at a time can use the GFX block.
+ * Returns: The dependency to wait on before the job can be pushed to the HW.
+ * The function is called multiple times until NULL is returned.
+ */
+struct dma_fence *amdgpu_device_enforce_isolation(struct amdgpu_device *adev,
+						  struct amdgpu_ring *ring,
+						  struct amdgpu_job *job)
+{
+	struct amdgpu_isolation *isolation = &adev->isolation[ring->xcp_id];
+	struct drm_sched_fence *f = job->base.s_fence;
+	struct dma_fence *dep;
+	void *owner;
+	int r;
+
+	/*
+	 * For now enforce isolation only for the GFX block since we only need
+	 * the cleaner shader on those rings.
+	 */
+	if (ring->funcs->type != AMDGPU_RING_TYPE_GFX &&
+	    ring->funcs->type != AMDGPU_RING_TYPE_COMPUTE)
+		return NULL;
+
+	/*
+	 * All submissions where enforce isolation is false are handled as if
+	 * they come from a single client. Use ~0l as the owner to distinct it
+	 * from kernel submissions where the owner is NULL.
+	 */
+	owner = job->enforce_isolation ? f->owner : (void *)~0l;
+
+	mutex_lock(&adev->enforce_isolation_mutex);
+
+	/*
+	 * The "spearhead" submission is the first one which changes the
+	 * ownership to its client. We always need to wait for it to be
+	 * pushed to the HW before proceeding with anything.
+	 */
+	if (&f->scheduled != isolation->spearhead &&
+	    !dma_fence_is_signaled(isolation->spearhead)) {
+		dep = isolation->spearhead;
+		goto out_grab_ref;
+	}
+
+	if (isolation->owner != owner) {
+
+		/*
+		 * Wait for any gang to be assembled before switching to a
+		 * different owner or otherwise we could deadlock the
+		 * submissions.
+		 */
+		if (!job->gang_submit) {
+			dep = amdgpu_device_get_gang(adev);
+			if (!dma_fence_is_signaled(dep))
+				goto out_return_dep;
+			dma_fence_put(dep);
+		}
+
+		dma_fence_put(isolation->spearhead);
+		isolation->spearhead = dma_fence_get(&f->scheduled);
+		amdgpu_sync_move(&isolation->active, &isolation->prev);
+		trace_amdgpu_isolation(isolation->owner, owner);
+		isolation->owner = owner;
+	}
+
+	/*
+	 * Specifying the ring here helps to pipeline submissions even when
+	 * isolation is enabled. If that is not desired for testing NULL can be
+	 * used instead of the ring to enforce a CPU round trip while switching
+	 * between clients.
+	 */
+	dep = amdgpu_sync_peek_fence(&isolation->prev, ring);
+	r = amdgpu_sync_fence(&isolation->active, &f->finished, GFP_NOWAIT);
+	if (r)
+		dev_warn(adev->dev, "OOM tracking isolation\n");
+
+out_grab_ref:
+	dma_fence_get(dep);
+out_return_dep:
+	mutex_unlock(&adev->enforce_isolation_mutex);
+	return dep;
+}
+
 bool amdgpu_device_has_display_hardware(struct amdgpu_device *adev)
 {
 	switch (adev->asic_type) {
@@ -6052,9 +7705,133 @@ bool amdgpu_device_has_display_hardware(struct amdgpu_device *adev)
 		return true;
 	default:
 		/* IP discovery */
-		if (!adev->ip_versions[DCE_HWIP][0] ||
+		if (!amdgpu_ip_version(adev, DCE_HWIP, 0) ||
 		    (adev->harvest_ip_mask & AMD_HARVEST_IP_DMU_MASK))
 			return false;
 		return true;
 	}
 }
+
+uint32_t amdgpu_device_wait_on_rreg(struct amdgpu_device *adev,
+		uint32_t inst, uint32_t reg_addr, char reg_name[],
+		uint32_t expected_value, uint32_t mask)
+{
+	uint32_t ret = 0;
+	uint32_t old_ = 0;
+	uint32_t tmp_ = RREG32(reg_addr);
+	uint32_t loop = adev->usec_timeout;
+
+	while ((tmp_ & (mask)) != (expected_value)) {
+		if (old_ != tmp_) {
+			loop = adev->usec_timeout;
+			old_ = tmp_;
+		} else
+			udelay(1);
+		tmp_ = RREG32(reg_addr);
+		loop--;
+		if (!loop) {
+			dev_warn(
+				adev->dev,
+				"Register(%d) [%s] failed to reach value 0x%08x != 0x%08xn",
+				inst, reg_name, (uint32_t)expected_value,
+				(uint32_t)(tmp_ & (mask)));
+			ret = -ETIMEDOUT;
+			break;
+		}
+	}
+	return ret;
+}
+
+ssize_t amdgpu_get_soft_full_reset_mask(struct amdgpu_ring *ring)
+{
+	ssize_t size = 0;
+
+	if (!ring || !ring->adev)
+		return size;
+
+	if (amdgpu_device_should_recover_gpu(ring->adev))
+		size |= AMDGPU_RESET_TYPE_FULL;
+
+	if (unlikely(!ring->adev->debug_disable_soft_recovery) &&
+	    !amdgpu_sriov_vf(ring->adev) && ring->funcs->soft_recovery)
+		size |= AMDGPU_RESET_TYPE_SOFT_RESET;
+
+	return size;
+}
+
+ssize_t amdgpu_show_reset_mask(char *buf, uint32_t supported_reset)
+{
+	ssize_t size = 0;
+
+	if (supported_reset == 0) {
+		size += sysfs_emit_at(buf, size, "unsupported");
+		size += sysfs_emit_at(buf, size, "\n");
+		return size;
+
+	}
+
+	if (supported_reset & AMDGPU_RESET_TYPE_SOFT_RESET)
+		size += sysfs_emit_at(buf, size, "soft ");
+
+	if (supported_reset & AMDGPU_RESET_TYPE_PER_QUEUE)
+		size += sysfs_emit_at(buf, size, "queue ");
+
+	if (supported_reset & AMDGPU_RESET_TYPE_PER_PIPE)
+		size += sysfs_emit_at(buf, size, "pipe ");
+
+	if (supported_reset & AMDGPU_RESET_TYPE_FULL)
+		size += sysfs_emit_at(buf, size, "full ");
+
+	size += sysfs_emit_at(buf, size, "\n");
+	return size;
+}
+
+void amdgpu_device_set_uid(struct amdgpu_uid *uid_info,
+			   enum amdgpu_uid_type type, uint8_t inst,
+			   uint64_t uid)
+{
+	if (!uid_info)
+		return;
+
+	if (type >= AMDGPU_UID_TYPE_MAX) {
+		dev_err_once(uid_info->adev->dev, "Invalid UID type %d\n",
+			     type);
+		return;
+	}
+
+	if (inst >= AMDGPU_UID_INST_MAX) {
+		dev_err_once(uid_info->adev->dev, "Invalid UID instance %d\n",
+			     inst);
+		return;
+	}
+
+	if (uid_info->uid[type][inst] != 0) {
+		dev_warn_once(
+			uid_info->adev->dev,
+			"Overwriting existing UID %llu for type %d instance %d\n",
+			uid_info->uid[type][inst], type, inst);
+	}
+
+	uid_info->uid[type][inst] = uid;
+}
+
+u64 amdgpu_device_get_uid(struct amdgpu_uid *uid_info,
+			  enum amdgpu_uid_type type, uint8_t inst)
+{
+	if (!uid_info)
+		return 0;
+
+	if (type >= AMDGPU_UID_TYPE_MAX) {
+		dev_err_once(uid_info->adev->dev, "Invalid UID type %d\n",
+			     type);
+		return 0;
+	}
+
+	if (inst >= AMDGPU_UID_INST_MAX) {
+		dev_err_once(uid_info->adev->dev, "Invalid UID instance %d\n",
+			     inst);
+		return 0;
+	}
+
+	return uid_info->uid[type][inst];
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_df.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_df.h
index 1538b2dbfff1..eb605e79ae0e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_df.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_df.h
@@ -33,6 +33,7 @@ struct amdgpu_df_hash_status {
 struct amdgpu_df_funcs {
 	void (*sw_init)(struct amdgpu_device *adev);
 	void (*sw_fini)(struct amdgpu_device *adev);
+	void (*hw_init)(struct amdgpu_device *adev);
 	void (*enable_broadcast_mode)(struct amdgpu_device *adev,
 				      bool enable);
 	u32 (*get_fb_channel_number)(struct amdgpu_device *adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c
index 5139334925ea..fa2a22dfa048 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2018 Advanced Micro Devices, Inc.
+ * Copyright 2018-2024 Advanced Micro Devices, Inc. All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -27,17 +27,22 @@
 #include "amdgpu_discovery.h"
 #include "soc15_hw_ip.h"
 #include "discovery.h"
+#include "amdgpu_ras.h"
 
 #include "soc15.h"
 #include "gfx_v9_0.h"
+#include "gfx_v9_4_3.h"
 #include "gmc_v9_0.h"
 #include "df_v1_7.h"
 #include "df_v3_6.h"
 #include "df_v4_3.h"
+#include "df_v4_6_2.h"
+#include "df_v4_15.h"
 #include "nbio_v6_1.h"
 #include "nbio_v7_0.h"
 #include "nbio_v7_4.h"
 #include "nbio_v7_9.h"
+#include "nbio_v7_11.h"
 #include "hdp_v4_0.h"
 #include "vega10_ih.h"
 #include "vega20_ih.h"
@@ -51,43 +56,77 @@
 #include "smuio_v9_0.h"
 #include "gmc_v10_0.h"
 #include "gmc_v11_0.h"
+#include "gmc_v12_0.h"
 #include "gfxhub_v2_0.h"
 #include "mmhub_v2_0.h"
 #include "nbio_v2_3.h"
 #include "nbio_v4_3.h"
 #include "nbio_v7_2.h"
 #include "nbio_v7_7.h"
+#include "nbif_v6_3_1.h"
 #include "hdp_v5_0.h"
 #include "hdp_v5_2.h"
 #include "hdp_v6_0.h"
+#include "hdp_v7_0.h"
 #include "nv.h"
 #include "soc21.h"
+#include "soc24.h"
 #include "navi10_ih.h"
 #include "ih_v6_0.h"
+#include "ih_v6_1.h"
+#include "ih_v7_0.h"
 #include "gfx_v10_0.h"
 #include "gfx_v11_0.h"
+#include "gfx_v12_0.h"
 #include "sdma_v5_0.h"
 #include "sdma_v5_2.h"
 #include "sdma_v6_0.h"
+#include "sdma_v7_0.h"
 #include "lsdma_v6_0.h"
+#include "lsdma_v7_0.h"
 #include "vcn_v2_0.h"
 #include "jpeg_v2_0.h"
 #include "vcn_v3_0.h"
 #include "jpeg_v3_0.h"
 #include "vcn_v4_0.h"
 #include "jpeg_v4_0.h"
+#include "vcn_v4_0_3.h"
+#include "jpeg_v4_0_3.h"
+#include "vcn_v4_0_5.h"
+#include "jpeg_v4_0_5.h"
 #include "amdgpu_vkms.h"
-#include "mes_v10_1.h"
 #include "mes_v11_0.h"
+#include "mes_v12_0.h"
 #include "smuio_v11_0.h"
 #include "smuio_v11_0_6.h"
 #include "smuio_v13_0.h"
+#include "smuio_v13_0_3.h"
 #include "smuio_v13_0_6.h"
+#include "smuio_v14_0_2.h"
+#include "vcn_v5_0_0.h"
+#include "vcn_v5_0_1.h"
+#include "jpeg_v5_0_0.h"
+#include "jpeg_v5_0_1.h"
+#include "amdgpu_ras_mgr.h"
+
+#include "amdgpu_vpe.h"
+#if defined(CONFIG_DRM_AMD_ISP)
+#include "amdgpu_isp.h"
+#endif
 
-#define FIRMWARE_IP_DISCOVERY "amdgpu/ip_discovery.bin"
-MODULE_FIRMWARE(FIRMWARE_IP_DISCOVERY);
-
+MODULE_FIRMWARE("amdgpu/ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/vega10_ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/vega12_ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/vega20_ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/raven_ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/raven2_ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/picasso_ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/arcturus_ip_discovery.bin");
+MODULE_FIRMWARE("amdgpu/aldebaran_ip_discovery.bin");
+
+#define mmIP_DISCOVERY_VERSION  0x16A00
 #define mmRCC_CONFIG_MEMSIZE	0xde3
+#define mmMP0_SMN_C2PMSG_33	0x16061
 #define mmMM_INDEX		0x0
 #define mmMM_INDEX_HI		0x6
 #define mmMM_DATA		0x1
@@ -169,6 +208,7 @@ static const char *hw_id_names[HW_ID_MAX] = {
 	[XGMI_HWID]		= "XGMI",
 	[XGBE_HWID]		= "XGBE",
 	[MP0_HWID]		= "MP0",
+	[VPE_HWID]		= "VPE",
 };
 
 static int hw_id_map[MAX_HWIP] = {
@@ -198,37 +238,117 @@ static int hw_id_map[MAX_HWIP] = {
 	[XGMI_HWIP]	= XGMI_HWID,
 	[DCI_HWIP]	= DCI_HWID,
 	[PCIE_HWIP]	= PCIE_HWID,
+	[VPE_HWIP]	= VPE_HWID,
+	[ISP_HWIP]	= ISP_HWID,
 };
 
-static int amdgpu_discovery_read_binary_from_vram(struct amdgpu_device *adev, uint8_t *binary)
+static int amdgpu_discovery_read_binary_from_sysmem(struct amdgpu_device *adev, uint8_t *binary)
 {
-	uint64_t vram_size = (uint64_t)RREG32(mmRCC_CONFIG_MEMSIZE) << 20;
-	uint64_t pos = vram_size - DISCOVERY_TMR_OFFSET;
+	u64 tmr_offset, tmr_size, pos;
+	void *discv_regn;
+	int ret;
 
-	amdgpu_device_vram_access(adev, pos, (uint32_t *)binary,
-				  adev->mman.discovery_tmr_size, false);
-	return 0;
+	ret = amdgpu_acpi_get_tmr_info(adev, &tmr_offset, &tmr_size);
+	if (ret)
+		return ret;
+
+	pos = tmr_offset + tmr_size - DISCOVERY_TMR_OFFSET;
+
+	/* This region is read-only and reserved from system use */
+	discv_regn = memremap(pos, adev->discovery.size, MEMREMAP_WC);
+	if (discv_regn) {
+		memcpy(binary, discv_regn, adev->discovery.size);
+		memunmap(discv_regn);
+		return 0;
+	}
+
+	return -ENOENT;
 }
 
-static int amdgpu_discovery_read_binary_from_file(struct amdgpu_device *adev, uint8_t *binary)
+#define IP_DISCOVERY_V2		2
+#define IP_DISCOVERY_V4		4
+
+static int amdgpu_discovery_read_binary_from_mem(struct amdgpu_device *adev,
+						 uint8_t *binary)
 {
-	const struct firmware *fw;
-	const char *fw_name;
-	int r;
+	bool sz_valid = true;
+	uint64_t vram_size;
+	int i, ret = 0;
+	u32 msg;
+
+	if (!amdgpu_sriov_vf(adev)) {
+		/* It can take up to two second for IFWI init to complete on some dGPUs,
+		 * but generally it should be in the 60-100ms range.  Normally this starts
+		 * as soon as the device gets power so by the time the OS loads this has long
+		 * completed.  However, when a card is hotplugged via e.g., USB4, we need to
+		 * wait for this to complete.  Once the C2PMSG is updated, we can
+		 * continue.
+		 */
 
-	switch (amdgpu_discovery) {
-	case 2:
-		fw_name = FIRMWARE_IP_DISCOVERY;
-		break;
-	default:
-		dev_warn(adev->dev, "amdgpu_discovery is not set properly\n");
-		return -EINVAL;
+		for (i = 0; i < 2000; i++) {
+			msg = RREG32(mmMP0_SMN_C2PMSG_33);
+			if (msg & 0x80000000)
+				break;
+			msleep(1);
+		}
 	}
 
-	r = request_firmware(&fw, fw_name, adev->dev);
+	vram_size = RREG32(mmRCC_CONFIG_MEMSIZE);
+	if (!vram_size || vram_size == U32_MAX)
+		sz_valid = false;
+	else
+		vram_size <<= 20;
+
+	/*
+	 * If in VRAM, discovery TMR is marked for reservation. If it is in system mem,
+	 * then it is not required to be reserved.
+	 */
+	if (sz_valid) {
+		if (amdgpu_sriov_vf(adev) && adev->virt.is_dynamic_crit_regn_enabled) {
+			/* For SRIOV VFs with dynamic critical region enabled,
+			 * we will get the IPD binary via below call.
+			 * If dynamic critical is disabled, fall through to normal seq.
+			 */
+			if (amdgpu_virt_get_dynamic_data_info(adev,
+						AMD_SRIOV_MSG_IPD_TABLE_ID, binary,
+						&adev->discovery.size)) {
+				dev_err(adev->dev,
+						"failed to read discovery info from dynamic critical region.");
+				ret = -EINVAL;
+				goto exit;
+			}
+		} else {
+			uint64_t pos = vram_size - DISCOVERY_TMR_OFFSET;
+
+			amdgpu_device_vram_access(adev, pos, (uint32_t *)binary,
+					adev->discovery.size, false);
+			adev->discovery.reserve_tmr = true;
+		}
+	} else {
+		ret = amdgpu_discovery_read_binary_from_sysmem(adev, binary);
+	}
+
+	if (ret)
+		dev_err(adev->dev,
+			"failed to read discovery info from memory, vram size read: %llx",
+			vram_size);
+exit:
+	return ret;
+}
+
+static int amdgpu_discovery_read_binary_from_file(struct amdgpu_device *adev,
+							uint8_t *binary,
+							const char *fw_name)
+{
+	const struct firmware *fw;
+	int r;
+
+	r = firmware_request_nowarn(&fw, fw_name, adev->dev);
 	if (r) {
-		dev_err(adev->dev, "can't load firmware \"%s\"\n",
-			fw_name);
+		if (amdgpu_discovery == 2)
+			dev_err(adev->dev, "can't load firmware \"%s\"\n", fw_name);
+		else
+			drm_info(&adev->ddev, "Optional firmware \"%s\" was not found\n", fw_name);
 		return r;
 	}
 
@@ -269,8 +389,8 @@ static void amdgpu_discovery_harvest_config_quirk(struct amdgpu_device *adev)
 	 * So far, apply this quirk only on those Navy Flounder boards which
 	 * have a bad harvest table of VCN config.
 	 */
-	if ((adev->ip_versions[UVD_HWIP][1] == IP_VERSION(3, 0, 1)) &&
-		(adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 2))) {
+	if ((amdgpu_ip_version(adev, UVD_HWIP, 1) == IP_VERSION(3, 0, 1)) &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 3, 2))) {
 		switch (adev->pdev->revision) {
 		case 0xC1:
 		case 0xC2:
@@ -280,6 +400,7 @@ static void amdgpu_discovery_harvest_config_quirk(struct amdgpu_device *adev)
 		case 0xCF:
 		case 0xDF:
 			adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN1;
+			adev->vcn.inst_mask &= ~AMDGPU_VCN_HARVEST_VCN1;
 			break;
 		default:
 			break;
@@ -287,58 +408,118 @@ static void amdgpu_discovery_harvest_config_quirk(struct amdgpu_device *adev)
 	}
 }
 
+static int amdgpu_discovery_verify_npsinfo(struct amdgpu_device *adev,
+					   struct binary_header *bhdr)
+{
+	uint8_t *discovery_bin = adev->discovery.bin;
+	struct table_info *info;
+	uint16_t checksum;
+	uint16_t offset;
+
+	info = &bhdr->table_list[NPS_INFO];
+	offset = le16_to_cpu(info->offset);
+	checksum = le16_to_cpu(info->checksum);
+
+	struct nps_info_header *nhdr =
+		(struct nps_info_header *)(discovery_bin + offset);
+
+	if (le32_to_cpu(nhdr->table_id) != NPS_INFO_TABLE_ID) {
+		dev_dbg(adev->dev, "invalid ip discovery nps info table id\n");
+		return -EINVAL;
+	}
+
+	if (!amdgpu_discovery_verify_checksum(discovery_bin + offset,
+					      le32_to_cpu(nhdr->size_bytes),
+					      checksum)) {
+		dev_dbg(adev->dev, "invalid nps info data table checksum\n");
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static const char *amdgpu_discovery_get_fw_name(struct amdgpu_device *adev)
+{
+	if (amdgpu_discovery == 2) {
+		/* Assume there is valid discovery TMR in VRAM even if binary is sideloaded */
+		adev->discovery.reserve_tmr = true;
+		return "amdgpu/ip_discovery.bin";
+	}
+
+	switch (adev->asic_type) {
+	case CHIP_VEGA10:
+		return "amdgpu/vega10_ip_discovery.bin";
+	case CHIP_VEGA12:
+		return "amdgpu/vega12_ip_discovery.bin";
+	case CHIP_RAVEN:
+		if (adev->apu_flags & AMD_APU_IS_RAVEN2)
+			return "amdgpu/raven2_ip_discovery.bin";
+		else if (adev->apu_flags & AMD_APU_IS_PICASSO)
+			return "amdgpu/picasso_ip_discovery.bin";
+		else
+			return "amdgpu/raven_ip_discovery.bin";
+	case CHIP_VEGA20:
+		return "amdgpu/vega20_ip_discovery.bin";
+	case CHIP_ARCTURUS:
+		return "amdgpu/arcturus_ip_discovery.bin";
+	case CHIP_ALDEBARAN:
+		return "amdgpu/aldebaran_ip_discovery.bin";
+	default:
+		return NULL;
+	}
+}
+
 static int amdgpu_discovery_init(struct amdgpu_device *adev)
 {
 	struct table_info *info;
 	struct binary_header *bhdr;
+	uint8_t *discovery_bin;
+	const char *fw_name;
 	uint16_t offset;
 	uint16_t size;
 	uint16_t checksum;
 	int r;
 
-	adev->mman.discovery_tmr_size = DISCOVERY_TMR_SIZE;
-	adev->mman.discovery_bin = kzalloc(adev->mman.discovery_tmr_size, GFP_KERNEL);
-	if (!adev->mman.discovery_bin)
+	adev->discovery.bin = kzalloc(DISCOVERY_TMR_SIZE, GFP_KERNEL);
+	if (!adev->discovery.bin)
 		return -ENOMEM;
+	adev->discovery.size = DISCOVERY_TMR_SIZE;
+	adev->discovery.debugfs_blob.data = adev->discovery.bin;
+	adev->discovery.debugfs_blob.size = adev->discovery.size;
+
+	discovery_bin = adev->discovery.bin;
+	/* Read from file if it is the preferred option */
+	fw_name = amdgpu_discovery_get_fw_name(adev);
+	if (fw_name != NULL) {
+		drm_dbg(&adev->ddev, "use ip discovery information from file");
+		r = amdgpu_discovery_read_binary_from_file(adev, discovery_bin,
+							   fw_name);
+		if (r)
+			goto out;
+	} else {
+		drm_dbg(&adev->ddev, "use ip discovery information from memory");
+		r = amdgpu_discovery_read_binary_from_mem(adev, discovery_bin);
+		if (r)
+			goto out;
+	}
 
-	r = amdgpu_discovery_read_binary_from_vram(adev, adev->mman.discovery_bin);
-	if (r) {
-		dev_err(adev->dev, "failed to read ip discovery binary from vram\n");
+	/* check the ip discovery binary signature */
+	if (!amdgpu_discovery_verify_binary_signature(discovery_bin)) {
+		dev_err(adev->dev,
+			"get invalid ip discovery binary signature\n");
 		r = -EINVAL;
 		goto out;
 	}
 
-	if (!amdgpu_discovery_verify_binary_signature(adev->mman.discovery_bin) || amdgpu_discovery == 2) {
-		/* ignore the discovery binary from vram if discovery=2 in kernel module parameter */
-		if (amdgpu_discovery == 2)
-			dev_info(adev->dev,"force read ip discovery binary from file");
-		else
-			dev_warn(adev->dev, "get invalid ip discovery binary signature from vram\n");
-
-		/* retry read ip discovery binary from file */
-		r = amdgpu_discovery_read_binary_from_file(adev, adev->mman.discovery_bin);
-		if (r) {
-			dev_err(adev->dev, "failed to read ip discovery binary from file\n");
-			r = -EINVAL;
-			goto out;
-		}
-		/* check the ip discovery binary signature */
-		if(!amdgpu_discovery_verify_binary_signature(adev->mman.discovery_bin)) {
-			dev_warn(adev->dev, "get invalid ip discovery binary signature from file\n");
-			r = -EINVAL;
-			goto out;
-		}
-	}
-
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
+	bhdr = (struct binary_header *)discovery_bin;
 
 	offset = offsetof(struct binary_header, binary_checksum) +
 		sizeof(bhdr->binary_checksum);
 	size = le16_to_cpu(bhdr->binary_size) - offset;
 	checksum = le16_to_cpu(bhdr->binary_checksum);
 
-	if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset,
-					      size, checksum)) {
+	if (!amdgpu_discovery_verify_checksum(discovery_bin + offset, size,
+					      checksum)) {
 		dev_err(adev->dev, "invalid ip discovery binary checksum\n");
 		r = -EINVAL;
 		goto out;
@@ -350,15 +531,16 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 
 	if (offset) {
 		struct ip_discovery_header *ihdr =
-			(struct ip_discovery_header *)(adev->mman.discovery_bin + offset);
+			(struct ip_discovery_header *)(discovery_bin + offset);
 		if (le32_to_cpu(ihdr->signature) != DISCOVERY_TABLE_SIGNATURE) {
 			dev_err(adev->dev, "invalid ip discovery data table signature\n");
 			r = -EINVAL;
 			goto out;
 		}
 
-		if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset,
-						      le16_to_cpu(ihdr->size), checksum)) {
+		if (!amdgpu_discovery_verify_checksum(discovery_bin + offset,
+						      le16_to_cpu(ihdr->size),
+						      checksum)) {
 			dev_err(adev->dev, "invalid ip discovery data table checksum\n");
 			r = -EINVAL;
 			goto out;
@@ -371,7 +553,7 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 
 	if (offset) {
 		struct gpu_info_header *ghdr =
-			(struct gpu_info_header *)(adev->mman.discovery_bin + offset);
+			(struct gpu_info_header *)(discovery_bin + offset);
 
 		if (le32_to_cpu(ghdr->table_id) != GC_TABLE_ID) {
 			dev_err(adev->dev, "invalid ip discovery gc table id\n");
@@ -379,8 +561,9 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 			goto out;
 		}
 
-		if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset,
-						      le32_to_cpu(ghdr->size), checksum)) {
+		if (!amdgpu_discovery_verify_checksum(discovery_bin + offset,
+						      le32_to_cpu(ghdr->size),
+						      checksum)) {
 			dev_err(adev->dev, "invalid gc data table checksum\n");
 			r = -EINVAL;
 			goto out;
@@ -393,7 +576,7 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 
 	if (offset) {
 		struct harvest_info_header *hhdr =
-			(struct harvest_info_header *)(adev->mman.discovery_bin + offset);
+			(struct harvest_info_header *)(discovery_bin + offset);
 
 		if (le32_to_cpu(hhdr->signature) != HARVEST_TABLE_SIGNATURE) {
 			dev_err(adev->dev, "invalid ip discovery harvest table signature\n");
@@ -401,8 +584,9 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 			goto out;
 		}
 
-		if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset,
-						      sizeof(struct harvest_table), checksum)) {
+		if (!amdgpu_discovery_verify_checksum(
+			    discovery_bin + offset,
+			    sizeof(struct harvest_table), checksum)) {
 			dev_err(adev->dev, "invalid harvest data table checksum\n");
 			r = -EINVAL;
 			goto out;
@@ -415,7 +599,7 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 
 	if (offset) {
 		struct vcn_info_header *vhdr =
-			(struct vcn_info_header *)(adev->mman.discovery_bin + offset);
+			(struct vcn_info_header *)(discovery_bin + offset);
 
 		if (le32_to_cpu(vhdr->table_id) != VCN_INFO_TABLE_ID) {
 			dev_err(adev->dev, "invalid ip discovery vcn table id\n");
@@ -423,8 +607,9 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 			goto out;
 		}
 
-		if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset,
-						      le32_to_cpu(vhdr->size_bytes), checksum)) {
+		if (!amdgpu_discovery_verify_checksum(
+			    discovery_bin + offset,
+			    le32_to_cpu(vhdr->size_bytes), checksum)) {
 			dev_err(adev->dev, "invalid vcn data table checksum\n");
 			r = -EINVAL;
 			goto out;
@@ -437,7 +622,7 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 
 	if (0 && offset) {
 		struct mall_info_header *mhdr =
-			(struct mall_info_header *)(adev->mman.discovery_bin + offset);
+			(struct mall_info_header *)(discovery_bin + offset);
 
 		if (le32_to_cpu(mhdr->table_id) != MALL_INFO_TABLE_ID) {
 			dev_err(adev->dev, "invalid ip discovery mall table id\n");
@@ -445,8 +630,9 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 			goto out;
 		}
 
-		if (!amdgpu_discovery_verify_checksum(adev->mman.discovery_bin + offset,
-						      le32_to_cpu(mhdr->size_bytes), checksum)) {
+		if (!amdgpu_discovery_verify_checksum(
+			    discovery_bin + offset,
+			    le32_to_cpu(mhdr->size_bytes), checksum)) {
 			dev_err(adev->dev, "invalid mall data table checksum\n");
 			r = -EINVAL;
 			goto out;
@@ -456,9 +642,11 @@ static int amdgpu_discovery_init(struct amdgpu_device *adev)
 	return 0;
 
 out:
-	kfree(adev->mman.discovery_bin);
-	adev->mman.discovery_bin = NULL;
-
+	kfree(adev->discovery.bin);
+	adev->discovery.bin = NULL;
+	if ((amdgpu_discovery != 2) &&
+	    (RREG32(mmIP_DISCOVERY_VERSION) == 4))
+		amdgpu_ras_query_boot_status(adev, 4);
 	return r;
 }
 
@@ -467,20 +655,23 @@ static void amdgpu_discovery_sysfs_fini(struct amdgpu_device *adev);
 void amdgpu_discovery_fini(struct amdgpu_device *adev)
 {
 	amdgpu_discovery_sysfs_fini(adev);
-	kfree(adev->mman.discovery_bin);
-	adev->mman.discovery_bin = NULL;
+	kfree(adev->discovery.bin);
+	adev->discovery.bin = NULL;
 }
 
-static int amdgpu_discovery_validate_ip(const struct ip *ip)
+static int amdgpu_discovery_validate_ip(struct amdgpu_device *adev,
+					uint8_t instance, uint16_t hw_id)
 {
-	if (ip->number_instance >= HWIP_MAX_INSTANCE) {
-		DRM_ERROR("Unexpected number_instance (%d) from ip discovery blob\n",
-			  ip->number_instance);
+	if (instance >= HWIP_MAX_INSTANCE) {
+		dev_err(adev->dev,
+			"Unexpected instance_number (%d) from ip discovery blob\n",
+			instance);
 		return -EINVAL;
 	}
-	if (le16_to_cpu(ip->hw_id) >= HW_ID_MAX) {
-		DRM_ERROR("Unexpected hw_id (%d) from ip discovery blob\n",
-			  le16_to_cpu(ip->hw_id));
+	if (hw_id >= HW_ID_MAX) {
+		dev_err(adev->dev,
+			"Unexpected hw_id (%d) from ip discovery blob\n",
+			hw_id);
 		return -EINVAL;
 	}
 
@@ -490,49 +681,64 @@ static int amdgpu_discovery_validate_ip(const struct ip *ip)
 static void amdgpu_discovery_read_harvest_bit_per_ip(struct amdgpu_device *adev,
 						uint32_t *vcn_harvest_count)
 {
+	uint8_t *discovery_bin = adev->discovery.bin;
 	struct binary_header *bhdr;
 	struct ip_discovery_header *ihdr;
 	struct die_header *dhdr;
 	struct ip *ip;
 	uint16_t die_offset, ip_offset, num_dies, num_ips;
+	uint16_t hw_id;
+	uint8_t inst;
 	int i, j;
 
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
-	ihdr = (struct ip_discovery_header *)(adev->mman.discovery_bin +
-			le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset));
+	bhdr = (struct binary_header *)discovery_bin;
+	ihdr = (struct ip_discovery_header
+			*)(discovery_bin +
+			   le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset));
 	num_dies = le16_to_cpu(ihdr->num_dies);
 
 	/* scan harvest bit of all IP data structures */
 	for (i = 0; i < num_dies; i++) {
 		die_offset = le16_to_cpu(ihdr->die_info[i].die_offset);
-		dhdr = (struct die_header *)(adev->mman.discovery_bin + die_offset);
+		dhdr = (struct die_header *)(discovery_bin + die_offset);
 		num_ips = le16_to_cpu(dhdr->num_ips);
 		ip_offset = die_offset + sizeof(*dhdr);
 
 		for (j = 0; j < num_ips; j++) {
-			ip = (struct ip *)(adev->mman.discovery_bin + ip_offset);
-
-			if (amdgpu_discovery_validate_ip(ip))
+			ip = (struct ip *)(discovery_bin + ip_offset);
+			inst = ip->number_instance;
+			hw_id = le16_to_cpu(ip->hw_id);
+			if (amdgpu_discovery_validate_ip(adev, inst, hw_id))
 				goto next_ip;
 
-			if (le16_to_cpu(ip->harvest) == 1) {
-				switch (le16_to_cpu(ip->hw_id)) {
+			if (ip->harvest == 1) {
+				switch (hw_id) {
 				case VCN_HWID:
 					(*vcn_harvest_count)++;
-					if (ip->number_instance == 0)
+					if (inst == 0) {
 						adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN0;
-					else
+						adev->vcn.inst_mask &=
+							~AMDGPU_VCN_HARVEST_VCN0;
+						adev->jpeg.inst_mask &=
+							~AMDGPU_VCN_HARVEST_VCN0;
+					} else {
 						adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN1;
+						adev->vcn.inst_mask &=
+							~AMDGPU_VCN_HARVEST_VCN1;
+						adev->jpeg.inst_mask &=
+							~AMDGPU_VCN_HARVEST_VCN1;
+					}
 					break;
 				case DMU_HWID:
 					adev->harvest_ip_mask |= AMD_HARVEST_IP_DMU_MASK;
 					break;
 				default:
 					break;
-                                }
-                        }
+				}
+			}
 next_ip:
-			ip_offset += struct_size(ip, base_address, ip->num_base_address);
+			ip_offset += struct_size(ip, base_address,
+						 ip->num_base_address);
 		}
 	}
 }
@@ -541,13 +747,14 @@ static void amdgpu_discovery_read_from_harvest_table(struct amdgpu_device *adev,
 						     uint32_t *vcn_harvest_count,
 						     uint32_t *umc_harvest_count)
 {
+	uint8_t *discovery_bin = adev->discovery.bin;
 	struct binary_header *bhdr;
 	struct harvest_table *harvest_info;
 	u16 offset;
 	int i;
 	uint32_t umc_harvest_config = 0;
 
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
+	bhdr = (struct binary_header *)discovery_bin;
 	offset = le16_to_cpu(bhdr->table_list[HARVEST_INFO].offset);
 
 	if (!offset) {
@@ -555,7 +762,7 @@ static void amdgpu_discovery_read_from_harvest_table(struct amdgpu_device *adev,
 		return;
 	}
 
-	harvest_info = (struct harvest_table *)(adev->mman.discovery_bin + offset);
+	harvest_info = (struct harvest_table *)(discovery_bin + offset);
 
 	for (i = 0; i < 32; i++) {
 		if (le16_to_cpu(harvest_info->list[i].hw_id) == 0)
@@ -564,10 +771,15 @@ static void amdgpu_discovery_read_from_harvest_table(struct amdgpu_device *adev,
 		switch (le16_to_cpu(harvest_info->list[i].hw_id)) {
 		case VCN_HWID:
 			(*vcn_harvest_count)++;
-			if (harvest_info->list[i].number_instance == 0)
-				adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN0;
-			else
-				adev->vcn.harvest_config |= AMDGPU_VCN_HARVEST_VCN1;
+			adev->vcn.harvest_config |=
+				(1 << harvest_info->list[i].number_instance);
+			adev->jpeg.harvest_config |=
+				(1 << harvest_info->list[i].number_instance);
+
+			adev->vcn.inst_mask &=
+				~(1U << harvest_info->list[i].number_instance);
+			adev->jpeg.inst_mask &=
+				~(1U << harvest_info->list[i].number_instance);
 			break;
 		case DMU_HWID:
 			adev->harvest_ip_mask |= AMD_HARVEST_IP_DMU_MASK;
@@ -577,6 +789,20 @@ static void amdgpu_discovery_read_from_harvest_table(struct amdgpu_device *adev,
 				1 << (le16_to_cpu(harvest_info->list[i].number_instance));
 			(*umc_harvest_count)++;
 			break;
+		case GC_HWID:
+			adev->gfx.xcc_mask &=
+				~(1U << harvest_info->list[i].number_instance);
+			break;
+		case SDMA0_HWID:
+			adev->sdma.sdma_mask &=
+				~(1U << harvest_info->list[i].number_instance);
+			break;
+#if defined(CONFIG_DRM_AMD_ISP)
+		case ISP_HWID:
+			adev->isp.harvest_config |=
+				~(1U << harvest_info->list[i].number_instance);
+			break;
+#endif
 		default:
 			break;
 		}
@@ -597,7 +823,7 @@ struct ip_hw_instance {
 	u8  harvest;
 
 	int num_base_addresses;
-	u32 base_addr[];
+	u32 base_addr[] __counted_by(num_base_addresses);
 };
 
 struct ip_hw_id {
@@ -832,15 +1058,51 @@ static void ip_disc_release(struct kobject *kobj)
 						       kobj);
 	struct amdgpu_device *adev = ip_top->adev;
 
-	adev->ip_top = NULL;
 	kfree(ip_top);
+	adev->discovery.ip_top = NULL;
+}
+
+static uint8_t amdgpu_discovery_get_harvest_info(struct amdgpu_device *adev,
+						 uint16_t hw_id, uint8_t inst)
+{
+	uint8_t harvest = 0;
+
+	/* Until a uniform way is figured, get mask based on hwid */
+	switch (hw_id) {
+	case VCN_HWID:
+		/* VCN vs UVD+VCE */
+		if (!amdgpu_ip_version(adev, VCE_HWIP, 0))
+			harvest = ((1 << inst) & adev->vcn.inst_mask) == 0;
+		break;
+	case DMU_HWID:
+		if (adev->harvest_ip_mask & AMD_HARVEST_IP_DMU_MASK)
+			harvest = 0x1;
+		break;
+	case UMC_HWID:
+		/* TODO: It needs another parsing; for now, ignore.*/
+		break;
+	case GC_HWID:
+		harvest = ((1 << inst) & adev->gfx.xcc_mask) == 0;
+		break;
+	case SDMA0_HWID:
+		harvest = ((1 << inst) & adev->sdma.sdma_mask) == 0;
+		break;
+	default:
+		break;
+	}
+
+	return harvest;
 }
 
 static int amdgpu_discovery_sysfs_ips(struct amdgpu_device *adev,
 				      struct ip_die_entry *ip_die_entry,
-				      const size_t _ip_offset, const int num_ips)
+				      const size_t _ip_offset, const int num_ips,
+				      bool reg_base_64)
 {
+	uint8_t *discovery_bin = adev->discovery.bin;
 	int ii, jj, kk, res;
+	uint16_t hw_id;
+	uint8_t inst;
 
 	DRM_DEBUG("num_ips:%d", num_ips);
 
@@ -852,12 +1114,14 @@ static int amdgpu_discovery_sysfs_ips(struct amdgpu_device *adev,
 		size_t ip_offset = _ip_offset;
 
 		for (jj = 0; jj < num_ips; jj++) {
-			struct ip *ip;
+			struct ip_v4 *ip;
 			struct ip_hw_instance *ip_hw_instance;
 
-			ip = (struct ip *)(adev->mman.discovery_bin + ip_offset);
-			if (amdgpu_discovery_validate_ip(ip) ||
-			    le16_to_cpu(ip->hw_id) != ii)
+			ip = (struct ip_v4 *)(discovery_bin + ip_offset);
+			inst = ip->instance_number;
+			hw_id = le16_to_cpu(ip->hw_id);
+			if (amdgpu_discovery_validate_ip(adev, inst, hw_id) ||
+			    hw_id != ii)
 				goto next_ip;
 
 			DRM_DEBUG("match:%d @ ip_offset:%zu", ii, ip_offset);
@@ -903,22 +1167,35 @@ static int amdgpu_discovery_sysfs_ips(struct amdgpu_device *adev,
 				return -ENOMEM;
 			}
 			ip_hw_instance->hw_id = le16_to_cpu(ip->hw_id); /* == ii */
-			ip_hw_instance->num_instance = ip->number_instance;
+			ip_hw_instance->num_instance = ip->instance_number;
 			ip_hw_instance->major = ip->major;
 			ip_hw_instance->minor = ip->minor;
 			ip_hw_instance->revision = ip->revision;
-			ip_hw_instance->harvest = ip->harvest;
+			ip_hw_instance->harvest =
+				amdgpu_discovery_get_harvest_info(
+					adev, ip_hw_instance->hw_id,
+					ip_hw_instance->num_instance);
 			ip_hw_instance->num_base_addresses = ip->num_base_address;
 
-			for (kk = 0; kk < ip_hw_instance->num_base_addresses; kk++)
-				ip_hw_instance->base_addr[kk] = ip->base_address[kk];
+			for (kk = 0; kk < ip_hw_instance->num_base_addresses; kk++) {
+				if (reg_base_64)
+					ip_hw_instance->base_addr[kk] =
+						lower_32_bits(le64_to_cpu(ip->base_address_64[kk])) & 0x3FFFFFFF;
+				else
+					ip_hw_instance->base_addr[kk] = ip->base_address[kk];
+			}
 
 			kobject_init(&ip_hw_instance->kobj, &ip_hw_instance_ktype);
 			ip_hw_instance->kobj.kset = &ip_hw_id->hw_id_kset;
 			res = kobject_add(&ip_hw_instance->kobj, NULL,
 					  "%d", ip_hw_instance->num_instance);
 next_ip:
-			ip_offset += struct_size(ip, base_address, ip->num_base_address);
+			if (reg_base_64)
+				ip_offset += struct_size(ip, base_address_64,
+							 ip->num_base_address);
+			else
+				ip_offset += struct_size(ip, base_address,
+							 ip->num_base_address);
 		}
 	}
 
@@ -927,17 +1204,20 @@ next_ip:
 
 static int amdgpu_discovery_sysfs_recurse(struct amdgpu_device *adev)
 {
+	struct ip_discovery_top *ip_top = adev->discovery.ip_top;
+	uint8_t *discovery_bin = adev->discovery.bin;
 	struct binary_header *bhdr;
 	struct ip_discovery_header *ihdr;
 	struct die_header *dhdr;
-	struct kset *die_kset = &adev->ip_top->die_kset;
+	struct kset *die_kset = &ip_top->die_kset;
 	u16 num_dies, die_offset, num_ips;
 	size_t ip_offset;
 	int ii, res;
 
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
-	ihdr = (struct ip_discovery_header *)(adev->mman.discovery_bin +
-					      le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset));
+	bhdr = (struct binary_header *)discovery_bin;
+	ihdr = (struct ip_discovery_header
+			*)(discovery_bin +
+			   le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset));
 	num_dies = le16_to_cpu(ihdr->num_dies);
 
 	DRM_DEBUG("number of dies: %d\n", num_dies);
@@ -946,7 +1226,7 @@ static int amdgpu_discovery_sysfs_recurse(struct amdgpu_device *adev)
 		struct ip_die_entry *ip_die_entry;
 
 		die_offset = le16_to_cpu(ihdr->die_info[ii].die_offset);
-		dhdr = (struct die_header *)(adev->mman.discovery_bin + die_offset);
+		dhdr = (struct die_header *)(discovery_bin + die_offset);
 		num_ips = le16_to_cpu(dhdr->num_ips);
 		ip_offset = die_offset + sizeof(*dhdr);
 
@@ -972,7 +1252,7 @@ static int amdgpu_discovery_sysfs_recurse(struct amdgpu_device *adev)
 			return res;
 		}
 
-		amdgpu_discovery_sysfs_ips(adev, ip_die_entry, ip_offset, num_ips);
+		amdgpu_discovery_sysfs_ips(adev, ip_die_entry, ip_offset, num_ips, !!ihdr->base_addr_64_bit);
 	}
 
 	return 0;
@@ -980,27 +1260,32 @@ static int amdgpu_discovery_sysfs_recurse(struct amdgpu_device *adev)
 
 static int amdgpu_discovery_sysfs_init(struct amdgpu_device *adev)
 {
+	uint8_t *discovery_bin = adev->discovery.bin;
+	struct ip_discovery_top *ip_top;
 	struct kset *die_kset;
 	int res, ii;
 
-	adev->ip_top = kzalloc(sizeof(*adev->ip_top), GFP_KERNEL);
-	if (!adev->ip_top)
-		return -ENOMEM;
+	if (!discovery_bin)
+		return -EINVAL;
 
-	adev->ip_top->adev = adev;
+	ip_top = kzalloc(sizeof(*ip_top), GFP_KERNEL);
+	if (!ip_top)
+		return -ENOMEM;
 
-	res = kobject_init_and_add(&adev->ip_top->kobj, &ip_discovery_ktype,
+	ip_top->adev = adev;
+	adev->discovery.ip_top = ip_top;
+	res = kobject_init_and_add(&ip_top->kobj, &ip_discovery_ktype,
 				   &adev->dev->kobj, "ip_discovery");
 	if (res) {
 		DRM_ERROR("Couldn't init and add ip_discovery/");
 		goto Err;
 	}
 
-	die_kset = &adev->ip_top->die_kset;
+	die_kset = &ip_top->die_kset;
 	kobject_set_name(&die_kset->kobj, "%s", "die");
-	die_kset->kobj.parent = &adev->ip_top->kobj;
+	die_kset->kobj.parent = &ip_top->kobj;
 	die_kset->kobj.ktype = &die_kobj_ktype;
-	res = kset_register(&adev->ip_top->die_kset);
+	res = kset_register(&ip_top->die_kset);
 	if (res) {
 		DRM_ERROR("Couldn't register die_kset");
 		goto Err;
@@ -1014,7 +1299,7 @@ static int amdgpu_discovery_sysfs_init(struct amdgpu_device *adev)
 
 	return res;
 Err:
-	kobject_put(&adev->ip_top->kobj);
+	kobject_put(&ip_top->kobj);
 	return res;
 }
 
@@ -1059,10 +1344,11 @@ static void amdgpu_discovery_sysfs_die_free(struct ip_die_entry *ip_die_entry)
 
 static void amdgpu_discovery_sysfs_fini(struct amdgpu_device *adev)
 {
+	struct ip_discovery_top *ip_top = adev->discovery.ip_top;
 	struct list_head *el, *tmp;
 	struct kset *die_kset;
 
-	die_kset = &adev->ip_top->die_kset;
+	die_kset = &ip_top->die_kset;
 	spin_lock(&die_kset->list_lock);
 	list_for_each_prev_safe(el, tmp, &die_kset->list) {
 		list_del_init(el);
@@ -1071,43 +1357,51 @@ static void amdgpu_discovery_sysfs_fini(struct amdgpu_device *adev)
 		spin_lock(&die_kset->list_lock);
 	}
 	spin_unlock(&die_kset->list_lock);
-	kobject_put(&adev->ip_top->die_kset.kobj);
-	kobject_put(&adev->ip_top->kobj);
+	kobject_put(&ip_top->die_kset.kobj);
+	kobject_put(&ip_top->kobj);
 }
 
 /* ================================================== */
 
 static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev)
 {
+	uint8_t num_base_address, subrev, variant;
 	struct binary_header *bhdr;
 	struct ip_discovery_header *ihdr;
 	struct die_header *dhdr;
-	struct ip *ip;
+	uint8_t *discovery_bin;
+	struct ip_v4 *ip;
 	uint16_t die_offset;
 	uint16_t ip_offset;
 	uint16_t num_dies;
+	uint32_t wafl_ver;
 	uint16_t num_ips;
-	uint8_t num_base_address;
+	uint16_t hw_id;
+	uint8_t inst;
 	int hw_ip;
 	int i, j, k;
 	int r;
 
 	r = amdgpu_discovery_init(adev);
-	if (r) {
-		DRM_ERROR("amdgpu_discovery_init failed\n");
+	if (r)
 		return r;
-	}
-
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
-	ihdr = (struct ip_discovery_header *)(adev->mman.discovery_bin +
-			le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset));
+	discovery_bin = adev->discovery.bin;
+	wafl_ver = 0;
+	adev->gfx.xcc_mask = 0;
+	adev->sdma.sdma_mask = 0;
+	adev->vcn.inst_mask = 0;
+	adev->jpeg.inst_mask = 0;
+	bhdr = (struct binary_header *)discovery_bin;
+	ihdr = (struct ip_discovery_header
+			*)(discovery_bin +
+			   le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset));
 	num_dies = le16_to_cpu(ihdr->num_dies);
 
 	DRM_DEBUG("number of dies: %d\n", num_dies);
 
 	for (i = 0; i < num_dies; i++) {
 		die_offset = le16_to_cpu(ihdr->die_info[i].die_offset);
-		dhdr = (struct die_header *)(adev->mman.discovery_bin + die_offset);
+		dhdr = (struct die_header *)(discovery_bin + die_offset);
 		num_ips = le16_to_cpu(dhdr->num_ips);
 		ip_offset = die_offset + sizeof(*dhdr);
 
@@ -1121,9 +1415,11 @@ static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev)
 				le16_to_cpu(dhdr->die_id), num_ips);
 
 		for (j = 0; j < num_ips; j++) {
-			ip = (struct ip *)(adev->mman.discovery_bin + ip_offset);
+			ip = (struct ip_v4 *)(discovery_bin + ip_offset);
 
-			if (amdgpu_discovery_validate_ip(ip))
+			inst = ip->instance_number;
+			hw_id = le16_to_cpu(ip->hw_id);
+			if (amdgpu_discovery_validate_ip(adev, inst, hw_id))
 				goto next_ip;
 
 			num_base_address = ip->num_base_address;
@@ -1131,7 +1427,7 @@ static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev)
 			DRM_DEBUG("%s(%d) #%d v%d.%d.%d:\n",
 				  hw_id_names[le16_to_cpu(ip->hw_id)],
 				  le16_to_cpu(ip->hw_id),
-				  ip->number_instance,
+				  ip->instance_number,
 				  ip->major, ip->minor,
 				  ip->revision);
 
@@ -1142,26 +1438,45 @@ static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev)
 				 *     0b10 : encode is disabled
 				 *     0b01 : decode is disabled
 				 */
-				adev->vcn.vcn_config[adev->vcn.num_vcn_inst] =
-					ip->revision & 0xc0;
-				ip->revision &= ~0xc0;
-				if (adev->vcn.num_vcn_inst < AMDGPU_MAX_VCN_INSTANCES)
+				if (adev->vcn.num_vcn_inst <
+				    AMDGPU_MAX_VCN_INSTANCES) {
+					adev->vcn.inst[adev->vcn.num_vcn_inst].vcn_config =
+						ip->revision & 0xc0;
 					adev->vcn.num_vcn_inst++;
-				else
+					adev->vcn.inst_mask |=
+						(1U << ip->instance_number);
+					adev->jpeg.inst_mask |=
+						(1U << ip->instance_number);
+				} else {
 					dev_err(adev->dev, "Too many VCN instances: %d vs %d\n",
 						adev->vcn.num_vcn_inst + 1,
 						AMDGPU_MAX_VCN_INSTANCES);
+				}
+				ip->revision &= ~0xc0;
 			}
 			if (le16_to_cpu(ip->hw_id) == SDMA0_HWID ||
 			    le16_to_cpu(ip->hw_id) == SDMA1_HWID ||
 			    le16_to_cpu(ip->hw_id) == SDMA2_HWID ||
 			    le16_to_cpu(ip->hw_id) == SDMA3_HWID) {
-				if (adev->sdma.num_instances < AMDGPU_MAX_SDMA_INSTANCES)
+				if (adev->sdma.num_instances <
+				    AMDGPU_MAX_SDMA_INSTANCES) {
 					adev->sdma.num_instances++;
-				else
+					adev->sdma.sdma_mask |=
+						(1U << ip->instance_number);
+				} else {
 					dev_err(adev->dev, "Too many SDMA instances: %d vs %d\n",
 						adev->sdma.num_instances + 1,
 						AMDGPU_MAX_SDMA_INSTANCES);
+				}
+			}
+
+			if (le16_to_cpu(ip->hw_id) == VPE_HWID) {
+				if (adev->vpe.num_instances < AMDGPU_MAX_VPE_INSTANCES)
+					adev->vpe.num_instances++;
+				else
+					dev_err(adev->dev, "Too many VPE instances: %d vs %d\n",
+						adev->vpe.num_instances + 1,
+						AMDGPU_MAX_VPE_INSTANCES);
 			}
 
 			if (le16_to_cpu(ip->hw_id) == UMC_HWID) {
@@ -1169,20 +1484,42 @@ static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev)
 				adev->umc.node_inst_num++;
 			}
 
+			if (le16_to_cpu(ip->hw_id) == GC_HWID)
+				adev->gfx.xcc_mask |=
+					(1U << ip->instance_number);
+
+			if (!wafl_ver && le16_to_cpu(ip->hw_id) == WAFLC_HWID)
+				wafl_ver = IP_VERSION_FULL(ip->major, ip->minor,
+							   ip->revision, 0, 0);
+
 			for (k = 0; k < num_base_address; k++) {
 				/*
 				 * convert the endianness of base addresses in place,
 				 * so that we don't need to convert them when accessing adev->reg_offset.
 				 */
-				ip->base_address[k] = le32_to_cpu(ip->base_address[k]);
+				if (ihdr->base_addr_64_bit)
+					/* Truncate the 64bit base address from ip discovery
+					 * and only store lower 32bit ip base in reg_offset[].
+					 * Bits > 32 follows ASIC specific format, thus just
+					 * discard them and handle it within specific ASIC.
+					 * By this way reg_offset[] and related helpers can
+					 * stay unchanged.
+					 * The base address is in dwords, thus clear the
+					 * highest 2 bits to store.
+					 */
+					ip->base_address[k] =
+						lower_32_bits(le64_to_cpu(ip->base_address_64[k])) & 0x3FFFFFFF;
+				else
+					ip->base_address[k] = le32_to_cpu(ip->base_address[k]);
 				DRM_DEBUG("\t0x%08x\n", ip->base_address[k]);
 			}
 
 			for (hw_ip = 0; hw_ip < MAX_HWIP; hw_ip++) {
-				if (hw_id_map[hw_ip] == le16_to_cpu(ip->hw_id)) {
+				if (hw_id_map[hw_ip] == le16_to_cpu(ip->hw_id) &&
+				    hw_id_map[hw_ip] != 0) {
 					DRM_DEBUG("set register base offset for %s\n",
 							hw_id_names[le16_to_cpu(ip->hw_id)]);
-					adev->reg_offset[hw_ip][ip->number_instance] =
+					adev->reg_offset[hw_ip][ip->instance_number] =
 						ip->base_address;
 					/* Instance support is somewhat inconsistent.
 					 * SDMA is a good example.  Sienna cichlid has 4 total
@@ -1193,80 +1530,59 @@ static int amdgpu_discovery_reg_base_init(struct amdgpu_device *adev)
 					 * example.  On most chips there are multiple instances
 					 * with the same HWID.
 					 */
-					adev->ip_versions[hw_ip][ip->number_instance] =
-						IP_VERSION(ip->major, ip->minor, ip->revision);
+
+					if (ihdr->version < 3) {
+						subrev = 0;
+						variant = 0;
+					} else {
+						subrev = ip->sub_revision;
+						variant = ip->variant;
+					}
+
+					adev->ip_versions[hw_ip]
+							 [ip->instance_number] =
+						IP_VERSION_FULL(ip->major,
+								ip->minor,
+								ip->revision,
+								variant,
+								subrev);
 				}
 			}
 
 next_ip:
-			ip_offset += struct_size(ip, base_address, ip->num_base_address);
+			if (ihdr->base_addr_64_bit)
+				ip_offset += struct_size(ip, base_address_64, ip->num_base_address);
+			else
+				ip_offset += struct_size(ip, base_address, ip->num_base_address);
 		}
 	}
 
-	amdgpu_discovery_sysfs_init(adev);
+	if (wafl_ver && !adev->ip_versions[XGMI_HWIP][0])
+		adev->ip_versions[XGMI_HWIP][0] = wafl_ver;
 
 	return 0;
 }
 
-int amdgpu_discovery_get_ip_version(struct amdgpu_device *adev, int hw_id, int number_instance,
-				    int *major, int *minor, int *revision)
-{
-	struct binary_header *bhdr;
-	struct ip_discovery_header *ihdr;
-	struct die_header *dhdr;
-	struct ip *ip;
-	uint16_t die_offset;
-	uint16_t ip_offset;
-	uint16_t num_dies;
-	uint16_t num_ips;
-	int i, j;
-
-	if (!adev->mman.discovery_bin) {
-		DRM_ERROR("ip discovery uninitialized\n");
-		return -EINVAL;
-	}
-
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
-	ihdr = (struct ip_discovery_header *)(adev->mman.discovery_bin +
-			le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset));
-	num_dies = le16_to_cpu(ihdr->num_dies);
-
-	for (i = 0; i < num_dies; i++) {
-		die_offset = le16_to_cpu(ihdr->die_info[i].die_offset);
-		dhdr = (struct die_header *)(adev->mman.discovery_bin + die_offset);
-		num_ips = le16_to_cpu(dhdr->num_ips);
-		ip_offset = die_offset + sizeof(*dhdr);
-
-		for (j = 0; j < num_ips; j++) {
-			ip = (struct ip *)(adev->mman.discovery_bin + ip_offset);
-
-			if ((le16_to_cpu(ip->hw_id) == hw_id) && (ip->number_instance == number_instance)) {
-				if (major)
-					*major = ip->major;
-				if (minor)
-					*minor = ip->minor;
-				if (revision)
-					*revision = ip->revision;
-				return 0;
-			}
-			ip_offset += struct_size(ip, base_address, ip->num_base_address);
-		}
-	}
-
-	return -EINVAL;
-}
-
 static void amdgpu_discovery_harvest_ip(struct amdgpu_device *adev)
 {
+	uint8_t *discovery_bin = adev->discovery.bin;
+	struct ip_discovery_header *ihdr;
+	struct binary_header *bhdr;
 	int vcn_harvest_count = 0;
 	int umc_harvest_count = 0;
+	uint16_t offset, ihdr_ver;
 
+	bhdr = (struct binary_header *)discovery_bin;
+	offset = le16_to_cpu(bhdr->table_list[IP_DISCOVERY].offset);
+	ihdr = (struct ip_discovery_header *)(discovery_bin + offset);
+	ihdr_ver = le16_to_cpu(ihdr->version);
 	/*
 	 * Harvest table does not fit Navi1x and legacy GPUs,
 	 * so read harvest bit per IP data structure to set
 	 * harvest configuration.
 	 */
-	if (adev->ip_versions[GC_HWIP][0] < IP_VERSION(10, 2, 0)) {
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(10, 2, 0) &&
+	    ihdr_ver <= 2) {
 		if ((adev->pdev->device == 0x731E &&
 			(adev->pdev->revision == 0xC6 ||
 			 adev->pdev->revision == 0xC7)) ||
@@ -1298,27 +1614,30 @@ union gc_info {
 	struct gc_info_v1_0 v1;
 	struct gc_info_v1_1 v1_1;
 	struct gc_info_v1_2 v1_2;
+	struct gc_info_v1_3 v1_3;
 	struct gc_info_v2_0 v2;
+	struct gc_info_v2_1 v2_1;
 };
 
 static int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev)
 {
+	uint8_t *discovery_bin = adev->discovery.bin;
 	struct binary_header *bhdr;
 	union gc_info *gc_info;
 	u16 offset;
 
-	if (!adev->mman.discovery_bin) {
+	if (!discovery_bin) {
 		DRM_ERROR("ip discovery uninitialized\n");
 		return -EINVAL;
 	}
 
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
+	bhdr = (struct binary_header *)discovery_bin;
 	offset = le16_to_cpu(bhdr->table_list[GC].offset);
 
 	if (!offset)
 		return 0;
 
-	gc_info = (union gc_info *)(adev->mman.discovery_bin + offset);
+	gc_info = (union gc_info *)(discovery_bin + offset);
 
 	switch (le16_to_cpu(gc_info->v1.header.version_major)) {
 	case 1:
@@ -1340,12 +1659,12 @@ static int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev)
 		adev->gfx.config.num_sc_per_sh = le32_to_cpu(gc_info->v1.gc_num_sc_per_se) /
 			le32_to_cpu(gc_info->v1.gc_num_sa_per_se);
 		adev->gfx.config.num_packer_per_sc = le32_to_cpu(gc_info->v1.gc_num_packer_per_sc);
-		if (gc_info->v1.header.version_minor >= 1) {
+		if (le16_to_cpu(gc_info->v1.header.version_minor) >= 1) {
 			adev->gfx.config.gc_num_tcp_per_sa = le32_to_cpu(gc_info->v1_1.gc_num_tcp_per_sa);
 			adev->gfx.config.gc_num_sdp_interface = le32_to_cpu(gc_info->v1_1.gc_num_sdp_interface);
 			adev->gfx.config.gc_num_tcps = le32_to_cpu(gc_info->v1_1.gc_num_tcps);
 		}
-		if (gc_info->v1.header.version_minor >= 2) {
+		if (le16_to_cpu(gc_info->v1.header.version_minor) >= 2) {
 			adev->gfx.config.gc_num_tcp_per_wpg = le32_to_cpu(gc_info->v1_2.gc_num_tcp_per_wpg);
 			adev->gfx.config.gc_tcp_l1_size = le32_to_cpu(gc_info->v1_2.gc_tcp_l1_size);
 			adev->gfx.config.gc_num_sqc_per_wgp = le32_to_cpu(gc_info->v1_2.gc_num_sqc_per_wgp);
@@ -1355,6 +1674,16 @@ static int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev)
 			adev->gfx.config.gc_gl1c_size_per_instance = le32_to_cpu(gc_info->v1_2.gc_gl1c_size_per_instance);
 			adev->gfx.config.gc_gl2c_per_gpu = le32_to_cpu(gc_info->v1_2.gc_gl2c_per_gpu);
 		}
+		if (le16_to_cpu(gc_info->v1.header.version_minor) >= 3) {
+			adev->gfx.config.gc_tcp_size_per_cu = le32_to_cpu(gc_info->v1_3.gc_tcp_size_per_cu);
+			adev->gfx.config.gc_tcp_cache_line_size = le32_to_cpu(gc_info->v1_3.gc_tcp_cache_line_size);
+			adev->gfx.config.gc_instruction_cache_size_per_sqc = le32_to_cpu(gc_info->v1_3.gc_instruction_cache_size_per_sqc);
+			adev->gfx.config.gc_instruction_cache_line_size = le32_to_cpu(gc_info->v1_3.gc_instruction_cache_line_size);
+			adev->gfx.config.gc_scalar_data_cache_size_per_sqc = le32_to_cpu(gc_info->v1_3.gc_scalar_data_cache_size_per_sqc);
+			adev->gfx.config.gc_scalar_data_cache_line_size = le32_to_cpu(gc_info->v1_3.gc_scalar_data_cache_line_size);
+			adev->gfx.config.gc_tcc_size = le32_to_cpu(gc_info->v1_3.gc_tcc_size);
+			adev->gfx.config.gc_tcc_cache_line_size = le32_to_cpu(gc_info->v1_3.gc_tcc_cache_line_size);
+		}
 		break;
 	case 2:
 		adev->gfx.config.max_shader_engines = le32_to_cpu(gc_info->v2.gc_num_se);
@@ -1374,6 +1703,15 @@ static int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev)
 		adev->gfx.config.num_sc_per_sh = le32_to_cpu(gc_info->v2.gc_num_sc_per_se) /
 			le32_to_cpu(gc_info->v2.gc_num_sh_per_se);
 		adev->gfx.config.num_packer_per_sc = le32_to_cpu(gc_info->v2.gc_num_packer_per_sc);
+		if (le16_to_cpu(gc_info->v2.header.version_minor) == 1) {
+			adev->gfx.config.gc_num_tcp_per_sa = le32_to_cpu(gc_info->v2_1.gc_num_tcp_per_sh);
+			adev->gfx.config.gc_tcp_size_per_cu = le32_to_cpu(gc_info->v2_1.gc_tcp_size_per_cu);
+			adev->gfx.config.gc_num_sdp_interface = le32_to_cpu(gc_info->v2_1.gc_num_sdp_interface); /* per XCD */
+			adev->gfx.config.gc_num_cu_per_sqc = le32_to_cpu(gc_info->v2_1.gc_num_cu_per_sqc);
+			adev->gfx.config.gc_l1_instruction_cache_size_per_sqc = le32_to_cpu(gc_info->v2_1.gc_instruction_cache_size_per_sqc);
+			adev->gfx.config.gc_l1_data_cache_size_per_sqc = le32_to_cpu(gc_info->v2_1.gc_scalar_data_cache_size_per_sqc);
+			adev->gfx.config.gc_tcc_size = le32_to_cpu(gc_info->v2_1.gc_tcc_size); /* per XCD */
+		}
 		break;
 	default:
 		dev_err(adev->dev,
@@ -1387,28 +1725,30 @@ static int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev)
 
 union mall_info {
 	struct mall_info_v1_0 v1;
+	struct mall_info_v2_0 v2;
 };
 
 static int amdgpu_discovery_get_mall_info(struct amdgpu_device *adev)
 {
+	uint8_t *discovery_bin = adev->discovery.bin;
 	struct binary_header *bhdr;
 	union mall_info *mall_info;
 	u32 u, mall_size_per_umc, m_s_present, half_use;
 	u64 mall_size;
 	u16 offset;
 
-	if (!adev->mman.discovery_bin) {
+	if (!discovery_bin) {
 		DRM_ERROR("ip discovery uninitialized\n");
 		return -EINVAL;
 	}
 
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
+	bhdr = (struct binary_header *)discovery_bin;
 	offset = le16_to_cpu(bhdr->table_list[MALL_INFO].offset);
 
 	if (!offset)
 		return 0;
 
-	mall_info = (union mall_info *)(adev->mman.discovery_bin + offset);
+	mall_info = (union mall_info *)(discovery_bin + offset);
 
 	switch (le16_to_cpu(mall_info->v1.header.version_major)) {
 	case 1:
@@ -1425,6 +1765,11 @@ static int amdgpu_discovery_get_mall_info(struct amdgpu_device *adev)
 				mall_size += mall_size_per_umc;
 		}
 		adev->gmc.mall_size = mall_size;
+		adev->gmc.m_half_use = half_use;
+		break;
+	case 2:
+		mall_size_per_umc = le32_to_cpu(mall_info->v2.mall_size_per_umc);
+		adev->gmc.mall_size = (uint64_t)mall_size_per_umc * adev->gmc.num_umc;
 		break;
 	default:
 		dev_err(adev->dev,
@@ -1442,12 +1787,13 @@ union vcn_info {
 
 static int amdgpu_discovery_get_vcn_info(struct amdgpu_device *adev)
 {
+	uint8_t *discovery_bin = adev->discovery.bin;
 	struct binary_header *bhdr;
 	union vcn_info *vcn_info;
 	u16 offset;
 	int v;
 
-	if (!adev->mman.discovery_bin) {
+	if (!discovery_bin) {
 		DRM_ERROR("ip discovery uninitialized\n");
 		return -EINVAL;
 	}
@@ -1462,13 +1808,13 @@ static int amdgpu_discovery_get_vcn_info(struct amdgpu_device *adev)
 		return -EINVAL;
 	}
 
-	bhdr = (struct binary_header *)adev->mman.discovery_bin;
+	bhdr = (struct binary_header *)discovery_bin;
 	offset = le16_to_cpu(bhdr->table_list[VCN_INFO].offset);
 
 	if (!offset)
 		return 0;
 
-	vcn_info = (union vcn_info *)(adev->mman.discovery_bin + offset);
+	vcn_info = (union vcn_info *)(discovery_bin + offset);
 
 	switch (le16_to_cpu(vcn_info->v1.header.version_major)) {
 	case 1:
@@ -1476,7 +1822,7 @@ static int amdgpu_discovery_get_vcn_info(struct amdgpu_device *adev)
 		 * so this won't overflow.
 		 */
 		for (v = 0; v < adev->vcn.num_vcn_inst; v++) {
-			adev->vcn.vcn_codec_disable_mask[v] =
+			adev->vcn.inst[v].vcn_codec_disable_mask =
 				le32_to_cpu(vcn_info->v1.instance_info[v].fuse_data.all_bits);
 		}
 		break;
@@ -1490,10 +1836,113 @@ static int amdgpu_discovery_get_vcn_info(struct amdgpu_device *adev)
 	return 0;
 }
 
+union nps_info {
+	struct nps_info_v1_0 v1;
+};
+
+static int amdgpu_discovery_refresh_nps_info(struct amdgpu_device *adev,
+					     union nps_info *nps_data)
+{
+	uint64_t vram_size, pos, offset;
+	struct nps_info_header *nhdr;
+	struct binary_header bhdr;
+	uint16_t checksum;
+
+	vram_size = (uint64_t)RREG32(mmRCC_CONFIG_MEMSIZE) << 20;
+	pos = vram_size - DISCOVERY_TMR_OFFSET;
+	amdgpu_device_vram_access(adev, pos, &bhdr, sizeof(bhdr), false);
+
+	offset = le16_to_cpu(bhdr.table_list[NPS_INFO].offset);
+	checksum = le16_to_cpu(bhdr.table_list[NPS_INFO].checksum);
+
+	amdgpu_device_vram_access(adev, (pos + offset), nps_data,
+				  sizeof(*nps_data), false);
+
+	nhdr = (struct nps_info_header *)(nps_data);
+	if (!amdgpu_discovery_verify_checksum((uint8_t *)nps_data,
+					      le32_to_cpu(nhdr->size_bytes),
+					      checksum)) {
+		dev_err(adev->dev, "nps data refresh, checksum mismatch\n");
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+int amdgpu_discovery_get_nps_info(struct amdgpu_device *adev,
+				  uint32_t *nps_type,
+				  struct amdgpu_gmc_memrange **ranges,
+				  int *range_cnt, bool refresh)
+{
+	uint8_t *discovery_bin = adev->discovery.bin;
+	struct amdgpu_gmc_memrange *mem_ranges;
+	struct binary_header *bhdr;
+	union nps_info *nps_info;
+	union nps_info nps_data;
+	u16 offset;
+	int i, r;
+
+	if (!nps_type || !range_cnt || !ranges)
+		return -EINVAL;
+
+	if (refresh) {
+		r = amdgpu_discovery_refresh_nps_info(adev, &nps_data);
+		if (r)
+			return r;
+		nps_info = &nps_data;
+	} else {
+		if (!discovery_bin) {
+			dev_err(adev->dev,
+				"fetch mem range failed, ip discovery uninitialized\n");
+			return -EINVAL;
+		}
+
+		bhdr = (struct binary_header *)discovery_bin;
+		offset = le16_to_cpu(bhdr->table_list[NPS_INFO].offset);
+
+		if (!offset)
+			return -ENOENT;
+
+		/* If verification fails, return as if NPS table doesn't exist */
+		if (amdgpu_discovery_verify_npsinfo(adev, bhdr))
+			return -ENOENT;
+
+		nps_info = (union nps_info *)(discovery_bin + offset);
+	}
+
+	switch (le16_to_cpu(nps_info->v1.header.version_major)) {
+	case 1:
+		mem_ranges = kvcalloc(nps_info->v1.count,
+				      sizeof(*mem_ranges),
+				      GFP_KERNEL);
+		if (!mem_ranges)
+			return -ENOMEM;
+		*nps_type = nps_info->v1.nps_type;
+		*range_cnt = nps_info->v1.count;
+		for (i = 0; i < *range_cnt; i++) {
+			mem_ranges[i].base_address =
+				nps_info->v1.instance_info[i].base_address;
+			mem_ranges[i].limit_address =
+				nps_info->v1.instance_info[i].limit_address;
+			mem_ranges[i].nid_mask = -1;
+			mem_ranges[i].flags = 0;
+		}
+		*ranges = mem_ranges;
+		break;
+	default:
+		dev_err(adev->dev, "Unhandled NPS info table %d.%d\n",
+			le16_to_cpu(nps_info->v1.header.version_major),
+			le16_to_cpu(nps_info->v1.header.version_minor));
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
 static int amdgpu_discovery_set_common_ip_blocks(struct amdgpu_device *adev)
 {
 	/* what IP to use for this? */
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 1, 0):
 	case IP_VERSION(9, 2, 1):
@@ -1502,6 +1951,9 @@ static int amdgpu_discovery_set_common_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(9, 4, 0):
 	case IP_VERSION(9, 4, 1):
 	case IP_VERSION(9, 4, 2):
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
 		amdgpu_device_ip_block_add(adev, &vega10_common_ip_block);
 		break;
 	case IP_VERSION(10, 1, 10):
@@ -1524,12 +1976,20 @@ static int amdgpu_discovery_set_common_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 		amdgpu_device_ip_block_add(adev, &soc21_common_ip_block);
 		break;
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		amdgpu_device_ip_block_add(adev, &soc24_common_ip_block);
+		break;
 	default:
 		dev_err(adev->dev,
 			"Failed to add common ip block(GC_HWIP:0x%x)\n",
-			adev->ip_versions[GC_HWIP][0]);
+			amdgpu_ip_version(adev, GC_HWIP, 0));
 		return -EINVAL;
 	}
 	return 0;
@@ -1538,7 +1998,7 @@ static int amdgpu_discovery_set_common_ip_blocks(struct amdgpu_device *adev)
 static int amdgpu_discovery_set_gmc_ip_blocks(struct amdgpu_device *adev)
 {
 	/* use GC or MMHUB IP version */
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 1, 0):
 	case IP_VERSION(9, 2, 1):
@@ -1548,6 +2008,8 @@ static int amdgpu_discovery_set_gmc_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(9, 4, 1):
 	case IP_VERSION(9, 4, 2):
 	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
 		amdgpu_device_ip_block_add(adev, &gmc_v9_0_ip_block);
 		break;
 	case IP_VERSION(10, 1, 10):
@@ -1570,12 +2032,19 @@ static int amdgpu_discovery_set_gmc_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 		amdgpu_device_ip_block_add(adev, &gmc_v11_0_ip_block);
 		break;
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		amdgpu_device_ip_block_add(adev, &gmc_v12_0_ip_block);
+		break;
 	default:
-		dev_err(adev->dev,
-			"Failed to add gmc ip block(GC_HWIP:0x%x)\n",
-			adev->ip_versions[GC_HWIP][0]);
+		dev_err(adev->dev, "Failed to add gmc ip block(GC_HWIP:0x%x)\n",
+			amdgpu_ip_version(adev, GC_HWIP, 0));
 		return -EINVAL;
 	}
 	return 0;
@@ -1583,7 +2052,7 @@ static int amdgpu_discovery_set_gmc_ip_blocks(struct amdgpu_device *adev)
 
 static int amdgpu_discovery_set_ih_ip_blocks(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[OSSSYS_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, OSSSYS_HWIP, 0)) {
 	case IP_VERSION(4, 0, 0):
 	case IP_VERSION(4, 0, 1):
 	case IP_VERSION(4, 1, 0):
@@ -1595,6 +2064,7 @@ static int amdgpu_discovery_set_ih_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(4, 2, 1):
 	case IP_VERSION(4, 4, 0):
 	case IP_VERSION(4, 4, 2):
+	case IP_VERSION(4, 4, 5):
 		amdgpu_device_ip_block_add(adev, &vega20_ih_ip_block);
 		break;
 	case IP_VERSION(5, 0, 0):
@@ -1610,10 +2080,16 @@ static int amdgpu_discovery_set_ih_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(6, 0, 2):
 		amdgpu_device_ip_block_add(adev, &ih_v6_0_ip_block);
 		break;
+	case IP_VERSION(6, 1, 0):
+		amdgpu_device_ip_block_add(adev, &ih_v6_1_ip_block);
+		break;
+	case IP_VERSION(7, 0, 0):
+		amdgpu_device_ip_block_add(adev, &ih_v7_0_ip_block);
+		break;
 	default:
 		dev_err(adev->dev,
 			"Failed to add ih ip block(OSSSYS_HWIP:0x%x)\n",
-			adev->ip_versions[OSSSYS_HWIP][0]);
+			amdgpu_ip_version(adev, OSSSYS_HWIP, 0));
 		return -EINVAL;
 	}
 	return 0;
@@ -1621,7 +2097,7 @@ static int amdgpu_discovery_set_ih_ip_blocks(struct amdgpu_device *adev)
 
 static int amdgpu_discovery_set_psp_ip_blocks(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[MP0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 	case IP_VERSION(9, 0, 0):
 		amdgpu_device_ip_block_add(adev, &psp_v3_1_ip_block);
 		break;
@@ -1639,6 +2115,7 @@ static int amdgpu_discovery_set_psp_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(11, 0, 12):
 	case IP_VERSION(11, 0, 13):
 	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 2):
 		amdgpu_device_ip_block_add(adev, &psp_v11_0_ip_block);
 		break;
 	case IP_VERSION(11, 0, 8):
@@ -1658,15 +2135,25 @@ static int amdgpu_discovery_set_psp_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(13, 0, 8):
 	case IP_VERSION(13, 0, 10):
 	case IP_VERSION(13, 0, 11):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
+	case IP_VERSION(14, 0, 0):
+	case IP_VERSION(14, 0, 1):
+	case IP_VERSION(14, 0, 4):
 		amdgpu_device_ip_block_add(adev, &psp_v13_0_ip_block);
 		break;
 	case IP_VERSION(13, 0, 4):
 		amdgpu_device_ip_block_add(adev, &psp_v13_0_4_ip_block);
 		break;
+	case IP_VERSION(14, 0, 2):
+	case IP_VERSION(14, 0, 3):
+	case IP_VERSION(14, 0, 5):
+		amdgpu_device_ip_block_add(adev, &psp_v14_0_ip_block);
+		break;
 	default:
 		dev_err(adev->dev,
 			"Failed to add psp ip block(MP0_HWIP:0x%x)\n",
-			adev->ip_versions[MP0_HWIP][0]);
+			amdgpu_ip_version(adev, MP0_HWIP, 0));
 		return -EINVAL;
 	}
 	return 0;
@@ -1674,7 +2161,7 @@ static int amdgpu_discovery_set_psp_ip_blocks(struct amdgpu_device *adev)
 
 static int amdgpu_discovery_set_smu_ip_blocks(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[MP1_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
 	case IP_VERSION(9, 0, 0):
 	case IP_VERSION(10, 0, 0):
 	case IP_VERSION(10, 0, 1):
@@ -1688,13 +2175,17 @@ static int amdgpu_discovery_set_smu_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(11, 0, 5):
 	case IP_VERSION(11, 0, 9):
 	case IP_VERSION(11, 0, 7):
-	case IP_VERSION(11, 0, 8):
 	case IP_VERSION(11, 0, 11):
 	case IP_VERSION(11, 0, 12):
 	case IP_VERSION(11, 0, 13):
 	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 2):
 		amdgpu_device_ip_block_add(adev, &smu_v11_0_ip_block);
 		break;
+	case IP_VERSION(11, 0, 8):
+		if (adev->apu_flags & AMD_APU_IS_CYAN_SKILLFISH2)
+			amdgpu_device_ip_block_add(adev, &smu_v11_0_ip_block);
+		break;
 	case IP_VERSION(12, 0, 0):
 	case IP_VERSION(12, 0, 1):
 		amdgpu_device_ip_block_add(adev, &smu_v12_0_ip_block);
@@ -1705,16 +2196,27 @@ static int amdgpu_discovery_set_smu_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(13, 0, 3):
 	case IP_VERSION(13, 0, 4):
 	case IP_VERSION(13, 0, 5):
+	case IP_VERSION(13, 0, 6):
 	case IP_VERSION(13, 0, 7):
 	case IP_VERSION(13, 0, 8):
 	case IP_VERSION(13, 0, 10):
 	case IP_VERSION(13, 0, 11):
+	case IP_VERSION(13, 0, 14):
+	case IP_VERSION(13, 0, 12):
 		amdgpu_device_ip_block_add(adev, &smu_v13_0_ip_block);
 		break;
+	case IP_VERSION(14, 0, 0):
+	case IP_VERSION(14, 0, 1):
+	case IP_VERSION(14, 0, 2):
+	case IP_VERSION(14, 0, 3):
+	case IP_VERSION(14, 0, 4):
+	case IP_VERSION(14, 0, 5):
+		amdgpu_device_ip_block_add(adev, &smu_v14_0_ip_block);
+		break;
 	default:
 		dev_err(adev->dev,
 			"Failed to add smu ip block(MP1_HWIP:0x%x)\n",
-			adev->ip_versions[MP1_HWIP][0]);
+			amdgpu_ip_version(adev, MP1_HWIP, 0));
 		return -EINVAL;
 	}
 	return 0;
@@ -1739,8 +2241,8 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev)
 		return 0;
 
 #if defined(CONFIG_DRM_AMD_DC)
-	if (adev->ip_versions[DCE_HWIP][0]) {
-		switch (adev->ip_versions[DCE_HWIP][0]) {
+	if (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
+		switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
 		case IP_VERSION(1, 0, 0):
 		case IP_VERSION(1, 0, 1):
 		case IP_VERSION(2, 0, 2):
@@ -1758,6 +2260,14 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev)
 		case IP_VERSION(3, 1, 6):
 		case IP_VERSION(3, 2, 0):
 		case IP_VERSION(3, 2, 1):
+		case IP_VERSION(3, 5, 0):
+		case IP_VERSION(3, 5, 1):
+		case IP_VERSION(3, 6, 0):
+		case IP_VERSION(4, 1, 0):
+			/* TODO: Fix IP version. DC code expects version 4.0.1 */
+			if (adev->ip_versions[DCE_HWIP][0] == IP_VERSION(4, 1, 0))
+				adev->ip_versions[DCE_HWIP][0] = IP_VERSION(4, 0, 1);
+
 			if (amdgpu_sriov_vf(adev))
 				amdgpu_discovery_set_sriov_display(adev);
 			else
@@ -1766,11 +2276,11 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev)
 		default:
 			dev_err(adev->dev,
 				"Failed to add dm ip block(DCE_HWIP:0x%x)\n",
-				adev->ip_versions[DCE_HWIP][0]);
+				amdgpu_ip_version(adev, DCE_HWIP, 0));
 			return -EINVAL;
 		}
-	} else if (adev->ip_versions[DCI_HWIP][0]) {
-		switch (adev->ip_versions[DCI_HWIP][0]) {
+	} else if (amdgpu_ip_version(adev, DCI_HWIP, 0)) {
+		switch (amdgpu_ip_version(adev, DCI_HWIP, 0)) {
 		case IP_VERSION(12, 0, 0):
 		case IP_VERSION(12, 0, 1):
 		case IP_VERSION(12, 1, 0):
@@ -1782,7 +2292,7 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev)
 		default:
 			dev_err(adev->dev,
 				"Failed to add dm ip block(DCI_HWIP:0x%x)\n",
-				adev->ip_versions[DCI_HWIP][0]);
+				amdgpu_ip_version(adev, DCI_HWIP, 0));
 			return -EINVAL;
 		}
 	}
@@ -1792,7 +2302,7 @@ static int amdgpu_discovery_set_display_ip_blocks(struct amdgpu_device *adev)
 
 static int amdgpu_discovery_set_gc_ip_blocks(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 1, 0):
 	case IP_VERSION(9, 2, 1):
@@ -1803,6 +2313,11 @@ static int amdgpu_discovery_set_gc_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(9, 4, 2):
 		amdgpu_device_ip_block_add(adev, &gfx_v9_0_ip_block);
 		break;
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
+		amdgpu_device_ip_block_add(adev, &gfx_v9_4_3_ip_block);
+		break;
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 2):
 	case IP_VERSION(10, 1, 1):
@@ -1823,12 +2338,19 @@ static int amdgpu_discovery_set_gc_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 		amdgpu_device_ip_block_add(adev, &gfx_v11_0_ip_block);
 		break;
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		amdgpu_device_ip_block_add(adev, &gfx_v12_0_ip_block);
+		break;
 	default:
-		dev_err(adev->dev,
-			"Failed to add gfx ip block(GC_HWIP:0x%x)\n",
-			adev->ip_versions[GC_HWIP][0]);
+		dev_err(adev->dev, "Failed to add gfx ip block(GC_HWIP:0x%x)\n",
+			amdgpu_ip_version(adev, GC_HWIP, 0));
 		return -EINVAL;
 	}
 	return 0;
@@ -1836,7 +2358,7 @@ static int amdgpu_discovery_set_gc_ip_blocks(struct amdgpu_device *adev)
 
 static int amdgpu_discovery_set_sdma_ip_blocks(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[SDMA0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
 	case IP_VERSION(4, 0, 0):
 	case IP_VERSION(4, 0, 1):
 	case IP_VERSION(4, 1, 0):
@@ -1848,6 +2370,8 @@ static int amdgpu_discovery_set_sdma_ip_blocks(struct amdgpu_device *adev)
 		amdgpu_device_ip_block_add(adev, &sdma_v4_0_ip_block);
 		break;
 	case IP_VERSION(4, 4, 2):
+	case IP_VERSION(4, 4, 5):
+	case IP_VERSION(4, 4, 4):
 		amdgpu_device_ip_block_add(adev, &sdma_v4_4_2_ip_block);
 		break;
 	case IP_VERSION(5, 0, 0):
@@ -1870,21 +2394,44 @@ static int amdgpu_discovery_set_sdma_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(6, 0, 1):
 	case IP_VERSION(6, 0, 2):
 	case IP_VERSION(6, 0, 3):
+	case IP_VERSION(6, 1, 0):
+	case IP_VERSION(6, 1, 1):
+	case IP_VERSION(6, 1, 2):
+	case IP_VERSION(6, 1, 3):
 		amdgpu_device_ip_block_add(adev, &sdma_v6_0_ip_block);
 		break;
+	case IP_VERSION(7, 0, 0):
+	case IP_VERSION(7, 0, 1):
+		amdgpu_device_ip_block_add(adev, &sdma_v7_0_ip_block);
+		break;
 	default:
 		dev_err(adev->dev,
 			"Failed to add sdma ip block(SDMA0_HWIP:0x%x)\n",
-			adev->ip_versions[SDMA0_HWIP][0]);
+			amdgpu_ip_version(adev, SDMA0_HWIP, 0));
 		return -EINVAL;
 	}
+
+	return 0;
+}
+
+static int amdgpu_discovery_set_ras_ip_blocks(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
+		amdgpu_device_ip_block_add(adev, &ras_v1_0_ip_block);
+		break;
+	default:
+		break;
+	}
 	return 0;
 }
 
 static int amdgpu_discovery_set_mm_ip_blocks(struct amdgpu_device *adev)
 {
-	if (adev->ip_versions[VCE_HWIP][0]) {
-		switch (adev->ip_versions[UVD_HWIP][0]) {
+	if (amdgpu_ip_version(adev, VCE_HWIP, 0)) {
+		switch (amdgpu_ip_version(adev, UVD_HWIP, 0)) {
 		case IP_VERSION(7, 0, 0):
 		case IP_VERSION(7, 2, 0):
 			/* UVD is not supported on vega20 SR-IOV */
@@ -1894,10 +2441,10 @@ static int amdgpu_discovery_set_mm_ip_blocks(struct amdgpu_device *adev)
 		default:
 			dev_err(adev->dev,
 				"Failed to add uvd v7 ip block(UVD_HWIP:0x%x)\n",
-				adev->ip_versions[UVD_HWIP][0]);
+				amdgpu_ip_version(adev, UVD_HWIP, 0));
 			return -EINVAL;
 		}
-		switch (adev->ip_versions[VCE_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, VCE_HWIP, 0)) {
 		case IP_VERSION(4, 0, 0):
 		case IP_VERSION(4, 1, 0):
 			/* VCE is not supported on vega20 SR-IOV */
@@ -1907,11 +2454,11 @@ static int amdgpu_discovery_set_mm_ip_blocks(struct amdgpu_device *adev)
 		default:
 			dev_err(adev->dev,
 				"Failed to add VCE v4 ip block(VCE_HWIP:0x%x)\n",
-				adev->ip_versions[VCE_HWIP][0]);
+				amdgpu_ip_version(adev, VCE_HWIP, 0));
 			return -EINVAL;
 		}
 	} else {
-		switch (adev->ip_versions[UVD_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, UVD_HWIP, 0)) {
 		case IP_VERSION(1, 0, 0):
 		case IP_VERSION(1, 0, 1):
 			amdgpu_device_ip_block_add(adev, &vcn_v1_0_ip_block);
@@ -1938,7 +2485,6 @@ static int amdgpu_discovery_set_mm_ip_blocks(struct amdgpu_device *adev)
 		case IP_VERSION(3, 1, 1):
 		case IP_VERSION(3, 1, 2):
 		case IP_VERSION(3, 0, 2):
-		case IP_VERSION(3, 0, 192):
 			amdgpu_device_ip_block_add(adev, &vcn_v3_0_ip_block);
 			if (!amdgpu_sriov_vf(adev))
 				amdgpu_device_ip_block_add(adev, &jpeg_v3_0_ip_block);
@@ -1951,11 +2497,28 @@ static int amdgpu_discovery_set_mm_ip_blocks(struct amdgpu_device *adev)
 		case IP_VERSION(4, 0, 4):
 			amdgpu_device_ip_block_add(adev, &vcn_v4_0_ip_block);
 			amdgpu_device_ip_block_add(adev, &jpeg_v4_0_ip_block);
-			return 0;
+			break;
+		case IP_VERSION(4, 0, 3):
+			amdgpu_device_ip_block_add(adev, &vcn_v4_0_3_ip_block);
+			amdgpu_device_ip_block_add(adev, &jpeg_v4_0_3_ip_block);
+			break;
+		case IP_VERSION(4, 0, 5):
+		case IP_VERSION(4, 0, 6):
+			amdgpu_device_ip_block_add(adev, &vcn_v4_0_5_ip_block);
+			amdgpu_device_ip_block_add(adev, &jpeg_v4_0_5_ip_block);
+			break;
+		case IP_VERSION(5, 0, 0):
+			amdgpu_device_ip_block_add(adev, &vcn_v5_0_0_ip_block);
+			amdgpu_device_ip_block_add(adev, &jpeg_v5_0_0_ip_block);
+			break;
+		case IP_VERSION(5, 0, 1):
+			amdgpu_device_ip_block_add(adev, &vcn_v5_0_1_ip_block);
+			amdgpu_device_ip_block_add(adev, &jpeg_v5_0_1_ip_block);
+			break;
 		default:
 			dev_err(adev->dev,
 				"Failed to add vcn/jpeg ip block(UVD_HWIP:0x%x)\n",
-				adev->ip_versions[UVD_HWIP][0]);
+				amdgpu_ip_version(adev, UVD_HWIP, 0));
 			return -EINVAL;
 		}
 	}
@@ -1964,38 +2527,94 @@ static int amdgpu_discovery_set_mm_ip_blocks(struct amdgpu_device *adev)
 
 static int amdgpu_discovery_set_mes_ip_blocks(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
-	case IP_VERSION(10, 1, 10):
-	case IP_VERSION(10, 1, 1):
-	case IP_VERSION(10, 1, 2):
-	case IP_VERSION(10, 1, 3):
-	case IP_VERSION(10, 1, 4):
-	case IP_VERSION(10, 3, 0):
-	case IP_VERSION(10, 3, 1):
-	case IP_VERSION(10, 3, 2):
-	case IP_VERSION(10, 3, 3):
-	case IP_VERSION(10, 3, 4):
-	case IP_VERSION(10, 3, 5):
-	case IP_VERSION(10, 3, 6):
-		if (amdgpu_mes) {
-			amdgpu_device_ip_block_add(adev, &mes_v10_1_ip_block);
-			adev->enable_mes = true;
-			if (amdgpu_mes_kiq)
-				adev->enable_mes_kiq = true;
-		}
-		break;
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
 	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 		amdgpu_device_ip_block_add(adev, &mes_v11_0_ip_block);
 		adev->enable_mes = true;
 		adev->enable_mes_kiq = true;
 		break;
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		amdgpu_device_ip_block_add(adev, &mes_v12_0_ip_block);
+		adev->enable_mes = true;
+		adev->enable_mes_kiq = true;
+		if (amdgpu_uni_mes)
+			adev->enable_uni_mes = true;
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
+static void amdgpu_discovery_init_soc_config(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
+		aqua_vanjaram_init_soc_config(adev);
+		break;
+	default:
+		break;
+	}
+}
+
+static int amdgpu_discovery_set_vpe_ip_blocks(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, VPE_HWIP, 0)) {
+	case IP_VERSION(6, 1, 0):
+	case IP_VERSION(6, 1, 1):
+	case IP_VERSION(6, 1, 3):
+		amdgpu_device_ip_block_add(adev, &vpe_v6_1_ip_block);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int amdgpu_discovery_set_umsch_mm_ip_blocks(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, VCN_HWIP, 0)) {
+	case IP_VERSION(4, 0, 5):
+	case IP_VERSION(4, 0, 6):
+		if (amdgpu_umsch_mm & 0x1) {
+			amdgpu_device_ip_block_add(adev, &umsch_mm_v4_0_ip_block);
+			adev->enable_umsch_mm = true;
+		}
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int amdgpu_discovery_set_isp_ip_blocks(struct amdgpu_device *adev)
+{
+#if defined(CONFIG_DRM_AMD_ISP)
+	switch (amdgpu_ip_version(adev, ISP_HWIP, 0)) {
+	case IP_VERSION(4, 1, 0):
+		amdgpu_device_ip_block_add(adev, &isp_v4_1_0_ip_block);
+		break;
+	case IP_VERSION(4, 1, 1):
+		amdgpu_device_ip_block_add(adev, &isp_v4_1_1_ip_block);
+		break;
 	default:
 		break;
 	}
+#endif
+
 	return 0;
 }
 
@@ -2005,9 +2624,16 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 
 	switch (adev->asic_type) {
 	case CHIP_VEGA10:
+		/* This is not fatal.  We only need the discovery
+		 * binary for sysfs.  We don't need it for a
+		 * functional system.
+		 */
+		amdgpu_discovery_init(adev);
 		vega10_reg_base_init(adev);
 		adev->sdma.num_instances = 2;
+		adev->sdma.sdma_mask = 3;
 		adev->gmc.num_umc = 4;
+		adev->gfx.xcc_mask = 1;
 		adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 0, 0);
 		adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 0, 0);
 		adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(4, 0, 0);
@@ -2027,9 +2653,16 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		adev->ip_versions[DCI_HWIP][0] = IP_VERSION(12, 0, 0);
 		break;
 	case CHIP_VEGA12:
+		/* This is not fatal.  We only need the discovery
+		 * binary for sysfs.  We don't need it for a
+		 * functional system.
+		 */
+		amdgpu_discovery_init(adev);
 		vega10_reg_base_init(adev);
 		adev->sdma.num_instances = 2;
+		adev->sdma.sdma_mask = 3;
 		adev->gmc.num_umc = 4;
+		adev->gfx.xcc_mask = 1;
 		adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 3, 0);
 		adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 3, 0);
 		adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(4, 0, 1);
@@ -2049,10 +2682,17 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		adev->ip_versions[DCI_HWIP][0] = IP_VERSION(12, 0, 1);
 		break;
 	case CHIP_RAVEN:
+		/* This is not fatal.  We only need the discovery
+		 * binary for sysfs.  We don't need it for a
+		 * functional system.
+		 */
+		amdgpu_discovery_init(adev);
 		vega10_reg_base_init(adev);
 		adev->sdma.num_instances = 1;
+		adev->sdma.sdma_mask = 1;
 		adev->vcn.num_vcn_inst = 1;
 		adev->gmc.num_umc = 2;
+		adev->gfx.xcc_mask = 1;
 		if (adev->apu_flags & AMD_APU_IS_RAVEN2) {
 			adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 2, 0);
 			adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 2, 0);
@@ -2069,6 +2709,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 			adev->ip_versions[GC_HWIP][0] = IP_VERSION(9, 2, 2);
 			adev->ip_versions[UVD_HWIP][0] = IP_VERSION(1, 0, 1);
 			adev->ip_versions[DCE_HWIP][0] = IP_VERSION(1, 0, 1);
+			adev->ip_versions[ISP_HWIP][0] = IP_VERSION(2, 0, 0);
 		} else {
 			adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 1, 0);
 			adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 1, 0);
@@ -2085,12 +2726,20 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 			adev->ip_versions[GC_HWIP][0] = IP_VERSION(9, 1, 0);
 			adev->ip_versions[UVD_HWIP][0] = IP_VERSION(1, 0, 0);
 			adev->ip_versions[DCE_HWIP][0] = IP_VERSION(1, 0, 0);
+			adev->ip_versions[ISP_HWIP][0] = IP_VERSION(2, 0, 0);
 		}
 		break;
 	case CHIP_VEGA20:
+		/* This is not fatal.  We only need the discovery
+		 * binary for sysfs.  We don't need it for a
+		 * functional system.
+		 */
+		amdgpu_discovery_init(adev);
 		vega20_reg_base_init(adev);
 		adev->sdma.num_instances = 2;
+		adev->sdma.sdma_mask = 3;
 		adev->gmc.num_umc = 8;
+		adev->gfx.xcc_mask = 1;
 		adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 4, 0);
 		adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 4, 0);
 		adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(4, 2, 0);
@@ -2111,10 +2760,17 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		adev->ip_versions[DCI_HWIP][0] = IP_VERSION(12, 1, 0);
 		break;
 	case CHIP_ARCTURUS:
+		/* This is not fatal.  We only need the discovery
+		 * binary for sysfs.  We don't need it for a
+		 * functional system.
+		 */
+		amdgpu_discovery_init(adev);
 		arct_reg_base_init(adev);
 		adev->sdma.num_instances = 8;
+		adev->sdma.sdma_mask = 0xff;
 		adev->vcn.num_vcn_inst = 2;
 		adev->gmc.num_umc = 8;
+		adev->gfx.xcc_mask = 1;
 		adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 4, 1);
 		adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 4, 1);
 		adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(4, 2, 1);
@@ -2139,10 +2795,17 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		adev->ip_versions[UVD_HWIP][1] = IP_VERSION(2, 5, 0);
 		break;
 	case CHIP_ALDEBARAN:
+		/* This is not fatal.  We only need the discovery
+		 * binary for sysfs.  We don't need it for a
+		 * functional system.
+		 */
+		amdgpu_discovery_init(adev);
 		aldebaran_reg_base_init(adev);
 		adev->sdma.num_instances = 5;
+		adev->sdma.sdma_mask = 0x1f;
 		adev->vcn.num_vcn_inst = 2;
 		adev->gmc.num_umc = 4;
+		adev->gfx.xcc_mask = 1;
 		adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 4, 2);
 		adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 4, 2);
 		adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(4, 4, 0);
@@ -2164,10 +2827,44 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		adev->ip_versions[UVD_HWIP][1] = IP_VERSION(2, 6, 0);
 		adev->ip_versions[XGMI_HWIP][0] = IP_VERSION(6, 1, 0);
 		break;
+	case CHIP_CYAN_SKILLFISH:
+		if (adev->apu_flags & AMD_APU_IS_CYAN_SKILLFISH2) {
+			r = amdgpu_discovery_reg_base_init(adev);
+			if (r)
+				return -EINVAL;
+
+			amdgpu_discovery_harvest_ip(adev);
+			amdgpu_discovery_get_gfx_info(adev);
+			amdgpu_discovery_get_mall_info(adev);
+			amdgpu_discovery_get_vcn_info(adev);
+		} else {
+			cyan_skillfish_reg_base_init(adev);
+			adev->sdma.num_instances = 2;
+			adev->sdma.sdma_mask = 3;
+			adev->gfx.xcc_mask = 1;
+			adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(2, 0, 3);
+			adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(2, 0, 3);
+			adev->ip_versions[OSSSYS_HWIP][0] = IP_VERSION(5, 0, 1);
+			adev->ip_versions[HDP_HWIP][0] = IP_VERSION(5, 0, 1);
+			adev->ip_versions[SDMA0_HWIP][0] = IP_VERSION(5, 0, 1);
+			adev->ip_versions[SDMA1_HWIP][1] = IP_VERSION(5, 0, 1);
+			adev->ip_versions[DF_HWIP][0] = IP_VERSION(3, 5, 0);
+			adev->ip_versions[NBIO_HWIP][0] = IP_VERSION(2, 1, 1);
+			adev->ip_versions[UMC_HWIP][0] = IP_VERSION(8, 1, 1);
+			adev->ip_versions[MP0_HWIP][0] = IP_VERSION(11, 0, 8);
+			adev->ip_versions[MP1_HWIP][0] = IP_VERSION(11, 0, 8);
+			adev->ip_versions[THM_HWIP][0] = IP_VERSION(11, 0, 1);
+			adev->ip_versions[SMUIO_HWIP][0] = IP_VERSION(11, 0, 8);
+			adev->ip_versions[GC_HWIP][0] = IP_VERSION(10, 1, 3);
+			adev->ip_versions[UVD_HWIP][0] = IP_VERSION(2, 0, 3);
+		}
+		break;
 	default:
 		r = amdgpu_discovery_reg_base_init(adev);
-		if (r)
-			return -EINVAL;
+		if (r) {
+			drm_err(&adev->ddev, "discovery failed: %d\n", r);
+			return r;
+		}
 
 		amdgpu_discovery_harvest_ip(adev);
 		amdgpu_discovery_get_gfx_info(adev);
@@ -2176,13 +2873,18 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		break;
 	}
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	amdgpu_discovery_init_soc_config(adev);
+	amdgpu_discovery_sysfs_init(adev);
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 2, 1):
 	case IP_VERSION(9, 4, 0):
 	case IP_VERSION(9, 4, 1):
 	case IP_VERSION(9, 4, 2):
 	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
 		adev->family = AMDGPU_FAMILY_AI;
 		break;
 	case IP_VERSION(9, 1, 0):
@@ -2223,11 +2925,21 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(11, 0, 4):
 		adev->family = AMDGPU_FAMILY_GC_11_0_1;
 		break;
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
+		adev->family = AMDGPU_FAMILY_GC_11_5_0;
+		break;
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		adev->family = AMDGPU_FAMILY_GC_12_0_0;
+		break;
 	default:
 		return -EINVAL;
 	}
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 1, 0):
 	case IP_VERSION(9, 2, 2):
 	case IP_VERSION(9, 3, 0):
@@ -2239,17 +2951,18 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(10, 3, 7):
 	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 		adev->flags |= AMD_IS_APU;
 		break;
 	default:
 		break;
 	}
 
-	if (adev->ip_versions[XGMI_HWIP][0] == IP_VERSION(4, 8, 0))
-		adev->gmc.xgmi.supported = true;
-
 	/* set NBIO version */
-	switch (adev->ip_versions[NBIO_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
 	case IP_VERSION(6, 1, 0):
 	case IP_VERSION(6, 2, 0):
 		adev->nbio.funcs = &nbio_v6_1_funcs;
@@ -2268,9 +2981,17 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		adev->nbio.hdp_flush_reg = &nbio_v7_4_hdp_flush_reg;
 		break;
 	case IP_VERSION(7, 9, 0):
+	case IP_VERSION(7, 9, 1):
 		adev->nbio.funcs = &nbio_v7_9_funcs;
 		adev->nbio.hdp_flush_reg = &nbio_v7_9_hdp_flush_reg;
 		break;
+	case IP_VERSION(7, 11, 0):
+	case IP_VERSION(7, 11, 1):
+	case IP_VERSION(7, 11, 2):
+	case IP_VERSION(7, 11, 3):
+		adev->nbio.funcs = &nbio_v7_11_funcs;
+		adev->nbio.hdp_flush_reg = &nbio_v7_11_hdp_flush_reg;
+		break;
 	case IP_VERSION(7, 2, 0):
 	case IP_VERSION(7, 2, 1):
 	case IP_VERSION(7, 3, 0):
@@ -2303,11 +3024,15 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		adev->nbio.funcs = &nbio_v7_7_funcs;
 		adev->nbio.hdp_flush_reg = &nbio_v7_7_hdp_flush_reg;
 		break;
+	case IP_VERSION(6, 3, 1):
+		adev->nbio.funcs = &nbif_v6_3_1_funcs;
+		adev->nbio.hdp_flush_reg = &nbif_v6_3_1_hdp_flush_reg;
+		break;
 	default:
 		break;
 	}
 
-	switch (adev->ip_versions[HDP_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, HDP_HWIP, 0)) {
 	case IP_VERSION(4, 0, 0):
 	case IP_VERSION(4, 0, 1):
 	case IP_VERSION(4, 1, 0):
@@ -2317,6 +3042,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(4, 2, 1):
 	case IP_VERSION(4, 4, 0):
 	case IP_VERSION(4, 4, 2):
+	case IP_VERSION(4, 4, 5):
 		adev->hdp.funcs = &hdp_v4_0_funcs;
 		break;
 	case IP_VERSION(5, 0, 0):
@@ -2332,13 +3058,17 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 		break;
 	case IP_VERSION(6, 0, 0):
 	case IP_VERSION(6, 0, 1):
+	case IP_VERSION(6, 1, 0):
 		adev->hdp.funcs = &hdp_v6_0_funcs;
 		break;
+	case IP_VERSION(7, 0, 0):
+		adev->hdp.funcs = &hdp_v7_0_funcs;
+		break;
 	default:
 		break;
 	}
 
-	switch (adev->ip_versions[DF_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, DF_HWIP, 0)) {
 	case IP_VERSION(3, 6, 0):
 	case IP_VERSION(3, 6, 1):
 	case IP_VERSION(3, 6, 2):
@@ -2354,11 +3084,18 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(4, 3, 0):
 		adev->df.funcs = &df_v4_3_funcs;
 		break;
+	case IP_VERSION(4, 6, 2):
+		adev->df.funcs = &df_v4_6_2_funcs;
+		break;
+	case IP_VERSION(4, 15, 0):
+	case IP_VERSION(4, 15, 1):
+		adev->df.funcs = &df_v4_15_funcs;
+		break;
 	default:
 		break;
 	}
 
-	switch (adev->ip_versions[SMUIO_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, SMUIO_HWIP, 0)) {
 	case IP_VERSION(9, 0, 0):
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(10, 0, 0):
@@ -2378,6 +3115,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(11, 0, 10):
 	case IP_VERSION(11, 0, 11):
 	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 2):
 	case IP_VERSION(13, 0, 1):
 	case IP_VERSION(13, 0, 9):
 	case IP_VERSION(13, 0, 10):
@@ -2386,21 +3124,37 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 	case IP_VERSION(13, 0, 2):
 		adev->smuio.funcs = &smuio_v13_0_funcs;
 		break;
+	case IP_VERSION(13, 0, 3):
+	case IP_VERSION(13, 0, 11):
+		adev->smuio.funcs = &smuio_v13_0_3_funcs;
+		if (adev->smuio.funcs->get_pkg_type(adev) == AMDGPU_PKG_TYPE_APU) {
+			adev->flags |= AMD_IS_APU;
+		}
+		break;
 	case IP_VERSION(13, 0, 6):
 	case IP_VERSION(13, 0, 8):
+	case IP_VERSION(14, 0, 0):
+	case IP_VERSION(14, 0, 1):
 		adev->smuio.funcs = &smuio_v13_0_6_funcs;
 		break;
+	case IP_VERSION(14, 0, 2):
+		adev->smuio.funcs = &smuio_v14_0_2_funcs;
+		break;
 	default:
 		break;
 	}
 
-	switch (adev->ip_versions[LSDMA_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, LSDMA_HWIP, 0)) {
 	case IP_VERSION(6, 0, 0):
 	case IP_VERSION(6, 0, 1):
 	case IP_VERSION(6, 0, 2):
 	case IP_VERSION(6, 0, 3):
 		adev->lsdma.funcs = &lsdma_v6_0_funcs;
 		break;
+	case IP_VERSION(7, 0, 0):
+	case IP_VERSION(7, 0, 1):
+		adev->lsdma.funcs = &lsdma_v7_0_funcs;
+		break;
 	default:
 		break;
 	}
@@ -2451,6 +3205,10 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 	if (r)
 		return r;
 
+	r = amdgpu_discovery_set_ras_ip_blocks(adev);
+	if (r)
+		return r;
+
 	if ((adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT &&
 	     !amdgpu_sriov_vf(adev)) ||
 	    (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO && amdgpu_dpm == 1)) {
@@ -2467,6 +3225,17 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev)
 	if (r)
 		return r;
 
+	r = amdgpu_discovery_set_vpe_ip_blocks(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_discovery_set_umsch_mm_ip_blocks(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_discovery_set_isp_ip_blocks(adev);
+	if (r)
+		return r;
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h
index 8563dd4a7dc2..4ce04486cc31 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.h
@@ -24,12 +24,27 @@
 #ifndef __AMDGPU_DISCOVERY__
 #define __AMDGPU_DISCOVERY__
 
-#define DISCOVERY_TMR_SIZE      (4 << 10)
+#include <linux/debugfs.h>
+
+#define DISCOVERY_TMR_SIZE      (10 << 10)
 #define DISCOVERY_TMR_OFFSET    (64 << 10)
 
+struct ip_discovery_top;
+
+struct amdgpu_discovery_info {
+	struct debugfs_blob_wrapper debugfs_blob;
+	struct ip_discovery_top *ip_top;
+	uint32_t size;
+	uint8_t *bin;
+	bool reserve_tmr;
+};
+
 void amdgpu_discovery_fini(struct amdgpu_device *adev);
-int amdgpu_discovery_get_ip_version(struct amdgpu_device *adev, int hw_id, int number_instance,
-                                    int *major, int *minor, int *revision);
 int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev);
 
+int amdgpu_discovery_get_nps_info(struct amdgpu_device *adev,
+				  uint32_t *nps_type,
+				  struct amdgpu_gmc_memrange **ranges,
+				  int *range_cnt, bool refresh);
+
 #endif /* __AMDGPU_DISCOVERY__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
index d60fe7eb5579..b5d34797d606 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
@@ -33,11 +33,14 @@
 #include "soc15_common.h"
 #include "gc/gc_11_0_0_offset.h"
 #include "gc/gc_11_0_0_sh_mask.h"
+#include "bif/bif_4_1_d.h"
 #include <asm/div64.h>
 
 #include <linux/pci.h>
 #include <linux/pm_runtime.h>
 #include <drm/drm_crtc_helper.h>
+#include <drm/drm_damage_helper.h>
+#include <drm/drm_drv.h>
 #include <drm/drm_edid.h>
 #include <drm/drm_fb_helper.h>
 #include <drm/drm_gem_framebuffer_helper.h>
@@ -98,7 +101,7 @@ static void amdgpu_display_flip_callback(struct dma_fence *f,
 static bool amdgpu_display_flip_handle_fence(struct amdgpu_flip_work *work,
 					     struct dma_fence **f)
 {
-	struct dma_fence *fence= *f;
+	struct dma_fence *fence = *f;
 
 	if (fence == NULL)
 		return false;
@@ -124,7 +127,7 @@ static void amdgpu_display_flip_work_func(struct work_struct *__work)
 
 	struct drm_crtc *crtc = &amdgpu_crtc->base;
 	unsigned long flags;
-	unsigned i;
+	unsigned int i;
 	int vpos, hpos;
 
 	for (i = 0; i < work->shared_count; ++i)
@@ -201,7 +204,7 @@ int amdgpu_display_crtc_page_flip_target(struct drm_crtc *crtc,
 	u64 tiling_flags;
 	int i, r;
 
-	work = kzalloc(sizeof *work, GFP_KERNEL);
+	work = kzalloc(sizeof(*work), GFP_KERNEL);
 	if (work == NULL)
 		return -ENOMEM;
 
@@ -231,6 +234,7 @@ int amdgpu_display_crtc_page_flip_target(struct drm_crtc *crtc,
 	}
 
 	if (!adev->enable_virtual_display) {
+		new_abo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 		r = amdgpu_bo_pin(new_abo,
 				  amdgpu_display_supported_domains(adev, new_abo->flags));
 		if (unlikely(r != 0)) {
@@ -328,22 +332,19 @@ int amdgpu_display_crtc_set_config(struct drm_mode_set *set,
 		if (crtc->enabled)
 			active = true;
 
-	pm_runtime_mark_last_busy(dev->dev);
-
 	adev = drm_to_adev(dev);
 	/* if we have active crtcs and we don't have a power ref,
-	   take the current one */
+	 * take the current one
+	 */
 	if (active && !adev->have_disp_power_ref) {
 		adev->have_disp_power_ref = true;
 		return ret;
 	}
-	/* if we have no active crtcs, then drop the power ref
-	   we got before */
-	if (!active && adev->have_disp_power_ref) {
-		pm_runtime_put_autosuspend(dev->dev);
+	/* if we have no active crtcs, then go to
+	 * drop the power ref we got before
+	 */
+	if (!active && adev->have_disp_power_ref)
 		adev->have_disp_power_ref = false;
-	}
-
 out:
 	/* drop the power reference we got coming in here */
 	pm_runtime_put_autosuspend(dev->dev);
@@ -507,11 +508,10 @@ bool amdgpu_display_ddc_probe(struct amdgpu_connector *amdgpu_connector,
 	if (amdgpu_connector->router.ddc_valid)
 		amdgpu_i2c_router_select_ddc_port(amdgpu_connector);
 
-	if (use_aux) {
+	if (use_aux)
 		ret = i2c_transfer(&amdgpu_connector->ddc_bus->aux.ddc, msgs, 2);
-	} else {
+	else
 		ret = i2c_transfer(&amdgpu_connector->ddc_bus->adapter, msgs, 2);
-	}
 
 	if (ret != 2)
 		/* Couldn't find an accessible DDC on this connector */
@@ -520,20 +520,40 @@ bool amdgpu_display_ddc_probe(struct amdgpu_connector *amdgpu_connector,
 	 * EDID header starts with:
 	 * 0x00,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0x00.
 	 * Only the first 6 bytes must be valid as
-	 * drm_edid_block_valid() can fix the last 2 bytes */
+	 * drm_edid_block_valid() can fix the last 2 bytes
+	 */
 	if (drm_edid_header_is_valid(buf) < 6) {
 		/* Couldn't find an accessible EDID on this
-		 * connector */
+		 * connector
+		 */
 		return false;
 	}
 	return true;
 }
 
+static int amdgpu_dirtyfb(struct drm_framebuffer *fb, struct drm_file *file,
+			  unsigned int flags, unsigned int color,
+			  struct drm_clip_rect *clips, unsigned int num_clips)
+{
+
+	if (file)
+		return -ENOSYS;
+
+	return drm_atomic_helper_dirtyfb(fb, file, flags, color, clips,
+					 num_clips);
+}
+
 static const struct drm_framebuffer_funcs amdgpu_fb_funcs = {
 	.destroy = drm_gem_fb_destroy,
 	.create_handle = drm_gem_fb_create_handle,
 };
 
+static const struct drm_framebuffer_funcs amdgpu_fb_funcs_atomic = {
+	.destroy = drm_gem_fb_destroy,
+	.create_handle = drm_gem_fb_create_handle,
+	.dirty = amdgpu_dirtyfb
+};
+
 uint32_t amdgpu_display_supported_domains(struct amdgpu_device *adev,
 					  uint64_t bo_flags)
 {
@@ -634,6 +654,10 @@ amdgpu_lookup_format_info(u32 format, uint64_t modifier)
 	if (!IS_AMD_FMT_MOD(modifier))
 		return NULL;
 
+	if (AMD_FMT_MOD_GET(TILE_VERSION, modifier) < AMD_FMT_MOD_TILE_VER_GFX9 ||
+	    AMD_FMT_MOD_GET(TILE_VERSION, modifier) >= AMD_FMT_MOD_TILE_VER_GFX12)
+		return NULL;
+
 	if (AMD_FMT_MOD_GET(DCC_RETILE, modifier))
 		return lookup_format_info(dcc_retile_formats,
 					  ARRAY_SIZE(dcc_retile_formats),
@@ -698,6 +722,30 @@ extract_render_dcc_offset(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int convert_tiling_flags_to_modifier_gfx12(struct amdgpu_framebuffer *afb)
+{
+	u64 modifier = 0;
+	int swizzle_mode = AMDGPU_TILING_GET(afb->tiling_flags, GFX12_SWIZZLE_MODE);
+
+	if (!swizzle_mode) {
+		modifier = DRM_FORMAT_MOD_LINEAR;
+	} else {
+		int max_comp_block =
+			AMDGPU_TILING_GET(afb->tiling_flags, GFX12_DCC_MAX_COMPRESSED_BLOCK);
+
+		modifier =
+			AMD_FMT_MOD |
+			AMD_FMT_MOD_SET(TILE_VERSION, AMD_FMT_MOD_TILE_VER_GFX12) |
+			AMD_FMT_MOD_SET(TILE, swizzle_mode) |
+			AMD_FMT_MOD_SET(DCC, afb->gfx12_dcc) |
+			AMD_FMT_MOD_SET(DCC_MAX_COMPRESSED_BLOCK, max_comp_block);
+	}
+
+	afb->base.modifier = modifier;
+	afb->base.flags |= DRM_MODE_FB_MODIFIERS;
+	return 0;
+}
+
 static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb)
 {
 	struct amdgpu_device *adev = drm_to_adev(afb->base.dev);
@@ -743,11 +791,13 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb)
 			return -EINVAL;
 		}
 
-		if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(11, 0, 0))
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(11, 0, 0))
 			version = AMD_FMT_MOD_TILE_VER_GFX11;
-		else if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0))
+		else if (amdgpu_ip_version(adev, GC_HWIP, 0) >=
+			 IP_VERSION(10, 3, 0))
 			version = AMD_FMT_MOD_TILE_VER_GFX10_RBPLUS;
-		else if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 0, 0))
+		else if (amdgpu_ip_version(adev, GC_HWIP, 0) >=
+			 IP_VERSION(10, 0, 0))
 			version = AMD_FMT_MOD_TILE_VER_GFX10;
 		else
 			version = AMD_FMT_MOD_TILE_VER_GFX9;
@@ -756,13 +806,15 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb)
 		case 0: /* Z microtiling */
 			return -EINVAL;
 		case 1: /* S microtiling */
-			if (adev->ip_versions[GC_HWIP][0] < IP_VERSION(11, 0, 0)) {
+			if (amdgpu_ip_version(adev, GC_HWIP, 0) <
+			    IP_VERSION(11, 0, 0)) {
 				if (!has_xor)
 					version = AMD_FMT_MOD_TILE_VER_GFX9;
 			}
 			break;
 		case 2:
-			if (adev->ip_versions[GC_HWIP][0] < IP_VERSION(11, 0, 0)) {
+			if (amdgpu_ip_version(adev, GC_HWIP, 0) <
+			    IP_VERSION(11, 0, 0)) {
 				if (!has_xor && afb->base.format->cpp[0] != 4)
 					version = AMD_FMT_MOD_TILE_VER_GFX9;
 			}
@@ -815,10 +867,12 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb)
 			u64 render_dcc_offset;
 
 			/* Enable constant encode on RAVEN2 and later. */
-			bool dcc_constant_encode = (adev->asic_type > CHIP_RAVEN ||
-						   (adev->asic_type == CHIP_RAVEN &&
-						    adev->external_rev_id >= 0x81)) &&
-						    adev->ip_versions[GC_HWIP][0] < IP_VERSION(11, 0, 0);
+			bool dcc_constant_encode =
+				(adev->asic_type > CHIP_RAVEN ||
+				 (adev->asic_type == CHIP_RAVEN &&
+				  adev->external_rev_id >= 0x81)) &&
+				amdgpu_ip_version(adev, GC_HWIP, 0) <
+					IP_VERSION(11, 0, 0);
 
 			int max_cblock_size = dcc_i64b ? AMD_FMT_MOD_DCC_BLOCK_64B :
 					      dcc_i128b ? AMD_FMT_MOD_DCC_BLOCK_128B :
@@ -855,7 +909,9 @@ static int convert_tiling_flags_to_modifier(struct amdgpu_framebuffer *afb)
 				if (adev->family >= AMDGPU_FAMILY_NV) {
 					int extra_pipe = 0;
 
-					if ((adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0)) &&
+					if ((amdgpu_ip_version(adev, GC_HWIP,
+							       0) >=
+					     IP_VERSION(10, 3, 0)) &&
 					    pipes == packers && pipes > 1)
 						extra_pipe = 1;
 
@@ -889,8 +945,7 @@ static int check_tiling_flags_gfx6(struct amdgpu_framebuffer *afb)
 {
 	u64 micro_tile_mode;
 
-	/* Zero swizzle mode means linear */
-	if (AMDGPU_TILING_GET(afb->tiling_flags, SWIZZLE_MODE) == 0)
+	if (AMDGPU_TILING_GET(afb->tiling_flags, ARRAY_MODE) == 1) /* LINEAR_ALIGNED */
 		return 0;
 
 	micro_tile_mode = AMDGPU_TILING_GET(afb->tiling_flags, MICRO_TILE_MODE);
@@ -1014,6 +1069,30 @@ static int amdgpu_display_verify_sizes(struct amdgpu_framebuffer *rfb)
 			block_width = 256 / format_info->cpp[i];
 			block_height = 1;
 			block_size_log2 = 8;
+		} else if (AMD_FMT_MOD_GET(TILE_VERSION, modifier) >= AMD_FMT_MOD_TILE_VER_GFX12) {
+			int swizzle = AMD_FMT_MOD_GET(TILE, modifier);
+
+			switch (swizzle) {
+			case AMD_FMT_MOD_TILE_GFX12_256B_2D:
+				block_size_log2 = 8;
+				break;
+			case AMD_FMT_MOD_TILE_GFX12_4K_2D:
+				block_size_log2 = 12;
+				break;
+			case AMD_FMT_MOD_TILE_GFX12_64K_2D:
+				block_size_log2 = 16;
+				break;
+			case AMD_FMT_MOD_TILE_GFX12_256K_2D:
+				block_size_log2 = 18;
+				break;
+			default:
+				drm_dbg_kms(rfb->base.dev,
+					    "Gfx12 swizzle mode with unknown block size: %d\n", swizzle);
+				return -EINVAL;
+			}
+
+			get_block_dimensions(block_size_log2, format_info->cpp[i],
+					     &block_width, &block_height);
 		} else {
 			int swizzle = AMD_FMT_MOD_GET(TILE, modifier);
 
@@ -1049,7 +1128,8 @@ static int amdgpu_display_verify_sizes(struct amdgpu_framebuffer *rfb)
 			return ret;
 	}
 
-	if (AMD_FMT_MOD_GET(DCC, modifier)) {
+	if (AMD_FMT_MOD_GET(TILE_VERSION, modifier) <= AMD_FMT_MOD_TILE_VER_GFX11 &&
+	    AMD_FMT_MOD_GET(DCC, modifier)) {
 		if (AMD_FMT_MOD_GET(DCC_RETILE, modifier)) {
 			block_size_log2 = get_dcc_block_size(modifier, false, false);
 			get_block_dimensions(block_size_log2 + 8, format_info->cpp[0],
@@ -1079,7 +1159,8 @@ static int amdgpu_display_verify_sizes(struct amdgpu_framebuffer *rfb)
 }
 
 static int amdgpu_display_get_fb_info(const struct amdgpu_framebuffer *amdgpu_fb,
-				      uint64_t *tiling_flags, bool *tmz_surface)
+				      uint64_t *tiling_flags, bool *tmz_surface,
+				      bool *gfx12_dcc)
 {
 	struct amdgpu_bo *rbo;
 	int r;
@@ -1087,6 +1168,7 @@ static int amdgpu_display_get_fb_info(const struct amdgpu_framebuffer *amdgpu_fb
 	if (!amdgpu_fb) {
 		*tiling_flags = 0;
 		*tmz_surface = false;
+		*gfx12_dcc = false;
 		return 0;
 	}
 
@@ -1100,11 +1182,9 @@ static int amdgpu_display_get_fb_info(const struct amdgpu_framebuffer *amdgpu_fb
 		return r;
 	}
 
-	if (tiling_flags)
-		amdgpu_bo_get_tiling_flags(rbo, tiling_flags);
-
-	if (tmz_surface)
-		*tmz_surface = amdgpu_bo_encrypted(rbo);
+	amdgpu_bo_get_tiling_flags(rbo, tiling_flags);
+	*tmz_surface = amdgpu_bo_encrypted(rbo);
+	*gfx12_dcc = rbo->flags & AMDGPU_GEM_CREATE_GFX12_DCC;
 
 	amdgpu_bo_unreserve(rbo);
 
@@ -1114,13 +1194,14 @@ static int amdgpu_display_get_fb_info(const struct amdgpu_framebuffer *amdgpu_fb
 static int amdgpu_display_gem_fb_verify_and_init(struct drm_device *dev,
 						 struct amdgpu_framebuffer *rfb,
 						 struct drm_file *file_priv,
+						 const struct drm_format_info *info,
 						 const struct drm_mode_fb_cmd2 *mode_cmd,
 						 struct drm_gem_object *obj)
 {
 	int ret;
 
 	rfb->base.obj[0] = obj;
-	drm_helper_mode_fill_fb_struct(dev, &rfb->base, mode_cmd);
+	drm_helper_mode_fill_fb_struct(dev, &rfb->base, info, mode_cmd);
 	/* Verify that the modifier is supported. */
 	if (!drm_any_plane_has_format(dev, mode_cmd->pixel_format,
 				      mode_cmd->modifier[0])) {
@@ -1136,7 +1217,11 @@ static int amdgpu_display_gem_fb_verify_and_init(struct drm_device *dev,
 	if (ret)
 		goto err;
 
-	ret = drm_framebuffer_init(dev, &rfb->base, &amdgpu_fb_funcs);
+	if (drm_drv_uses_atomic_modeset(dev))
+		ret = drm_framebuffer_init(dev, &rfb->base,
+					   &amdgpu_fb_funcs_atomic);
+	else
+		ret = drm_framebuffer_init(dev, &rfb->base, &amdgpu_fb_funcs);
 
 	if (ret)
 		goto err;
@@ -1169,7 +1254,8 @@ static int amdgpu_display_framebuffer_init(struct drm_device *dev,
 		}
 	}
 
-	ret = amdgpu_display_get_fb_info(rfb, &rfb->tiling_flags, &rfb->tmz_surface);
+	ret = amdgpu_display_get_fb_info(rfb, &rfb->tiling_flags, &rfb->tmz_surface,
+					 &rfb->gfx12_dcc);
 	if (ret)
 		return ret;
 
@@ -1183,7 +1269,11 @@ static int amdgpu_display_framebuffer_init(struct drm_device *dev,
 
 	if (!dev->mode_config.fb_modifiers_not_supported &&
 	    !(rfb->base.flags & DRM_MODE_FB_MODIFIERS)) {
-		ret = convert_tiling_flags_to_modifier(rfb);
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(12, 0, 0))
+			ret = convert_tiling_flags_to_modifier_gfx12(rfb);
+		else
+			ret = convert_tiling_flags_to_modifier(rfb);
+
 		if (ret) {
 			drm_dbg_kms(dev, "Failed to convert tiling flags 0x%llX to a modifier",
 				    rfb->tiling_flags);
@@ -1206,6 +1296,7 @@ static int amdgpu_display_framebuffer_init(struct drm_device *dev,
 struct drm_framebuffer *
 amdgpu_display_user_framebuffer_create(struct drm_device *dev,
 				       struct drm_file *file_priv,
+				       const struct drm_format_info *info,
 				       const struct drm_mode_fb_cmd2 *mode_cmd)
 {
 	struct amdgpu_framebuffer *amdgpu_fb;
@@ -1216,15 +1307,17 @@ amdgpu_display_user_framebuffer_create(struct drm_device *dev,
 
 	obj = drm_gem_object_lookup(file_priv, mode_cmd->handles[0]);
 	if (obj ==  NULL) {
-		drm_dbg_kms(dev, "No GEM object associated to handle 0x%08X, "
-			    "can't create framebuffer\n", mode_cmd->handles[0]);
+		drm_dbg_kms(dev,
+			    "No GEM object associated to handle 0x%08X, can't create framebuffer\n",
+			    mode_cmd->handles[0]);
+
 		return ERR_PTR(-ENOENT);
 	}
 
 	/* Handle is imported dma-buf, so cannot be migrated to VRAM for scanout */
 	bo = gem_to_amdgpu_bo(obj);
 	domains = amdgpu_display_supported_domains(drm_to_adev(dev), bo->flags);
-	if (obj->import_attach && !(domains & AMDGPU_GEM_DOMAIN_GTT)) {
+	if (drm_gem_is_imported(obj) && !(domains & AMDGPU_GEM_DOMAIN_GTT)) {
 		drm_dbg_kms(dev, "Cannot create framebuffer from imported dma_buf\n");
 		drm_gem_object_put(obj);
 		return ERR_PTR(-EINVAL);
@@ -1237,7 +1330,7 @@ amdgpu_display_user_framebuffer_create(struct drm_device *dev,
 	}
 
 	ret = amdgpu_display_gem_fb_verify_and_init(dev, amdgpu_fb, file_priv,
-						    mode_cmd, obj);
+						    info, mode_cmd, obj);
 	if (ret) {
 		kfree(amdgpu_fb);
 		drm_gem_object_put(obj);
@@ -1252,24 +1345,82 @@ const struct drm_mode_config_funcs amdgpu_mode_funcs = {
 	.fb_create = amdgpu_display_user_framebuffer_create,
 };
 
-static const struct drm_prop_enum_list amdgpu_underscan_enum_list[] =
-{	{ UNDERSCAN_OFF, "off" },
+static const struct drm_prop_enum_list amdgpu_underscan_enum_list[] = {
+	{ UNDERSCAN_OFF, "off" },
 	{ UNDERSCAN_ON, "on" },
 	{ UNDERSCAN_AUTO, "auto" },
 };
 
-static const struct drm_prop_enum_list amdgpu_audio_enum_list[] =
-{	{ AMDGPU_AUDIO_DISABLE, "off" },
+static const struct drm_prop_enum_list amdgpu_audio_enum_list[] = {
+	{ AMDGPU_AUDIO_DISABLE, "off" },
 	{ AMDGPU_AUDIO_ENABLE, "on" },
 	{ AMDGPU_AUDIO_AUTO, "auto" },
 };
 
 /* XXX support different dither options? spatial, temporal, both, etc. */
-static const struct drm_prop_enum_list amdgpu_dither_enum_list[] =
-{	{ AMDGPU_FMT_DITHER_DISABLE, "off" },
+static const struct drm_prop_enum_list amdgpu_dither_enum_list[] = {
+	{ AMDGPU_FMT_DITHER_DISABLE, "off" },
 	{ AMDGPU_FMT_DITHER_ENABLE, "on" },
 };
 
+/**
+ * DOC: property for adaptive backlight modulation
+ *
+ * The 'adaptive backlight modulation' property is used for the compositor to
+ * directly control the adaptive backlight modulation power savings feature
+ * that is part of DCN hardware.
+ *
+ * The property will be attached specifically to eDP panels that support it.
+ *
+ * The property is by default set to 'sysfs' to allow the sysfs file 'panel_power_savings'
+ * to be able to control it.
+ * If set to 'off' the compositor will ensure it stays off.
+ * The other values 'min', 'bias min', 'bias max', and 'max' will control the
+ * intensity of the power savings.
+ *
+ * Modifying this value can have implications on color accuracy, so tread
+ * carefully.
+ */
+static int amdgpu_display_setup_abm_prop(struct amdgpu_device *adev)
+{
+	const struct drm_prop_enum_list props[] = {
+		{ ABM_SYSFS_CONTROL, "sysfs" },
+		{ ABM_LEVEL_OFF, "off" },
+		{ ABM_LEVEL_MIN, "min" },
+		{ ABM_LEVEL_BIAS_MIN, "bias min" },
+		{ ABM_LEVEL_BIAS_MAX, "bias max" },
+		{ ABM_LEVEL_MAX, "max" },
+	};
+	struct drm_property *prop;
+	int i;
+
+	if (!adev->dc_enabled)
+		return 0;
+
+	prop = drm_property_create(adev_to_drm(adev), DRM_MODE_PROP_ENUM,
+				"adaptive backlight modulation",
+				6);
+	if (!prop)
+		return -ENOMEM;
+
+	for (i = 0; i < ARRAY_SIZE(props); i++) {
+		int ret;
+
+		ret = drm_property_add_enum(prop, props[i].type,
+						props[i].name);
+
+		if (ret) {
+			drm_property_destroy(adev_to_drm(adev), prop);
+
+			return ret;
+		}
+	}
+
+	adev->mode_info.abm_level_property = prop;
+
+	return 0;
+}
+
 int amdgpu_display_modeset_create_props(struct amdgpu_device *adev)
 {
 	int sz;
@@ -1316,15 +1467,7 @@ int amdgpu_display_modeset_create_props(struct amdgpu_device *adev)
 					 "dither",
 					 amdgpu_dither_enum_list, sz);
 
-	if (adev->dc_enabled) {
-		adev->mode_info.abm_level_property =
-			drm_property_create_range(adev_to_drm(adev), 0,
-						  "abm level", 0, 4);
-		if (!adev->mode_info.abm_level_property)
-			return -ENOMEM;
-	}
-
-	return 0;
+	return amdgpu_display_setup_abm_prop(adev);
 }
 
 void amdgpu_display_update_priority(struct amdgpu_device *adev)
@@ -1391,7 +1534,7 @@ bool amdgpu_display_crtc_scaling_mode_fixup(struct drm_crtc *crtc,
 		if ((!(mode->flags & DRM_MODE_FLAG_INTERLACE)) &&
 		    ((amdgpu_encoder->underscan_type == UNDERSCAN_ON) ||
 		     ((amdgpu_encoder->underscan_type == UNDERSCAN_AUTO) &&
-		      connector->display_info.is_hdmi &&
+		      connector && connector->display_info.is_hdmi &&
 		      amdgpu_display_is_hdtv_mode(mode)))) {
 			if (amdgpu_encoder->underscan_hborder != 0)
 				amdgpu_crtc->h_border = amdgpu_encoder->underscan_hborder;
@@ -1410,6 +1553,7 @@ bool amdgpu_display_crtc_scaling_mode_fixup(struct drm_crtc *crtc,
 	}
 	if (amdgpu_crtc->rmx_type != RMX_OFF) {
 		fixed20_12 a, b;
+
 		a.full = dfixed_const(src_v);
 		b.full = dfixed_const(dst_v);
 		amdgpu_crtc->vsc.full = dfixed_div(a, b);
@@ -1429,7 +1573,7 @@ bool amdgpu_display_crtc_scaling_mode_fixup(struct drm_crtc *crtc,
  *
  * \param dev Device to query.
  * \param pipe Crtc to query.
- * \param flags Flags from caller (DRM_CALLED_FROM_VBLIRQ or 0).
+ * \param flags from caller (DRM_CALLED_FROM_VBLIRQ or 0).
  *              For driver internal use only also supports these flags:
  *
  *              USE_REAL_VBLANKSTART to use the real start of vblank instead
@@ -1496,8 +1640,7 @@ int amdgpu_display_get_crtc_scanoutpos(struct drm_device *dev,
 		ret |= DRM_SCANOUTPOS_ACCURATE;
 		vbl_start = vbl & 0x1fff;
 		vbl_end = (vbl >> 16) & 0x1fff;
-	}
-	else {
+	} else {
 		/* No: Fake something reasonable which gives at least ok results. */
 		vbl_start = mode->crtc_vdisplay;
 		vbl_end = 0;
@@ -1505,8 +1648,8 @@ int amdgpu_display_get_crtc_scanoutpos(struct drm_device *dev,
 
 	/* Called from driver internal vblank counter query code? */
 	if (flags & GET_DISTANCE_TO_VBLANKSTART) {
-	    /* Caller wants distance from real vbl_start in *hpos */
-	    *hpos = *vpos - vbl_start;
+		/* Caller wants distance from real vbl_start in *hpos */
+		*hpos = *vpos - vbl_start;
 	}
 
 	/* Fudge vblank to start a few scanlines earlier to handle the
@@ -1528,7 +1671,7 @@ int amdgpu_display_get_crtc_scanoutpos(struct drm_device *dev,
 
 	/* In vblank? */
 	if (in_vbl)
-	    ret |= DRM_SCANOUTPOS_IN_VBLANK;
+		ret |= DRM_SCANOUTPOS_IN_VBLANK;
 
 	/* Called from driver internal vblank counter query code? */
 	if (flags & GET_DISTANCE_TO_VBLANKSTART) {
@@ -1636,6 +1779,7 @@ int amdgpu_display_suspend_helper(struct amdgpu_device *adev)
 
 		if (amdgpu_crtc->cursor_bo && !adev->enable_virtual_display) {
 			struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
+
 			r = amdgpu_bo_reserve(aobj, true);
 			if (r == 0) {
 				amdgpu_bo_unpin(aobj);
@@ -1643,9 +1787,9 @@ int amdgpu_display_suspend_helper(struct amdgpu_device *adev)
 			}
 		}
 
-		if (fb == NULL || fb->obj[0] == NULL) {
+		if (!fb || !fb->obj[0])
 			continue;
-		}
+
 		robj = gem_to_amdgpu_bo(fb->obj[0]);
 		if (!amdgpu_display_robj_is_fb(adev, robj)) {
 			r = amdgpu_bo_reserve(robj, true);
@@ -1672,8 +1816,10 @@ int amdgpu_display_resume_helper(struct amdgpu_device *adev)
 
 		if (amdgpu_crtc->cursor_bo && !adev->enable_virtual_display) {
 			struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
+
 			r = amdgpu_bo_reserve(aobj, true);
 			if (r == 0) {
+				aobj->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 				r = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM);
 				if (r != 0)
 					dev_err(adev->dev, "Failed to pin cursor BO (%d)\n", r);
@@ -1701,3 +1847,82 @@ int amdgpu_display_resume_helper(struct amdgpu_device *adev)
 	return 0;
 }
 
+/* panic_bo is set in amdgpu_dm_plane_get_scanout_buffer() and only used in amdgpu_dm_set_pixel()
+ * they are called from the panic handler, and protected by the drm_panic spinlock.
+ */
+static struct amdgpu_bo *panic_abo;
+
+/* Use the indirect MMIO to write each pixel to the GPU VRAM,
+ * This is a simplified version of amdgpu_device_mm_access()
+ */
+static void amdgpu_display_set_pixel(struct drm_scanout_buffer *sb,
+				     unsigned int x,
+				     unsigned int y,
+				     u32 color)
+{
+	struct amdgpu_res_cursor cursor;
+	unsigned long offset;
+	struct amdgpu_bo *abo = panic_abo;
+	struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev);
+	uint32_t tmp;
+
+	offset = x * 4 + y * sb->pitch[0];
+	amdgpu_res_first(abo->tbo.resource, offset, 4, &cursor);
+
+	tmp = cursor.start >> 31;
+	WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t) cursor.start) | 0x80000000);
+	if (tmp != 0xffffffff)
+		WREG32_NO_KIQ(mmMM_INDEX_HI, tmp);
+	WREG32_NO_KIQ(mmMM_DATA, color);
+}
+
+int amdgpu_display_get_scanout_buffer(struct drm_plane *plane,
+				      struct drm_scanout_buffer *sb)
+{
+	struct amdgpu_bo *abo;
+	struct drm_framebuffer *fb = plane->state->fb;
+
+	if (!fb)
+		return -EINVAL;
+
+	DRM_DEBUG_KMS("Framebuffer %dx%d %p4cc\n", fb->width, fb->height, &fb->format->format);
+
+	abo = gem_to_amdgpu_bo(fb->obj[0]);
+	if (!abo)
+		return -EINVAL;
+
+	sb->width = fb->width;
+	sb->height = fb->height;
+	/* Use the generic linear format, because tiling will be disabled in panic_flush() */
+	sb->format = drm_format_info(fb->format->format);
+	if (!sb->format)
+		return -EINVAL;
+
+	sb->pitch[0] = fb->pitches[0];
+
+	if (abo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) {
+		if (abo->tbo.resource->mem_type != TTM_PL_VRAM) {
+			drm_warn(plane->dev, "amdgpu panic, framebuffer not in VRAM\n");
+			return -EINVAL;
+		}
+		/* Only handle 32bits format, to simplify mmio access */
+		if (fb->format->cpp[0] != 4) {
+			drm_warn(plane->dev, "amdgpu panic, pixel format is not 32bits\n");
+			return -EINVAL;
+		}
+		sb->set_pixel = amdgpu_display_set_pixel;
+		panic_abo = abo;
+		return 0;
+	}
+	if (!abo->kmap.virtual &&
+	    ttm_bo_kmap(&abo->tbo, 0, PFN_UP(abo->tbo.base.size), &abo->kmap)) {
+		drm_warn(plane->dev, "amdgpu bo map failed, panic won't be displayed\n");
+		return -ENOMEM;
+	}
+	if (abo->kmap.bo_kmap_type & TTM_BO_MAP_IOMEM_MASK)
+		iosys_map_set_vaddr_iomem(&sb->map[0], abo->kmap.virtual);
+	else
+		iosys_map_set_vaddr(&sb->map[0], abo->kmap.virtual);
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h
index 9d19940f73c8..49a29bf47a37 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h
@@ -23,6 +23,8 @@
 #ifndef __AMDGPU_DISPLAY_H__
 #define __AMDGPU_DISPLAY_H__
 
+#include <drm/drm_panic.h>
+
 #define amdgpu_display_vblank_get_counter(adev, crtc) (adev)->mode_info.funcs->vblank_get_counter((adev), (crtc))
 #define amdgpu_display_backlight_set_level(adev, e, l) (adev)->mode_info.funcs->backlight_set_level((e), (l))
 #define amdgpu_display_backlight_get_level(adev, e) (adev)->mode_info.funcs->backlight_get_level((e))
@@ -42,6 +44,7 @@ uint32_t amdgpu_display_supported_domains(struct amdgpu_device *adev,
 struct drm_framebuffer *
 amdgpu_display_user_framebuffer_create(struct drm_device *dev,
 				       struct drm_file *file_priv,
+				       const struct drm_format_info *info,
 				       const struct drm_mode_fb_cmd2 *mode_cmd);
 const struct drm_format_info *
 amdgpu_lookup_format_info(u32 format, uint64_t modifier);
@@ -49,4 +52,14 @@ amdgpu_lookup_format_info(u32 format, uint64_t modifier);
 int amdgpu_display_suspend_helper(struct amdgpu_device *adev);
 int amdgpu_display_resume_helper(struct amdgpu_device *adev);
 
+int amdgpu_display_get_scanout_buffer(struct drm_plane *plane,
+				      struct drm_scanout_buffer *sb);
+
+#define ABM_SYSFS_CONTROL	-1
+#define ABM_LEVEL_OFF		0
+#define ABM_LEVEL_MIN		1
+#define ABM_LEVEL_BIAS_MIN	2
+#define ABM_LEVEL_BIAS_MAX	3
+#define ABM_LEVEL_MAX		4
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c
index 0c001bb8fc2b..e22cfa7c6d32 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dma_buf.c
@@ -36,12 +36,35 @@
 #include "amdgpu_gem.h"
 #include "amdgpu_dma_buf.h"
 #include "amdgpu_xgmi.h"
+#include "amdgpu_vm.h"
 #include <drm/amdgpu_drm.h>
 #include <drm/ttm/ttm_tt.h>
 #include <linux/dma-buf.h>
 #include <linux/dma-fence-array.h>
 #include <linux/pci-p2pdma.h>
-#include <linux/pm_runtime.h>
+
+static const struct dma_buf_attach_ops amdgpu_dma_buf_attach_ops;
+
+/**
+ * dma_buf_attach_adev - Helper to get adev of an attachment
+ *
+ * @attach: attachment
+ *
+ * Returns:
+ * A struct amdgpu_device * if the attaching device is an amdgpu device or
+ * partition, NULL otherwise.
+ */
+static struct amdgpu_device *dma_buf_attach_adev(struct dma_buf_attachment *attach)
+{
+	if (attach->importer_ops == &amdgpu_dma_buf_attach_ops) {
+		struct drm_gem_object *obj = attach->importer_priv;
+		struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+
+		return amdgpu_ttm_adev(bo->tbo.bdev);
+	}
+
+	return NULL;
+}
 
 /**
  * amdgpu_dma_buf_attach - &dma_buf_ops.attach implementation
@@ -54,42 +77,49 @@
 static int amdgpu_dma_buf_attach(struct dma_buf *dmabuf,
 				 struct dma_buf_attachment *attach)
 {
+	struct amdgpu_device *attach_adev = dma_buf_attach_adev(attach);
 	struct drm_gem_object *obj = dmabuf->priv;
 	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
 	int r;
 
-	if (pci_p2pdma_distance(adev->pdev, attach->dev, false) < 0)
+	/*
+	 * Disable peer-to-peer access for DCC-enabled VRAM surfaces on GFX12+.
+	 * Such buffers cannot be safely accessed over P2P due to device-local
+	 * compression metadata. Fallback to system-memory path instead.
+	 * Device supports GFX12 (GC 12.x or newer)
+	 * BO was created with the AMDGPU_GEM_CREATE_GFX12_DCC flag
+	 *
+	 */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(12, 0, 0) &&
+	    bo->flags & AMDGPU_GEM_CREATE_GFX12_DCC)
 		attach->peer2peer = false;
 
-	r = pm_runtime_get_sync(adev_to_drm(adev)->dev);
-	if (r < 0)
-		goto out;
+	/*
+	 * Disable peer-to-peer access for DCC-enabled VRAM surfaces on GFX12+.
+	 * Such buffers cannot be safely accessed over P2P due to device-local
+	 * compression metadata. Fallback to system-memory path instead.
+	 * Device supports GFX12 (GC 12.x or newer)
+	 * BO was created with the AMDGPU_GEM_CREATE_GFX12_DCC flag
+	 *
+	 */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(12, 0, 0) &&
+	    bo->flags & AMDGPU_GEM_CREATE_GFX12_DCC)
+		attach->peer2peer = false;
 
-	return 0;
+	if (!amdgpu_dmabuf_is_xgmi_accessible(attach_adev, bo) &&
+	    pci_p2pdma_distance(adev->pdev, attach->dev, false) < 0)
+		attach->peer2peer = false;
 
-out:
-	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
-	return r;
-}
+	r = dma_resv_lock(bo->tbo.base.resv, NULL);
+	if (r)
+		return r;
 
-/**
- * amdgpu_dma_buf_detach - &dma_buf_ops.detach implementation
- *
- * @dmabuf: DMA-buf where we remove the attachment from
- * @attach: the attachment to remove
- *
- * Called when an attachment is removed from the DMA-buf.
- */
-static void amdgpu_dma_buf_detach(struct dma_buf *dmabuf,
-				  struct dma_buf_attachment *attach)
-{
-	struct drm_gem_object *obj = dmabuf->priv;
-	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
-	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
+	amdgpu_vm_bo_update_shared(bo);
 
-	pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
-	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
+	dma_resv_unlock(bo->tbo.base.resv);
+
+	return 0;
 }
 
 /**
@@ -101,11 +131,35 @@ static void amdgpu_dma_buf_detach(struct dma_buf *dmabuf,
  */
 static int amdgpu_dma_buf_pin(struct dma_buf_attachment *attach)
 {
-	struct drm_gem_object *obj = attach->dmabuf->priv;
-	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+	struct dma_buf *dmabuf = attach->dmabuf;
+	struct amdgpu_bo *bo = gem_to_amdgpu_bo(dmabuf->priv);
+	u32 domains = bo->allowed_domains;
+
+	dma_resv_assert_held(dmabuf->resv);
+
+	/* Try pinning into VRAM to allow P2P with RDMA NICs without ODP
+	 * support if all attachments can do P2P. If any attachment can't do
+	 * P2P just pin into GTT instead.
+	 *
+	 * To avoid with conflicting pinnings between GPUs and RDMA when move
+	 * notifiers are disabled, only allow pinning in VRAM when move
+	 * notiers are enabled.
+	 */
+	if (!IS_ENABLED(CONFIG_DMABUF_MOVE_NOTIFY)) {
+		domains &= ~AMDGPU_GEM_DOMAIN_VRAM;
+	} else {
+		list_for_each_entry(attach, &dmabuf->attachments, node)
+			if (!attach->peer2peer)
+				domains &= ~AMDGPU_GEM_DOMAIN_VRAM;
+	}
+
+	if (domains & AMDGPU_GEM_DOMAIN_VRAM)
+		bo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
 
-	/* pin buffer into GTT */
-	return amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT);
+	if (WARN_ON(!domains))
+		return -EINVAL;
+
+	return amdgpu_bo_pin(bo, domains);
 }
 
 /**
@@ -149,7 +203,7 @@ static struct sg_table *amdgpu_dma_buf_map(struct dma_buf_attachment *attach,
 	if (!bo->tbo.pin_count) {
 		/* move buffer into GTT or VRAM */
 		struct ttm_operation_ctx ctx = { false, false };
-		unsigned domains = AMDGPU_GEM_DOMAIN_GTT;
+		unsigned int domains = AMDGPU_GEM_DOMAIN_GTT;
 
 		if (bo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM &&
 		    attach->peer2peer) {
@@ -160,10 +214,6 @@ static struct sg_table *amdgpu_dma_buf_map(struct dma_buf_attachment *attach,
 		r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 		if (r)
 			return ERR_PTR(r);
-
-	} else if (!(amdgpu_mem_type_to_domain(bo->tbo.resource->mem_type) &
-		     AMDGPU_GEM_DOMAIN_GTT)) {
-		return ERR_PTR(-EBUSY);
 	}
 
 	switch (bo->tbo.resource->mem_type) {
@@ -180,6 +230,11 @@ static struct sg_table *amdgpu_dma_buf_map(struct dma_buf_attachment *attach,
 		break;
 
 	case TTM_PL_VRAM:
+		/* XGMI-accessible memory should never be DMA-mapped */
+		if (WARN_ON(amdgpu_dmabuf_is_xgmi_accessible(
+				dma_buf_attach_adev(attach), bo)))
+			return ERR_PTR(-EINVAL);
+
 		r = amdgpu_vram_mgr_alloc_sgt(adev, bo->tbo.resource, 0,
 					      bo->tbo.base.size, attach->dev,
 					      dir, &sgt);
@@ -211,7 +266,7 @@ static void amdgpu_dma_buf_unmap(struct dma_buf_attachment *attach,
 				 struct sg_table *sgt,
 				 enum dma_data_direction dir)
 {
-	if (sgt->sgl->page_link) {
+	if (sg_page(sgt->sgl)) {
 		dma_unmap_sgtable(attach->dev, sgt, dir, 0);
 		sg_free_table(sgt);
 		kfree(sgt);
@@ -261,9 +316,38 @@ static int amdgpu_dma_buf_begin_cpu_access(struct dma_buf *dma_buf,
 	return ret;
 }
 
+static int amdgpu_dma_buf_vmap(struct dma_buf *dma_buf, struct iosys_map *map)
+{
+	struct drm_gem_object *obj = dma_buf->priv;
+	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+	int ret;
+
+	/*
+	 * Pin to keep buffer in place while it's vmap'ed. The actual
+	 * domain is not that important as long as it's mapable. Using
+	 * GTT and VRAM should be compatible with most use cases.
+	 */
+	ret = amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT | AMDGPU_GEM_DOMAIN_VRAM);
+	if (ret)
+		return ret;
+	ret = drm_gem_dmabuf_vmap(dma_buf, map);
+	if (ret)
+		amdgpu_bo_unpin(bo);
+
+	return ret;
+}
+
+static void amdgpu_dma_buf_vunmap(struct dma_buf *dma_buf, struct iosys_map *map)
+{
+	struct drm_gem_object *obj = dma_buf->priv;
+	struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+
+	drm_gem_dmabuf_vunmap(dma_buf, map);
+	amdgpu_bo_unpin(bo);
+}
+
 const struct dma_buf_ops amdgpu_dmabuf_ops = {
 	.attach = amdgpu_dma_buf_attach,
-	.detach = amdgpu_dma_buf_detach,
 	.pin = amdgpu_dma_buf_pin,
 	.unpin = amdgpu_dma_buf_unpin,
 	.map_dma_buf = amdgpu_dma_buf_map,
@@ -271,8 +355,8 @@ const struct dma_buf_ops amdgpu_dmabuf_ops = {
 	.release = drm_gem_dmabuf_release,
 	.begin_cpu_access = amdgpu_dma_buf_begin_cpu_access,
 	.mmap = drm_gem_dmabuf_mmap,
-	.vmap = drm_gem_dmabuf_vmap,
-	.vunmap = drm_gem_dmabuf_vunmap,
+	.vmap = amdgpu_dma_buf_vmap,
+	.vunmap = amdgpu_dma_buf_vunmap,
 };
 
 /**
@@ -290,11 +374,23 @@ struct dma_buf *amdgpu_gem_prime_export(struct drm_gem_object *gobj,
 {
 	struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj);
 	struct dma_buf *buf;
+	struct ttm_operation_ctx ctx = {
+		.interruptible = true,
+		.no_wait_gpu = true,
+		/* We opt to avoid OOM on system pages allocations */
+		.gfp_retry_mayfail = true,
+		.allow_res_evict = false,
+	};
+	int ret;
 
 	if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) ||
 	    bo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID)
 		return ERR_PTR(-EPERM);
 
+	ret = ttm_bo_setup_export(&bo->tbo, &ctx);
+	if (ret)
+		return ERR_PTR(ret);
+
 	buf = drm_gem_prime_export(gobj, flags);
 	if (!IS_ERR(buf))
 		buf->ops = &amdgpu_dmabuf_ops;
@@ -331,12 +427,13 @@ amdgpu_dma_buf_create_obj(struct drm_device *dev, struct dma_buf *dma_buf)
 
 		flags |= other->flags & (AMDGPU_GEM_CREATE_CPU_GTT_USWC |
 					 AMDGPU_GEM_CREATE_COHERENT |
+					 AMDGPU_GEM_CREATE_EXT_COHERENT |
 					 AMDGPU_GEM_CREATE_UNCACHED);
 	}
 
 	ret = amdgpu_gem_object_create(adev, dma_buf->size, PAGE_SIZE,
 				       AMDGPU_GEM_DOMAIN_CPU, flags,
-				       ttm_bo_type_sg, resv, &gobj);
+				       ttm_bo_type_sg, resv, &gobj, 0);
 	if (ret)
 		goto error;
 
@@ -372,6 +469,10 @@ amdgpu_dma_buf_move_notify(struct dma_buf_attachment *attach)
 	struct amdgpu_vm_bo_base *bo_base;
 	int r;
 
+	/* FIXME: This should be after the "if", but needs a fix to make sure
+	 * DMABuf imports are initialized in the right VM list.
+	 */
+	amdgpu_vm_bo_invalidate(bo, false);
 	if (!bo->tbo.resource || bo->tbo.resource->mem_type == TTM_PL_SYSTEM)
 		return;
 
@@ -403,9 +504,12 @@ amdgpu_dma_buf_move_notify(struct dma_buf_attachment *attach)
 				continue;
 		}
 
-		r = amdgpu_vm_clear_freed(adev, vm, NULL);
+		/* Reserve fences for two SDMA page table updates */
+		r = dma_resv_reserve_fences(resv, 2);
+		if (!r)
+			r = amdgpu_vm_clear_freed(adev, vm, NULL);
 		if (!r)
-			r = amdgpu_vm_handle_moved(adev, vm);
+			r = amdgpu_vm_handle_moved(adev, vm, ticket);
 
 		if (r && r != -EBUSY)
 			DRM_ERROR("Failed to invalidate VM page tables (%d))\n",
@@ -479,7 +583,10 @@ bool amdgpu_dmabuf_is_xgmi_accessible(struct amdgpu_device *adev,
 	struct drm_gem_object *obj = &bo->tbo.base;
 	struct drm_gem_object *gobj;
 
-	if (obj->import_attach) {
+	if (!adev)
+		return false;
+
+	if (drm_gem_is_imported(obj)) {
 		struct dma_buf *dma_buf = obj->import_attach->dmabuf;
 
 		if (dma_buf->ops != &amdgpu_dmabuf_ops)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h
index 7199b6b0be81..2675689ef70f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h
@@ -21,6 +21,9 @@
  *
  */
 
+#ifndef AMDGPU_DOORBELL_H
+#define AMDGPU_DOORBELL_H
+
 /*
  * GPU doorbell structures, functions & helpers
  */
@@ -28,8 +31,15 @@ struct amdgpu_doorbell {
 	/* doorbell mmio */
 	resource_size_t		base;
 	resource_size_t		size;
-	u32 __iomem		*ptr;
-	u32			num_doorbells;	/* Number of doorbells actually reserved for amdgpu. */
+
+	/* Number of doorbells reserved for amdgpu kernel driver */
+	u32 num_kernel_doorbells;
+
+	/* Kernel doorbells */
+	struct amdgpu_bo *kernel_doorbells;
+
+	/* For CPU access of doorbells */
+	uint32_t *cpu_addr;
 };
 
 /* Reserved doorbells for amdgpu (including multimedia).
@@ -54,7 +64,7 @@ struct amdgpu_doorbell_index {
 	uint32_t gfx_ring1;
 	uint32_t gfx_userqueue_start;
 	uint32_t gfx_userqueue_end;
-	uint32_t sdma_engine[8];
+	uint32_t sdma_engine[16];
 	uint32_t mes_ring0;
 	uint32_t mes_ring1;
 	uint32_t ih;
@@ -76,15 +86,17 @@ struct amdgpu_doorbell_index {
 			uint32_t vce_ring6_7;
 		} uvd_vce;
 	};
+	uint32_t vpe_ring;
 	uint32_t first_non_cp;
 	uint32_t last_non_cp;
 	uint32_t max_assignment;
 	/* Per engine SDMA doorbell size in dword */
 	uint32_t sdma_doorbell_range;
+	/* Per xcc doorbell size for KIQ/KCQ */
+	uint32_t xcc_doorbell_range;
 };
 
-typedef enum _AMDGPU_DOORBELL_ASSIGNMENT
-{
+enum AMDGPU_DOORBELL_ASSIGNMENT {
 	AMDGPU_DOORBELL_KIQ                     = 0x000,
 	AMDGPU_DOORBELL_HIQ                     = 0x001,
 	AMDGPU_DOORBELL_DIQ                     = 0x002,
@@ -102,10 +114,10 @@ typedef enum _AMDGPU_DOORBELL_ASSIGNMENT
 	AMDGPU_DOORBELL_IH                      = 0x1E8,
 	AMDGPU_DOORBELL_MAX_ASSIGNMENT          = 0x3FF,
 	AMDGPU_DOORBELL_INVALID                 = 0xFFFF
-} AMDGPU_DOORBELL_ASSIGNMENT;
+};
+
+enum AMDGPU_VEGA20_DOORBELL_ASSIGNMENT {
 
-typedef enum _AMDGPU_VEGA20_DOORBELL_ASSIGNMENT
-{
 	/* Compute + GFX: 0~255 */
 	AMDGPU_VEGA20_DOORBELL_KIQ                     = 0x000,
 	AMDGPU_VEGA20_DOORBELL_HIQ                     = 0x001,
@@ -159,12 +171,20 @@ typedef enum _AMDGPU_VEGA20_DOORBELL_ASSIGNMENT
 	AMDGPU_VEGA20_DOORBELL64_FIRST_NON_CP            = AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE0,
 	AMDGPU_VEGA20_DOORBELL64_LAST_NON_CP             = AMDGPU_VEGA20_DOORBELL64_VCE_RING6_7,
 
-	AMDGPU_VEGA20_DOORBELL_MAX_ASSIGNMENT            = 0x18F,
+	/* kiq/kcq from second XCD. Max 8 XCDs */
+	AMDGPU_VEGA20_DOORBELL_XCC1_KIQ_START             = 0x190,
+	/* 8 compute rings per GC. Max to 0x1CE */
+	AMDGPU_VEGA20_DOORBELL_XCC1_MEC_RING0_START       = 0x197,
+
+	/* AID1 SDMA: 0x1D0 ~ 0x1F7 */
+	AMDGPU_VEGA20_DOORBELL_AID1_sDMA_START           = 0x1D0,
+
+	AMDGPU_VEGA20_DOORBELL_MAX_ASSIGNMENT            = 0x1F7,
 	AMDGPU_VEGA20_DOORBELL_INVALID                   = 0xFFFF
-} AMDGPU_VEGA20_DOORBELL_ASSIGNMENT;
+};
+
+enum AMDGPU_NAVI10_DOORBELL_ASSIGNMENT {
 
-typedef enum _AMDGPU_NAVI10_DOORBELL_ASSIGNMENT
-{
 	/* Compute + GFX: 0~255 */
 	AMDGPU_NAVI10_DOORBELL_KIQ			= 0x000,
 	AMDGPU_NAVI10_DOORBELL_HIQ			= 0x001,
@@ -207,18 +227,19 @@ typedef enum _AMDGPU_NAVI10_DOORBELL_ASSIGNMENT
 	AMDGPU_NAVI10_DOORBELL64_VCNc_d			= 0x18E,
 	AMDGPU_NAVI10_DOORBELL64_VCNe_f			= 0x18F,
 
+	AMDGPU_NAVI10_DOORBELL64_VPE			= 0x190,
+
 	AMDGPU_NAVI10_DOORBELL64_FIRST_NON_CP		= AMDGPU_NAVI10_DOORBELL_sDMA_ENGINE0,
-	AMDGPU_NAVI10_DOORBELL64_LAST_NON_CP		= AMDGPU_NAVI10_DOORBELL64_VCNe_f,
+	AMDGPU_NAVI10_DOORBELL64_LAST_NON_CP		= AMDGPU_NAVI10_DOORBELL64_VPE,
 
-	AMDGPU_NAVI10_DOORBELL_MAX_ASSIGNMENT		= 0x18F,
+	AMDGPU_NAVI10_DOORBELL_MAX_ASSIGNMENT		= AMDGPU_NAVI10_DOORBELL64_VPE,
 	AMDGPU_NAVI10_DOORBELL_INVALID			= 0xFFFF
-} AMDGPU_NAVI10_DOORBELL_ASSIGNMENT;
+};
 
 /*
  * 64bit doorbell, offset are in QWORD, occupy 2KB doorbell space
  */
-typedef enum _AMDGPU_DOORBELL64_ASSIGNMENT
-{
+enum AMDGPU_DOORBELL64_ASSIGNMENT {
 	/*
 	 * All compute related doorbells: kiq, hiq, diq, traditional compute queue, user queue, should locate in
 	 * a continues range so that programming CP_MEC_DOORBELL_RANGE_LOWER/UPPER can cover this range.
@@ -294,15 +315,58 @@ typedef enum _AMDGPU_DOORBELL64_ASSIGNMENT
 
 	AMDGPU_DOORBELL64_MAX_ASSIGNMENT          = 0xFF,
 	AMDGPU_DOORBELL64_INVALID                 = 0xFFFF
-} AMDGPU_DOORBELL64_ASSIGNMENT;
+};
+
+enum AMDGPU_DOORBELL_ASSIGNMENT_LAYOUT1 {
+
+	/* XCC0: 0x00 ~20, XCC1: 20 ~ 2F ... */
+
+	/* KIQ/HIQ/DIQ */
+	AMDGPU_DOORBELL_LAYOUT1_KIQ_START		= 0x000,
+	AMDGPU_DOORBELL_LAYOUT1_HIQ			= 0x001,
+	AMDGPU_DOORBELL_LAYOUT1_DIQ			= 0x002,
+	/* Compute: 0x08 ~ 0x20  */
+	AMDGPU_DOORBELL_LAYOUT1_MEC_RING_START		= 0x008,
+	AMDGPU_DOORBELL_LAYOUT1_MEC_RING_END		= 0x00F,
+	AMDGPU_DOORBELL_LAYOUT1_USERQUEUE_START		= 0x010,
+	AMDGPU_DOORBELL_LAYOUT1_USERQUEUE_END		= 0x01F,
+	AMDGPU_DOORBELL_LAYOUT1_XCC_RANGE		= 0x020,
+
+	/* SDMA: 0x100 ~ 0x19F */
+	AMDGPU_DOORBELL_LAYOUT1_sDMA_ENGINE_START	= 0x100,
+	AMDGPU_DOORBELL_LAYOUT1_sDMA_ENGINE_END		= 0x19F,
+	/* IH: 0x1A0 ~ 0x1AF */
+	AMDGPU_DOORBELL_LAYOUT1_IH                      = 0x1A0,
+	/* VCN: 0x1B0 ~ 0x1E8 */
+	AMDGPU_DOORBELL_LAYOUT1_VCN_START               = 0x1B0,
+	AMDGPU_DOORBELL_LAYOUT1_VCN_END                 = 0x1E8,
+
+	AMDGPU_DOORBELL_LAYOUT1_FIRST_NON_CP		= AMDGPU_DOORBELL_LAYOUT1_sDMA_ENGINE_START,
+	AMDGPU_DOORBELL_LAYOUT1_LAST_NON_CP		= AMDGPU_DOORBELL_LAYOUT1_VCN_END,
+
+	AMDGPU_DOORBELL_LAYOUT1_MAX_ASSIGNMENT          = 0x1E8,
+	AMDGPU_DOORBELL_LAYOUT1_INVALID                 = 0xFFFF
+};
 
 u32 amdgpu_mm_rdoorbell(struct amdgpu_device *adev, u32 index);
 void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v);
 u64 amdgpu_mm_rdoorbell64(struct amdgpu_device *adev, u32 index);
 void amdgpu_mm_wdoorbell64(struct amdgpu_device *adev, u32 index, u64 v);
 
+/*
+ * GPU doorbell aperture helpers function.
+ */
+int amdgpu_doorbell_init(struct amdgpu_device *adev);
+void amdgpu_doorbell_fini(struct amdgpu_device *adev);
+int amdgpu_doorbell_create_kernel_doorbells(struct amdgpu_device *adev);
+uint32_t amdgpu_doorbell_index_on_bar(struct amdgpu_device *adev,
+				      struct amdgpu_bo *db_bo,
+				      uint32_t doorbell_index,
+				      uint32_t db_size);
+
 #define RDOORBELL32(index) amdgpu_mm_rdoorbell(adev, (index))
 #define WDOORBELL32(index, v) amdgpu_mm_wdoorbell(adev, (index), (v))
 #define RDOORBELL64(index) amdgpu_mm_rdoorbell64(adev, (index))
 #define WDOORBELL64(index, v) amdgpu_mm_wdoorbell64(adev, (index), (v))
 
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell_mgr.c
new file mode 100644
index 000000000000..3040437d99c2
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell_mgr.c
@@ -0,0 +1,244 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+
+/**
+ * amdgpu_mm_rdoorbell - read a doorbell dword
+ *
+ * @adev: amdgpu_device pointer
+ * @index: doorbell index
+ *
+ * Returns the value in the doorbell aperture at the
+ * requested doorbell index (CIK).
+ */
+u32 amdgpu_mm_rdoorbell(struct amdgpu_device *adev, u32 index)
+{
+	if (amdgpu_device_skip_hw_access(adev))
+		return 0;
+
+	if (index < adev->doorbell.num_kernel_doorbells)
+		return readl(adev->doorbell.cpu_addr + index);
+
+	dev_err(adev->dev, "reading beyond doorbell aperture: 0x%08x!\n",
+		index);
+	return 0;
+}
+
+/**
+ * amdgpu_mm_wdoorbell - write a doorbell dword
+ *
+ * @adev: amdgpu_device pointer
+ * @index: doorbell index
+ * @v: value to write
+ *
+ * Writes @v to the doorbell aperture at the
+ * requested doorbell index (CIK).
+ */
+void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v)
+{
+	if (amdgpu_device_skip_hw_access(adev))
+		return;
+
+	if (index < adev->doorbell.num_kernel_doorbells)
+		writel(v, adev->doorbell.cpu_addr + index);
+	else
+		dev_err(adev->dev,
+			"writing beyond doorbell aperture: 0x%08x!\n", index);
+}
+
+/**
+ * amdgpu_mm_rdoorbell64 - read a doorbell Qword
+ *
+ * @adev: amdgpu_device pointer
+ * @index: doorbell index
+ *
+ * Returns the value in the doorbell aperture at the
+ * requested doorbell index (VEGA10+).
+ */
+u64 amdgpu_mm_rdoorbell64(struct amdgpu_device *adev, u32 index)
+{
+	if (amdgpu_device_skip_hw_access(adev))
+		return 0;
+
+	if (index < adev->doorbell.num_kernel_doorbells)
+		return atomic64_read((atomic64_t *)(adev->doorbell.cpu_addr + index));
+
+	dev_err(adev->dev, "reading beyond doorbell aperture: 0x%08x!\n",
+		index);
+	return 0;
+}
+
+/**
+ * amdgpu_mm_wdoorbell64 - write a doorbell Qword
+ *
+ * @adev: amdgpu_device pointer
+ * @index: doorbell index
+ * @v: value to write
+ *
+ * Writes @v to the doorbell aperture at the
+ * requested doorbell index (VEGA10+).
+ */
+void amdgpu_mm_wdoorbell64(struct amdgpu_device *adev, u32 index, u64 v)
+{
+	if (amdgpu_device_skip_hw_access(adev))
+		return;
+
+	if (index < adev->doorbell.num_kernel_doorbells)
+		atomic64_set((atomic64_t *)(adev->doorbell.cpu_addr + index), v);
+	else
+		dev_err(adev->dev,
+			"writing beyond doorbell aperture: 0x%08x!\n", index);
+}
+
+/**
+ * amdgpu_doorbell_index_on_bar - Find doorbell's absolute offset in BAR
+ *
+ * @adev: amdgpu_device pointer
+ * @db_bo: doorbell object's bo
+ * @doorbell_index: doorbell relative index in this doorbell object
+ * @db_size: doorbell size is in byte
+ *
+ * returns doorbell's absolute index in BAR
+ */
+uint32_t amdgpu_doorbell_index_on_bar(struct amdgpu_device *adev,
+				      struct amdgpu_bo *db_bo,
+				      uint32_t doorbell_index,
+				      uint32_t db_size)
+{
+	int db_bo_offset;
+
+	db_bo_offset = amdgpu_bo_gpu_offset_no_check(db_bo);
+
+	/* doorbell index is 32 bit but doorbell's size can be 32 bit
+	 * or 64 bit, so *db_size(in byte)/4 for alignment.
+	 */
+	return db_bo_offset / sizeof(u32) + doorbell_index *
+	       DIV_ROUND_UP(db_size, 4);
+}
+
+/**
+ * amdgpu_doorbell_create_kernel_doorbells - Create kernel doorbells for graphics
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Creates doorbells for graphics driver usages.
+ * returns 0 on success, error otherwise.
+ */
+int amdgpu_doorbell_create_kernel_doorbells(struct amdgpu_device *adev)
+{
+	int r;
+	int size;
+
+	/* SI HW does not have doorbells, skip allocation */
+	if (adev->doorbell.num_kernel_doorbells == 0)
+		return 0;
+
+	/* Reserve first num_kernel_doorbells (page-aligned) for kernel ops */
+	size = ALIGN(adev->doorbell.num_kernel_doorbells * sizeof(u32), PAGE_SIZE);
+
+	/* Allocate an extra page for MES kernel usages (ring test) */
+	adev->mes.db_start_dw_offset = size / sizeof(u32);
+	size += PAGE_SIZE;
+
+	r = amdgpu_bo_create_kernel(adev,
+				    size,
+				    PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_DOORBELL,
+				    &adev->doorbell.kernel_doorbells,
+				    NULL,
+				    (void **)&adev->doorbell.cpu_addr);
+	if (r) {
+		dev_err(adev->dev,
+			"Failed to allocate kernel doorbells, err=%d\n", r);
+		return r;
+	}
+
+	adev->doorbell.num_kernel_doorbells = size / sizeof(u32);
+	return 0;
+}
+
+/*
+ * GPU doorbell aperture helpers function.
+ */
+/**
+ * amdgpu_doorbell_init - Init doorbell driver information.
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Init doorbell driver information (CIK)
+ * Returns 0 on success, error on failure.
+ */
+int amdgpu_doorbell_init(struct amdgpu_device *adev)
+{
+
+	/* No doorbell on SI hardware generation */
+	if (adev->asic_type < CHIP_BONAIRE) {
+		adev->doorbell.base = 0;
+		adev->doorbell.size = 0;
+		adev->doorbell.num_kernel_doorbells = 0;
+		return 0;
+	}
+
+	if (pci_resource_flags(adev->pdev, 2) & IORESOURCE_UNSET)
+		return -EINVAL;
+
+	amdgpu_asic_init_doorbell_index(adev);
+
+	/* doorbell bar mapping */
+	adev->doorbell.base = pci_resource_start(adev->pdev, 2);
+	adev->doorbell.size = pci_resource_len(adev->pdev, 2);
+
+	adev->doorbell.num_kernel_doorbells =
+		min_t(u32, adev->doorbell.size / sizeof(u32),
+		      adev->doorbell_index.max_assignment + 1);
+	if (adev->doorbell.num_kernel_doorbells == 0)
+		return -EINVAL;
+
+	/*
+	 * For Vega, reserve and map two pages on doorbell BAR since SDMA
+	 * paging queue doorbell use the second page. The
+	 * AMDGPU_DOORBELL64_MAX_ASSIGNMENT definition assumes all the
+	 * doorbells are in the first page. So with paging queue enabled,
+	 * the max num_kernel_doorbells should + 1 page (0x400 in dword)
+	 */
+	if (adev->asic_type >= CHIP_VEGA10)
+		adev->doorbell.num_kernel_doorbells += 0x400;
+
+	return 0;
+}
+
+/**
+ * amdgpu_doorbell_fini - Tear down doorbell driver information.
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Tear down doorbell driver information (CIK)
+ */
+void amdgpu_doorbell_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_free_kernel(&adev->doorbell.kernel_doorbells,
+			      NULL,
+			      (void **)&adev->doorbell.cpu_addr);
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
index b4189d669b54..2dfbddcef9ab 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
@@ -23,33 +23,37 @@
  */
 
 #include <drm/amdgpu_drm.h>
+#include <drm/clients/drm_client_setup.h>
 #include <drm/drm_drv.h>
-#include <drm/drm_fbdev_generic.h>
+#include <drm/drm_fbdev_ttm.h>
 #include <drm/drm_gem.h>
-#include <drm/drm_vblank.h>
 #include <drm/drm_managed.h>
-#include "amdgpu_drv.h"
-
 #include <drm/drm_pciids.h>
-#include <linux/module.h>
-#include <linux/pm_runtime.h>
-#include <linux/vga_switcheroo.h>
 #include <drm/drm_probe_helper.h>
-#include <linux/mmu_notifier.h>
-#include <linux/suspend.h>
+#include <drm/drm_vblank.h>
+
 #include <linux/cc_platform.h>
 #include <linux/dynamic_debug.h>
+#include <linux/module.h>
+#include <linux/mmu_notifier.h>
+#include <linux/pm_runtime.h>
+#include <linux/suspend.h>
+#include <linux/vga_switcheroo.h>
 
 #include "amdgpu.h"
-#include "amdgpu_irq.h"
+#include "amdgpu_amdkfd.h"
 #include "amdgpu_dma_buf.h"
-#include "amdgpu_sched.h"
+#include "amdgpu_drv.h"
 #include "amdgpu_fdinfo.h"
-#include "amdgpu_amdkfd.h"
-
+#include "amdgpu_irq.h"
+#include "amdgpu_psp.h"
 #include "amdgpu_ras.h"
-#include "amdgpu_xgmi.h"
 #include "amdgpu_reset.h"
+#include "amdgpu_sched.h"
+#include "amdgpu_xgmi.h"
+#include "amdgpu_userq.h"
+#include "amdgpu_userq_fence.h"
+#include "../amdxcp/amdgpu_xcp_drv.h"
 
 /*
  * KMS wrapper.
@@ -110,11 +114,40 @@
  *   3.52.0 - Add AMDGPU_IDS_FLAGS_CONFORMANT_TRUNC_COORD, add device_info fields:
  *            tcp_cache_size, num_sqc_per_wgp, sqc_data_cache_size, sqc_inst_cache_size,
  *            gl1c_cache_size, gl2c_cache_size, mall_size, enabled_rb_pipes_mask_hi
+ *   3.53.0 - Support for GFX11 CP GFX shadowing
+ *   3.54.0 - Add AMDGPU_CTX_QUERY2_FLAGS_RESET_IN_PROGRESS support
+ * - 3.55.0 - Add AMDGPU_INFO_GPUVM_FAULT query
+ * - 3.56.0 - Update IB start address and size alignment for decode and encode
+ * - 3.57.0 - Compute tunneling on GFX10+
+ * - 3.58.0 - Add GFX12 DCC support
+ * - 3.59.0 - Cleared VRAM
+ * - 3.60.0 - Add AMDGPU_TILING_GFX12_DCC_WRITE_COMPRESS_DISABLE (Vulkan requirement)
+ * - 3.61.0 - Contains fix for RV/PCO compute queues
+ * - 3.62.0 - Add AMDGPU_IDS_FLAGS_MODE_PF, AMDGPU_IDS_FLAGS_MODE_VF & AMDGPU_IDS_FLAGS_MODE_PT
+ * - 3.63.0 - GFX12 display DCC supports 256B max compressed block size
+ * - 3.64.0 - Userq IP support query
  */
 #define KMS_DRIVER_MAJOR	3
-#define KMS_DRIVER_MINOR	52
+#define KMS_DRIVER_MINOR	64
 #define KMS_DRIVER_PATCHLEVEL	0
 
+/*
+ * amdgpu.debug module options. Are all disabled by default
+ */
+enum AMDGPU_DEBUG_MASK {
+	AMDGPU_DEBUG_VM = BIT(0),
+	AMDGPU_DEBUG_LARGEBAR = BIT(1),
+	AMDGPU_DEBUG_DISABLE_GPU_SOFT_RECOVERY = BIT(2),
+	AMDGPU_DEBUG_USE_VRAM_FW_BUF = BIT(3),
+	AMDGPU_DEBUG_ENABLE_RAS_ACA = BIT(4),
+	AMDGPU_DEBUG_ENABLE_EXP_RESETS = BIT(5),
+	AMDGPU_DEBUG_DISABLE_GPU_RING_RESET = BIT(6),
+	AMDGPU_DEBUG_SMU_POOL = BIT(7),
+	AMDGPU_DEBUG_VM_USERPTR = BIT(8),
+	AMDGPU_DEBUG_DISABLE_RAS_CE_LOG = BIT(9),
+	AMDGPU_DEBUG_ENABLE_CE_CS = BIT(10)
+};
+
 unsigned int amdgpu_vram_limit = UINT_MAX;
 int amdgpu_vis_vram_limit;
 int amdgpu_gart_size = -1; /* auto */
@@ -137,7 +170,6 @@ int amdgpu_vm_size = -1;
 int amdgpu_vm_fragment_size = -1;
 int amdgpu_vm_block_size = -1;
 int amdgpu_vm_fault_stop;
-int amdgpu_vm_debug;
 int amdgpu_vm_update_mode = -1;
 int amdgpu_exp_hw_support;
 int amdgpu_dc = -1;
@@ -150,6 +182,17 @@ uint amdgpu_pg_mask = 0xffffffff;
 uint amdgpu_sdma_phase_quantum = 32;
 char *amdgpu_disable_cu;
 char *amdgpu_virtual_display;
+int amdgpu_enforce_isolation = -1;
+int amdgpu_modeset = -1;
+
+/* Specifies the default granularity for SVM, used in buffer
+ * migration and restoration of backing memory when handling
+ * recoverable page faults.
+ *
+ * The value is given as log(numPages(buffer)); for a 2 MiB
+ * buffer it computes to be 9
+ */
+uint amdgpu_svm_default_granularity = 9;
 
 /*
  * OverDrive(bit 14) disabled by default
@@ -157,7 +200,6 @@ char *amdgpu_virtual_display;
  */
 uint amdgpu_pp_feature_mask = 0xfff7bfff;
 uint amdgpu_force_long_training;
-int amdgpu_job_hang_limit;
 int amdgpu_lbpw = -1;
 int amdgpu_compute_multipipe = -1;
 int amdgpu_gpu_recovery = -1; /* auto */
@@ -178,10 +220,12 @@ uint amdgpu_dc_feature_mask = 2;
 uint amdgpu_dc_debug_mask;
 uint amdgpu_dc_visual_confirm;
 int amdgpu_async_gfx_ring = 1;
-int amdgpu_mcbp;
+int amdgpu_mcbp = -1;
 int amdgpu_discovery = -1;
 int amdgpu_mes;
+int amdgpu_mes_log_enable = 0;
 int amdgpu_mes_kiq;
+int amdgpu_uni_mes = 1;
 int amdgpu_noretry = -1;
 int amdgpu_force_asic_type = -1;
 int amdgpu_tmz = -1; /* auto */
@@ -192,8 +236,16 @@ int amdgpu_smartshift_bias;
 int amdgpu_use_xgmi_p2p = 1;
 int amdgpu_vcnfw_log;
 int amdgpu_sg_display = -1; /* auto */
-
-static void amdgpu_drv_delayed_reset_work_handler(struct work_struct *work);
+int amdgpu_user_partt_mode = AMDGPU_AUTO_COMPUTE_PARTITION_MODE;
+int amdgpu_umsch_mm;
+int amdgpu_seamless = -1; /* auto */
+uint amdgpu_debug_mask;
+int amdgpu_agp = -1; /* auto */
+int amdgpu_wbrf = -1;
+int amdgpu_damage_clips = -1; /* auto */
+int amdgpu_umsch_mm_fwlog;
+int amdgpu_rebar = -1; /* auto */
+int amdgpu_user_queue = -1;
 
 DECLARE_DYNDBG_CLASSMAP(drm_debug_classes, DD_CLASS_TYPE_DISJOINT_BITS, 0,
 			"DRM_UT_CORE",
@@ -209,9 +261,6 @@ DECLARE_DYNDBG_CLASSMAP(drm_debug_classes, DD_CLASS_TYPE_DISJOINT_BITS, 0,
 
 struct amdgpu_mgpu_info mgpu_info = {
 	.mutex = __MUTEX_INITIALIZER(mgpu_info.mutex),
-	.delayed_reset_work = __DELAYED_WORK_INITIALIZER(
-			mgpu_info.delayed_reset_work,
-			amdgpu_drv_delayed_reset_work_handler, 0),
 };
 int amdgpu_ras_enable = -1;
 uint amdgpu_ras_mask = 0xffffffff;
@@ -246,7 +295,8 @@ module_param_named(gartsize, amdgpu_gart_size, uint, 0600);
 /**
  * DOC: gttsize (int)
  * Restrict the size of GTT domain (for userspace use) in MiB for testing.
- * The default is -1 (Use 1/2 RAM, minimum value is 3GB).
+ * The default is -1 (Use value specified by TTM).
+ * This parameter is deprecated and will be removed in the future.
  */
 MODULE_PARM_DESC(gttsize, "Size of the GTT userspace domain in megabytes (-1 = auto)");
 module_param_named(gttsize, amdgpu_gtt_size, int, 0600);
@@ -262,7 +312,7 @@ module_param_named(moverate, amdgpu_moverate, int, 0600);
  * DOC: audio (int)
  * Set HDMI/DPAudio. Only affects non-DC display handling. The default is -1 (Enabled), set 0 to disabled it.
  */
-MODULE_PARM_DESC(audio, "Audio enable (-1 = auto, 0 = disable, 1 = enable)");
+MODULE_PARM_DESC(audio, "HDMI/DP Audio enable for non DC displays (-1 = auto, 0 = disable, 1 = enable)");
 module_param_named(audio, amdgpu_audio, int, 0444);
 
 /**
@@ -294,27 +344,26 @@ MODULE_PARM_DESC(msi, "MSI support (1 = enable, 0 = disable, -1 = auto)");
 module_param_named(msi, amdgpu_msi, int, 0444);
 
 /**
+ * DOC: svm_default_granularity (uint)
+ * Used in buffer migration and handling of recoverable page faults
+ */
+MODULE_PARM_DESC(svm_default_granularity, "SVM's default granularity in log(2^Pages), default 9 = 2^9 = 2 MiB");
+module_param_named(svm_default_granularity, amdgpu_svm_default_granularity, uint, 0644);
+
+/**
  * DOC: lockup_timeout (string)
  * Set GPU scheduler timeout value in ms.
  *
- * The format can be [Non-Compute] or [GFX,Compute,SDMA,Video]. That is there can be one or
- * multiple values specified. 0 and negative values are invalidated. They will be adjusted
- * to the default timeout.
- *
- * - With one value specified, the setting will apply to all non-compute jobs.
- * - With multiple values specified, the first one will be for GFX.
- *   The second one is for Compute. The third and fourth ones are
- *   for SDMA and Video.
+ * The format can be [single value] for setting all timeouts at once or
+ * [GFX,Compute,SDMA,Video] to set individual timeouts.
+ * Negative values mean infinity.
  *
- * By default(with no lockup_timeout settings), the timeout for all non-compute(GFX, SDMA and Video)
- * jobs is 10000. The timeout for compute is 60000.
+ * By default(with no lockup_timeout settings), the timeout for all queues is 2000.
  */
-MODULE_PARM_DESC(lockup_timeout, "GPU lockup timeout in ms (default: for bare metal 10000 for non-compute jobs and 60000 for compute jobs; "
-		"for passthrough or sriov, 10000 for all jobs."
-		" 0: keep default value. negative: infinity timeout), "
-		"format: for bare metal [Non-Compute] or [GFX,Compute,SDMA,Video]; "
-		"for passthrough or sriov [all jobs] or [GFX,Compute,SDMA,Video].");
-module_param_string(lockup_timeout, amdgpu_lockup_timeout, sizeof(amdgpu_lockup_timeout), 0444);
+MODULE_PARM_DESC(lockup_timeout,
+		 "GPU lockup timeout in ms (default: 2000. 0: keep default value. negative: infinity timeout), format: [single value for all] or [GFX,Compute,SDMA,Video].");
+module_param_string(lockup_timeout, amdgpu_lockup_timeout,
+		    sizeof(amdgpu_lockup_timeout), 0444);
 
 /**
  * DOC: dpm (int)
@@ -347,8 +396,9 @@ module_param_named(aspm, amdgpu_aspm, int, 0444);
  * Override for runtime power management control for dGPUs. The amdgpu driver can dynamically power down
  * the dGPUs when they are idle if supported. The default is -1 (auto enable).
  * Setting the value to 0 disables this functionality.
+ * Setting the value to -2 is auto enabled with power down when displays are attached.
  */
-MODULE_PARM_DESC(runpm, "PX runtime pm (2 = force enable with BAMACO, 1 = force enable with BACO, 0 = disable, -1 = auto)");
+MODULE_PARM_DESC(runpm, "PX runtime pm (2 = force enable with BAMACO, 1 = force enable with BACO, 0 = disable, -1 = auto, -2 = auto with displays)");
 module_param_named(runpm, amdgpu_runtime_pm, int, 0444);
 
 /**
@@ -359,7 +409,7 @@ module_param_named(runpm, amdgpu_runtime_pm, int, 0444);
  * the kernel log for the list of IPs on the asic. The default is 0xffffffff (enable all blocks on a device).
  */
 MODULE_PARM_DESC(ip_block_mask, "IP Block Mask (all blocks enabled (default))");
-module_param_named(ip_block_mask, amdgpu_ip_block_mask, uint, 0444);
+module_param_named_unsafe(ip_block_mask, amdgpu_ip_block_mask, uint, 0444);
 
 /**
  * DOC: bapm (int)
@@ -405,13 +455,6 @@ MODULE_PARM_DESC(vm_fault_stop, "Stop on VM fault (0 = never (default), 1 = prin
 module_param_named(vm_fault_stop, amdgpu_vm_fault_stop, int, 0444);
 
 /**
- * DOC: vm_debug (int)
- * Debug VM handling (0 = disabled, 1 = enabled). The default is 0 (Disabled).
- */
-MODULE_PARM_DESC(vm_debug, "Debug VM handling (0 = disabled (default), 1 = enabled)");
-module_param_named(vm_debug, amdgpu_vm_debug, int, 0644);
-
-/**
  * DOC: vm_update_mode (int)
  * Override VM update mode. VM updated by using CPU (0 = never, 1 = Graphics only, 2 = Compute only, 3 = Both). The default
  * is -1 (Only in large BAR(LB) systems Compute VM tables will be updated by CPU, otherwise 0, never).
@@ -424,7 +467,7 @@ module_param_named(vm_update_mode, amdgpu_vm_update_mode, int, 0444);
  * Enable experimental hw support (1 = enable). The default is 0 (disabled).
  */
 MODULE_PARM_DESC(exp_hw_support, "experimental hw support (1 = enable, 0 = disable (default))");
-module_param_named(exp_hw_support, amdgpu_exp_hw_support, int, 0444);
+module_param_named_unsafe(exp_hw_support, amdgpu_exp_hw_support, int, 0444);
 
 /**
  * DOC: dc (int)
@@ -521,13 +564,6 @@ MODULE_PARM_DESC(virtual_display,
 module_param_named(virtual_display, amdgpu_virtual_display, charp, 0444);
 
 /**
- * DOC: job_hang_limit (int)
- * Set how much time allow a job hang and not drop it. The default is 0.
- */
-MODULE_PARM_DESC(job_hang_limit, "how much time allow a job hang and not drop it (default 0)");
-module_param_named(job_hang_limit, amdgpu_job_hang_limit, int ,0444);
-
-/**
  * DOC: lbpw (int)
  * Override Load Balancing Per Watt (LBPW) support (1 = enable, 0 = disable). The default is -1 (auto, enabled).
  */
@@ -542,14 +578,14 @@ module_param_named(compute_multipipe, amdgpu_compute_multipipe, int, 0444);
  * Set to enable GPU recovery mechanism (1 = enable, 0 = disable). The default is -1 (auto, disabled except SRIOV).
  */
 MODULE_PARM_DESC(gpu_recovery, "Enable GPU recovery mechanism, (1 = enable, 0 = disable, -1 = auto)");
-module_param_named(gpu_recovery, amdgpu_gpu_recovery, int, 0444);
+module_param_named_unsafe(gpu_recovery, amdgpu_gpu_recovery, int, 0444);
 
 /**
  * DOC: emu_mode (int)
  * Set value 1 to enable emulation mode. This is only needed when running on an emulator. The default is 0 (disabled).
  */
 MODULE_PARM_DESC(emu_mode, "Emulation mode, (1 = enable, 0 = disable)");
-module_param_named(emu_mode, amdgpu_emu_mode, int, 0444);
+module_param_named_unsafe(emu_mode, amdgpu_emu_mode, int, 0444);
 
 /**
  * DOC: ras_enable (int)
@@ -582,39 +618,39 @@ module_param_named(timeout_period, amdgpu_watchdog_timer.period, uint, 0644);
 
 /**
  * DOC: si_support (int)
- * Set SI support driver. This parameter works after set config CONFIG_DRM_AMDGPU_SI. For SI asic, when radeon driver is enabled,
- * set value 0 to use radeon driver, while set value 1 to use amdgpu driver. The default is using radeon driver when it available,
- * otherwise using amdgpu driver.
- */
+ * 1 = enabled, 0 = disabled, -1 = default
+ *
+ * SI (Southern Islands) are first generation GCN GPUs, supported by both
+ * drivers: radeon (old) and amdgpu (new). This parameter controls whether
+ * amdgpu should support SI.
+ * By default, SI dedicated GPUs are supported by amdgpu.
+ * Only relevant when CONFIG_DRM_AMDGPU_SI is enabled to build SI support in amdgpu.
+ * See also radeon.si_support which should be disabled when amdgpu.si_support is
+ * enabled, and vice versa.
+ */
+int amdgpu_si_support = -1;
 #ifdef CONFIG_DRM_AMDGPU_SI
-
-#if defined(CONFIG_DRM_RADEON) || defined(CONFIG_DRM_RADEON_MODULE)
-int amdgpu_si_support = 0;
-MODULE_PARM_DESC(si_support, "SI support (1 = enabled, 0 = disabled (default))");
-#else
-int amdgpu_si_support = 1;
-MODULE_PARM_DESC(si_support, "SI support (1 = enabled (default), 0 = disabled)");
-#endif
-
+MODULE_PARM_DESC(si_support, "SI support (1 = enabled, 0 = disabled, -1 = default)");
 module_param_named(si_support, amdgpu_si_support, int, 0444);
 #endif
 
 /**
  * DOC: cik_support (int)
- * Set CIK support driver. This parameter works after set config CONFIG_DRM_AMDGPU_CIK. For CIK asic, when radeon driver is enabled,
- * set value 0 to use radeon driver, while set value 1 to use amdgpu driver. The default is using radeon driver when it available,
- * otherwise using amdgpu driver.
- */
+ * 1 = enabled, 0 = disabled, -1 = default
+ *
+ * CIK (Sea Islands) are second generation GCN GPUs, supported by both
+ * drivers: radeon (old) and amdgpu (new). This parameter controls whether
+ * amdgpu should support CIK.
+ * By default:
+ * - CIK dedicated GPUs are supported by amdgpu.
+ * - CIK APUs are supported by radeon (except when radeon is not built).
+ * Only relevant when CONFIG_DRM_AMDGPU_CIK is enabled to build CIK support in amdgpu.
+ * See also radeon.cik_support which should be disabled when amdgpu.cik_support is
+ * enabled, and vice versa.
+ */
+int amdgpu_cik_support = -1;
 #ifdef CONFIG_DRM_AMDGPU_CIK
-
-#if defined(CONFIG_DRM_RADEON) || defined(CONFIG_DRM_RADEON_MODULE)
-int amdgpu_cik_support = 0;
-MODULE_PARM_DESC(cik_support, "CIK support (1 = enabled, 0 = disabled (default))");
-#else
-int amdgpu_cik_support = 1;
-MODULE_PARM_DESC(cik_support, "CIK support (1 = enabled (default), 0 = disabled)");
-#endif
-
+MODULE_PARM_DESC(cik_support, "CIK support  (1 = enabled, 0 = disabled, -1 = default)");
 module_param_named(cik_support, amdgpu_cik_support, int, 0444);
 #endif
 
@@ -624,8 +660,7 @@ module_param_named(cik_support, amdgpu_cik_support, int, 0444);
  * E.g. 0x1 = 256Mbyte, 0x2 = 512Mbyte, 0x4 = 1 Gbyte, 0x8 = 2GByte. The default is 0 (disabled).
  */
 MODULE_PARM_DESC(smu_memory_pool_size,
-	"reserve gtt for smu debug usage, 0 = disable,"
-		"0x1 = 256Mbyte, 0x2 = 512Mbyte, 0x4 = 1 Gbyte, 0x8 = 2GByte");
+	"reserve gtt for smu debug usage, 0 = disable,0x1 = 256Mbyte, 0x2 = 512Mbyte, 0x4 = 1 Gbyte, 0x8 = 2GByte");
 module_param_named(smu_memory_pool_size, amdgpu_smu_memory_pool_size, uint, 0444);
 
 /**
@@ -638,10 +673,10 @@ module_param_named(async_gfx_ring, amdgpu_async_gfx_ring, int, 0444);
 
 /**
  * DOC: mcbp (int)
- * It is used to enable mid command buffer preemption. (0 = disabled (default), 1 = enabled)
+ * It is used to enable mid command buffer preemption. (0 = disabled, 1 = enabled, -1 auto (default))
  */
 MODULE_PARM_DESC(mcbp,
-	"Enable Mid-command buffer preemption (0 = disabled (default), 1 = enabled)");
+	"Enable Mid-command buffer preemption (0 = disabled, 1 = enabled), -1 = auto (default)");
 module_param_named(mcbp, amdgpu_mcbp, int, 0444);
 
 /**
@@ -663,6 +698,15 @@ MODULE_PARM_DESC(mes,
 module_param_named(mes, amdgpu_mes, int, 0444);
 
 /**
+ * DOC: mes_log_enable (int)
+ * Enable Micro Engine Scheduler log. This is used to enable/disable MES internal log.
+ * (0 = disabled (default), 1 = enabled)
+ */
+MODULE_PARM_DESC(mes_log_enable,
+	"Enable Micro Engine Scheduler log (0 = disabled (default), 1 = enabled)");
+module_param_named(mes_log_enable, amdgpu_mes_log_enable, int, 0444);
+
+/**
  * DOC: mes_kiq (int)
  * Enable Micro Engine Scheduler KIQ. This is a new engine pipe for kiq.
  * (0 = disabled (default), 1 = enabled)
@@ -672,6 +716,15 @@ MODULE_PARM_DESC(mes_kiq,
 module_param_named(mes_kiq, amdgpu_mes_kiq, int, 0444);
 
 /**
+ * DOC: uni_mes (int)
+ * Enable Unified Micro Engine Scheduler. This is a new engine pipe for unified scheduler.
+ * (0 = disabled (default), 1 = enabled)
+ */
+MODULE_PARM_DESC(uni_mes,
+	"Enable Unified Micro Engine Scheduler (0 = disabled, 1 = enabled(default)");
+module_param_named(uni_mes, amdgpu_uni_mes, int, 0444);
+
+/**
  * DOC: noretry (int)
  * Disable XNACK retry in the SQ by default on GFXv9 hardware. On ASICs that
  * do not support per-process XNACK this also disables retry page faults.
@@ -687,7 +740,7 @@ module_param_named(noretry, amdgpu_noretry, int, 0644);
  */
 MODULE_PARM_DESC(force_asic_type,
 	"A non negative value used to specify the asic type for all supported GPUs");
-module_param_named(force_asic_type, amdgpu_force_asic_type, int, 0444);
+module_param_named_unsafe(force_asic_type, amdgpu_force_asic_type, int, 0444);
 
 /**
  * DOC: use_xgmi_p2p (int)
@@ -706,7 +759,7 @@ module_param_named(use_xgmi_p2p, amdgpu_use_xgmi_p2p, int, 0444);
  * assigns queues to HQDs.
  */
 int sched_policy = KFD_SCHED_POLICY_HWS;
-module_param(sched_policy, int, 0444);
+module_param_unsafe(sched_policy, int, 0444);
 MODULE_PARM_DESC(sched_policy,
 	"Scheduling policy (0 = HWS (Default), 1 = HWS without over-subscription, 2 = Non-HWS (Used for debugging only)");
 
@@ -751,38 +804,12 @@ MODULE_PARM_DESC(send_sigterm,
 	"Send sigterm to HSA process on unhandled exception (0 = disable, 1 = enable)");
 
 /**
- * DOC: debug_largebar (int)
- * Set debug_largebar as 1 to enable simulating large-bar capability on non-large bar
- * system. This limits the VRAM size reported to ROCm applications to the visible
- * size, usually 256MB.
- * Default value is 0, diabled.
- */
-int debug_largebar;
-module_param(debug_largebar, int, 0444);
-MODULE_PARM_DESC(debug_largebar,
-	"Debug large-bar flag used to simulate large-bar capability on non-large bar machine (0 = disable, 1 = enable)");
-
-/**
- * DOC: ignore_crat (int)
- * Ignore CRAT table during KFD initialization. By default, KFD uses the ACPI CRAT
- * table to get information about AMD APUs. This option can serve as a workaround on
- * systems with a broken CRAT table.
- *
- * Default is auto (according to asic type, iommu_v2, and crat table, to decide
- * whether use CRAT)
- */
-int ignore_crat;
-module_param(ignore_crat, int, 0444);
-MODULE_PARM_DESC(ignore_crat,
-	"Ignore CRAT table during KFD initialization (0 = auto (default), 1 = ignore CRAT)");
-
-/**
  * DOC: halt_if_hws_hang (int)
  * Halt if HWS hang is detected. Default value, 0, disables the halt on hang.
  * Setting 1 enables halt on hang.
  */
 int halt_if_hws_hang;
-module_param(halt_if_hws_hang, int, 0644);
+module_param_unsafe(halt_if_hws_hang, int, 0644);
 MODULE_PARM_DESC(halt_if_hws_hang, "Halt if HWS hang is detected (0 = off (default), 1 = on)");
 
 /**
@@ -791,13 +818,13 @@ MODULE_PARM_DESC(halt_if_hws_hang, "Halt if HWS hang is detected (0 = off (defau
  * check says. Default value: false (rely on MEC2 firmware version check).
  */
 bool hws_gws_support;
-module_param(hws_gws_support, bool, 0444);
+module_param_unsafe(hws_gws_support, bool, 0444);
 MODULE_PARM_DESC(hws_gws_support, "Assume MEC2 FW supports GWS barriers (false = rely on FW version check (Default), true = force supported)");
 
 /**
-  * DOC: queue_preemption_timeout_ms (int)
-  * queue preemption timeout in ms (1 = Minimum, 9000 = default)
-  */
+ * DOC: queue_preemption_timeout_ms (int)
+ * queue preemption timeout in ms (1 = Minimum, 9000 = default)
+ */
 int queue_preemption_timeout_ms = 9000;
 module_param(queue_preemption_timeout_ms, int, 0644);
 MODULE_PARM_DESC(queue_preemption_timeout_ms, "queue preemption timeout in ms (1 = Minimum, 9000 = default)");
@@ -824,10 +851,17 @@ MODULE_PARM_DESC(no_system_mem_limit, "disable system memory limit (false = defa
  */
 int amdgpu_no_queue_eviction_on_vm_fault;
 MODULE_PARM_DESC(no_queue_eviction_on_vm_fault, "No queue eviction on VM fault (0 = queue eviction, 1 = no queue eviction)");
-module_param_named(no_queue_eviction_on_vm_fault, amdgpu_no_queue_eviction_on_vm_fault, int, 0444);
+module_param_named_unsafe(no_queue_eviction_on_vm_fault, amdgpu_no_queue_eviction_on_vm_fault, int, 0444);
 #endif
 
 /**
+ * DOC: mtype_local (int)
+ */
+int amdgpu_mtype_local;
+MODULE_PARM_DESC(mtype_local, "MTYPE for local memory (0 = MTYPE_RW (default), 1 = MTYPE_NC, 2 = MTYPE_CC)");
+module_param_named_unsafe(mtype_local, amdgpu_mtype_local, int, 0444);
+
+/**
  * DOC: pcie_p2p (bool)
  * Enable PCIe P2P (requires large-BAR). Default value: true (on)
  */
@@ -847,7 +881,7 @@ module_param_named(dcfeaturemask, amdgpu_dc_feature_mask, uint, 0444);
 
 /**
  * DOC: dcdebugmask (uint)
- * Override display features enabled. See enum DC_DEBUG_MASK in drivers/gpu/drm/amd/include/amd_shared.h.
+ * Display debug options. See enum DC_DEBUG_MASK in drivers/gpu/drm/amd/include/amd_shared.h.
  */
 MODULE_PARM_DESC(dcdebugmask, "all debug options disabled (default))");
 module_param_named(dcdebugmask, amdgpu_dc_debug_mask, uint, 0444);
@@ -864,18 +898,31 @@ module_param_named(visualconfirm, amdgpu_dc_visual_confirm, uint, 0444);
  * the ABM algorithm, with 1 being the least reduction and 4 being the most
  * reduction.
  *
- * Defaults to 0, or disabled. Userspace can still override this level later
- * after boot.
+ * Defaults to -1, or auto. Userspace can only override this level after
+ * boot if it's set to auto.
  */
-uint amdgpu_dm_abm_level;
-MODULE_PARM_DESC(abmlevel, "ABM level (0 = off (default), 1-4 = backlight reduction level) ");
-module_param_named(abmlevel, amdgpu_dm_abm_level, uint, 0444);
+int amdgpu_dm_abm_level = -1;
+MODULE_PARM_DESC(abmlevel,
+		 "ABM level (0 = off, 1-4 = backlight reduction level, -1 auto (default))");
+module_param_named(abmlevel, amdgpu_dm_abm_level, int, 0444);
 
 int amdgpu_backlight = -1;
 MODULE_PARM_DESC(backlight, "Backlight control (0 = pwm, 1 = aux, -1 auto (default))");
 module_param_named(backlight, amdgpu_backlight, bint, 0444);
 
 /**
+ * DOC: damageclips (int)
+ * Enable or disable damage clips support. If damage clips support is disabled,
+ * we will force full frame updates, irrespective of what user space sends to
+ * us.
+ *
+ * Defaults to -1 (where it is enabled unless a PSR-SU display is detected).
+ */
+MODULE_PARM_DESC(damageclips,
+		 "Damage clips support (0 = disable, 1 = enable, -1 auto (default))");
+module_param_named(damageclips, amdgpu_damage_clips, int, 0444);
+
+/**
  * DOC: tmz (int)
  * Trusted Memory Zone (TMZ) is a method to protect data being written
  * to or read from memory.
@@ -908,7 +955,7 @@ module_param_named(tmz, amdgpu_tmz, int, 0444);
  */
 MODULE_PARM_DESC(
 	freesync_video,
-	"Enable freesync modesetting optimization feature (0 = off (default), 1 = on)");
+	"Adds additional modes via VRR for refresh changes without a full modeset (0 = off (default), 1 = on)");
 module_param_named(freesync_video, amdgpu_freesync_vid_mode, uint, 0444);
 
 /**
@@ -916,7 +963,7 @@ module_param_named(freesync_video, amdgpu_freesync_vid_mode, uint, 0444);
  * GPU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 = mode2, 4 = baco)
  */
 MODULE_PARM_DESC(reset_method, "GPU reset method (-1 = auto (default), 0 = legacy, 1 = mode0, 2 = mode1, 3 = mode2, 4 = baco/bamaco)");
-module_param_named(reset_method, amdgpu_reset_method, int, 0444);
+module_param_named_unsafe(reset_method, amdgpu_reset_method, int, 0644);
 
 /**
  * DOC: bad_page_threshold (int) Bad page threshold is specifies the
@@ -924,7 +971,7 @@ module_param_named(reset_method, amdgpu_reset_method, int, 0444);
  * result in the GPU entering bad status when the number of total
  * faulty pages by ECC exceeds the threshold value.
  */
-MODULE_PARM_DESC(bad_page_threshold, "Bad page threshold(-1 = ignore threshold (default value), 0 = disable bad page retirement, -2 = driver sets threshold)");
+MODULE_PARM_DESC(bad_page_threshold, "Bad page threshold(-1 = ignore threshold (default value), 0 = disable bad page retirement, -2 = threshold determined by a formula, 0 < threshold < max records, user-defined threshold)");
 module_param_named(bad_page_threshold, amdgpu_bad_page_threshold, int, 0444);
 
 MODULE_PARM_DESC(num_kcq, "number of kernel compute queue user want to setup (8 if set to greater than 8 or less than 0, only affect gfx 8+)");
@@ -948,6 +995,22 @@ MODULE_PARM_DESC(sg_display, "S/G Display (-1 = auto (default), 0 = disable)");
 module_param_named(sg_display, amdgpu_sg_display, int, 0444);
 
 /**
+ * DOC: umsch_mm (int)
+ * Enable Multi Media User Mode Scheduler. This is a HW scheduling engine for VCN and VPE.
+ * (0 = disabled (default), 1 = enabled)
+ */
+MODULE_PARM_DESC(umsch_mm,
+	"Enable Multi Media User Mode Scheduler (0 = disabled (default), 1 = enabled)");
+module_param_named(umsch_mm, amdgpu_umsch_mm, int, 0444);
+
+/**
+ * DOC: umsch_mm_fwlog (int)
+ * Enable umschfw log output for debugging, the default is disabled.
+ */
+MODULE_PARM_DESC(umsch_mm_fwlog, "Enable umschfw log(0 = disable (default value), 1 = enable)");
+module_param_named(umsch_mm_fwlog, amdgpu_umsch_mm_fwlog, int, 0444);
+
+/**
  * DOC: smu_pptable_id (int)
  * Used to override pptable id. id = 0 use VBIOS pptable.
  * id > 0 use the soft pptable with specicfied id.
@@ -956,6 +1019,109 @@ MODULE_PARM_DESC(smu_pptable_id,
 	"specify pptable id to be used (-1 = auto(default) value, 0 = use pptable from vbios, > 0 = soft pptable id)");
 module_param_named(smu_pptable_id, amdgpu_smu_pptable_id, int, 0444);
 
+/**
+ * DOC: partition_mode (int)
+ * Used to override the default SPX mode.
+ */
+MODULE_PARM_DESC(
+	user_partt_mode,
+	"specify partition mode to be used (-2 = AMDGPU_AUTO_COMPUTE_PARTITION_MODE(default value) \
+						0 = AMDGPU_SPX_PARTITION_MODE, \
+						1 = AMDGPU_DPX_PARTITION_MODE, \
+						2 = AMDGPU_TPX_PARTITION_MODE, \
+						3 = AMDGPU_QPX_PARTITION_MODE, \
+						4 = AMDGPU_CPX_PARTITION_MODE)");
+module_param_named(user_partt_mode, amdgpu_user_partt_mode, uint, 0444);
+
+
+/**
+ * DOC: enforce_isolation (int)
+ * enforce process isolation between graphics and compute.
+ * (-1 = auto, 0 = disable, 1 = enable, 2 = enable legacy mode, 3 = enable without cleaner shader)
+ */
+module_param_named(enforce_isolation, amdgpu_enforce_isolation, int, 0444);
+MODULE_PARM_DESC(enforce_isolation,
+"enforce process isolation between graphics and compute. (-1 = auto, 0 = disable, 1 = enable, 2 = enable legacy mode, 3 = enable without cleaner shader)");
+
+/**
+ * DOC: modeset (int)
+ * Override nomodeset (1 = override, -1 = auto). The default is -1 (auto).
+ */
+MODULE_PARM_DESC(modeset, "Override nomodeset (1 = enable, -1 = auto)");
+module_param_named(modeset, amdgpu_modeset, int, 0444);
+
+/**
+ * DOC: seamless (int)
+ * Seamless boot will keep the image on the screen during the boot process.
+ */
+MODULE_PARM_DESC(seamless, "Seamless boot (-1 = auto (default), 0 = disable, 1 = enable)");
+module_param_named(seamless, amdgpu_seamless, int, 0444);
+
+/**
+ * DOC: debug_mask (uint)
+ * Debug options for amdgpu, work as a binary mask with the following options:
+ *
+ * - 0x1: Debug VM handling
+ * - 0x2: Enable simulating large-bar capability on non-large bar system. This
+ *   limits the VRAM size reported to ROCm applications to the visible
+ *   size, usually 256MB.
+ * - 0x4: Disable GPU soft recovery, always do a full reset
+ * - 0x8: Use VRAM for firmware loading
+ * - 0x10: Enable ACA based RAS logging
+ * - 0x20: Enable experimental resets
+ * - 0x40: Disable ring resets
+ * - 0x80: Use VRAM for SMU pool
+ */
+MODULE_PARM_DESC(debug_mask, "debug options for amdgpu, disabled by default");
+module_param_named_unsafe(debug_mask, amdgpu_debug_mask, uint, 0444);
+
+/**
+ * DOC: agp (int)
+ * Enable the AGP aperture.  This provides an aperture in the GPU's internal
+ * address space for direct access to system memory.  Note that these accesses
+ * are non-snooped, so they are only used for access to uncached memory.
+ */
+MODULE_PARM_DESC(agp, "AGP (-1 = auto (default), 0 = disable, 1 = enable)");
+module_param_named(agp, amdgpu_agp, int, 0444);
+
+/**
+ * DOC: wbrf (int)
+ * Enable Wifi RFI interference mitigation feature.
+ * Due to electrical and mechanical constraints there may be likely interference of
+ * relatively high-powered harmonics of the (G-)DDR memory clocks with local radio
+ * module frequency bands used by Wifi 6/6e/7. To mitigate the possible RFI interference,
+ * with this feature enabled, PMFW will use either “shadowed P-State” or “P-State” based
+ * on active list of frequencies in-use (to be avoided) as part of initial setting or
+ * P-state transition. However, there may be potential performance impact with this
+ * feature enabled.
+ * (0 = disabled, 1 = enabled, -1 = auto (default setting, will be enabled if supported))
+ */
+MODULE_PARM_DESC(wbrf,
+	"Enable Wifi RFI interference mitigation (0 = disabled, 1 = enabled, -1 = auto(default)");
+module_param_named(wbrf, amdgpu_wbrf, int, 0444);
+
+/**
+ * DOC: rebar (int)
+ * Allow BAR resizing.  Disable this to prevent the driver from attempting
+ * to resize the BAR if the GPU supports it and there is available MMIO space.
+ * Note that this just prevents the driver from resizing the BAR.  The BIOS
+ * may have already resized the BAR at boot time.
+ */
+MODULE_PARM_DESC(rebar, "Resizable BAR (-1 = auto (default), 0 = disable, 1 = enable)");
+module_param_named(rebar, amdgpu_rebar, int, 0444);
+
+/**
+ * DOC: user_queue (int)
+ * Enable user queues on systems that support user queues. Possible values:
+ *
+ * - -1 = auto (ASIC specific default)
+ * -  0 = user queues disabled
+ * -  1 = user queues enabled and kernel queues enabled (if supported)
+ * -  2 = user queues enabled and kernel queues disabled
+ */
+MODULE_PARM_DESC(user_queue, "Enable user queues (-1 = auto (default), 0 = disable, 1 = enable, 2 = enable UQs and disable KQs)");
+module_param_named(user_queue, amdgpu_user_queue, int, 0444);
+
 /* These devices are not supported by amdgpu.
  * They are supported by the mach64, r128, radeon drivers
  */
@@ -1623,6 +1789,7 @@ static const u16 amdgpu_unsupported_pciidlist[] = {
 	0x5874,
 	0x5940,
 	0x5941,
+	0x5b70,
 	0x5b72,
 	0x5b73,
 	0x5b74,
@@ -1668,7 +1835,6 @@ static const u16 amdgpu_unsupported_pciidlist[] = {
 };
 
 static const struct pci_device_id pciidlist[] = {
-#ifdef  CONFIG_DRM_AMDGPU_SI
 	{0x1002, 0x6780, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TAHITI},
 	{0x1002, 0x6784, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TAHITI},
 	{0x1002, 0x6788, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TAHITI},
@@ -1741,8 +1907,6 @@ static const struct pci_device_id pciidlist[] = {
 	{0x1002, 0x6665, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_HAINAN|AMD_IS_MOBILITY},
 	{0x1002, 0x6667, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_HAINAN|AMD_IS_MOBILITY},
 	{0x1002, 0x666F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_HAINAN|AMD_IS_MOBILITY},
-#endif
-#ifdef CONFIG_DRM_AMDGPU_CIK
 	/* Kaveri */
 	{0x1002, 0x1304, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_KAVERI|AMD_IS_MOBILITY|AMD_IS_APU},
 	{0x1002, 0x1305, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_KAVERI|AMD_IS_APU},
@@ -1825,7 +1989,6 @@ static const struct pci_device_id pciidlist[] = {
 	{0x1002, 0x985D, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_MULLINS|AMD_IS_MOBILITY|AMD_IS_APU},
 	{0x1002, 0x985E, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_MULLINS|AMD_IS_MOBILITY|AMD_IS_APU},
 	{0x1002, 0x985F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_MULLINS|AMD_IS_MOBILITY|AMD_IS_APU},
-#endif
 	/* topaz */
 	{0x1002, 0x6900, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TOPAZ},
 	{0x1002, 0x6901, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TOPAZ},
@@ -2004,6 +2167,11 @@ static const struct pci_device_id pciidlist[] = {
 	{0x1002, 0x7410, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_ALDEBARAN},
 
 	/* CYAN_SKILLFISH */
+	{0x1002, 0x13DB, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CYAN_SKILLFISH|AMD_IS_APU},
+	{0x1002, 0x13F9, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CYAN_SKILLFISH|AMD_IS_APU},
+	{0x1002, 0x13FA, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CYAN_SKILLFISH|AMD_IS_APU},
+	{0x1002, 0x13FB, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CYAN_SKILLFISH|AMD_IS_APU},
+	{0x1002, 0x13FC, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CYAN_SKILLFISH|AMD_IS_APU},
 	{0x1002, 0x13FE, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CYAN_SKILLFISH|AMD_IS_APU},
 	{0x1002, 0x143F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_CYAN_SKILLFISH|AMD_IS_APU},
 
@@ -2025,11 +2193,24 @@ static const struct pci_device_id pciidlist[] = {
 	  .class_mask = 0xffffff,
 	  .driver_data = CHIP_IP_DISCOVERY },
 
+	{ PCI_DEVICE(0x1002, PCI_ANY_ID),
+	  .class = PCI_CLASS_ACCELERATOR_PROCESSING << 8,
+	  .class_mask = 0xffffff,
+	  .driver_data = CHIP_IP_DISCOVERY },
+
 	{0, 0, 0}
 };
 
 MODULE_DEVICE_TABLE(pci, pciidlist);
 
+static const struct amdgpu_asic_type_quirk asic_type_quirks[] = {
+	/* differentiate between P10 and P11 asics with the same DID */
+	{0x67FF, 0xE3, CHIP_POLARIS10},
+	{0x67FF, 0xE7, CHIP_POLARIS10},
+	{0x67FF, 0xF3, CHIP_POLARIS10},
+	{0x67FF, 0xF7, CHIP_POLARIS10},
+};
+
 static const struct drm_driver amdgpu_kms_driver;
 
 static void amdgpu_get_secondary_funcs(struct amdgpu_device *adev)
@@ -2047,13 +2228,150 @@ static void amdgpu_get_secondary_funcs(struct amdgpu_device *adev)
 						adev->pdev->bus->number, i);
 		if (p) {
 			pm_runtime_get_sync(&p->dev);
-			pm_runtime_mark_last_busy(&p->dev);
 			pm_runtime_put_autosuspend(&p->dev);
 			pci_dev_put(p);
 		}
 	}
 }
 
+static void amdgpu_init_debug_options(struct amdgpu_device *adev)
+{
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_VM) {
+		pr_info("debug: VM handling debug enabled\n");
+		adev->debug_vm = true;
+	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_LARGEBAR) {
+		pr_info("debug: enabled simulating large-bar capability on non-large bar system\n");
+		adev->debug_largebar = true;
+	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_DISABLE_GPU_SOFT_RECOVERY) {
+		pr_info("debug: soft reset for GPU recovery disabled\n");
+		adev->debug_disable_soft_recovery = true;
+	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_USE_VRAM_FW_BUF) {
+		pr_info("debug: place fw in vram for frontdoor loading\n");
+		adev->debug_use_vram_fw_buf = true;
+	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_ENABLE_RAS_ACA) {
+		pr_info("debug: enable RAS ACA\n");
+		adev->debug_enable_ras_aca = true;
+	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_ENABLE_EXP_RESETS) {
+		pr_info("debug: enable experimental reset features\n");
+		adev->debug_exp_resets = true;
+	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_DISABLE_GPU_RING_RESET) {
+		pr_info("debug: ring reset disabled\n");
+		adev->debug_disable_gpu_ring_reset = true;
+	}
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_SMU_POOL) {
+		pr_info("debug: use vram for smu pool\n");
+		adev->pm.smu_debug_mask |= SMU_DEBUG_POOL_USE_VRAM;
+	}
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_VM_USERPTR) {
+		pr_info("debug: VM mode debug for userptr is enabled\n");
+		adev->debug_vm_userptr = true;
+	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_DISABLE_RAS_CE_LOG) {
+		pr_info("debug: disable kernel logs of correctable errors\n");
+		adev->debug_disable_ce_logs = true;
+	}
+
+	if (amdgpu_debug_mask & AMDGPU_DEBUG_ENABLE_CE_CS) {
+		pr_info("debug: allowing command submission to CE engine\n");
+		adev->debug_enable_ce_cs = true;
+	}
+}
+
+static unsigned long amdgpu_fix_asic_type(struct pci_dev *pdev, unsigned long flags)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(asic_type_quirks); i++) {
+		if (pdev->device == asic_type_quirks[i].device &&
+			pdev->revision == asic_type_quirks[i].revision) {
+				flags &= ~AMD_ASIC_MASK;
+				flags |= asic_type_quirks[i].type;
+				break;
+			}
+	}
+
+	return flags;
+}
+
+static bool amdgpu_support_enabled(struct device *dev,
+				   const enum amd_asic_type family)
+{
+	const char *gen;
+	const char *param;
+	int module_param = -1;
+	bool radeon_support_built = IS_ENABLED(CONFIG_DRM_RADEON);
+	bool amdgpu_support_built = false;
+	bool support_by_default = false;
+
+	switch (family) {
+	case CHIP_TAHITI:
+	case CHIP_PITCAIRN:
+	case CHIP_VERDE:
+	case CHIP_OLAND:
+	case CHIP_HAINAN:
+		gen = "SI";
+		param = "si_support";
+		module_param = amdgpu_si_support;
+		amdgpu_support_built = IS_ENABLED(CONFIG_DRM_AMDGPU_SI);
+		support_by_default = true;
+		break;
+
+	case CHIP_BONAIRE:
+	case CHIP_HAWAII:
+		support_by_default = true;
+		fallthrough;
+	case CHIP_KAVERI:
+	case CHIP_KABINI:
+	case CHIP_MULLINS:
+		gen = "CIK";
+		param = "cik_support";
+		module_param = amdgpu_cik_support;
+		amdgpu_support_built = IS_ENABLED(CONFIG_DRM_AMDGPU_CIK);
+		break;
+
+	default:
+		/* All other chips are supported by amdgpu only */
+		return true;
+	}
+
+	if (!amdgpu_support_built) {
+		dev_info(dev, "amdgpu built without %s support\n", gen);
+		return false;
+	}
+
+	if ((module_param == -1 && (support_by_default || !radeon_support_built)) ||
+	    module_param == 1) {
+		if (radeon_support_built)
+			dev_info(dev, "%s support provided by amdgpu.\n"
+				 "Use radeon.%s=1 amdgpu.%s=0 to override.\n",
+				 gen, param, param);
+
+		return true;
+	}
+
+	if (radeon_support_built)
+		dev_info(dev, "%s support provided by radeon.\n"
+			 "Use radeon.%s=0 amdgpu.%s=1 to override.\n",
+			 gen, param, param);
+	else if (module_param == 0)
+		dev_info(dev, "%s support disabled by module param\n", gen);
+
+	return false;
+}
+
 static int amdgpu_pci_probe(struct pci_dev *pdev,
 			    const struct pci_device_id *ent)
 {
@@ -2063,6 +2381,12 @@ static int amdgpu_pci_probe(struct pci_dev *pdev,
 	int ret, retry = 0, i;
 	bool supports_atomic = false;
 
+	if ((pdev->class >> 8) == PCI_CLASS_DISPLAY_VGA ||
+	    (pdev->class >> 8) == PCI_CLASS_DISPLAY_OTHER) {
+		if (drm_firmware_drivers_only() && amdgpu_modeset == -1)
+			return -EINVAL;
+	}
+
 	/* skip devices which are owned by radeon */
 	for (i = 0; i < ARRAY_SIZE(amdgpu_unsupported_pciidlist); i++) {
 		if (amdgpu_unsupported_pciidlist[i] == pdev->device)
@@ -2073,7 +2397,7 @@ static int amdgpu_pci_probe(struct pci_dev *pdev,
 		amdgpu_aspm = 0;
 
 	if (amdgpu_virtual_display ||
-	    amdgpu_device_asic_has_dc_support(flags & AMD_ASIC_MASK))
+	    amdgpu_device_asic_has_dc_support(pdev, flags & AMD_ASIC_MASK))
 		supports_atomic = true;
 
 	if ((flags & AMD_EXP_HW_SUPPORT) && !amdgpu_exp_hw_support) {
@@ -2081,15 +2405,8 @@ static int amdgpu_pci_probe(struct pci_dev *pdev,
 			 "See modparam exp_hw_support\n");
 		return -ENODEV;
 	}
-	/* differentiate between P10 and P11 asics with the same DID */
-	if (pdev->device == 0x67FF &&
-	    (pdev->revision == 0xE3 ||
-	     pdev->revision == 0xE7 ||
-	     pdev->revision == 0xF3 ||
-	     pdev->revision == 0xF7)) {
-		flags &= ~AMD_ASIC_MASK;
-		flags |= CHIP_POLARIS10;
-	}
+
+	flags = amdgpu_fix_asic_type(pdev, flags);
 
 	/* Due to hardware bugs, S/G Display on raven requires a 1:1 IOMMU mapping,
 	 * however, SME requires an indirect IOMMU mapping because the encryption
@@ -2102,40 +2419,8 @@ static int amdgpu_pci_probe(struct pci_dev *pdev,
 		return -ENOTSUPP;
 	}
 
-#ifdef CONFIG_DRM_AMDGPU_SI
-	if (!amdgpu_si_support) {
-		switch (flags & AMD_ASIC_MASK) {
-		case CHIP_TAHITI:
-		case CHIP_PITCAIRN:
-		case CHIP_VERDE:
-		case CHIP_OLAND:
-		case CHIP_HAINAN:
-			dev_info(&pdev->dev,
-				 "SI support provided by radeon.\n");
-			dev_info(&pdev->dev,
-				 "Use radeon.si_support=0 amdgpu.si_support=1 to override.\n"
-				);
-			return -ENODEV;
-		}
-	}
-#endif
-#ifdef CONFIG_DRM_AMDGPU_CIK
-	if (!amdgpu_cik_support) {
-		switch (flags & AMD_ASIC_MASK) {
-		case CHIP_KAVERI:
-		case CHIP_BONAIRE:
-		case CHIP_HAWAII:
-		case CHIP_KABINI:
-		case CHIP_MULLINS:
-			dev_info(&pdev->dev,
-				 "CIK support provided by radeon.\n");
-			dev_info(&pdev->dev,
-				 "Use radeon.cik_support=0 amdgpu.cik_support=1 to override.\n"
-				);
-			return -ENODEV;
-		}
-	}
-#endif
+	if (!amdgpu_support_enabled(&pdev->dev, flags & AMD_ASIC_MASK))
+		return -ENODEV;
 
 	adev = devm_drm_dev_alloc(&pdev->dev, &amdgpu_kms_driver, typeof(*adev), ddev);
 	if (IS_ERR(adev))
@@ -2154,6 +2439,8 @@ static int amdgpu_pci_probe(struct pci_dev *pdev,
 
 	pci_set_drvdata(pdev, ddev);
 
+	amdgpu_init_debug_options(adev);
+
 	ret = amdgpu_driver_load_kms(adev, flags);
 	if (ret)
 		goto err_pci;
@@ -2169,17 +2456,29 @@ retry_init:
 		goto err_pci;
 	}
 
+	ret = amdgpu_xcp_dev_register(adev, ent);
+	if (ret)
+		goto err_pci;
+
+	ret = amdgpu_amdkfd_drm_client_create(adev);
+	if (ret)
+		goto err_pci;
+
 	/*
 	 * 1. don't init fbdev on hw without DCE
 	 * 2. don't init fbdev if there are no connectors
 	 */
 	if (adev->mode_info.mode_config_initialized &&
 	    !list_empty(&adev_to_drm(adev)->mode_config.connector_list)) {
+		const struct drm_format_info *format;
+
 		/* select 8 bpp console on low vram cards */
 		if (adev->gmc.real_vram_size <= (32*1024*1024))
-			drm_fbdev_generic_setup(adev_to_drm(adev), 8);
+			format = drm_format_info(DRM_FORMAT_C8);
 		else
-			drm_fbdev_generic_setup(adev_to_drm(adev), 32);
+			format = NULL;
+
+		drm_client_setup(adev_to_drm(adev), format);
 	}
 
 	ret = amdgpu_debugfs_init(adev);
@@ -2188,10 +2487,10 @@ retry_init:
 
 	if (adev->pm.rpm_mode != AMDGPU_RUNPM_NONE) {
 		/* only need to skip on ATPX */
-		if (amdgpu_device_supports_px(ddev))
+		if (amdgpu_device_supports_px(adev))
 			dev_pm_set_driver_flags(ddev->dev, DPM_FLAG_NO_DIRECT_COMPLETE);
 		/* we want direct complete for BOCO */
-		if (amdgpu_device_supports_boco(ddev))
+		if (amdgpu_device_supports_boco(adev))
 			dev_pm_set_driver_flags(ddev->dev, DPM_FLAG_SMART_PREPARE |
 						DPM_FLAG_SMART_SUSPEND |
 						DPM_FLAG_MAY_SKIP_RESUME);
@@ -2200,9 +2499,10 @@ retry_init:
 
 		pm_runtime_allow(ddev->dev);
 
-		pm_runtime_mark_last_busy(ddev->dev);
 		pm_runtime_put_autosuspend(ddev->dev);
 
+		pci_wake_from_d3(pdev, TRUE);
+
 		/*
 		 * For runpm implemented via BACO, PMFW will handle the
 		 * timing for BACO in and out:
@@ -2222,9 +2522,9 @@ retry_init:
 		 * into D0 state. Then there will be a PMFW-aware D-state
 		 * transition(D0->D3) on runpm suspend.
 		 */
-		if (amdgpu_device_supports_baco(ddev) &&
+		if (amdgpu_device_supports_baco(adev) &&
 		    !(adev->flags & AMD_IS_APU) &&
-		    (adev->asic_type >= CHIP_NAVI10))
+		    adev->asic_type >= CHIP_NAVI10)
 			amdgpu_get_secondary_funcs(adev);
 	}
 
@@ -2241,6 +2541,9 @@ amdgpu_pci_remove(struct pci_dev *pdev)
 	struct drm_device *dev = pci_get_drvdata(pdev);
 	struct amdgpu_device *adev = drm_to_adev(dev);
 
+	amdgpu_ras_eeprom_check_and_recover(adev);
+	amdgpu_xcp_dev_unplug(adev);
+	amdgpu_gmc_prepare_nps_mode_change(adev);
 	drm_dev_unplug(dev);
 
 	if (adev->pm.rpm_mode != AMDGPU_RUNPM_NONE) {
@@ -2248,38 +2551,6 @@ amdgpu_pci_remove(struct pci_dev *pdev)
 		pm_runtime_forbid(dev->dev);
 	}
 
-	if (adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2) &&
-	    !amdgpu_sriov_vf(adev)) {
-		bool need_to_reset_gpu = false;
-
-		if (adev->gmc.xgmi.num_physical_nodes > 1) {
-			struct amdgpu_hive_info *hive;
-
-			hive = amdgpu_get_xgmi_hive(adev);
-			if (hive->device_remove_count == 0)
-				need_to_reset_gpu = true;
-			hive->device_remove_count++;
-			amdgpu_put_xgmi_hive(hive);
-		} else {
-			need_to_reset_gpu = true;
-		}
-
-		/* Workaround for ASICs need to reset SMU.
-		 * Called only when the first device is removed.
-		 */
-		if (need_to_reset_gpu) {
-			struct amdgpu_reset_context reset_context;
-
-			adev->shutdown = true;
-			memset(&reset_context, 0, sizeof(reset_context));
-			reset_context.method = AMD_RESET_METHOD_NONE;
-			reset_context.reset_req_dev = adev;
-			set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
-			set_bit(AMDGPU_RESET_FOR_DEVICE_REMOVE, &reset_context.flags);
-			amdgpu_device_gpu_recover(adev, NULL, &reset_context);
-		}
-	}
-
 	amdgpu_driver_unload_kms(dev);
 
 	/*
@@ -2300,6 +2571,10 @@ amdgpu_pci_shutdown(struct pci_dev *pdev)
 	if (amdgpu_ras_intr_triggered())
 		return;
 
+	/* device maybe not resumed here, return immediately in this case */
+	if (adev->in_s4 && adev->in_suspend)
+		return;
+
 	/* if we are running in a VM, make sure the device
 	 * torn down properly on reboot/shutdown.
 	 * unfortunately we can't detect certain
@@ -2307,93 +2582,25 @@ amdgpu_pci_shutdown(struct pci_dev *pdev)
 	 */
 	if (!amdgpu_passthrough(adev))
 		adev->mp1_state = PP_MP1_STATE_UNLOAD;
-	amdgpu_device_ip_suspend(adev);
+	amdgpu_device_prepare(dev);
+	amdgpu_device_suspend(dev, true);
 	adev->mp1_state = PP_MP1_STATE_NONE;
 }
 
-/**
- * amdgpu_drv_delayed_reset_work_handler - work handler for reset
- *
- * @work: work_struct.
- */
-static void amdgpu_drv_delayed_reset_work_handler(struct work_struct *work)
-{
-	struct list_head device_list;
-	struct amdgpu_device *adev;
-	int i, r;
-	struct amdgpu_reset_context reset_context;
-
-	memset(&reset_context, 0, sizeof(reset_context));
-
-	mutex_lock(&mgpu_info.mutex);
-	if (mgpu_info.pending_reset == true) {
-		mutex_unlock(&mgpu_info.mutex);
-		return;
-	}
-	mgpu_info.pending_reset = true;
-	mutex_unlock(&mgpu_info.mutex);
-
-	/* Use a common context, just need to make sure full reset is done */
-	reset_context.method = AMD_RESET_METHOD_NONE;
-	set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
-
-	for (i = 0; i < mgpu_info.num_dgpu; i++) {
-		adev = mgpu_info.gpu_ins[i].adev;
-		reset_context.reset_req_dev = adev;
-		r = amdgpu_device_pre_asic_reset(adev, &reset_context);
-		if (r) {
-			dev_err(adev->dev, "GPU pre asic reset failed with err, %d for drm dev, %s ",
-				r, adev_to_drm(adev)->unique);
-		}
-		if (!queue_work(system_unbound_wq, &adev->xgmi_reset_work))
-			r = -EALREADY;
-	}
-	for (i = 0; i < mgpu_info.num_dgpu; i++) {
-		adev = mgpu_info.gpu_ins[i].adev;
-		flush_work(&adev->xgmi_reset_work);
-		adev->gmc.xgmi.pending_reset = false;
-	}
-
-	/* reset function will rebuild the xgmi hive info , clear it now */
-	for (i = 0; i < mgpu_info.num_dgpu; i++)
-		amdgpu_xgmi_remove_device(mgpu_info.gpu_ins[i].adev);
-
-	INIT_LIST_HEAD(&device_list);
-
-	for (i = 0; i < mgpu_info.num_dgpu; i++)
-		list_add_tail(&mgpu_info.gpu_ins[i].adev->reset_list, &device_list);
-
-	/* unregister the GPU first, reset function will add them back */
-	list_for_each_entry(adev, &device_list, reset_list)
-		amdgpu_unregister_gpu_instance(adev);
-
-	/* Use a common context, just need to make sure full reset is done */
-	set_bit(AMDGPU_SKIP_HW_RESET, &reset_context.flags);
-	r = amdgpu_do_asic_reset(&device_list, &reset_context);
-
-	if (r) {
-		DRM_ERROR("reinit gpus failure");
-		return;
-	}
-	for (i = 0; i < mgpu_info.num_dgpu; i++) {
-		adev = mgpu_info.gpu_ins[i].adev;
-		if (!adev->kfd.init_complete)
-			amdgpu_amdkfd_device_init(adev);
-		amdgpu_ttm_set_buffer_funcs_status(adev, true);
-	}
-	return;
-}
-
 static int amdgpu_pmops_prepare(struct device *dev)
 {
 	struct drm_device *drm_dev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(drm_dev);
 
+	/* device maybe not resumed here, return immediately in this case */
+	if (adev->in_s4 && adev->in_suspend)
+		return 0;
+
 	/* Return a positive number here so
 	 * DPM_FLAG_SMART_SUSPEND works properly
 	 */
-	if (amdgpu_device_supports_boco(drm_dev))
-		return pm_runtime_suspended(dev);
+	if (amdgpu_device_supports_boco(adev) && pm_runtime_suspended(dev))
+		return 1;
 
 	/* if we will not support s3 or s2i for the device
 	 *  then skip suspend
@@ -2402,12 +2609,12 @@ static int amdgpu_pmops_prepare(struct device *dev)
 	    !amdgpu_acpi_is_s3_active(adev))
 		return 1;
 
-	return 0;
+	return amdgpu_device_prepare(drm_dev);
 }
 
 static void amdgpu_pmops_complete(struct device *dev)
 {
-	/* nothing to do */
+	amdgpu_device_complete(dev_get_drvdata(dev));
 }
 
 static int amdgpu_pmops_suspend(struct device *dev)
@@ -2419,8 +2626,24 @@ static int amdgpu_pmops_suspend(struct device *dev)
 		adev->in_s0ix = true;
 	else if (amdgpu_acpi_is_s3_active(adev))
 		adev->in_s3 = true;
-	if (!adev->in_s0ix && !adev->in_s3)
+	if (!adev->in_s0ix && !adev->in_s3) {
+#if IS_ENABLED(CONFIG_SUSPEND)
+		/* don't allow going deep first time followed by s2idle the next time */
+		if (adev->last_suspend_state != PM_SUSPEND_ON &&
+		    adev->last_suspend_state != pm_suspend_target_state) {
+			drm_err_once(drm_dev, "Unsupported suspend state %d\n",
+				     pm_suspend_target_state);
+			return -EINVAL;
+		}
+#endif
 		return 0;
+	}
+
+#if IS_ENABLED(CONFIG_SUSPEND)
+	/* cache the state last used for suspend */
+	adev->last_suspend_state = pm_suspend_target_state;
+#endif
+
 	return amdgpu_device_suspend(drm_dev, true);
 }
 
@@ -2428,9 +2651,14 @@ static int amdgpu_pmops_suspend_noirq(struct device *dev)
 {
 	struct drm_device *drm_dev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(drm_dev);
+	int r;
 
-	if (amdgpu_acpi_should_gpu_reset(adev))
-		return amdgpu_asic_reset(adev);
+	if (amdgpu_acpi_should_gpu_reset(adev)) {
+		amdgpu_device_lock_reset_domain(adev->reset_domain);
+		r = amdgpu_asic_reset(adev);
+		amdgpu_device_unlock_reset_domain(adev->reset_domain);
+		return r;
+	}
 
 	return 0;
 }
@@ -2462,9 +2690,7 @@ static int amdgpu_pmops_freeze(struct device *dev)
 	struct amdgpu_device *adev = drm_to_adev(drm_dev);
 	int r;
 
-	adev->in_s4 = true;
 	r = amdgpu_device_suspend(drm_dev, true);
-	adev->in_s4 = false;
 	if (r)
 		return r;
 
@@ -2477,12 +2703,21 @@ static int amdgpu_pmops_thaw(struct device *dev)
 {
 	struct drm_device *drm_dev = dev_get_drvdata(dev);
 
+	/* do not resume device if it's normal hibernation */
+	if (!pm_hibernate_is_recovering() && !pm_hibernation_mode_is_suspend())
+		return 0;
+
 	return amdgpu_device_resume(drm_dev, true);
 }
 
 static int amdgpu_pmops_poweroff(struct device *dev)
 {
 	struct drm_device *drm_dev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(drm_dev);
+
+	/* device maybe not resumed here, return immediately in this case */
+	if (adev->in_s4 && adev->in_suspend)
+		return 0;
 
 	return amdgpu_device_suspend(drm_dev, true);
 }
@@ -2505,24 +2740,26 @@ static int amdgpu_runtime_idle_check_display(struct device *dev)
 		struct drm_connector_list_iter iter;
 		int ret = 0;
 
-		/* XXX: Return busy if any displays are connected to avoid
-		 * possible display wakeups after runtime resume due to
-		 * hotplug events in case any displays were connected while
-		 * the GPU was in suspend.  Remove this once that is fixed.
-		 */
-		mutex_lock(&drm_dev->mode_config.mutex);
-		drm_connector_list_iter_begin(drm_dev, &iter);
-		drm_for_each_connector_iter(list_connector, &iter) {
-			if (list_connector->status == connector_status_connected) {
-				ret = -EBUSY;
-				break;
+		if (amdgpu_runtime_pm != -2) {
+			/* XXX: Return busy if any displays are connected to avoid
+			 * possible display wakeups after runtime resume due to
+			 * hotplug events in case any displays were connected while
+			 * the GPU was in suspend.  Remove this once that is fixed.
+			 */
+			mutex_lock(&drm_dev->mode_config.mutex);
+			drm_connector_list_iter_begin(drm_dev, &iter);
+			drm_for_each_connector_iter(list_connector, &iter) {
+				if (list_connector->status == connector_status_connected) {
+					ret = -EBUSY;
+					break;
+				}
 			}
-		}
-		drm_connector_list_iter_end(&iter);
-		mutex_unlock(&drm_dev->mode_config.mutex);
+			drm_connector_list_iter_end(&iter);
+			mutex_unlock(&drm_dev->mode_config.mutex);
 
-		if (ret)
-			return ret;
+			if (ret)
+				return ret;
+		}
 
 		if (adev->dc_enabled) {
 			struct drm_crtc *crtc;
@@ -2559,6 +2796,15 @@ static int amdgpu_runtime_idle_check_display(struct device *dev)
 	return 0;
 }
 
+static int amdgpu_runtime_idle_check_userq(struct device *dev)
+{
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct drm_device *drm_dev = pci_get_drvdata(pdev);
+	struct amdgpu_device *adev = drm_to_adev(drm_dev);
+
+	return xa_empty(&adev->userq_doorbell_xa) ? 0 : -EBUSY;
+}
+
 static int amdgpu_pmops_runtime_suspend(struct device *dev)
 {
 	struct pci_dev *pdev = to_pci_dev(dev);
@@ -2574,10 +2820,14 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev)
 	ret = amdgpu_runtime_idle_check_display(dev);
 	if (ret)
 		return ret;
+	ret = amdgpu_runtime_idle_check_userq(dev);
+	if (ret)
+		return ret;
 
 	/* wait for all rings to drain before suspending */
 	for (i = 0; i < AMDGPU_MAX_RINGS; i++) {
 		struct amdgpu_ring *ring = adev->rings[i];
+
 		if (ring && ring->sched.ready) {
 			ret = amdgpu_fence_wait_empty(ring);
 			if (ret)
@@ -2586,7 +2836,7 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev)
 	}
 
 	adev->in_runpm = true;
-	if (amdgpu_device_supports_px(drm_dev))
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX)
 		drm_dev->switch_power_state = DRM_SWITCH_POWER_CHANGING;
 
 	/*
@@ -2596,21 +2846,24 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev)
 	 * platforms.
 	 * TODO: this may be also needed for PX capable platform.
 	 */
-	if (amdgpu_device_supports_boco(drm_dev))
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO)
 		adev->mp1_state = PP_MP1_STATE_UNLOAD;
 
+	ret = amdgpu_device_prepare(drm_dev);
+	if (ret)
+		return ret;
 	ret = amdgpu_device_suspend(drm_dev, false);
 	if (ret) {
 		adev->in_runpm = false;
-		if (amdgpu_device_supports_boco(drm_dev))
+		if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO)
 			adev->mp1_state = PP_MP1_STATE_NONE;
 		return ret;
 	}
 
-	if (amdgpu_device_supports_boco(drm_dev))
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO)
 		adev->mp1_state = PP_MP1_STATE_NONE;
 
-	if (amdgpu_device_supports_px(drm_dev)) {
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX) {
 		/* Only need to handle PCI state in the driver for ATPX
 		 * PCI core handles it for _PR3.
 		 */
@@ -2619,10 +2872,11 @@ static int amdgpu_pmops_runtime_suspend(struct device *dev)
 		pci_ignore_hotplug(pdev);
 		pci_set_power_state(pdev, PCI_D3cold);
 		drm_dev->switch_power_state = DRM_SWITCH_POWER_DYNAMIC_OFF;
-	} else if (amdgpu_device_supports_boco(drm_dev)) {
+	} else if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO) {
 		/* nothing to do */
-	} else if (amdgpu_device_supports_baco(drm_dev)) {
-		amdgpu_device_baco_enter(drm_dev);
+	} else if ((adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) ||
+			(adev->pm.rpm_mode == AMDGPU_RUNPM_BAMACO)) {
+		amdgpu_device_baco_enter(adev);
 	}
 
 	dev_dbg(&pdev->dev, "asic/device is runtime suspended\n");
@@ -2644,7 +2898,7 @@ static int amdgpu_pmops_runtime_resume(struct device *dev)
 	if (!pci_device_is_present(adev->pdev))
 		adev->no_hw_access = true;
 
-	if (amdgpu_device_supports_px(drm_dev)) {
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX) {
 		drm_dev->switch_power_state = DRM_SWITCH_POWER_CHANGING;
 
 		/* Only need to handle PCI state in the driver for ATPX
@@ -2656,22 +2910,23 @@ static int amdgpu_pmops_runtime_resume(struct device *dev)
 		if (ret)
 			return ret;
 		pci_set_master(pdev);
-	} else if (amdgpu_device_supports_boco(drm_dev)) {
+	} else if (adev->pm.rpm_mode == AMDGPU_RUNPM_BOCO) {
 		/* Only need to handle PCI state in the driver for ATPX
 		 * PCI core handles it for _PR3.
 		 */
 		pci_set_master(pdev);
-	} else if (amdgpu_device_supports_baco(drm_dev)) {
-		amdgpu_device_baco_exit(drm_dev);
+	} else if ((adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) ||
+			(adev->pm.rpm_mode == AMDGPU_RUNPM_BAMACO)) {
+		amdgpu_device_baco_exit(adev);
 	}
 	ret = amdgpu_device_resume(drm_dev, false);
 	if (ret) {
-		if (amdgpu_device_supports_px(drm_dev))
+		if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX)
 			pci_disable_device(pdev);
 		return ret;
 	}
 
-	if (amdgpu_device_supports_px(drm_dev))
+	if (adev->pm.rpm_mode == AMDGPU_RUNPM_PX)
 		drm_dev->switch_power_state = DRM_SWITCH_POWER_ON;
 	adev->in_runpm = false;
 	return 0;
@@ -2681,8 +2936,7 @@ static int amdgpu_pmops_runtime_idle(struct device *dev)
 {
 	struct drm_device *drm_dev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(drm_dev);
-	/* we don't want the main rpm_idle to call suspend - we want to autosuspend */
-	int ret = 1;
+	int ret;
 
 	if (adev->pm.rpm_mode == AMDGPU_RUNPM_NONE) {
 		pm_runtime_forbid(dev);
@@ -2690,18 +2944,39 @@ static int amdgpu_pmops_runtime_idle(struct device *dev)
 	}
 
 	ret = amdgpu_runtime_idle_check_display(dev);
+	if (ret)
+		goto done;
 
-	pm_runtime_mark_last_busy(dev);
+	ret = amdgpu_runtime_idle_check_userq(dev);
+done:
 	pm_runtime_autosuspend(dev);
 	return ret;
 }
 
+static int amdgpu_drm_release(struct inode *inode, struct file *filp)
+{
+	struct drm_file *file_priv = filp->private_data;
+	struct amdgpu_fpriv *fpriv = file_priv->driver_priv;
+	struct drm_device *dev = file_priv->minor->dev;
+	int idx;
+
+	if (fpriv && drm_dev_enter(dev, &idx)) {
+		fpriv->evf_mgr.fd_closing = true;
+		amdgpu_eviction_fence_destroy(&fpriv->evf_mgr);
+		amdgpu_userq_mgr_fini(&fpriv->userq_mgr);
+		drm_dev_exit(idx);
+	}
+
+	return drm_release(inode, filp);
+}
+
 long amdgpu_drm_ioctl(struct file *filp,
 		      unsigned int cmd, unsigned long arg)
 {
 	struct drm_file *file_priv = filp->private_data;
 	struct drm_device *dev;
 	long ret;
+
 	dev = file_priv->minor->dev;
 	ret = pm_runtime_get_sync(dev->dev);
 	if (ret < 0)
@@ -2709,22 +2984,21 @@ long amdgpu_drm_ioctl(struct file *filp,
 
 	ret = drm_ioctl(filp, cmd, arg);
 
-	pm_runtime_mark_last_busy(dev->dev);
 out:
 	pm_runtime_put_autosuspend(dev->dev);
 	return ret;
 }
 
 static const struct dev_pm_ops amdgpu_pm_ops = {
-	.prepare = amdgpu_pmops_prepare,
-	.complete = amdgpu_pmops_complete,
-	.suspend = amdgpu_pmops_suspend,
-	.suspend_noirq = amdgpu_pmops_suspend_noirq,
-	.resume = amdgpu_pmops_resume,
-	.freeze = amdgpu_pmops_freeze,
-	.thaw = amdgpu_pmops_thaw,
-	.poweroff = amdgpu_pmops_poweroff,
-	.restore = amdgpu_pmops_restore,
+	.prepare = pm_sleep_ptr(amdgpu_pmops_prepare),
+	.complete = pm_sleep_ptr(amdgpu_pmops_complete),
+	.suspend = pm_sleep_ptr(amdgpu_pmops_suspend),
+	.suspend_noirq = pm_sleep_ptr(amdgpu_pmops_suspend_noirq),
+	.resume = pm_sleep_ptr(amdgpu_pmops_resume),
+	.freeze = pm_sleep_ptr(amdgpu_pmops_freeze),
+	.thaw = pm_sleep_ptr(amdgpu_pmops_thaw),
+	.poweroff = pm_sleep_ptr(amdgpu_pmops_poweroff),
+	.restore = pm_sleep_ptr(amdgpu_pmops_restore),
 	.runtime_suspend = amdgpu_pmops_runtime_suspend,
 	.runtime_resume = amdgpu_pmops_runtime_resume,
 	.runtime_idle = amdgpu_pmops_runtime_idle,
@@ -2746,7 +3020,7 @@ static const struct file_operations amdgpu_driver_kms_fops = {
 	.owner = THIS_MODULE,
 	.open = drm_open,
 	.flush = amdgpu_flush,
-	.release = drm_release,
+	.release = amdgpu_drm_release,
 	.unlocked_ioctl = amdgpu_drm_ioctl,
 	.mmap = drm_gem_mmap,
 	.poll = drm_poll,
@@ -2755,8 +3029,9 @@ static const struct file_operations amdgpu_driver_kms_fops = {
 	.compat_ioctl = amdgpu_kms_compat_ioctl,
 #endif
 #ifdef CONFIG_PROC_FS
-	.show_fdinfo = amdgpu_show_fdinfo
+	.show_fdinfo = drm_show_fdinfo,
 #endif
+	.fop_flags = FOP_UNSIGNED_OFFSET,
 };
 
 int amdgpu_file_to_fpriv(struct file *filp, struct amdgpu_fpriv **fpriv)
@@ -2766,9 +3041,8 @@ int amdgpu_file_to_fpriv(struct file *filp, struct amdgpu_fpriv **fpriv)
 	if (!filp)
 		return -EINVAL;
 
-	if (filp->f_op != &amdgpu_driver_kms_fops) {
+	if (filp->f_op != &amdgpu_driver_kms_fops)
 		return -EINVAL;
-	}
 
 	file = filp->private_data;
 	*fpriv = file->driver_priv;
@@ -2793,6 +3067,10 @@ const struct drm_ioctl_desc amdgpu_ioctls_kms[] = {
 	DRM_IOCTL_DEF_DRV(AMDGPU_GEM_VA, amdgpu_gem_va_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF_DRV(AMDGPU_GEM_OP, amdgpu_gem_op_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 	DRM_IOCTL_DEF_DRV(AMDGPU_GEM_USERPTR, amdgpu_gem_userptr_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(AMDGPU_USERQ, amdgpu_userq_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(AMDGPU_USERQ_SIGNAL, amdgpu_userq_signal_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(AMDGPU_USERQ_WAIT, amdgpu_userq_wait_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(AMDGPU_GEM_LIST_HANDLES, amdgpu_gem_list_handles_ioctl, DRM_AUTH|DRM_RENDER_ALLOW),
 };
 
 static const struct drm_driver amdgpu_kms_driver = {
@@ -2803,22 +3081,44 @@ static const struct drm_driver amdgpu_kms_driver = {
 	    DRIVER_SYNCOBJ_TIMELINE,
 	.open = amdgpu_driver_open_kms,
 	.postclose = amdgpu_driver_postclose_kms,
-	.lastclose = amdgpu_driver_lastclose_kms,
 	.ioctls = amdgpu_ioctls_kms,
 	.num_ioctls = ARRAY_SIZE(amdgpu_ioctls_kms),
 	.dumb_create = amdgpu_mode_dumb_create,
 	.dumb_map_offset = amdgpu_mode_dumb_mmap,
+	DRM_FBDEV_TTM_DRIVER_OPS,
+	.fops = &amdgpu_driver_kms_fops,
+	.release = &amdgpu_driver_release_kms,
+#ifdef CONFIG_PROC_FS
+	.show_fdinfo = amdgpu_show_fdinfo,
+#endif
+
+	.gem_prime_import = amdgpu_gem_prime_import,
+
+	.name = DRIVER_NAME,
+	.desc = DRIVER_DESC,
+	.major = KMS_DRIVER_MAJOR,
+	.minor = KMS_DRIVER_MINOR,
+	.patchlevel = KMS_DRIVER_PATCHLEVEL,
+};
+
+const struct drm_driver amdgpu_partition_driver = {
+	.driver_features =
+	    DRIVER_GEM | DRIVER_RENDER | DRIVER_SYNCOBJ |
+	    DRIVER_SYNCOBJ_TIMELINE,
+	.open = amdgpu_driver_open_kms,
+	.postclose = amdgpu_driver_postclose_kms,
+	.ioctls = amdgpu_ioctls_kms,
+	.num_ioctls = ARRAY_SIZE(amdgpu_ioctls_kms),
+	.dumb_create = amdgpu_mode_dumb_create,
+	.dumb_map_offset = amdgpu_mode_dumb_mmap,
+	DRM_FBDEV_TTM_DRIVER_OPS,
 	.fops = &amdgpu_driver_kms_fops,
 	.release = &amdgpu_driver_release_kms,
 
-	.prime_handle_to_fd = drm_gem_prime_handle_to_fd,
-	.prime_fd_to_handle = drm_gem_prime_fd_to_handle,
 	.gem_prime_import = amdgpu_gem_prime_import,
-	.gem_prime_mmap = drm_gem_prime_mmap,
 
 	.name = DRIVER_NAME,
 	.desc = DRIVER_DESC,
-	.date = DRIVER_DATE,
 	.major = KMS_DRIVER_MAJOR,
 	.minor = KMS_DRIVER_MINOR,
 	.patchlevel = KMS_DRIVER_PATCHLEVEL,
@@ -2831,25 +3131,20 @@ static struct pci_error_handlers amdgpu_pci_err_handler = {
 	.resume		= amdgpu_pci_resume,
 };
 
-extern const struct attribute_group amdgpu_vram_mgr_attr_group;
-extern const struct attribute_group amdgpu_gtt_mgr_attr_group;
-extern const struct attribute_group amdgpu_vbios_version_attr_group;
-
 static const struct attribute_group *amdgpu_sysfs_groups[] = {
 	&amdgpu_vram_mgr_attr_group,
 	&amdgpu_gtt_mgr_attr_group,
-	&amdgpu_vbios_version_attr_group,
+	&amdgpu_flash_attr_group,
 	NULL,
 };
 
-
 static struct pci_driver amdgpu_kms_pci_driver = {
 	.name = DRIVER_NAME,
 	.id_table = pciidlist,
 	.probe = amdgpu_pci_probe,
 	.remove = amdgpu_pci_remove,
 	.shutdown = amdgpu_pci_shutdown,
-	.driver.pm = &amdgpu_pm_ops,
+	.driver.pm = pm_ptr(&amdgpu_pm_ops),
 	.err_handler = &amdgpu_pci_err_handler,
 	.dev_groups = amdgpu_sysfs_groups,
 };
@@ -2858,14 +3153,11 @@ static int __init amdgpu_init(void)
 {
 	int r;
 
-	if (drm_firmware_drivers_only())
-		return -EINVAL;
-
 	r = amdgpu_sync_init();
 	if (r)
 		goto error_sync;
 
-	r = amdgpu_fence_slab_init();
+	r = amdgpu_userq_fence_slab_init();
 	if (r)
 		goto error_fence;
 
@@ -2876,6 +3168,12 @@ static int __init amdgpu_init(void)
 	/* Ignore KFD init failures. Normal when CONFIG_HSA_AMD is not set. */
 	amdgpu_amdkfd_init();
 
+	if (amdgpu_pp_feature_mask & PP_OVERDRIVE_MASK) {
+		add_taint(TAINT_CPU_OUT_OF_SPEC, LOCKDEP_STILL_OK);
+		pr_crit("Overdrive is enabled, please disable it before "
+			"reporting any bugs unrelated to overdrive.\n");
+	}
+
 	/* let modprobe override vga console setting */
 	return pci_register_driver(&amdgpu_kms_pci_driver);
 
@@ -2891,9 +3189,11 @@ static void __exit amdgpu_exit(void)
 	amdgpu_amdkfd_fini();
 	pci_unregister_driver(&amdgpu_kms_pci_driver);
 	amdgpu_unregister_atpx_handler();
+	amdgpu_acpi_release();
 	amdgpu_sync_fini();
-	amdgpu_fence_slab_fini();
+	amdgpu_userq_fence_slab_fini();
 	mmu_notifier_synchronize();
+	amdgpu_xcp_drv_release();
 }
 
 module_init(amdgpu_init);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.h
index 8178323e4bef..2d86cc6f7f4d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.h
@@ -40,7 +40,8 @@
 
 #define DRIVER_NAME		"amdgpu"
 #define DRIVER_DESC		"AMD GPU"
-#define DRIVER_DATE		"20150101"
+
+extern const struct drm_driver amdgpu_partition_driver;
 
 long amdgpu_drm_ioctl(struct file *filp,
 		      unsigned int cmd, unsigned long arg);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.c
index 7d2a908438e9..8cd69836dd99 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.c
@@ -90,7 +90,7 @@
 #define MAKE_I2C_ADDR(_aa) ((0xA << 3) | (((_aa) >> 16) & 0xF))
 
 static int __amdgpu_eeprom_xfer(struct i2c_adapter *i2c_adap, u32 eeprom_addr,
-				u8 *eeprom_buf, u16 buf_size, bool read)
+				u8 *eeprom_buf, u32 buf_size, bool read)
 {
 	u8 eeprom_offset_buf[EEPROM_OFFSET_SIZE];
 	struct i2c_msg msgs[] = {
@@ -133,15 +133,15 @@ static int __amdgpu_eeprom_xfer(struct i2c_adapter *i2c_adap, u32 eeprom_addr,
 			 * cycle begins. This is implied for the
 			 * "i2c_transfer()" abstraction.
 			 */
-			len = min(EEPROM_PAGE_SIZE - (eeprom_addr &
-						      EEPROM_PAGE_MASK),
-				  (u32)buf_size);
+			len = min(EEPROM_PAGE_SIZE - (eeprom_addr & EEPROM_PAGE_MASK),
+					buf_size);
 		} else {
 			/* Reading from the EEPROM has no limitation
 			 * on the number of bytes read from the EEPROM
 			 * device--they are simply sequenced out.
+			 * Keep in mind that i2c_msg.len is u16 type.
 			 */
-			len = buf_size;
+			len = min(U16_MAX, buf_size);
 		}
 		msgs[1].len = len;
 		msgs[1].buf = eeprom_buf;
@@ -179,10 +179,12 @@ static int __amdgpu_eeprom_xfer(struct i2c_adapter *i2c_adap, u32 eeprom_addr,
  * Returns the number of bytes read/written; -errno on error.
  */
 static int amdgpu_eeprom_xfer(struct i2c_adapter *i2c_adap, u32 eeprom_addr,
-			      u8 *eeprom_buf, u16 buf_size, bool read)
+			      u8 *eeprom_buf, u32 buf_size, bool read)
 {
 	const struct i2c_adapter_quirks *quirks = i2c_adap->quirks;
 	u16 limit;
+	u16 ps; /* Partial size */
+	int res = 0, r;
 
 	if (!quirks)
 		limit = 0;
@@ -198,35 +200,32 @@ static int amdgpu_eeprom_xfer(struct i2c_adapter *i2c_adap, u32 eeprom_addr,
 		dev_err_ratelimited(&i2c_adap->dev,
 				    "maddr:0x%04X size:0x%02X:quirk max_%s_len must be > %d",
 				    eeprom_addr, buf_size,
-				    read ? "read" : "write", EEPROM_OFFSET_SIZE);
+				    str_read_write(read), EEPROM_OFFSET_SIZE);
 		return -EINVAL;
-	} else {
-		u16 ps; /* Partial size */
-		int res = 0, r;
-
-		/* The "limit" includes all data bytes sent/received,
-		 * which would include the EEPROM_OFFSET_SIZE bytes.
-		 * Account for them here.
-		 */
-		limit -= EEPROM_OFFSET_SIZE;
-		for ( ; buf_size > 0;
-		      buf_size -= ps, eeprom_addr += ps, eeprom_buf += ps) {
-			ps = min(limit, buf_size);
-
-			r = __amdgpu_eeprom_xfer(i2c_adap, eeprom_addr,
-						 eeprom_buf, ps, read);
-			if (r < 0)
-				return r;
-			res += r;
-		}
+	}
 
-		return res;
+	/* The "limit" includes all data bytes sent/received,
+	 * which would include the EEPROM_OFFSET_SIZE bytes.
+	 * Account for them here.
+	 */
+	limit -= EEPROM_OFFSET_SIZE;
+	for ( ; buf_size > 0;
+	      buf_size -= ps, eeprom_addr += ps, eeprom_buf += ps) {
+		ps = min(limit, buf_size);
+
+		r = __amdgpu_eeprom_xfer(i2c_adap, eeprom_addr,
+					 eeprom_buf, ps, read);
+		if (r < 0)
+			return r;
+		res += r;
 	}
+
+	return res;
 }
 
 int amdgpu_eeprom_read(struct i2c_adapter *i2c_adap,
 		       u32 eeprom_addr, u8 *eeprom_buf,
-		       u16 bytes)
+		       u32 bytes)
 {
 	return amdgpu_eeprom_xfer(i2c_adap, eeprom_addr, eeprom_buf, bytes,
 				  true);
@@ -234,7 +233,7 @@ int amdgpu_eeprom_read(struct i2c_adapter *i2c_adap,
 
 int amdgpu_eeprom_write(struct i2c_adapter *i2c_adap,
 			u32 eeprom_addr, u8 *eeprom_buf,
-			u16 bytes)
+			u32 bytes)
 {
 	return amdgpu_eeprom_xfer(i2c_adap, eeprom_addr, eeprom_buf, bytes,
 				  false);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.h
index 6935adb2be1f..8083b8253ef4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_eeprom.h
@@ -28,10 +28,10 @@
 
 int amdgpu_eeprom_read(struct i2c_adapter *i2c_adap,
 		       u32 eeprom_addr, u8 *eeprom_buf,
-		       u16 bytes);
+		       u32 bytes);
 
 int amdgpu_eeprom_write(struct i2c_adapter *i2c_adap,
 			u32 eeprom_addr, u8 *eeprom_buf,
-			u16 bytes);
+			u32 bytes);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_encoders.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_encoders.c
index 27a782a9dc72..3aaeed2d3562 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_encoders.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_encoders.c
@@ -70,6 +70,7 @@ void amdgpu_encoder_set_active_device(struct drm_encoder *encoder)
 	drm_for_each_connector_iter(connector, &iter) {
 		if (connector->encoder == encoder) {
 			struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
+
 			amdgpu_encoder->active_device = amdgpu_encoder->devices & amdgpu_connector->devices;
 			DRM_DEBUG_KMS("setting active device to %08x from %08x %08x for encoder %d\n",
 				  amdgpu_encoder->active_device, amdgpu_encoder->devices,
@@ -165,12 +166,12 @@ void amdgpu_panel_mode_fixup(struct drm_encoder *encoder,
 {
 	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
 	struct drm_display_mode *native_mode = &amdgpu_encoder->native_mode;
-	unsigned hblank = native_mode->htotal - native_mode->hdisplay;
-	unsigned vblank = native_mode->vtotal - native_mode->vdisplay;
-	unsigned hover = native_mode->hsync_start - native_mode->hdisplay;
-	unsigned vover = native_mode->vsync_start - native_mode->vdisplay;
-	unsigned hsync_width = native_mode->hsync_end - native_mode->hsync_start;
-	unsigned vsync_width = native_mode->vsync_end - native_mode->vsync_start;
+	unsigned int hblank = native_mode->htotal - native_mode->hdisplay;
+	unsigned int vblank = native_mode->vtotal - native_mode->vdisplay;
+	unsigned int hover = native_mode->hsync_start - native_mode->hdisplay;
+	unsigned int vover = native_mode->vsync_start - native_mode->vdisplay;
+	unsigned int hsync_width = native_mode->hsync_end - native_mode->hsync_start;
+	unsigned int vsync_width = native_mode->vsync_end - native_mode->vsync_start;
 
 	adjusted_mode->clock = native_mode->clock;
 	adjusted_mode->flags = native_mode->flags;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c
new file mode 100644
index 000000000000..23d7d0b0d625
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.c
@@ -0,0 +1,241 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include <linux/sched.h>
+#include <drm/drm_exec.h>
+#include "amdgpu.h"
+
+#define work_to_evf_mgr(w, name) container_of(w, struct amdgpu_eviction_fence_mgr, name)
+#define evf_mgr_to_fpriv(e) container_of(e, struct amdgpu_fpriv, evf_mgr)
+
+static const char *
+amdgpu_eviction_fence_get_driver_name(struct dma_fence *fence)
+{
+	return "amdgpu_eviction_fence";
+}
+
+static const char *
+amdgpu_eviction_fence_get_timeline_name(struct dma_fence *f)
+{
+	struct amdgpu_eviction_fence *ef;
+
+	ef = container_of(f, struct amdgpu_eviction_fence, base);
+	return ef->timeline_name;
+}
+
+int
+amdgpu_eviction_fence_replace_fence(struct amdgpu_eviction_fence_mgr *evf_mgr,
+				    struct drm_exec *exec)
+{
+	struct amdgpu_eviction_fence *old_ef, *new_ef;
+	struct drm_gem_object *obj;
+	unsigned long index;
+	int ret;
+
+	if (evf_mgr->ev_fence &&
+	    !dma_fence_is_signaled(&evf_mgr->ev_fence->base))
+		return 0;
+	/*
+	 * Steps to replace eviction fence:
+	 * * lock all objects in exec (caller)
+	 * * create a new eviction fence
+	 * * update new eviction fence in evf_mgr
+	 * * attach the new eviction fence to BOs
+	 * * release the old fence
+	 * * unlock the objects (caller)
+	 */
+	new_ef = amdgpu_eviction_fence_create(evf_mgr);
+	if (!new_ef) {
+		DRM_ERROR("Failed to create new eviction fence\n");
+		return -ENOMEM;
+	}
+
+	/* Update the eviction fence now */
+	spin_lock(&evf_mgr->ev_fence_lock);
+	old_ef = evf_mgr->ev_fence;
+	evf_mgr->ev_fence = new_ef;
+	spin_unlock(&evf_mgr->ev_fence_lock);
+
+	/* Attach the new fence */
+	drm_exec_for_each_locked_object(exec, index, obj) {
+		struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj);
+
+		if (!bo)
+			continue;
+		ret = amdgpu_eviction_fence_attach(evf_mgr, bo);
+		if (ret) {
+			DRM_ERROR("Failed to attch new eviction fence\n");
+			goto free_err;
+		}
+	}
+
+	/* Free old fence */
+	if (old_ef)
+		dma_fence_put(&old_ef->base);
+	return 0;
+
+free_err:
+	kfree(new_ef);
+	return ret;
+}
+
+static void
+amdgpu_eviction_fence_suspend_worker(struct work_struct *work)
+{
+	struct amdgpu_eviction_fence_mgr *evf_mgr = work_to_evf_mgr(work, suspend_work.work);
+	struct amdgpu_fpriv *fpriv = evf_mgr_to_fpriv(evf_mgr);
+	struct amdgpu_userq_mgr *uq_mgr = &fpriv->userq_mgr;
+	struct amdgpu_eviction_fence *ev_fence;
+
+	mutex_lock(&uq_mgr->userq_mutex);
+	spin_lock(&evf_mgr->ev_fence_lock);
+	ev_fence = evf_mgr->ev_fence;
+	if (ev_fence)
+		dma_fence_get(&ev_fence->base);
+	else
+		goto unlock;
+	spin_unlock(&evf_mgr->ev_fence_lock);
+
+	amdgpu_userq_evict(uq_mgr, ev_fence);
+
+	mutex_unlock(&uq_mgr->userq_mutex);
+	dma_fence_put(&ev_fence->base);
+	return;
+
+unlock:
+	spin_unlock(&evf_mgr->ev_fence_lock);
+	mutex_unlock(&uq_mgr->userq_mutex);
+}
+
+static bool amdgpu_eviction_fence_enable_signaling(struct dma_fence *f)
+{
+	struct amdgpu_eviction_fence_mgr *evf_mgr;
+	struct amdgpu_eviction_fence *ev_fence;
+
+	if (!f)
+		return true;
+
+	ev_fence = to_ev_fence(f);
+	evf_mgr = ev_fence->evf_mgr;
+
+	schedule_delayed_work(&evf_mgr->suspend_work, 0);
+	return true;
+}
+
+static const struct dma_fence_ops amdgpu_eviction_fence_ops = {
+	.get_driver_name = amdgpu_eviction_fence_get_driver_name,
+	.get_timeline_name = amdgpu_eviction_fence_get_timeline_name,
+	.enable_signaling = amdgpu_eviction_fence_enable_signaling,
+};
+
+void amdgpu_eviction_fence_signal(struct amdgpu_eviction_fence_mgr *evf_mgr,
+				  struct amdgpu_eviction_fence *ev_fence)
+{
+	spin_lock(&evf_mgr->ev_fence_lock);
+	dma_fence_signal(&ev_fence->base);
+	spin_unlock(&evf_mgr->ev_fence_lock);
+}
+
+struct amdgpu_eviction_fence *
+amdgpu_eviction_fence_create(struct amdgpu_eviction_fence_mgr *evf_mgr)
+{
+	struct amdgpu_eviction_fence *ev_fence;
+
+	ev_fence = kzalloc(sizeof(*ev_fence), GFP_KERNEL);
+	if (!ev_fence)
+		return NULL;
+
+	ev_fence->evf_mgr = evf_mgr;
+	get_task_comm(ev_fence->timeline_name, current);
+	spin_lock_init(&ev_fence->lock);
+	dma_fence_init64(&ev_fence->base, &amdgpu_eviction_fence_ops,
+			 &ev_fence->lock, evf_mgr->ev_fence_ctx,
+			 atomic_inc_return(&evf_mgr->ev_fence_seq));
+	return ev_fence;
+}
+
+void amdgpu_eviction_fence_destroy(struct amdgpu_eviction_fence_mgr *evf_mgr)
+{
+	struct amdgpu_eviction_fence *ev_fence;
+
+	/* Wait for any pending work to execute */
+	flush_delayed_work(&evf_mgr->suspend_work);
+
+	spin_lock(&evf_mgr->ev_fence_lock);
+	ev_fence = evf_mgr->ev_fence;
+	spin_unlock(&evf_mgr->ev_fence_lock);
+
+	if (!ev_fence)
+		return;
+
+	dma_fence_wait(&ev_fence->base, false);
+
+	/* Last unref of ev_fence */
+	dma_fence_put(&ev_fence->base);
+}
+
+int amdgpu_eviction_fence_attach(struct amdgpu_eviction_fence_mgr *evf_mgr,
+				 struct amdgpu_bo *bo)
+{
+	struct amdgpu_eviction_fence *ev_fence;
+	struct dma_resv *resv = bo->tbo.base.resv;
+	int ret;
+
+	if (!resv)
+		return 0;
+
+	ret = dma_resv_reserve_fences(resv, 1);
+	if (ret) {
+		DRM_DEBUG_DRIVER("Failed to resv fence space\n");
+		return ret;
+	}
+
+	spin_lock(&evf_mgr->ev_fence_lock);
+	ev_fence = evf_mgr->ev_fence;
+	if (ev_fence)
+		dma_resv_add_fence(resv, &ev_fence->base, DMA_RESV_USAGE_BOOKKEEP);
+	spin_unlock(&evf_mgr->ev_fence_lock);
+
+	return 0;
+}
+
+void amdgpu_eviction_fence_detach(struct amdgpu_eviction_fence_mgr *evf_mgr,
+				  struct amdgpu_bo *bo)
+{
+	struct dma_fence *stub = dma_fence_get_stub();
+
+	dma_resv_replace_fences(bo->tbo.base.resv, evf_mgr->ev_fence_ctx,
+				stub, DMA_RESV_USAGE_BOOKKEEP);
+	dma_fence_put(stub);
+}
+
+int amdgpu_eviction_fence_init(struct amdgpu_eviction_fence_mgr *evf_mgr)
+{
+	/* This needs to be done one time per open */
+	atomic_set(&evf_mgr->ev_fence_seq, 0);
+	evf_mgr->ev_fence_ctx = dma_fence_context_alloc(1);
+	spin_lock_init(&evf_mgr->ev_fence_lock);
+
+	INIT_DELAYED_WORK(&evf_mgr->suspend_work, amdgpu_eviction_fence_suspend_worker);
+	return 0;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.h
new file mode 100644
index 000000000000..fcd867b7147d
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_eviction_fence.h
@@ -0,0 +1,69 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef AMDGPU_EV_FENCE_H_
+#define AMDGPU_EV_FENCE_H_
+
+struct amdgpu_eviction_fence {
+	struct dma_fence base;
+	spinlock_t	 lock;
+	char		 timeline_name[TASK_COMM_LEN];
+	struct amdgpu_eviction_fence_mgr *evf_mgr;
+};
+
+struct amdgpu_eviction_fence_mgr {
+	u64			ev_fence_ctx;
+	atomic_t		ev_fence_seq;
+	spinlock_t		ev_fence_lock;
+	struct amdgpu_eviction_fence *ev_fence;
+	struct delayed_work	suspend_work;
+	uint8_t fd_closing;
+};
+
+/* Eviction fence helper functions */
+struct amdgpu_eviction_fence *
+amdgpu_eviction_fence_create(struct amdgpu_eviction_fence_mgr *evf_mgr);
+
+void
+amdgpu_eviction_fence_destroy(struct amdgpu_eviction_fence_mgr *evf_mgr);
+
+int
+amdgpu_eviction_fence_attach(struct amdgpu_eviction_fence_mgr *evf_mgr,
+			     struct amdgpu_bo *bo);
+
+void
+amdgpu_eviction_fence_detach(struct amdgpu_eviction_fence_mgr *evf_mgr,
+			     struct amdgpu_bo *bo);
+
+int
+amdgpu_eviction_fence_init(struct amdgpu_eviction_fence_mgr *evf_mgr);
+
+void
+amdgpu_eviction_fence_signal(struct amdgpu_eviction_fence_mgr *evf_mgr,
+			     struct amdgpu_eviction_fence *ev_fence);
+
+int
+amdgpu_eviction_fence_replace_fence(struct amdgpu_eviction_fence_mgr *evf_mgr,
+				    struct drm_exec *exec);
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c
index c57252f004e8..b349bb3676d5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.c
@@ -33,6 +33,7 @@
 #include <drm/amdgpu_drm.h>
 #include <drm/drm_debugfs.h>
 #include <drm/drm_drv.h>
+#include <drm/drm_file.h>
 
 #include "amdgpu.h"
 #include "amdgpu_vm.h"
@@ -51,34 +52,29 @@ static const char *amdgpu_ip_name[AMDGPU_HW_IP_NUM] = {
 	[AMDGPU_HW_IP_VCN_DEC]	=	"dec",
 	[AMDGPU_HW_IP_VCN_ENC]	=	"enc",
 	[AMDGPU_HW_IP_VCN_JPEG]	=	"jpeg",
+	[AMDGPU_HW_IP_VPE]	=	"vpe",
 };
 
-void amdgpu_show_fdinfo(struct seq_file *m, struct file *f)
+void amdgpu_show_fdinfo(struct drm_printer *p, struct drm_file *file)
 {
-	struct drm_file *file = f->private_data;
-	struct amdgpu_device *adev = drm_to_adev(file->minor->dev);
 	struct amdgpu_fpriv *fpriv = file->driver_priv;
 	struct amdgpu_vm *vm = &fpriv->vm;
 
-	struct amdgpu_mem_stats stats;
+	struct amdgpu_mem_stats stats[__AMDGPU_PL_NUM];
 	ktime_t usage[AMDGPU_HW_IP_NUM];
-	uint32_t bus, dev, fn, domain;
-	unsigned int hw_ip;
-	int ret;
-
-	memset(&stats, 0, sizeof(stats));
-	bus = adev->pdev->bus->number;
-	domain = pci_domain_nr(adev->pdev->bus);
-	dev = PCI_SLOT(adev->pdev->devfn);
-	fn = PCI_FUNC(adev->pdev->devfn);
-
-	ret = amdgpu_bo_reserve(vm->root.bo, false);
-	if (ret)
-		return;
-
-	amdgpu_vm_get_memory(vm, &stats);
-	amdgpu_bo_unreserve(vm->root.bo);
+	const char *pl_name[] = {
+		[TTM_PL_VRAM] = "vram",
+		[TTM_PL_TT] = "gtt",
+		[TTM_PL_SYSTEM] = "cpu",
+		[AMDGPU_PL_GDS] = "gds",
+		[AMDGPU_PL_GWS] = "gws",
+		[AMDGPU_PL_OA] = "oa",
+		[AMDGPU_PL_DOORBELL] = "doorbell",
+		[AMDGPU_PL_MMIO_REMAP] = "mmioremap",
+	};
+	unsigned int hw_ip, i;
 
+	amdgpu_vm_get_memory(vm, stats);
 	amdgpu_ctx_mgr_usage(&fpriv->ctx_mgr, usage);
 
 	/*
@@ -87,31 +83,42 @@ void amdgpu_show_fdinfo(struct seq_file *m, struct file *f)
 	 * ******************************************************************
 	 */
 
-	seq_printf(m, "pasid:\t%u\n", fpriv->vm.pasid);
-	seq_printf(m, "drm-driver:\t%s\n", file->minor->dev->driver->name);
-	seq_printf(m, "drm-pdev:\t%04x:%02x:%02x.%d\n", domain, bus, dev, fn);
-	seq_printf(m, "drm-client-id:\t%Lu\n", vm->immediate.fence_context);
-	seq_printf(m, "drm-memory-vram:\t%llu KiB\n", stats.vram/1024UL);
-	seq_printf(m, "drm-memory-gtt: \t%llu KiB\n", stats.gtt/1024UL);
-	seq_printf(m, "drm-memory-cpu: \t%llu KiB\n", stats.cpu/1024UL);
-	seq_printf(m, "amd-memory-visible-vram:\t%llu KiB\n",
-		   stats.visible_vram/1024UL);
-	seq_printf(m, "amd-evicted-vram:\t%llu KiB\n",
-		   stats.evicted_vram/1024UL);
-	seq_printf(m, "amd-evicted-visible-vram:\t%llu KiB\n",
-		   stats.evicted_visible_vram/1024UL);
-	seq_printf(m, "amd-requested-vram:\t%llu KiB\n",
-		   stats.requested_vram/1024UL);
-	seq_printf(m, "amd-requested-visible-vram:\t%llu KiB\n",
-		   stats.requested_visible_vram/1024UL);
-	seq_printf(m, "amd-requested-gtt:\t%llu KiB\n",
-		   stats.requested_gtt/1024UL);
+	drm_printf(p, "pasid:\t%u\n", fpriv->vm.pasid);
+
+	for (i = 0; i < ARRAY_SIZE(pl_name); i++) {
+		if (!pl_name[i])
+			continue;
+
+		drm_print_memory_stats(p,
+				       &stats[i].drm,
+				       DRM_GEM_OBJECT_RESIDENT |
+				       DRM_GEM_OBJECT_PURGEABLE,
+				       pl_name[i]);
+	}
+
+	/* Legacy amdgpu keys, alias to drm-resident-memory-: */
+	drm_printf(p, "drm-memory-vram:\t%llu KiB\n",
+		   stats[TTM_PL_VRAM].drm.resident/1024UL);
+	drm_printf(p, "drm-memory-gtt: \t%llu KiB\n",
+		   stats[TTM_PL_TT].drm.resident/1024UL);
+	drm_printf(p, "drm-memory-cpu: \t%llu KiB\n",
+		   stats[TTM_PL_SYSTEM].drm.resident/1024UL);
+
+	/* Amdgpu specific memory accounting keys: */
+	drm_printf(p, "amd-evicted-vram:\t%llu KiB\n",
+		   stats[TTM_PL_VRAM].evicted/1024UL);
+	drm_printf(p, "amd-requested-vram:\t%llu KiB\n",
+		   (stats[TTM_PL_VRAM].drm.shared +
+		    stats[TTM_PL_VRAM].drm.private) / 1024UL);
+	drm_printf(p, "amd-requested-gtt:\t%llu KiB\n",
+		   (stats[TTM_PL_TT].drm.shared +
+		    stats[TTM_PL_TT].drm.private) / 1024UL);
 
 	for (hw_ip = 0; hw_ip < AMDGPU_HW_IP_NUM; ++hw_ip) {
 		if (!usage[hw_ip])
 			continue;
 
-		seq_printf(m, "drm-engine-%s:\t%Ld ns\n", amdgpu_ip_name[hw_ip],
+		drm_printf(p, "drm-engine-%s:\t%lld ns\n", amdgpu_ip_name[hw_ip],
 			   ktime_to_ns(usage[hw_ip]));
 	}
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.h
index e86834bfea1d..0398f5a159ef 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fdinfo.h
@@ -37,6 +37,6 @@
 #include "amdgpu_ids.h"
 
 uint32_t amdgpu_get_ip_count(struct amdgpu_device *adev, int id);
-void amdgpu_show_fdinfo(struct seq_file *m, struct file *f);
+void amdgpu_show_fdinfo(struct drm_printer *p, struct drm_file *file);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
index f52d0ba91a77..c7843e336310 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
@@ -42,53 +42,14 @@
 #include "amdgpu_reset.h"
 
 /*
- * Fences
- * Fences mark an event in the GPUs pipeline and are used
- * for GPU/CPU synchronization.  When the fence is written,
- * it is expected that all buffers associated with that fence
- * are no longer in use by the associated ring on the GPU and
- * that the relevant GPU caches have been flushed.
- */
-
-struct amdgpu_fence {
-	struct dma_fence base;
-
-	/* RB, DMA, etc. */
-	struct amdgpu_ring		*ring;
-	ktime_t				start_timestamp;
-};
-
-static struct kmem_cache *amdgpu_fence_slab;
-
-int amdgpu_fence_slab_init(void)
-{
-	amdgpu_fence_slab = kmem_cache_create(
-		"amdgpu_fence", sizeof(struct amdgpu_fence), 0,
-		SLAB_HWCACHE_ALIGN, NULL);
-	if (!amdgpu_fence_slab)
-		return -ENOMEM;
-	return 0;
-}
-
-void amdgpu_fence_slab_fini(void)
-{
-	rcu_barrier();
-	kmem_cache_destroy(amdgpu_fence_slab);
-}
-/*
  * Cast helper
  */
 static const struct dma_fence_ops amdgpu_fence_ops;
-static const struct dma_fence_ops amdgpu_job_fence_ops;
 static inline struct amdgpu_fence *to_amdgpu_fence(struct dma_fence *f)
 {
 	struct amdgpu_fence *__f = container_of(f, struct amdgpu_fence, base);
 
-	if (__f->base.ops == &amdgpu_fence_ops ||
-	    __f->base.ops == &amdgpu_job_fence_ops)
-		return __f;
-
-	return NULL;
+	return __f;
 }
 
 /**
@@ -132,57 +93,32 @@ static u32 amdgpu_fence_read(struct amdgpu_ring *ring)
  * amdgpu_fence_emit - emit a fence on the requested ring
  *
  * @ring: ring the fence is associated with
- * @f: resulting fence object
- * @job: job the fence is embedded in
+ * @af: amdgpu fence input
  * @flags: flags to pass into the subordinate .emit_fence() call
  *
  * Emits a fence command on the requested ring (all asics).
  * Returns 0 on success, -ENOMEM on failure.
  */
-int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **f, struct amdgpu_job *job,
-		      unsigned flags)
+int amdgpu_fence_emit(struct amdgpu_ring *ring, struct amdgpu_fence *af,
+		      unsigned int flags)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct dma_fence *fence;
-	struct amdgpu_fence *am_fence;
 	struct dma_fence __rcu **ptr;
 	uint32_t seq;
 	int r;
 
-	if (job == NULL) {
-		/* create a sperate hw fence */
-		am_fence = kmem_cache_alloc(amdgpu_fence_slab, GFP_ATOMIC);
-		if (am_fence == NULL)
-			return -ENOMEM;
-		fence = &am_fence->base;
-		am_fence->ring = ring;
-	} else {
-		/* take use of job-embedded fence */
-		fence = &job->hw_fence;
-	}
+	fence = &af->base;
+	af->ring = ring;
 
 	seq = ++ring->fence_drv.sync_seq;
-	if (job && job->job_run_counter) {
-		/* reinit seq for resubmitted jobs */
-		fence->seqno = seq;
-		/* TO be inline with external fence creation and other drivers */
-		dma_fence_get(fence);
-	} else {
-		if (job) {
-			dma_fence_init(fence, &amdgpu_job_fence_ops,
-				       &ring->fence_drv.lock,
-				       adev->fence_context + ring->idx, seq);
-			/* Against remove in amdgpu_job_{free, free_cb} */
-			dma_fence_get(fence);
-		}
-		else
-			dma_fence_init(fence, &amdgpu_fence_ops,
-				       &ring->fence_drv.lock,
-				       adev->fence_context + ring->idx, seq);
-	}
+	dma_fence_init(fence, &amdgpu_fence_ops,
+		       &ring->fence_drv.lock,
+		       adev->fence_context + ring->idx, seq);
 
 	amdgpu_ring_emit_fence(ring, ring->fence_drv.gpu_addr,
 			       seq, flags | AMDGPU_FENCE_FLAG_INT);
+	amdgpu_fence_save_wptr(af);
 	pm_runtime_get_noresume(adev_to_drm(adev)->dev);
 	ptr = &ring->fence_drv.fences[seq & ring->fence_drv.num_fences_mask];
 	if (unlikely(rcu_dereference_protected(*ptr, 1))) {
@@ -207,8 +143,6 @@ int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **f, struct amd
 	 */
 	rcu_assign_pointer(*ptr, dma_fence_get(fence));
 
-	*f = fence;
-
 	return 0;
 }
 
@@ -283,7 +217,7 @@ bool amdgpu_fence_process(struct amdgpu_ring *ring)
 
 	} while (atomic_cmpxchg(&drv->last_seq, last_seq, seq) != last_seq);
 
-	if (del_timer(&ring->fence_drv.fallback_timer) &&
+	if (timer_delete(&ring->fence_drv.fallback_timer) &&
 	    seq != ring->fence_drv.sync_seq)
 		amdgpu_fence_schedule_fallback(ring);
 
@@ -295,6 +229,7 @@ bool amdgpu_fence_process(struct amdgpu_ring *ring)
 
 	do {
 		struct dma_fence *fence, **ptr;
+		struct amdgpu_fence *am_fence;
 
 		++last_seq;
 		last_seq &= drv->num_fences_mask;
@@ -307,9 +242,14 @@ bool amdgpu_fence_process(struct amdgpu_ring *ring)
 		if (!fence)
 			continue;
 
+		/* Save the wptr in the fence driver so we know what the last processed
+		 * wptr was.  This is required for re-emitting the ring state for
+		 * queues that are reset but are not guilty and thus have no guilty fence.
+		 */
+		am_fence = container_of(fence, struct amdgpu_fence, base);
+		drv->signalled_wptr = am_fence->wptr;
 		dma_fence_signal(fence);
 		dma_fence_put(fence);
-		pm_runtime_mark_last_busy(adev_to_drm(adev)->dev);
 		pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
 	} while (last_seq != seq);
 
@@ -325,11 +265,13 @@ bool amdgpu_fence_process(struct amdgpu_ring *ring)
  */
 static void amdgpu_fence_fallback(struct timer_list *t)
 {
-	struct amdgpu_ring *ring = from_timer(ring, t,
-					      fence_drv.fallback_timer);
+	struct amdgpu_ring *ring = timer_container_of(ring, t,
+						      fence_drv.fallback_timer);
 
 	if (amdgpu_fence_process(ring))
-		DRM_WARN("Fence fallback timer expired on ring %s\n", ring->name);
+		dev_warn(ring->adev->dev,
+			 "Fence fallback timer expired on ring %s\n",
+			 ring->name);
 }
 
 /**
@@ -377,14 +319,11 @@ signed long amdgpu_fence_wait_polling(struct amdgpu_ring *ring,
 				      uint32_t wait_seq,
 				      signed long timeout)
 {
-	uint32_t seq;
-
-	do {
-		seq = amdgpu_fence_read(ring);
-		udelay(5);
-		timeout -= 5;
-	} while ((int32_t)(wait_seq - seq) > 0 && timeout > 0);
 
+	while ((int32_t)(wait_seq - amdgpu_fence_read(ring)) > 0 && timeout > 0) {
+		udelay(2);
+		timeout -= 2;
+	}
 	return timeout > 0 ? timeout : 0;
 }
 /**
@@ -396,7 +335,7 @@ signed long amdgpu_fence_wait_polling(struct amdgpu_ring *ring,
  * Returns the number of emitted fences on the ring.  Used by the
  * dynpm code to ring track activity.
  */
-unsigned amdgpu_fence_count_emitted(struct amdgpu_ring *ring)
+unsigned int amdgpu_fence_count_emitted(struct amdgpu_ring *ring)
 {
 	uint64_t emitted;
 
@@ -475,7 +414,7 @@ void amdgpu_fence_update_start_timestamp(struct amdgpu_ring *ring, uint32_t seq,
  */
 int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring,
 				   struct amdgpu_irq_src *irq_src,
-				   unsigned irq_type)
+				   unsigned int irq_type)
 {
 	struct amdgpu_device *adev = ring->adev;
 	uint64_t index;
@@ -556,6 +495,42 @@ int amdgpu_fence_driver_sw_init(struct amdgpu_device *adev)
 }
 
 /**
+ * amdgpu_fence_need_ring_interrupt_restore - helper function to check whether
+ * fence driver interrupts need to be restored.
+ *
+ * @ring: ring that to be checked
+ *
+ * Interrupts for rings that belong to GFX IP don't need to be restored
+ * when the target power state is s0ix.
+ *
+ * Return true if need to restore interrupts, false otherwise.
+ */
+static bool amdgpu_fence_need_ring_interrupt_restore(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	bool is_gfx_power_domain = false;
+
+	switch (ring->funcs->type) {
+	case AMDGPU_RING_TYPE_SDMA:
+	/* SDMA 5.x+ is part of GFX power domain so it's covered by GFXOFF */
+		if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) >=
+		    IP_VERSION(5, 0, 0))
+			is_gfx_power_domain = true;
+		break;
+	case AMDGPU_RING_TYPE_GFX:
+	case AMDGPU_RING_TYPE_COMPUTE:
+	case AMDGPU_RING_TYPE_KIQ:
+	case AMDGPU_RING_TYPE_MES:
+		is_gfx_power_domain = true;
+		break;
+	default:
+		break;
+	}
+
+	return !(adev->in_s0ix && is_gfx_power_domain);
+}
+
+/**
  * amdgpu_fence_driver_hw_fini - tear down the fence driver
  * for all possible rings.
  *
@@ -582,11 +557,13 @@ void amdgpu_fence_driver_hw_fini(struct amdgpu_device *adev)
 		if (r)
 			amdgpu_fence_driver_force_completion(ring);
 
-		if (ring->fence_drv.irq_src)
+		if (!drm_dev_is_unplugged(adev_to_drm(adev)) &&
+		    ring->fence_drv.irq_src &&
+		    amdgpu_fence_need_ring_interrupt_restore(ring))
 			amdgpu_irq_put(adev, ring->fence_drv.irq_src,
 				       ring->fence_drv.irq_type);
 
-		del_timer_sync(&ring->fence_drv.fallback_timer);
+		timer_delete_sync(&ring->fence_drv.fallback_timer);
 	}
 }
 
@@ -653,44 +630,40 @@ void amdgpu_fence_driver_hw_init(struct amdgpu_device *adev)
 
 	for (i = 0; i < AMDGPU_MAX_RINGS; i++) {
 		struct amdgpu_ring *ring = adev->rings[i];
+
 		if (!ring || !ring->fence_drv.initialized)
 			continue;
 
 		/* enable the interrupt */
-		if (ring->fence_drv.irq_src)
+		if (ring->fence_drv.irq_src &&
+		    amdgpu_fence_need_ring_interrupt_restore(ring))
 			amdgpu_irq_get(adev, ring->fence_drv.irq_src,
 				       ring->fence_drv.irq_type);
 	}
 }
 
 /**
- * amdgpu_fence_driver_clear_job_fences - clear job embedded fences of ring
- *
- * @ring: fence of the ring to be cleared
+ * amdgpu_fence_driver_set_error - set error code on fences
+ * @ring: the ring which contains the fences
+ * @error: the error code to set
  *
+ * Set an error code to all the fences pending on the ring.
  */
-void amdgpu_fence_driver_clear_job_fences(struct amdgpu_ring *ring)
+void amdgpu_fence_driver_set_error(struct amdgpu_ring *ring, int error)
 {
-	int i;
-	struct dma_fence *old, **ptr;
+	struct amdgpu_fence_driver *drv = &ring->fence_drv;
+	unsigned long flags;
 
-	for (i = 0; i <= ring->fence_drv.num_fences_mask; i++) {
-		ptr = &ring->fence_drv.fences[i];
-		old = rcu_dereference_protected(*ptr, 1);
-		if (old && old->ops == &amdgpu_job_fence_ops) {
-			struct amdgpu_job *job;
+	spin_lock_irqsave(&drv->lock, flags);
+	for (unsigned int i = 0; i <= drv->num_fences_mask; ++i) {
+		struct dma_fence *fence;
 
-			/* For non-scheduler bad job, i.e. failed ib test, we need to signal
-			 * it right here or we won't be able to track them in fence_drv
-			 * and they will remain unsignaled during sa_bo free.
-			 */
-			job = container_of(old, struct amdgpu_job, hw_fence);
-			if (!job->base.s_fence && !dma_fence_is_signaled(old))
-				dma_fence_signal(old);
-			RCU_INIT_POINTER(*ptr, NULL);
-			dma_fence_put(old);
-		}
+		fence = rcu_dereference_protected(drv->fences[i],
+						  lockdep_is_held(&drv->lock));
+		if (fence && !dma_fence_is_signaled_locked(fence))
+			dma_fence_set_error(fence, error);
 	}
+	spin_unlock_irqrestore(&drv->lock, flags);
 }
 
 /**
@@ -701,10 +674,125 @@ void amdgpu_fence_driver_clear_job_fences(struct amdgpu_ring *ring)
  */
 void amdgpu_fence_driver_force_completion(struct amdgpu_ring *ring)
 {
+	amdgpu_fence_driver_set_error(ring, -ECANCELED);
 	amdgpu_fence_write(ring, ring->fence_drv.sync_seq);
 	amdgpu_fence_process(ring);
 }
 
+
+/*
+ * Kernel queue reset handling
+ *
+ * The driver can reset individual queues for most engines, but those queues
+ * may contain work from multiple contexts.  Resetting the queue will reset
+ * lose all of that state.  In order to minimize the collateral damage, the
+ * driver will save the ring contents which are not associated with the guilty
+ * context prior to resetting the queue.  After resetting the queue the queue
+ * contents from the other contexts is re-emitted to the rings so that it can
+ * be processed by the engine.  To handle this, we save the queue's write
+ * pointer (wptr) in the fences associated with each context.  If we get a
+ * queue timeout, we can then use the wptrs from the fences to determine
+ * which data needs to be saved out of the queue's ring buffer.
+ */
+
+/**
+ * amdgpu_fence_driver_guilty_force_completion - force signal of specified sequence
+ *
+ * @af: fence of the ring to signal
+ *
+ */
+void amdgpu_fence_driver_guilty_force_completion(struct amdgpu_fence *af)
+{
+	struct dma_fence *unprocessed;
+	struct dma_fence __rcu **ptr;
+	struct amdgpu_fence *fence;
+	struct amdgpu_ring *ring = af->ring;
+	unsigned long flags;
+	u32 seq, last_seq;
+
+	last_seq = amdgpu_fence_read(ring) & ring->fence_drv.num_fences_mask;
+	seq = ring->fence_drv.sync_seq & ring->fence_drv.num_fences_mask;
+
+	/* mark all fences from the guilty context with an error */
+	spin_lock_irqsave(&ring->fence_drv.lock, flags);
+	do {
+		last_seq++;
+		last_seq &= ring->fence_drv.num_fences_mask;
+
+		ptr = &ring->fence_drv.fences[last_seq];
+		rcu_read_lock();
+		unprocessed = rcu_dereference(*ptr);
+
+		if (unprocessed && !dma_fence_is_signaled_locked(unprocessed)) {
+			fence = container_of(unprocessed, struct amdgpu_fence, base);
+
+			if (fence == af)
+				dma_fence_set_error(&fence->base, -ETIME);
+			else if (fence->context == af->context)
+				dma_fence_set_error(&fence->base, -ECANCELED);
+		}
+		rcu_read_unlock();
+	} while (last_seq != seq);
+	spin_unlock_irqrestore(&ring->fence_drv.lock, flags);
+	/* signal the guilty fence */
+	amdgpu_fence_write(ring, (u32)af->base.seqno);
+	amdgpu_fence_process(ring);
+}
+
+void amdgpu_fence_save_wptr(struct amdgpu_fence *af)
+{
+	af->wptr = af->ring->wptr;
+}
+
+static void amdgpu_ring_backup_unprocessed_command(struct amdgpu_ring *ring,
+						   u64 start_wptr, u32 end_wptr)
+{
+	unsigned int first_idx = start_wptr & ring->buf_mask;
+	unsigned int last_idx = end_wptr & ring->buf_mask;
+	unsigned int i;
+
+	/* Backup the contents of the ring buffer. */
+	for (i = first_idx; i != last_idx; ++i, i &= ring->buf_mask)
+		ring->ring_backup[ring->ring_backup_entries_to_copy++] = ring->ring[i];
+}
+
+void amdgpu_ring_backup_unprocessed_commands(struct amdgpu_ring *ring,
+					     struct amdgpu_fence *guilty_fence)
+{
+	struct dma_fence *unprocessed;
+	struct dma_fence __rcu **ptr;
+	struct amdgpu_fence *fence;
+	u64 wptr;
+	u32 seq, last_seq;
+
+	last_seq = amdgpu_fence_read(ring) & ring->fence_drv.num_fences_mask;
+	seq = ring->fence_drv.sync_seq & ring->fence_drv.num_fences_mask;
+	wptr = ring->fence_drv.signalled_wptr;
+	ring->ring_backup_entries_to_copy = 0;
+
+	do {
+		last_seq++;
+		last_seq &= ring->fence_drv.num_fences_mask;
+
+		ptr = &ring->fence_drv.fences[last_seq];
+		rcu_read_lock();
+		unprocessed = rcu_dereference(*ptr);
+
+		if (unprocessed && !dma_fence_is_signaled(unprocessed)) {
+			fence = container_of(unprocessed, struct amdgpu_fence, base);
+
+			/* save everything if the ring is not guilty, otherwise
+			 * just save the content from other contexts.
+			 */
+			if (!guilty_fence || (fence->context != guilty_fence->context))
+				amdgpu_ring_backup_unprocessed_command(ring, wptr,
+								       fence->wptr);
+			wptr = fence->wptr;
+		}
+		rcu_read_unlock();
+	} while (last_seq != seq);
+}
+
 /*
  * Common fence implementation
  */
@@ -719,13 +807,6 @@ static const char *amdgpu_fence_get_timeline_name(struct dma_fence *f)
 	return (const char *)to_amdgpu_fence(f)->ring->name;
 }
 
-static const char *amdgpu_job_fence_get_timeline_name(struct dma_fence *f)
-{
-	struct amdgpu_job *job = container_of(f, struct amdgpu_job, hw_fence);
-
-	return (const char *)to_amdgpu_ring(job->base.sched)->name;
-}
-
 /**
  * amdgpu_fence_enable_signaling - enable signalling on fence
  * @f: fence
@@ -743,23 +824,6 @@ static bool amdgpu_fence_enable_signaling(struct dma_fence *f)
 }
 
 /**
- * amdgpu_job_fence_enable_signaling - enable signalling on job fence
- * @f: fence
- *
- * This is the simliar function with amdgpu_fence_enable_signaling above, it
- * only handles the job embedded fence.
- */
-static bool amdgpu_job_fence_enable_signaling(struct dma_fence *f)
-{
-	struct amdgpu_job *job = container_of(f, struct amdgpu_job, hw_fence);
-
-	if (!timer_pending(&to_amdgpu_ring(job->base.sched)->fence_drv.fallback_timer))
-		amdgpu_fence_schedule_fallback(to_amdgpu_ring(job->base.sched));
-
-	return true;
-}
-
-/**
  * amdgpu_fence_free - free up the fence memory
  *
  * @rcu: RCU callback head
@@ -771,22 +835,7 @@ static void amdgpu_fence_free(struct rcu_head *rcu)
 	struct dma_fence *f = container_of(rcu, struct dma_fence, rcu);
 
 	/* free fence_slab if it's separated fence*/
-	kmem_cache_free(amdgpu_fence_slab, to_amdgpu_fence(f));
-}
-
-/**
- * amdgpu_job_fence_free - free up the job with embedded fence
- *
- * @rcu: RCU callback head
- *
- * Free up the job with embedded fence after the RCU grace period.
- */
-static void amdgpu_job_fence_free(struct rcu_head *rcu)
-{
-	struct dma_fence *f = container_of(rcu, struct dma_fence, rcu);
-
-	/* free job if fence has a parent job */
-	kfree(container_of(f, struct amdgpu_job, hw_fence));
+	kfree(to_amdgpu_fence(f));
 }
 
 /**
@@ -802,19 +851,6 @@ static void amdgpu_fence_release(struct dma_fence *f)
 	call_rcu(&f->rcu, amdgpu_fence_free);
 }
 
-/**
- * amdgpu_job_fence_release - callback that job embedded fence can be freed
- *
- * @f: fence
- *
- * This is the simliar function with amdgpu_fence_release above, it
- * only handles the job embedded fence.
- */
-static void amdgpu_job_fence_release(struct dma_fence *f)
-{
-	call_rcu(&f->rcu, amdgpu_job_fence_free);
-}
-
 static const struct dma_fence_ops amdgpu_fence_ops = {
 	.get_driver_name = amdgpu_fence_get_driver_name,
 	.get_timeline_name = amdgpu_fence_get_timeline_name,
@@ -822,24 +858,18 @@ static const struct dma_fence_ops amdgpu_fence_ops = {
 	.release = amdgpu_fence_release,
 };
 
-static const struct dma_fence_ops amdgpu_job_fence_ops = {
-	.get_driver_name = amdgpu_fence_get_driver_name,
-	.get_timeline_name = amdgpu_job_fence_get_timeline_name,
-	.enable_signaling = amdgpu_job_fence_enable_signaling,
-	.release = amdgpu_job_fence_release,
-};
-
 /*
  * Fence debugfs
  */
 #if defined(CONFIG_DEBUG_FS)
 static int amdgpu_debugfs_fence_info_show(struct seq_file *m, void *unused)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct amdgpu_device *adev = m->private;
 	int i;
 
 	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
 		struct amdgpu_ring *ring = adev->rings[i];
+
 		if (!ring || !ring->fence_drv.initialized)
 			continue;
 
@@ -897,7 +927,6 @@ static int gpu_recover_get(void *data, u64 *val)
 
 	*val = atomic_read(&adev->reset_domain->reset_res);
 
-	pm_runtime_mark_last_busy(dev->dev);
 	pm_runtime_put_autosuspend(dev->dev);
 
 	return 0;
@@ -913,11 +942,14 @@ static void amdgpu_debugfs_reset_work(struct work_struct *work)
 						  reset_work);
 
 	struct amdgpu_reset_context reset_context;
+
 	memset(&reset_context, 0, sizeof(reset_context));
 
 	reset_context.method = AMD_RESET_METHOD_NONE;
 	reset_context.reset_req_dev = adev;
+	reset_context.src = AMDGPU_RESET_SRC_USER;
 	set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
+	set_bit(AMDGPU_SKIP_COREDUMP, &reset_context.flags);
 
 	amdgpu_device_gpu_recover(adev, NULL, &reset_context);
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c
index 4620c4712ce3..b0082aa7f3c6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.c
@@ -31,6 +31,7 @@
 
 #define FRU_EEPROM_MADDR_6      0x60000
 #define FRU_EEPROM_MADDR_8      0x80000
+#define FRU_EEPROM_MADDR_INV    0xFFFFF
 
 static bool is_fru_eeprom_supported(struct amdgpu_device *adev, u32 *fru_addr)
 {
@@ -42,8 +43,9 @@ static bool is_fru_eeprom_supported(struct amdgpu_device *adev, u32 *fru_addr)
 
 	/* The i2c access is blocked on VF
 	 * TODO: Need other way to get the info
+	 * Also, FRU not valid for APU devices.
 	 */
-	if (amdgpu_sriov_vf(adev))
+	if (amdgpu_sriov_vf(adev) || (adev->flags & AMD_IS_APU))
 		return false;
 
 	/* The default I2C EEPROM address of the FRU.
@@ -57,40 +59,56 @@ static bool is_fru_eeprom_supported(struct amdgpu_device *adev, u32 *fru_addr)
 	 * for ease/speed/readability. For now, 2 string comparisons are
 	 * reasonable and not too expensive
 	 */
-	switch (adev->asic_type) {
-	case CHIP_VEGA20:
-		/* D161 and D163 are the VG20 server SKUs */
-		if (strnstr(atom_ctx->vbios_version, "D161",
-			    sizeof(atom_ctx->vbios_version)) ||
-		    strnstr(atom_ctx->vbios_version, "D163",
-			    sizeof(atom_ctx->vbios_version))) {
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
+	case IP_VERSION(11, 0, 2):
+		switch (adev->asic_type) {
+		case CHIP_VEGA20:
+			/* D161 and D163 are the VG20 server SKUs */
+			if (atom_ctx && (strnstr(atom_ctx->vbios_pn, "D161",
+						 sizeof(atom_ctx->vbios_pn)) ||
+					 strnstr(atom_ctx->vbios_pn, "D163",
+						 sizeof(atom_ctx->vbios_pn)))) {
+				if (fru_addr)
+					*fru_addr = FRU_EEPROM_MADDR_6;
+				return true;
+			} else {
+				return false;
+			}
+		case CHIP_ARCTURUS:
+		default:
+			return false;
+		}
+	case IP_VERSION(11, 0, 7):
+		if (atom_ctx && strnstr(atom_ctx->vbios_pn, "D603",
+					sizeof(atom_ctx->vbios_pn))) {
+			if (strnstr(atom_ctx->vbios_pn, "D603GLXE",
+				    sizeof(atom_ctx->vbios_pn))) {
+				return false;
+			}
+
 			if (fru_addr)
 				*fru_addr = FRU_EEPROM_MADDR_6;
 			return true;
+
 		} else {
 			return false;
 		}
-	case CHIP_ALDEBARAN:
+	case IP_VERSION(13, 0, 2):
 		/* All Aldebaran SKUs have an FRU */
-		if (!strnstr(atom_ctx->vbios_version, "D673",
-			     sizeof(atom_ctx->vbios_version)))
+		if (atom_ctx && !strnstr(atom_ctx->vbios_pn, "D673",
+					 sizeof(atom_ctx->vbios_pn)))
 			if (fru_addr)
 				*fru_addr = FRU_EEPROM_MADDR_6;
 		return true;
-	case CHIP_SIENNA_CICHLID:
-		if (strnstr(atom_ctx->vbios_version, "D603",
-			    sizeof(atom_ctx->vbios_version))) {
-			if (strnstr(atom_ctx->vbios_version, "D603GLXE",
-				    sizeof(atom_ctx->vbios_version))) {
-				return false;
-			} else {
-				if (fru_addr)
-					*fru_addr = FRU_EEPROM_MADDR_6;
-				return true;
-			}
-		} else {
-			return false;
-		}
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 14):
+			if (fru_addr)
+				*fru_addr = FRU_EEPROM_MADDR_8;
+			return true;
+	case IP_VERSION(13, 0, 12):
+			if (fru_addr)
+				*fru_addr = FRU_EEPROM_MADDR_INV;
+			return true;
 	default:
 		return false;
 	}
@@ -98,6 +116,7 @@ static bool is_fru_eeprom_supported(struct amdgpu_device *adev, u32 *fru_addr)
 
 int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 {
+	struct amdgpu_fru_info *fru_info;
 	unsigned char buf[8], *pia;
 	u32 addr, fru_addr;
 	int size, len;
@@ -106,9 +125,27 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 	if (!is_fru_eeprom_supported(adev, &fru_addr))
 		return 0;
 
+	/* FRU data avaialble, but no direct EEPROM access */
+	if (fru_addr == FRU_EEPROM_MADDR_INV)
+		return 0;
+
+	if (!adev->fru_info) {
+		adev->fru_info = kzalloc(sizeof(*adev->fru_info), GFP_KERNEL);
+		if (!adev->fru_info)
+			return -ENOMEM;
+	}
+
+	fru_info = adev->fru_info;
+	/* For Arcturus-and-later, default value of serial_number is unique_id
+	 * so convert it to a 16-digit HEX string for convenience and
+	 * backwards-compatibility.
+	 */
+	sprintf(fru_info->serial, "%llx", adev->unique_id);
+
 	/* If algo exists, it means that the i2c_adapter's initialized */
 	if (!adev->pm.fru_eeprom_i2c_bus || !adev->pm.fru_eeprom_i2c_bus->algo) {
-		DRM_WARN("Cannot access FRU, EEPROM accessor not initialized");
+		dev_warn(adev->dev,
+			 "Cannot access FRU, EEPROM accessor not initialized");
 		return -ENODEV;
 	}
 
@@ -116,19 +153,22 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 	len = amdgpu_eeprom_read(adev->pm.fru_eeprom_i2c_bus, fru_addr, buf,
 				 sizeof(buf));
 	if (len != 8) {
-		DRM_ERROR("Couldn't read the IPMI Common Header: %d", len);
+		dev_err(adev->dev, "Couldn't read the IPMI Common Header: %d",
+			len);
 		return len < 0 ? len : -EIO;
 	}
 
 	if (buf[0] != 1) {
-		DRM_ERROR("Bad IPMI Common Header version: 0x%02x", buf[0]);
+		dev_err(adev->dev, "Bad IPMI Common Header version: 0x%02x",
+			buf[0]);
 		return -EIO;
 	}
 
 	for (csum = 0; len > 0; len--)
 		csum += buf[len - 1];
 	if (csum) {
-		DRM_ERROR("Bad IPMI Common Header checksum: 0x%02x", csum);
+		dev_err(adev->dev, "Bad IPMI Common Header checksum: 0x%02x",
+			csum);
 		return -EIO;
 	}
 
@@ -143,12 +183,14 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 	/* Read the header of the PIA. */
 	len = amdgpu_eeprom_read(adev->pm.fru_eeprom_i2c_bus, addr, buf, 3);
 	if (len != 3) {
-		DRM_ERROR("Couldn't read the Product Info Area header: %d", len);
+		dev_err(adev->dev,
+			"Couldn't read the Product Info Area header: %d", len);
 		return len < 0 ? len : -EIO;
 	}
 
 	if (buf[0] != 1) {
-		DRM_ERROR("Bad IPMI Product Info Area version: 0x%02x", buf[0]);
+		dev_err(adev->dev, "Bad IPMI Product Info Area version: 0x%02x",
+			buf[0]);
 		return -EIO;
 	}
 
@@ -161,40 +203,49 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 	len = amdgpu_eeprom_read(adev->pm.fru_eeprom_i2c_bus, addr, pia, size);
 	if (len != size) {
 		kfree(pia);
-		DRM_ERROR("Couldn't read the Product Info Area: %d", len);
+		dev_err(adev->dev, "Couldn't read the Product Info Area: %d",
+			len);
 		return len < 0 ? len : -EIO;
 	}
 
 	for (csum = 0; size > 0; size--)
 		csum += pia[size - 1];
 	if (csum) {
-		DRM_ERROR("Bad Product Info Area checksum: 0x%02x", csum);
+		dev_err(adev->dev, "Bad Product Info Area checksum: 0x%02x",
+			csum);
+		kfree(pia);
 		return -EIO;
 	}
 
 	/* Now extract useful information from the PIA.
 	 *
-	 * Skip the Manufacturer Name at [3] and go directly to
-	 * the Product Name field.
+	 * Read Manufacturer Name field whose length is [3].
 	 */
-	addr = 3 + 1 + (pia[3] & 0x3F);
+	addr = 3;
 	if (addr + 1 >= len)
 		goto Out;
-	memcpy(adev->product_name, pia + addr + 1,
-	       min_t(size_t,
-		     sizeof(adev->product_name),
+	memcpy(fru_info->manufacturer_name, pia + addr + 1,
+	       min_t(size_t, sizeof(fru_info->manufacturer_name),
 		     pia[addr] & 0x3F));
-	adev->product_name[sizeof(adev->product_name) - 1] = '\0';
+	fru_info->manufacturer_name[sizeof(fru_info->manufacturer_name) - 1] =
+		'\0';
+
+	/* Read Product Name field. */
+	addr += 1 + (pia[addr] & 0x3F);
+	if (addr + 1 >= len)
+		goto Out;
+	memcpy(fru_info->product_name, pia + addr + 1,
+	       min_t(size_t, sizeof(fru_info->product_name), pia[addr] & 0x3F));
+	fru_info->product_name[sizeof(fru_info->product_name) - 1] = '\0';
 
 	/* Go to the Product Part/Model Number field. */
 	addr += 1 + (pia[addr] & 0x3F);
 	if (addr + 1 >= len)
 		goto Out;
-	memcpy(adev->product_number, pia + addr + 1,
-	       min_t(size_t,
-		     sizeof(adev->product_number),
+	memcpy(fru_info->product_number, pia + addr + 1,
+	       min_t(size_t, sizeof(fru_info->product_number),
 		     pia[addr] & 0x3F));
-	adev->product_number[sizeof(adev->product_number) - 1] = '\0';
+	fru_info->product_number[sizeof(fru_info->product_number) - 1] = '\0';
 
 	/* Go to the Product Version field. */
 	addr += 1 + (pia[addr] & 0x3F);
@@ -203,11 +254,155 @@ int amdgpu_fru_get_product_info(struct amdgpu_device *adev)
 	addr += 1 + (pia[addr] & 0x3F);
 	if (addr + 1 >= len)
 		goto Out;
-	memcpy(adev->serial, pia + addr + 1, min_t(size_t,
-						   sizeof(adev->serial),
-						   pia[addr] & 0x3F));
-	adev->serial[sizeof(adev->serial) - 1] = '\0';
+	memcpy(fru_info->serial, pia + addr + 1,
+	       min_t(size_t, sizeof(fru_info->serial), pia[addr] & 0x3F));
+	fru_info->serial[sizeof(fru_info->serial) - 1] = '\0';
+
+	/* Asset Tag field */
+	addr += 1 + (pia[addr] & 0x3F);
+
+	/* FRU File Id field. This could be 'null'. */
+	addr += 1 + (pia[addr] & 0x3F);
+	if ((addr + 1 >= len) || !(pia[addr] & 0x3F))
+		goto Out;
+	memcpy(fru_info->fru_id, pia + addr + 1,
+	       min_t(size_t, sizeof(fru_info->fru_id), pia[addr] & 0x3F));
+	fru_info->fru_id[sizeof(fru_info->fru_id) - 1] = '\0';
+
 Out:
 	kfree(pia);
 	return 0;
 }
+
+/**
+ * DOC: product_name
+ *
+ * The amdgpu driver provides a sysfs API for reporting the product name
+ * for the device
+ * The file product_name is used for this and returns the product name
+ * as returned from the FRU.
+ * NOTE: This is only available for certain server cards
+ */
+
+static ssize_t amdgpu_fru_product_name_show(struct device *dev,
+					    struct device_attribute *attr,
+					    char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	return sysfs_emit(buf, "%s\n", adev->fru_info->product_name);
+}
+
+static DEVICE_ATTR(product_name, 0444, amdgpu_fru_product_name_show, NULL);
+
+/**
+ * DOC: product_number
+ *
+ * The amdgpu driver provides a sysfs API for reporting the part number
+ * for the device
+ * The file product_number is used for this and returns the part number
+ * as returned from the FRU.
+ * NOTE: This is only available for certain server cards
+ */
+
+static ssize_t amdgpu_fru_product_number_show(struct device *dev,
+					      struct device_attribute *attr,
+					      char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	return sysfs_emit(buf, "%s\n", adev->fru_info->product_number);
+}
+
+static DEVICE_ATTR(product_number, 0444, amdgpu_fru_product_number_show, NULL);
+
+/**
+ * DOC: serial_number
+ *
+ * The amdgpu driver provides a sysfs API for reporting the serial number
+ * for the device
+ * The file serial_number is used for this and returns the serial number
+ * as returned from the FRU.
+ * NOTE: This is only available for certain server cards
+ */
+
+static ssize_t amdgpu_fru_serial_number_show(struct device *dev,
+					     struct device_attribute *attr,
+					     char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	return sysfs_emit(buf, "%s\n", adev->fru_info->serial);
+}
+
+static DEVICE_ATTR(serial_number, 0444, amdgpu_fru_serial_number_show, NULL);
+
+/**
+ * DOC: fru_id
+ *
+ * The amdgpu driver provides a sysfs API for reporting FRU File Id
+ * for the device.
+ * The file fru_id is used for this and returns the File Id value
+ * as returned from the FRU.
+ * NOTE: This is only available for certain server cards
+ */
+
+static ssize_t amdgpu_fru_id_show(struct device *dev,
+				  struct device_attribute *attr, char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	return sysfs_emit(buf, "%s\n", adev->fru_info->fru_id);
+}
+
+static DEVICE_ATTR(fru_id, 0444, amdgpu_fru_id_show, NULL);
+
+/**
+ * DOC: manufacturer
+ *
+ * The amdgpu driver provides a sysfs API for reporting manufacturer name from
+ * FRU information.
+ * The file manufacturer returns the value as returned from the FRU.
+ * NOTE: This is only available for certain server cards
+ */
+
+static ssize_t amdgpu_fru_manufacturer_name_show(struct device *dev,
+						 struct device_attribute *attr,
+						 char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	return sysfs_emit(buf, "%s\n", adev->fru_info->manufacturer_name);
+}
+
+static DEVICE_ATTR(manufacturer, 0444, amdgpu_fru_manufacturer_name_show, NULL);
+
+static const struct attribute *amdgpu_fru_attributes[] = {
+	&dev_attr_product_name.attr,
+	&dev_attr_product_number.attr,
+	&dev_attr_serial_number.attr,
+	&dev_attr_fru_id.attr,
+	&dev_attr_manufacturer.attr,
+	NULL
+};
+
+int amdgpu_fru_sysfs_init(struct amdgpu_device *adev)
+{
+	if (!is_fru_eeprom_supported(adev, NULL) || !adev->fru_info)
+		return 0;
+
+	return sysfs_create_files(&adev->dev->kobj, amdgpu_fru_attributes);
+}
+
+void amdgpu_fru_sysfs_fini(struct amdgpu_device *adev)
+{
+	if (!adev->fru_info)
+		return;
+
+	sysfs_remove_files(&adev->dev->kobj, amdgpu_fru_attributes);
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.h
index 1308d976d60e..98f3196599ef 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fru_eeprom.h
@@ -24,6 +24,19 @@
 #ifndef __AMDGPU_FRU_EEPROM_H__
 #define __AMDGPU_FRU_EEPROM_H__
 
+#define AMDGPU_PRODUCT_NAME_LEN 64
+
+/* FRU product information */
+struct amdgpu_fru_info {
+	char				product_number[20];
+	char				product_name[AMDGPU_PRODUCT_NAME_LEN];
+	char				serial[20];
+	char				manufacturer_name[32];
+	char				fru_id[50];
+};
+
 int amdgpu_fru_get_product_info(struct amdgpu_device *adev);
+int amdgpu_fru_sysfs_init(struct amdgpu_device *adev);
+void amdgpu_fru_sysfs_fini(struct amdgpu_device *adev);
 
 #endif  // __AMDGPU_FRU_EEPROM_H__
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fw_attestation.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fw_attestation.c
index 2ca3c329de6d..328a1b963548 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fw_attestation.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fw_attestation.c
@@ -32,17 +32,15 @@
 #include "soc15_common.h"
 
 #define FW_ATTESTATION_DB_COOKIE        0x143b6a37
-#define FW_ATTESTATION_RECORD_VALID  	1
+#define FW_ATTESTATION_RECORD_VALID	1
 #define FW_ATTESTATION_MAX_SIZE		4096
 
-typedef struct FW_ATT_DB_HEADER
-{
+struct FW_ATT_DB_HEADER {
 	uint32_t AttDbVersion;           /* version of the fwar feature */
 	uint32_t AttDbCookie;            /* cookie as an extra check for corrupt data */
-} FW_ATT_DB_HEADER;
+};
 
-typedef struct FW_ATT_RECORD
-{
+struct FW_ATT_RECORD {
 	uint16_t AttFwIdV1;              /* Legacy FW Type field */
 	uint16_t AttFwIdV2;              /* V2 FW ID field */
 	uint32_t AttFWVersion;           /* FW Version */
@@ -50,7 +48,7 @@ typedef struct FW_ATT_RECORD
 	uint8_t  AttSource;              /* FW source indicator */
 	uint8_t  RecordValid;            /* Indicates whether the record is a valid entry */
 	uint32_t AttFwTaId;              /* Ta ID (only in TA Attestation Table) */
-} FW_ATT_RECORD;
+};
 
 static ssize_t amdgpu_fw_attestation_debugfs_read(struct file *f,
 						  char __user *buf,
@@ -60,15 +58,15 @@ static ssize_t amdgpu_fw_attestation_debugfs_read(struct file *f,
 	struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(f)->i_private;
 	uint64_t records_addr = 0;
 	uint64_t vram_pos = 0;
-	FW_ATT_DB_HEADER fw_att_hdr = {0};
-	FW_ATT_RECORD fw_att_record = {0};
+	struct FW_ATT_DB_HEADER fw_att_hdr = {0};
+	struct FW_ATT_RECORD fw_att_record = {0};
 
-	if (size < sizeof(FW_ATT_RECORD)) {
+	if (size < sizeof(struct FW_ATT_RECORD)) {
 		DRM_WARN("FW attestation input buffer not enough memory");
 		return -EINVAL;
 	}
 
-	if ((*pos + sizeof(FW_ATT_DB_HEADER)) >= FW_ATTESTATION_MAX_SIZE) {
+	if ((*pos + sizeof(struct FW_ATT_DB_HEADER)) >= FW_ATTESTATION_MAX_SIZE) {
 		DRM_WARN("FW attestation out of bounds");
 		return 0;
 	}
@@ -83,8 +81,8 @@ static ssize_t amdgpu_fw_attestation_debugfs_read(struct file *f,
 	if (*pos == 0) {
 		amdgpu_device_vram_access(adev,
 					  vram_pos,
-					  (uint32_t*)&fw_att_hdr,
-					  sizeof(FW_ATT_DB_HEADER),
+					  (uint32_t *)&fw_att_hdr,
+					  sizeof(struct FW_ATT_DB_HEADER),
 					  false);
 
 		if (fw_att_hdr.AttDbCookie != FW_ATTESTATION_DB_COOKIE) {
@@ -96,20 +94,20 @@ static ssize_t amdgpu_fw_attestation_debugfs_read(struct file *f,
 	}
 
 	amdgpu_device_vram_access(adev,
-				  vram_pos + sizeof(FW_ATT_DB_HEADER) + *pos,
-				  (uint32_t*)&fw_att_record,
-				  sizeof(FW_ATT_RECORD),
+				  vram_pos + sizeof(struct FW_ATT_DB_HEADER) + *pos,
+				  (uint32_t *)&fw_att_record,
+				  sizeof(struct FW_ATT_RECORD),
 				  false);
 
 	if (fw_att_record.RecordValid != FW_ATTESTATION_RECORD_VALID)
 		return 0;
 
-	if (copy_to_user(buf, (void*)&fw_att_record, sizeof(FW_ATT_RECORD)))
+	if (copy_to_user(buf, (void *)&fw_att_record, sizeof(struct FW_ATT_RECORD)))
 		return -EINVAL;
 
-	*pos += sizeof(FW_ATT_RECORD);
+	*pos += sizeof(struct FW_ATT_RECORD);
 
-	return sizeof(FW_ATT_RECORD);
+	return sizeof(struct FW_ATT_RECORD);
 }
 
 static const struct file_operations amdgpu_fw_attestation_debugfs_ops = {
@@ -124,6 +122,10 @@ static int amdgpu_is_fw_attestation_supported(struct amdgpu_device *adev)
 	if (adev->flags & AMD_IS_APU)
 		return 0;
 
+	if (amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(14, 0, 2) ||
+	    amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(14, 0, 3))
+		return 0;
+
 	if (adev->asic_type >= CHIP_SIENNA_CICHLID)
 		return 1;
 
@@ -136,7 +138,7 @@ void amdgpu_fw_attestation_debugfs_init(struct amdgpu_device *adev)
 		return;
 
 	debugfs_create_file("amdgpu_fw_attestation",
-			    S_IRUSR,
+			    0400,
 			    adev_to_drm(adev)->primary->debugfs_root,
 			    adev,
 			    &amdgpu_fw_attestation_debugfs_ops);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c
index 01cb89ffbd56..d2237ce9da70 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c
@@ -34,7 +34,9 @@
 #include <asm/set_memory.h>
 #endif
 #include "amdgpu.h"
+#include "amdgpu_reset.h"
 #include <drm/drm_drv.h>
+#include <drm/ttm/ttm_tt.h>
 
 /*
  * GART
@@ -76,8 +78,9 @@ static int amdgpu_gart_dummy_page_init(struct amdgpu_device *adev)
 
 	if (adev->dummy_page_addr)
 		return 0;
-	adev->dummy_page_addr = dma_map_page(&adev->pdev->dev, dummy_page, 0,
-					     PAGE_SIZE, DMA_BIDIRECTIONAL);
+	adev->dummy_page_addr = dma_map_page_attrs(&adev->pdev->dev, dummy_page, 0,
+							PAGE_SIZE, DMA_BIDIRECTIONAL,
+							DMA_ATTR_SKIP_CPU_SYNC);
 	if (dma_mapping_error(&adev->pdev->dev, adev->dummy_page_addr)) {
 		dev_err(&adev->pdev->dev, "Failed to DMA MAP the dummy page\n");
 		adev->dummy_page_addr = 0;
@@ -97,12 +100,157 @@ void amdgpu_gart_dummy_page_fini(struct amdgpu_device *adev)
 {
 	if (!adev->dummy_page_addr)
 		return;
-	dma_unmap_page(&adev->pdev->dev, adev->dummy_page_addr, PAGE_SIZE,
-		       DMA_BIDIRECTIONAL);
+	dma_unmap_page_attrs(&adev->pdev->dev, adev->dummy_page_addr, PAGE_SIZE,
+				DMA_BIDIRECTIONAL,
+				DMA_ATTR_SKIP_CPU_SYNC);
 	adev->dummy_page_addr = 0;
 }
 
 /**
+ * amdgpu_gart_table_ram_alloc - allocate system ram for gart page table
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Allocate system memory for GART page table for ASICs that don't have
+ * dedicated VRAM.
+ * Returns 0 for success, error for failure.
+ */
+int amdgpu_gart_table_ram_alloc(struct amdgpu_device *adev)
+{
+	unsigned int order = get_order(adev->gart.table_size);
+	gfp_t gfp_flags = GFP_KERNEL | __GFP_ZERO;
+	struct amdgpu_bo *bo = NULL;
+	struct sg_table *sg = NULL;
+	struct amdgpu_bo_param bp;
+	dma_addr_t dma_addr;
+	struct page *p;
+	unsigned long x;
+	int ret;
+
+	if (adev->gart.bo != NULL)
+		return 0;
+
+	p = alloc_pages(gfp_flags, order);
+	if (!p)
+		return -ENOMEM;
+
+	/* assign pages to this device */
+	for (x = 0; x < (1UL << order); x++)
+		p[x].mapping = adev->mman.bdev.dev_mapping;
+
+	/* If the hardware does not support UTCL2 snooping of the CPU caches
+	 * then set_memory_wc() could be used as a workaround to mark the pages
+	 * as write combine memory.
+	 */
+	dma_addr = dma_map_page(&adev->pdev->dev, p, 0, adev->gart.table_size,
+				DMA_BIDIRECTIONAL);
+	if (dma_mapping_error(&adev->pdev->dev, dma_addr)) {
+		dev_err(&adev->pdev->dev, "Failed to DMA MAP the GART BO page\n");
+		__free_pages(p, order);
+		p = NULL;
+		return -EFAULT;
+	}
+
+	dev_info(adev->dev, "%s dma_addr:%pad\n", __func__, &dma_addr);
+	/* Create SG table */
+	sg = kmalloc(sizeof(*sg), GFP_KERNEL);
+	if (!sg) {
+		ret = -ENOMEM;
+		goto error;
+	}
+	ret = sg_alloc_table(sg, 1, GFP_KERNEL);
+	if (ret)
+		goto error;
+
+	sg_dma_address(sg->sgl) = dma_addr;
+	sg->sgl->length = adev->gart.table_size;
+#ifdef CONFIG_NEED_SG_DMA_LENGTH
+	sg->sgl->dma_length = adev->gart.table_size;
+#endif
+	/* Create SG BO */
+	memset(&bp, 0, sizeof(bp));
+	bp.size = adev->gart.table_size;
+	bp.byte_align = PAGE_SIZE;
+	bp.domain = AMDGPU_GEM_DOMAIN_CPU;
+	bp.type = ttm_bo_type_sg;
+	bp.resv = NULL;
+	bp.bo_ptr_size = sizeof(struct amdgpu_bo);
+	bp.flags = 0;
+	ret = amdgpu_bo_create(adev, &bp, &bo);
+	if (ret)
+		goto error;
+
+	bo->tbo.sg = sg;
+	bo->tbo.ttm->sg = sg;
+	bo->allowed_domains = AMDGPU_GEM_DOMAIN_GTT;
+	bo->preferred_domains = AMDGPU_GEM_DOMAIN_GTT;
+
+	ret = amdgpu_bo_reserve(bo, true);
+	if (ret) {
+		dev_err(adev->dev, "(%d) failed to reserve bo for GART system bo\n", ret);
+		goto error;
+	}
+
+	ret = amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT);
+	WARN(ret, "Pinning the GART table failed");
+	if (ret)
+		goto error_resv;
+
+	adev->gart.bo = bo;
+	adev->gart.ptr = page_to_virt(p);
+	/* Make GART table accessible in VMID0 */
+	ret = amdgpu_ttm_alloc_gart(&adev->gart.bo->tbo);
+	if (ret)
+		amdgpu_gart_table_ram_free(adev);
+	amdgpu_bo_unreserve(bo);
+
+	return 0;
+
+error_resv:
+	amdgpu_bo_unreserve(bo);
+error:
+	amdgpu_bo_unref(&bo);
+	if (sg) {
+		sg_free_table(sg);
+		kfree(sg);
+	}
+	__free_pages(p, order);
+	return ret;
+}
+
+/**
+ * amdgpu_gart_table_ram_free - free gart page table system ram
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Free the system memory used for the GART page tableon ASICs that don't
+ * have dedicated VRAM.
+ */
+void amdgpu_gart_table_ram_free(struct amdgpu_device *adev)
+{
+	unsigned int order = get_order(adev->gart.table_size);
+	struct sg_table *sg = adev->gart.bo->tbo.sg;
+	struct page *p;
+	unsigned long x;
+	int ret;
+
+	ret = amdgpu_bo_reserve(adev->gart.bo, false);
+	if (!ret) {
+		amdgpu_bo_unpin(adev->gart.bo);
+		amdgpu_bo_unreserve(adev->gart.bo);
+	}
+	amdgpu_bo_unref(&adev->gart.bo);
+	sg_free_table(sg);
+	kfree(sg);
+	p = virt_to_page(adev->gart.ptr);
+	for (x = 0; x < (1UL << order); x++)
+		p[x].mapping = NULL;
+	__free_pages(p, order);
+
+	adev->gart.ptr = NULL;
+}
+
+/**
  * amdgpu_gart_table_vram_alloc - allocate vram for gart page table
  *
  * @adev: amdgpu_device pointer
@@ -154,7 +302,6 @@ void amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset,
 			int pages)
 {
 	unsigned t;
-	unsigned p;
 	int i, j;
 	u64 page_base;
 	/* Starting from VEGA10, system bit must be 0 to mean invalid. */
@@ -168,8 +315,7 @@ void amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset,
 		return;
 
 	t = offset / AMDGPU_GPU_PAGE_SIZE;
-	p = t / AMDGPU_GPU_PAGES_IN_CPU_PAGE;
-	for (i = 0; i < pages; i++, p++) {
+	for (i = 0; i < pages; i++) {
 		page_base = adev->dummy_page_addr;
 		if (!adev->gart.ptr)
 			continue;
@@ -180,10 +326,7 @@ void amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset,
 			page_base += AMDGPU_GPU_PAGE_SIZE;
 		}
 	}
-	mb();
-	amdgpu_device_flush_hdp(adev, NULL);
-	for (i = 0; i < adev->num_vmhubs; i++)
-		amdgpu_gmc_flush_gpu_tlb(adev, 0, i, 0);
+	amdgpu_gart_invalidate_tlb(adev);
 
 	drm_dev_exit(idx);
 }
@@ -225,6 +368,42 @@ void amdgpu_gart_map(struct amdgpu_device *adev, uint64_t offset,
 }
 
 /**
+ * amdgpu_gart_map_vram_range - map VRAM pages into the GART page table
+ *
+ * @adev: amdgpu_device pointer
+ * @pa: physical address of the first page to be mapped
+ * @start_page: first page to map in the GART aperture
+ * @num_pages: number of pages to be mapped
+ * @flags: page table entry flags
+ * @dst: CPU address of the GART table
+ *
+ * Binds a BO that is allocated in VRAM to the GART page table
+ * (all ASICs).
+ *
+ * Useful when a kernel BO is located in VRAM but
+ * needs to be accessed from the GART address space.
+ */
+void amdgpu_gart_map_vram_range(struct amdgpu_device *adev, uint64_t pa,
+				uint64_t start_page, uint64_t num_pages,
+				uint64_t flags, void *dst)
+{
+	u32 i, idx;
+
+	/* The SYSTEM flag indicates the pages aren't in VRAM. */
+	WARN_ON_ONCE(flags & AMDGPU_PTE_SYSTEM);
+
+	if (!drm_dev_enter(adev_to_drm(adev), &idx))
+		return;
+
+	for (i = 0; i < num_pages; ++i) {
+		amdgpu_gmc_set_pte_pde(adev, adev->gart.ptr,
+			start_page + i, pa + AMDGPU_GPU_PAGE_SIZE * i, flags);
+	}
+
+	drm_dev_exit(idx);
+}
+
+/**
  * amdgpu_gart_bind - bind pages into the gart page table
  *
  * @adev: amdgpu_device pointer
@@ -263,8 +442,11 @@ void amdgpu_gart_invalidate_tlb(struct amdgpu_device *adev)
 		return;
 
 	mb();
-	amdgpu_device_flush_hdp(adev, NULL);
-	for (i = 0; i < adev->num_vmhubs; i++)
+	if (down_read_trylock(&adev->reset_domain->sem)) {
+		amdgpu_device_flush_hdp(adev, NULL);
+		up_read(&adev->reset_domain->sem);
+	}
+	for_each_set_bit(i, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS)
 		amdgpu_gmc_flush_gpu_tlb(adev, 0, i, 0);
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h
index 8fea3e04e411..d3118275ddae 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h
@@ -51,10 +51,10 @@ struct amdgpu_gart {
 	uint64_t			gart_pte_flags;
 };
 
+int amdgpu_gart_table_ram_alloc(struct amdgpu_device *adev);
+void amdgpu_gart_table_ram_free(struct amdgpu_device *adev);
 int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev);
 void amdgpu_gart_table_vram_free(struct amdgpu_device *adev);
-int amdgpu_gart_table_vram_pin(struct amdgpu_device *adev);
-void amdgpu_gart_table_vram_unpin(struct amdgpu_device *adev);
 int amdgpu_gart_init(struct amdgpu_device *adev);
 void amdgpu_gart_dummy_page_fini(struct amdgpu_device *adev);
 void amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset,
@@ -64,5 +64,8 @@ void amdgpu_gart_map(struct amdgpu_device *adev, uint64_t offset,
 		     void *dst);
 void amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset,
 		      int pages, dma_addr_t *dma_addr, uint64_t flags);
+void amdgpu_gart_map_vram_range(struct amdgpu_device *adev, uint64_t pa,
+				uint64_t start_page, uint64_t num_pages,
+				uint64_t flags, void *dst);
 void amdgpu_gart_invalidate_tlb(struct amdgpu_device *adev);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
index 863cb668e000..3e38c5db2987 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
@@ -33,16 +33,125 @@
 
 #include <drm/amdgpu_drm.h>
 #include <drm/drm_drv.h>
+#include <drm/drm_exec.h>
 #include <drm/drm_gem_ttm_helper.h>
 #include <drm/ttm/ttm_tt.h>
+#include <drm/drm_syncobj.h>
 
 #include "amdgpu.h"
 #include "amdgpu_display.h"
 #include "amdgpu_dma_buf.h"
 #include "amdgpu_hmm.h"
 #include "amdgpu_xgmi.h"
+#include "amdgpu_vm.h"
 
-static const struct drm_gem_object_funcs amdgpu_gem_object_funcs;
+static int
+amdgpu_gem_add_input_fence(struct drm_file *filp,
+			   uint64_t syncobj_handles_array,
+			   uint32_t num_syncobj_handles)
+{
+	struct dma_fence *fence;
+	uint32_t *syncobj_handles;
+	int ret, i;
+
+	if (!num_syncobj_handles)
+		return 0;
+
+	syncobj_handles = memdup_user(u64_to_user_ptr(syncobj_handles_array),
+				      size_mul(sizeof(uint32_t), num_syncobj_handles));
+	if (IS_ERR(syncobj_handles))
+		return PTR_ERR(syncobj_handles);
+
+	for (i = 0; i < num_syncobj_handles; i++) {
+
+		if (!syncobj_handles[i]) {
+			ret = -EINVAL;
+			goto free_memdup;
+		}
+
+		ret = drm_syncobj_find_fence(filp, syncobj_handles[i], 0, 0, &fence);
+		if (ret)
+			goto free_memdup;
+
+		dma_fence_wait(fence, false);
+
+		/* TODO: optimize async handling */
+		dma_fence_put(fence);
+	}
+
+free_memdup:
+	kfree(syncobj_handles);
+	return ret;
+}
+
+static int
+amdgpu_gem_update_timeline_node(struct drm_file *filp,
+				uint32_t syncobj_handle,
+				uint64_t point,
+				struct drm_syncobj **syncobj,
+				struct dma_fence_chain **chain)
+{
+	if (!syncobj_handle)
+		return 0;
+
+	/* Find the sync object */
+	*syncobj = drm_syncobj_find(filp, syncobj_handle);
+	if (!*syncobj)
+		return -ENOENT;
+
+	if (!point)
+		return 0;
+
+	/* Allocate the chain node */
+	*chain = dma_fence_chain_alloc();
+	if (!*chain) {
+		drm_syncobj_put(*syncobj);
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static void
+amdgpu_gem_update_bo_mapping(struct drm_file *filp,
+			     struct amdgpu_bo_va *bo_va,
+			     uint32_t operation,
+			     uint64_t point,
+			     struct dma_fence *fence,
+			     struct drm_syncobj *syncobj,
+			     struct dma_fence_chain *chain)
+{
+	struct amdgpu_bo *bo = bo_va ? bo_va->base.bo : NULL;
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
+	struct amdgpu_vm *vm = &fpriv->vm;
+	struct dma_fence *last_update;
+
+	if (!syncobj)
+		return;
+
+	/* Find the last update fence */
+	switch (operation) {
+	case AMDGPU_VA_OP_MAP:
+	case AMDGPU_VA_OP_REPLACE:
+		if (bo && (bo->tbo.base.resv == vm->root.bo->tbo.base.resv))
+			last_update = vm->last_update;
+		else
+			last_update = bo_va->last_pt_update;
+		break;
+	case AMDGPU_VA_OP_UNMAP:
+	case AMDGPU_VA_OP_CLEAR:
+		last_update = fence;
+		break;
+	default:
+		return;
+	}
+
+	/* Add fence to timeline */
+	if (!point)
+		drm_syncobj_replace_fence(syncobj, last_update);
+	else
+		drm_syncobj_add_point(syncobj, chain, last_update, point);
+}
 
 static vm_fault_t amdgpu_gem_fault(struct vm_fault *vmf)
 {
@@ -86,19 +195,17 @@ static const struct vm_operations_struct amdgpu_gem_vm_ops = {
 
 static void amdgpu_gem_object_free(struct drm_gem_object *gobj)
 {
-	struct amdgpu_bo *robj = gem_to_amdgpu_bo(gobj);
+	struct amdgpu_bo *aobj = gem_to_amdgpu_bo(gobj);
 
-	if (robj) {
-		amdgpu_hmm_unregister(robj);
-		amdgpu_bo_unref(&robj);
-	}
+	amdgpu_hmm_unregister(aobj);
+	ttm_bo_fini(&aobj->tbo);
 }
 
 int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
 			     int alignment, u32 initial_domain,
 			     u64 flags, enum ttm_bo_type type,
 			     struct dma_resv *resv,
-			     struct drm_gem_object **obj)
+			     struct drm_gem_object **obj, int8_t xcp_id_plus1)
 {
 	struct amdgpu_bo *bo;
 	struct amdgpu_bo_user *ubo;
@@ -107,6 +214,7 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
 
 	memset(&bp, 0, sizeof(bp));
 	*obj = NULL;
+	flags |= AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE;
 
 	bp.size = size;
 	bp.byte_align = alignment;
@@ -116,6 +224,7 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
 	bp.flags = flags;
 	bp.domain = initial_domain;
 	bp.bo_ptr_size = sizeof(struct amdgpu_bo);
+	bp.xcp_id_plus1 = xcp_id_plus1;
 
 	r = amdgpu_bo_create_user(adev, &bp, &ubo);
 	if (r)
@@ -123,7 +232,6 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
 
 	bo = &ubo->bo;
 	*obj = &bo->tbo.base;
-	(*obj)->funcs = &amdgpu_gem_object_funcs;
 
 	return 0;
 }
@@ -172,21 +280,63 @@ static int amdgpu_gem_object_open(struct drm_gem_object *obj,
 		return -EPERM;
 
 	if (abo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID &&
-	    abo->tbo.base.resv != vm->root.bo->tbo.base.resv)
+	    !amdgpu_vm_is_bo_always_valid(vm, abo))
 		return -EPERM;
 
 	r = amdgpu_bo_reserve(abo, false);
 	if (r)
 		return r;
 
+	amdgpu_vm_bo_update_shared(abo);
 	bo_va = amdgpu_vm_bo_find(vm, abo);
-	if (!bo_va) {
+	if (!bo_va)
 		bo_va = amdgpu_vm_bo_add(adev, vm, abo);
-	} else {
+	else
 		++bo_va->ref_count;
+
+	/* attach gfx eviction fence */
+	r = amdgpu_eviction_fence_attach(&fpriv->evf_mgr, abo);
+	if (r) {
+		DRM_DEBUG_DRIVER("Failed to attach eviction fence to BO\n");
+		amdgpu_bo_unreserve(abo);
+		return r;
 	}
+
 	amdgpu_bo_unreserve(abo);
-	return 0;
+
+	/* Validate and add eviction fence to DMABuf imports with dynamic
+	 * attachment in compute VMs. Re-validation will be done by
+	 * amdgpu_vm_validate. Fences are on the reservation shared with the
+	 * export, which is currently required to be validated and fenced
+	 * already by amdgpu_amdkfd_gpuvm_restore_process_bos.
+	 *
+	 * Nested locking below for the case that a GEM object is opened in
+	 * kfd_mem_export_dmabuf. Since the lock below is only taken for imports,
+	 * but not for export, this is a different lock class that cannot lead to
+	 * circular lock dependencies.
+	 */
+	if (!vm->is_compute_context || !vm->process_info)
+		return 0;
+	if (!drm_gem_is_imported(obj) ||
+	    !dma_buf_is_dynamic(obj->import_attach->dmabuf))
+		return 0;
+	mutex_lock_nested(&vm->process_info->lock, 1);
+	if (!WARN_ON(!vm->process_info->eviction_fence)) {
+		r = amdgpu_amdkfd_bo_validate_and_fence(abo, AMDGPU_GEM_DOMAIN_GTT,
+							&vm->process_info->eviction_fence->base);
+		if (r) {
+			struct amdgpu_task_info *ti = amdgpu_vm_get_task_info_vm(vm);
+
+			dev_warn(adev->dev, "validate_and_fence failed: %d\n", r);
+			if (ti) {
+				dev_warn(adev->dev, "pid %d\n", ti->task.pid);
+				amdgpu_vm_put_task_info(ti);
+			}
+		}
+	}
+	mutex_unlock(&vm->process_info->lock);
+
+	return r;
 }
 
 static void amdgpu_gem_object_close(struct drm_gem_object *obj,
@@ -197,38 +347,40 @@ static void amdgpu_gem_object_close(struct drm_gem_object *obj,
 	struct amdgpu_fpriv *fpriv = file_priv->driver_priv;
 	struct amdgpu_vm *vm = &fpriv->vm;
 
-	struct amdgpu_bo_list_entry vm_pd;
-	struct list_head list, duplicates;
 	struct dma_fence *fence = NULL;
-	struct ttm_validate_buffer tv;
-	struct ww_acquire_ctx ticket;
 	struct amdgpu_bo_va *bo_va;
+	struct drm_exec exec;
 	long r;
 
-	INIT_LIST_HEAD(&list);
-	INIT_LIST_HEAD(&duplicates);
-
-	tv.bo = &bo->tbo;
-	tv.num_shared = 2;
-	list_add(&tv.head, &list);
+	drm_exec_init(&exec, DRM_EXEC_IGNORE_DUPLICATES, 0);
+	drm_exec_until_all_locked(&exec) {
+		r = drm_exec_prepare_obj(&exec, &bo->tbo.base, 1);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(r))
+			goto out_unlock;
+
+		r = amdgpu_vm_lock_pd(vm, &exec, 0);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(r))
+			goto out_unlock;
+	}
 
-	amdgpu_vm_get_pd_bo(vm, &list, &vm_pd);
+	if (!amdgpu_vm_is_bo_always_valid(vm, bo))
+		amdgpu_eviction_fence_detach(&fpriv->evf_mgr, bo);
 
-	r = ttm_eu_reserve_buffers(&ticket, &list, false, &duplicates);
-	if (r) {
-		dev_err(adev->dev, "leaking bo va because "
-			"we fail to reserve bo (%ld)\n", r);
-		return;
-	}
 	bo_va = amdgpu_vm_bo_find(vm, bo);
 	if (!bo_va || --bo_va->ref_count)
 		goto out_unlock;
 
 	amdgpu_vm_bo_del(adev, bo_va);
+	amdgpu_vm_bo_update_shared(bo);
 	if (!amdgpu_vm_ready(vm))
 		goto out_unlock;
 
 	r = amdgpu_vm_clear_freed(adev, vm, &fence);
+	if (unlikely(r < 0))
+		dev_err(adev->dev, "failed to clear page "
+			"tables on GEM object close (%ld)\n", r);
 	if (r || !fence)
 		goto out_unlock;
 
@@ -236,10 +388,9 @@ static void amdgpu_gem_object_close(struct drm_gem_object *obj,
 	dma_fence_put(fence);
 
 out_unlock:
-	if (unlikely(r < 0))
-		dev_err(adev->dev, "failed to clear page "
-			"tables on GEM object close (%ld)\n", r);
-	ttm_eu_backoff_reservation(&ticket, &list);
+	if (r)
+		dev_err(adev->dev, "leaking bo va (%ld)\n", r);
+	drm_exec_fini(&exec);
 }
 
 static int amdgpu_gem_object_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma)
@@ -263,7 +414,7 @@ static int amdgpu_gem_object_mmap(struct drm_gem_object *obj, struct vm_area_str
 	return drm_gem_ttm_mmap(obj, vma);
 }
 
-static const struct drm_gem_object_funcs amdgpu_gem_object_funcs = {
+const struct drm_gem_object_funcs amdgpu_gem_object_funcs = {
 	.free = amdgpu_gem_object_free,
 	.open = amdgpu_gem_object_open,
 	.close = amdgpu_gem_object_close,
@@ -292,14 +443,7 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
 	int r;
 
 	/* reject invalid gem flags */
-	if (flags & ~(AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
-		      AMDGPU_GEM_CREATE_NO_CPU_ACCESS |
-		      AMDGPU_GEM_CREATE_CPU_GTT_USWC |
-		      AMDGPU_GEM_CREATE_VRAM_CLEARED |
-		      AMDGPU_GEM_CREATE_VM_ALWAYS_VALID |
-		      AMDGPU_GEM_CREATE_EXPLICIT_SYNC |
-		      AMDGPU_GEM_CREATE_ENCRYPTED |
-		      AMDGPU_GEM_CREATE_DISCARDABLE))
+	if (flags & ~AMDGPU_GEM_CREATE_SETTABLE_MASK)
 		return -EINVAL;
 
 	/* reject invalid gem domains */
@@ -311,6 +455,12 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
 		return -EINVAL;
 	}
 
+	/* always clear VRAM */
+	flags |= AMDGPU_GEM_CREATE_VRAM_CLEARED;
+
+	if (args->in.domains & AMDGPU_GEM_DOMAIN_MMIO_REMAP)
+		return -EINVAL;
+
 	/* create a gem object to contain this object in */
 	if (args->in.domains & (AMDGPU_GEM_DOMAIN_GDS |
 	    AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA)) {
@@ -336,7 +486,7 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data,
 retry:
 	r = amdgpu_gem_object_create(adev, size, args->in.alignment,
 				     initial_domain,
-				     flags, ttm_bo_type_device, resv, &gobj);
+				     flags, ttm_bo_type_device, resv, &gobj, fpriv->xcp_id + 1);
 	if (r && r != -ERESTARTSYS) {
 		if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) {
 			flags &= ~AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
@@ -379,8 +529,9 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 	struct ttm_operation_ctx ctx = { true, false };
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct drm_amdgpu_gem_userptr *args = data;
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
 	struct drm_gem_object *gobj;
-	struct hmm_range *range;
+	struct amdgpu_hmm_range *range;
 	struct amdgpu_bo *bo;
 	uint32_t handle;
 	int r;
@@ -405,7 +556,7 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 
 	/* create a gem object to contain this object in */
 	r = amdgpu_gem_object_create(adev, args->size, 0, AMDGPU_GEM_DOMAIN_CPU,
-				     0, ttm_bo_type_device, NULL, &gobj);
+				     0, ttm_bo_type_device, NULL, &gobj, fpriv->xcp_id + 1);
 	if (r)
 		return r;
 
@@ -421,15 +572,20 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 		goto release_object;
 
 	if (args->flags & AMDGPU_GEM_USERPTR_VALIDATE) {
-		r = amdgpu_ttm_tt_get_user_pages(bo, bo->tbo.ttm->pages,
-						 &range);
-		if (r)
+		range = amdgpu_hmm_range_alloc(NULL);
+		if (unlikely(!range))
+			return -ENOMEM;
+		r = amdgpu_ttm_tt_get_user_pages(bo, range);
+		if (r) {
+			amdgpu_hmm_range_free(range);
 			goto release_object;
-
+		}
 		r = amdgpu_bo_reserve(bo, true);
 		if (r)
 			goto user_pages_done;
 
+		amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, range);
+
 		amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT);
 		r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 		amdgpu_bo_unreserve(bo);
@@ -445,8 +601,7 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data,
 
 user_pages_done:
 	if (args->flags & AMDGPU_GEM_USERPTR_VALIDATE)
-		amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm, range);
-
+		amdgpu_hmm_range_free(range);
 release_object:
 	drm_gem_object_put(gobj);
 
@@ -461,9 +616,9 @@ int amdgpu_mode_dumb_mmap(struct drm_file *filp,
 	struct amdgpu_bo *robj;
 
 	gobj = drm_gem_object_lookup(filp, handle);
-	if (gobj == NULL) {
+	if (!gobj)
 		return -ENOENT;
-	}
+
 	robj = gem_to_amdgpu_bo(gobj);
 	if (amdgpu_ttm_tt_get_usermm(robj->tbo.ttm) ||
 	    (robj->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)) {
@@ -480,6 +635,7 @@ int amdgpu_gem_mmap_ioctl(struct drm_device *dev, void *data,
 {
 	union drm_amdgpu_gem_mmap *args = data;
 	uint32_t handle = args->in.handle;
+
 	memset(args, 0, sizeof(*args));
 	return amdgpu_mode_dumb_mmap(filp, dev, handle, &args->out.addr_ptr);
 }
@@ -506,7 +662,7 @@ unsigned long amdgpu_gem_timeout(uint64_t timeout_ns)
 
 	timeout_jiffies = nsecs_to_jiffies(ktime_to_ns(timeout));
 	/*  clamp timeout to avoid unsigned-> signed overflow */
-	if (timeout_jiffies > MAX_SCHEDULE_TIMEOUT )
+	if (timeout_jiffies > MAX_SCHEDULE_TIMEOUT)
 		return MAX_SCHEDULE_TIMEOUT - 1;
 
 	return timeout_jiffies;
@@ -524,9 +680,9 @@ int amdgpu_gem_wait_idle_ioctl(struct drm_device *dev, void *data,
 	long ret;
 
 	gobj = drm_gem_object_lookup(filp, handle);
-	if (gobj == NULL) {
+	if (!gobj)
 		return -ENOENT;
-	}
+
 	robj = gem_to_amdgpu_bo(gobj);
 	ret = dma_resv_wait_timeout(robj->tbo.base.resv, DMA_RESV_USAGE_READ,
 				    true, timeout);
@@ -553,7 +709,7 @@ int amdgpu_gem_metadata_ioctl(struct drm_device *dev, void *data,
 	struct amdgpu_bo *robj;
 	int r = -1;
 
-	DRM_DEBUG("%d \n", args->handle);
+	DRM_DEBUG("%d\n", args->handle);
 	gobj = drm_gem_object_lookup(filp, args->handle);
 	if (gobj == NULL)
 		return -ENOENT;
@@ -598,18 +754,23 @@ out:
  *
  * Update the bo_va directly after setting its address. Errors are not
  * vital here, so they are not reported back to userspace.
+ *
+ * Returns resulting fence if freed BO(s) got cleared from the PT.
+ * otherwise stub fence in case of error.
  */
-static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
-				    struct amdgpu_vm *vm,
-				    struct amdgpu_bo_va *bo_va,
-				    uint32_t operation)
+static struct dma_fence *
+amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
+			struct amdgpu_vm *vm,
+			struct amdgpu_bo_va *bo_va,
+			uint32_t operation)
 {
+	struct dma_fence *fence = dma_fence_get_stub();
 	int r;
 
 	if (!amdgpu_vm_ready(vm))
-		return;
+		return fence;
 
-	r = amdgpu_vm_clear_freed(adev, vm, NULL);
+	r = amdgpu_vm_clear_freed(adev, vm, &fence);
 	if (r)
 		goto error;
 
@@ -625,36 +786,8 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
 error:
 	if (r && r != -ERESTARTSYS)
 		DRM_ERROR("Couldn't update BO_VA (%d)\n", r);
-}
 
-/**
- * amdgpu_gem_va_map_flags - map GEM UAPI flags into hardware flags
- *
- * @adev: amdgpu_device pointer
- * @flags: GEM UAPI flags
- *
- * Returns the GEM UAPI flags mapped into hardware for the ASIC.
- */
-uint64_t amdgpu_gem_va_map_flags(struct amdgpu_device *adev, uint32_t flags)
-{
-	uint64_t pte_flag = 0;
-
-	if (flags & AMDGPU_VM_PAGE_EXECUTABLE)
-		pte_flag |= AMDGPU_PTE_EXECUTABLE;
-	if (flags & AMDGPU_VM_PAGE_READABLE)
-		pte_flag |= AMDGPU_PTE_READABLE;
-	if (flags & AMDGPU_VM_PAGE_WRITEABLE)
-		pte_flag |= AMDGPU_PTE_WRITEABLE;
-	if (flags & AMDGPU_VM_PAGE_PRT)
-		pte_flag |= AMDGPU_PTE_PRT;
-	if (flags & AMDGPU_VM_PAGE_NOALLOC)
-		pte_flag |= AMDGPU_PTE_NOALLOC;
-
-	if (adev->gmc.gmc_funcs->map_mtype)
-		pte_flag |= amdgpu_gmc_map_mtype(adev,
-						 flags & AMDGPU_VM_MTYPE_MASK);
-
-	return pte_flag;
+	return fence;
 }
 
 int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
@@ -673,25 +806,24 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 	struct amdgpu_fpriv *fpriv = filp->driver_priv;
 	struct amdgpu_bo *abo;
 	struct amdgpu_bo_va *bo_va;
-	struct amdgpu_bo_list_entry vm_pd;
-	struct ttm_validate_buffer tv;
-	struct ww_acquire_ctx ticket;
-	struct list_head list, duplicates;
-	uint64_t va_flags;
+	struct drm_syncobj *timeline_syncobj = NULL;
+	struct dma_fence_chain *timeline_chain = NULL;
+	struct dma_fence *fence;
+	struct drm_exec exec;
 	uint64_t vm_size;
 	int r = 0;
 
-	if (args->va_address < AMDGPU_VA_RESERVED_SIZE) {
+	if (args->va_address < AMDGPU_VA_RESERVED_BOTTOM) {
 		dev_dbg(dev->dev,
-			"va_address 0x%LX is in reserved area 0x%LX\n",
-			args->va_address, AMDGPU_VA_RESERVED_SIZE);
+			"va_address 0x%llx is in reserved area 0x%llx\n",
+			args->va_address, AMDGPU_VA_RESERVED_BOTTOM);
 		return -EINVAL;
 	}
 
 	if (args->va_address >= AMDGPU_GMC_HOLE_START &&
 	    args->va_address < AMDGPU_GMC_HOLE_END) {
 		dev_dbg(dev->dev,
-			"va_address 0x%LX is in VA hole 0x%LX-0x%LX\n",
+			"va_address 0x%llx is in VA hole 0x%llx-0x%llx\n",
 			args->va_address, AMDGPU_GMC_HOLE_START,
 			AMDGPU_GMC_HOLE_END);
 		return -EINVAL;
@@ -700,7 +832,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 	args->va_address &= AMDGPU_GMC_HOLE_MASK;
 
 	vm_size = adev->vm_manager.max_pfn * AMDGPU_GPU_PAGE_SIZE;
-	vm_size -= AMDGPU_VA_RESERVED_SIZE;
+	vm_size -= AMDGPU_VA_RESERVED_TOP;
 	if (args->va_address + args->map_size > vm_size) {
 		dev_dbg(dev->dev,
 			"va_address 0x%llx is in top reserved area 0x%llx\n",
@@ -726,36 +858,44 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 		return -EINVAL;
 	}
 
-	INIT_LIST_HEAD(&list);
-	INIT_LIST_HEAD(&duplicates);
 	if ((args->operation != AMDGPU_VA_OP_CLEAR) &&
 	    !(args->flags & AMDGPU_VM_PAGE_PRT)) {
 		gobj = drm_gem_object_lookup(filp, args->handle);
 		if (gobj == NULL)
 			return -ENOENT;
 		abo = gem_to_amdgpu_bo(gobj);
-		tv.bo = &abo->tbo;
-		if (abo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID)
-			tv.num_shared = 1;
-		else
-			tv.num_shared = 0;
-		list_add(&tv.head, &list);
 	} else {
 		gobj = NULL;
 		abo = NULL;
 	}
 
-	amdgpu_vm_get_pd_bo(&fpriv->vm, &list, &vm_pd);
-
-	r = ttm_eu_reserve_buffers(&ticket, &list, true, &duplicates);
+	r = amdgpu_gem_add_input_fence(filp,
+				       args->input_fence_syncobj_handles,
+				       args->num_syncobj_handles);
 	if (r)
-		goto error_unref;
+		goto error_put_gobj;
+
+	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT |
+		      DRM_EXEC_IGNORE_DUPLICATES, 0);
+	drm_exec_until_all_locked(&exec) {
+		if (gobj) {
+			r = drm_exec_lock_obj(&exec, gobj);
+			drm_exec_retry_on_contention(&exec);
+			if (unlikely(r))
+				goto error;
+		}
+
+		r = amdgpu_vm_lock_pd(&fpriv->vm, &exec, 2);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(r))
+			goto error;
+	}
 
 	if (abo) {
 		bo_va = amdgpu_vm_bo_find(&fpriv->vm, abo);
 		if (!bo_va) {
 			r = -ENOENT;
-			goto error_backoff;
+			goto error;
 		}
 	} else if (args->operation != AMDGPU_VA_OP_CLEAR) {
 		bo_va = fpriv->prt_va;
@@ -763,12 +903,19 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 		bo_va = NULL;
 	}
 
+	r = amdgpu_gem_update_timeline_node(filp,
+					    args->vm_timeline_syncobj_out,
+					    args->vm_timeline_point,
+					    &timeline_syncobj,
+					    &timeline_chain);
+	if (r)
+		goto error;
+
 	switch (args->operation) {
 	case AMDGPU_VA_OP_MAP:
-		va_flags = amdgpu_gem_va_map_flags(adev, args->flags);
 		r = amdgpu_vm_bo_map(adev, bo_va, args->va_address,
 				     args->offset_in_bo, args->map_size,
-				     va_flags);
+				     args->flags);
 		break;
 	case AMDGPU_VA_OP_UNMAP:
 		r = amdgpu_vm_bo_unmap(adev, bo_va, args->va_address);
@@ -780,22 +927,31 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 						args->map_size);
 		break;
 	case AMDGPU_VA_OP_REPLACE:
-		va_flags = amdgpu_gem_va_map_flags(adev, args->flags);
 		r = amdgpu_vm_bo_replace_map(adev, bo_va, args->va_address,
 					     args->offset_in_bo, args->map_size,
-					     va_flags);
+					     args->flags);
 		break;
 	default:
 		break;
 	}
-	if (!r && !(args->flags & AMDGPU_VM_DELAY_UPDATE) && !amdgpu_vm_debug)
-		amdgpu_gem_va_update_vm(adev, &fpriv->vm, bo_va,
-					args->operation);
+	if (!r && !(args->flags & AMDGPU_VM_DELAY_UPDATE) && !adev->debug_vm) {
+		fence = amdgpu_gem_va_update_vm(adev, &fpriv->vm, bo_va,
+						args->operation);
+
+		if (timeline_syncobj)
+			amdgpu_gem_update_bo_mapping(filp, bo_va,
+					     args->operation,
+					     args->vm_timeline_point,
+					     fence, timeline_syncobj,
+					     timeline_chain);
+		else
+			dma_fence_put(fence);
 
-error_backoff:
-	ttm_eu_backoff_reservation(&ticket, &list);
+	}
 
-error_unref:
+error:
+	drm_exec_fini(&exec);
+error_put_gobj:
 	drm_gem_object_put(gobj);
 	return r;
 }
@@ -803,22 +959,38 @@ error_unref:
 int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *filp)
 {
-	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct drm_amdgpu_gem_op *args = data;
 	struct drm_gem_object *gobj;
 	struct amdgpu_vm_bo_base *base;
 	struct amdgpu_bo *robj;
+	struct drm_exec exec;
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
 	int r;
 
+	if (args->padding)
+		return -EINVAL;
+
 	gobj = drm_gem_object_lookup(filp, args->handle);
-	if (gobj == NULL) {
+	if (!gobj)
 		return -ENOENT;
-	}
+
 	robj = gem_to_amdgpu_bo(gobj);
 
-	r = amdgpu_bo_reserve(robj, false);
-	if (unlikely(r))
-		goto out;
+	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT |
+			  DRM_EXEC_IGNORE_DUPLICATES, 0);
+	drm_exec_until_all_locked(&exec) {
+		r = drm_exec_lock_obj(&exec, gobj);
+		drm_exec_retry_on_contention(&exec);
+		if (r)
+			goto out_exec;
+
+		if (args->op == AMDGPU_GEM_OP_GET_MAPPING_INFO) {
+			r = amdgpu_vm_lock_pd(&fpriv->vm, &exec, 0);
+			drm_exec_retry_on_contention(&exec);
+			if (r)
+				goto out_exec;
+		}
+	}
 
 	switch (args->op) {
 	case AMDGPU_GEM_OP_GET_GEM_CREATE_INFO: {
@@ -829,29 +1001,26 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 		info.alignment = robj->tbo.page_alignment << PAGE_SHIFT;
 		info.domains = robj->preferred_domains;
 		info.domain_flags = robj->flags;
-		amdgpu_bo_unreserve(robj);
+		drm_exec_fini(&exec);
 		if (copy_to_user(out, &info, sizeof(info)))
 			r = -EFAULT;
 		break;
 	}
 	case AMDGPU_GEM_OP_SET_PLACEMENT:
-		if (robj->tbo.base.import_attach &&
+		if (drm_gem_is_imported(&robj->tbo.base) &&
 		    args->value & AMDGPU_GEM_DOMAIN_VRAM) {
 			r = -EINVAL;
-			amdgpu_bo_unreserve(robj);
-			break;
+			goto out_exec;
 		}
 		if (amdgpu_ttm_tt_get_usermm(robj->tbo.ttm)) {
 			r = -EPERM;
-			amdgpu_bo_unreserve(robj);
-			break;
+			goto out_exec;
 		}
 		for (base = robj->vm_bo; base; base = base->next)
 			if (amdgpu_xgmi_same_hive(amdgpu_ttm_adev(robj->tbo.bdev),
 				amdgpu_ttm_adev(base->vm->root.bo->tbo.bdev))) {
 				r = -EINVAL;
-				amdgpu_bo_unreserve(robj);
-				goto out;
+				goto out_exec;
 			}
 
 
@@ -863,18 +1032,147 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 			robj->allowed_domains |= AMDGPU_GEM_DOMAIN_GTT;
 
 		if (robj->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID)
-			amdgpu_vm_bo_invalidate(adev, robj, true);
+			amdgpu_vm_bo_invalidate(robj, true);
+		drm_exec_fini(&exec);
+		break;
+	case AMDGPU_GEM_OP_GET_MAPPING_INFO: {
+		struct amdgpu_bo_va *bo_va = amdgpu_vm_bo_find(&fpriv->vm, robj);
+		struct drm_amdgpu_gem_vm_entry *vm_entries;
+		struct amdgpu_bo_va_mapping *mapping;
+		int num_mappings = 0;
+		/*
+		 * num_entries is set as an input to the size of the user-allocated array of
+		 * drm_amdgpu_gem_vm_entry stored at args->value.
+		 * num_entries is sent back as output as the number of mappings the bo has.
+		 * If that number is larger than the size of the array, the ioctl must
+		 * be retried.
+		 */
+		vm_entries = kvcalloc(args->num_entries, sizeof(*vm_entries), GFP_KERNEL);
+		if (!vm_entries)
+			return -ENOMEM;
+
+		amdgpu_vm_bo_va_for_each_valid_mapping(bo_va, mapping) {
+			if (num_mappings < args->num_entries) {
+				vm_entries[num_mappings].addr = mapping->start * AMDGPU_GPU_PAGE_SIZE;
+				vm_entries[num_mappings].size = (mapping->last - mapping->start + 1) * AMDGPU_GPU_PAGE_SIZE;
+				vm_entries[num_mappings].offset = mapping->offset;
+				vm_entries[num_mappings].flags = mapping->flags;
+			}
+			num_mappings += 1;
+		}
+
+		amdgpu_vm_bo_va_for_each_invalid_mapping(bo_va, mapping) {
+			if (num_mappings < args->num_entries) {
+				vm_entries[num_mappings].addr = mapping->start * AMDGPU_GPU_PAGE_SIZE;
+				vm_entries[num_mappings].size = (mapping->last - mapping->start + 1) * AMDGPU_GPU_PAGE_SIZE;
+				vm_entries[num_mappings].offset = mapping->offset;
+				vm_entries[num_mappings].flags = mapping->flags;
+			}
+			num_mappings += 1;
+		}
 
-		amdgpu_bo_unreserve(robj);
+		drm_exec_fini(&exec);
+
+		if (num_mappings > 0 && num_mappings <= args->num_entries)
+			if (copy_to_user(u64_to_user_ptr(args->value), vm_entries, num_mappings * sizeof(*vm_entries)))
+				r = -EFAULT;
+
+		args->num_entries = num_mappings;
+
+		kvfree(vm_entries);
 		break;
+	}
 	default:
-		amdgpu_bo_unreserve(robj);
+		drm_exec_fini(&exec);
 		r = -EINVAL;
 	}
 
-out:
 	drm_gem_object_put(gobj);
 	return r;
+out_exec:
+	drm_exec_fini(&exec);
+	drm_gem_object_put(gobj);
+	return r;
+}
+
+/**
+ * amdgpu_gem_list_handles_ioctl - get information about a process' buffer objects
+ *
+ * @dev: drm device pointer
+ * @data: drm_amdgpu_gem_list_handles
+ * @filp: drm file pointer
+ *
+ * num_entries is set as an input to the size of the entries array.
+ * num_entries is sent back as output as the number of bos in the process.
+ * If that number is larger than the size of the array, the ioctl must
+ * be retried.
+ *
+ * Returns:
+ * 0 for success, -errno for errors.
+ */
+int amdgpu_gem_list_handles_ioctl(struct drm_device *dev, void *data,
+				  struct drm_file *filp)
+{
+	struct drm_amdgpu_gem_list_handles *args = data;
+	struct drm_amdgpu_gem_list_handles_entry *bo_entries;
+	struct drm_gem_object *gobj;
+	int id, ret = 0;
+	int bo_index = 0;
+	int num_bos = 0;
+
+	spin_lock(&filp->table_lock);
+	idr_for_each_entry(&filp->object_idr, gobj, id)
+		num_bos += 1;
+	spin_unlock(&filp->table_lock);
+
+	if (args->num_entries < num_bos) {
+		args->num_entries = num_bos;
+		return 0;
+	}
+
+	if (num_bos == 0) {
+		args->num_entries = 0;
+		return 0;
+	}
+
+	bo_entries = kvcalloc(num_bos, sizeof(*bo_entries), GFP_KERNEL);
+	if (!bo_entries)
+		return -ENOMEM;
+
+	spin_lock(&filp->table_lock);
+	idr_for_each_entry(&filp->object_idr, gobj, id) {
+		struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj);
+		struct drm_amdgpu_gem_list_handles_entry *bo_entry;
+
+		if (bo_index >= num_bos) {
+			ret = -EAGAIN;
+			break;
+		}
+
+		bo_entry = &bo_entries[bo_index];
+
+		bo_entry->size = amdgpu_bo_size(bo);
+		bo_entry->alloc_flags = bo->flags & AMDGPU_GEM_CREATE_SETTABLE_MASK;
+		bo_entry->preferred_domains = bo->preferred_domains;
+		bo_entry->gem_handle = id;
+		bo_entry->alignment = bo->tbo.page_alignment;
+
+		if (bo->tbo.base.import_attach)
+			bo_entry->flags |= AMDGPU_GEM_LIST_HANDLES_FLAG_IS_IMPORT;
+
+		bo_index += 1;
+	}
+	spin_unlock(&filp->table_lock);
+
+	args->num_entries = bo_index;
+
+	if (!ret)
+		if (copy_to_user(u64_to_user_ptr(args->entries), bo_entries, num_bos * sizeof(*bo_entries)))
+			ret = -EFAULT;
+
+	kvfree(bo_entries);
+
+	return ret;
 }
 
 static int amdgpu_gem_align_pitch(struct amdgpu_device *adev,
@@ -908,6 +1206,7 @@ int amdgpu_mode_dumb_create(struct drm_file *file_priv,
 			    struct drm_mode_create_dumb *args)
 {
 	struct amdgpu_device *adev = drm_to_adev(dev);
+	struct amdgpu_fpriv *fpriv = file_priv->driver_priv;
 	struct drm_gem_object *gobj;
 	uint32_t handle;
 	u64 flags = AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
@@ -931,16 +1230,16 @@ int amdgpu_mode_dumb_create(struct drm_file *file_priv,
 	domain = amdgpu_bo_get_preferred_domain(adev,
 				amdgpu_display_supported_domains(adev, flags));
 	r = amdgpu_gem_object_create(adev, args->size, 0, domain, flags,
-				     ttm_bo_type_device, NULL, &gobj);
+				     ttm_bo_type_device, NULL, &gobj, fpriv->xcp_id + 1);
 	if (r)
 		return -ENOMEM;
 
 	r = drm_gem_handle_create(file_priv, gobj, &handle);
 	/* drop reference from allocate - handle holds it now */
 	drm_gem_object_put(gobj);
-	if (r) {
+	if (r)
 		return r;
-	}
+
 	args->handle = handle;
 	return 0;
 }
@@ -948,7 +1247,7 @@ int amdgpu_mode_dumb_create(struct drm_file *file_priv,
 #if defined(CONFIG_DEBUG_FS)
 static int amdgpu_debugfs_gem_info_show(struct seq_file *m, void *unused)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct amdgpu_device *adev = m->private;
 	struct drm_device *dev = adev_to_drm(adev);
 	struct drm_file *file;
 	int r;
@@ -960,6 +1259,7 @@ static int amdgpu_debugfs_gem_info_show(struct seq_file *m, void *unused)
 	list_for_each_entry(file, &dev->filelist, lhead) {
 		struct task_struct *task;
 		struct drm_gem_object *gobj;
+		struct pid *pid;
 		int id;
 
 		/*
@@ -969,8 +1269,9 @@ static int amdgpu_debugfs_gem_info_show(struct seq_file *m, void *unused)
 		 * Therefore, we need to protect this ->comm access using RCU.
 		 */
 		rcu_read_lock();
-		task = pid_task(file->pid, PIDTYPE_TGID);
-		seq_printf(m, "pid %8d command %s:\n", pid_nr(file->pid),
+		pid = rcu_dereference(file->pid);
+		task = pid_task(pid, PIDTYPE_TGID);
+		seq_printf(m, "pid %8d command %s:\n", pid_nr(pid),
 			   task ? task->comm : "<unknown>");
 		rcu_read_unlock();
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h
index 637bf51dbf06..b558336bc4c6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h
@@ -33,6 +33,8 @@
 #define AMDGPU_GEM_DOMAIN_MAX		0x3
 #define gem_to_amdgpu_bo(gobj) container_of((gobj), struct amdgpu_bo, tbo.base)
 
+extern const struct drm_gem_object_funcs amdgpu_gem_object_funcs;
+
 unsigned long amdgpu_gem_timeout(uint64_t timeout_ns);
 
 /*
@@ -43,8 +45,7 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
 			     int alignment, u32 initial_domain,
 			     u64 flags, enum ttm_bo_type type,
 			     struct dma_resv *resv,
-			     struct drm_gem_object **obj);
-
+			     struct drm_gem_object **obj, int8_t xcp_id_plus1);
 int amdgpu_mode_dumb_create(struct drm_file *file_priv,
 			    struct drm_device *dev,
 			    struct drm_mode_create_dumb *args);
@@ -62,13 +63,28 @@ int amdgpu_gem_mmap_ioctl(struct drm_device *dev, void *data,
 			  struct drm_file *filp);
 int amdgpu_gem_wait_idle_ioctl(struct drm_device *dev, void *data,
 			      struct drm_file *filp);
-uint64_t amdgpu_gem_va_map_flags(struct amdgpu_device *adev, uint32_t flags);
 int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data,
 			  struct drm_file *filp);
 int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data,
 			struct drm_file *filp);
+int amdgpu_gem_list_handles_ioctl(struct drm_device *dev, void *data,
+				  struct drm_file *filp);
 
 int amdgpu_gem_metadata_ioctl(struct drm_device *dev, void *data,
 				struct drm_file *filp);
 
+#define AMDGPU_GEM_CREATE_SETTABLE_MASK	(AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | \
+	AMDGPU_GEM_CREATE_NO_CPU_ACCESS | \
+	AMDGPU_GEM_CREATE_CPU_GTT_USWC | \
+	AMDGPU_GEM_CREATE_VRAM_CLEARED | \
+	AMDGPU_GEM_CREATE_VM_ALWAYS_VALID | \
+	AMDGPU_GEM_CREATE_EXPLICIT_SYNC | \
+	AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE | \
+	AMDGPU_GEM_CREATE_ENCRYPTED | \
+	AMDGPU_GEM_CREATE_GFX12_DCC | \
+	AMDGPU_GEM_CREATE_DISCARDABLE | \
+	AMDGPU_GEM_CREATE_COHERENT | \
+	AMDGPU_GEM_CREATE_UNCACHED | \
+	AMDGPU_GEM_CREATE_EXT_COHERENT)
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index c50d59855011..8b118c53f351 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -24,10 +24,17 @@
  */
 
 #include <linux/firmware.h>
+#include <linux/pm_runtime.h>
+
 #include "amdgpu.h"
 #include "amdgpu_gfx.h"
 #include "amdgpu_rlc.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_reset.h"
+#include "amdgpu_xcp.h"
+#include "amdgpu_xgmi.h"
+#include "amdgpu_mes.h"
+#include "nvd.h"
 
 /* delay 0.1 second to enable gfx off feature */
 #define GFX_OFF_DELAY_ENABLE         msecs_to_jiffies(100)
@@ -63,35 +70,26 @@ void amdgpu_queue_mask_bit_to_mec_queue(struct amdgpu_device *adev, int bit,
 }
 
 bool amdgpu_gfx_is_mec_queue_enabled(struct amdgpu_device *adev,
-				     int mec, int pipe, int queue)
+				     int xcc_id, int mec, int pipe, int queue)
 {
 	return test_bit(amdgpu_gfx_mec_queue_to_bit(adev, mec, pipe, queue),
-			adev->gfx.mec.queue_bitmap);
+			adev->gfx.mec_bitmap[xcc_id].queue_bitmap);
 }
 
-int amdgpu_gfx_me_queue_to_bit(struct amdgpu_device *adev,
-			       int me, int pipe, int queue)
+static int amdgpu_gfx_me_queue_to_bit(struct amdgpu_device *adev,
+				      int me, int pipe, int queue)
 {
+	int num_queue_per_pipe = 1; /* we only enable 1 KGQ per pipe */
 	int bit = 0;
 
 	bit += me * adev->gfx.me.num_pipe_per_me
-		* adev->gfx.me.num_queue_per_pipe;
-	bit += pipe * adev->gfx.me.num_queue_per_pipe;
+		* num_queue_per_pipe;
+	bit += pipe * num_queue_per_pipe;
 	bit += queue;
 
 	return bit;
 }
 
-void amdgpu_gfx_bit_to_me_queue(struct amdgpu_device *adev, int bit,
-				int *me, int *pipe, int *queue)
-{
-	*queue = bit % adev->gfx.me.num_queue_per_pipe;
-	*pipe = (bit / adev->gfx.me.num_queue_per_pipe)
-		% adev->gfx.me.num_pipe_per_me;
-	*me = (bit / adev->gfx.me.num_queue_per_pipe)
-		/ adev->gfx.me.num_pipe_per_me;
-}
-
 bool amdgpu_gfx_is_me_queue_enabled(struct amdgpu_device *adev,
 				    int me, int pipe, int queue)
 {
@@ -109,9 +107,9 @@ bool amdgpu_gfx_is_me_queue_enabled(struct amdgpu_device *adev,
  * The bitmask of CUs to be disabled in the shader array determined by se and
  * sh is stored in mask[se * max_sh + sh].
  */
-void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se, unsigned max_sh)
+void amdgpu_gfx_parse_disable_cu(unsigned int *mask, unsigned int max_se, unsigned int max_sh)
 {
-	unsigned se, sh, cu;
+	unsigned int se, sh, cu;
 	const char *p;
 
 	memset(mask, 0, sizeof(*mask) * max_se * max_sh);
@@ -123,6 +121,7 @@ void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se, unsigned max_s
 	for (;;) {
 		char *next;
 		int ret = sscanf(p, "%u.%u.%u", &se, &sh, &cu);
+
 		if (ret < 3) {
 			DRM_ERROR("amdgpu: could not parse disable_cu\n");
 			return;
@@ -151,12 +150,12 @@ static bool amdgpu_gfx_is_graphics_multipipe_capable(struct amdgpu_device *adev)
 static bool amdgpu_gfx_is_compute_multipipe_capable(struct amdgpu_device *adev)
 {
 	if (amdgpu_compute_multipipe != -1) {
-		DRM_INFO("amdgpu: forcing compute pipe policy %d\n",
+		dev_info(adev->dev, "amdgpu: forcing compute pipe policy %d\n",
 			 amdgpu_compute_multipipe);
 		return amdgpu_compute_multipipe == 1;
 	}
 
-	if (adev->ip_versions[GC_HWIP][0] > IP_VERSION(9, 0, 0))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) > IP_VERSION(9, 0, 0))
 		return true;
 
 	/* FIXME: spreading the queues across pipes causes perf regressions
@@ -204,37 +203,46 @@ bool amdgpu_gfx_is_high_priority_compute_queue(struct amdgpu_device *adev,
 
 void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev)
 {
-	int i, queue, pipe;
+	int i, j, queue, pipe;
 	bool multipipe_policy = amdgpu_gfx_is_compute_multipipe_capable(adev);
 	int max_queues_per_mec = min(adev->gfx.mec.num_pipe_per_mec *
 				     adev->gfx.mec.num_queue_per_pipe,
 				     adev->gfx.num_compute_rings);
+	int num_xcc = adev->gfx.xcc_mask ? NUM_XCC(adev->gfx.xcc_mask) : 1;
 
 	if (multipipe_policy) {
-		/* policy: make queues evenly cross all pipes on MEC1 only */
-		for (i = 0; i < max_queues_per_mec; i++) {
-			pipe = i % adev->gfx.mec.num_pipe_per_mec;
-			queue = (i / adev->gfx.mec.num_pipe_per_mec) %
-				adev->gfx.mec.num_queue_per_pipe;
-
-			set_bit(pipe * adev->gfx.mec.num_queue_per_pipe + queue,
-					adev->gfx.mec.queue_bitmap);
+		/* policy: make queues evenly cross all pipes on MEC1 only
+		 * for multiple xcc, just use the original policy for simplicity */
+		for (j = 0; j < num_xcc; j++) {
+			for (i = 0; i < max_queues_per_mec; i++) {
+				pipe = i % adev->gfx.mec.num_pipe_per_mec;
+				queue = (i / adev->gfx.mec.num_pipe_per_mec) %
+					 adev->gfx.mec.num_queue_per_pipe;
+
+				set_bit(pipe * adev->gfx.mec.num_queue_per_pipe + queue,
+					adev->gfx.mec_bitmap[j].queue_bitmap);
+			}
 		}
 	} else {
 		/* policy: amdgpu owns all queues in the given pipe */
-		for (i = 0; i < max_queues_per_mec; ++i)
-			set_bit(i, adev->gfx.mec.queue_bitmap);
+		for (j = 0; j < num_xcc; j++) {
+			for (i = 0; i < max_queues_per_mec; ++i)
+				set_bit(i, adev->gfx.mec_bitmap[j].queue_bitmap);
+		}
 	}
 
-	dev_dbg(adev->dev, "mec queue bitmap weight=%d\n", bitmap_weight(adev->gfx.mec.queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES));
+	for (j = 0; j < num_xcc; j++) {
+		dev_dbg(adev->dev, "mec queue bitmap weight=%d\n",
+			bitmap_weight(adev->gfx.mec_bitmap[j].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES));
+	}
 }
 
 void amdgpu_gfx_graphics_queue_acquire(struct amdgpu_device *adev)
 {
 	int i, queue, pipe;
 	bool multipipe_policy = amdgpu_gfx_is_graphics_multipipe_capable(adev);
-	int max_queues_per_me = adev->gfx.me.num_pipe_per_me *
-					adev->gfx.me.num_queue_per_pipe;
+	int num_queue_per_pipe = 1; /* we only enable 1 KGQ per pipe */
+	int max_queues_per_me = adev->gfx.me.num_pipe_per_me * num_queue_per_pipe;
 
 	if (multipipe_policy) {
 		/* policy: amdgpu owns the first queue per pipe at this stage
@@ -242,9 +250,9 @@ void amdgpu_gfx_graphics_queue_acquire(struct amdgpu_device *adev)
 		for (i = 0; i < max_queues_per_me; i++) {
 			pipe = i % adev->gfx.me.num_pipe_per_me;
 			queue = (i / adev->gfx.me.num_pipe_per_me) %
-				adev->gfx.me.num_queue_per_pipe;
+				num_queue_per_pipe;
 
-			set_bit(pipe * adev->gfx.me.num_queue_per_pipe + queue,
+			set_bit(pipe * num_queue_per_pipe + queue,
 				adev->gfx.me.queue_bitmap);
 		}
 	} else {
@@ -253,12 +261,13 @@ void amdgpu_gfx_graphics_queue_acquire(struct amdgpu_device *adev)
 	}
 
 	/* update the number of active graphics rings */
-	adev->gfx.num_gfx_rings =
-		bitmap_weight(adev->gfx.me.queue_bitmap, AMDGPU_MAX_GFX_QUEUES);
+	if (adev->gfx.num_gfx_rings)
+		adev->gfx.num_gfx_rings =
+			bitmap_weight(adev->gfx.me.queue_bitmap, AMDGPU_MAX_GFX_QUEUES);
 }
 
 static int amdgpu_gfx_kiq_acquire(struct amdgpu_device *adev,
-				  struct amdgpu_ring *ring)
+				  struct amdgpu_ring *ring, int xcc_id)
 {
 	int queue_bit;
 	int mec, pipe, queue;
@@ -268,7 +277,7 @@ static int amdgpu_gfx_kiq_acquire(struct amdgpu_device *adev,
 		    * adev->gfx.mec.num_queue_per_pipe;
 
 	while (--queue_bit >= 0) {
-		if (test_bit(queue_bit, adev->gfx.mec.queue_bitmap))
+		if (test_bit(queue_bit, adev->gfx.mec_bitmap[xcc_id].queue_bitmap))
 			continue;
 
 		amdgpu_queue_mask_bit_to_mec_queue(adev, queue_bit, &mec, &pipe, &queue);
@@ -292,11 +301,11 @@ static int amdgpu_gfx_kiq_acquire(struct amdgpu_device *adev,
 	return -EINVAL;
 }
 
-int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,
-			     struct amdgpu_ring *ring,
-			     struct amdgpu_irq_src *irq)
+int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, int xcc_id)
 {
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
+	struct amdgpu_irq_src *irq = &kiq->irq;
+	struct amdgpu_ring *ring = &kiq->ring;
 	int r = 0;
 
 	spin_lock_init(&kiq->ring_lock);
@@ -304,15 +313,22 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,
 	ring->adev = NULL;
 	ring->ring_obj = NULL;
 	ring->use_doorbell = true;
-	ring->doorbell_index = adev->doorbell_index.kiq;
-
-	r = amdgpu_gfx_kiq_acquire(adev, ring);
+	ring->xcc_id = xcc_id;
+	ring->vm_hub = AMDGPU_GFXHUB(xcc_id);
+	ring->doorbell_index =
+		(adev->doorbell_index.kiq +
+		 xcc_id * adev->doorbell_index.xcc_doorbell_range)
+		<< 1;
+
+	r = amdgpu_gfx_kiq_acquire(adev, ring, xcc_id);
 	if (r)
 		return r;
 
 	ring->eop_gpu_addr = kiq->eop_gpu_addr;
 	ring->no_scheduler = true;
-	sprintf(ring->name, "kiq_%d.%d.%d", ring->me, ring->pipe, ring->queue);
+	snprintf(ring->name, sizeof(ring->name), "kiq_%hhu.%hhu.%hhu.%hhu",
+		 (unsigned char)xcc_id, (unsigned char)ring->me,
+		 (unsigned char)ring->pipe, (unsigned char)ring->queue);
 	r = amdgpu_ring_init(adev, ring, 1024, irq, AMDGPU_CP_KIQ_IRQ_DRIVER0,
 			     AMDGPU_RING_PRIO_DEFAULT, NULL);
 	if (r)
@@ -326,19 +342,19 @@ void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring)
 	amdgpu_ring_fini(ring);
 }
 
-void amdgpu_gfx_kiq_fini(struct amdgpu_device *adev)
+void amdgpu_gfx_kiq_fini(struct amdgpu_device *adev, int xcc_id)
 {
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
 
 	amdgpu_bo_free_kernel(&kiq->eop_obj, &kiq->eop_gpu_addr, NULL);
 }
 
 int amdgpu_gfx_kiq_init(struct amdgpu_device *adev,
-			unsigned hpd_size)
+			unsigned int hpd_size, int xcc_id)
 {
 	int r;
 	u32 *hpd;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
 
 	r = amdgpu_bo_create_kernel(adev, hpd_size, PAGE_SIZE,
 				    AMDGPU_GEM_DOMAIN_GTT, &kiq->eop_obj,
@@ -361,13 +377,20 @@ int amdgpu_gfx_kiq_init(struct amdgpu_device *adev,
 
 /* create MQD for each compute/gfx queue */
 int amdgpu_gfx_mqd_sw_init(struct amdgpu_device *adev,
-			   unsigned mqd_size)
+			   unsigned int mqd_size, int xcc_id)
 {
-	struct amdgpu_ring *ring = NULL;
-	int r, i;
+	int r, i, j;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
+	struct amdgpu_ring *ring = &kiq->ring;
+	u32 domain = AMDGPU_GEM_DOMAIN_GTT;
+
+#if !defined(CONFIG_ARM) && !defined(CONFIG_ARM64)
+	/* Only enable on gfx10 and 11 for now to avoid changing behavior on older chips */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 0, 0))
+		domain |= AMDGPU_GEM_DOMAIN_VRAM;
+#endif
 
 	/* create MQD for KIQ */
-	ring = &adev->gfx.kiq.ring;
 	if (!adev->enable_mes_kiq && !ring->mqd_obj) {
 		/* originaly the KIQ MQD is put in GTT domain, but for SRIOV VRAM domain is a must
 		 * otherwise hypervisor trigger SAVE_VF fail after driver unloaded which mean MQD
@@ -386,9 +409,12 @@ int amdgpu_gfx_mqd_sw_init(struct amdgpu_device *adev,
 		}
 
 		/* prepare MQD backup */
-		adev->gfx.mec.mqd_backup[AMDGPU_MAX_COMPUTE_RINGS] = kmalloc(mqd_size, GFP_KERNEL);
-		if (!adev->gfx.mec.mqd_backup[AMDGPU_MAX_COMPUTE_RINGS])
-				dev_warn(adev->dev, "no memory to create MQD backup for ring %s\n", ring->name);
+		kiq->mqd_backup = kzalloc(mqd_size, GFP_KERNEL);
+		if (!kiq->mqd_backup) {
+			dev_warn(adev->dev,
+				 "no memory to create MQD backup for ring %s\n", ring->name);
+			return -ENOMEM;
+		}
 	}
 
 	if (adev->asic_type >= CHIP_NAVI10 && amdgpu_async_gfx_ring) {
@@ -397,47 +423,55 @@ int amdgpu_gfx_mqd_sw_init(struct amdgpu_device *adev,
 			ring = &adev->gfx.gfx_ring[i];
 			if (!ring->mqd_obj) {
 				r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE,
-							    AMDGPU_GEM_DOMAIN_GTT, &ring->mqd_obj,
+							    domain, &ring->mqd_obj,
 							    &ring->mqd_gpu_addr, &ring->mqd_ptr);
 				if (r) {
 					dev_warn(adev->dev, "failed to create ring mqd bo (%d)", r);
 					return r;
 				}
 
+				ring->mqd_size = mqd_size;
 				/* prepare MQD backup */
-				adev->gfx.me.mqd_backup[i] = kmalloc(mqd_size, GFP_KERNEL);
-				if (!adev->gfx.me.mqd_backup[i])
+				adev->gfx.me.mqd_backup[i] = kzalloc(mqd_size, GFP_KERNEL);
+				if (!adev->gfx.me.mqd_backup[i]) {
 					dev_warn(adev->dev, "no memory to create MQD backup for ring %s\n", ring->name);
+					return -ENOMEM;
+				}
 			}
 		}
 	}
 
 	/* create MQD for each KCQ */
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-		ring = &adev->gfx.compute_ring[i];
+		j = i + xcc_id * adev->gfx.num_compute_rings;
+		ring = &adev->gfx.compute_ring[j];
 		if (!ring->mqd_obj) {
 			r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE,
-						    AMDGPU_GEM_DOMAIN_GTT, &ring->mqd_obj,
+						    domain, &ring->mqd_obj,
 						    &ring->mqd_gpu_addr, &ring->mqd_ptr);
 			if (r) {
 				dev_warn(adev->dev, "failed to create ring mqd bo (%d)", r);
 				return r;
 			}
 
+			ring->mqd_size = mqd_size;
 			/* prepare MQD backup */
-			adev->gfx.mec.mqd_backup[i] = kmalloc(mqd_size, GFP_KERNEL);
-			if (!adev->gfx.mec.mqd_backup[i])
+			adev->gfx.mec.mqd_backup[j] = kzalloc(mqd_size, GFP_KERNEL);
+			if (!adev->gfx.mec.mqd_backup[j]) {
 				dev_warn(adev->dev, "no memory to create MQD backup for ring %s\n", ring->name);
+				return -ENOMEM;
+			}
 		}
 	}
 
 	return 0;
 }
 
-void amdgpu_gfx_mqd_sw_fini(struct amdgpu_device *adev)
+void amdgpu_gfx_mqd_sw_fini(struct amdgpu_device *adev, int xcc_id)
 {
 	struct amdgpu_ring *ring = NULL;
-	int i;
+	int i, j;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
 
 	if (adev->asic_type >= CHIP_NAVI10 && amdgpu_async_gfx_ring) {
 		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
@@ -450,43 +484,121 @@ void amdgpu_gfx_mqd_sw_fini(struct amdgpu_device *adev)
 	}
 
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-		ring = &adev->gfx.compute_ring[i];
-		kfree(adev->gfx.mec.mqd_backup[i]);
+		j = i + xcc_id * adev->gfx.num_compute_rings;
+		ring = &adev->gfx.compute_ring[j];
+		kfree(adev->gfx.mec.mqd_backup[j]);
 		amdgpu_bo_free_kernel(&ring->mqd_obj,
 				      &ring->mqd_gpu_addr,
 				      &ring->mqd_ptr);
 	}
 
-	ring = &adev->gfx.kiq.ring;
-	kfree(adev->gfx.mec.mqd_backup[AMDGPU_MAX_COMPUTE_RINGS]);
+	ring = &kiq->ring;
+	kfree(kiq->mqd_backup);
 	amdgpu_bo_free_kernel(&ring->mqd_obj,
 			      &ring->mqd_gpu_addr,
 			      &ring->mqd_ptr);
 }
 
-int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev)
+int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev, int xcc_id)
 {
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
 	struct amdgpu_ring *kiq_ring = &kiq->ring;
 	int i, r = 0;
+	int j;
+
+	if (adev->enable_mes) {
+		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+			j = i + xcc_id * adev->gfx.num_compute_rings;
+			amdgpu_mes_unmap_legacy_queue(adev,
+						   &adev->gfx.compute_ring[j],
+						   RESET_QUEUES, 0, 0);
+		}
+		return 0;
+	}
 
 	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
 		return -EINVAL;
 
-	spin_lock(&adev->gfx.kiq.ring_lock);
+	if (!kiq_ring->sched.ready || amdgpu_in_reset(adev))
+		return 0;
+
+	spin_lock(&kiq->ring_lock);
 	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size *
 					adev->gfx.num_compute_rings)) {
-		spin_unlock(&adev->gfx.kiq.ring_lock);
+		spin_unlock(&kiq->ring_lock);
 		return -ENOMEM;
 	}
 
-	for (i = 0; i < adev->gfx.num_compute_rings; i++)
-		kiq->pmf->kiq_unmap_queues(kiq_ring, &adev->gfx.compute_ring[i],
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		j = i + xcc_id * adev->gfx.num_compute_rings;
+		kiq->pmf->kiq_unmap_queues(kiq_ring,
+					   &adev->gfx.compute_ring[j],
 					   RESET_QUEUES, 0, 0);
+	}
+	/* Submit unmap queue packet */
+	amdgpu_ring_commit(kiq_ring);
+	/*
+	 * Ring test will do a basic scratch register change check. Just run
+	 * this to ensure that unmap queues that is submitted before got
+	 * processed successfully before returning.
+	 */
+	r = amdgpu_ring_test_helper(kiq_ring);
+
+	spin_unlock(&kiq->ring_lock);
+
+	return r;
+}
+
+int amdgpu_gfx_disable_kgq(struct amdgpu_device *adev, int xcc_id)
+{
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	int i, r = 0;
+	int j;
+
+	if (adev->enable_mes) {
+		if (amdgpu_gfx_is_master_xcc(adev, xcc_id)) {
+			for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+				j = i + xcc_id * adev->gfx.num_gfx_rings;
+				amdgpu_mes_unmap_legacy_queue(adev,
+						      &adev->gfx.gfx_ring[j],
+						      PREEMPT_QUEUES, 0, 0);
+			}
+		}
+		return 0;
+	}
+
+	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
+		return -EINVAL;
+
+	if (!adev->gfx.kiq[0].ring.sched.ready || amdgpu_in_reset(adev))
+		return 0;
+
+	if (amdgpu_gfx_is_master_xcc(adev, xcc_id)) {
+		spin_lock(&kiq->ring_lock);
+		if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size *
+						adev->gfx.num_gfx_rings)) {
+			spin_unlock(&kiq->ring_lock);
+			return -ENOMEM;
+		}
+
+		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+			j = i + xcc_id * adev->gfx.num_gfx_rings;
+			kiq->pmf->kiq_unmap_queues(kiq_ring,
+						   &adev->gfx.gfx_ring[j],
+						   PREEMPT_QUEUES, 0, 0);
+		}
+		/* Submit unmap queue packet */
+		amdgpu_ring_commit(kiq_ring);
 
-	if (adev->gfx.kiq.ring.sched.ready && !adev->job_hang)
+		/*
+		 * Ring test will do a basic scratch register change check.
+		 * Just run this to ensure that unmap queues that is submitted
+		 * before got processed successfully before returning.
+		 */
 		r = amdgpu_ring_test_helper(kiq_ring);
-	spin_unlock(&adev->gfx.kiq.ring_lock);
+		spin_unlock(&kiq->ring_lock);
+	}
 
 	return r;
 }
@@ -504,70 +616,167 @@ int amdgpu_queue_mask_bit_to_set_resource_bit(struct amdgpu_device *adev,
 	return set_resource_bit;
 }
 
-int amdgpu_gfx_enable_kcq(struct amdgpu_device *adev)
+static int amdgpu_gfx_mes_enable_kcq(struct amdgpu_device *adev, int xcc_id)
 {
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	uint64_t queue_mask = ~0ULL;
+	int r, i, j;
+
+	amdgpu_device_flush_hdp(adev, NULL);
+
+	if (!adev->enable_uni_mes) {
+		spin_lock(&kiq->ring_lock);
+		r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->set_resources_size);
+		if (r) {
+			dev_err(adev->dev, "Failed to lock KIQ (%d).\n", r);
+			spin_unlock(&kiq->ring_lock);
+			return r;
+		}
+
+		kiq->pmf->kiq_set_resources(kiq_ring, queue_mask);
+		r = amdgpu_ring_test_helper(kiq_ring);
+		spin_unlock(&kiq->ring_lock);
+		if (r)
+			dev_err(adev->dev, "KIQ failed to set resources\n");
+	}
+
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		j = i + xcc_id * adev->gfx.num_compute_rings;
+		r = amdgpu_mes_map_legacy_queue(adev,
+						&adev->gfx.compute_ring[j]);
+		if (r) {
+			dev_err(adev->dev, "failed to map compute queue\n");
+			return r;
+		}
+	}
+
+	return 0;
+}
+
+int amdgpu_gfx_enable_kcq(struct amdgpu_device *adev, int xcc_id)
+{
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
 	uint64_t queue_mask = 0;
-	int r, i;
+	int r, i, j;
+
+	if (adev->mes.enable_legacy_queue_map)
+		return amdgpu_gfx_mes_enable_kcq(adev, xcc_id);
 
 	if (!kiq->pmf || !kiq->pmf->kiq_map_queues || !kiq->pmf->kiq_set_resources)
 		return -EINVAL;
 
 	for (i = 0; i < AMDGPU_MAX_COMPUTE_QUEUES; ++i) {
-		if (!test_bit(i, adev->gfx.mec.queue_bitmap))
+		if (!test_bit(i, adev->gfx.mec_bitmap[xcc_id].queue_bitmap))
 			continue;
 
 		/* This situation may be hit in the future if a new HW
 		 * generation exposes more than 64 queues. If so, the
 		 * definition of queue_mask needs updating */
 		if (WARN_ON(i > (sizeof(queue_mask)*8))) {
-			DRM_ERROR("Invalid KCQ enabled: %d\n", i);
+			dev_err(adev->dev, "Invalid KCQ enabled: %d\n", i);
 			break;
 		}
 
 		queue_mask |= (1ull << amdgpu_queue_mask_bit_to_set_resource_bit(adev, i));
 	}
 
-	DRM_INFO("kiq ring mec %d pipe %d q %d\n", kiq_ring->me, kiq_ring->pipe,
-							kiq_ring->queue);
-	spin_lock(&adev->gfx.kiq.ring_lock);
+	amdgpu_device_flush_hdp(adev, NULL);
+
+	dev_info(adev->dev, "kiq ring mec %d pipe %d q %d\n", kiq_ring->me,
+		 kiq_ring->pipe, kiq_ring->queue);
+
+	spin_lock(&kiq->ring_lock);
 	r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size *
 					adev->gfx.num_compute_rings +
 					kiq->pmf->set_resources_size);
 	if (r) {
-		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
-		spin_unlock(&adev->gfx.kiq.ring_lock);
+		dev_err(adev->dev, "Failed to lock KIQ (%d).\n", r);
+		spin_unlock(&kiq->ring_lock);
 		return r;
 	}
 
-	if (adev->enable_mes)
-		queue_mask = ~0ULL;
-
 	kiq->pmf->kiq_set_resources(kiq_ring, queue_mask);
-	for (i = 0; i < adev->gfx.num_compute_rings; i++)
-		kiq->pmf->kiq_map_queues(kiq_ring, &adev->gfx.compute_ring[i]);
-
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		j = i + xcc_id * adev->gfx.num_compute_rings;
+		kiq->pmf->kiq_map_queues(kiq_ring,
+					 &adev->gfx.compute_ring[j]);
+	}
+	/* Submit map queue packet */
+	amdgpu_ring_commit(kiq_ring);
+	/*
+	 * Ring test will do a basic scratch register change check. Just run
+	 * this to ensure that map queues that is submitted before got
+	 * processed successfully before returning.
+	 */
 	r = amdgpu_ring_test_helper(kiq_ring);
-	spin_unlock(&adev->gfx.kiq.ring_lock);
+	spin_unlock(&kiq->ring_lock);
 	if (r)
-		DRM_ERROR("KCQ enable failed\n");
+		dev_err(adev->dev, "KCQ enable failed\n");
 
 	return r;
 }
 
-/* amdgpu_gfx_off_ctrl - Handle gfx off feature enable/disable
- *
- * @adev: amdgpu_device pointer
- * @bool enable true: enable gfx off feature, false: disable gfx off feature
- *
- * 1. gfx off feature will be enabled by gfx ip after gfx cg gp enabled.
- * 2. other client can send request to disable gfx off feature, the request should be honored.
- * 3. other client can cancel their request of disable gfx off feature
- * 4. other client should not send request to enable gfx off feature before disable gfx off feature.
- */
+int amdgpu_gfx_enable_kgq(struct amdgpu_device *adev, int xcc_id)
+{
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	int r, i, j;
 
-void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable)
+	if (!kiq->pmf || !kiq->pmf->kiq_map_queues)
+		return -EINVAL;
+
+	amdgpu_device_flush_hdp(adev, NULL);
+
+	if (adev->mes.enable_legacy_queue_map) {
+		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+			j = i + xcc_id * adev->gfx.num_gfx_rings;
+			r = amdgpu_mes_map_legacy_queue(adev,
+							&adev->gfx.gfx_ring[j]);
+			if (r) {
+				dev_err(adev->dev, "failed to map gfx queue\n");
+				return r;
+			}
+		}
+
+		return 0;
+	}
+
+	spin_lock(&kiq->ring_lock);
+	/* No need to map kcq on the slave */
+	if (amdgpu_gfx_is_master_xcc(adev, xcc_id)) {
+		r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size *
+						adev->gfx.num_gfx_rings);
+		if (r) {
+			dev_err(adev->dev, "Failed to lock KIQ (%d).\n", r);
+			spin_unlock(&kiq->ring_lock);
+			return r;
+		}
+
+		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+			j = i + xcc_id * adev->gfx.num_gfx_rings;
+			kiq->pmf->kiq_map_queues(kiq_ring,
+						 &adev->gfx.gfx_ring[j]);
+		}
+	}
+	/* Submit map queue packet */
+	amdgpu_ring_commit(kiq_ring);
+	/*
+	 * Ring test will do a basic scratch register change check. Just run
+	 * this to ensure that map queues that is submitted before got
+	 * processed successfully before returning.
+	 */
+	r = amdgpu_ring_test_helper(kiq_ring);
+	spin_unlock(&kiq->ring_lock);
+	if (r)
+		dev_err(adev->dev, "KGQ enable failed\n");
+
+	return r;
+}
+
+static void amdgpu_gfx_do_off_ctrl(struct amdgpu_device *adev, bool enable,
+				   bool no_delay)
 {
 	unsigned long delay = GFX_OFF_DELAY_ENABLE;
 
@@ -589,9 +798,9 @@ void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable)
 		if (adev->gfx.gfx_off_req_count == 0 &&
 		    !adev->gfx.gfx_off_state) {
 			/* If going to s2idle, no need to wait */
-			if (adev->in_s0ix) {
+			if (no_delay) {
 				if (!amdgpu_dpm_set_powergating_by_smu(adev,
-						AMD_IP_BLOCK_TYPE_GFX, true))
+						AMD_IP_BLOCK_TYPE_GFX, true, 0))
 					adev->gfx.gfx_off_state = true;
 			} else {
 				schedule_delayed_work(&adev->gfx.gfx_off_delay_work,
@@ -603,7 +812,7 @@ void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable)
 			cancel_delayed_work_sync(&adev->gfx.gfx_off_delay_work);
 
 			if (adev->gfx.gfx_off_state &&
-			    !amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, false)) {
+			    !amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, false, 0)) {
 				adev->gfx.gfx_off_state = false;
 
 				if (adev->gfx.funcs->init_spm_golden) {
@@ -621,6 +830,43 @@ unlock:
 	mutex_unlock(&adev->gfx.gfx_off_mutex);
 }
 
+/* amdgpu_gfx_off_ctrl - Handle gfx off feature enable/disable
+ *
+ * @adev: amdgpu_device pointer
+ * @bool enable true: enable gfx off feature, false: disable gfx off feature
+ *
+ * 1. gfx off feature will be enabled by gfx ip after gfx cg pg enabled.
+ * 2. other client can send request to disable gfx off feature, the request should be honored.
+ * 3. other client can cancel their request of disable gfx off feature
+ * 4. other client should not send request to enable gfx off feature before disable gfx off feature.
+ *
+ * gfx off allow will be delayed by GFX_OFF_DELAY_ENABLE ms.
+ */
+void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable)
+{
+	/* If going to s2idle, no need to wait */
+	bool no_delay = adev->in_s0ix ? true : false;
+
+	amdgpu_gfx_do_off_ctrl(adev, enable, no_delay);
+}
+
+/* amdgpu_gfx_off_ctrl_immediate - Handle gfx off feature enable/disable
+ *
+ * @adev: amdgpu_device pointer
+ * @bool enable true: enable gfx off feature, false: disable gfx off feature
+ *
+ * 1. gfx off feature will be enabled by gfx ip after gfx cg pg enabled.
+ * 2. other client can send request to disable gfx off feature, the request should be honored.
+ * 3. other client can cancel their request of disable gfx off feature
+ * 4. other client should not send request to enable gfx off feature before disable gfx off feature.
+ *
+ * gfx off allow will be issued immediately.
+ */
+void amdgpu_gfx_off_ctrl_immediate(struct amdgpu_device *adev, bool enable)
+{
+	amdgpu_gfx_do_off_ctrl(adev, enable, true);
+}
+
 int amdgpu_set_gfx_off_residency(struct amdgpu_device *adev, bool value)
 {
 	int r = 0;
@@ -679,16 +925,24 @@ int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *r
 	int r;
 
 	if (amdgpu_ras_is_supported(adev, ras_block->block)) {
-		if (!amdgpu_persistent_edc_harvesting_supported(adev))
-			amdgpu_ras_reset_error_status(adev, AMDGPU_RAS_BLOCK__GFX);
+		if (!amdgpu_persistent_edc_harvesting_supported(adev)) {
+			r = amdgpu_ras_reset_error_status(adev, AMDGPU_RAS_BLOCK__GFX);
+			if (r)
+				return r;
+		}
 
 		r = amdgpu_ras_block_late_init(adev, ras_block);
 		if (r)
 			return r;
 
-		r = amdgpu_irq_get(adev, &adev->gfx.cp_ecc_error_irq, 0);
-		if (r)
-			goto late_fini;
+		if (amdgpu_sriov_vf(adev))
+			return r;
+
+		if (adev->gfx.cp_ecc_error_irq.funcs) {
+			r = amdgpu_irq_get(adev, &adev->gfx.cp_ecc_error_irq, 0);
+			if (r)
+				goto late_fini;
+		}
 	} else {
 		amdgpu_ras_feature_enable_on_boot(adev, ras_block, 0);
 	}
@@ -777,23 +1031,42 @@ int amdgpu_gfx_cp_ecc_error_irq(struct amdgpu_device *adev,
 
 	ih_data.head = *ras_if;
 
-	DRM_ERROR("CP ECC ERROR IRQ\n");
+	dev_err(adev->dev, "CP ECC ERROR IRQ\n");
 	amdgpu_ras_interrupt_dispatch(adev, &ih_data);
 	return 0;
 }
 
-uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
+void amdgpu_gfx_ras_error_func(struct amdgpu_device *adev,
+		void *ras_error_status,
+		void (*func)(struct amdgpu_device *adev, void *ras_error_status,
+				int xcc_id))
+{
+	int i;
+	int num_xcc = adev->gfx.xcc_mask ? NUM_XCC(adev->gfx.xcc_mask) : 1;
+	uint32_t xcc_mask = GENMASK(num_xcc - 1, 0);
+	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
+
+	if (err_data) {
+		err_data->ue_count = 0;
+		err_data->ce_count = 0;
+	}
+
+	for_each_inst(i, xcc_mask)
+		func(adev, ras_error_status, i);
+}
+
+uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg, uint32_t xcc_id)
 {
 	signed long r, cnt = 0;
 	unsigned long flags;
 	uint32_t seq, reg_val_offs = 0, value = 0;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
 	struct amdgpu_ring *ring = &kiq->ring;
 
 	if (amdgpu_device_skip_hw_access(adev))
 		return 0;
 
-	if (adev->mes.ring.sched.ready)
+	if (adev->mes.ring[0].sched.ready)
 		return amdgpu_mes_rreg(adev, reg);
 
 	BUG_ON(!ring->funcs->emit_rreg);
@@ -803,7 +1076,10 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
 		pr_err("critical bug! too many kiq readers\n");
 		goto failed_unlock;
 	}
-	amdgpu_ring_alloc(ring, 32);
+	r = amdgpu_ring_alloc(ring, 32);
+	if (r)
+		goto failed_unlock;
+
 	amdgpu_ring_emit_rreg(ring, reg, reg_val_offs);
 	r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
 	if (r)
@@ -827,6 +1103,9 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg)
 
 	might_sleep();
 	while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
+		if (amdgpu_in_reset(adev))
+			goto failed_kiq_read;
+
 		msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
 		r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
 	}
@@ -850,12 +1129,12 @@ failed_kiq_read:
 	return ~0;
 }
 
-void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v)
+void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v, uint32_t xcc_id)
 {
 	signed long r, cnt = 0;
 	unsigned long flags;
 	uint32_t seq;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
 	struct amdgpu_ring *ring = &kiq->ring;
 
 	BUG_ON(!ring->funcs->emit_wreg);
@@ -863,13 +1142,16 @@ void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v)
 	if (amdgpu_device_skip_hw_access(adev))
 		return;
 
-	if (adev->mes.ring.sched.ready) {
+	if (adev->mes.ring[0].sched.ready) {
 		amdgpu_mes_wreg(adev, reg, v);
 		return;
 	}
 
 	spin_lock_irqsave(&kiq->ring_lock, flags);
-	amdgpu_ring_alloc(ring, 32);
+	r = amdgpu_ring_alloc(ring, 32);
+	if (r)
+		goto failed_unlock;
+
 	amdgpu_ring_emit_wreg(ring, reg, v);
 	r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
 	if (r)
@@ -893,6 +1175,8 @@ void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v)
 
 	might_sleep();
 	while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
+		if (amdgpu_in_reset(adev))
+			goto failed_kiq_write;
 
 		msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
 		r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
@@ -905,11 +1189,81 @@ void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v)
 
 failed_undo:
 	amdgpu_ring_undo(ring);
+failed_unlock:
 	spin_unlock_irqrestore(&kiq->ring_lock, flags);
 failed_kiq_write:
 	dev_err(adev->dev, "failed to write reg:%x\n", reg);
 }
 
+int amdgpu_kiq_hdp_flush(struct amdgpu_device *adev)
+{
+	signed long r, cnt = 0;
+	unsigned long flags;
+	uint32_t seq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
+	struct amdgpu_ring *ring = &kiq->ring;
+
+	if (amdgpu_device_skip_hw_access(adev))
+		return 0;
+
+	if (adev->enable_mes_kiq && adev->mes.ring[0].sched.ready)
+		return amdgpu_mes_hdp_flush(adev);
+
+	if (!ring->funcs->emit_hdp_flush) {
+		return -EOPNOTSUPP;
+	}
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+	r = amdgpu_ring_alloc(ring, 32);
+	if (r)
+		goto failed_unlock;
+
+	amdgpu_ring_emit_hdp_flush(ring);
+	r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
+	if (r)
+		goto failed_undo;
+
+	amdgpu_ring_commit(ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+
+	r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
+
+	/* don't wait anymore for gpu reset case because this way may
+	 * block gpu_recover() routine forever, e.g. this virt_kiq_rreg
+	 * is triggered in TTM and ttm_bo_lock_delayed_workqueue() will
+	 * never return if we keep waiting in virt_kiq_rreg, which cause
+	 * gpu_recover() hang there.
+	 *
+	 * also don't wait anymore for IRQ context
+	 * */
+	if (r < 1 && (amdgpu_in_reset(adev) || in_interrupt()))
+		goto failed_kiq_hdp_flush;
+
+	might_sleep();
+	while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
+		if (amdgpu_in_reset(adev))
+			goto failed_kiq_hdp_flush;
+
+		msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
+		r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
+	}
+
+	if (cnt > MAX_KIQ_REG_TRY) {
+		dev_err(adev->dev, "failed to flush HDP via KIQ timeout\n");
+		return -ETIMEDOUT;
+	}
+
+	return 0;
+
+failed_undo:
+	amdgpu_ring_undo(ring);
+failed_unlock:
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+failed_kiq_hdp_flush:
+	dev_err(adev->dev, "failed to flush HDP via KIQ\n");
+	return r < 0 ? r : -EIO;
+}
+
 int amdgpu_gfx_get_num_kcq(struct amdgpu_device *adev)
 {
 	if (amdgpu_num_kcq == -1) {
@@ -1049,7 +1403,8 @@ void amdgpu_gfx_cp_init_microcode(struct amdgpu_device *adev,
 		fw_size = le32_to_cpu(cp_hdr_v2_0->data_size_bytes);
 		break;
 	default:
-		break;
+		dev_err(adev->dev, "Invalid ucode id %u\n", ucode_id);
+		return;
 	}
 
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
@@ -1059,3 +1414,1144 @@ void amdgpu_gfx_cp_init_microcode(struct amdgpu_device *adev,
 		adev->firmware.fw_size += ALIGN(fw_size, PAGE_SIZE);
 	}
 }
+
+bool amdgpu_gfx_is_master_xcc(struct amdgpu_device *adev, int xcc_id)
+{
+	return !(xcc_id % (adev->gfx.num_xcc_per_xcp ?
+			adev->gfx.num_xcc_per_xcp : 1));
+}
+
+static ssize_t amdgpu_gfx_get_current_compute_partition(struct device *dev,
+						struct device_attribute *addr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	int mode;
+
+	/* Only minimal precaution taken to reject requests while in reset.*/
+	if (amdgpu_in_reset(adev))
+		return -EPERM;
+
+	mode = amdgpu_xcp_query_partition_mode(adev->xcp_mgr,
+					       AMDGPU_XCP_FL_NONE);
+
+	return sysfs_emit(buf, "%s\n", amdgpu_gfx_compute_mode_desc(mode));
+}
+
+static ssize_t amdgpu_gfx_set_compute_partition(struct device *dev,
+						struct device_attribute *addr,
+						const char *buf, size_t count)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	enum amdgpu_gfx_partition mode;
+	int ret = 0, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	if (num_xcc % 2 != 0)
+		return -EINVAL;
+
+	if (!strncasecmp("SPX", buf, strlen("SPX"))) {
+		mode = AMDGPU_SPX_PARTITION_MODE;
+	} else if (!strncasecmp("DPX", buf, strlen("DPX"))) {
+		/*
+		 * DPX mode needs AIDs to be in multiple of 2.
+		 * Each AID connects 2 XCCs.
+		 */
+		if (num_xcc%4)
+			return -EINVAL;
+		mode = AMDGPU_DPX_PARTITION_MODE;
+	} else if (!strncasecmp("TPX", buf, strlen("TPX"))) {
+		if (num_xcc != 6)
+			return -EINVAL;
+		mode = AMDGPU_TPX_PARTITION_MODE;
+	} else if (!strncasecmp("QPX", buf, strlen("QPX"))) {
+		if (num_xcc != 8)
+			return -EINVAL;
+		mode = AMDGPU_QPX_PARTITION_MODE;
+	} else if (!strncasecmp("CPX", buf, strlen("CPX"))) {
+		mode = AMDGPU_CPX_PARTITION_MODE;
+	} else {
+		return -EINVAL;
+	}
+
+	/* Don't allow a switch while under reset */
+	if (!down_read_trylock(&adev->reset_domain->sem))
+		return -EPERM;
+
+	ret = amdgpu_xcp_switch_partition_mode(adev->xcp_mgr, mode);
+
+	up_read(&adev->reset_domain->sem);
+
+	if (ret)
+		return ret;
+
+	return count;
+}
+
+static const char *xcp_desc[] = {
+	[AMDGPU_SPX_PARTITION_MODE] = "SPX",
+	[AMDGPU_DPX_PARTITION_MODE] = "DPX",
+	[AMDGPU_TPX_PARTITION_MODE] = "TPX",
+	[AMDGPU_QPX_PARTITION_MODE] = "QPX",
+	[AMDGPU_CPX_PARTITION_MODE] = "CPX",
+};
+
+static ssize_t amdgpu_gfx_get_available_compute_partition(struct device *dev,
+						struct device_attribute *addr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	struct amdgpu_xcp_mgr *xcp_mgr = adev->xcp_mgr;
+	int size = 0, mode;
+	char *sep = "";
+
+	if (!xcp_mgr || !xcp_mgr->avail_xcp_modes)
+		return sysfs_emit(buf, "Not supported\n");
+
+	for_each_inst(mode, xcp_mgr->avail_xcp_modes) {
+		size += sysfs_emit_at(buf, size, "%s%s", sep, xcp_desc[mode]);
+		sep = ", ";
+	}
+
+	size += sysfs_emit_at(buf, size, "\n");
+
+	return size;
+}
+
+static int amdgpu_gfx_run_cleaner_shader_job(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct drm_gpu_scheduler *sched = &ring->sched;
+	struct drm_sched_entity entity;
+	static atomic_t counter;
+	struct dma_fence *f;
+	struct amdgpu_job *job;
+	struct amdgpu_ib *ib;
+	void *owner;
+	int i, r;
+
+	/* Initialize the scheduler entity */
+	r = drm_sched_entity_init(&entity, DRM_SCHED_PRIORITY_NORMAL,
+				  &sched, 1, NULL);
+	if (r) {
+		dev_err(adev->dev, "Failed setting up GFX kernel entity.\n");
+		goto err;
+	}
+
+	/*
+	 * Use some unique dummy value as the owner to make sure we execute
+	 * the cleaner shader on each submission. The value just need to change
+	 * for each submission and is otherwise meaningless.
+	 */
+	owner = (void *)(unsigned long)atomic_inc_return(&counter);
+
+	r = amdgpu_job_alloc_with_ib(ring->adev, &entity, owner,
+				     64, 0, &job,
+				     AMDGPU_KERNEL_JOB_ID_CLEANER_SHADER);
+	if (r)
+		goto err;
+
+	job->enforce_isolation = true;
+	/* always run the cleaner shader */
+	job->run_cleaner_shader = true;
+
+	ib = &job->ibs[0];
+	for (i = 0; i <= ring->funcs->align_mask; ++i)
+		ib->ptr[i] = ring->funcs->nop;
+	ib->length_dw = ring->funcs->align_mask + 1;
+
+	f = amdgpu_job_submit(job);
+
+	r = dma_fence_wait(f, false);
+	if (r)
+		goto err;
+
+	dma_fence_put(f);
+
+	/* Clean up the scheduler entity */
+	drm_sched_entity_destroy(&entity);
+	return 0;
+
+err:
+	return r;
+}
+
+static int amdgpu_gfx_run_cleaner_shader(struct amdgpu_device *adev, int xcp_id)
+{
+	int num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	struct amdgpu_ring *ring;
+	int num_xcc_to_clear;
+	int i, r, xcc_id;
+
+	if (adev->gfx.num_xcc_per_xcp)
+		num_xcc_to_clear = adev->gfx.num_xcc_per_xcp;
+	else
+		num_xcc_to_clear = 1;
+
+	for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) {
+		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+			ring = &adev->gfx.compute_ring[i + xcc_id * adev->gfx.num_compute_rings];
+			if ((ring->xcp_id == xcp_id) && ring->sched.ready) {
+				r = amdgpu_gfx_run_cleaner_shader_job(ring);
+				if (r)
+					return r;
+				num_xcc_to_clear--;
+				break;
+			}
+		}
+	}
+
+	if (num_xcc_to_clear)
+		return -ENOENT;
+
+	return 0;
+}
+
+/**
+ * amdgpu_gfx_set_run_cleaner_shader - Execute the AMDGPU GFX Cleaner Shader
+ * @dev: The device structure
+ * @attr: The device attribute structure
+ * @buf: The buffer containing the input data
+ * @count: The size of the input data
+ *
+ * Provides the sysfs interface to manually run a cleaner shader, which is
+ * used to clear the GPU state between different tasks. Writing a value to the
+ * 'run_cleaner_shader' sysfs file triggers the cleaner shader execution.
+ * The value written corresponds to the partition index on multi-partition
+ * devices. On single-partition devices, the value should be '0'.
+ *
+ * The cleaner shader clears the Local Data Store (LDS) and General Purpose
+ * Registers (GPRs) to ensure data isolation between GPU workloads.
+ *
+ * Return: The number of bytes written to the sysfs file.
+ */
+static ssize_t amdgpu_gfx_set_run_cleaner_shader(struct device *dev,
+						 struct device_attribute *attr,
+						 const char *buf,
+						 size_t count)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	int ret;
+	long value;
+
+	if (amdgpu_in_reset(adev))
+		return -EPERM;
+	if (adev->in_suspend && !adev->in_runpm)
+		return -EPERM;
+
+	if (adev->gfx.disable_kq)
+		return -EPERM;
+
+	ret = kstrtol(buf, 0, &value);
+
+	if (ret)
+		return -EINVAL;
+
+	if (value < 0)
+		return -EINVAL;
+
+	if (adev->xcp_mgr) {
+		if (value >= adev->xcp_mgr->num_xcps)
+			return -EINVAL;
+	} else {
+		if (value > 1)
+			return -EINVAL;
+	}
+
+	ret = pm_runtime_get_sync(ddev->dev);
+	if (ret < 0) {
+		pm_runtime_put_autosuspend(ddev->dev);
+		return ret;
+	}
+
+	ret = amdgpu_gfx_run_cleaner_shader(adev, value);
+
+	pm_runtime_put_autosuspend(ddev->dev);
+
+	if (ret)
+		return ret;
+
+	return count;
+}
+
+/**
+ * amdgpu_gfx_get_enforce_isolation - Query AMDGPU GFX Enforce Isolation Settings
+ * @dev: The device structure
+ * @attr: The device attribute structure
+ * @buf: The buffer to store the output data
+ *
+ * Provides the sysfs read interface to get the current settings of the 'enforce_isolation'
+ * feature for each GPU partition. Reading from the 'enforce_isolation'
+ * sysfs file returns the isolation settings for all partitions, where '0'
+ * indicates disabled, '1' indicates enabled, and '2' indicates enabled in legacy mode,
+ * and '3' indicates enabled without cleaner shader.
+ *
+ * Return: The number of bytes read from the sysfs file.
+ */
+static ssize_t amdgpu_gfx_get_enforce_isolation(struct device *dev,
+						struct device_attribute *attr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	int i;
+	ssize_t size = 0;
+
+	if (adev->xcp_mgr) {
+		for (i = 0; i < adev->xcp_mgr->num_xcps; i++) {
+			size += sysfs_emit_at(buf, size, "%u", adev->enforce_isolation[i]);
+			if (i < (adev->xcp_mgr->num_xcps - 1))
+				size += sysfs_emit_at(buf, size, " ");
+		}
+		buf[size++] = '\n';
+	} else {
+		size = sysfs_emit_at(buf, 0, "%u\n", adev->enforce_isolation[0]);
+	}
+
+	return size;
+}
+
+/**
+ * amdgpu_gfx_set_enforce_isolation - Control AMDGPU GFX Enforce Isolation
+ * @dev: The device structure
+ * @attr: The device attribute structure
+ * @buf: The buffer containing the input data
+ * @count: The size of the input data
+ *
+ * This function allows control over the 'enforce_isolation' feature, which
+ * serializes access to the graphics engine. Writing '0' to disable, '1' to
+ * enable isolation with cleaner shader, '2' to enable legacy isolation without
+ * cleaner shader, or '3' to enable process isolation without submitting the
+ * cleaner shader to the 'enforce_isolation' sysfs file sets the isolation mode
+ * for each partition. The input should specify the setting for all
+ * partitions.
+ *
+ * Return: The number of bytes written to the sysfs file.
+ */
+static ssize_t amdgpu_gfx_set_enforce_isolation(struct device *dev,
+						struct device_attribute *attr,
+						const char *buf, size_t count)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	long partition_values[MAX_XCP] = {0};
+	int ret, i, num_partitions;
+	const char *input_buf = buf;
+
+	for (i = 0; i < (adev->xcp_mgr ? adev->xcp_mgr->num_xcps : 1); i++) {
+		ret = sscanf(input_buf, "%ld", &partition_values[i]);
+		if (ret <= 0)
+			break;
+
+		/* Move the pointer to the next value in the string */
+		input_buf = strchr(input_buf, ' ');
+		if (input_buf) {
+			input_buf++;
+		} else {
+			i++;
+			break;
+		}
+	}
+	num_partitions = i;
+
+	if (adev->xcp_mgr && num_partitions != adev->xcp_mgr->num_xcps)
+		return -EINVAL;
+
+	if (!adev->xcp_mgr && num_partitions != 1)
+		return -EINVAL;
+
+	for (i = 0; i < num_partitions; i++) {
+		if (partition_values[i] != 0 &&
+		    partition_values[i] != 1 &&
+		    partition_values[i] != 2 &&
+		    partition_values[i] != 3)
+			return -EINVAL;
+	}
+
+	mutex_lock(&adev->enforce_isolation_mutex);
+	for (i = 0; i < num_partitions; i++) {
+		switch (partition_values[i]) {
+		case 0:
+		default:
+			adev->enforce_isolation[i] = AMDGPU_ENFORCE_ISOLATION_DISABLE;
+			break;
+		case 1:
+			adev->enforce_isolation[i] =
+				AMDGPU_ENFORCE_ISOLATION_ENABLE;
+			break;
+		case 2:
+			adev->enforce_isolation[i] =
+				AMDGPU_ENFORCE_ISOLATION_ENABLE_LEGACY;
+			break;
+		case 3:
+			adev->enforce_isolation[i] =
+				AMDGPU_ENFORCE_ISOLATION_NO_CLEANER_SHADER;
+			break;
+		}
+	}
+	mutex_unlock(&adev->enforce_isolation_mutex);
+
+	amdgpu_mes_update_enforce_isolation(adev);
+
+	return count;
+}
+
+static ssize_t amdgpu_gfx_get_gfx_reset_mask(struct device *dev,
+						struct device_attribute *attr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (!adev)
+		return -ENODEV;
+
+	return amdgpu_show_reset_mask(buf, adev->gfx.gfx_supported_reset);
+}
+
+static ssize_t amdgpu_gfx_get_compute_reset_mask(struct device *dev,
+						struct device_attribute *attr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (!adev)
+		return -ENODEV;
+
+	return amdgpu_show_reset_mask(buf, adev->gfx.compute_supported_reset);
+}
+
+static DEVICE_ATTR(run_cleaner_shader, 0200,
+		   NULL, amdgpu_gfx_set_run_cleaner_shader);
+
+static DEVICE_ATTR(enforce_isolation, 0644,
+		   amdgpu_gfx_get_enforce_isolation,
+		   amdgpu_gfx_set_enforce_isolation);
+
+static DEVICE_ATTR(current_compute_partition, 0644,
+		   amdgpu_gfx_get_current_compute_partition,
+		   amdgpu_gfx_set_compute_partition);
+
+static DEVICE_ATTR(available_compute_partition, 0444,
+		   amdgpu_gfx_get_available_compute_partition, NULL);
+static DEVICE_ATTR(gfx_reset_mask, 0444,
+		   amdgpu_gfx_get_gfx_reset_mask, NULL);
+
+static DEVICE_ATTR(compute_reset_mask, 0444,
+		   amdgpu_gfx_get_compute_reset_mask, NULL);
+
+static int amdgpu_gfx_sysfs_xcp_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_xcp_mgr *xcp_mgr = adev->xcp_mgr;
+	bool xcp_switch_supported;
+	int r;
+
+	if (!xcp_mgr)
+		return 0;
+
+	xcp_switch_supported =
+		(xcp_mgr->funcs && xcp_mgr->funcs->switch_partition_mode);
+
+	if (!xcp_switch_supported)
+		dev_attr_current_compute_partition.attr.mode &=
+			~(S_IWUSR | S_IWGRP | S_IWOTH);
+
+	r = device_create_file(adev->dev, &dev_attr_current_compute_partition);
+	if (r)
+		return r;
+
+	if (xcp_switch_supported)
+		r = device_create_file(adev->dev,
+				       &dev_attr_available_compute_partition);
+
+	return r;
+}
+
+static void amdgpu_gfx_sysfs_xcp_fini(struct amdgpu_device *adev)
+{
+	struct amdgpu_xcp_mgr *xcp_mgr = adev->xcp_mgr;
+	bool xcp_switch_supported;
+
+	if (!xcp_mgr)
+		return;
+
+	xcp_switch_supported =
+		(xcp_mgr->funcs && xcp_mgr->funcs->switch_partition_mode);
+	device_remove_file(adev->dev, &dev_attr_current_compute_partition);
+
+	if (xcp_switch_supported)
+		device_remove_file(adev->dev,
+				   &dev_attr_available_compute_partition);
+}
+
+static int amdgpu_gfx_sysfs_isolation_shader_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = device_create_file(adev->dev, &dev_attr_enforce_isolation);
+	if (r)
+		return r;
+	if (adev->gfx.enable_cleaner_shader)
+		r = device_create_file(adev->dev, &dev_attr_run_cleaner_shader);
+
+	return r;
+}
+
+static void amdgpu_gfx_sysfs_isolation_shader_fini(struct amdgpu_device *adev)
+{
+	device_remove_file(adev->dev, &dev_attr_enforce_isolation);
+	if (adev->gfx.enable_cleaner_shader)
+		device_remove_file(adev->dev, &dev_attr_run_cleaner_shader);
+}
+
+static int amdgpu_gfx_sysfs_reset_mask_init(struct amdgpu_device *adev)
+{
+	int r = 0;
+
+	if (!amdgpu_gpu_recovery)
+		return r;
+
+	if (adev->gfx.num_gfx_rings) {
+		r = device_create_file(adev->dev, &dev_attr_gfx_reset_mask);
+		if (r)
+			return r;
+	}
+
+	if (adev->gfx.num_compute_rings) {
+		r = device_create_file(adev->dev, &dev_attr_compute_reset_mask);
+		if (r)
+			return r;
+	}
+
+	return r;
+}
+
+static void amdgpu_gfx_sysfs_reset_mask_fini(struct amdgpu_device *adev)
+{
+	if (!amdgpu_gpu_recovery)
+		return;
+
+	if (adev->gfx.num_gfx_rings)
+		device_remove_file(adev->dev, &dev_attr_gfx_reset_mask);
+
+	if (adev->gfx.num_compute_rings)
+		device_remove_file(adev->dev, &dev_attr_compute_reset_mask);
+}
+
+int amdgpu_gfx_sysfs_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = amdgpu_gfx_sysfs_xcp_init(adev);
+	if (r) {
+		dev_err(adev->dev, "failed to create xcp sysfs files");
+		return r;
+	}
+
+	r = amdgpu_gfx_sysfs_isolation_shader_init(adev);
+	if (r)
+		dev_err(adev->dev, "failed to create isolation sysfs files");
+
+	r = amdgpu_gfx_sysfs_reset_mask_init(adev);
+	if (r)
+		dev_err(adev->dev, "failed to create reset mask sysfs files");
+
+	return r;
+}
+
+void amdgpu_gfx_sysfs_fini(struct amdgpu_device *adev)
+{
+	if (adev->dev->kobj.sd) {
+		amdgpu_gfx_sysfs_xcp_fini(adev);
+		amdgpu_gfx_sysfs_isolation_shader_fini(adev);
+		amdgpu_gfx_sysfs_reset_mask_fini(adev);
+	}
+}
+
+int amdgpu_gfx_cleaner_shader_sw_init(struct amdgpu_device *adev,
+				      unsigned int cleaner_shader_size)
+{
+	if (!adev->gfx.enable_cleaner_shader)
+		return -EOPNOTSUPP;
+
+	return amdgpu_bo_create_kernel(adev, cleaner_shader_size, PAGE_SIZE,
+				       AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT,
+				       &adev->gfx.cleaner_shader_obj,
+				       &adev->gfx.cleaner_shader_gpu_addr,
+				       (void **)&adev->gfx.cleaner_shader_cpu_ptr);
+}
+
+void amdgpu_gfx_cleaner_shader_sw_fini(struct amdgpu_device *adev)
+{
+	if (!adev->gfx.enable_cleaner_shader)
+		return;
+
+	amdgpu_bo_free_kernel(&adev->gfx.cleaner_shader_obj,
+			      &adev->gfx.cleaner_shader_gpu_addr,
+			      (void **)&adev->gfx.cleaner_shader_cpu_ptr);
+}
+
+void amdgpu_gfx_cleaner_shader_init(struct amdgpu_device *adev,
+				    unsigned int cleaner_shader_size,
+				    const void *cleaner_shader_ptr)
+{
+	if (!adev->gfx.enable_cleaner_shader)
+		return;
+
+	if (adev->gfx.cleaner_shader_cpu_ptr && cleaner_shader_ptr)
+		memcpy_toio(adev->gfx.cleaner_shader_cpu_ptr, cleaner_shader_ptr,
+			    cleaner_shader_size);
+}
+
+/**
+ * amdgpu_gfx_kfd_sch_ctrl - Control the KFD scheduler from the KGD (Graphics Driver)
+ * @adev: amdgpu_device pointer
+ * @idx: Index of the scheduler to control
+ * @enable: Whether to enable or disable the KFD scheduler
+ *
+ * This function is used to control the KFD (Kernel Fusion Driver) scheduler
+ * from the KGD. It is part of the cleaner shader feature. This function plays
+ * a key role in enforcing process isolation on the GPU.
+ *
+ * The function uses a reference count mechanism (kfd_sch_req_count) to keep
+ * track of the number of requests to enable the KFD scheduler. When a request
+ * to enable the KFD scheduler is made, the reference count is decremented.
+ * When the reference count reaches zero, a delayed work is scheduled to
+ * enforce isolation after a delay of GFX_SLICE_PERIOD.
+ *
+ * When a request to disable the KFD scheduler is made, the function first
+ * checks if the reference count is zero. If it is, it cancels the delayed work
+ * for enforcing isolation and checks if the KFD scheduler is active. If the
+ * KFD scheduler is active, it sends a request to stop the KFD scheduler and
+ * sets the KFD scheduler state to inactive. Then, it increments the reference
+ * count.
+ *
+ * The function is synchronized using the kfd_sch_mutex to ensure that the KFD
+ * scheduler state and reference count are updated atomically.
+ *
+ * Note: If the reference count is already zero when a request to enable the
+ * KFD scheduler is made, it means there's an imbalance bug somewhere. The
+ * function triggers a warning in this case.
+ */
+static void amdgpu_gfx_kfd_sch_ctrl(struct amdgpu_device *adev, u32 idx,
+				    bool enable)
+{
+	mutex_lock(&adev->gfx.userq_sch_mutex);
+
+	if (enable) {
+		/* If the count is already 0, it means there's an imbalance bug somewhere.
+		 * Note that the bug may be in a different caller than the one which triggers the
+		 * WARN_ON_ONCE.
+		 */
+		if (WARN_ON_ONCE(adev->gfx.userq_sch_req_count[idx] == 0)) {
+			dev_err(adev->dev, "Attempted to enable KFD scheduler when reference count is already zero\n");
+			goto unlock;
+		}
+
+		adev->gfx.userq_sch_req_count[idx]--;
+
+		if (adev->gfx.userq_sch_req_count[idx] == 0 &&
+		    adev->gfx.userq_sch_inactive[idx]) {
+			schedule_delayed_work(&adev->gfx.enforce_isolation[idx].work,
+					      msecs_to_jiffies(adev->gfx.enforce_isolation_time[idx]));
+		}
+	} else {
+		if (adev->gfx.userq_sch_req_count[idx] == 0) {
+			cancel_delayed_work_sync(&adev->gfx.enforce_isolation[idx].work);
+			if (!adev->gfx.userq_sch_inactive[idx]) {
+				amdgpu_userq_stop_sched_for_enforce_isolation(adev, idx);
+				if (adev->kfd.init_complete)
+					amdgpu_amdkfd_stop_sched(adev, idx);
+				adev->gfx.userq_sch_inactive[idx] = true;
+			}
+		}
+
+		adev->gfx.userq_sch_req_count[idx]++;
+	}
+
+unlock:
+	mutex_unlock(&adev->gfx.userq_sch_mutex);
+}
+
+/**
+ * amdgpu_gfx_enforce_isolation_handler - work handler for enforcing shader isolation
+ *
+ * @work: work_struct.
+ *
+ * This function is the work handler for enforcing shader isolation on AMD GPUs.
+ * It counts the number of emitted fences for each GFX and compute ring. If there
+ * are any fences, it schedules the `enforce_isolation_work` to be run after a
+ * delay of `GFX_SLICE_PERIOD`. If there are no fences, it signals the Kernel Fusion
+ * Driver (KFD) to resume the runqueue. The function is synchronized using the
+ * `enforce_isolation_mutex`.
+ */
+void amdgpu_gfx_enforce_isolation_handler(struct work_struct *work)
+{
+	struct amdgpu_isolation_work *isolation_work =
+		container_of(work, struct amdgpu_isolation_work, work.work);
+	struct amdgpu_device *adev = isolation_work->adev;
+	u32 i, idx, fences = 0;
+
+	if (isolation_work->xcp_id == AMDGPU_XCP_NO_PARTITION)
+		idx = 0;
+	else
+		idx = isolation_work->xcp_id;
+
+	if (idx >= MAX_XCP)
+		return;
+
+	mutex_lock(&adev->enforce_isolation_mutex);
+	for (i = 0; i < AMDGPU_MAX_GFX_RINGS; ++i) {
+		if (isolation_work->xcp_id == adev->gfx.gfx_ring[i].xcp_id)
+			fences += amdgpu_fence_count_emitted(&adev->gfx.gfx_ring[i]);
+	}
+	for (i = 0; i < (AMDGPU_MAX_COMPUTE_RINGS * AMDGPU_MAX_GC_INSTANCES); ++i) {
+		if (isolation_work->xcp_id == adev->gfx.compute_ring[i].xcp_id)
+			fences += amdgpu_fence_count_emitted(&adev->gfx.compute_ring[i]);
+	}
+	if (fences) {
+		/* we've already had our timeslice, so let's wrap this up */
+		schedule_delayed_work(&adev->gfx.enforce_isolation[idx].work,
+				      msecs_to_jiffies(1));
+	} else {
+		/* Tell KFD to resume the runqueue */
+		WARN_ON_ONCE(!adev->gfx.userq_sch_inactive[idx]);
+		WARN_ON_ONCE(adev->gfx.userq_sch_req_count[idx]);
+
+		amdgpu_userq_start_sched_for_enforce_isolation(adev, idx);
+		if (adev->kfd.init_complete)
+			amdgpu_amdkfd_start_sched(adev, idx);
+		adev->gfx.userq_sch_inactive[idx] = false;
+	}
+	mutex_unlock(&adev->enforce_isolation_mutex);
+}
+
+/**
+ * amdgpu_gfx_enforce_isolation_wait_for_kfd - Manage KFD wait period for process isolation
+ * @adev: amdgpu_device pointer
+ * @idx: Index of the GPU partition
+ *
+ * When kernel submissions come in, the jobs are given a time slice and once
+ * that time slice is up, if there are KFD user queues active, kernel
+ * submissions are blocked until KFD has had its time slice. Once the KFD time
+ * slice is up, KFD user queues are preempted and kernel submissions are
+ * unblocked and allowed to run again.
+ */
+static void
+amdgpu_gfx_enforce_isolation_wait_for_kfd(struct amdgpu_device *adev,
+					  u32 idx)
+{
+	unsigned long cjiffies;
+	bool wait = false;
+
+	mutex_lock(&adev->enforce_isolation_mutex);
+	if (adev->enforce_isolation[idx] == AMDGPU_ENFORCE_ISOLATION_ENABLE) {
+		/* set the initial values if nothing is set */
+		if (!adev->gfx.enforce_isolation_jiffies[idx]) {
+			adev->gfx.enforce_isolation_jiffies[idx] = jiffies;
+			adev->gfx.enforce_isolation_time[idx] =	GFX_SLICE_PERIOD_MS;
+		}
+		/* Make sure KFD gets a chance to run */
+		if (amdgpu_amdkfd_compute_active(adev, idx)) {
+			cjiffies = jiffies;
+			if (time_after(cjiffies, adev->gfx.enforce_isolation_jiffies[idx])) {
+				cjiffies -= adev->gfx.enforce_isolation_jiffies[idx];
+				if ((jiffies_to_msecs(cjiffies) >= GFX_SLICE_PERIOD_MS)) {
+					/* if our time is up, let KGD work drain before scheduling more */
+					wait = true;
+					/* reset the timer period */
+					adev->gfx.enforce_isolation_time[idx] =	GFX_SLICE_PERIOD_MS;
+				} else {
+					/* set the timer period to what's left in our time slice */
+					adev->gfx.enforce_isolation_time[idx] =
+						GFX_SLICE_PERIOD_MS - jiffies_to_msecs(cjiffies);
+				}
+			} else {
+				/* if jiffies wrap around we will just wait a little longer */
+				adev->gfx.enforce_isolation_jiffies[idx] = jiffies;
+			}
+		} else {
+			/* if there is no KFD work, then set the full slice period */
+			adev->gfx.enforce_isolation_jiffies[idx] = jiffies;
+			adev->gfx.enforce_isolation_time[idx] = GFX_SLICE_PERIOD_MS;
+		}
+	}
+	mutex_unlock(&adev->enforce_isolation_mutex);
+
+	if (wait)
+		msleep(GFX_SLICE_PERIOD_MS);
+}
+
+/**
+ * amdgpu_gfx_enforce_isolation_ring_begin_use - Begin use of a ring with enforced isolation
+ * @ring: Pointer to the amdgpu_ring structure
+ *
+ * Ring begin_use helper implementation for gfx which serializes access to the
+ * gfx IP between kernel submission IOCTLs and KFD user queues when isolation
+ * enforcement is enabled. The kernel submission IOCTLs and KFD user queues
+ * each get a time slice when both are active.
+ */
+void amdgpu_gfx_enforce_isolation_ring_begin_use(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 idx;
+	bool sched_work = false;
+
+	if (!adev->gfx.enable_cleaner_shader)
+		return;
+
+	if (ring->xcp_id == AMDGPU_XCP_NO_PARTITION)
+		idx = 0;
+	else
+		idx = ring->xcp_id;
+
+	if (idx >= MAX_XCP)
+		return;
+
+	/* Don't submit more work until KFD has had some time */
+	amdgpu_gfx_enforce_isolation_wait_for_kfd(adev, idx);
+
+	mutex_lock(&adev->enforce_isolation_mutex);
+	if (adev->enforce_isolation[idx] == AMDGPU_ENFORCE_ISOLATION_ENABLE) {
+		if (adev->kfd.init_complete)
+			sched_work = true;
+	}
+	mutex_unlock(&adev->enforce_isolation_mutex);
+
+	if (sched_work)
+		amdgpu_gfx_kfd_sch_ctrl(adev, idx, false);
+}
+
+/**
+ * amdgpu_gfx_enforce_isolation_ring_end_use - End use of a ring with enforced isolation
+ * @ring: Pointer to the amdgpu_ring structure
+ *
+ * Ring end_use helper implementation for gfx which serializes access to the
+ * gfx IP between kernel submission IOCTLs and KFD user queues when isolation
+ * enforcement is enabled. The kernel submission IOCTLs and KFD user queues
+ * each get a time slice when both are active.
+ */
+void amdgpu_gfx_enforce_isolation_ring_end_use(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 idx;
+	bool sched_work = false;
+
+	if (!adev->gfx.enable_cleaner_shader)
+		return;
+
+	if (ring->xcp_id == AMDGPU_XCP_NO_PARTITION)
+		idx = 0;
+	else
+		idx = ring->xcp_id;
+
+	if (idx >= MAX_XCP)
+		return;
+
+	mutex_lock(&adev->enforce_isolation_mutex);
+	if (adev->enforce_isolation[idx] == AMDGPU_ENFORCE_ISOLATION_ENABLE) {
+		if (adev->kfd.init_complete)
+			sched_work = true;
+	}
+	mutex_unlock(&adev->enforce_isolation_mutex);
+
+	if (sched_work)
+		amdgpu_gfx_kfd_sch_ctrl(adev, idx, true);
+}
+
+void amdgpu_gfx_profile_idle_work_handler(struct work_struct *work)
+{
+	struct amdgpu_device *adev =
+		container_of(work, struct amdgpu_device, gfx.idle_work.work);
+	enum PP_SMC_POWER_PROFILE profile;
+	u32 i, fences = 0;
+	int r;
+
+	if (adev->gfx.num_gfx_rings)
+		profile = PP_SMC_POWER_PROFILE_FULLSCREEN3D;
+	else
+		profile = PP_SMC_POWER_PROFILE_COMPUTE;
+
+	for (i = 0; i < AMDGPU_MAX_GFX_RINGS; ++i)
+		fences += amdgpu_fence_count_emitted(&adev->gfx.gfx_ring[i]);
+	for (i = 0; i < (AMDGPU_MAX_COMPUTE_RINGS * AMDGPU_MAX_GC_INSTANCES); ++i)
+		fences += amdgpu_fence_count_emitted(&adev->gfx.compute_ring[i]);
+	if (!fences && !atomic_read(&adev->gfx.total_submission_cnt)) {
+		mutex_lock(&adev->gfx.workload_profile_mutex);
+		if (adev->gfx.workload_profile_active) {
+			r = amdgpu_dpm_switch_power_profile(adev, profile, false);
+			if (r)
+				dev_warn(adev->dev, "(%d) failed to disable %s power profile mode\n", r,
+					 profile == PP_SMC_POWER_PROFILE_FULLSCREEN3D ?
+					 "fullscreen 3D" : "compute");
+			adev->gfx.workload_profile_active = false;
+		}
+		mutex_unlock(&adev->gfx.workload_profile_mutex);
+	} else {
+		schedule_delayed_work(&adev->gfx.idle_work, GFX_PROFILE_IDLE_TIMEOUT);
+	}
+}
+
+void amdgpu_gfx_profile_ring_begin_use(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	enum PP_SMC_POWER_PROFILE profile;
+	int r;
+
+	if (amdgpu_dpm_is_overdrive_enabled(adev))
+		return;
+
+	if (adev->gfx.num_gfx_rings)
+		profile = PP_SMC_POWER_PROFILE_FULLSCREEN3D;
+	else
+		profile = PP_SMC_POWER_PROFILE_COMPUTE;
+
+	atomic_inc(&adev->gfx.total_submission_cnt);
+
+	cancel_delayed_work_sync(&adev->gfx.idle_work);
+
+	/* We can safely return early here because we've cancelled the
+	 * the delayed work so there is no one else to set it to false
+	 * and we don't care if someone else sets it to true.
+	 */
+	if (adev->gfx.workload_profile_active)
+		return;
+
+	mutex_lock(&adev->gfx.workload_profile_mutex);
+	if (!adev->gfx.workload_profile_active) {
+		r = amdgpu_dpm_switch_power_profile(adev, profile, true);
+		if (r)
+			dev_warn(adev->dev, "(%d) failed to disable %s power profile mode\n", r,
+				 profile == PP_SMC_POWER_PROFILE_FULLSCREEN3D ?
+				 "fullscreen 3D" : "compute");
+		adev->gfx.workload_profile_active = true;
+	}
+	mutex_unlock(&adev->gfx.workload_profile_mutex);
+}
+
+void amdgpu_gfx_profile_ring_end_use(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (amdgpu_dpm_is_overdrive_enabled(adev))
+		return;
+
+	atomic_dec(&ring->adev->gfx.total_submission_cnt);
+
+	schedule_delayed_work(&ring->adev->gfx.idle_work, GFX_PROFILE_IDLE_TIMEOUT);
+}
+
+/**
+ * amdgpu_gfx_csb_preamble_start - Set CSB preamble start
+ *
+ * @buffer: This is an output variable that gets the PACKET3 preamble setup.
+ *
+ * Return:
+ * return the latest index.
+ */
+u32 amdgpu_gfx_csb_preamble_start(u32 *buffer)
+{
+	u32 count = 0;
+
+	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
+	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
+
+	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1));
+	buffer[count++] = cpu_to_le32(0x80000000);
+	buffer[count++] = cpu_to_le32(0x80000000);
+
+	return count;
+}
+
+/**
+ * amdgpu_gfx_csb_data_parser - Parser CS data
+ *
+ * @adev: amdgpu_device pointer used to get the CS data and other gfx info.
+ * @buffer: This is an output variable that gets the PACKET3 preamble end.
+ * @count: Index to start set the preemble end.
+ *
+ * Return:
+ * return the latest index.
+ */
+u32 amdgpu_gfx_csb_data_parser(struct amdgpu_device *adev, u32 *buffer, u32 count)
+{
+	const struct cs_section_def *sect = NULL;
+	const struct cs_extent_def *ext = NULL;
+	u32 i;
+
+	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
+		for (ext = sect->section; ext->extent != NULL; ++ext) {
+			if (sect->id == SECT_CONTEXT) {
+				buffer[count++] = cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
+				buffer[count++] = cpu_to_le32(ext->reg_index - PACKET3_SET_CONTEXT_REG_START);
+
+				for (i = 0; i < ext->reg_count; i++)
+					buffer[count++] = cpu_to_le32(ext->extent[i]);
+			}
+		}
+	}
+
+	return count;
+}
+
+/**
+ * amdgpu_gfx_csb_preamble_end - Set CSB preamble end
+ *
+ * @buffer: This is an output variable that gets the PACKET3 preamble end.
+ * @count: Index to start set the preemble end.
+ */
+void amdgpu_gfx_csb_preamble_end(u32 *buffer, u32 count)
+{
+	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
+	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE);
+
+	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0));
+	buffer[count++] = cpu_to_le32(0);
+}
+
+/*
+ * debugfs for to enable/disable gfx job submission to specific core.
+ */
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_debugfs_gfx_sched_mask_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+
+	mask = (1ULL << adev->gfx.num_gfx_rings) - 1;
+	if ((val & mask) == 0)
+		return -EINVAL;
+
+	for (i = 0; i < adev->gfx.num_gfx_rings; ++i) {
+		ring = &adev->gfx.gfx_ring[i];
+		if (val & (1 << i))
+			ring->sched.ready = true;
+		else
+			ring->sched.ready = false;
+	}
+	/* publish sched.ready flag update effective immediately across smp */
+	smp_rmb();
+	return 0;
+}
+
+static int amdgpu_debugfs_gfx_sched_mask_get(void *data, u64 *val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+	for (i = 0; i < adev->gfx.num_gfx_rings; ++i) {
+		ring = &adev->gfx.gfx_ring[i];
+		if (ring->sched.ready)
+			mask |= 1ULL << i;
+	}
+
+	*val = mask;
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_debugfs_gfx_sched_mask_fops,
+			 amdgpu_debugfs_gfx_sched_mask_get,
+			 amdgpu_debugfs_gfx_sched_mask_set, "%llx\n");
+
+#endif
+
+void amdgpu_debugfs_gfx_sched_mask_init(struct amdgpu_device *adev)
+{
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+	char name[32];
+
+	if (!(adev->gfx.num_gfx_rings > 1))
+		return;
+	sprintf(name, "amdgpu_gfx_sched_mask");
+	debugfs_create_file(name, 0600, root, adev,
+			    &amdgpu_debugfs_gfx_sched_mask_fops);
+#endif
+}
+
+/*
+ * debugfs for to enable/disable compute job submission to specific core.
+ */
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_debugfs_compute_sched_mask_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+
+	mask = (1ULL << adev->gfx.num_compute_rings) - 1;
+	if ((val & mask) == 0)
+		return -EINVAL;
+
+	for (i = 0; i < adev->gfx.num_compute_rings; ++i) {
+		ring = &adev->gfx.compute_ring[i];
+		if (val & (1 << i))
+			ring->sched.ready = true;
+		else
+			ring->sched.ready = false;
+	}
+
+	/* publish sched.ready flag update effective immediately across smp */
+	smp_rmb();
+	return 0;
+}
+
+static int amdgpu_debugfs_compute_sched_mask_get(void *data, u64 *val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+	for (i = 0; i < adev->gfx.num_compute_rings; ++i) {
+		ring = &adev->gfx.compute_ring[i];
+		if (ring->sched.ready)
+			mask |= 1ULL << i;
+	}
+
+	*val = mask;
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_debugfs_compute_sched_mask_fops,
+			 amdgpu_debugfs_compute_sched_mask_get,
+			 amdgpu_debugfs_compute_sched_mask_set, "%llx\n");
+
+#endif
+
+void amdgpu_debugfs_compute_sched_mask_init(struct amdgpu_device *adev)
+{
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+	char name[32];
+
+	if (!(adev->gfx.num_compute_rings > 1))
+		return;
+	sprintf(name, "amdgpu_compute_sched_mask");
+	debugfs_create_file(name, 0600, root, adev,
+			    &amdgpu_debugfs_compute_sched_mask_fops);
+#endif
+}
+
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
index de9e7a00bb15..efd61a1ccc66 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
@@ -34,6 +34,7 @@
 #include "soc15.h"
 #include "amdgpu_ras.h"
 #include "amdgpu_ring_mux.h"
+#include "amdgpu_xcp.h"
 
 /* GFX current status */
 #define AMDGPU_GFX_NORMAL_MODE			0x00000000L
@@ -42,8 +43,11 @@
 #define AMDGPU_GFX_CG_DISABLED_MODE		0x00000004L
 #define AMDGPU_GFX_LBPW_DISABLED_MODE		0x00000008L
 
-#define AMDGPU_MAX_GFX_QUEUES KGD_MAX_QUEUES
-#define AMDGPU_MAX_COMPUTE_QUEUES KGD_MAX_QUEUES
+#define AMDGPU_MAX_GC_INSTANCES		8
+#define AMDGPU_MAX_QUEUES		128
+
+#define AMDGPU_MAX_GFX_QUEUES AMDGPU_MAX_QUEUES
+#define AMDGPU_MAX_COMPUTE_QUEUES AMDGPU_MAX_QUEUES
 
 enum amdgpu_gfx_pipe_priority {
 	AMDGPU_GFX_PIPE_PRIO_NORMAL = AMDGPU_RING_PRIO_1,
@@ -53,6 +57,48 @@ enum amdgpu_gfx_pipe_priority {
 #define AMDGPU_GFX_QUEUE_PRIORITY_MINIMUM  0
 #define AMDGPU_GFX_QUEUE_PRIORITY_MAXIMUM  15
 
+/* 1 second timeout */
+#define GFX_PROFILE_IDLE_TIMEOUT	msecs_to_jiffies(1000)
+
+enum amdgpu_gfx_partition {
+	AMDGPU_SPX_PARTITION_MODE = 0,
+	AMDGPU_DPX_PARTITION_MODE = 1,
+	AMDGPU_TPX_PARTITION_MODE = 2,
+	AMDGPU_QPX_PARTITION_MODE = 3,
+	AMDGPU_CPX_PARTITION_MODE = 4,
+	AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE = -1,
+	/* Automatically choose the right mode */
+	AMDGPU_AUTO_COMPUTE_PARTITION_MODE = -2,
+};
+
+#define NUM_XCC(x) hweight16(x)
+
+enum amdgpu_gfx_ras_mem_id_type {
+	AMDGPU_GFX_CP_MEM = 0,
+	AMDGPU_GFX_GCEA_MEM,
+	AMDGPU_GFX_GC_CANE_MEM,
+	AMDGPU_GFX_GCUTCL2_MEM,
+	AMDGPU_GFX_GDS_MEM,
+	AMDGPU_GFX_LDS_MEM,
+	AMDGPU_GFX_RLC_MEM,
+	AMDGPU_GFX_SP_MEM,
+	AMDGPU_GFX_SPI_MEM,
+	AMDGPU_GFX_SQC_MEM,
+	AMDGPU_GFX_SQ_MEM,
+	AMDGPU_GFX_TA_MEM,
+	AMDGPU_GFX_TCC_MEM,
+	AMDGPU_GFX_TCA_MEM,
+	AMDGPU_GFX_TCI_MEM,
+	AMDGPU_GFX_TCP_MEM,
+	AMDGPU_GFX_TD_MEM,
+	AMDGPU_GFX_TCX_MEM,
+	AMDGPU_GFX_ATC_L2_MEM,
+	AMDGPU_GFX_UTCL2_MEM,
+	AMDGPU_GFX_VML2_MEM,
+	AMDGPU_GFX_VML2_WALKER_MEM,
+	AMDGPU_GFX_MEM_TYPE_NUM
+};
+
 struct amdgpu_mec {
 	struct amdgpu_bo	*hpd_eop_obj;
 	u64			hpd_eop_gpu_addr;
@@ -64,8 +110,10 @@ struct amdgpu_mec {
 	u32 num_mec;
 	u32 num_pipe_per_mec;
 	u32 num_queue_per_pipe;
-	void			*mqd_backup[AMDGPU_MAX_COMPUTE_RINGS + 1];
+	void			*mqd_backup[AMDGPU_MAX_COMPUTE_RINGS * AMDGPU_MAX_GC_INSTANCES];
+};
 
+struct amdgpu_mec_bitmap {
 	/* These are the resources for which amdgpu takes ownership */
 	DECLARE_BITMAP(queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
 };
@@ -94,6 +142,10 @@ struct kiq_pm4_funcs {
 	void (*kiq_invalidate_tlbs)(struct amdgpu_ring *kiq_ring,
 				uint16_t pasid, uint32_t flush_type,
 				bool all_hub);
+	void (*kiq_reset_hw_queue)(struct amdgpu_ring *kiq_ring,
+				   uint32_t queue_type, uint32_t me_id,
+				   uint32_t pipe_id, uint32_t queue_id,
+				   uint32_t xcc_id, uint32_t vmid);
 	/* Packet sizes */
 	int set_resources_size;
 	int map_queues_size;
@@ -109,6 +161,7 @@ struct amdgpu_kiq {
 	struct amdgpu_ring	ring;
 	struct amdgpu_irq_src	irq;
 	const struct kiq_pm4_funcs *pmf;
+	void			*mqd_backup;
 };
 
 /*
@@ -117,10 +170,46 @@ struct amdgpu_kiq {
 #define AMDGPU_GFX_MAX_SE 4
 #define AMDGPU_GFX_MAX_SH_PER_SE 2
 
+/**
+ * amdgpu_rb_config - Configure a single Render Backend (RB)
+ *
+ * Bad RBs are fused off and there is a harvest register the driver reads to
+ * determine which RB(s) are fused off so that the driver can configure the
+ * hardware state so that nothing gets sent to them. There are also user
+ * harvest registers that the driver can program to disable additional RBs,
+ * etc., for testing purposes.
+ */
 struct amdgpu_rb_config {
+	/**
+	 * @rb_backend_disable:
+	 *
+	 * The value captured from register RB_BACKEND_DISABLE indicates if the
+	 * RB backend is disabled or not.
+	 */
 	uint32_t rb_backend_disable;
+
+	/**
+	 * @user_rb_backend_disable:
+	 *
+	 * The value captured from register USER_RB_BACKEND_DISABLE indicates
+	 * if the User RB backend is disabled or not.
+	 */
 	uint32_t user_rb_backend_disable;
+
+	/**
+	 * @raster_config:
+	 *
+	 * To set up all of the states, it is necessary to have two registers
+	 * to keep all of the states. This field holds the first register.
+	 */
 	uint32_t raster_config;
+
+	/**
+	 * @raster_config_1:
+	 *
+	 * To set up all of the states, it is necessary to have two registers
+	 * to keep all of the states. This field holds the second register.
+	 */
 	uint32_t raster_config_1;
 };
 
@@ -168,6 +257,13 @@ struct amdgpu_gfx_config {
 	uint32_t macrotile_mode_array[16];
 
 	struct gb_addr_config gb_addr_config_fields;
+
+	/**
+	 * @rb_config:
+	 *
+	 * Matrix that keeps all the Render Backend (color and depth buffer
+	 * handling) configuration on the 3D engine.
+	 */
 	struct amdgpu_rb_config rb_config[AMDGPU_GFX_MAX_SE][AMDGPU_GFX_MAX_SH_PER_SE];
 
 	/* gfx configure feature */
@@ -192,6 +288,15 @@ struct amdgpu_gfx_config {
 	uint32_t gc_gl1c_per_sa;
 	uint32_t gc_gl1c_size_per_instance;
 	uint32_t gc_gl2c_per_gpu;
+	uint32_t gc_tcp_size_per_cu;
+	uint32_t gc_num_cu_per_sqc;
+	uint32_t gc_tcc_size;
+	uint32_t gc_tcp_cache_line_size;
+	uint32_t gc_instruction_cache_size_per_sqc;
+	uint32_t gc_instruction_cache_line_size;
+	uint32_t gc_scalar_data_cache_size_per_sqc;
+	uint32_t gc_scalar_data_cache_line_size;
+	uint32_t gc_tcc_cache_line_size;
 };
 
 struct amdgpu_cu_info {
@@ -205,13 +310,12 @@ struct amdgpu_cu_info {
 	uint32_t number;
 	uint32_t ao_cu_mask;
 	uint32_t ao_cu_bitmap[4][4];
-	uint32_t bitmap[4][4];
+	uint32_t bitmap[AMDGPU_MAX_GC_INSTANCES][4][4];
 };
 
 struct amdgpu_gfx_ras {
 	struct amdgpu_ras_block_object  ras_block;
 	void (*enable_watchdog_timer)(struct amdgpu_device *adev);
-	bool (*query_utcl2_poison_status)(struct amdgpu_device *adev);
 	int (*rlc_gc_fed_irq)(struct amdgpu_device *adev,
 				struct amdgpu_irq_src *source,
 				struct amdgpu_iv_entry *entry);
@@ -219,23 +323,39 @@ struct amdgpu_gfx_ras {
 						struct amdgpu_iv_entry *entry);
 };
 
+struct amdgpu_gfx_shadow_info {
+	u32 shadow_size;
+	u32 shadow_alignment;
+	u32 csa_size;
+	u32 csa_alignment;
+};
+
 struct amdgpu_gfx_funcs {
 	/* get the gpu clock counter */
 	uint64_t (*get_gpu_clock_counter)(struct amdgpu_device *adev);
 	void (*select_se_sh)(struct amdgpu_device *adev, u32 se_num,
-			     u32 sh_num, u32 instance);
-	void (*read_wave_data)(struct amdgpu_device *adev, uint32_t simd,
+			     u32 sh_num, u32 instance, int xcc_id);
+	void (*read_wave_data)(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 			       uint32_t wave, uint32_t *dst, int *no_fields);
-	void (*read_wave_vgprs)(struct amdgpu_device *adev, uint32_t simd,
+	void (*read_wave_vgprs)(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				uint32_t wave, uint32_t thread, uint32_t start,
 				uint32_t size, uint32_t *dst);
-	void (*read_wave_sgprs)(struct amdgpu_device *adev, uint32_t simd,
+	void (*read_wave_sgprs)(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				uint32_t wave, uint32_t start, uint32_t size,
 				uint32_t *dst);
 	void (*select_me_pipe_q)(struct amdgpu_device *adev, u32 me, u32 pipe,
-				 u32 queue, u32 vmid);
+				 u32 queue, u32 vmid, u32 xcc_id);
 	void (*init_spm_golden)(struct amdgpu_device *adev);
 	void (*update_perfmon_mgcg)(struct amdgpu_device *adev, bool enable);
+	int (*get_gfx_shadow_info)(struct amdgpu_device *adev,
+				   struct amdgpu_gfx_shadow_info *shadow_info,
+				   bool skip_check);
+	enum amdgpu_gfx_partition
+			(*query_partition_mode)(struct amdgpu_device *adev);
+	int (*switch_partition_mode)(struct amdgpu_device *adev,
+				     int num_xccs_per_xcp);
+	int (*ih_node_to_logical_xcc)(struct amdgpu_device *adev, int ih_node);
+	int (*get_xccs_per_xcp)(struct amdgpu_device *adev);
 };
 
 struct sq_work {
@@ -277,6 +397,12 @@ struct amdgpu_me {
 	DECLARE_BITMAP(queue_bitmap, AMDGPU_MAX_GFX_QUEUES);
 };
 
+struct amdgpu_isolation_work {
+	struct amdgpu_device		*adev;
+	u32				xcp_id;
+	struct delayed_work		work;
+};
+
 struct amdgpu_gfx {
 	struct mutex			gpu_clock_mutex;
 	struct amdgpu_gfx_config	config;
@@ -285,7 +411,8 @@ struct amdgpu_gfx {
 	struct amdgpu_ce		ce;
 	struct amdgpu_me		me;
 	struct amdgpu_mec		mec;
-	struct amdgpu_kiq		kiq;
+	struct amdgpu_mec_bitmap	mec_bitmap[AMDGPU_MAX_GC_INSTANCES];
+	struct amdgpu_kiq		kiq[AMDGPU_MAX_GC_INSTANCES];
 	struct amdgpu_imu		imu;
 	bool				rs64_enable; /* firmware format */
 	const struct firmware		*me_fw;	/* ME firmware */
@@ -323,11 +450,12 @@ struct amdgpu_gfx {
 	bool				cp_fw_write_wait;
 	struct amdgpu_ring		gfx_ring[AMDGPU_MAX_GFX_RINGS];
 	unsigned			num_gfx_rings;
-	struct amdgpu_ring		compute_ring[AMDGPU_MAX_COMPUTE_RINGS];
+	struct amdgpu_ring		compute_ring[AMDGPU_MAX_COMPUTE_RINGS * AMDGPU_MAX_GC_INSTANCES];
 	unsigned			num_compute_rings;
 	struct amdgpu_irq_src		eop_irq;
 	struct amdgpu_irq_src		priv_reg_irq;
 	struct amdgpu_irq_src		priv_inst_irq;
+	struct amdgpu_irq_src		bad_op_irq;
 	struct amdgpu_irq_src		cp_ecc_error_irq;
 	struct amdgpu_irq_src		sq_irq;
 	struct amdgpu_irq_src		rlc_gc_fed_irq;
@@ -343,6 +471,8 @@ struct amdgpu_gfx {
 	/* reset mask */
 	uint32_t                        grbm_soft_reset;
 	uint32_t                        srbm_soft_reset;
+	uint32_t 			gfx_supported_reset;
+	uint32_t 			compute_supported_reset;
 
 	/* gfx off */
 	bool                            gfx_off_state;      /* true: enabled, false: disabled */
@@ -364,12 +494,63 @@ struct amdgpu_gfx {
 
 	struct amdgpu_ring		sw_gfx_ring[AMDGPU_MAX_SW_GFX_RINGS];
 	struct amdgpu_ring_mux          muxer;
+
+	bool				cp_gfx_shadow; /* for gfx11 */
+
+	uint16_t 			xcc_mask;
+	uint32_t			num_xcc_per_xcp;
+	struct mutex			partition_mutex;
+	bool				mcbp; /* mid command buffer preemption */
+
+	/* IP reg dump */
+	uint32_t			*ip_dump_core;
+	uint32_t			*ip_dump_compute_queues;
+	uint32_t			*ip_dump_gfx_queues;
+
+	struct mutex			reset_sem_mutex;
+
+	/* cleaner shader */
+	struct amdgpu_bo		*cleaner_shader_obj;
+	unsigned int                    cleaner_shader_size;
+	u64				cleaner_shader_gpu_addr;
+	void				*cleaner_shader_cpu_ptr;
+	const void			*cleaner_shader_ptr;
+	bool				enable_cleaner_shader;
+	struct amdgpu_isolation_work	enforce_isolation[MAX_XCP];
+	/* Mutex for synchronizing KFD scheduler operations */
+	struct mutex                    userq_sch_mutex;
+	u64				userq_sch_req_count[MAX_XCP];
+	bool				userq_sch_inactive[MAX_XCP];
+	unsigned long			enforce_isolation_jiffies[MAX_XCP];
+	unsigned long			enforce_isolation_time[MAX_XCP];
+
+	atomic_t			total_submission_cnt;
+	struct delayed_work		idle_work;
+	bool				workload_profile_active;
+	struct mutex                    workload_profile_mutex;
+
+	bool				disable_kq;
+	bool				disable_uq;
 };
 
+struct amdgpu_gfx_ras_reg_entry {
+	struct amdgpu_ras_err_status_reg_entry reg_entry;
+	enum amdgpu_gfx_ras_mem_id_type mem_id_type;
+	uint32_t se_num;
+};
+
+struct amdgpu_gfx_ras_mem_id_entry {
+	const struct amdgpu_ras_memory_id_entry *mem_id_ent;
+	uint32_t size;
+};
+
+#define AMDGPU_GFX_MEMID_ENT(x) {(x), ARRAY_SIZE(x)},
+
 #define amdgpu_gfx_get_gpu_clock_counter(adev) (adev)->gfx.funcs->get_gpu_clock_counter((adev))
-#define amdgpu_gfx_select_se_sh(adev, se, sh, instance) (adev)->gfx.funcs->select_se_sh((adev), (se), (sh), (instance))
-#define amdgpu_gfx_select_me_pipe_q(adev, me, pipe, q, vmid) (adev)->gfx.funcs->select_me_pipe_q((adev), (me), (pipe), (q), (vmid))
+#define amdgpu_gfx_select_se_sh(adev, se, sh, instance, xcc_id) ((adev)->gfx.funcs->select_se_sh((adev), (se), (sh), (instance), (xcc_id)))
+#define amdgpu_gfx_select_me_pipe_q(adev, me, pipe, q, vmid, xcc_id) ((adev)->gfx.funcs->select_me_pipe_q((adev), (me), (pipe), (q), (vmid), (xcc_id)))
 #define amdgpu_gfx_init_spm_golden(adev) (adev)->gfx.funcs->init_spm_golden((adev))
+#define amdgpu_gfx_get_gfx_shadow_info(adev, si) ((adev)->gfx.funcs->get_gfx_shadow_info((adev), (si), false))
 
 /**
  * amdgpu_gfx_create_bitmask - create a bitmask
@@ -387,21 +568,21 @@ static inline u32 amdgpu_gfx_create_bitmask(u32 bit_width)
 void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se,
 				 unsigned max_sh);
 
-int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,
-			     struct amdgpu_ring *ring,
-			     struct amdgpu_irq_src *irq);
+int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, int xcc_id);
 
 void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring);
 
-void amdgpu_gfx_kiq_fini(struct amdgpu_device *adev);
+void amdgpu_gfx_kiq_fini(struct amdgpu_device *adev, int xcc_id);
 int amdgpu_gfx_kiq_init(struct amdgpu_device *adev,
-			unsigned hpd_size);
+			unsigned hpd_size, int xcc_id);
 
 int amdgpu_gfx_mqd_sw_init(struct amdgpu_device *adev,
-			   unsigned mqd_size);
-void amdgpu_gfx_mqd_sw_fini(struct amdgpu_device *adev);
-int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev);
-int amdgpu_gfx_enable_kcq(struct amdgpu_device *adev);
+			   unsigned mqd_size, int xcc_id);
+void amdgpu_gfx_mqd_sw_fini(struct amdgpu_device *adev, int xcc_id);
+int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev, int xcc_id);
+int amdgpu_gfx_enable_kcq(struct amdgpu_device *adev, int xcc_id);
+int amdgpu_gfx_disable_kgq(struct amdgpu_device *adev, int xcc_id);
+int amdgpu_gfx_enable_kgq(struct amdgpu_device *adev, int xcc_id);
 
 void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev);
 void amdgpu_gfx_graphics_queue_acquire(struct amdgpu_device *adev);
@@ -410,19 +591,16 @@ int amdgpu_gfx_mec_queue_to_bit(struct amdgpu_device *adev, int mec,
 				int pipe, int queue);
 void amdgpu_queue_mask_bit_to_mec_queue(struct amdgpu_device *adev, int bit,
 				 int *mec, int *pipe, int *queue);
-bool amdgpu_gfx_is_mec_queue_enabled(struct amdgpu_device *adev, int mec,
-				     int pipe, int queue);
+bool amdgpu_gfx_is_mec_queue_enabled(struct amdgpu_device *adev, int xcc_id,
+				     int mec, int pipe, int queue);
 bool amdgpu_gfx_is_high_priority_compute_queue(struct amdgpu_device *adev,
 					       struct amdgpu_ring *ring);
 bool amdgpu_gfx_is_high_priority_graphics_queue(struct amdgpu_device *adev,
 						struct amdgpu_ring *ring);
-int amdgpu_gfx_me_queue_to_bit(struct amdgpu_device *adev, int me,
-			       int pipe, int queue);
-void amdgpu_gfx_bit_to_me_queue(struct amdgpu_device *adev, int bit,
-				int *me, int *pipe, int *queue);
 bool amdgpu_gfx_is_me_queue_enabled(struct amdgpu_device *adev, int me,
 				    int pipe, int queue);
 void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable);
+void amdgpu_gfx_off_ctrl_immediate(struct amdgpu_device *adev, bool enable);
 int amdgpu_get_gfx_off_status(struct amdgpu_device *adev, uint32_t *value);
 int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block);
 void amdgpu_gfx_ras_fini(struct amdgpu_device *adev);
@@ -435,12 +613,59 @@ int amdgpu_gfx_process_ras_data_cb(struct amdgpu_device *adev,
 int amdgpu_gfx_cp_ecc_error_irq(struct amdgpu_device *adev,
 				  struct amdgpu_irq_src *source,
 				  struct amdgpu_iv_entry *entry);
-uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg);
-void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v);
+uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg, uint32_t xcc_id);
+void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v, uint32_t xcc_id);
+int amdgpu_kiq_hdp_flush(struct amdgpu_device *adev);
 int amdgpu_gfx_get_num_kcq(struct amdgpu_device *adev);
 void amdgpu_gfx_cp_init_microcode(struct amdgpu_device *adev, uint32_t ucode_id);
 
 int amdgpu_gfx_ras_sw_init(struct amdgpu_device *adev);
 int amdgpu_gfx_poison_consumption_handler(struct amdgpu_device *adev,
 						struct amdgpu_iv_entry *entry);
+
+bool amdgpu_gfx_is_master_xcc(struct amdgpu_device *adev, int xcc_id);
+int amdgpu_gfx_sysfs_init(struct amdgpu_device *adev);
+void amdgpu_gfx_sysfs_fini(struct amdgpu_device *adev);
+void amdgpu_gfx_ras_error_func(struct amdgpu_device *adev,
+		void *ras_error_status,
+		void (*func)(struct amdgpu_device *adev, void *ras_error_status,
+				int xcc_id));
+int amdgpu_gfx_cleaner_shader_sw_init(struct amdgpu_device *adev,
+				      unsigned int cleaner_shader_size);
+void amdgpu_gfx_cleaner_shader_sw_fini(struct amdgpu_device *adev);
+void amdgpu_gfx_cleaner_shader_init(struct amdgpu_device *adev,
+				    unsigned int cleaner_shader_size,
+				    const void *cleaner_shader_ptr);
+void amdgpu_gfx_enforce_isolation_handler(struct work_struct *work);
+void amdgpu_gfx_enforce_isolation_ring_begin_use(struct amdgpu_ring *ring);
+void amdgpu_gfx_enforce_isolation_ring_end_use(struct amdgpu_ring *ring);
+
+void amdgpu_gfx_profile_idle_work_handler(struct work_struct *work);
+void amdgpu_gfx_profile_ring_begin_use(struct amdgpu_ring *ring);
+void amdgpu_gfx_profile_ring_end_use(struct amdgpu_ring *ring);
+u32 amdgpu_gfx_csb_preamble_start(u32 *buffer);
+u32 amdgpu_gfx_csb_data_parser(struct amdgpu_device *adev, u32 *buffer, u32 count);
+void amdgpu_gfx_csb_preamble_end(u32 *buffer, u32 count);
+
+void amdgpu_debugfs_gfx_sched_mask_init(struct amdgpu_device *adev);
+void amdgpu_debugfs_compute_sched_mask_init(struct amdgpu_device *adev);
+
+static inline const char *amdgpu_gfx_compute_mode_desc(int mode)
+{
+	switch (mode) {
+	case AMDGPU_SPX_PARTITION_MODE:
+		return "SPX";
+	case AMDGPU_DPX_PARTITION_MODE:
+		return "DPX";
+	case AMDGPU_TPX_PARTITION_MODE:
+		return "TPX";
+	case AMDGPU_QPX_PARTITION_MODE:
+		return "QPX";
+	case AMDGPU_CPX_PARTITION_MODE:
+		return "CPX";
+	default:
+		return "UNKNOWN";
+	}
+}
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
index 88bc7f5f46e6..869bceb0fe2c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
@@ -32,11 +32,19 @@
 #include "amdgpu.h"
 #include "amdgpu_gmc.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_reset.h"
 #include "amdgpu_xgmi.h"
 
 #include <drm/drm_drv.h>
 #include <drm/ttm/ttm_tt.h>
 
+static const u64 four_gb = 0x100000000ULL;
+
+bool amdgpu_gmc_is_pdb0_enabled(struct amdgpu_device *adev)
+{
+	return adev->gmc.xgmi.connected_to_cpu || amdgpu_virt_xgmi_migrate_enabled(adev);
+}
+
 /**
  * amdgpu_gmc_pdb0_alloc - allocate vram for pdb0
  *
@@ -51,7 +59,7 @@ int amdgpu_gmc_pdb0_alloc(struct amdgpu_device *adev)
 	struct amdgpu_bo_param bp;
 	u64 vram_size = adev->gmc.xgmi.node_segment_size * adev->gmc.xgmi.num_physical_nodes;
 	uint32_t pde0_page_shift = adev->gmc.vmid0_page_table_block_size + 21;
-	uint32_t npdes = (vram_size + (1ULL << pde0_page_shift) -1) >> pde0_page_shift;
+	uint32_t npdes = (vram_size + (1ULL << pde0_page_shift) - 1) >> pde0_page_shift;
 
 	memset(&bp, 0, sizeof(bp));
 	bp.size = PAGE_ALIGN((npdes + 1) * 8);
@@ -180,6 +188,9 @@ uint64_t amdgpu_gmc_agp_addr(struct ttm_buffer_object *bo)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);
 
+	if (!bo->ttm)
+		return AMDGPU_BO_INVALID_OFFSET;
+
 	if (bo->ttm->num_pages != 1 || bo->ttm->caching == ttm_cached)
 		return AMDGPU_BO_INVALID_OFFSET;
 
@@ -247,10 +258,20 @@ void amdgpu_gmc_sysvm_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc
 	u64 hive_vram_end = mc->xgmi.node_segment_size * mc->xgmi.num_physical_nodes - 1;
 	mc->vram_start = mc->xgmi.node_segment_size * mc->xgmi.physical_node_id;
 	mc->vram_end = mc->vram_start + mc->xgmi.node_segment_size - 1;
-	mc->gart_start = hive_vram_end + 1;
+	/* node_segment_size may not 4GB aligned on SRIOV, align up is needed. */
+	mc->gart_start = ALIGN(hive_vram_end + 1, four_gb);
 	mc->gart_end = mc->gart_start + mc->gart_size - 1;
-	mc->fb_start = hive_vram_start;
-	mc->fb_end = hive_vram_end;
+	if (amdgpu_virt_xgmi_migrate_enabled(adev)) {
+		/* set mc->vram_start to 0 to switch the returned GPU address of
+		 * amdgpu_bo_create_reserved() from FB aperture to GART aperture.
+		 */
+		mc->vram_start = 0;
+		mc->vram_end = mc->vram_start + mc->mc_vram_size - 1;
+		mc->visible_vram_size = min(mc->visible_vram_size, mc->real_vram_size);
+	} else {
+		mc->fb_start = hive_vram_start;
+		mc->fb_end = hive_vram_end;
+	}
 	dev_info(adev->dev, "VRAM: %lluM 0x%016llX - 0x%016llX (%lluM used)\n",
 			mc->mc_vram_size >> 20, mc->vram_start,
 			mc->vram_end, mc->real_vram_size >> 20);
@@ -263,14 +284,15 @@ void amdgpu_gmc_sysvm_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc
  *
  * @adev: amdgpu device structure holding all necessary information
  * @mc: memory controller structure holding memory information
+ * @gart_placement: GART placement policy with respect to VRAM
  *
- * Function will place try to place GART before or after VRAM.
+ * Function will try to place GART before or after VRAM.
  * If GART size is bigger than space left then we ajust GART size.
  * Thus function will never fails.
  */
-void amdgpu_gmc_gart_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc)
+void amdgpu_gmc_gart_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc,
+			      enum amdgpu_gart_placement gart_placement)
 {
-	const uint64_t four_gb = 0x100000000ULL;
 	u64 size_af, size_bf;
 	/*To avoid the hole, limit the max mc address to AMDGPU_GMC_HOLE_START*/
 	u64 max_mc_address = min(adev->gmc.mc_mask, AMDGPU_GMC_HOLE_START - 1);
@@ -286,11 +308,22 @@ void amdgpu_gmc_gart_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc)
 		mc->gart_size = max(size_bf, size_af);
 	}
 
-	if ((size_bf >= mc->gart_size && size_bf < size_af) ||
-	    (size_af < mc->gart_size))
-		mc->gart_start = 0;
-	else
+	switch (gart_placement) {
+	case AMDGPU_GART_PLACEMENT_HIGH:
 		mc->gart_start = max_mc_address - mc->gart_size + 1;
+		break;
+	case AMDGPU_GART_PLACEMENT_LOW:
+		mc->gart_start = 0;
+		break;
+	case AMDGPU_GART_PLACEMENT_BEST_FIT:
+	default:
+		if ((size_bf >= mc->gart_size && size_bf < size_af) ||
+		    (size_af < mc->gart_size))
+			mc->gart_start = 0;
+		else
+			mc->gart_start = max_mc_address - mc->gart_size + 1;
+		break;
+	}
 
 	mc->gart_start &= ~(four_gb - 1);
 	mc->gart_end = mc->gart_start + mc->gart_size - 1;
@@ -315,14 +348,6 @@ void amdgpu_gmc_agp_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc)
 	const uint64_t sixteen_gb_mask = ~(sixteen_gb - 1);
 	u64 size_af, size_bf;
 
-	if (amdgpu_sriov_vf(adev)) {
-		mc->agp_start = 0xffffffffffff;
-		mc->agp_end = 0x0;
-		mc->agp_size = 0;
-
-		return;
-	}
-
 	if (mc->fb_start > mc->gart_start) {
 		size_bf = (mc->fb_start & sixteen_gb_mask) -
 			ALIGN(mc->gart_end + 1, sixteen_gb);
@@ -347,6 +372,25 @@ void amdgpu_gmc_agp_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc)
 }
 
 /**
+ * amdgpu_gmc_set_agp_default - Set the default AGP aperture value.
+ * @adev: amdgpu device structure holding all necessary information
+ * @mc: memory controller structure holding memory information
+ *
+ * To disable the AGP aperture, you need to set the start to a larger
+ * value than the end.  This function sets the default value which
+ * can then be overridden using amdgpu_gmc_agp_location() if you want
+ * to enable the AGP aperture on a specific chip.
+ *
+ */
+void amdgpu_gmc_set_agp_default(struct amdgpu_device *adev,
+				struct amdgpu_gmc *mc)
+{
+	mc->agp_start = 0xffffffffffff;
+	mc->agp_end = 0;
+	mc->agp_size = 0;
+}
+
+/**
  * amdgpu_gmc_fault_key - get hask key from vm fault address and pasid
  *
  * @addr: 48 bit physical address, page aligned (36 significant bits)
@@ -395,8 +439,21 @@ bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev,
 	while (fault->timestamp >= stamp) {
 		uint64_t tmp;
 
-		if (atomic64_read(&fault->key) == key)
-			return true;
+		if (atomic64_read(&fault->key) == key) {
+			/*
+			 * if we get a fault which is already present in
+			 * the fault_ring and the timestamp of
+			 * the fault is after the expired timestamp,
+			 * then this is a new fault that needs to be added
+			 * into the fault ring.
+			 */
+			if (fault->timestamp_expiry != 0 &&
+			    amdgpu_ih_ts_after(fault->timestamp_expiry,
+					       timestamp))
+				break;
+			else
+				return true;
+		}
 
 		tmp = fault->timestamp;
 		fault = &gmc->fault_ring[fault->next];
@@ -432,15 +489,35 @@ void amdgpu_gmc_filter_faults_remove(struct amdgpu_device *adev, uint64_t addr,
 {
 	struct amdgpu_gmc *gmc = &adev->gmc;
 	uint64_t key = amdgpu_gmc_fault_key(addr, pasid);
+	struct amdgpu_ih_ring *ih;
 	struct amdgpu_gmc_fault *fault;
+	uint32_t last_wptr;
+	uint64_t last_ts;
 	uint32_t hash;
 	uint64_t tmp;
 
+	if (adev->irq.retry_cam_enabled)
+		return;
+
+	ih = &adev->irq.ih1;
+	/* Get the WPTR of the last entry in IH ring */
+	last_wptr = amdgpu_ih_get_wptr(adev, ih);
+	/* Order wptr with ring data. */
+	rmb();
+	/* Get the timetamp of the last entry in IH ring */
+	last_ts = amdgpu_ih_decode_iv_ts(adev, ih, last_wptr, -1);
+
 	hash = hash_64(key, AMDGPU_GMC_FAULT_HASH_ORDER);
 	fault = &gmc->fault_ring[gmc->fault_hash[hash].idx];
 	do {
-		if (atomic64_cmpxchg(&fault->key, key, 0) == key)
+		if (atomic64_read(&fault->key) == key) {
+			/*
+			 * Update the timestamp when this fault
+			 * expired.
+			 */
+			fault->timestamp_expiry = last_ts;
 			break;
+		}
 
 		tmp = fault->timestamp;
 		fault = &gmc->fault_ring[fault->next];
@@ -504,29 +581,42 @@ void amdgpu_gmc_ras_fini(struct amdgpu_device *adev)
 	 *                    subject to change when ring number changes
 	 * Engine 17: Gart flushes
 	 */
-#define GFXHUB_FREE_VM_INV_ENGS_BITMAP		0x1FFF3
-#define MMHUB_FREE_VM_INV_ENGS_BITMAP		0x1FFF3
+#define AMDGPU_VMHUB_INV_ENG_BITMAP		0x1FFF3
 
 int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev)
 {
 	struct amdgpu_ring *ring;
-	unsigned vm_inv_engs[AMDGPU_MAX_VMHUBS] =
-		{GFXHUB_FREE_VM_INV_ENGS_BITMAP, MMHUB_FREE_VM_INV_ENGS_BITMAP,
-		GFXHUB_FREE_VM_INV_ENGS_BITMAP};
+	unsigned vm_inv_engs[AMDGPU_MAX_VMHUBS] = {0};
 	unsigned i;
 	unsigned vmhub, inv_eng;
+	struct amdgpu_ring *shared_ring;
 
-	if (adev->enable_mes) {
+	/* init the vm inv eng for all vmhubs */
+	for_each_set_bit(i, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS) {
+		vm_inv_engs[i] = AMDGPU_VMHUB_INV_ENG_BITMAP;
 		/* reserve engine 5 for firmware */
-		for (vmhub = 0; vmhub < AMDGPU_MAX_VMHUBS; vmhub++)
-			vm_inv_engs[vmhub] &= ~(1 << 5);
+		if (adev->enable_mes)
+			vm_inv_engs[i] &= ~(1 << 5);
+		/* reserve engine 6 for uni mes */
+		if (adev->enable_uni_mes)
+			vm_inv_engs[i] &= ~(1 << 6);
+		/* reserve mmhub engine 3 for firmware */
+		if (adev->enable_umsch_mm)
+			vm_inv_engs[i] &= ~(1 << 3);
 	}
 
 	for (i = 0; i < adev->num_rings; ++i) {
 		ring = adev->rings[i];
-		vmhub = ring->funcs->vmhub;
+		vmhub = ring->vm_hub;
 
-		if (ring == &adev->mes.ring)
+		if (ring == &adev->mes.ring[0] ||
+		    ring == &adev->mes.ring[1] ||
+		    ring == &adev->umsch_mm.ring ||
+		    ring == &adev->cper.ring_buf)
+			continue;
+
+		/* Skip if the ring is a shared ring */
+		if (amdgpu_sdma_is_shared_inv_eng(adev, ring))
 			continue;
 
 		inv_eng = ffs(vm_inv_engs[vmhub]);
@@ -540,12 +630,230 @@ int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev)
 		vm_inv_engs[vmhub] &= ~(1 << ring->vm_inv_eng);
 
 		dev_info(adev->dev, "ring %s uses VM inv eng %u on hub %u\n",
-			 ring->name, ring->vm_inv_eng, ring->funcs->vmhub);
+			 ring->name, ring->vm_inv_eng, ring->vm_hub);
+		/* SDMA has a special packet which allows it to use the same
+		 * invalidation engine for all the rings in one instance.
+		 * Therefore, we do not allocate a separate VM invalidation engine
+		 * for SDMA page rings. Instead, they share the VM invalidation
+		 * engine with the SDMA gfx ring. This change ensures efficient
+		 * resource management and avoids the issue of insufficient VM
+		 * invalidation engines.
+		 */
+		shared_ring = amdgpu_sdma_get_shared_ring(adev, ring);
+		if (shared_ring) {
+			shared_ring->vm_inv_eng = ring->vm_inv_eng;
+			dev_info(adev->dev, "ring %s shares VM invalidation engine %u with ring %s on hub %u\n",
+					ring->name, ring->vm_inv_eng, shared_ring->name, ring->vm_hub);
+			continue;
+		}
 	}
 
 	return 0;
 }
 
+void amdgpu_gmc_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
+			      uint32_t vmhub, uint32_t flush_type)
+{
+	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
+	struct amdgpu_vmhub *hub = &adev->vmhub[vmhub];
+	struct dma_fence *fence;
+	struct amdgpu_job *job;
+	int r;
+
+	if (!hub->sdma_invalidation_workaround || vmid ||
+	    !adev->mman.buffer_funcs_enabled || !adev->ib_pool_ready ||
+	    !ring->sched.ready) {
+		/*
+		 * A GPU reset should flush all TLBs anyway, so no need to do
+		 * this while one is ongoing.
+		 */
+		if (!down_read_trylock(&adev->reset_domain->sem))
+			return;
+
+		if (adev->gmc.flush_tlb_needs_extra_type_2)
+			adev->gmc.gmc_funcs->flush_gpu_tlb(adev, vmid,
+							   vmhub, 2);
+
+		if (adev->gmc.flush_tlb_needs_extra_type_0 && flush_type == 2)
+			adev->gmc.gmc_funcs->flush_gpu_tlb(adev, vmid,
+							   vmhub, 0);
+
+		adev->gmc.gmc_funcs->flush_gpu_tlb(adev, vmid, vmhub,
+						   flush_type);
+		up_read(&adev->reset_domain->sem);
+		return;
+	}
+
+	/* The SDMA on Navi 1x has a bug which can theoretically result in memory
+	 * corruption if an invalidation happens at the same time as an VA
+	 * translation. Avoid this by doing the invalidation from the SDMA
+	 * itself at least for GART.
+	 */
+	mutex_lock(&adev->mman.gtt_window_lock);
+	r = amdgpu_job_alloc_with_ib(ring->adev, &adev->mman.high_pr,
+				     AMDGPU_FENCE_OWNER_UNDEFINED,
+				     16 * 4, AMDGPU_IB_POOL_IMMEDIATE,
+				     &job, AMDGPU_KERNEL_JOB_ID_FLUSH_GPU_TLB);
+	if (r)
+		goto error_alloc;
+
+	job->vm_pd_addr = amdgpu_gmc_pd_addr(adev->gart.bo);
+	job->vm_needs_flush = true;
+	job->ibs->ptr[job->ibs->length_dw++] = ring->funcs->nop;
+	amdgpu_ring_pad_ib(ring, &job->ibs[0]);
+	fence = amdgpu_job_submit(job);
+	mutex_unlock(&adev->mman.gtt_window_lock);
+
+	dma_fence_wait(fence, false);
+	dma_fence_put(fence);
+
+	return;
+
+error_alloc:
+	mutex_unlock(&adev->mman.gtt_window_lock);
+	dev_err(adev->dev, "Error flushing GPU TLB using the SDMA (%d)!\n", r);
+}
+
+int amdgpu_gmc_flush_gpu_tlb_pasid(struct amdgpu_device *adev, uint16_t pasid,
+				   uint32_t flush_type, bool all_hub,
+				   uint32_t inst)
+{
+	struct amdgpu_ring *ring = &adev->gfx.kiq[inst].ring;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[inst];
+	unsigned int ndw;
+	int r, cnt = 0;
+	uint32_t seq;
+
+	/*
+	 * A GPU reset should flush all TLBs anyway, so no need to do
+	 * this while one is ongoing.
+	 */
+	if (!down_read_trylock(&adev->reset_domain->sem))
+		return 0;
+
+	if (!adev->gmc.flush_pasid_uses_kiq || !ring->sched.ready) {
+		if (adev->gmc.flush_tlb_needs_extra_type_2)
+			adev->gmc.gmc_funcs->flush_gpu_tlb_pasid(adev, pasid,
+								 2, all_hub,
+								 inst);
+
+		if (adev->gmc.flush_tlb_needs_extra_type_0 && flush_type == 2)
+			adev->gmc.gmc_funcs->flush_gpu_tlb_pasid(adev, pasid,
+								 0, all_hub,
+								 inst);
+
+		adev->gmc.gmc_funcs->flush_gpu_tlb_pasid(adev, pasid,
+							 flush_type, all_hub,
+							 inst);
+		r = 0;
+	} else {
+		/* 2 dwords flush + 8 dwords fence */
+		ndw = kiq->pmf->invalidate_tlbs_size + 8;
+
+		if (adev->gmc.flush_tlb_needs_extra_type_2)
+			ndw += kiq->pmf->invalidate_tlbs_size;
+
+		if (adev->gmc.flush_tlb_needs_extra_type_0)
+			ndw += kiq->pmf->invalidate_tlbs_size;
+
+		spin_lock(&adev->gfx.kiq[inst].ring_lock);
+		r = amdgpu_ring_alloc(ring, ndw);
+		if (r) {
+			spin_unlock(&adev->gfx.kiq[inst].ring_lock);
+			goto error_unlock_reset;
+		}
+		if (adev->gmc.flush_tlb_needs_extra_type_2)
+			kiq->pmf->kiq_invalidate_tlbs(ring, pasid, 2, all_hub);
+
+		if (flush_type == 2 && adev->gmc.flush_tlb_needs_extra_type_0)
+			kiq->pmf->kiq_invalidate_tlbs(ring, pasid, 0, all_hub);
+
+		kiq->pmf->kiq_invalidate_tlbs(ring, pasid, flush_type, all_hub);
+		r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
+		if (r) {
+			amdgpu_ring_undo(ring);
+			spin_unlock(&adev->gfx.kiq[inst].ring_lock);
+			goto error_unlock_reset;
+		}
+
+		amdgpu_ring_commit(ring);
+		spin_unlock(&adev->gfx.kiq[inst].ring_lock);
+
+		r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
+
+		might_sleep();
+		while (r < 1 && cnt++ < MAX_KIQ_REG_TRY &&
+		       !amdgpu_reset_pending(adev->reset_domain)) {
+			msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
+			r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
+		}
+
+		if (cnt > MAX_KIQ_REG_TRY) {
+			dev_err(adev->dev, "timeout waiting for kiq fence\n");
+			r = -ETIME;
+		} else
+			r = 0;
+	}
+
+error_unlock_reset:
+	up_read(&adev->reset_domain->sem);
+	return r;
+}
+
+void amdgpu_gmc_fw_reg_write_reg_wait(struct amdgpu_device *adev,
+				      uint32_t reg0, uint32_t reg1,
+				      uint32_t ref, uint32_t mask,
+				      uint32_t xcc_inst)
+{
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_inst];
+	struct amdgpu_ring *ring = &kiq->ring;
+	signed long r, cnt = 0;
+	unsigned long flags;
+	uint32_t seq;
+
+	if (adev->mes.ring[0].sched.ready) {
+		amdgpu_mes_reg_write_reg_wait(adev, reg0, reg1,
+					      ref, mask);
+		return;
+	}
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+	amdgpu_ring_alloc(ring, 32);
+	amdgpu_ring_emit_reg_write_reg_wait(ring, reg0, reg1,
+					    ref, mask);
+	r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
+	if (r)
+		goto failed_undo;
+
+	amdgpu_ring_commit(ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+
+	r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
+
+	/* don't wait anymore for IRQ context */
+	if (r < 1 && in_interrupt())
+		goto failed_kiq;
+
+	might_sleep();
+	while (r < 1 && cnt++ < MAX_KIQ_REG_TRY &&
+	       !amdgpu_reset_pending(adev->reset_domain)) {
+
+		msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
+		r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
+	}
+
+	if (cnt > MAX_KIQ_REG_TRY)
+		goto failed_kiq;
+
+	return;
+
+failed_undo:
+	amdgpu_ring_undo(ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+failed_kiq:
+	dev_err(adev->dev, "failed to write reg %x wait reg %x\n", reg0, reg1);
+}
+
 /**
  * amdgpu_gmc_tmz_set -- check and set if a device supports TMZ
  * @adev: amdgpu_device pointer
@@ -555,7 +863,7 @@ int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev)
  */
 void amdgpu_gmc_tmz_set(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	/* RAVEN */
 	case IP_VERSION(9, 2, 2):
 	case IP_VERSION(9, 1, 0):
@@ -563,6 +871,8 @@ void amdgpu_gmc_tmz_set(struct amdgpu_device *adev)
 	case IP_VERSION(9, 3, 0):
 	/* GC 10.3.7 */
 	case IP_VERSION(10, 3, 7):
+	/* GC 11.0.1 */
+	case IP_VERSION(11, 0, 1):
 		if (amdgpu_tmz == 0) {
 			adev->gmc.tmz_enabled = false;
 			dev_info(adev->dev,
@@ -586,8 +896,11 @@ void amdgpu_gmc_tmz_set(struct amdgpu_device *adev)
 	case IP_VERSION(10, 3, 1):
 	/* YELLOW_CARP*/
 	case IP_VERSION(10, 3, 3):
-	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 		/* Don't enable it by default yet.
 		 */
 		if (amdgpu_tmz < 1) {
@@ -618,16 +931,20 @@ void amdgpu_gmc_tmz_set(struct amdgpu_device *adev)
 void amdgpu_gmc_noretry_set(struct amdgpu_device *adev)
 {
 	struct amdgpu_gmc *gmc = &adev->gmc;
-	uint32_t gc_ver = adev->ip_versions[GC_HWIP][0];
+	uint32_t gc_ver = amdgpu_ip_version(adev, GC_HWIP, 0);
 	bool noretry_default = (gc_ver == IP_VERSION(9, 0, 1) ||
-				gc_ver == IP_VERSION(9, 3, 0) ||
 				gc_ver == IP_VERSION(9, 4, 0) ||
 				gc_ver == IP_VERSION(9, 4, 1) ||
 				gc_ver == IP_VERSION(9, 4, 2) ||
 				gc_ver == IP_VERSION(9, 4, 3) ||
+				gc_ver == IP_VERSION(9, 4, 4) ||
+				gc_ver == IP_VERSION(9, 5, 0) ||
 				gc_ver >= IP_VERSION(10, 3, 0));
 
-	gmc->noretry = (amdgpu_noretry == -1) ? noretry_default : amdgpu_noretry;
+	if (!amdgpu_sriov_xnack_support(adev))
+		gmc->noretry = 1;
+	else
+		gmc->noretry = (amdgpu_noretry == -1) ? noretry_default : amdgpu_noretry;
 }
 
 void amdgpu_gmc_set_vm_fault_masks(struct amdgpu_device *adev, int hub_type,
@@ -640,7 +957,7 @@ void amdgpu_gmc_set_vm_fault_masks(struct amdgpu_device *adev, int hub_type,
 	for (i = 0; i < 16; i++) {
 		reg = hub->vm_context0_cntl + hub->ctx_distance * i;
 
-		tmp = (hub_type == AMDGPU_GFXHUB_0) ?
+		tmp = (hub_type == AMDGPU_GFXHUB(0)) ?
 			RREG32_SOC15_IP(GC, reg) :
 			RREG32_SOC15_IP(MMHUB, reg);
 
@@ -649,7 +966,7 @@ void amdgpu_gmc_set_vm_fault_masks(struct amdgpu_device *adev, int hub_type,
 		else
 			tmp &= ~hub->vm_cntx_cntl_vm_fault;
 
-		(hub_type == AMDGPU_GFXHUB_0) ?
+		(hub_type == AMDGPU_GFXHUB(0)) ?
 			WREG32_SOC15_IP(GC, reg, tmp) :
 			WREG32_SOC15_IP(MMHUB, reg, tmp);
 	}
@@ -691,12 +1008,6 @@ void amdgpu_gmc_get_vbios_allocations(struct amdgpu_device *adev)
 	case CHIP_RENOIR:
 		adev->mman.keep_stolen_vga_memory = true;
 		break;
-	case CHIP_YELLOW_CARP:
-		if (amdgpu_discovery == 0) {
-			adev->mman.stolen_reserved_offset = 0x1ffb0000;
-			adev->mman.stolen_reserved_size = 64 * PAGE_SIZE;
-		}
-		break;
 	default:
 		adev->mman.keep_stolen_vga_memory = false;
 		break;
@@ -749,9 +1060,7 @@ void amdgpu_gmc_init_pdb0(struct amdgpu_device *adev)
 	 */
 	u64 vram_size = adev->gmc.xgmi.node_segment_size * adev->gmc.xgmi.num_physical_nodes;
 	u64 pde0_page_size = (1ULL<<adev->gmc.vmid0_page_table_block_size)<<21;
-	u64 vram_addr = adev->vm_manager.vram_base_offset -
-		adev->gmc.xgmi.physical_node_id * adev->gmc.xgmi.node_segment_size;
-	u64 vram_end = vram_addr + vram_size;
+	u64 vram_addr, vram_end;
 	u64 gart_ptb_gpu_pa = amdgpu_gmc_vram_pa(adev, adev->gart.bo);
 	int idx;
 
@@ -762,7 +1071,12 @@ void amdgpu_gmc_init_pdb0(struct amdgpu_device *adev)
 	flags |= AMDGPU_PTE_WRITEABLE;
 	flags |= AMDGPU_PTE_SNOOPED;
 	flags |= AMDGPU_PTE_FRAG((adev->gmc.vmid0_page_table_block_size + 9*1));
-	flags |= AMDGPU_PDE_PTE;
+	flags |= AMDGPU_PDE_PTE_FLAG(adev);
+
+	vram_addr = adev->vm_manager.vram_base_offset;
+	if (!amdgpu_virt_xgmi_migrate_enabled(adev))
+		vram_addr -= adev->gmc.xgmi.physical_node_id * adev->gmc.xgmi.node_segment_size;
+	vram_end = vram_addr + vram_size;
 
 	/* The first n PDE0 entries are used as PTE,
 	 * pointing to vram
@@ -775,7 +1089,7 @@ void amdgpu_gmc_init_pdb0(struct amdgpu_device *adev)
 	 * pointing to a 4K system page
 	 */
 	flags = AMDGPU_PTE_VALID;
-	flags |= AMDGPU_PDE_BFS(0) | AMDGPU_PTE_SNOOPED;
+	flags |= AMDGPU_PTE_SNOOPED | AMDGPU_PDE_BFS_FLAG(adev, 0);
 	/* Requires gart_ptb_gpu_pa to be 4K aligned */
 	amdgpu_gmc_set_pte_pde(adev, adev->gmc.ptr_pdb0, i, gart_ptb_gpu_pa, flags);
 	drm_dev_exit(idx);
@@ -805,18 +1119,6 @@ uint64_t amdgpu_gmc_vram_pa(struct amdgpu_device *adev, struct amdgpu_bo *bo)
 	return amdgpu_gmc_vram_mc2pa(adev, amdgpu_bo_gpu_offset(bo));
 }
 
-/**
- * amdgpu_gmc_vram_cpu_pa - calculate vram buffer object's physical address
- * from CPU's view
- *
- * @adev: amdgpu_device pointer
- * @bo: amdgpu buffer object
- */
-uint64_t amdgpu_gmc_vram_cpu_pa(struct amdgpu_device *adev, struct amdgpu_bo *bo)
-{
-	return amdgpu_bo_gpu_offset(bo) - adev->gmc.vram_start + adev->gmc.aper_base;
-}
-
 int amdgpu_gmc_vram_checking(struct amdgpu_device *adev)
 {
 	struct amdgpu_bo *vram_bo = NULL;
@@ -846,19 +1148,535 @@ int amdgpu_gmc_vram_checking(struct amdgpu_device *adev)
 	 * seconds, so here, we just pick up three parts for emulation.
 	 */
 	ret = memcmp(vram_ptr, cptr, 10);
-	if (ret)
-		return ret;
+	if (ret) {
+		ret = -EIO;
+		goto release_buffer;
+	}
 
 	ret = memcmp(vram_ptr + (size / 2), cptr, 10);
-	if (ret)
-		return ret;
+	if (ret) {
+		ret = -EIO;
+		goto release_buffer;
+	}
 
 	ret = memcmp(vram_ptr + size - 10, cptr, 10);
-	if (ret)
-		return ret;
+	if (ret) {
+		ret = -EIO;
+		goto release_buffer;
+	}
 
+release_buffer:
 	amdgpu_bo_free_kernel(&vram_bo, &vram_gpu,
 			&vram_ptr);
 
+	return ret;
+}
+
+static const char *nps_desc[] = {
+	[AMDGPU_NPS1_PARTITION_MODE] = "NPS1",
+	[AMDGPU_NPS2_PARTITION_MODE] = "NPS2",
+	[AMDGPU_NPS3_PARTITION_MODE] = "NPS3",
+	[AMDGPU_NPS4_PARTITION_MODE] = "NPS4",
+	[AMDGPU_NPS6_PARTITION_MODE] = "NPS6",
+	[AMDGPU_NPS8_PARTITION_MODE] = "NPS8",
+};
+
+static ssize_t available_memory_partition_show(struct device *dev,
+					       struct device_attribute *addr,
+					       char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	int size = 0, mode;
+	char *sep = "";
+
+	for_each_inst(mode, adev->gmc.supported_nps_modes) {
+		size += sysfs_emit_at(buf, size, "%s%s", sep, nps_desc[mode]);
+		sep = ", ";
+	}
+	size += sysfs_emit_at(buf, size, "\n");
+
+	return size;
+}
+
+static ssize_t current_memory_partition_store(struct device *dev,
+					      struct device_attribute *attr,
+					      const char *buf, size_t count)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	enum amdgpu_memory_partition mode;
+	struct amdgpu_hive_info *hive;
+	int i;
+
+	mode = UNKNOWN_MEMORY_PARTITION_MODE;
+	for_each_inst(i, adev->gmc.supported_nps_modes) {
+		if (!strncasecmp(nps_desc[i], buf, strlen(nps_desc[i]))) {
+			mode = i;
+			break;
+		}
+	}
+
+	if (mode == UNKNOWN_MEMORY_PARTITION_MODE)
+		return -EINVAL;
+
+	if (mode == adev->gmc.gmc_funcs->query_mem_partition_mode(adev)) {
+		dev_info(
+			adev->dev,
+			"requested NPS mode is same as current NPS mode, skipping\n");
+		return count;
+	}
+
+	/* If device is part of hive, all devices in the hive should request the
+	 * same mode. Hence store the requested mode in hive.
+	 */
+	hive = amdgpu_get_xgmi_hive(adev);
+	if (hive) {
+		atomic_set(&hive->requested_nps_mode, mode);
+		amdgpu_put_xgmi_hive(hive);
+	} else {
+		adev->gmc.requested_nps_mode = mode;
+	}
+
+	dev_info(
+		adev->dev,
+		"NPS mode change requested, please remove and reload the driver\n");
+
+	return count;
+}
+
+static ssize_t current_memory_partition_show(
+	struct device *dev, struct device_attribute *addr, char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	enum amdgpu_memory_partition mode;
+
+	/* Only minimal precaution taken to reject requests while in reset */
+	if (amdgpu_in_reset(adev))
+		return -EPERM;
+
+	mode = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+	if ((mode >= ARRAY_SIZE(nps_desc)) ||
+	    (BIT(mode) & AMDGPU_ALL_NPS_MASK) != BIT(mode))
+		return sysfs_emit(buf, "UNKNOWN\n");
+
+	return sysfs_emit(buf, "%s\n", nps_desc[mode]);
+}
+
+static DEVICE_ATTR_RW(current_memory_partition);
+static DEVICE_ATTR_RO(available_memory_partition);
+
+int amdgpu_gmc_sysfs_init(struct amdgpu_device *adev)
+{
+	bool nps_switch_support;
+	int r = 0;
+
+	if (!adev->gmc.gmc_funcs->query_mem_partition_mode)
+		return 0;
+
+	nps_switch_support = (hweight32(adev->gmc.supported_nps_modes &
+					AMDGPU_ALL_NPS_MASK) > 1);
+	if (!nps_switch_support)
+		dev_attr_current_memory_partition.attr.mode &=
+			~(S_IWUSR | S_IWGRP | S_IWOTH);
+	else
+		r = device_create_file(adev->dev,
+				       &dev_attr_available_memory_partition);
+
+	if (r)
+		return r;
+
+	return device_create_file(adev->dev,
+				  &dev_attr_current_memory_partition);
+}
+
+void amdgpu_gmc_sysfs_fini(struct amdgpu_device *adev)
+{
+	if (!adev->gmc.gmc_funcs->query_mem_partition_mode)
+		return;
+
+	device_remove_file(adev->dev, &dev_attr_current_memory_partition);
+	device_remove_file(adev->dev, &dev_attr_available_memory_partition);
+}
+
+int amdgpu_gmc_get_nps_memranges(struct amdgpu_device *adev,
+				 struct amdgpu_mem_partition_info *mem_ranges,
+				 uint8_t *exp_ranges)
+{
+	struct amdgpu_gmc_memrange *ranges;
+	int range_cnt, ret, i, j;
+	uint32_t nps_type;
+	bool refresh;
+
+	if (!mem_ranges || !exp_ranges)
+		return -EINVAL;
+
+	refresh = (adev->init_lvl->level != AMDGPU_INIT_LEVEL_MINIMAL_XGMI) &&
+		  (adev->gmc.reset_flags & AMDGPU_GMC_INIT_RESET_NPS);
+	ret = amdgpu_discovery_get_nps_info(adev, &nps_type, &ranges,
+					    &range_cnt, refresh);
+
+	if (ret)
+		return ret;
+
+	/* TODO: For now, expect ranges and partition count to be the same.
+	 * Adjust if there are holes expected in any NPS domain.
+	 */
+	if (*exp_ranges && (range_cnt != *exp_ranges)) {
+		dev_warn(
+			adev->dev,
+			"NPS config mismatch - expected ranges: %d discovery - nps mode: %d, nps ranges: %d",
+			*exp_ranges, nps_type, range_cnt);
+		ret = -EINVAL;
+		goto err;
+	}
+
+	for (i = 0; i < range_cnt; ++i) {
+		if (ranges[i].base_address >= ranges[i].limit_address) {
+			dev_warn(
+				adev->dev,
+				"Invalid NPS range - nps mode: %d, range[%d]: base: %llx limit: %llx",
+				nps_type, i, ranges[i].base_address,
+				ranges[i].limit_address);
+			ret = -EINVAL;
+			goto err;
+		}
+
+		/* Check for overlaps, not expecting any now */
+		for (j = i - 1; j >= 0; j--) {
+			if (max(ranges[j].base_address,
+				ranges[i].base_address) <=
+			    min(ranges[j].limit_address,
+				ranges[i].limit_address)) {
+				dev_warn(
+					adev->dev,
+					"overlapping ranges detected [ %llx - %llx ] | [%llx - %llx]",
+					ranges[j].base_address,
+					ranges[j].limit_address,
+					ranges[i].base_address,
+					ranges[i].limit_address);
+				ret = -EINVAL;
+				goto err;
+			}
+		}
+
+		mem_ranges[i].range.fpfn =
+			(ranges[i].base_address -
+			 adev->vm_manager.vram_base_offset) >>
+			AMDGPU_GPU_PAGE_SHIFT;
+		mem_ranges[i].range.lpfn =
+			(ranges[i].limit_address -
+			 adev->vm_manager.vram_base_offset) >>
+			AMDGPU_GPU_PAGE_SHIFT;
+		mem_ranges[i].size =
+			ranges[i].limit_address - ranges[i].base_address + 1;
+	}
+
+	if (!*exp_ranges)
+		*exp_ranges = range_cnt;
+err:
+	kfree(ranges);
+
+	return ret;
+}
+
+int amdgpu_gmc_request_memory_partition(struct amdgpu_device *adev,
+					int nps_mode)
+{
+	/* Not supported on VF devices and APUs */
+	if (amdgpu_sriov_vf(adev) || (adev->flags & AMD_IS_APU))
+		return -EOPNOTSUPP;
+
+	if (!adev->psp.funcs) {
+		dev_err(adev->dev,
+			"PSP interface not available for nps mode change request");
+		return -EINVAL;
+	}
+
+	return psp_memory_partition(&adev->psp, nps_mode);
+}
+
+static inline bool amdgpu_gmc_need_nps_switch_req(struct amdgpu_device *adev,
+						  int req_nps_mode,
+						  int cur_nps_mode)
+{
+	return (((BIT(req_nps_mode) & adev->gmc.supported_nps_modes) ==
+			BIT(req_nps_mode)) &&
+		req_nps_mode != cur_nps_mode);
+}
+
+void amdgpu_gmc_prepare_nps_mode_change(struct amdgpu_device *adev)
+{
+	int req_nps_mode, cur_nps_mode, r;
+	struct amdgpu_hive_info *hive;
+
+	if (amdgpu_sriov_vf(adev) || !adev->gmc.supported_nps_modes ||
+	    !adev->gmc.gmc_funcs->request_mem_partition_mode)
+		return;
+
+	cur_nps_mode = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+	hive = amdgpu_get_xgmi_hive(adev);
+	if (hive) {
+		req_nps_mode = atomic_read(&hive->requested_nps_mode);
+		if (!amdgpu_gmc_need_nps_switch_req(adev, req_nps_mode,
+						    cur_nps_mode)) {
+			amdgpu_put_xgmi_hive(hive);
+			return;
+		}
+		r = amdgpu_xgmi_request_nps_change(adev, hive, req_nps_mode);
+		amdgpu_put_xgmi_hive(hive);
+		goto out;
+	}
+
+	req_nps_mode = adev->gmc.requested_nps_mode;
+	if (!amdgpu_gmc_need_nps_switch_req(adev, req_nps_mode, cur_nps_mode))
+		return;
+
+	/* even if this fails, we should let driver unload w/o blocking */
+	r = adev->gmc.gmc_funcs->request_mem_partition_mode(adev, req_nps_mode);
+out:
+	if (r)
+		dev_err(adev->dev, "NPS mode change request failed\n");
+	else
+		dev_info(
+			adev->dev,
+			"NPS mode change request done, reload driver to complete the change\n");
+}
+
+bool amdgpu_gmc_need_reset_on_init(struct amdgpu_device *adev)
+{
+	if (adev->gmc.gmc_funcs->need_reset_on_init)
+		return adev->gmc.gmc_funcs->need_reset_on_init(adev);
+
+	return false;
+}
+
+enum amdgpu_memory_partition
+amdgpu_gmc_get_vf_memory_partition(struct amdgpu_device *adev)
+{
+	switch (adev->gmc.num_mem_partitions) {
+	case 0:
+		return UNKNOWN_MEMORY_PARTITION_MODE;
+	case 1:
+		return AMDGPU_NPS1_PARTITION_MODE;
+	case 2:
+		return AMDGPU_NPS2_PARTITION_MODE;
+	case 4:
+		return AMDGPU_NPS4_PARTITION_MODE;
+	case 8:
+		return AMDGPU_NPS8_PARTITION_MODE;
+	default:
+		return AMDGPU_NPS1_PARTITION_MODE;
+	}
+}
+
+enum amdgpu_memory_partition
+amdgpu_gmc_get_memory_partition(struct amdgpu_device *adev, u32 *supp_modes)
+{
+	enum amdgpu_memory_partition mode = UNKNOWN_MEMORY_PARTITION_MODE;
+
+	if (adev->nbio.funcs &&
+	    adev->nbio.funcs->get_memory_partition_mode)
+		mode = adev->nbio.funcs->get_memory_partition_mode(adev,
+								   supp_modes);
+	else
+		dev_warn(adev->dev, "memory partition mode query is not supported\n");
+
+	return mode;
+}
+
+enum amdgpu_memory_partition
+amdgpu_gmc_query_memory_partition(struct amdgpu_device *adev)
+{
+	if (amdgpu_sriov_vf(adev))
+		return amdgpu_gmc_get_vf_memory_partition(adev);
+	else
+		return amdgpu_gmc_get_memory_partition(adev, NULL);
+}
+
+static bool amdgpu_gmc_validate_partition_info(struct amdgpu_device *adev)
+{
+	enum amdgpu_memory_partition mode;
+	u32 supp_modes;
+	bool valid;
+
+	mode = amdgpu_gmc_get_memory_partition(adev, &supp_modes);
+
+	/* Mode detected by hardware not present in supported modes */
+	if ((mode != UNKNOWN_MEMORY_PARTITION_MODE) &&
+	    !(BIT(mode - 1) & supp_modes))
+		return false;
+
+	switch (mode) {
+	case UNKNOWN_MEMORY_PARTITION_MODE:
+	case AMDGPU_NPS1_PARTITION_MODE:
+		valid = (adev->gmc.num_mem_partitions == 1);
+		break;
+	case AMDGPU_NPS2_PARTITION_MODE:
+		valid = (adev->gmc.num_mem_partitions == 2);
+		break;
+	case AMDGPU_NPS4_PARTITION_MODE:
+		valid = (adev->gmc.num_mem_partitions == 3 ||
+			 adev->gmc.num_mem_partitions == 4);
+		break;
+	case AMDGPU_NPS8_PARTITION_MODE:
+		valid = (adev->gmc.num_mem_partitions == 8);
+		break;
+	default:
+		valid = false;
+	}
+
+	return valid;
+}
+
+static bool amdgpu_gmc_is_node_present(int *node_ids, int num_ids, int nid)
+{
+	int i;
+
+	/* Check if node with id 'nid' is present in 'node_ids' array */
+	for (i = 0; i < num_ids; ++i)
+		if (node_ids[i] == nid)
+			return true;
+
+	return false;
+}
+
+static void
+amdgpu_gmc_init_acpi_mem_ranges(struct amdgpu_device *adev,
+				struct amdgpu_mem_partition_info *mem_ranges)
+{
+	struct amdgpu_numa_info numa_info;
+	int node_ids[AMDGPU_MAX_MEM_RANGES];
+	int num_ranges = 0, ret;
+	int num_xcc, xcc_id;
+	uint32_t xcc_mask;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	xcc_mask = (1U << num_xcc) - 1;
+
+	for_each_inst(xcc_id, xcc_mask)	{
+		ret = amdgpu_acpi_get_mem_info(adev, xcc_id, &numa_info);
+		if (ret)
+			continue;
+
+		if (numa_info.nid == NUMA_NO_NODE) {
+			mem_ranges[0].size = numa_info.size;
+			mem_ranges[0].numa.node = numa_info.nid;
+			num_ranges = 1;
+			break;
+		}
+
+		if (amdgpu_gmc_is_node_present(node_ids, num_ranges,
+					     numa_info.nid))
+			continue;
+
+		node_ids[num_ranges] = numa_info.nid;
+		mem_ranges[num_ranges].numa.node = numa_info.nid;
+		mem_ranges[num_ranges].size = numa_info.size;
+		++num_ranges;
+	}
+
+	adev->gmc.num_mem_partitions = num_ranges;
+}
+
+void amdgpu_gmc_init_sw_mem_ranges(struct amdgpu_device *adev,
+				   struct amdgpu_mem_partition_info *mem_ranges)
+{
+	enum amdgpu_memory_partition mode;
+	u32 start_addr = 0, size;
+	int i, r, l;
+
+	mode = amdgpu_gmc_query_memory_partition(adev);
+
+	switch (mode) {
+	case UNKNOWN_MEMORY_PARTITION_MODE:
+		adev->gmc.num_mem_partitions = 0;
+		break;
+	case AMDGPU_NPS1_PARTITION_MODE:
+		adev->gmc.num_mem_partitions = 1;
+		break;
+	case AMDGPU_NPS2_PARTITION_MODE:
+		adev->gmc.num_mem_partitions = 2;
+		break;
+	case AMDGPU_NPS4_PARTITION_MODE:
+		if (adev->flags & AMD_IS_APU)
+			adev->gmc.num_mem_partitions = 3;
+		else
+			adev->gmc.num_mem_partitions = 4;
+		break;
+	case AMDGPU_NPS8_PARTITION_MODE:
+		adev->gmc.num_mem_partitions = 8;
+		break;
+	default:
+		adev->gmc.num_mem_partitions = 1;
+		break;
+	}
+
+	/* Use NPS range info, if populated */
+	r = amdgpu_gmc_get_nps_memranges(adev, mem_ranges,
+					 &adev->gmc.num_mem_partitions);
+	if (!r) {
+		l = 0;
+		for (i = 1; i < adev->gmc.num_mem_partitions; ++i) {
+			if (mem_ranges[i].range.lpfn >
+			    mem_ranges[i - 1].range.lpfn)
+				l = i;
+		}
+
+	} else {
+		if (!adev->gmc.num_mem_partitions) {
+			dev_warn(adev->dev,
+				 "Not able to detect NPS mode, fall back to NPS1\n");
+			adev->gmc.num_mem_partitions = 1;
+		}
+		/* Fallback to sw based calculation */
+		size = (adev->gmc.real_vram_size + SZ_16M) >> AMDGPU_GPU_PAGE_SHIFT;
+		size /= adev->gmc.num_mem_partitions;
+
+		for (i = 0; i < adev->gmc.num_mem_partitions; ++i) {
+			mem_ranges[i].range.fpfn = start_addr;
+			mem_ranges[i].size =
+				((u64)size << AMDGPU_GPU_PAGE_SHIFT);
+			mem_ranges[i].range.lpfn = start_addr + size - 1;
+			start_addr += size;
+		}
+
+		l = adev->gmc.num_mem_partitions - 1;
+	}
+
+	/* Adjust the last one */
+	mem_ranges[l].range.lpfn =
+		(adev->gmc.real_vram_size >> AMDGPU_GPU_PAGE_SHIFT) - 1;
+	mem_ranges[l].size =
+		adev->gmc.real_vram_size -
+		((u64)mem_ranges[l].range.fpfn << AMDGPU_GPU_PAGE_SHIFT);
+}
+
+int amdgpu_gmc_init_mem_ranges(struct amdgpu_device *adev)
+{
+	bool valid;
+
+	adev->gmc.mem_partitions = kcalloc(AMDGPU_MAX_MEM_RANGES,
+					   sizeof(struct amdgpu_mem_partition_info),
+					   GFP_KERNEL);
+	if (!adev->gmc.mem_partitions)
+		return -ENOMEM;
+
+	if (adev->gmc.is_app_apu)
+		amdgpu_gmc_init_acpi_mem_ranges(adev, adev->gmc.mem_partitions);
+	else
+		amdgpu_gmc_init_sw_mem_ranges(adev, adev->gmc.mem_partitions);
+
+	if (amdgpu_sriov_vf(adev))
+		valid = true;
+	else
+		valid = amdgpu_gmc_validate_partition_info(adev);
+	if (!valid) {
+		/* TODO: handle invalid case */
+		dev_warn(adev->dev,
+			 "Mem ranges not matching with hardware config\n");
+	}
+
 	return 0;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h
index 232523e3e270..727342689d4b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h
@@ -29,6 +29,7 @@
 #include <linux/types.h>
 
 #include "amdgpu_irq.h"
+#include "amdgpu_xgmi.h"
 #include "amdgpu_ras.h"
 
 /* VA hole for 48bit addresses on Vega10 */
@@ -61,8 +62,35 @@
  */
 #define AMDGPU_GMC_FAULT_TIMEOUT	5000ULL
 
+/* XNACK flags */
+#define AMDGPU_GMC_XNACK_FLAG_CHAIN BIT(0)
+
 struct firmware;
 
+enum amdgpu_memory_partition {
+	UNKNOWN_MEMORY_PARTITION_MODE = 0,
+	AMDGPU_NPS1_PARTITION_MODE = 1,
+	AMDGPU_NPS2_PARTITION_MODE = 2,
+	AMDGPU_NPS3_PARTITION_MODE = 3,
+	AMDGPU_NPS4_PARTITION_MODE = 4,
+	AMDGPU_NPS6_PARTITION_MODE = 6,
+	AMDGPU_NPS8_PARTITION_MODE = 8,
+};
+
+#define AMDGPU_ALL_NPS_MASK                                                  \
+	(BIT(AMDGPU_NPS1_PARTITION_MODE) | BIT(AMDGPU_NPS2_PARTITION_MODE) | \
+	 BIT(AMDGPU_NPS3_PARTITION_MODE) | BIT(AMDGPU_NPS4_PARTITION_MODE) | \
+	 BIT(AMDGPU_NPS6_PARTITION_MODE) | BIT(AMDGPU_NPS8_PARTITION_MODE))
+
+#define AMDGPU_GMC_INIT_RESET_NPS  BIT(0)
+
+#define AMDGPU_MAX_MEM_RANGES 8
+
+#define AMDGPU_GMC9_FAULT_SOURCE_DATA_RETRY 0x80
+#define AMDGPU_GMC9_FAULT_SOURCE_DATA_READ  0x40
+#define AMDGPU_GMC9_FAULT_SOURCE_DATA_WRITE 0x20
+#define AMDGPU_GMC9_FAULT_SOURCE_DATA_EXE   0x10
+
 /*
  * GMC page fault information
  */
@@ -70,6 +98,7 @@ struct amdgpu_gmc_fault {
 	uint64_t	timestamp:48;
 	uint64_t	next:AMDGPU_GMC_FAULT_RING_ORDER;
 	atomic64_t	key;
+	uint64_t	timestamp_expiry:48;
 };
 
 /*
@@ -106,6 +135,8 @@ struct amdgpu_vmhub {
 
 	uint32_t	vm_contexts_disable;
 
+	bool		sdma_invalidation_workaround;
+
 	const struct amdgpu_vmhub_funcs *vmhub_funcs;
 };
 
@@ -117,8 +148,9 @@ struct amdgpu_gmc_funcs {
 	void (*flush_gpu_tlb)(struct amdgpu_device *adev, uint32_t vmid,
 				uint32_t vmhub, uint32_t flush_type);
 	/* flush the vm tlb via pasid */
-	int (*flush_gpu_tlb_pasid)(struct amdgpu_device *adev, uint16_t pasid,
-					uint32_t flush_type, bool all_hub);
+	void (*flush_gpu_tlb_pasid)(struct amdgpu_device *adev, uint16_t pasid,
+				    uint32_t flush_type, bool all_hub,
+				    uint32_t inst);
 	/* flush the vm tlb via ring */
 	uint64_t (*emit_flush_gpu_tlb)(struct amdgpu_ring *ring, unsigned vmid,
 				       uint64_t pd_addr);
@@ -127,40 +159,58 @@ struct amdgpu_gmc_funcs {
 				   unsigned pasid);
 	/* enable/disable PRT support */
 	void (*set_prt)(struct amdgpu_device *adev, bool enable);
-	/* map mtype to hardware flags */
-	uint64_t (*map_mtype)(struct amdgpu_device *adev, uint32_t flags);
 	/* get the pde for a given mc addr */
 	void (*get_vm_pde)(struct amdgpu_device *adev, int level,
 			   u64 *dst, u64 *flags);
-	/* get the pte flags to use for a BO VA mapping */
+	/* get the pte flags to use for PTEs */
 	void (*get_vm_pte)(struct amdgpu_device *adev,
-			   struct amdgpu_bo_va_mapping *mapping,
-			   uint64_t *flags);
+			   struct amdgpu_vm *vm,
+			   struct amdgpu_bo *bo,
+			   uint32_t vm_flags,
+			   uint64_t *pte_flags);
+	/* override per-page pte flags */
+	void (*override_vm_pte_flags)(struct amdgpu_device *dev,
+				      struct amdgpu_vm *vm,
+				      uint64_t addr, uint64_t *flags);
 	/* get the amount of memory used by the vbios for pre-OS console */
 	unsigned int (*get_vbios_fb_size)(struct amdgpu_device *adev);
+	/* get the DCC buffer alignment */
+	unsigned int (*get_dcc_alignment)(struct amdgpu_device *adev);
+
+	enum amdgpu_memory_partition (*query_mem_partition_mode)(
+		struct amdgpu_device *adev);
+	/* Request NPS mode */
+	int (*request_mem_partition_mode)(struct amdgpu_device *adev,
+					  int nps_mode);
+	bool (*need_reset_on_init)(struct amdgpu_device *adev);
+};
+
+struct amdgpu_mem_partition_info {
+	union {
+		struct {
+			uint32_t fpfn;
+			uint32_t lpfn;
+		} range;
+		struct {
+			int node;
+		} numa;
+	};
+	uint64_t size;
 };
 
-struct amdgpu_xgmi_ras {
-	struct amdgpu_ras_block_object ras_block;
+#define INVALID_PFN    -1
+
+struct amdgpu_gmc_memrange {
+	uint64_t base_address;
+	uint64_t limit_address;
+	uint32_t flags;
+	int nid_mask;
 };
 
-struct amdgpu_xgmi {
-	/* from psp */
-	u64 node_id;
-	u64 hive_id;
-	/* fixed per family */
-	u64 node_segment_size;
-	/* physical node (0-3) */
-	unsigned physical_node_id;
-	/* number of nodes (0-4) */
-	unsigned num_physical_nodes;
-	/* gpu list in the same hive */
-	struct list_head head;
-	bool supported;
-	struct ras_common_if *ras_if;
-	bool connected_to_cpu;
-	bool pending_reset;
-	struct amdgpu_xgmi_ras *ras;
+enum amdgpu_gart_placement {
+	AMDGPU_GART_PLACEMENT_BEST_FIT = 0,
+	AMDGPU_GART_PLACEMENT_HIGH,
+	AMDGPU_GART_PLACEMENT_LOW,
 };
 
 struct amdgpu_gmc {
@@ -249,12 +299,19 @@ struct amdgpu_gmc {
 	uint64_t		last_fault:AMDGPU_GMC_FAULT_RING_ORDER;
 
 	bool tmz_enabled;
+	bool is_app_apu;
 
+	struct amdgpu_mem_partition_info *mem_partitions;
+	uint8_t num_mem_partitions;
 	const struct amdgpu_gmc_funcs	*gmc_funcs;
+	enum amdgpu_memory_partition	requested_nps_mode;
+	uint32_t supported_nps_modes;
+	uint32_t reset_flags;
 
 	struct amdgpu_xgmi xgmi;
 	struct amdgpu_irq_src	ecc_irq;
 	int noretry;
+	uint32_t xnack_flags;
 
 	uint32_t	vmid0_page_table_block_size;
 	uint32_t	vmid0_page_table_depth;
@@ -264,6 +321,8 @@ struct amdgpu_gmc {
 
 	/* MALL size */
 	u64 mall_size;
+	uint32_t m_half_use;
+
 	/* number of UMC instances */
 	int num_umc;
 	/* mode2 save restore */
@@ -292,18 +351,28 @@ struct amdgpu_gmc {
 	u64 VM_CONTEXT_PAGE_TABLE_END_ADDR_LO32[16];
 	u64 VM_CONTEXT_PAGE_TABLE_END_ADDR_HI32[16];
 	u64 MC_VM_MX_L1_TLB_CNTL;
+
+	u64 noretry_flags;
+
+	bool flush_tlb_needs_extra_type_0;
+	bool flush_tlb_needs_extra_type_2;
+	bool flush_pasid_uses_kiq;
 };
 
-#define amdgpu_gmc_flush_gpu_tlb(adev, vmid, vmhub, type) ((adev)->gmc.gmc_funcs->flush_gpu_tlb((adev), (vmid), (vmhub), (type)))
-#define amdgpu_gmc_flush_gpu_tlb_pasid(adev, pasid, type, allhub) \
-	((adev)->gmc.gmc_funcs->flush_gpu_tlb_pasid \
-	((adev), (pasid), (type), (allhub)))
 #define amdgpu_gmc_emit_flush_gpu_tlb(r, vmid, addr) (r)->adev->gmc.gmc_funcs->emit_flush_gpu_tlb((r), (vmid), (addr))
 #define amdgpu_gmc_emit_pasid_mapping(r, vmid, pasid) (r)->adev->gmc.gmc_funcs->emit_pasid_mapping((r), (vmid), (pasid))
-#define amdgpu_gmc_map_mtype(adev, flags) (adev)->gmc.gmc_funcs->map_mtype((adev),(flags))
 #define amdgpu_gmc_get_vm_pde(adev, level, dst, flags) (adev)->gmc.gmc_funcs->get_vm_pde((adev), (level), (dst), (flags))
-#define amdgpu_gmc_get_vm_pte(adev, mapping, flags) (adev)->gmc.gmc_funcs->get_vm_pte((adev), (mapping), (flags))
+#define amdgpu_gmc_get_vm_pte(adev, vm, bo, vm_flags, pte_flags) \
+	((adev)->gmc.gmc_funcs->get_vm_pte((adev), (vm), (bo), (vm_flags), \
+					   (pte_flags)))
+#define amdgpu_gmc_override_vm_pte_flags(adev, vm, addr, pte_flags)	\
+	(adev)->gmc.gmc_funcs->override_vm_pte_flags			\
+		((adev), (vm), (addr), (pte_flags))
 #define amdgpu_gmc_get_vbios_fb_size(adev) (adev)->gmc.gmc_funcs->get_vbios_fb_size((adev))
+#define amdgpu_gmc_get_dcc_alignment(adev) ({			\
+	typeof(adev) _adev = (adev);				\
+	_adev->gmc.gmc_funcs->get_dcc_alignment(_adev);		\
+})
 
 /**
  * amdgpu_gmc_vram_full_visible - Check if full VRAM is visible through the BAR
@@ -333,6 +402,7 @@ static inline uint64_t amdgpu_gmc_sign_extend(uint64_t addr)
 	return addr;
 }
 
+bool amdgpu_gmc_is_pdb0_enabled(struct amdgpu_device *adev);
 int amdgpu_gmc_pdb0_alloc(struct amdgpu_device *adev);
 void amdgpu_gmc_get_pde_for_bo(struct amdgpu_bo *bo, int level,
 			       uint64_t *addr, uint64_t *flags);
@@ -345,9 +415,12 @@ void amdgpu_gmc_sysvm_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc
 void amdgpu_gmc_vram_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc,
 			      u64 base);
 void amdgpu_gmc_gart_location(struct amdgpu_device *adev,
-			      struct amdgpu_gmc *mc);
+			      struct amdgpu_gmc *mc,
+			      enum amdgpu_gart_placement gart_placement);
 void amdgpu_gmc_agp_location(struct amdgpu_device *adev,
 			     struct amdgpu_gmc *mc);
+void amdgpu_gmc_set_agp_default(struct amdgpu_device *adev,
+				struct amdgpu_gmc *mc);
 bool amdgpu_gmc_filter_faults(struct amdgpu_device *adev,
 			      struct amdgpu_ih_ring *ih, uint64_t addr,
 			      uint16_t pasid, uint64_t timestamp);
@@ -357,6 +430,15 @@ int amdgpu_gmc_ras_sw_init(struct amdgpu_device *adev);
 int amdgpu_gmc_ras_late_init(struct amdgpu_device *adev);
 void amdgpu_gmc_ras_fini(struct amdgpu_device *adev);
 int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev);
+void amdgpu_gmc_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
+			      uint32_t vmhub, uint32_t flush_type);
+int amdgpu_gmc_flush_gpu_tlb_pasid(struct amdgpu_device *adev, uint16_t pasid,
+				   uint32_t flush_type, bool all_hub,
+				   uint32_t inst);
+void amdgpu_gmc_fw_reg_write_reg_wait(struct amdgpu_device *adev,
+				      uint32_t reg0, uint32_t reg1,
+				      uint32_t ref, uint32_t mask,
+				      uint32_t xcc_inst);
 
 extern void amdgpu_gmc_tmz_set(struct amdgpu_device *adev);
 extern void amdgpu_gmc_noretry_set(struct amdgpu_device *adev);
@@ -370,6 +452,25 @@ void amdgpu_gmc_get_vbios_allocations(struct amdgpu_device *adev);
 void amdgpu_gmc_init_pdb0(struct amdgpu_device *adev);
 uint64_t amdgpu_gmc_vram_mc2pa(struct amdgpu_device *adev, uint64_t mc_addr);
 uint64_t amdgpu_gmc_vram_pa(struct amdgpu_device *adev, struct amdgpu_bo *bo);
-uint64_t amdgpu_gmc_vram_cpu_pa(struct amdgpu_device *adev, struct amdgpu_bo *bo);
 int amdgpu_gmc_vram_checking(struct amdgpu_device *adev);
+int amdgpu_gmc_sysfs_init(struct amdgpu_device *adev);
+void amdgpu_gmc_sysfs_fini(struct amdgpu_device *adev);
+
+int amdgpu_gmc_get_nps_memranges(struct amdgpu_device *adev,
+				 struct amdgpu_mem_partition_info *mem_ranges,
+				 uint8_t *exp_ranges);
+
+int amdgpu_gmc_request_memory_partition(struct amdgpu_device *adev,
+					int nps_mode);
+void amdgpu_gmc_prepare_nps_mode_change(struct amdgpu_device *adev);
+bool amdgpu_gmc_need_reset_on_init(struct amdgpu_device *adev);
+enum amdgpu_memory_partition
+amdgpu_gmc_get_vf_memory_partition(struct amdgpu_device *adev);
+enum amdgpu_memory_partition
+amdgpu_gmc_get_memory_partition(struct amdgpu_device *adev, u32 *supp_modes);
+enum amdgpu_memory_partition
+amdgpu_gmc_query_memory_partition(struct amdgpu_device *adev);
+int amdgpu_gmc_init_mem_ranges(struct amdgpu_device *adev);
+void amdgpu_gmc_init_sw_mem_ranges(struct amdgpu_device *adev,
+				   struct amdgpu_mem_partition_info *mem_ranges);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c
index 44367f03316f..895c1e4c6747 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c
@@ -200,8 +200,6 @@ void amdgpu_gtt_mgr_recover(struct amdgpu_gtt_mgr *mgr)
 		amdgpu_ttm_recover_gart(node->base.bo);
 	}
 	spin_unlock(&mgr->lock);
-
-	amdgpu_gart_invalidate_tlb(adev);
 }
 
 /**
@@ -286,6 +284,7 @@ int amdgpu_gtt_mgr_init(struct amdgpu_device *adev, uint64_t gtt_size)
 	ttm_resource_manager_init(man, &adev->mman.bdev, gtt_size);
 
 	start = AMDGPU_GTT_MAX_TRANSFER_SIZE * AMDGPU_GTT_NUM_TRANSFER_WINDOWS;
+	start += amdgpu_vce_required_gart_pages(adev);
 	size = (adev->gmc.gart_size >> PAGE_SHIFT) - start;
 	drm_mm_init(&mgr->mm, start, size);
 	spin_lock_init(&mgr->lock);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c
index b6cf801939aa..5a60d69a3e1f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.c
@@ -22,6 +22,7 @@
  */
 #include "amdgpu.h"
 #include "amdgpu_ras.h"
+#include <uapi/linux/kfd_ioctl.h>
 
 int amdgpu_hdp_ras_sw_init(struct amdgpu_device *adev)
 {
@@ -46,3 +47,38 @@ int amdgpu_hdp_ras_sw_init(struct amdgpu_device *adev)
 	/* hdp ras follows amdgpu_ras_block_late_init_default for late init */
 	return 0;
 }
+
+void amdgpu_hdp_generic_flush(struct amdgpu_device *adev,
+			      struct amdgpu_ring *ring)
+{
+	if (!ring || !ring->funcs->emit_wreg) {
+		WREG32((adev->rmmio_remap.reg_offset +
+			KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >>
+			       2,
+		       0);
+		if (adev->nbio.funcs->get_memsize)
+			adev->nbio.funcs->get_memsize(adev);
+	} else {
+		amdgpu_ring_emit_wreg(ring,
+				      (adev->rmmio_remap.reg_offset +
+				       KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >>
+					      2,
+				      0);
+	}
+}
+
+void amdgpu_hdp_invalidate(struct amdgpu_device *adev, struct amdgpu_ring *ring)
+{
+	if (adev->asic_funcs && adev->asic_funcs->invalidate_hdp)
+		adev->asic_funcs->invalidate_hdp(adev, ring);
+	else if (adev->hdp.funcs && adev->hdp.funcs->invalidate_hdp)
+		adev->hdp.funcs->invalidate_hdp(adev, ring);
+}
+
+void amdgpu_hdp_flush(struct amdgpu_device *adev, struct amdgpu_ring *ring)
+{
+	if (adev->asic_funcs && adev->asic_funcs->flush_hdp)
+		adev->asic_funcs->flush_hdp(adev, ring);
+	else if (adev->hdp.funcs && adev->hdp.funcs->flush_hdp)
+		adev->hdp.funcs->flush_hdp(adev, ring);
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h
index 7b8a6152dc8d..d9f488fa76b9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hdp.h
@@ -44,4 +44,10 @@ struct amdgpu_hdp {
 };
 
 int amdgpu_hdp_ras_sw_init(struct amdgpu_device *adev);
+void amdgpu_hdp_generic_flush(struct amdgpu_device *adev,
+			      struct amdgpu_ring *ring);
+void amdgpu_hdp_invalidate(struct amdgpu_device *adev,
+			   struct amdgpu_ring *ring);
+void amdgpu_hdp_flush(struct amdgpu_device *adev,
+		      struct amdgpu_ring *ring);
 #endif /* __AMDGPU_HDP_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c
index 2dadcfe43d03..90d26d820bac 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c
@@ -129,13 +129,25 @@ static const struct mmu_interval_notifier_ops amdgpu_hmm_hsa_ops = {
  */
 int amdgpu_hmm_register(struct amdgpu_bo *bo, unsigned long addr)
 {
+	int r;
+
 	if (bo->kfd_bo)
-		return mmu_interval_notifier_insert(&bo->notifier, current->mm,
+		r = mmu_interval_notifier_insert(&bo->notifier, current->mm,
 						    addr, amdgpu_bo_size(bo),
 						    &amdgpu_hmm_hsa_ops);
-	return mmu_interval_notifier_insert(&bo->notifier, current->mm, addr,
-					    amdgpu_bo_size(bo),
-					    &amdgpu_hmm_gfx_ops);
+	else
+		r = mmu_interval_notifier_insert(&bo->notifier, current->mm, addr,
+							amdgpu_bo_size(bo),
+							&amdgpu_hmm_gfx_ops);
+	if (r)
+		/*
+		 * Make sure amdgpu_hmm_unregister() doesn't call
+		 * mmu_interval_notifier_remove() when the notifier isn't properly
+		 * initialized.
+		 */
+		bo->notifier.mm = NULL;
+
+	return r;
 }
 
 /**
@@ -155,19 +167,14 @@ void amdgpu_hmm_unregister(struct amdgpu_bo *bo)
 
 int amdgpu_hmm_range_get_pages(struct mmu_interval_notifier *notifier,
 			       uint64_t start, uint64_t npages, bool readonly,
-			       void *owner, struct page **pages,
-			       struct hmm_range **phmm_range)
+			       void *owner,
+			       struct amdgpu_hmm_range *range)
 {
-	struct hmm_range *hmm_range;
 	unsigned long end;
 	unsigned long timeout;
-	unsigned long i;
 	unsigned long *pfns;
 	int r = 0;
-
-	hmm_range = kzalloc(sizeof(*hmm_range), GFP_KERNEL);
-	if (unlikely(!hmm_range))
-		return -ENOMEM;
+	struct hmm_range *hmm_range = &range->hmm_range;
 
 	pfns = kvmalloc_array(npages, sizeof(*pfns), GFP_KERNEL);
 	if (unlikely(!pfns)) {
@@ -190,19 +197,12 @@ int amdgpu_hmm_range_get_pages(struct mmu_interval_notifier *notifier,
 		pr_debug("hmm range: start = 0x%lx, end = 0x%lx",
 			hmm_range->start, hmm_range->end);
 
-		/* Assuming 512MB takes maxmium 1 second to fault page address */
-		timeout = max((hmm_range->end - hmm_range->start) >> 29, 1UL);
-		timeout *= HMM_RANGE_DEFAULT_TIMEOUT;
-		timeout = jiffies + msecs_to_jiffies(timeout);
+		timeout = jiffies + msecs_to_jiffies(HMM_RANGE_DEFAULT_TIMEOUT);
 
 retry:
 		hmm_range->notifier_seq = mmu_interval_read_begin(notifier);
 		r = hmm_range_fault(hmm_range);
 		if (unlikely(r)) {
-			/*
-			 * FIXME: This timeout should encompass the retry from
-			 * mmu_interval_read_retry() as well.
-			 */
 			if (r == -EBUSY && !time_after(jiffies, timeout))
 				goto retry;
 			goto out_free_pfns;
@@ -212,40 +212,82 @@ retry:
 			break;
 		hmm_range->hmm_pfns += MAX_WALK_BYTE >> PAGE_SHIFT;
 		hmm_range->start = hmm_range->end;
-		schedule();
 	} while (hmm_range->end < end);
 
 	hmm_range->start = start;
 	hmm_range->hmm_pfns = pfns;
 
-	/*
-	 * Due to default_flags, all pages are HMM_PFN_VALID or
-	 * hmm_range_fault() fails. FIXME: The pages cannot be touched outside
-	 * the notifier_lock, and mmu_interval_read_retry() must be done first.
-	 */
-	for (i = 0; pages && i < npages; i++)
-		pages[i] = hmm_pfn_to_page(pfns[i]);
-
-	*phmm_range = hmm_range;
-
 	return 0;
 
 out_free_pfns:
 	kvfree(pfns);
+	hmm_range->hmm_pfns = NULL;
 out_free_range:
-	kfree(hmm_range);
-
+	if (r == -EBUSY)
+		r = -EAGAIN;
 	return r;
 }
 
-bool amdgpu_hmm_range_get_pages_done(struct hmm_range *hmm_range)
+/**
+ * amdgpu_hmm_range_valid - check if an HMM range is still valid
+ * @range: pointer to the &struct amdgpu_hmm_range to validate
+ *
+ * Determines whether the given HMM range @range is still valid by
+ * checking for invalidations via the MMU notifier sequence. This is
+ * typically used to verify that the range has not been invalidated
+ * by concurrent address space updates before it is accessed.
+ *
+ * Return:
+ * * true if @range is valid and can be used safely
+ * * false if @range is NULL or has been invalidated
+ */
+bool amdgpu_hmm_range_valid(struct amdgpu_hmm_range *range)
 {
-	bool r;
+	if (!range)
+		return false;
 
-	r = mmu_interval_read_retry(hmm_range->notifier,
-				    hmm_range->notifier_seq);
-	kvfree(hmm_range->hmm_pfns);
-	kfree(hmm_range);
+	return !mmu_interval_read_retry(range->hmm_range.notifier,
+					range->hmm_range.notifier_seq);
+}
 
-	return r;
+/**
+ * amdgpu_hmm_range_alloc - allocate and initialize an AMDGPU HMM range
+ * @bo: optional buffer object to associate with this HMM range
+ *
+ * Allocates memory for amdgpu_hmm_range and associates it with the @bo passed.
+ * The reference count of the @bo is incremented.
+ *
+ * Return:
+ * Pointer to a newly allocated struct amdgpu_hmm_range on success,
+ * or NULL if memory allocation fails.
+ */
+struct amdgpu_hmm_range *amdgpu_hmm_range_alloc(struct amdgpu_bo *bo)
+{
+	struct amdgpu_hmm_range *range;
+
+	range = kzalloc(sizeof(*range), GFP_KERNEL);
+	if (!range)
+		return NULL;
+
+	range->bo = amdgpu_bo_ref(bo);
+	return range;
+}
+
+/**
+ * amdgpu_hmm_range_free - release an AMDGPU HMM range
+ * @range: pointer to the range object to free
+ *
+ * Releases all resources held by @range, including the associated
+ * hmm_pfns and the dropping reference of associated bo if any.
+ *
+ * Return: void
+ */
+void amdgpu_hmm_range_free(struct amdgpu_hmm_range *range)
+{
+	if (!range)
+		return;
+
+	kvfree(range->hmm_range.hmm_pfns);
+	amdgpu_bo_unref(&range->bo);
+	kfree(range);
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.h
index e2edcd010ccc..140bc9cd57b4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.h
@@ -31,13 +31,20 @@
 #include <linux/interval_tree.h>
 #include <linux/mmu_notifier.h>
 
+struct amdgpu_hmm_range {
+	struct hmm_range hmm_range;
+	struct amdgpu_bo *bo;
+};
+
 int amdgpu_hmm_range_get_pages(struct mmu_interval_notifier *notifier,
 			       uint64_t start, uint64_t npages, bool readonly,
-			       void *owner, struct page **pages,
-			       struct hmm_range **phmm_range);
-bool amdgpu_hmm_range_get_pages_done(struct hmm_range *hmm_range);
+			       void *owner,
+			       struct amdgpu_hmm_range *range);
 
 #if defined(CONFIG_HMM_MIRROR)
+bool amdgpu_hmm_range_valid(struct amdgpu_hmm_range *range);
+struct amdgpu_hmm_range *amdgpu_hmm_range_alloc(struct amdgpu_bo *bo);
+void amdgpu_hmm_range_free(struct amdgpu_hmm_range *range);
 int amdgpu_hmm_register(struct amdgpu_bo *bo, unsigned long addr);
 void amdgpu_hmm_unregister(struct amdgpu_bo *bo);
 #else
@@ -47,7 +54,20 @@ static inline int amdgpu_hmm_register(struct amdgpu_bo *bo, unsigned long addr)
 		      "add CONFIG_ZONE_DEVICE=y in config file to fix this\n");
 	return -ENODEV;
 }
+
 static inline void amdgpu_hmm_unregister(struct amdgpu_bo *bo) {}
+
+static inline bool amdgpu_hmm_range_valid(struct amdgpu_hmm_range *range)
+{
+	return false;
+}
+
+static inline struct amdgpu_hmm_range *amdgpu_hmm_range_alloc(struct amdgpu_bo *bo)
+{
+	return NULL;
+}
+
+static inline void amdgpu_hmm_range_free(struct amdgpu_hmm_range *range) {}
 #endif
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c
index 82608df43396..9cb72f0c5277 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.c
@@ -24,7 +24,6 @@
  *          Alex Deucher
  */
 
-#include <linux/export.h>
 #include <linux/pci.h>
 
 #include <drm/drm_edid.h>
@@ -175,7 +174,6 @@ struct amdgpu_i2c_chan *amdgpu_i2c_create(struct drm_device *dev,
 
 	i2c->rec = *rec;
 	i2c->adapter.owner = THIS_MODULE;
-	i2c->adapter.class = I2C_CLASS_DDC;
 	i2c->adapter.dev.parent = dev->dev;
 	i2c->dev = dev;
 	i2c_set_adapdata(&i2c->adapter, i2c);
@@ -186,7 +184,7 @@ struct amdgpu_i2c_chan *amdgpu_i2c_create(struct drm_device *dev,
 		snprintf(i2c->adapter.name, sizeof(i2c->adapter.name),
 			 "AMDGPU i2c hw bus %s", name);
 		i2c->adapter.algo = &amdgpu_atombios_i2c_algo;
-		ret = i2c_add_adapter(&i2c->adapter);
+		ret = devm_i2c_add_adapter(dev->dev, &i2c->adapter);
 		if (ret)
 			goto out_free;
 	} else {
@@ -217,22 +215,23 @@ out_free:
 
 }
 
-void amdgpu_i2c_destroy(struct amdgpu_i2c_chan *i2c)
-{
-	if (!i2c)
-		return;
-	WARN_ON(i2c->has_aux);
-	i2c_del_adapter(&i2c->adapter);
-	kfree(i2c);
-}
-
-/* Add the default buses */
 void amdgpu_i2c_init(struct amdgpu_device *adev)
 {
-	if (amdgpu_hw_i2c)
-		DRM_INFO("hw_i2c forced on, you may experience display detection problems!\n");
-
-	amdgpu_atombios_i2c_init(adev);
+	if (!adev->is_atom_fw) {
+		if (!amdgpu_device_has_dc_support(adev)) {
+			amdgpu_atombios_i2c_init(adev);
+		} else {
+			switch (adev->asic_type) {
+			case CHIP_POLARIS10:
+			case CHIP_POLARIS11:
+			case CHIP_POLARIS12:
+				amdgpu_atombios_oem_i2c_init(adev, 0x97);
+				break;
+			default:
+				break;
+			}
+		}
+	}
 }
 
 /* remove all the buses */
@@ -240,28 +239,9 @@ void amdgpu_i2c_fini(struct amdgpu_device *adev)
 {
 	int i;
 
-	for (i = 0; i < AMDGPU_MAX_I2C_BUS; i++) {
-		if (adev->i2c_bus[i]) {
-			amdgpu_i2c_destroy(adev->i2c_bus[i]);
+	for (i = 0; i < AMDGPU_MAX_I2C_BUS; i++)
+		if (adev->i2c_bus[i])
 			adev->i2c_bus[i] = NULL;
-		}
-	}
-}
-
-/* Add additional buses */
-void amdgpu_i2c_add(struct amdgpu_device *adev,
-		    const struct amdgpu_i2c_bus_rec *rec,
-		    const char *name)
-{
-	struct drm_device *dev = adev_to_drm(adev);
-	int i;
-
-	for (i = 0; i < AMDGPU_MAX_I2C_BUS; i++) {
-		if (!adev->i2c_bus[i]) {
-			adev->i2c_bus[i] = amdgpu_i2c_create(dev, rec, name);
-			return;
-		}
-	}
 }
 
 /* looks up bus based on id */
@@ -280,7 +260,7 @@ amdgpu_i2c_lookup(struct amdgpu_device *adev,
 	return NULL;
 }
 
-static void amdgpu_i2c_get_byte(struct amdgpu_i2c_chan *i2c_bus,
+static int amdgpu_i2c_get_byte(struct amdgpu_i2c_chan *i2c_bus,
 				 u8 slave_addr,
 				 u8 addr,
 				 u8 *val)
@@ -305,16 +285,18 @@ static void amdgpu_i2c_get_byte(struct amdgpu_i2c_chan *i2c_bus,
 	out_buf[0] = addr;
 	out_buf[1] = 0;
 
-	if (i2c_transfer(&i2c_bus->adapter, msgs, 2) == 2) {
-		*val = in_buf[0];
-		DRM_DEBUG("val = 0x%02x\n", *val);
-	} else {
-		DRM_DEBUG("i2c 0x%02x 0x%02x read failed\n",
-			  addr, *val);
+	if (i2c_transfer(&i2c_bus->adapter, msgs, 2) != 2) {
+		DRM_DEBUG("i2c 0x%02x read failed\n", addr);
+		return -EIO;
 	}
+
+	*val = in_buf[0];
+	DRM_DEBUG("val = 0x%02x\n", *val);
+
+	return 0;
 }
 
-static void amdgpu_i2c_put_byte(struct amdgpu_i2c_chan *i2c_bus,
+static int amdgpu_i2c_put_byte(struct amdgpu_i2c_chan *i2c_bus,
 				 u8 slave_addr,
 				 u8 addr,
 				 u8 val)
@@ -330,9 +312,12 @@ static void amdgpu_i2c_put_byte(struct amdgpu_i2c_chan *i2c_bus,
 	out_buf[0] = addr;
 	out_buf[1] = val;
 
-	if (i2c_transfer(&i2c_bus->adapter, &msg, 1) != 1)
-		DRM_DEBUG("i2c 0x%02x 0x%02x write failed\n",
-			  addr, val);
+	if (i2c_transfer(&i2c_bus->adapter, &msg, 1) != 1) {
+		DRM_DEBUG("i2c 0x%02x 0x%02x write failed\n", addr, val);
+		return -EIO;
+	}
+
+	return 0;
 }
 
 /* ddc router switching */
@@ -347,16 +332,18 @@ amdgpu_i2c_router_select_ddc_port(const struct amdgpu_connector *amdgpu_connecto
 	if (!amdgpu_connector->router_bus)
 		return;
 
-	amdgpu_i2c_get_byte(amdgpu_connector->router_bus,
+	if (amdgpu_i2c_get_byte(amdgpu_connector->router_bus,
 			    amdgpu_connector->router.i2c_addr,
-			    0x3, &val);
+			    0x3, &val))
+		return;
 	val &= ~amdgpu_connector->router.ddc_mux_control_pin;
 	amdgpu_i2c_put_byte(amdgpu_connector->router_bus,
 			    amdgpu_connector->router.i2c_addr,
 			    0x3, val);
-	amdgpu_i2c_get_byte(amdgpu_connector->router_bus,
+	if (amdgpu_i2c_get_byte(amdgpu_connector->router_bus,
 			    amdgpu_connector->router.i2c_addr,
-			    0x1, &val);
+			    0x1, &val))
+		return;
 	val &= ~amdgpu_connector->router.ddc_mux_control_pin;
 	val |= amdgpu_connector->router.ddc_mux_state;
 	amdgpu_i2c_put_byte(amdgpu_connector->router_bus,
@@ -376,16 +363,18 @@ amdgpu_i2c_router_select_cd_port(const struct amdgpu_connector *amdgpu_connector
 	if (!amdgpu_connector->router_bus)
 		return;
 
-	amdgpu_i2c_get_byte(amdgpu_connector->router_bus,
+	if (amdgpu_i2c_get_byte(amdgpu_connector->router_bus,
 			    amdgpu_connector->router.i2c_addr,
-			    0x3, &val);
+			    0x3, &val))
+		return;
 	val &= ~amdgpu_connector->router.cd_mux_control_pin;
 	amdgpu_i2c_put_byte(amdgpu_connector->router_bus,
 			    amdgpu_connector->router.i2c_addr,
 			    0x3, val);
-	amdgpu_i2c_get_byte(amdgpu_connector->router_bus,
+	if (amdgpu_i2c_get_byte(amdgpu_connector->router_bus,
 			    amdgpu_connector->router.i2c_addr,
-			    0x1, &val);
+			    0x1, &val))
+		return;
 	val &= ~amdgpu_connector->router.cd_mux_control_pin;
 	val |= amdgpu_connector->router.cd_mux_state;
 	amdgpu_i2c_put_byte(amdgpu_connector->router_bus,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.h
index 63c2ff7499e1..1d3d3806e0dd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_i2c.h
@@ -30,9 +30,6 @@ struct amdgpu_i2c_chan *amdgpu_i2c_create(struct drm_device *dev,
 void amdgpu_i2c_destroy(struct amdgpu_i2c_chan *i2c);
 void amdgpu_i2c_init(struct amdgpu_device *adev);
 void amdgpu_i2c_fini(struct amdgpu_device *adev);
-void amdgpu_i2c_add(struct amdgpu_device *adev,
-		    const struct amdgpu_i2c_bus_rec *rec,
-		    const char *name);
 struct amdgpu_i2c_chan *
 amdgpu_i2c_lookup(struct amdgpu_device *adev,
 		  const struct amdgpu_i2c_bus_rec *i2c_bus);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
index df7eb0b7c4b9..586a58facca1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
@@ -62,7 +62,7 @@
  * Returns 0 on success, error on failure.
  */
 int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-		  unsigned size, enum amdgpu_ib_pool_type pool_type,
+		  unsigned int size, enum amdgpu_ib_pool_type pool_type,
 		  struct amdgpu_ib *ib)
 {
 	int r;
@@ -89,16 +89,14 @@ int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 /**
  * amdgpu_ib_free - free an IB (Indirect Buffer)
  *
- * @adev: amdgpu_device pointer
  * @ib: IB object to free
  * @f: the fence SA bo need wait on for the ib alloation
  *
  * Free an IB (all asics).
  */
-void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib,
-		    struct dma_fence *f)
+void amdgpu_ib_free(struct amdgpu_ib *ib, struct dma_fence *f)
 {
-	amdgpu_sa_bo_free(adev, &ib->sa_bo, f);
+	amdgpu_sa_bo_free(&ib->sa_bo, f);
 }
 
 /**
@@ -123,24 +121,26 @@ void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib,
  * a CONST_IB), it will be put on the ring prior to the DE IB.  Prior
  * to SI there was just a DE IB.
  */
-int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
+int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned int num_ibs,
 		       struct amdgpu_ib *ibs, struct amdgpu_job *job,
 		       struct dma_fence **f)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_ib *ib = &ibs[0];
 	struct dma_fence *tmp = NULL;
+	struct amdgpu_fence *af;
 	bool need_ctx_switch;
-	unsigned patch_offset = ~0;
 	struct amdgpu_vm *vm;
 	uint64_t fence_ctx;
 	uint32_t status = 0, alloc_size;
-	unsigned fence_flags = 0;
-	bool secure;
-
-	unsigned i;
-	int r = 0;
+	unsigned int fence_flags = 0;
+	bool secure, init_shadow;
+	u64 shadow_va, csa_va, gds_va;
+	int vmid = AMDGPU_JOB_GET_VMID(job);
 	bool need_pipe_sync = false;
+	unsigned int cond_exec;
+	unsigned int i;
+	int r = 0;
 
 	if (num_ibs == 0)
 		return -EINVAL;
@@ -149,26 +149,48 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 	if (job) {
 		vm = job->vm;
 		fence_ctx = job->base.s_fence ?
-			job->base.s_fence->scheduled.context : 0;
+			job->base.s_fence->finished.context : 0;
+		shadow_va = job->shadow_va;
+		csa_va = job->csa_va;
+		gds_va = job->gds_va;
+		init_shadow = job->init_shadow;
+		af = job->hw_fence;
+		/* Save the context of the job for reset handling.
+		 * The driver needs this so it can skip the ring
+		 * contents for guilty contexts.
+		 */
+		af->context = fence_ctx;
+		/* the vm fence is also part of the job's context */
+		job->hw_vm_fence->context = fence_ctx;
 	} else {
 		vm = NULL;
 		fence_ctx = 0;
+		shadow_va = 0;
+		csa_va = 0;
+		gds_va = 0;
+		init_shadow = false;
+		af = kzalloc(sizeof(*af), GFP_ATOMIC);
+		if (!af)
+			return -ENOMEM;
 	}
 
-	if (!ring->sched.ready && !ring->is_mes_queue) {
+	if (!ring->sched.ready) {
 		dev_err(adev->dev, "couldn't schedule ib on ring <%s>\n", ring->name);
-		return -EINVAL;
+		r = -EINVAL;
+		goto free_fence;
 	}
 
-	if (vm && !job->vmid && !ring->is_mes_queue) {
+	if (vm && !job->vmid) {
 		dev_err(adev->dev, "VM IB without ID\n");
-		return -EINVAL;
+		r = -EINVAL;
+		goto free_fence;
 	}
 
 	if ((ib->flags & AMDGPU_IB_FLAGS_SECURE) &&
 	    (!ring->funcs->secure_submission_supported)) {
 		dev_err(adev->dev, "secure submissions not supported on ring <%s>\n", ring->name);
-		return -EINVAL;
+		r = -EINVAL;
+		goto free_fence;
 	}
 
 	alloc_size = ring->funcs->emit_frame_size + num_ibs *
@@ -177,14 +199,14 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 	r = amdgpu_ring_alloc(ring, alloc_size);
 	if (r) {
 		dev_err(adev->dev, "scheduling IB failed (%d).\n", r);
-		return r;
+		goto free_fence;
 	}
 
 	need_ctx_switch = ring->current_ctx != fence_ctx;
 	if (ring->funcs->emit_pipeline_sync && job &&
 	    ((tmp = amdgpu_sync_get_fence(&job->explicit_sync)) ||
-	     (amdgpu_sriov_vf(adev) && need_ctx_switch) ||
-	     amdgpu_vm_need_pipeline_sync(ring, job))) {
+	     need_ctx_switch || amdgpu_vm_need_pipeline_sync(ring, job))) {
+
 		need_pipe_sync = true;
 
 		if (tmp)
@@ -212,8 +234,14 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 	}
 
 	amdgpu_ring_ib_begin(ring);
-	if (job && ring->funcs->init_cond_exec)
-		patch_offset = amdgpu_ring_init_cond_exec(ring);
+
+	if (ring->funcs->emit_gfx_shadow)
+		amdgpu_ring_emit_gfx_shadow(ring, shadow_va, csa_va, gds_va,
+					    init_shadow, vmid);
+
+	if (ring->funcs->init_cond_exec)
+		cond_exec = amdgpu_ring_init_cond_exec(ring,
+						       ring->cond_exe_gpu_addr);
 
 	amdgpu_device_flush_hdp(adev, ring);
 
@@ -263,32 +291,54 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 				       fence_flags | AMDGPU_FENCE_FLAG_64BIT);
 	}
 
-	r = amdgpu_fence_emit(ring, f, job, fence_flags);
+	if (ring->funcs->emit_gfx_shadow && ring->funcs->init_cond_exec) {
+		amdgpu_ring_emit_gfx_shadow(ring, 0, 0, 0, false, 0);
+		amdgpu_ring_init_cond_exec(ring, ring->cond_exe_gpu_addr);
+	}
+
+	r = amdgpu_fence_emit(ring, af, fence_flags);
 	if (r) {
 		dev_err(adev->dev, "failed to emit fence (%d)\n", r);
 		if (job && job->vmid)
-			amdgpu_vmid_reset(adev, ring->funcs->vmhub, job->vmid);
+			amdgpu_vmid_reset(adev, ring->vm_hub, job->vmid);
 		amdgpu_ring_undo(ring);
 		return r;
 	}
+	*f = &af->base;
+	/* get a ref for the job */
+	if (job)
+		dma_fence_get(*f);
 
 	if (ring->funcs->insert_end)
 		ring->funcs->insert_end(ring);
 
-	if (patch_offset != ~0 && ring->funcs->patch_cond_exec)
-		amdgpu_ring_patch_cond_exec(ring, patch_offset);
+	amdgpu_ring_patch_cond_exec(ring, cond_exec);
 
 	ring->current_ctx = fence_ctx;
-	if (vm && ring->funcs->emit_switch_buffer)
+	if (job && ring->funcs->emit_switch_buffer)
 		amdgpu_ring_emit_switch_buffer(ring);
 
 	if (ring->funcs->emit_wave_limit &&
 	    ring->hw_prio == AMDGPU_GFX_PIPE_PRIO_HIGH)
 		ring->funcs->emit_wave_limit(ring, false);
 
+	/* Save the wptr associated with this fence.
+	 * This must be last for resets to work properly
+	 * as we need to save the wptr associated with this
+	 * fence so we know what rings contents to backup
+	 * after we reset the queue.
+	 */
+	amdgpu_fence_save_wptr(af);
+
 	amdgpu_ring_ib_end(ring);
 	amdgpu_ring_commit(ring);
+
 	return 0;
+
+free_fence:
+	if (!job)
+		kfree(af);
+	return r;
 }
 
 /**
@@ -358,7 +408,7 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev)
 {
 	long tmo_gfx, tmo_mm;
 	int r, ret = 0;
-	unsigned i;
+	unsigned int i;
 
 	tmo_mm = tmo_gfx = AMDGPU_IB_TEST_TIMEOUT;
 	if (amdgpu_sriov_vf(adev)) {
@@ -375,7 +425,7 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev)
 		/* for CP & SDMA engines since they are scheduled together so
 		 * need to make the timeout width enough to cover the time
 		 * cost waiting for it coming back under RUNTIME only
-		*/
+		 */
 		tmo_gfx = 8 * AMDGPU_IB_TEST_TIMEOUT;
 	} else if (adev->gmc.xgmi.hive_id) {
 		tmo_gfx = AMDGPU_IB_TEST_GFX_XGMI_TIMEOUT;
@@ -436,15 +486,15 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev)
 
 static int amdgpu_debugfs_sa_info_show(struct seq_file *m, void *unused)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct amdgpu_device *adev = m->private;
 
-	seq_printf(m, "--------------------- DELAYED --------------------- \n");
+	seq_puts(m, "--------------------- DELAYED ---------------------\n");
 	amdgpu_sa_bo_dump_debug_info(&adev->ib_pools[AMDGPU_IB_POOL_DELAYED],
 				     m);
-	seq_printf(m, "-------------------- IMMEDIATE -------------------- \n");
+	seq_puts(m, "-------------------- IMMEDIATE --------------------\n");
 	amdgpu_sa_bo_dump_debug_info(&adev->ib_pools[AMDGPU_IB_POOL_IMMEDIATE],
 				     m);
-	seq_printf(m, "--------------------- DIRECT ---------------------- \n");
+	seq_puts(m, "--------------------- DIRECT ----------------------\n");
 	amdgpu_sa_bo_dump_debug_info(&adev->ib_pools[AMDGPU_IB_POOL_DIRECT], m);
 
 	return 0;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
index 3f07b1a2ce47..9cab36322c16 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
@@ -62,9 +62,8 @@ int amdgpu_pasid_alloc(unsigned int bits)
 	int pasid = -EINVAL;
 
 	for (bits = min(bits, 31U); bits > 0; bits--) {
-		pasid = ida_simple_get(&amdgpu_pasid_ida,
-				       1U << (bits - 1), 1U << bits,
-				       GFP_KERNEL);
+		pasid = ida_alloc_range(&amdgpu_pasid_ida, 1U << (bits - 1),
+					(1U << bits) - 1, GFP_KERNEL);
 		if (pasid != -ENOSPC)
 			break;
 	}
@@ -82,7 +81,7 @@ int amdgpu_pasid_alloc(unsigned int bits)
 void amdgpu_pasid_free(u32 pasid)
 {
 	trace_amdgpu_pasid_freed(pasid);
-	ida_simple_remove(&amdgpu_pasid_ida, pasid);
+	ida_free(&amdgpu_pasid_ida, pasid);
 }
 
 static void amdgpu_pasid_free_cb(struct dma_fence *fence,
@@ -188,7 +187,6 @@ static bool amdgpu_vmid_compatible(struct amdgpu_vmid *id,
 /**
  * amdgpu_vmid_grab_idle - grab idle VMID
  *
- * @vm: vm to allocate id for
  * @ring: ring we want to submit job to
  * @idle: resulting idle VMID
  * @fence: fence to wait for if no id could be grabbed
@@ -196,66 +194,41 @@ static bool amdgpu_vmid_compatible(struct amdgpu_vmid *id,
  * Try to find an idle VMID, if none is idle add a fence to wait to the sync
  * object. Returns -ENOMEM when we are out of memory.
  */
-static int amdgpu_vmid_grab_idle(struct amdgpu_vm *vm,
-				 struct amdgpu_ring *ring,
+static int amdgpu_vmid_grab_idle(struct amdgpu_ring *ring,
 				 struct amdgpu_vmid **idle,
 				 struct dma_fence **fence)
 {
 	struct amdgpu_device *adev = ring->adev;
-	unsigned vmhub = ring->funcs->vmhub;
+	unsigned vmhub = ring->vm_hub;
 	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
-	struct dma_fence **fences;
-	unsigned i;
 
+	/* If anybody is waiting for a VMID let everybody wait for fairness */
 	if (!dma_fence_is_signaled(ring->vmid_wait)) {
 		*fence = dma_fence_get(ring->vmid_wait);
 		return 0;
 	}
 
-	fences = kmalloc_array(id_mgr->num_ids, sizeof(void *), GFP_KERNEL);
-	if (!fences)
-		return -ENOMEM;
-
 	/* Check if we have an idle VMID */
-	i = 0;
-	list_for_each_entry((*idle), &id_mgr->ids_lru, list) {
+	list_for_each_entry_reverse((*idle), &id_mgr->ids_lru, list) {
 		/* Don't use per engine and per process VMID at the same time */
 		struct amdgpu_ring *r = adev->vm_manager.concurrent_flush ?
 			NULL : ring;
 
-		fences[i] = amdgpu_sync_peek_fence(&(*idle)->active, r);
-		if (!fences[i])
-			break;
-		++i;
+		*fence = amdgpu_sync_peek_fence(&(*idle)->active, r);
+		if (!(*fence))
+			return 0;
 	}
 
-	/* If we can't find a idle VMID to use, wait till one becomes available */
-	if (&(*idle)->list == &id_mgr->ids_lru) {
-		u64 fence_context = adev->vm_manager.fence_context + ring->idx;
-		unsigned seqno = ++adev->vm_manager.seqno[ring->idx];
-		struct dma_fence_array *array;
-		unsigned j;
-
-		*idle = NULL;
-		for (j = 0; j < i; ++j)
-			dma_fence_get(fences[j]);
-
-		array = dma_fence_array_create(i, fences, fence_context,
-					       seqno, true);
-		if (!array) {
-			for (j = 0; j < i; ++j)
-				dma_fence_put(fences[j]);
-			kfree(fences);
-			return -ENOMEM;
-		}
-
-		*fence = dma_fence_get(&array->base);
-		dma_fence_put(ring->vmid_wait);
-		ring->vmid_wait = &array->base;
-		return 0;
-	}
-	kfree(fences);
+	/*
+	 * If we can't find a idle VMID to use, wait on a fence from the least
+	 * recently used in the hope that it will be available soon.
+	 */
+	*idle = NULL;
+	dma_fence_put(ring->vmid_wait);
+	ring->vmid_wait = dma_fence_get(*fence);
 
+	/* This is the reference we return */
+	dma_fence_get(*fence);
 	return 0;
 }
 
@@ -277,23 +250,28 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm,
 				     struct dma_fence **fence)
 {
 	struct amdgpu_device *adev = ring->adev;
-	unsigned vmhub = ring->funcs->vmhub;
-	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
+	unsigned vmhub = ring->vm_hub;
 	uint64_t fence_context = adev->fence_context + ring->idx;
 	bool needs_flush = vm->use_cpu_for_update;
 	uint64_t updates = amdgpu_vm_tlb_seq(vm);
 	int r;
 
-	*id = id_mgr->reserved;
+	*id = vm->reserved_vmid[vmhub];
 	if ((*id)->owner != vm->immediate.fence_context ||
 	    !amdgpu_vmid_compatible(*id, job) ||
 	    (*id)->flushed_updates < updates ||
 	    !(*id)->last_flush ||
 	    ((*id)->last_flush->context != fence_context &&
-	     !dma_fence_is_signaled((*id)->last_flush))) {
+	     !dma_fence_is_signaled((*id)->last_flush)))
+		needs_flush = true;
+
+	if ((*id)->owner != vm->immediate.fence_context ||
+	    (!adev->vm_manager.concurrent_flush && needs_flush)) {
 		struct dma_fence *tmp;
 
-		/* Don't use per engine and per process VMID at the same time */
+		/* Don't use per engine and per process VMID at the
+		 * same time
+		 */
 		if (adev->vm_manager.concurrent_flush)
 			ring = NULL;
 
@@ -305,13 +283,13 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm,
 			*fence = dma_fence_get(tmp);
 			return 0;
 		}
-		needs_flush = true;
 	}
 
 	/* Good we can use this VMID. Remember this submission as
 	* user of the VMID.
 	*/
-	r = amdgpu_sync_fence(&(*id)->active, &job->base.s_fence->finished);
+	r = amdgpu_sync_fence(&(*id)->active, &job->base.s_fence->finished,
+			      GFP_ATOMIC);
 	if (r)
 		return r;
 
@@ -327,18 +305,16 @@ static int amdgpu_vmid_grab_reserved(struct amdgpu_vm *vm,
  * @ring: ring we want to submit job to
  * @job: job who wants to use the VMID
  * @id: resulting VMID
- * @fence: fence to wait for if no id could be grabbed
  *
  * Try to reuse a VMID for this submission.
  */
 static int amdgpu_vmid_grab_used(struct amdgpu_vm *vm,
 				 struct amdgpu_ring *ring,
 				 struct amdgpu_job *job,
-				 struct amdgpu_vmid **id,
-				 struct dma_fence **fence)
+				 struct amdgpu_vmid **id)
 {
 	struct amdgpu_device *adev = ring->adev;
-	unsigned vmhub = ring->funcs->vmhub;
+	unsigned vmhub = ring->vm_hub;
 	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
 	uint64_t fence_context = adev->fence_context + ring->idx;
 	uint64_t updates = amdgpu_vm_tlb_seq(vm);
@@ -372,7 +348,8 @@ static int amdgpu_vmid_grab_used(struct amdgpu_vm *vm,
 		 * user of the VMID.
 		 */
 		r = amdgpu_sync_fence(&(*id)->active,
-				      &job->base.s_fence->finished);
+				      &job->base.s_fence->finished,
+				      GFP_ATOMIC);
 		if (r)
 			return r;
 
@@ -398,23 +375,23 @@ int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
 		     struct amdgpu_job *job, struct dma_fence **fence)
 {
 	struct amdgpu_device *adev = ring->adev;
-	unsigned vmhub = ring->funcs->vmhub;
+	unsigned vmhub = ring->vm_hub;
 	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
 	struct amdgpu_vmid *idle = NULL;
 	struct amdgpu_vmid *id = NULL;
 	int r = 0;
 
 	mutex_lock(&id_mgr->lock);
-	r = amdgpu_vmid_grab_idle(vm, ring, &idle, fence);
+	r = amdgpu_vmid_grab_idle(ring, &idle, fence);
 	if (r || !idle)
 		goto error;
 
-	if (vm->reserved_vmid[vmhub]) {
+	if (amdgpu_vmid_uses_reserved(vm, vmhub)) {
 		r = amdgpu_vmid_grab_reserved(vm, ring, job, &id, fence);
 		if (r || !id)
 			goto error;
 	} else {
-		r = amdgpu_vmid_grab_used(vm, ring, job, &id, fence);
+		r = amdgpu_vmid_grab_used(vm, ring, job, &id);
 		if (r)
 			goto error;
 
@@ -424,7 +401,8 @@ int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
 
 			/* Remember this submission as user of the VMID */
 			r = amdgpu_sync_fence(&id->active,
-					      &job->base.s_fence->finished);
+					      &job->base.s_fence->finished,
+					      GFP_ATOMIC);
 			if (r)
 				goto error;
 
@@ -459,47 +437,73 @@ error:
 	return r;
 }
 
-int amdgpu_vmid_alloc_reserved(struct amdgpu_device *adev,
-			       struct amdgpu_vm *vm,
+/*
+ * amdgpu_vmid_uses_reserved - check if a VM will use a reserved VMID
+ * @vm: the VM to check
+ * @vmhub: the VMHUB which will be used
+ *
+ * Returns: True if the VM will use a reserved VMID.
+ */
+bool amdgpu_vmid_uses_reserved(struct amdgpu_vm *vm, unsigned int vmhub)
+{
+	return vm->reserved_vmid[vmhub];
+}
+
+/*
+ * amdgpu_vmid_alloc_reserved - reserve a specific VMID for this vm
+ * @adev: amdgpu device structure
+ * @vm: the VM to reserve an ID for
+ * @vmhub: the VMHUB which should be used
+ *
+ * Mostly used to have a reserved VMID for debugging and SPM.
+ *
+ * Returns: 0 for success, -ENOENT if an ID is already reserved.
+ */
+int amdgpu_vmid_alloc_reserved(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			       unsigned vmhub)
 {
 	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
+	struct amdgpu_vmid *id;
+	int r = 0;
 
 	mutex_lock(&id_mgr->lock);
 	if (vm->reserved_vmid[vmhub])
 		goto unlock;
-
-	++id_mgr->reserved_use_count;
-	if (!id_mgr->reserved) {
-		struct amdgpu_vmid *id;
-
-		id = list_first_entry(&id_mgr->ids_lru, struct amdgpu_vmid,
-				      list);
-		/* Remove from normal round robin handling */
-		list_del_init(&id->list);
-		id_mgr->reserved = id;
+	if (id_mgr->reserved_vmid) {
+		r = -ENOENT;
+		goto unlock;
 	}
-	vm->reserved_vmid[vmhub] = true;
+	/* Remove from normal round robin handling */
+	id = list_first_entry(&id_mgr->ids_lru, struct amdgpu_vmid, list);
+	list_del_init(&id->list);
+	vm->reserved_vmid[vmhub] = id;
+	id_mgr->reserved_vmid = true;
+	mutex_unlock(&id_mgr->lock);
 
+	return 0;
 unlock:
 	mutex_unlock(&id_mgr->lock);
-	return 0;
+	return r;
 }
 
-void amdgpu_vmid_free_reserved(struct amdgpu_device *adev,
-			       struct amdgpu_vm *vm,
+/*
+ * amdgpu_vmid_free_reserved - free up a reserved VMID again
+ * @adev: amdgpu device structure
+ * @vm: the VM with the reserved ID
+ * @vmhub: the VMHUB which should be used
+ */
+void amdgpu_vmid_free_reserved(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			       unsigned vmhub)
 {
 	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
 
 	mutex_lock(&id_mgr->lock);
-	if (vm->reserved_vmid[vmhub] &&
-	    !--id_mgr->reserved_use_count) {
-		/* give the reserved ID back to normal round robin */
-		list_add(&id_mgr->reserved->list, &id_mgr->ids_lru);
-		id_mgr->reserved = NULL;
+	if (vm->reserved_vmid[vmhub]) {
+		list_add(&vm->reserved_vmid[vmhub]->list,
+			&id_mgr->ids_lru);
+		vm->reserved_vmid[vmhub] = NULL;
+		id_mgr->reserved_vmid = false;
 	}
-	vm->reserved_vmid[vmhub] = false;
 	mutex_unlock(&id_mgr->lock);
 }
 
@@ -566,10 +570,17 @@ void amdgpu_vmid_mgr_init(struct amdgpu_device *adev)
 
 		mutex_init(&id_mgr->lock);
 		INIT_LIST_HEAD(&id_mgr->ids_lru);
-		id_mgr->reserved_use_count = 0;
 
-		/* manage only VMIDs not used by KFD */
-		id_mgr->num_ids = adev->vm_manager.first_kfd_vmid;
+		/* for GC <10, SDMA uses MMHUB so use first_kfd_vmid for both GC and MM */
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(10, 0, 0))
+			/* manage only VMIDs not used by KFD */
+			id_mgr->num_ids = adev->vm_manager.first_kfd_vmid;
+		else if (AMDGPU_IS_MMHUB0(i) ||
+			 AMDGPU_IS_MMHUB1(i))
+			id_mgr->num_ids = 16;
+		else
+			/* manage only VMIDs not used by KFD */
+			id_mgr->num_ids = adev->vm_manager.first_kfd_vmid;
 
 		/* skip over VMID 0, since it is the system VM */
 		for (j = 1; j < id_mgr->num_ids; ++j) {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h
index d1cc09b45da4..b3649cd3af56 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.h
@@ -67,8 +67,7 @@ struct amdgpu_vmid_mgr {
 	unsigned		num_ids;
 	struct list_head	ids_lru;
 	struct amdgpu_vmid	ids[AMDGPU_NUM_VMID];
-	struct amdgpu_vmid	*reserved;
-	unsigned int		reserved_use_count;
+	bool			reserved_vmid;
 };
 
 int amdgpu_pasid_alloc(unsigned int bits);
@@ -78,11 +77,10 @@ void amdgpu_pasid_free_delayed(struct dma_resv *resv,
 
 bool amdgpu_vmid_had_gpu_reset(struct amdgpu_device *adev,
 			       struct amdgpu_vmid *id);
-int amdgpu_vmid_alloc_reserved(struct amdgpu_device *adev,
-			       struct amdgpu_vm *vm,
+bool amdgpu_vmid_uses_reserved(struct amdgpu_vm *vm, unsigned int vmhub);
+int amdgpu_vmid_alloc_reserved(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			       unsigned vmhub);
-void amdgpu_vmid_free_reserved(struct amdgpu_device *adev,
-			       struct amdgpu_vm *vm,
+void amdgpu_vmid_free_reserved(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			       unsigned vmhub);
 int amdgpu_vmid_grab(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
 		     struct amdgpu_job *job, struct dma_fence **fence);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c
index 1d5af50331e4..a6419246e9c2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c
@@ -25,6 +25,7 @@
 
 #include "amdgpu.h"
 #include "amdgpu_ih.h"
+#include "amdgpu_reset.h"
 
 /**
  * amdgpu_ih_ring_init - initialize the IH state
@@ -138,6 +139,7 @@ void amdgpu_ih_ring_fini(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih)
 /**
  * amdgpu_ih_ring_write - write IV to the ring buffer
  *
+ * @adev: amdgpu_device pointer
  * @ih: ih ring to write to
  * @iv: the iv to write
  * @num_dw: size of the iv in dw
@@ -145,8 +147,8 @@ void amdgpu_ih_ring_fini(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih)
  * Writes an IV to the ring buffer using the CPU and increment the wptr.
  * Used for testing and delegating IVs to a software ring.
  */
-void amdgpu_ih_ring_write(struct amdgpu_ih_ring *ih, const uint32_t *iv,
-			  unsigned int num_dw)
+void amdgpu_ih_ring_write(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih,
+			  const uint32_t *iv, unsigned int num_dw)
 {
 	uint32_t wptr = le32_to_cpu(*ih->wptr_cpu) >> 2;
 	unsigned int i;
@@ -161,6 +163,9 @@ void amdgpu_ih_ring_write(struct amdgpu_ih_ring *ih, const uint32_t *iv,
 	if (wptr != READ_ONCE(ih->rptr)) {
 		wmb();
 		WRITE_ONCE(*ih->wptr_cpu, cpu_to_le32(wptr));
+	} else if (adev->irq.retry_cam_enabled) {
+		dev_warn_once(adev->dev, "IH soft ring buffer overflow 0x%X, 0x%X\n",
+			      wptr, ih->rptr);
 	}
 }
 
@@ -213,7 +218,7 @@ int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih)
 
 restart_ih:
 	count  = AMDGPU_IH_MAX_NUM_IVS;
-	DRM_DEBUG("%s: rptr %d, wptr %d\n", __func__, ih->rptr, wptr);
+	dev_dbg(adev->dev, "%s: rptr %d, wptr %d\n", __func__, ih->rptr, wptr);
 
 	/* Order reading of wptr vs. reading of IH ring data */
 	rmb();
@@ -223,13 +228,23 @@ restart_ih:
 		ih->rptr &= ih->ptr_mask;
 	}
 
-	amdgpu_ih_set_rptr(adev, ih);
+	if (!ih->overflow)
+		amdgpu_ih_set_rptr(adev, ih);
+
 	wake_up_all(&ih->wait_process);
 
 	/* make sure wptr hasn't changed while processing */
 	wptr = amdgpu_ih_get_wptr(adev, ih);
 	if (wptr != ih->rptr)
-		goto restart_ih;
+		if (!ih->overflow)
+			goto restart_ih;
+
+	if (ih->overflow)
+		if (amdgpu_sriov_runtime(adev))
+			WARN_ONCE(!amdgpu_reset_domain_schedule(adev->reset_domain,
+				   &adev->virt.flr_work),
+				  "Failed to queue work! at %s",
+				  __func__);
 
 	return IRQ_HANDLED;
 }
@@ -270,7 +285,7 @@ void amdgpu_ih_decode_iv_helper(struct amdgpu_device *adev,
 	entry->timestamp = dw[1] | ((u64)(dw[2] & 0xffff) << 32);
 	entry->timestamp_src = dw[2] >> 31;
 	entry->pasid = dw[3] & 0xffff;
-	entry->pasid_src = dw[3] >> 31;
+	entry->node_id = (dw[3] >> 16) & 0xff;
 	entry->src_data[0] = dw[4];
 	entry->src_data[1] = dw[5];
 	entry->src_data[2] = dw[6];
@@ -294,3 +309,9 @@ uint64_t amdgpu_ih_decode_iv_ts_helper(struct amdgpu_ih_ring *ih, u32 rptr,
 	dw2 = le32_to_cpu(ih->ring[ring_index + 2]);
 	return dw1 | ((u64)(dw2 & 0xffff) << 32);
 }
+
+const char *amdgpu_ih_ring_name(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih)
+{
+	return ih == &adev->irq.ih ? "ih" : ih == &adev->irq.ih_soft ? "sw ih" :
+	       ih == &adev->irq.ih1 ? "ih1" : ih == &adev->irq.ih2 ? "ih2" : "unknown";
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h
index dd1c2eded6b9..f58b6be7fccc 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h
@@ -27,6 +27,9 @@
 /* Maximum number of IVs processed at once */
 #define AMDGPU_IH_MAX_NUM_IVS	32
 
+#define IH_RING_SIZE	(256 * 1024)
+#define IH_SW_RING_SIZE	(16 * 1024)	/* enough for 512 CAM entries */
+
 struct amdgpu_device;
 struct amdgpu_iv_entry;
 
@@ -53,14 +56,14 @@ struct amdgpu_ih_ring {
 	bool			use_bus_addr;
 
 	struct amdgpu_bo	*ring_obj;
-	volatile uint32_t	*ring;
+	uint32_t		*ring;
 	uint64_t		gpu_addr;
 
 	uint64_t		wptr_addr;
-	volatile uint32_t	*wptr_cpu;
+	uint32_t		*wptr_cpu;
 
 	uint64_t		rptr_addr;
-	volatile uint32_t	*rptr_cpu;
+	uint32_t		*rptr_cpu;
 
 	bool                    enabled;
 	unsigned		rptr;
@@ -69,12 +72,16 @@ struct amdgpu_ih_ring {
 	/* For waiting on IH processing at checkpoint. */
 	wait_queue_head_t wait_process;
 	uint64_t		processed_timestamp;
+	bool overflow;
 };
 
 /* return true if time stamp t2 is after t1 with 48bit wrap around */
 #define amdgpu_ih_ts_after(t1, t2) \
 		(((int64_t)((t2) << 16) - (int64_t)((t1) << 16)) > 0LL)
 
+#define amdgpu_ih_ts_after_or_equal(t1, t2) \
+		(((int64_t)((t2) << 16) - (int64_t)((t1) << 16)) >= 0LL)
+
 /* provided by the ih block */
 struct amdgpu_ih_funcs {
 	/* ring read/write ptr handling, called from interrupt context */
@@ -97,8 +104,8 @@ struct amdgpu_ih_funcs {
 int amdgpu_ih_ring_init(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih,
 			unsigned ring_size, bool use_bus_addr);
 void amdgpu_ih_ring_fini(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih);
-void amdgpu_ih_ring_write(struct amdgpu_ih_ring *ih, const uint32_t *iv,
-			  unsigned int num_dw);
+void amdgpu_ih_ring_write(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih,
+			  const uint32_t *iv, unsigned int num_dw);
 int amdgpu_ih_wait_on_checkpoint_process_ts(struct amdgpu_device *adev,
 					    struct amdgpu_ih_ring *ih);
 int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih);
@@ -107,4 +114,5 @@ void amdgpu_ih_decode_iv_helper(struct amdgpu_device *adev,
 				struct amdgpu_iv_entry *entry);
 uint64_t amdgpu_ih_decode_iv_ts_helper(struct amdgpu_ih_ring *ih, u32 rptr,
 				       signed int offset);
+const char *amdgpu_ih_ring_name(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ip.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ip.c
new file mode 100644
index 000000000000..99e1cf4fc955
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ip.c
@@ -0,0 +1,96 @@
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+#include "amdgpu_ip.h"
+
+static int8_t amdgpu_logical_to_dev_inst(struct amdgpu_device *adev,
+					 enum amd_hw_ip_block_type block,
+					 int8_t inst)
+{
+	int8_t dev_inst;
+
+	switch (block) {
+	case GC_HWIP:
+	case SDMA0_HWIP:
+	/* Both JPEG and VCN as JPEG is only alias of VCN */
+	case VCN_HWIP:
+		dev_inst = adev->ip_map.dev_inst[block][inst];
+		break;
+	default:
+		/* For rest of the IPs, no look up required.
+		 * Assume 'logical instance == physical instance' for all configs. */
+		dev_inst = inst;
+		break;
+	}
+
+	return dev_inst;
+}
+
+static uint32_t amdgpu_logical_to_dev_mask(struct amdgpu_device *adev,
+					   enum amd_hw_ip_block_type block,
+					   uint32_t mask)
+{
+	uint32_t dev_mask = 0;
+	int8_t log_inst, dev_inst;
+
+	while (mask) {
+		log_inst = ffs(mask) - 1;
+		dev_inst = amdgpu_logical_to_dev_inst(adev, block, log_inst);
+		dev_mask |= (1 << dev_inst);
+		mask &= ~(1 << log_inst);
+	}
+
+	return dev_mask;
+}
+
+static void amdgpu_populate_ip_map(struct amdgpu_device *adev,
+				   enum amd_hw_ip_block_type ip_block,
+				   uint32_t inst_mask)
+{
+	int l = 0, i;
+
+	while (inst_mask) {
+		i = ffs(inst_mask) - 1;
+		adev->ip_map.dev_inst[ip_block][l++] = i;
+		inst_mask &= ~(1 << i);
+	}
+	for (; l < HWIP_MAX_INSTANCE; l++)
+		adev->ip_map.dev_inst[ip_block][l] = -1;
+}
+
+void amdgpu_ip_map_init(struct amdgpu_device *adev)
+{
+	u32 ip_map[][2] = {
+		{ GC_HWIP, adev->gfx.xcc_mask },
+		{ SDMA0_HWIP, adev->sdma.sdma_mask },
+		{ VCN_HWIP, adev->vcn.inst_mask },
+	};
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(ip_map); ++i)
+		amdgpu_populate_ip_map(adev, ip_map[i][0], ip_map[i][1]);
+
+	adev->ip_map.logical_to_dev_inst = amdgpu_logical_to_dev_inst;
+	adev->ip_map.logical_to_dev_mask = amdgpu_logical_to_dev_mask;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ip.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ip.h
new file mode 100644
index 000000000000..2490fd322aec
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ip.h
@@ -0,0 +1,29 @@
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __AMDGPU_IP_H__
+#define __AMDGPU_IP_H__
+
+void amdgpu_ip_map_init(struct amdgpu_device *adev);
+
+#endif /* __AMDGPU_IP_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
index d0a1cc88832c..8112ffc85995 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
@@ -99,6 +99,21 @@ const char *soc15_ih_clientid_name[] = {
 	"MP1"
 };
 
+const int node_id_to_phys_map[NODEID_MAX] = {
+	[AID0_NODEID] = 0,
+	[XCD0_NODEID] = 0,
+	[XCD1_NODEID] = 1,
+	[AID1_NODEID] = 1,
+	[XCD2_NODEID] = 2,
+	[XCD3_NODEID] = 3,
+	[AID2_NODEID] = 2,
+	[XCD4_NODEID] = 4,
+	[XCD5_NODEID] = 5,
+	[AID3_NODEID] = 3,
+	[XCD6_NODEID] = 6,
+	[XCD7_NODEID] = 7,
+};
+
 /**
  * amdgpu_irq_disable_all - disable *all* interrupts
  *
@@ -109,7 +124,7 @@ const char *soc15_ih_clientid_name[] = {
 void amdgpu_irq_disable_all(struct amdgpu_device *adev)
 {
 	unsigned long irqflags;
-	unsigned i, j, k;
+	unsigned int i, j, k;
 	int r;
 
 	spin_lock_irqsave(&adev->irq.lock, irqflags);
@@ -124,12 +139,12 @@ void amdgpu_irq_disable_all(struct amdgpu_device *adev)
 				continue;
 
 			for (k = 0; k < src->num_types; ++k) {
-				atomic_set(&src->enabled_types[k], 0);
 				r = src->funcs->set(adev, src, k,
 						    AMDGPU_IRQ_STATE_DISABLE);
 				if (r)
-					DRM_ERROR("error disabling interrupt (%d)\n",
-						  r);
+					dev_err(adev->dev,
+						"error disabling interrupt (%d)\n",
+						r);
 			}
 		}
 	}
@@ -228,7 +243,7 @@ static bool amdgpu_msi_ok(struct amdgpu_device *adev)
 	return true;
 }
 
-static void amdgpu_restore_msix(struct amdgpu_device *adev)
+void amdgpu_restore_msix(struct amdgpu_device *adev)
 {
 	u16 ctrl;
 
@@ -256,29 +271,29 @@ static void amdgpu_restore_msix(struct amdgpu_device *adev)
  */
 int amdgpu_irq_init(struct amdgpu_device *adev)
 {
-	int r = 0;
-	unsigned int irq;
+	unsigned int irq, flags;
+	int r;
 
 	spin_lock_init(&adev->irq.lock);
 
 	/* Enable MSI if not disabled by module parameter */
 	adev->irq.msi_enabled = false;
 
-	if (amdgpu_msi_ok(adev)) {
-		int nvec = pci_msix_vec_count(adev->pdev);
-		unsigned int flags;
+	if (!amdgpu_msi_ok(adev))
+		flags = PCI_IRQ_INTX;
+	else
+		flags = PCI_IRQ_ALL_TYPES;
 
-		if (nvec <= 0) {
-			flags = PCI_IRQ_MSI;
-		} else {
-			flags = PCI_IRQ_MSI | PCI_IRQ_MSIX;
-		}
-		/* we only need one vector */
-		nvec = pci_alloc_irq_vectors(adev->pdev, 1, 1, flags);
-		if (nvec > 0) {
-			adev->irq.msi_enabled = true;
-			dev_dbg(adev->dev, "using MSI/MSI-X.\n");
-		}
+	/* we only need one vector */
+	r = pci_alloc_irq_vectors(adev->pdev, 1, 1, flags);
+	if (r < 0) {
+		dev_err(adev->dev, "Failed to alloc msi vectors\n");
+		return r;
+	}
+
+	if (amdgpu_msi_ok(adev)) {
+		adev->irq.msi_enabled = true;
+		dev_dbg(adev->dev, "using MSI/MSI-X.\n");
 	}
 
 	INIT_WORK(&adev->irq.ih1_work, amdgpu_irq_handle_ih1);
@@ -288,22 +303,29 @@ int amdgpu_irq_init(struct amdgpu_device *adev)
 	/* Use vector 0 for MSI-X. */
 	r = pci_irq_vector(adev->pdev, 0);
 	if (r < 0)
-		return r;
+		goto free_vectors;
 	irq = r;
 
 	/* PCI devices require shared interrupts. */
 	r = request_irq(irq, amdgpu_irq_handler, IRQF_SHARED, adev_to_drm(adev)->driver->name,
 			adev_to_drm(adev));
 	if (r)
-		return r;
+		goto free_vectors;
+
 	adev->irq.installed = true;
 	adev->irq.irq = irq;
 	adev_to_drm(adev)->max_vblank_count = 0x00ffffff;
 
-	DRM_DEBUG("amdgpu: irq initialized.\n");
+	dev_dbg(adev->dev, "amdgpu: irq initialized.\n");
 	return 0;
-}
 
+free_vectors:
+	if (adev->irq.msi_enabled)
+		pci_free_irq_vectors(adev->pdev);
+
+	adev->irq.msi_enabled = false;
+	return r;
+}
 
 void amdgpu_irq_fini_hw(struct amdgpu_device *adev)
 {
@@ -331,7 +353,7 @@ void amdgpu_irq_fini_hw(struct amdgpu_device *adev)
  */
 void amdgpu_irq_fini_sw(struct amdgpu_device *adev)
 {
-	unsigned i, j;
+	unsigned int i, j;
 
 	for (i = 0; i < AMDGPU_IRQ_CLIENTID_MAX; ++i) {
 		if (!adev->irq.client[i].sources)
@@ -365,7 +387,7 @@ void amdgpu_irq_fini_sw(struct amdgpu_device *adev)
  * 0 on success or error code otherwise
  */
 int amdgpu_irq_add_id(struct amdgpu_device *adev,
-		      unsigned client_id, unsigned src_id,
+		      unsigned int client_id, unsigned int src_id,
 		      struct amdgpu_irq_src *source)
 {
 	if (client_id >= AMDGPU_IRQ_CLIENTID_MAX)
@@ -417,13 +439,21 @@ void amdgpu_irq_dispatch(struct amdgpu_device *adev,
 {
 	u32 ring_index = ih->rptr >> 2;
 	struct amdgpu_iv_entry entry;
-	unsigned client_id, src_id;
+	unsigned int client_id, src_id;
 	struct amdgpu_irq_src *src;
 	bool handled = false;
 	int r;
 
 	entry.ih = ih;
 	entry.iv_entry = (const uint32_t *)&ih->ring[ring_index];
+
+	/*
+	 * timestamp is not supported on some legacy SOCs (cik, cz, iceland,
+	 * si and tonga), so initialize timestamp and timestamp_src to 0
+	 */
+	entry.timestamp = 0;
+	entry.timestamp_src = 0;
+
 	amdgpu_ih_decode_iv(adev, &entry);
 
 	trace_amdgpu_iv(ih - &adev->irq.ih, &entry);
@@ -432,28 +462,33 @@ void amdgpu_irq_dispatch(struct amdgpu_device *adev,
 	src_id = entry.src_id;
 
 	if (client_id >= AMDGPU_IRQ_CLIENTID_MAX) {
-		DRM_DEBUG("Invalid client_id in IV: %d\n", client_id);
+		dev_dbg(adev->dev, "Invalid client_id in IV: %d\n", client_id);
 
 	} else	if (src_id >= AMDGPU_MAX_IRQ_SRC_ID) {
-		DRM_DEBUG("Invalid src_id in IV: %d\n", src_id);
+		dev_dbg(adev->dev, "Invalid src_id in IV: %d\n", src_id);
 
-	} else if ((client_id == AMDGPU_IRQ_CLIENTID_LEGACY) &&
+	} else if (((client_id == AMDGPU_IRQ_CLIENTID_LEGACY) ||
+		    (client_id == SOC15_IH_CLIENTID_ISP)) &&
 		   adev->irq.virq[src_id]) {
 		generic_handle_domain_irq(adev->irq.domain, src_id);
 
 	} else if (!adev->irq.client[client_id].sources) {
-		DRM_DEBUG("Unregistered interrupt client_id: %d src_id: %d\n",
-			  client_id, src_id);
+		dev_dbg(adev->dev,
+			"Unregistered interrupt client_id: %d src_id: %d\n",
+			client_id, src_id);
 
 	} else if ((src = adev->irq.client[client_id].sources[src_id])) {
 		r = src->funcs->process(adev, src, &entry);
 		if (r < 0)
-			DRM_ERROR("error processing interrupt (%d)\n", r);
+			dev_err(adev->dev, "error processing interrupt (%d)\n",
+				r);
 		else if (r)
 			handled = true;
 
 	} else {
-		DRM_DEBUG("Unhandled interrupt src_id: %d\n", src_id);
+		dev_dbg(adev->dev,
+			"Unregistered interrupt src_id: %d of client_id:%d\n",
+			src_id, client_id);
 	}
 
 	/* Send it to amdkfd as well if it isn't already handled */
@@ -478,7 +513,7 @@ void amdgpu_irq_delegate(struct amdgpu_device *adev,
 			 struct amdgpu_iv_entry *entry,
 			 unsigned int num_dw)
 {
-	amdgpu_ih_ring_write(&adev->irq.ih_soft, entry->iv_entry, num_dw);
+	amdgpu_ih_ring_write(adev, &adev->irq.ih_soft, entry->iv_entry, num_dw);
 	schedule_work(&adev->irq.ih_soft_work);
 }
 
@@ -492,7 +527,7 @@ void amdgpu_irq_delegate(struct amdgpu_device *adev,
  * Updates interrupt state for the specific source (all ASICs).
  */
 int amdgpu_irq_update(struct amdgpu_device *adev,
-			     struct amdgpu_irq_src *src, unsigned type)
+			     struct amdgpu_irq_src *src, unsigned int type)
 {
 	unsigned long irqflags;
 	enum amdgpu_interrupt_state state;
@@ -501,7 +536,8 @@ int amdgpu_irq_update(struct amdgpu_device *adev,
 	spin_lock_irqsave(&adev->irq.lock, irqflags);
 
 	/* We need to determine after taking the lock, otherwise
-	   we might disable just enabled interrupts again */
+	 * we might disable just enabled interrupts again
+	 */
 	if (amdgpu_irq_enabled(adev, src, type))
 		state = AMDGPU_IRQ_STATE_ENABLE;
 	else
@@ -555,7 +591,7 @@ void amdgpu_irq_gpu_reset_resume_helper(struct amdgpu_device *adev)
  * 0 on success or error code otherwise
  */
 int amdgpu_irq_get(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
-		   unsigned type)
+		   unsigned int type)
 {
 	if (!adev->irq.installed)
 		return -ENOENT;
@@ -585,8 +621,12 @@ int amdgpu_irq_get(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
  * 0 on success or error code otherwise
  */
 int amdgpu_irq_put(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
-		   unsigned type)
+		   unsigned int type)
 {
+	/* When the threshold is reached,the interrupt source may not be enabled.return -EINVAL */
+	if (amdgpu_ras_is_rma(adev) && !amdgpu_irq_enabled(adev, src, type))
+		return -EINVAL;
+
 	if (!adev->irq.installed)
 		return -ENOENT;
 
@@ -596,6 +636,9 @@ int amdgpu_irq_put(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
 	if (!src->enabled_types || !src->funcs->set)
 		return -EINVAL;
 
+	if (WARN_ON(!amdgpu_irq_enabled(adev, src, type)))
+		return -EINVAL;
+
 	if (atomic_dec_and_test(&src->enabled_types[type]))
 		return amdgpu_irq_update(adev, src, type);
 
@@ -616,7 +659,7 @@ int amdgpu_irq_put(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
  * invalid parameters
  */
 bool amdgpu_irq_enabled(struct amdgpu_device *adev, struct amdgpu_irq_src *src,
-			unsigned type)
+			unsigned int type)
 {
 	if (!adev->irq.installed)
 		return false;
@@ -690,10 +733,10 @@ static const struct irq_domain_ops amdgpu_hw_irqdomain_ops = {
  */
 int amdgpu_irq_add_domain(struct amdgpu_device *adev)
 {
-	adev->irq.domain = irq_domain_add_linear(NULL, AMDGPU_MAX_IRQ_SRC_ID,
-						 &amdgpu_hw_irqdomain_ops, adev);
+	adev->irq.domain = irq_domain_create_linear(NULL, AMDGPU_MAX_IRQ_SRC_ID,
+						    &amdgpu_hw_irqdomain_ops, adev);
 	if (!adev->irq.domain) {
-		DRM_ERROR("GPU irq add domain failed\n");
+		dev_err(adev->dev, "GPU irq add domain failed\n");
 		return -ENODEV;
 	}
 
@@ -729,7 +772,7 @@ void amdgpu_irq_remove_domain(struct amdgpu_device *adev)
  * Returns:
  * Linux IRQ
  */
-unsigned amdgpu_irq_create_mapping(struct amdgpu_device *adev, unsigned src_id)
+unsigned int amdgpu_irq_create_mapping(struct amdgpu_device *adev, unsigned int src_id)
 {
 	adev->irq.virq[src_id] = irq_create_mapping(adev->irq.domain, src_id);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
index e9f2c11ea416..9f0417456abd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
@@ -53,7 +53,7 @@ struct amdgpu_iv_entry {
 	uint64_t timestamp;
 	unsigned timestamp_src;
 	unsigned pasid;
-	unsigned pasid_src;
+	unsigned node_id;
 	unsigned src_data[AMDGPU_IRQ_SRC_DATA_MAX_SIZE_DW];
 	const uint32_t *iv_entry;
 };
@@ -98,8 +98,28 @@ struct amdgpu_irq {
 	struct irq_domain		*domain; /* GPU irq controller domain */
 	unsigned			virq[AMDGPU_MAX_IRQ_SRC_ID];
 	uint32_t                        srbm_soft_reset;
+	u32                             retry_cam_doorbell_index;
+	bool                            retry_cam_enabled;
 };
 
+enum interrupt_node_id_per_aid {
+	AID0_NODEID = 0,
+	XCD0_NODEID = 1,
+	XCD1_NODEID = 2,
+	AID1_NODEID = 4,
+	XCD2_NODEID = 5,
+	XCD3_NODEID = 6,
+	AID2_NODEID = 8,
+	XCD4_NODEID = 9,
+	XCD5_NODEID = 10,
+	AID3_NODEID = 12,
+	XCD6_NODEID = 13,
+	XCD7_NODEID = 14,
+	NODEID_MAX,
+};
+
+extern const int node_id_to_phys_map[NODEID_MAX];
+
 void amdgpu_irq_disable_all(struct amdgpu_device *adev);
 
 int amdgpu_irq_init(struct amdgpu_device *adev);
@@ -126,5 +146,6 @@ void amdgpu_irq_gpu_reset_resume_helper(struct amdgpu_device *adev);
 int amdgpu_irq_add_domain(struct amdgpu_device *adev);
 void amdgpu_irq_remove_domain(struct amdgpu_device *adev);
 unsigned amdgpu_irq_create_mapping(struct amdgpu_device *adev, unsigned src_id);
+void amdgpu_restore_msix(struct amdgpu_device *adev);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_isp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_isp.c
new file mode 100644
index 000000000000..37270c4dab8d
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_isp.c
@@ -0,0 +1,345 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright (C) 2024 Advanced Micro Devices, Inc. All rights reserved.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ */
+
+#include <linux/firmware.h>
+#include <linux/mfd/core.h>
+
+#include "amdgpu.h"
+#include "amdgpu_isp.h"
+#include "isp_v4_1_0.h"
+#include "isp_v4_1_1.h"
+
+#define ISP_MC_ADDR_ALIGN (1024 * 32)
+
+/**
+ * isp_hw_init - start and test isp block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ */
+static int isp_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_isp *isp = &adev->isp;
+
+	if (isp->funcs->hw_init != NULL)
+		return isp->funcs->hw_init(isp);
+
+	return -ENODEV;
+}
+
+/**
+ * isp_hw_fini - stop the hardware block
+ *
+ * @ip_block: Pointer to the amdgpu_ip_block for this hw instance.
+ *
+ */
+static int isp_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_isp *isp = &ip_block->adev->isp;
+
+	if (isp->funcs->hw_fini != NULL)
+		return isp->funcs->hw_fini(isp);
+
+	return -ENODEV;
+}
+
+static int isp_load_fw_by_psp(struct amdgpu_device *adev)
+{
+	const struct common_firmware_header *hdr;
+	char ucode_prefix[10];
+	int r = 0;
+
+	/* get isp fw binary name and path */
+	amdgpu_ucode_ip_version_decode(adev, ISP_HWIP, ucode_prefix,
+				       sizeof(ucode_prefix));
+
+	/* read isp fw */
+	r = amdgpu_ucode_request(adev, &adev->isp.fw, AMDGPU_UCODE_OPTIONAL,
+				"amdgpu/%s.bin", ucode_prefix);
+	if (r) {
+		amdgpu_ucode_release(&adev->isp.fw);
+		return r;
+	}
+
+	hdr = (const struct common_firmware_header *)adev->isp.fw->data;
+
+	adev->firmware.ucode[AMDGPU_UCODE_ID_ISP].ucode_id =
+		AMDGPU_UCODE_ID_ISP;
+	adev->firmware.ucode[AMDGPU_UCODE_ID_ISP].fw = adev->isp.fw;
+
+	adev->firmware.fw_size +=
+		ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE);
+
+	return r;
+}
+
+static int isp_early_init(struct amdgpu_ip_block *ip_block)
+{
+
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_isp *isp = &adev->isp;
+
+	switch (amdgpu_ip_version(adev, ISP_HWIP, 0)) {
+	case IP_VERSION(4, 1, 0):
+		isp_v4_1_0_set_isp_funcs(isp);
+		break;
+	case IP_VERSION(4, 1, 1):
+		isp_v4_1_1_set_isp_funcs(isp);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	isp->adev = adev;
+	isp->parent = adev->dev;
+
+	if (isp_load_fw_by_psp(adev)) {
+		DRM_DEBUG_DRIVER("%s: isp fw load failed\n", __func__);
+		return -ENOENT;
+	}
+
+	return 0;
+}
+
+static bool isp_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	return true;
+}
+
+static int isp_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+				     enum amd_clockgating_state state)
+{
+	return 0;
+}
+
+static int isp_set_powergating_state(struct amdgpu_ip_block *ip_block,
+				     enum amd_powergating_state state)
+{
+	return 0;
+}
+
+static int is_valid_isp_device(struct device *isp_parent, struct device *amdgpu_dev)
+{
+	if (isp_parent != amdgpu_dev)
+		return -EINVAL;
+
+	return 0;
+}
+
+/**
+ * isp_user_buffer_alloc - create user buffer object (BO) for isp
+ *
+ * @dev: isp device handle
+ * @dmabuf: DMABUF handle for isp buffer allocated in system memory
+ * @buf_obj: GPU buffer object handle to initialize
+ * @buf_addr: GPU addr of the pinned BO to initialize
+ *
+ * Imports isp DMABUF to allocate and pin a user BO for isp internal use. It does
+ * GART alloc to generate GPU addr for BO to make it accessible through the
+ * GART aperture for ISP HW.
+ *
+ * This function is exported to allow the V4L2 isp device external to drm device
+ * to create and access the isp user BO.
+ *
+ * Returns:
+ * 0 on success, negative error code otherwise.
+ */
+int isp_user_buffer_alloc(struct device *dev, void *dmabuf,
+			  void **buf_obj, u64 *buf_addr)
+{
+	struct platform_device *ispdev = to_platform_device(dev);
+	const struct isp_platform_data *isp_pdata;
+	struct amdgpu_device *adev;
+	struct mfd_cell *mfd_cell;
+	struct amdgpu_bo *bo;
+	u64 gpu_addr;
+	int ret;
+
+	if (WARN_ON(!ispdev))
+		return -ENODEV;
+
+	if (WARN_ON(!buf_obj))
+		return -EINVAL;
+
+	if (WARN_ON(!buf_addr))
+		return -EINVAL;
+
+	mfd_cell = &ispdev->mfd_cell[0];
+	if (!mfd_cell)
+		return -ENODEV;
+
+	isp_pdata = mfd_cell->platform_data;
+	adev = isp_pdata->adev;
+
+	ret = is_valid_isp_device(ispdev->dev.parent, adev->dev);
+	if (ret)
+		return ret;
+
+	ret = amdgpu_bo_create_isp_user(adev, dmabuf,
+					AMDGPU_GEM_DOMAIN_GTT, &bo, &gpu_addr);
+	if (ret) {
+		drm_err(&adev->ddev, "failed to alloc gart user buffer (%d)", ret);
+		return ret;
+	}
+
+	*buf_obj = (void *)bo;
+	*buf_addr = gpu_addr;
+
+	return 0;
+}
+EXPORT_SYMBOL(isp_user_buffer_alloc);
+
+/**
+ * isp_user_buffer_free - free isp user buffer object (BO)
+ *
+ * @buf_obj: amdgpu isp user BO to free
+ *
+ * unpin and unref BO for isp internal use.
+ *
+ * This function is exported to allow the V4L2 isp device
+ * external to drm device to free the isp user BO.
+ */
+void isp_user_buffer_free(void *buf_obj)
+{
+	amdgpu_bo_free_isp_user(buf_obj);
+}
+EXPORT_SYMBOL(isp_user_buffer_free);
+
+/**
+ * isp_kernel_buffer_alloc - create kernel buffer object (BO) for isp
+ *
+ * @dev: isp device handle
+ * @size: size for the new BO
+ * @buf_obj: GPU BO handle to initialize
+ * @gpu_addr: GPU addr of the pinned BO
+ * @cpu_addr: CPU address mapping of BO
+ *
+ * Allocates and pins a kernel BO for internal isp firmware use.
+ *
+ * This function is exported to allow the V4L2 isp device
+ * external to drm device to create and access the kernel BO.
+ *
+ * Returns:
+ * 0 on success, negative error code otherwise.
+ */
+int isp_kernel_buffer_alloc(struct device *dev, u64 size,
+			    void **buf_obj, u64 *gpu_addr, void **cpu_addr)
+{
+	struct platform_device *ispdev = to_platform_device(dev);
+	struct amdgpu_bo **bo = (struct amdgpu_bo **)buf_obj;
+	const struct isp_platform_data *isp_pdata;
+	struct amdgpu_device *adev;
+	struct mfd_cell *mfd_cell;
+	int ret;
+
+	if (WARN_ON(!ispdev))
+		return -ENODEV;
+
+	if (WARN_ON(!buf_obj))
+		return -EINVAL;
+
+	if (WARN_ON(!gpu_addr))
+		return -EINVAL;
+
+	if (WARN_ON(!cpu_addr))
+		return -EINVAL;
+
+	mfd_cell = &ispdev->mfd_cell[0];
+	if (!mfd_cell)
+		return -ENODEV;
+
+	isp_pdata = mfd_cell->platform_data;
+	adev = isp_pdata->adev;
+
+	ret = is_valid_isp_device(ispdev->dev.parent, adev->dev);
+	if (ret)
+		return ret;
+
+	/* Ensure *bo is NULL so a new BO will be created */
+	*bo = NULL;
+	ret = amdgpu_bo_create_kernel(adev,
+				      size,
+				      ISP_MC_ADDR_ALIGN,
+				      AMDGPU_GEM_DOMAIN_GTT,
+				      bo,
+				      gpu_addr,
+				      cpu_addr);
+	if (!cpu_addr || ret) {
+		drm_err(&adev->ddev, "failed to alloc gart kernel buffer (%d)", ret);
+		return ret;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL(isp_kernel_buffer_alloc);
+
+/**
+ * isp_kernel_buffer_free - free isp kernel buffer object (BO)
+ *
+ * @buf_obj: amdgpu isp user BO to free
+ * @gpu_addr: GPU addr of isp kernel BO
+ * @cpu_addr: CPU addr of isp kernel BO
+ *
+ * unmaps and unpin a isp kernel BO.
+ *
+ * This function is exported to allow the V4L2 isp device
+ * external to drm device to free the kernel BO.
+ */
+void isp_kernel_buffer_free(void **buf_obj, u64 *gpu_addr, void **cpu_addr)
+{
+	struct amdgpu_bo **bo = (struct amdgpu_bo **)buf_obj;
+
+	amdgpu_bo_free_kernel(bo, gpu_addr, cpu_addr);
+}
+EXPORT_SYMBOL(isp_kernel_buffer_free);
+
+static const struct amd_ip_funcs isp_ip_funcs = {
+	.name = "isp_ip",
+	.early_init = isp_early_init,
+	.hw_init = isp_hw_init,
+	.hw_fini = isp_hw_fini,
+	.is_idle = isp_is_idle,
+	.set_clockgating_state = isp_set_clockgating_state,
+	.set_powergating_state = isp_set_powergating_state,
+};
+
+const struct amdgpu_ip_block_version isp_v4_1_0_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_ISP,
+	.major = 4,
+	.minor = 1,
+	.rev = 0,
+	.funcs = &isp_ip_funcs,
+};
+
+const struct amdgpu_ip_block_version isp_v4_1_1_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_ISP,
+	.major = 4,
+	.minor = 1,
+	.rev = 1,
+	.funcs = &isp_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_isp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_isp.h
new file mode 100644
index 000000000000..d6f4ffa4c97c
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_isp.h
@@ -0,0 +1,60 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright (C) 2024 Advanced Micro Devices, Inc. All rights reserved.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ */
+
+#ifndef __AMDGPU_ISP_H__
+#define __AMDGPU_ISP_H__
+
+#include <drm/amd/isp.h>
+#include <linux/pm_domain.h>
+
+#define ISP_REGS_OFFSET_END 0x629A4
+
+struct amdgpu_isp;
+
+struct isp_funcs {
+	int (*hw_init)(struct amdgpu_isp *isp);
+	int (*hw_fini)(struct amdgpu_isp *isp);
+};
+
+struct amdgpu_isp {
+	struct device *parent;
+	struct amdgpu_device	*adev;
+	const struct isp_funcs	*funcs;
+	struct mfd_cell *isp_cell;
+	struct resource *isp_res;
+	struct resource *isp_i2c_res;
+	struct resource *isp_gpio_res;
+	struct isp_platform_data *isp_pdata;
+	unsigned int harvest_config;
+	const struct firmware	*fw;
+	struct generic_pm_domain ispgpd;
+};
+
+extern const struct amdgpu_ip_block_version isp_v4_1_0_ip_block;
+extern const struct amdgpu_ip_block_version isp_v4_1_1_ip_block;
+
+#endif /* __AMDGPU_ISP_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
index c3d9d75143f4..0a0dcbf0798d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
@@ -30,40 +30,124 @@
 #include "amdgpu.h"
 #include "amdgpu_trace.h"
 #include "amdgpu_reset.h"
+#include "amdgpu_dev_coredump.h"
+#include "amdgpu_xgmi.h"
+
+static void amdgpu_job_do_core_dump(struct amdgpu_device *adev,
+				    struct amdgpu_job *job)
+{
+	int i;
+
+	dev_info(adev->dev, "Dumping IP State\n");
+	for (i = 0; i < adev->num_ip_blocks; i++)
+		if (adev->ip_blocks[i].version->funcs->dump_ip_state)
+			adev->ip_blocks[i].version->funcs
+				->dump_ip_state((void *)&adev->ip_blocks[i]);
+	dev_info(adev->dev, "Dumping IP State Completed\n");
+
+	amdgpu_coredump(adev, true, false, job);
+}
+
+static void amdgpu_job_core_dump(struct amdgpu_device *adev,
+				 struct amdgpu_job *job)
+{
+	struct list_head device_list, *device_list_handle =  NULL;
+	struct amdgpu_device *tmp_adev = NULL;
+	struct amdgpu_hive_info *hive = NULL;
+
+	if (!amdgpu_sriov_vf(adev))
+		hive = amdgpu_get_xgmi_hive(adev);
+	if (hive)
+		mutex_lock(&hive->hive_lock);
+	/*
+	 * Reuse the logic in amdgpu_device_gpu_recover() to build list of
+	 * devices for code dump
+	 */
+	INIT_LIST_HEAD(&device_list);
+	if (!amdgpu_sriov_vf(adev) && (adev->gmc.xgmi.num_physical_nodes > 1) && hive) {
+		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head)
+			list_add_tail(&tmp_adev->reset_list, &device_list);
+		if (!list_is_first(&adev->reset_list, &device_list))
+			list_rotate_to_front(&adev->reset_list, &device_list);
+		device_list_handle = &device_list;
+	} else {
+		list_add_tail(&adev->reset_list, &device_list);
+		device_list_handle = &device_list;
+	}
+
+	/* Do the coredump for each device */
+	list_for_each_entry(tmp_adev, device_list_handle, reset_list)
+		amdgpu_job_do_core_dump(tmp_adev, job);
+
+	if (hive) {
+		mutex_unlock(&hive->hive_lock);
+		amdgpu_put_xgmi_hive(hive);
+	}
+}
 
 static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 {
 	struct amdgpu_ring *ring = to_amdgpu_ring(s_job->sched);
 	struct amdgpu_job *job = to_amdgpu_job(s_job);
-	struct amdgpu_task_info ti;
+	struct drm_wedge_task_info *info = NULL;
+	struct amdgpu_task_info *ti = NULL;
 	struct amdgpu_device *adev = ring->adev;
-	int idx;
-	int r;
+	int idx, r;
 
 	if (!drm_dev_enter(adev_to_drm(adev), &idx)) {
-		DRM_INFO("%s - device unplugged skipping recovery on scheduler:%s",
+		dev_info(adev->dev, "%s - device unplugged skipping recovery on scheduler:%s",
 			 __func__, s_job->sched->name);
 
 		/* Effectively the job is aborted as the device is gone */
 		return DRM_GPU_SCHED_STAT_ENODEV;
 	}
 
-	memset(&ti, 0, sizeof(struct amdgpu_task_info));
-	adev->job_hang = true;
+	/*
+	 * Do the coredump immediately after a job timeout to get a very
+	 * close dump/snapshot/representation of GPU's current error status
+	 * Skip it for SRIOV, since VF FLR will be triggered by host driver
+	 * before job timeout
+	 */
+	if (!amdgpu_sriov_vf(adev))
+		amdgpu_job_core_dump(adev, job);
 
 	if (amdgpu_gpu_recovery &&
+	    amdgpu_ring_is_reset_type_supported(ring, AMDGPU_RESET_TYPE_SOFT_RESET) &&
 	    amdgpu_ring_soft_recovery(ring, job->vmid, s_job->s_fence->parent)) {
-		DRM_ERROR("ring %s timeout, but soft recovered\n",
-			  s_job->sched->name);
+		dev_err(adev->dev, "ring %s timeout, but soft recovered\n",
+			s_job->sched->name);
 		goto exit;
 	}
 
-	amdgpu_vm_get_task_info(ring->adev, job->pasid, &ti);
-	DRM_ERROR("ring %s timeout, signaled seq=%u, emitted seq=%u\n",
-		  job->base.sched->name, atomic_read(&ring->fence_drv.last_seq),
-		  ring->fence_drv.sync_seq);
-	DRM_ERROR("Process information: process %s pid %d thread %s pid %d\n",
-		  ti.process_name, ti.tgid, ti.task_name, ti.pid);
+	dev_err(adev->dev, "ring %s timeout, signaled seq=%u, emitted seq=%u\n",
+		job->base.sched->name, atomic_read(&ring->fence_drv.last_seq),
+		ring->fence_drv.sync_seq);
+
+	ti = amdgpu_vm_get_task_info_pasid(ring->adev, job->pasid);
+	if (ti) {
+		amdgpu_vm_print_task_info(adev, ti);
+		info = &ti->task;
+	}
+
+	/* attempt a per ring reset */
+	if (amdgpu_gpu_recovery &&
+	    amdgpu_ring_is_reset_type_supported(ring, AMDGPU_RESET_TYPE_PER_QUEUE) &&
+	    ring->funcs->reset) {
+		dev_err(adev->dev, "Starting %s ring reset\n",
+			s_job->sched->name);
+		r = amdgpu_ring_reset(ring, job->vmid, job->hw_fence);
+		if (!r) {
+			atomic_inc(&ring->adev->gpu_reset_counter);
+			dev_err(adev->dev, "Ring %s reset succeeded\n",
+				ring->sched.name);
+			drm_dev_wedged_event(adev_to_drm(adev),
+					     DRM_WEDGE_RECOVERY_NONE, info);
+			goto exit;
+		}
+		dev_err(adev->dev, "Ring %s reset failed\n", ring->sched.name);
+	}
+
+	dma_fence_set_error(&s_job->s_fence->finished, -ETIME);
 
 	if (amdgpu_device_should_recover_gpu(ring->adev)) {
 		struct amdgpu_reset_context reset_context;
@@ -71,11 +155,18 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 
 		reset_context.method = AMD_RESET_METHOD_NONE;
 		reset_context.reset_req_dev = adev;
+		reset_context.src = AMDGPU_RESET_SRC_JOB;
 		clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
 
+		/*
+		 * To avoid an unnecessary extra coredump, as we have already
+		 * got the very close representation of GPU's error status
+		 */
+		set_bit(AMDGPU_SKIP_COREDUMP, &reset_context.flags);
+
 		r = amdgpu_device_gpu_recover(ring->adev, job, &reset_context);
 		if (r)
-			DRM_ERROR("GPU Recovery Failed: %d\n", r);
+			dev_err(adev->dev, "GPU Recovery Failed: %d\n", r);
 	} else {
 		drm_sched_suspend_timeout(&ring->sched);
 		if (amdgpu_sriov_vf(adev))
@@ -83,15 +174,19 @@ static enum drm_gpu_sched_stat amdgpu_job_timedout(struct drm_sched_job *s_job)
 	}
 
 exit:
-	adev->job_hang = false;
+	amdgpu_vm_put_task_info(ti);
 	drm_dev_exit(idx);
-	return DRM_GPU_SCHED_STAT_NOMINAL;
+	return DRM_GPU_SCHED_STAT_RESET;
 }
 
 int amdgpu_job_alloc(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		     struct drm_sched_entity *entity, void *owner,
-		     unsigned int num_ibs, struct amdgpu_job **job)
+		     unsigned int num_ibs, struct amdgpu_job **job,
+		     u64 drm_client_id)
 {
+	struct amdgpu_fence *af;
+	int r;
+
 	if (num_ibs == 0)
 		return -EINVAL;
 
@@ -99,31 +194,50 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	if (!*job)
 		return -ENOMEM;
 
-	/*
-	 * Initialize the scheduler to at least some ring so that we always
-	 * have a pointer to adev.
-	 */
-	(*job)->base.sched = &adev->rings[0]->sched;
+	af = kzalloc(sizeof(struct amdgpu_fence), GFP_KERNEL);
+	if (!af) {
+		r = -ENOMEM;
+		goto err_job;
+	}
+	(*job)->hw_fence = af;
+
+	af = kzalloc(sizeof(struct amdgpu_fence), GFP_KERNEL);
+	if (!af) {
+		r = -ENOMEM;
+		goto err_fence;
+	}
+	(*job)->hw_vm_fence = af;
+
 	(*job)->vm = vm;
 
 	amdgpu_sync_create(&(*job)->explicit_sync);
-	(*job)->vram_lost_counter = atomic_read(&adev->vram_lost_counter);
+	(*job)->generation = amdgpu_vm_generation(adev, vm);
 	(*job)->vm_pd_addr = AMDGPU_BO_INVALID_OFFSET;
 
 	if (!entity)
 		return 0;
 
-	return drm_sched_job_init(&(*job)->base, entity, owner);
+	return drm_sched_job_init(&(*job)->base, entity, 1, owner,
+				  drm_client_id);
+
+err_fence:
+	kfree((*job)->hw_fence);
+err_job:
+	kfree(*job);
+	*job = NULL;
+
+	return r;
 }
 
 int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev,
 			     struct drm_sched_entity *entity, void *owner,
 			     size_t size, enum amdgpu_ib_pool_type pool_type,
-			     struct amdgpu_job **job)
+			     struct amdgpu_job **job, u64 k_job_id)
 {
 	int r;
 
-	r = amdgpu_job_alloc(adev, NULL, entity, owner, 1, job);
+	r = amdgpu_job_alloc(adev, NULL, entity, owner, 1, job,
+			     k_job_id);
 	if (r)
 		return r;
 
@@ -132,7 +246,10 @@ int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev,
 	if (r) {
 		if (entity)
 			drm_sched_job_cleanup(&(*job)->base);
+		kfree((*job)->hw_vm_fence);
+		kfree((*job)->hw_fence);
 		kfree(*job);
+		*job = NULL;
 	}
 
 	return r;
@@ -157,20 +274,19 @@ void amdgpu_job_set_resources(struct amdgpu_job *job, struct amdgpu_bo *gds,
 
 void amdgpu_job_free_resources(struct amdgpu_job *job)
 {
-	struct amdgpu_ring *ring = to_amdgpu_ring(job->base.sched);
 	struct dma_fence *f;
 	unsigned i;
 
-	/* Check if any fences where initialized */
+	/* Check if any fences were initialized */
 	if (job->base.s_fence && job->base.s_fence->finished.ops)
 		f = &job->base.s_fence->finished;
-	else if (job->hw_fence.ops)
-		f = &job->hw_fence;
+	else if (job->hw_fence && job->hw_fence->base.ops)
+		f = &job->hw_fence->base;
 	else
 		f = NULL;
 
 	for (i = 0; i < job->num_ibs; ++i)
-		amdgpu_ib_free(ring->adev, &job->ibs[i], f);
+		amdgpu_ib_free(&job->ibs[i], f);
 }
 
 static void amdgpu_job_free_cb(struct drm_sched_job *s_job)
@@ -181,11 +297,16 @@ static void amdgpu_job_free_cb(struct drm_sched_job *s_job)
 
 	amdgpu_sync_free(&job->explicit_sync);
 
-	/* only put the hw fence if has embedded fence */
-	if (!job->hw_fence.ops)
-		kfree(job);
+	if (job->hw_fence->base.ops)
+		dma_fence_put(&job->hw_fence->base);
 	else
-		dma_fence_put(&job->hw_fence);
+		kfree(job->hw_fence);
+	if (job->hw_vm_fence->base.ops)
+		dma_fence_put(&job->hw_vm_fence->base);
+	else
+		kfree(job->hw_vm_fence);
+
+	kfree(job);
 }
 
 void amdgpu_job_set_gang_leader(struct amdgpu_job *job,
@@ -214,10 +335,16 @@ void amdgpu_job_free(struct amdgpu_job *job)
 	if (job->gang_submit != &job->base.s_fence->scheduled)
 		dma_fence_put(job->gang_submit);
 
-	if (!job->hw_fence.ops)
-		kfree(job);
+	if (job->hw_fence->base.ops)
+		dma_fence_put(&job->hw_fence->base);
 	else
-		dma_fence_put(&job->hw_fence);
+		kfree(job->hw_fence);
+	if (job->hw_vm_fence->base.ops)
+		dma_fence_put(&job->hw_vm_fence->base);
+	else
+		kfree(job->hw_vm_fence);
+
+	kfree(job);
 }
 
 struct dma_fence *amdgpu_job_submit(struct amdgpu_job *job)
@@ -253,19 +380,37 @@ amdgpu_job_prepare_job(struct drm_sched_job *sched_job,
 {
 	struct amdgpu_ring *ring = to_amdgpu_ring(s_entity->rq->sched);
 	struct amdgpu_job *job = to_amdgpu_job(sched_job);
-	struct dma_fence *fence = NULL;
+	struct dma_fence *fence;
 	int r;
 
-	if (!fence && job->gang_submit)
+	r = drm_sched_entity_error(s_entity);
+	if (r)
+		goto error;
+
+	if (job->gang_submit) {
 		fence = amdgpu_device_switch_gang(ring->adev, job->gang_submit);
+		if (fence)
+			return fence;
+	}
 
-	while (!fence && job->vm && !job->vmid) {
+	fence = amdgpu_device_enforce_isolation(ring->adev, ring, job);
+	if (fence)
+		return fence;
+
+	if (job->vm && !job->vmid) {
 		r = amdgpu_vmid_grab(job->vm, ring, job, &fence);
-		if (r)
-			DRM_ERROR("Error getting VM ID (%d)\n", r);
+		if (r) {
+			dev_err(ring->adev->dev, "Error getting VM ID (%d)\n", r);
+			goto error;
+		}
+		return fence;
 	}
 
-	return fence;
+	return NULL;
+
+error:
+	dma_fence_set_error(&job->base.s_fence->finished, r);
+	return NULL;
 }
 
 static struct dma_fence *amdgpu_job_run(struct drm_sched_job *sched_job)
@@ -282,17 +427,20 @@ static struct dma_fence *amdgpu_job_run(struct drm_sched_job *sched_job)
 	trace_amdgpu_sched_run_job(job);
 
 	/* Skip job if VRAM is lost and never resubmit gangs */
-	if (job->vram_lost_counter != atomic_read(&adev->vram_lost_counter) ||
+	if (job->generation != amdgpu_vm_generation(adev, job->vm) ||
 	    (job->job_run_counter && job->gang_submit))
 		dma_fence_set_error(finished, -ECANCELED);
 
 	if (finished->error < 0) {
-		DRM_INFO("Skip scheduling IBs!\n");
+		dev_dbg(adev->dev, "Skip scheduling IBs in ring(%s)",
+			ring->name);
 	} else {
 		r = amdgpu_ib_schedule(ring, job->num_ibs, job->ibs, job,
 				       &fence);
 		if (r)
-			DRM_ERROR("Error scheduling IBs (%d)\n", r);
+			dev_err(adev->dev,
+				"Error scheduling IBs (%d) in ring(%s)", r,
+				ring->name);
 	}
 
 	job->job_run_counter++;
@@ -302,8 +450,24 @@ static struct dma_fence *amdgpu_job_run(struct drm_sched_job *sched_job)
 	return fence;
 }
 
-#define to_drm_sched_job(sched_job)		\
-		container_of((sched_job), struct drm_sched_job, queue_node)
+/*
+ * This is a duplicate function from DRM scheduler sched_internal.h.
+ * Plan is to remove it when amdgpu_job_stop_all_jobs_on_sched is removed, due
+ * latter being incorrect and racy.
+ *
+ * See https://lore.kernel.org/amd-gfx/44edde63-7181-44fb-a4f7-94e50514f539@amd.com/
+ */
+static struct drm_sched_job *
+drm_sched_entity_queue_pop(struct drm_sched_entity *entity)
+{
+	struct spsc_node *node;
+
+	node = spsc_queue_pop(&entity->job_queue);
+	if (!node)
+		return NULL;
+
+	return container_of(node, struct drm_sched_job, queue_node);
+}
 
 void amdgpu_job_stop_all_jobs_on_sched(struct drm_gpu_scheduler *sched)
 {
@@ -312,11 +476,11 @@ void amdgpu_job_stop_all_jobs_on_sched(struct drm_gpu_scheduler *sched)
 	int i;
 
 	/* Signal all jobs not yet scheduled */
-	for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
-		struct drm_sched_rq *rq = &sched->sched_rq[i];
+	for (i = DRM_SCHED_PRIORITY_KERNEL; i < sched->num_rqs; i++) {
+		struct drm_sched_rq *rq = sched->sched_rq[i];
 		spin_lock(&rq->lock);
 		list_for_each_entry(s_entity, &rq->entities, list) {
-			while ((s_job = to_drm_sched_job(spsc_queue_pop(&s_entity->job_queue)))) {
+			while ((s_job = drm_sched_entity_queue_pop(s_entity))) {
 				struct drm_sched_fence *s_fence = s_job->s_fence;
 
 				dma_fence_signal(&s_fence->scheduled);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h
index 52f2e313ea17..7abf069d17d4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h
@@ -44,11 +44,28 @@
 struct amdgpu_fence;
 enum amdgpu_ib_pool_type;
 
+/* Internal kernel job ids. (decreasing values, starting from U64_MAX). */
+#define AMDGPU_KERNEL_JOB_ID_VM_UPDATE              (18446744073709551615ULL)
+#define AMDGPU_KERNEL_JOB_ID_VM_UPDATE_PDES         (18446744073709551614ULL)
+#define AMDGPU_KERNEL_JOB_ID_VM_UPDATE_RANGE        (18446744073709551613ULL)
+#define AMDGPU_KERNEL_JOB_ID_VM_PT_CLEAR            (18446744073709551612ULL)
+#define AMDGPU_KERNEL_JOB_ID_TTM_MAP_BUFFER         (18446744073709551611ULL)
+#define AMDGPU_KERNEL_JOB_ID_TTM_ACCESS_MEMORY_SDMA (18446744073709551610ULL)
+#define AMDGPU_KERNEL_JOB_ID_TTM_COPY_BUFFER        (18446744073709551609ULL)
+#define AMDGPU_KERNEL_JOB_ID_CLEAR_ON_RELEASE       (18446744073709551608ULL)
+#define AMDGPU_KERNEL_JOB_ID_MOVE_BLIT              (18446744073709551607ULL)
+#define AMDGPU_KERNEL_JOB_ID_TTM_CLEAR_BUFFER       (18446744073709551606ULL)
+#define AMDGPU_KERNEL_JOB_ID_CLEANER_SHADER         (18446744073709551605ULL)
+#define AMDGPU_KERNEL_JOB_ID_FLUSH_GPU_TLB          (18446744073709551604ULL)
+#define AMDGPU_KERNEL_JOB_ID_KFD_GART_MAP           (18446744073709551603ULL)
+#define AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST          (18446744073709551602ULL)
+
 struct amdgpu_job {
 	struct drm_sched_job    base;
 	struct amdgpu_vm	*vm;
 	struct amdgpu_sync	explicit_sync;
-	struct dma_fence	hw_fence;
+	struct amdgpu_fence	*hw_fence;
+	struct amdgpu_fence	*hw_vm_fence;
 	struct dma_fence	*gang_submit;
 	uint32_t		preamble_status;
 	uint32_t                preemption_status;
@@ -61,15 +78,25 @@ struct amdgpu_job {
 	uint32_t		gds_base, gds_size;
 	uint32_t		gws_base, gws_size;
 	uint32_t		oa_base, oa_size;
-	uint32_t		vram_lost_counter;
+	uint64_t		generation;
 
 	/* user fence handling */
 	uint64_t		uf_addr;
 	uint64_t		uf_sequence;
 
+	/* virtual addresses for shadow/GDS/CSA */
+	uint64_t		shadow_va;
+	uint64_t		csa_va;
+	uint64_t		gds_va;
+	bool			init_shadow;
+
 	/* job_run_counter >= 1 means a resubmit job */
 	uint32_t		job_run_counter;
 
+	/* enforce isolation */
+	bool			enforce_isolation;
+	bool			run_cleaner_shader;
+
 	uint32_t		num_ibs;
 	struct amdgpu_ib	ibs[];
 };
@@ -81,11 +108,13 @@ static inline struct amdgpu_ring *amdgpu_job_ring(struct amdgpu_job *job)
 
 int amdgpu_job_alloc(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		     struct drm_sched_entity *entity, void *owner,
-		     unsigned int num_ibs, struct amdgpu_job **job);
+		     unsigned int num_ibs, struct amdgpu_job **job,
+		     u64 drm_client_id);
 int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev,
 			     struct drm_sched_entity *entity, void *owner,
 			     size_t size, enum amdgpu_ib_pool_type pool_type,
-			     struct amdgpu_job **job);
+			     struct amdgpu_job **job,
+			     u64 k_job_id);
 void amdgpu_job_set_resources(struct amdgpu_job *job, struct amdgpu_bo *gds,
 			      struct amdgpu_bo *gws, struct amdgpu_bo *oa);
 void amdgpu_job_free_resources(struct amdgpu_job *job);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.c
index b07c000fc8ba..63ee6ba6a931 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.c
@@ -33,27 +33,62 @@
 #define JPEG_IDLE_TIMEOUT	msecs_to_jiffies(1000)
 
 static void amdgpu_jpeg_idle_work_handler(struct work_struct *work);
+static void amdgpu_jpeg_reg_dump_fini(struct amdgpu_device *adev);
 
 int amdgpu_jpeg_sw_init(struct amdgpu_device *adev)
 {
+	int i, r;
+
 	INIT_DELAYED_WORK(&adev->jpeg.idle_work, amdgpu_jpeg_idle_work_handler);
 	mutex_init(&adev->jpeg.jpeg_pg_lock);
 	atomic_set(&adev->jpeg.total_submission_cnt, 0);
 
+	if ((adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) &&
+	    (adev->pg_flags & AMD_PG_SUPPORT_JPEG_DPG))
+		adev->jpeg.indirect_sram = true;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; i++) {
+		if (adev->jpeg.harvest_config & (1U << i))
+			continue;
+
+		if (adev->jpeg.indirect_sram) {
+			r = amdgpu_bo_create_kernel(adev, 64 * 2 * 4, PAGE_SIZE,
+					AMDGPU_GEM_DOMAIN_VRAM |
+					AMDGPU_GEM_DOMAIN_GTT,
+					&adev->jpeg.inst[i].dpg_sram_bo,
+					&adev->jpeg.inst[i].dpg_sram_gpu_addr,
+					&adev->jpeg.inst[i].dpg_sram_cpu_addr);
+			if (r) {
+				dev_err(adev->dev,
+				"JPEG %d (%d) failed to allocate DPG bo\n", i, r);
+				return r;
+			}
+		}
+	}
+
 	return 0;
 }
 
 int amdgpu_jpeg_sw_fini(struct amdgpu_device *adev)
 {
-	int i;
+	int i, j;
 
 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
-		if (adev->jpeg.harvest_config & (1 << i))
+		if (adev->jpeg.harvest_config & (1U << i))
 			continue;
 
-		amdgpu_ring_fini(&adev->jpeg.inst[i].ring_dec);
+		amdgpu_bo_free_kernel(
+			&adev->jpeg.inst[i].dpg_sram_bo,
+			&adev->jpeg.inst[i].dpg_sram_gpu_addr,
+			(void **)&adev->jpeg.inst[i].dpg_sram_cpu_addr);
+
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j)
+			amdgpu_ring_fini(&adev->jpeg.inst[i].ring_dec[j]);
 	}
 
+	if (adev->jpeg.reg_list)
+		amdgpu_jpeg_reg_dump_fini(adev);
+
 	mutex_destroy(&adev->jpeg.jpeg_pg_lock);
 
 	return 0;
@@ -76,19 +111,22 @@ static void amdgpu_jpeg_idle_work_handler(struct work_struct *work)
 	struct amdgpu_device *adev =
 		container_of(work, struct amdgpu_device, jpeg.idle_work.work);
 	unsigned int fences = 0;
-	unsigned int i;
+	unsigned int i, j;
 
 	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
-		if (adev->jpeg.harvest_config & (1 << i))
+		if (adev->jpeg.harvest_config & (1U << i))
 			continue;
 
-		fences += amdgpu_fence_count_emitted(&adev->jpeg.inst[i].ring_dec);
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j)
+			fences += amdgpu_fence_count_emitted(&adev->jpeg.inst[i].ring_dec[j]);
 	}
 
-	if (!fences && !atomic_read(&adev->jpeg.total_submission_cnt))
+	if (!fences && !atomic_read(&adev->jpeg.total_submission_cnt)) {
+		mutex_lock(&adev->jpeg.jpeg_pg_lock);
 		amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_JPEG,
 						       AMD_PG_STATE_GATE);
-	else
+		mutex_unlock(&adev->jpeg.jpeg_pg_lock);
+	} else
 		schedule_delayed_work(&adev->jpeg.idle_work, JPEG_IDLE_TIMEOUT);
 }
 
@@ -122,18 +160,21 @@ int amdgpu_jpeg_dec_ring_test_ring(struct amdgpu_ring *ring)
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	WREG32(adev->jpeg.inst[ring->me].external.jpeg_pitch, 0xCAFEDEAD);
 	r = amdgpu_ring_alloc(ring, 3);
 	if (r)
 		return r;
 
-	amdgpu_ring_write(ring, PACKET0(adev->jpeg.internal.jpeg_pitch, 0));
-	amdgpu_ring_write(ring, 0xDEADBEEF);
+	WREG32(adev->jpeg.inst[ring->me].external.jpeg_pitch[ring->pipe], 0xCAFEDEAD);
+	/* Add a read register to make sure the write register is executed. */
+	RREG32(adev->jpeg.inst[ring->me].external.jpeg_pitch[ring->pipe]);
+
+	amdgpu_ring_write(ring, PACKET0(adev->jpeg.internal.jpeg_pitch[ring->pipe], 0));
+	amdgpu_ring_write(ring, 0xABADCAFE);
 	amdgpu_ring_commit(ring);
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(adev->jpeg.inst[ring->me].external.jpeg_pitch);
-		if (tmp == 0xDEADBEEF)
+		tmp = RREG32(adev->jpeg.inst[ring->me].external.jpeg_pitch[ring->pipe]);
+		if (tmp == 0xABADCAFE)
 			break;
 		udelay(1);
 	}
@@ -155,14 +196,14 @@ static int amdgpu_jpeg_dec_set_reg(struct amdgpu_ring *ring, uint32_t handle,
 	int i, r;
 
 	r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL, ib_size_dw * 4,
-				     AMDGPU_IB_POOL_DIRECT, &job);
+				     AMDGPU_IB_POOL_DIRECT, &job,
+				     AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		return r;
 
 	ib = &job->ibs[0];
 
-	ib->ptr[0] = PACKETJ(adev->jpeg.internal.jpeg_pitch, 0, 0,
-			     PACKETJ_TYPE0);
+	ib->ptr[0] = PACKETJ(adev->jpeg.internal.jpeg_pitch[ring->pipe], 0, 0, PACKETJ_TYPE0);
 	ib->ptr[1] = 0xDEADBEEF;
 	for (i = 2; i < 16; i += 2) {
 		ib->ptr[i] = PACKETJ(0, 0, 0, PACKETJ_TYPE6);
@@ -206,12 +247,15 @@ int amdgpu_jpeg_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	} else {
 		r = 0;
 	}
+
 	if (!amdgpu_sriov_vf(adev)) {
 		for (i = 0; i < adev->usec_timeout; i++) {
-			tmp = RREG32(adev->jpeg.inst[ring->me].external.jpeg_pitch);
+			tmp = RREG32(adev->jpeg.inst[ring->me].external.jpeg_pitch[ring->pipe]);
 			if (tmp == 0xDEADBEEF)
 				break;
 			udelay(1);
+			if (amdgpu_emu_mode == 1)
+				udelay(10);
 		}
 
 		if (i >= adev->usec_timeout)
@@ -241,6 +285,32 @@ int amdgpu_jpeg_process_poison_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
+int amdgpu_jpeg_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
+{
+	int r, i;
+
+	r = amdgpu_ras_block_late_init(adev, ras_block);
+	if (r)
+		return r;
+
+	if (amdgpu_ras_is_supported(adev, ras_block->block)) {
+		for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+			if (adev->jpeg.harvest_config & (1 << i) ||
+			    !adev->jpeg.inst[i].ras_poison_irq.funcs)
+				continue;
+
+			r = amdgpu_irq_get(adev, &adev->jpeg.inst[i].ras_poison_irq, 0);
+			if (r)
+				goto late_fini;
+		}
+	}
+	return 0;
+
+late_fini:
+	amdgpu_ras_block_late_fini(adev, ras_block);
+	return r;
+}
+
 int amdgpu_jpeg_ras_sw_init(struct amdgpu_device *adev)
 {
 	int err;
@@ -262,7 +332,276 @@ int amdgpu_jpeg_ras_sw_init(struct amdgpu_device *adev)
 	adev->jpeg.ras_if = &ras->ras_block.ras_comm;
 
 	if (!ras->ras_block.ras_late_init)
-		ras->ras_block.ras_late_init = amdgpu_ras_block_late_init;
+		ras->ras_block.ras_late_init = amdgpu_jpeg_ras_late_init;
+
+	return 0;
+}
+
+int amdgpu_jpeg_psp_update_sram(struct amdgpu_device *adev, int inst_idx,
+			       enum AMDGPU_UCODE_ID ucode_id)
+{
+	struct amdgpu_firmware_info ucode = {
+		.ucode_id = AMDGPU_UCODE_ID_JPEG_RAM,
+		.mc_addr = adev->jpeg.inst[inst_idx].dpg_sram_gpu_addr,
+		.ucode_size = ((uintptr_t)adev->jpeg.inst[inst_idx].dpg_sram_curr_addr -
+			      (uintptr_t)adev->jpeg.inst[inst_idx].dpg_sram_cpu_addr),
+	};
+
+	return psp_execute_ip_fw_load(&adev->psp, &ucode);
+}
+
+/*
+ * debugfs for to enable/disable jpeg job submission to specific core.
+ */
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_debugfs_jpeg_sched_mask_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i, j;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+
+	mask = (1ULL << (adev->jpeg.num_jpeg_inst * adev->jpeg.num_jpeg_rings)) - 1;
+	if ((val & mask) == 0)
+		return -EINVAL;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+			ring = &adev->jpeg.inst[i].ring_dec[j];
+			if (val & (BIT_ULL((i * adev->jpeg.num_jpeg_rings) + j)))
+				ring->sched.ready = true;
+			else
+				ring->sched.ready = false;
+		}
+	}
+	/* publish sched.ready flag update effective immediately across smp */
+	smp_rmb();
+	return 0;
+}
+
+static int amdgpu_debugfs_jpeg_sched_mask_get(void *data, u64 *val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i, j;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) {
+		for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) {
+			ring = &adev->jpeg.inst[i].ring_dec[j];
+			if (ring->sched.ready)
+				mask |= 1ULL << ((i * adev->jpeg.num_jpeg_rings) + j);
+		}
+	}
+	*val = mask;
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_debugfs_jpeg_sched_mask_fops,
+			 amdgpu_debugfs_jpeg_sched_mask_get,
+			 amdgpu_debugfs_jpeg_sched_mask_set, "%llx\n");
+
+#endif
+
+void amdgpu_debugfs_jpeg_sched_mask_init(struct amdgpu_device *adev)
+{
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+	char name[32];
+
+	if (!(adev->jpeg.num_jpeg_inst > 1) && !(adev->jpeg.num_jpeg_rings > 1))
+		return;
+	sprintf(name, "amdgpu_jpeg_sched_mask");
+	debugfs_create_file(name, 0600, root, adev,
+			    &amdgpu_debugfs_jpeg_sched_mask_fops);
+#endif
+}
+
+static ssize_t amdgpu_get_jpeg_reset_mask(struct device *dev,
+						struct device_attribute *attr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (!adev)
+		return -ENODEV;
+
+	return amdgpu_show_reset_mask(buf, adev->jpeg.supported_reset);
+}
+
+static DEVICE_ATTR(jpeg_reset_mask, 0444,
+		   amdgpu_get_jpeg_reset_mask, NULL);
+
+int amdgpu_jpeg_sysfs_reset_mask_init(struct amdgpu_device *adev)
+{
+	int r = 0;
+
+	if (adev->jpeg.num_jpeg_inst) {
+		r = device_create_file(adev->dev, &dev_attr_jpeg_reset_mask);
+		if (r)
+			return r;
+	}
+
+	return r;
+}
+
+void amdgpu_jpeg_sysfs_reset_mask_fini(struct amdgpu_device *adev)
+{
+	if (adev->dev->kobj.sd) {
+		if (adev->jpeg.num_jpeg_inst)
+			device_remove_file(adev->dev, &dev_attr_jpeg_reset_mask);
+	}
+}
+
+int amdgpu_jpeg_reg_dump_init(struct amdgpu_device *adev,
+			       const struct amdgpu_hwip_reg_entry *reg, u32 count)
+{
+	adev->jpeg.ip_dump = kcalloc(adev->jpeg.num_jpeg_inst * count,
+				     sizeof(uint32_t), GFP_KERNEL);
+	if (!adev->jpeg.ip_dump) {
+		dev_err(adev->dev,
+			"Failed to allocate memory for JPEG IP Dump\n");
+		return -ENOMEM;
+	}
+	adev->jpeg.reg_list = reg;
+	adev->jpeg.reg_count = count;
+
+	return 0;
+}
+
+static void amdgpu_jpeg_reg_dump_fini(struct amdgpu_device *adev)
+{
+	kfree(adev->jpeg.ip_dump);
+	adev->jpeg.reg_list = NULL;
+	adev->jpeg.reg_count = 0;
+}
+
+void amdgpu_jpeg_dump_ip_state(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	u32 inst_off, inst_id, is_powered;
+	int i, j;
+
+	if (!adev->jpeg.ip_dump)
+		return;
+
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; i++) {
+		if (adev->jpeg.harvest_config & (1 << i))
+			continue;
+
+		inst_id = GET_INST(JPEG, i);
+		inst_off = i * adev->jpeg.reg_count;
+		/* check power status from UVD_JPEG_POWER_STATUS */
+		adev->jpeg.ip_dump[inst_off] =
+			RREG32(SOC15_REG_ENTRY_OFFSET_INST(adev->jpeg.reg_list[0],
+							   inst_id));
+		is_powered = ((adev->jpeg.ip_dump[inst_off] & 0x1) != 1);
+
+		if (is_powered)
+			for (j = 1; j < adev->jpeg.reg_count; j++)
+				adev->jpeg.ip_dump[inst_off + j] =
+					RREG32(SOC15_REG_ENTRY_OFFSET_INST(adev->jpeg.reg_list[j],
+									   inst_id));
+	}
+}
+
+void amdgpu_jpeg_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	u32 inst_off, is_powered;
+	int i, j;
+
+	if (!adev->jpeg.ip_dump)
+		return;
+
+	drm_printf(p, "num_instances:%d\n", adev->jpeg.num_jpeg_inst);
+	for (i = 0; i < adev->jpeg.num_jpeg_inst; i++) {
+		if (adev->jpeg.harvest_config & (1 << i)) {
+			drm_printf(p, "\nHarvested Instance:JPEG%d Skipping dump\n", i);
+			continue;
+		}
+
+		inst_off = i * adev->jpeg.reg_count;
+		is_powered = ((adev->jpeg.ip_dump[inst_off] & 0x1) != 1);
+
+		if (is_powered) {
+			drm_printf(p, "Active Instance:JPEG%d\n", i);
+			for (j = 0; j < adev->jpeg.reg_count; j++)
+				drm_printf(p, "%-50s \t 0x%08x\n", adev->jpeg.reg_list[j].reg_name,
+					   adev->jpeg.ip_dump[inst_off + j]);
+		} else
+			drm_printf(p, "\nInactive Instance:JPEG%d\n", i);
+	}
+}
+
+static inline bool amdgpu_jpeg_reg_valid(u32 reg)
+{
+	if (reg < JPEG_REG_RANGE_START || reg > JPEG_REG_RANGE_END ||
+	    (reg >= JPEG_ATOMIC_RANGE_START && reg <= JPEG_ATOMIC_RANGE_END))
+		return false;
+	else
+		return true;
+}
+
+/**
+ * amdgpu_jpeg_dec_parse_cs - command submission parser
+ *
+ * @parser: Command submission parser context
+ * @job: the job to parse
+ * @ib: the IB to parse
+ *
+ * Parse the command stream, return -EINVAL for invalid packet,
+ * 0 otherwise
+ */
+
+int amdgpu_jpeg_dec_parse_cs(struct amdgpu_cs_parser *parser,
+			      struct amdgpu_job *job,
+			      struct amdgpu_ib *ib)
+{
+	u32 i, reg, res, cond, type;
+	struct amdgpu_device *adev = parser->adev;
+
+	for (i = 0; i < ib->length_dw ; i += 2) {
+		reg  = CP_PACKETJ_GET_REG(ib->ptr[i]);
+		res  = CP_PACKETJ_GET_RES(ib->ptr[i]);
+		cond = CP_PACKETJ_GET_COND(ib->ptr[i]);
+		type = CP_PACKETJ_GET_TYPE(ib->ptr[i]);
+
+		if (res) /* only support 0 at the moment */
+			return -EINVAL;
+
+		switch (type) {
+		case PACKETJ_TYPE0:
+			if (cond != PACKETJ_CONDITION_CHECK0 ||
+			    !amdgpu_jpeg_reg_valid(reg)) {
+				dev_err(adev->dev, "Invalid packet [0x%08x]!\n", ib->ptr[i]);
+				return -EINVAL;
+			}
+			break;
+		case PACKETJ_TYPE3:
+			if (cond != PACKETJ_CONDITION_CHECK3 ||
+			    !amdgpu_jpeg_reg_valid(reg)) {
+				dev_err(adev->dev, "Invalid packet [0x%08x]!\n", ib->ptr[i]);
+				return -EINVAL;
+			}
+			break;
+		case PACKETJ_TYPE6:
+			if (ib->ptr[i] == CP_PACKETJ_NOP)
+				continue;
+			dev_err(adev->dev, "Invalid packet [0x%08x]!\n", ib->ptr[i]);
+			return -EINVAL;
+		default:
+			dev_err(adev->dev, "Unknown packet type %d !\n", type);
+			return -EINVAL;
+		}
+	}
 
 	return 0;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h
index 0ca76f0f23e9..346ae0ab09d3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h
@@ -25,20 +25,103 @@
 #define __AMDGPU_JPEG_H__
 
 #include "amdgpu_ras.h"
+#include "amdgpu_cs.h"
+
+#define AMDGPU_MAX_JPEG_INSTANCES	4
+#define AMDGPU_MAX_JPEG_RINGS           10
+#define AMDGPU_MAX_JPEG_RINGS_4_0_3     8
+
+#define JPEG_REG_RANGE_START            0x4000
+#define JPEG_REG_RANGE_END              0x41c2
+#define JPEG_ATOMIC_RANGE_START         0x4120
+#define JPEG_ATOMIC_RANGE_END           0x412A
 
-#define AMDGPU_MAX_JPEG_INSTANCES	2
 
 #define AMDGPU_JPEG_HARVEST_JPEG0 (1 << 0)
 #define AMDGPU_JPEG_HARVEST_JPEG1 (1 << 1)
 
+#define WREG32_SOC15_JPEG_DPG_MODE(inst_idx, offset, value, indirect)			\
+	do {										\
+		if (!indirect) {							\
+			WREG32_SOC15(JPEG, GET_INST(JPEG, inst_idx),			\
+				     mmUVD_DPG_LMA_DATA, value);			\
+			WREG32_SOC15(							\
+				JPEG, GET_INST(JPEG, inst_idx),				\
+				mmUVD_DPG_LMA_CTL,					\
+				(UVD_DPG_LMA_CTL__READ_WRITE_MASK |			\
+				 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT |	\
+				 indirect << UVD_DPG_LMA_CTL__SRAM_SEL__SHIFT));	\
+		} else {								\
+			*adev->jpeg.inst[inst_idx].dpg_sram_curr_addr++ =		\
+				offset;							\
+			*adev->jpeg.inst[inst_idx].dpg_sram_curr_addr++ =		\
+				value;							\
+		}									\
+	} while (0)
+
+#define RREG32_SOC15_JPEG_DPG_MODE(inst_idx, offset, mask_en)					\
+	({											\
+		WREG32_SOC15(JPEG, inst_idx, mmUVD_DPG_LMA_CTL,					\
+			(0x0 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT |				\
+			mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT |				\
+			offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT));			\
+		RREG32_SOC15(JPEG, inst_idx, mmUVD_DPG_LMA_DATA);				\
+	})
+
+#define WREG32_SOC24_JPEG_DPG_MODE(inst_idx, offset, value, indirect)		\
+	do {									\
+		WREG32_SOC15(JPEG, GET_INST(JPEG, inst_idx),			\
+			     regUVD_DPG_LMA_DATA, value);			\
+		WREG32_SOC15(JPEG, GET_INST(JPEG, inst_idx),			\
+			     regUVD_DPG_LMA_MASK, 0xFFFFFFFF);			\
+		WREG32_SOC15(							\
+			JPEG, GET_INST(JPEG, inst_idx),				\
+			regUVD_DPG_LMA_CTL,					\
+			(UVD_DPG_LMA_CTL__READ_WRITE_MASK |			\
+			 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT |	\
+			 indirect << UVD_DPG_LMA_CTL__SRAM_SEL__SHIFT));	\
+	} while (0)
+
+#define RREG32_SOC24_JPEG_DPG_MODE(inst_idx, offset, mask_en)			\
+	do {									\
+		WREG32_SOC15(JPEG, GET_INST(JPEG, inst_idx),			\
+			regUVD_DPG_LMA_MASK, 0xFFFFFFFF);			\
+		WREG32_SOC15(JPEG, GET_INST(JPEG, inst_idx),			\
+			regUVD_DPG_LMA_CTL,					\
+			(UVD_DPG_LMA_CTL__MASK_EN_MASK |			\
+			offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT));	\
+		RREG32_SOC15(JPEG, inst_idx, regUVD_DPG_LMA_DATA);		\
+	} while (0)
+
+#define ADD_SOC24_JPEG_TO_DPG_SRAM(inst_idx, offset, value, indirect)		\
+	do {									\
+		*adev->jpeg.inst[inst_idx].dpg_sram_curr_addr++ = offset;	\
+		*adev->jpeg.inst[inst_idx].dpg_sram_curr_addr++ = value;	\
+	} while (0)
+
+struct amdgpu_hwip_reg_entry;
+
+enum amdgpu_jpeg_caps {
+	AMDGPU_JPEG_RRMT_ENABLED,
+};
+
+#define AMDGPU_JPEG_CAPS(caps) BIT(AMDGPU_JPEG_##caps)
+
 struct amdgpu_jpeg_reg{
-	unsigned jpeg_pitch;
+	unsigned jpeg_pitch[AMDGPU_MAX_JPEG_RINGS];
 };
 
 struct amdgpu_jpeg_inst {
-	struct amdgpu_ring ring_dec;
+	struct amdgpu_ring ring_dec[AMDGPU_MAX_JPEG_RINGS];
 	struct amdgpu_irq_src irq;
+	struct amdgpu_irq_src ras_poison_irq;
 	struct amdgpu_jpeg_reg external;
+	struct amdgpu_bo	*dpg_sram_bo;
+	struct dpg_pause_state	pause_state;
+	void			*dpg_sram_cpu_addr;
+	uint64_t		dpg_sram_gpu_addr;
+	uint32_t		*dpg_sram_curr_addr;
+	uint8_t aid_id;
 };
 
 struct amdgpu_jpeg_ras {
@@ -48,6 +131,7 @@ struct amdgpu_jpeg_ras {
 struct amdgpu_jpeg {
 	uint8_t	num_jpeg_inst;
 	struct amdgpu_jpeg_inst inst[AMDGPU_MAX_JPEG_INSTANCES];
+	unsigned num_jpeg_rings;
 	struct amdgpu_jpeg_reg internal;
 	unsigned harvest_config;
 	struct delayed_work idle_work;
@@ -56,6 +140,15 @@ struct amdgpu_jpeg {
 	atomic_t total_submission_cnt;
 	struct ras_common_if	*ras_if;
 	struct amdgpu_jpeg_ras	*ras;
+
+	uint16_t inst_mask;
+	uint8_t num_inst_per_aid;
+	bool	indirect_sram;
+	uint32_t supported_reset;
+	uint32_t caps;
+	u32 *ip_dump;
+	u32 reg_count;
+	const struct amdgpu_hwip_reg_entry *reg_list;
 };
 
 int amdgpu_jpeg_sw_init(struct amdgpu_device *adev);
@@ -72,6 +165,20 @@ int amdgpu_jpeg_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout);
 int amdgpu_jpeg_process_poison_irq(struct amdgpu_device *adev,
 				struct amdgpu_irq_src *source,
 				struct amdgpu_iv_entry *entry);
+int amdgpu_jpeg_ras_late_init(struct amdgpu_device *adev,
+				struct ras_common_if *ras_block);
 int amdgpu_jpeg_ras_sw_init(struct amdgpu_device *adev);
+int amdgpu_jpeg_psp_update_sram(struct amdgpu_device *adev, int inst_idx,
+			       enum AMDGPU_UCODE_ID ucode_id);
+void amdgpu_debugfs_jpeg_sched_mask_init(struct amdgpu_device *adev);
+int amdgpu_jpeg_sysfs_reset_mask_init(struct amdgpu_device *adev);
+void amdgpu_jpeg_sysfs_reset_mask_fini(struct amdgpu_device *adev);
+int amdgpu_jpeg_reg_dump_init(struct amdgpu_device *adev,
+			       const struct amdgpu_hwip_reg_entry *reg, u32 count);
+void amdgpu_jpeg_dump_ip_state(struct amdgpu_ip_block *ip_block);
+void amdgpu_jpeg_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p);
+int amdgpu_jpeg_dec_parse_cs(struct amdgpu_cs_parser *parser,
+			     struct amdgpu_job *job,
+			     struct amdgpu_ib *ib);
 
 #endif /*__AMDGPU_JPEG_H__*/
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
index 0efb38539d70..6ee77f431d56 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
@@ -43,7 +43,9 @@
 #include "amdgpu_gem.h"
 #include "amdgpu_display.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_reset.h"
 #include "amd_pcie.h"
+#include "amdgpu_userq.h"
 
 void amdgpu_unregister_gpu_instance(struct amdgpu_device *adev)
 {
@@ -89,7 +91,7 @@ void amdgpu_driver_unload_kms(struct drm_device *dev)
 	if (adev->rmmio == NULL)
 		return;
 
-	if (amdgpu_acpi_smart_shift_update(dev, AMDGPU_SS_DRV_UNLOAD))
+	if (amdgpu_acpi_smart_shift_update(adev, AMDGPU_SS_DRV_UNLOAD))
 		DRM_WARN("smart shift update failed\n");
 
 	amdgpu_acpi_fini(adev);
@@ -149,38 +151,7 @@ int amdgpu_driver_load_kms(struct amdgpu_device *adev, unsigned long flags)
 		goto out;
 	}
 
-	adev->pm.rpm_mode = AMDGPU_RUNPM_NONE;
-	if (amdgpu_device_supports_px(dev) &&
-	    (amdgpu_runtime_pm != 0)) { /* enable PX as runtime mode */
-		adev->pm.rpm_mode = AMDGPU_RUNPM_PX;
-		dev_info(adev->dev, "Using ATPX for runtime pm\n");
-	} else if (amdgpu_device_supports_boco(dev) &&
-		   (amdgpu_runtime_pm != 0)) { /* enable boco as runtime mode */
-		adev->pm.rpm_mode = AMDGPU_RUNPM_BOCO;
-		dev_info(adev->dev, "Using BOCO for runtime pm\n");
-	} else if (amdgpu_device_supports_baco(dev) &&
-		   (amdgpu_runtime_pm != 0)) {
-		switch (adev->asic_type) {
-		case CHIP_VEGA20:
-		case CHIP_ARCTURUS:
-			/* enable BACO as runpm mode if runpm=1 */
-			if (amdgpu_runtime_pm > 0)
-				adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
-			break;
-		case CHIP_VEGA10:
-			/* enable BACO as runpm mode if noretry=0 */
-			if (!adev->gmc.noretry)
-				adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
-			break;
-		default:
-			/* enable BACO as runpm mode on CI+ */
-			adev->pm.rpm_mode = AMDGPU_RUNPM_BACO;
-			break;
-		}
-
-		if (adev->pm.rpm_mode == AMDGPU_RUNPM_BACO)
-			dev_info(adev->dev, "Using BACO for runtime pm\n");
-	}
+	amdgpu_device_detect_runtime_pm_mode(adev);
 
 	/* Call ACPI methods: require modeset init
 	 * but failure is not fatal
@@ -190,7 +161,7 @@ int amdgpu_driver_load_kms(struct amdgpu_device *adev, unsigned long flags)
 	if (acpi_status)
 		dev_dbg(dev->dev, "Error during ACPI methods call\n");
 
-	if (amdgpu_acpi_smart_shift_update(dev, AMDGPU_SS_DRV_LOAD))
+	if (amdgpu_acpi_smart_shift_update(adev, AMDGPU_SS_DRV_LOAD))
 		DRM_WARN("smart shift update failed\n");
 
 out:
@@ -200,6 +171,44 @@ out:
 	return r;
 }
 
+static enum amd_ip_block_type
+	amdgpu_ip_get_block_type(struct amdgpu_device *adev, uint32_t ip)
+{
+	enum amd_ip_block_type type;
+
+	switch (ip) {
+	case AMDGPU_HW_IP_GFX:
+		type = AMD_IP_BLOCK_TYPE_GFX;
+		break;
+	case AMDGPU_HW_IP_COMPUTE:
+		type = AMD_IP_BLOCK_TYPE_GFX;
+		break;
+	case AMDGPU_HW_IP_DMA:
+		type = AMD_IP_BLOCK_TYPE_SDMA;
+		break;
+	case AMDGPU_HW_IP_UVD:
+	case AMDGPU_HW_IP_UVD_ENC:
+		type = AMD_IP_BLOCK_TYPE_UVD;
+		break;
+	case AMDGPU_HW_IP_VCE:
+		type = AMD_IP_BLOCK_TYPE_VCE;
+		break;
+	case AMDGPU_HW_IP_VCN_DEC:
+	case AMDGPU_HW_IP_VCN_ENC:
+		type = AMD_IP_BLOCK_TYPE_VCN;
+		break;
+	case AMDGPU_HW_IP_VCN_JPEG:
+		type = (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_JPEG)) ?
+				   AMD_IP_BLOCK_TYPE_JPEG : AMD_IP_BLOCK_TYPE_VCN;
+		break;
+	default:
+		type = AMD_IP_BLOCK_TYPE_NUM;
+		break;
+	}
+
+	return type;
+}
+
 static int amdgpu_firmware_info(struct drm_amdgpu_info_firmware *fw_info,
 				struct drm_amdgpu_query_fw *query_fw,
 				struct amdgpu_device *adev)
@@ -352,12 +361,36 @@ static int amdgpu_firmware_info(struct drm_amdgpu_info_firmware *fw_info,
 		fw_info->ver = adev->gfx.imu_fw_version;
 		fw_info->feature = 0;
 		break;
+	case AMDGPU_INFO_FW_VPE:
+		fw_info->ver = adev->vpe.fw_version;
+		fw_info->feature = adev->vpe.feature_version;
+		break;
 	default:
 		return -EINVAL;
 	}
 	return 0;
 }
 
+static int amdgpu_userq_metadata_info_gfx(struct amdgpu_device *adev,
+					  struct drm_amdgpu_info *info,
+					  struct drm_amdgpu_info_uq_metadata_gfx *meta)
+{
+	int ret = -EOPNOTSUPP;
+
+	if (adev->gfx.funcs->get_gfx_shadow_info) {
+		struct amdgpu_gfx_shadow_info shadow = {};
+
+		adev->gfx.funcs->get_gfx_shadow_info(adev, &shadow, true);
+		meta->shadow_size = shadow.shadow_size;
+		meta->shadow_alignment = shadow.shadow_alignment;
+		meta->csa_size = shadow.csa_size;
+		meta->csa_alignment = shadow.csa_alignment;
+		ret = 0;
+	}
+
+	return ret;
+}
+
 static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 			     struct drm_amdgpu_info *info,
 			     struct drm_amdgpu_info_hw_ip *result)
@@ -366,6 +399,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 	uint32_t ib_size_alignment = 0;
 	enum amd_ip_block_type type;
 	unsigned int num_rings = 0;
+	uint32_t num_slots = 0;
 	unsigned int i, j;
 
 	if (info->query_hw_ip.ip_instance >= AMDGPU_HW_IP_INSTANCE_MAX_COUNT)
@@ -375,24 +409,45 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 	case AMDGPU_HW_IP_GFX:
 		type = AMD_IP_BLOCK_TYPE_GFX;
 		for (i = 0; i < adev->gfx.num_gfx_rings; i++)
-			if (adev->gfx.gfx_ring[i].sched.ready)
+			if (adev->gfx.gfx_ring[i].sched.ready &&
+			    !adev->gfx.gfx_ring[i].no_user_submission)
 				++num_rings;
+
+		if (!adev->gfx.disable_uq) {
+			for (i = 0; i < AMDGPU_MES_MAX_GFX_PIPES; i++)
+				num_slots += hweight32(adev->mes.gfx_hqd_mask[i]);
+		}
+
 		ib_start_alignment = 32;
 		ib_size_alignment = 32;
 		break;
 	case AMDGPU_HW_IP_COMPUTE:
 		type = AMD_IP_BLOCK_TYPE_GFX;
 		for (i = 0; i < adev->gfx.num_compute_rings; i++)
-			if (adev->gfx.compute_ring[i].sched.ready)
+			if (adev->gfx.compute_ring[i].sched.ready &&
+			    !adev->gfx.compute_ring[i].no_user_submission)
 				++num_rings;
+
+		if (!adev->sdma.disable_uq) {
+			for (i = 0; i < AMDGPU_MES_MAX_COMPUTE_PIPES; i++)
+				num_slots += hweight32(adev->mes.compute_hqd_mask[i]);
+		}
+
 		ib_start_alignment = 32;
 		ib_size_alignment = 32;
 		break;
 	case AMDGPU_HW_IP_DMA:
 		type = AMD_IP_BLOCK_TYPE_SDMA;
 		for (i = 0; i < adev->sdma.num_instances; i++)
-			if (adev->sdma.instance[i].ring.sched.ready)
+			if (adev->sdma.instance[i].ring.sched.ready &&
+			    !adev->sdma.instance[i].ring.no_user_submission)
 				++num_rings;
+
+		if (!adev->gfx.disable_uq) {
+			for (i = 0; i < AMDGPU_MES_MAX_SDMA_PIPES; i++)
+				num_slots += hweight32(adev->mes.sdma_hqd_mask[i]);
+		}
+
 		ib_start_alignment = 256;
 		ib_size_alignment = 4;
 		break;
@@ -402,19 +457,21 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 			if (adev->uvd.harvest_config & (1 << i))
 				continue;
 
-			if (adev->uvd.inst[i].ring.sched.ready)
+			if (adev->uvd.inst[i].ring.sched.ready &&
+			    !adev->uvd.inst[i].ring.no_user_submission)
 				++num_rings;
 		}
-		ib_start_alignment = 64;
+		ib_start_alignment = 256;
 		ib_size_alignment = 64;
 		break;
 	case AMDGPU_HW_IP_VCE:
 		type = AMD_IP_BLOCK_TYPE_VCE;
 		for (i = 0; i < adev->vce.num_rings; i++)
-			if (adev->vce.ring[i].sched.ready)
+			if (adev->vce.ring[i].sched.ready &&
+			    !adev->vce.ring[i].no_user_submission)
 				++num_rings;
-		ib_start_alignment = 4;
-		ib_size_alignment = 1;
+		ib_start_alignment = 256;
+		ib_size_alignment = 4;
 		break;
 	case AMDGPU_HW_IP_UVD_ENC:
 		type = AMD_IP_BLOCK_TYPE_UVD;
@@ -423,11 +480,12 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 				continue;
 
 			for (j = 0; j < adev->uvd.num_enc_rings; j++)
-				if (adev->uvd.inst[i].ring_enc[j].sched.ready)
+				if (adev->uvd.inst[i].ring_enc[j].sched.ready &&
+				    !adev->uvd.inst[i].ring_enc[j].no_user_submission)
 					++num_rings;
 		}
-		ib_start_alignment = 64;
-		ib_size_alignment = 64;
+		ib_start_alignment = 256;
+		ib_size_alignment = 4;
 		break;
 	case AMDGPU_HW_IP_VCN_DEC:
 		type = AMD_IP_BLOCK_TYPE_VCN;
@@ -435,11 +493,12 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 			if (adev->vcn.harvest_config & (1 << i))
 				continue;
 
-			if (adev->vcn.inst[i].ring_dec.sched.ready)
+			if (adev->vcn.inst[i].ring_dec.sched.ready &&
+			    !adev->vcn.inst[i].ring_dec.no_user_submission)
 				++num_rings;
 		}
-		ib_start_alignment = 16;
-		ib_size_alignment = 16;
+		ib_start_alignment = 256;
+		ib_size_alignment = 64;
 		break;
 	case AMDGPU_HW_IP_VCN_ENC:
 		type = AMD_IP_BLOCK_TYPE_VCN;
@@ -447,12 +506,13 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 			if (adev->vcn.harvest_config & (1 << i))
 				continue;
 
-			for (j = 0; j < adev->vcn.num_enc_rings; j++)
-				if (adev->vcn.inst[i].ring_enc[j].sched.ready)
+			for (j = 0; j < adev->vcn.inst[i].num_enc_rings; j++)
+				if (adev->vcn.inst[i].ring_enc[j].sched.ready &&
+				    !adev->vcn.inst[i].ring_enc[j].no_user_submission)
 					++num_rings;
 		}
-		ib_start_alignment = 64;
-		ib_size_alignment = 1;
+		ib_start_alignment = 256;
+		ib_size_alignment = 4;
 		break;
 	case AMDGPU_HW_IP_VCN_JPEG:
 		type = (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_JPEG)) ?
@@ -462,11 +522,21 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 			if (adev->jpeg.harvest_config & (1 << i))
 				continue;
 
-			if (adev->jpeg.inst[i].ring_dec.sched.ready)
-				++num_rings;
+			for (j = 0; j < adev->jpeg.num_jpeg_rings; j++)
+				if (adev->jpeg.inst[i].ring_dec[j].sched.ready &&
+				    !adev->jpeg.inst[i].ring_dec[j].no_user_submission)
+					++num_rings;
 		}
-		ib_start_alignment = 16;
-		ib_size_alignment = 16;
+		ib_start_alignment = 256;
+		ib_size_alignment = 64;
+		break;
+	case AMDGPU_HW_IP_VPE:
+		type = AMD_IP_BLOCK_TYPE_VPE;
+		if (adev->vpe.ring.sched.ready &&
+		    !adev->vpe.ring.no_user_submission)
+			++num_rings;
+		ib_start_alignment = 256;
+		ib_size_alignment = 4;
 		break;
 	default:
 		return -EINVAL;
@@ -489,18 +559,26 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 	if (adev->asic_type >= CHIP_VEGA10) {
 		switch (type) {
 		case AMD_IP_BLOCK_TYPE_GFX:
-			result->ip_discovery_version = adev->ip_versions[GC_HWIP][0];
+			result->ip_discovery_version =
+				IP_VERSION_MAJ_MIN_REV(amdgpu_ip_version(adev, GC_HWIP, 0));
 			break;
 		case AMD_IP_BLOCK_TYPE_SDMA:
-			result->ip_discovery_version = adev->ip_versions[SDMA0_HWIP][0];
+			result->ip_discovery_version =
+				IP_VERSION_MAJ_MIN_REV(amdgpu_ip_version(adev, SDMA0_HWIP, 0));
 			break;
 		case AMD_IP_BLOCK_TYPE_UVD:
 		case AMD_IP_BLOCK_TYPE_VCN:
 		case AMD_IP_BLOCK_TYPE_JPEG:
-			result->ip_discovery_version = adev->ip_versions[UVD_HWIP][0];
+			result->ip_discovery_version =
+				IP_VERSION_MAJ_MIN_REV(amdgpu_ip_version(adev, UVD_HWIP, 0));
 			break;
 		case AMD_IP_BLOCK_TYPE_VCE:
-			result->ip_discovery_version = adev->ip_versions[VCE_HWIP][0];
+			result->ip_discovery_version =
+				IP_VERSION_MAJ_MIN_REV(amdgpu_ip_version(adev, VCE_HWIP, 0));
+			break;
+		case AMD_IP_BLOCK_TYPE_VPE:
+			result->ip_discovery_version =
+				IP_VERSION_MAJ_MIN_REV(amdgpu_ip_version(adev, VPE_HWIP, 0));
 			break;
 		default:
 			result->ip_discovery_version = 0;
@@ -511,6 +589,7 @@ static int amdgpu_hw_ip_info(struct amdgpu_device *adev,
 	}
 	result->capabilities_flags = 0;
 	result->available_rings = (1 << num_rings) - 1;
+	result->userq_num_slots = num_slots;
 	result->ib_start_alignment = ib_start_alignment;
 	result->ib_size_alignment = ib_size_alignment;
 	return 0;
@@ -537,11 +616,16 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	struct drm_amdgpu_info *info = data;
 	struct amdgpu_mode_info *minfo = &adev->mode_info;
 	void __user *out = (void __user *)(uintptr_t)info->return_pointer;
+	struct amdgpu_fpriv *fpriv;
+	struct amdgpu_ip_block *ip_block;
+	enum amd_ip_block_type type;
+	struct amdgpu_xcp *xcp;
+	u32 count, inst_mask;
 	uint32_t size = info->return_size;
 	struct drm_crtc *crtc;
 	uint32_t ui32 = 0;
 	uint64_t ui64 = 0;
-	int i, found;
+	int i, found, ret;
 	int ui32_size = sizeof(ui32);
 
 	if (!info->return_size || !info->return_pointer)
@@ -556,6 +640,7 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 			crtc = (struct drm_crtc *)minfo->crtcs[i];
 			if (crtc && crtc->base.id == info->mode_crtc.id) {
 				struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+
 				ui32 = amdgpu_crtc->crtc_id;
 				found = 1;
 				break;
@@ -568,56 +653,82 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		return copy_to_user(out, &ui32, min(size, 4u)) ? -EFAULT : 0;
 	case AMDGPU_INFO_HW_IP_INFO: {
 		struct drm_amdgpu_info_hw_ip ip = {};
-		int ret;
 
 		ret = amdgpu_hw_ip_info(adev, info, &ip);
 		if (ret)
 			return ret;
 
-		ret = copy_to_user(out, &ip, min((size_t)size, sizeof(ip)));
+		ret = copy_to_user(out, &ip, min_t(size_t, size, sizeof(ip)));
 		return ret ? -EFAULT : 0;
 	}
 	case AMDGPU_INFO_HW_IP_COUNT: {
-		enum amd_ip_block_type type;
-		uint32_t count = 0;
+		fpriv = (struct amdgpu_fpriv *)filp->driver_priv;
+		type = amdgpu_ip_get_block_type(adev, info->query_hw_ip.type);
+		ip_block = amdgpu_device_ip_get_ip_block(adev, type);
 
-		switch (info->query_hw_ip.type) {
-		case AMDGPU_HW_IP_GFX:
-			type = AMD_IP_BLOCK_TYPE_GFX;
-			break;
-		case AMDGPU_HW_IP_COMPUTE:
-			type = AMD_IP_BLOCK_TYPE_GFX;
-			break;
-		case AMDGPU_HW_IP_DMA:
-			type = AMD_IP_BLOCK_TYPE_SDMA;
-			break;
-		case AMDGPU_HW_IP_UVD:
-			type = AMD_IP_BLOCK_TYPE_UVD;
+		if (!ip_block || !ip_block->status.valid)
+			return -EINVAL;
+
+		if (adev->xcp_mgr && adev->xcp_mgr->num_xcps > 0 &&
+		    fpriv->xcp_id < adev->xcp_mgr->num_xcps) {
+			xcp = &adev->xcp_mgr->xcp[fpriv->xcp_id];
+			switch (type) {
+			case AMD_IP_BLOCK_TYPE_GFX:
+				ret = amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_GFX, &inst_mask);
+				if (ret)
+					return ret;
+				count = hweight32(inst_mask);
+				break;
+			case AMD_IP_BLOCK_TYPE_SDMA:
+				ret = amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_SDMA, &inst_mask);
+				if (ret)
+					return ret;
+				count = hweight32(inst_mask);
+				break;
+			case AMD_IP_BLOCK_TYPE_JPEG:
+				ret = amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_VCN, &inst_mask);
+				if (ret)
+					return ret;
+				count = hweight32(inst_mask) * adev->jpeg.num_jpeg_rings;
+				break;
+			case AMD_IP_BLOCK_TYPE_VCN:
+				ret = amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_VCN, &inst_mask);
+				if (ret)
+					return ret;
+				count = hweight32(inst_mask);
+				break;
+			default:
+				return -EINVAL;
+			}
+
+			return copy_to_user(out, &count, min(size, 4u)) ? -EFAULT : 0;
+		}
+
+		switch (type) {
+		case AMD_IP_BLOCK_TYPE_GFX:
+		case AMD_IP_BLOCK_TYPE_VCE:
+			count = 1;
 			break;
-		case AMDGPU_HW_IP_VCE:
-			type = AMD_IP_BLOCK_TYPE_VCE;
+		case AMD_IP_BLOCK_TYPE_SDMA:
+			count = adev->sdma.num_instances;
 			break;
-		case AMDGPU_HW_IP_UVD_ENC:
-			type = AMD_IP_BLOCK_TYPE_UVD;
+		case AMD_IP_BLOCK_TYPE_JPEG:
+			count = adev->jpeg.num_jpeg_inst * adev->jpeg.num_jpeg_rings;
 			break;
-		case AMDGPU_HW_IP_VCN_DEC:
-		case AMDGPU_HW_IP_VCN_ENC:
-			type = AMD_IP_BLOCK_TYPE_VCN;
+		case AMD_IP_BLOCK_TYPE_VCN:
+			count = adev->vcn.num_vcn_inst;
 			break;
-		case AMDGPU_HW_IP_VCN_JPEG:
-			type = (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_JPEG)) ?
-				AMD_IP_BLOCK_TYPE_JPEG : AMD_IP_BLOCK_TYPE_VCN;
+		case AMD_IP_BLOCK_TYPE_UVD:
+			count = adev->uvd.num_uvd_inst;
 			break;
+		/* For all other IP block types not listed in the switch statement
+		 * the ip status is valid here and the instance count is one.
+		 */
 		default:
-			return -EINVAL;
+			count = 1;
+			break;
 		}
 
-		for (i = 0; i < adev->num_ip_blocks; i++)
-			if (adev->ip_blocks[i].version->type == type &&
-			    adev->ip_blocks[i].status.valid &&
-			    count < AMDGPU_HW_IP_INSTANCE_MAX_COUNT)
-				count++;
-
 		return copy_to_user(out, &count, min(size, 4u)) ? -EFAULT : 0;
 	}
 	case AMDGPU_INFO_TIMESTAMP:
@@ -625,7 +736,6 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0;
 	case AMDGPU_INFO_FW_VERSION: {
 		struct drm_amdgpu_info_firmware fw_info;
-		int ret;
 
 		/* We only support one instance of each IP block right now. */
 		if (info->query_fw.ip_instance != 0)
@@ -648,7 +758,8 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		ui64 = atomic64_read(&adev->num_vram_cpu_page_faults);
 		return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0;
 	case AMDGPU_INFO_VRAM_USAGE:
-		ui64 = ttm_resource_manager_usage(&adev->mman.vram_mgr.manager);
+		ui64 = ttm_resource_manager_used(&adev->mman.vram_mgr.manager) ?
+			ttm_resource_manager_usage(&adev->mman.vram_mgr.manager) : 0;
 		return copy_to_user(out, &ui64, min(size, 8u)) ? -EFAULT : 0;
 	case AMDGPU_INFO_VIS_VRAM_USAGE:
 		ui64 = amdgpu_vram_mgr_vis_usage(&adev->mman.vram_mgr);
@@ -694,8 +805,8 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		mem.vram.usable_heap_size = adev->gmc.real_vram_size -
 			atomic64_read(&adev->vram_pin_size) -
 			AMDGPU_VM_RESERVED_VRAM;
-		mem.vram.heap_usage =
-			ttm_resource_manager_usage(vram_man);
+		mem.vram.heap_usage = ttm_resource_manager_used(&adev->mman.vram_mgr.manager) ?
+				ttm_resource_manager_usage(vram_man) : 0;
 		mem.vram.max_allocation = mem.vram.usable_heap_size * 3 / 4;
 
 		mem.cpu_accessible_vram.total_heap_size =
@@ -720,32 +831,47 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 				    ? -EFAULT : 0;
 	}
 	case AMDGPU_INFO_READ_MMR_REG: {
-		unsigned n, alloc_size;
+		int ret = 0;
+		unsigned int n, alloc_size;
 		uint32_t *regs;
-		unsigned se_num = (info->read_mmr_reg.instance >>
+		unsigned int se_num = (info->read_mmr_reg.instance >>
 				   AMDGPU_INFO_MMR_SE_INDEX_SHIFT) &
 				  AMDGPU_INFO_MMR_SE_INDEX_MASK;
-		unsigned sh_num = (info->read_mmr_reg.instance >>
+		unsigned int sh_num = (info->read_mmr_reg.instance >>
 				   AMDGPU_INFO_MMR_SH_INDEX_SHIFT) &
 				  AMDGPU_INFO_MMR_SH_INDEX_MASK;
 
+		if (!down_read_trylock(&adev->reset_domain->sem))
+			return -ENOENT;
+
 		/* set full masks if the userspace set all bits
-		 * in the bitfields */
-		if (se_num == AMDGPU_INFO_MMR_SE_INDEX_MASK)
+		 * in the bitfields
+		 */
+		if (se_num == AMDGPU_INFO_MMR_SE_INDEX_MASK) {
 			se_num = 0xffffffff;
-		else if (se_num >= AMDGPU_GFX_MAX_SE)
-			return -EINVAL;
-		if (sh_num == AMDGPU_INFO_MMR_SH_INDEX_MASK)
+		} else if (se_num >= AMDGPU_GFX_MAX_SE) {
+			ret = -EINVAL;
+			goto out;
+		}
+
+		if (sh_num == AMDGPU_INFO_MMR_SH_INDEX_MASK) {
 			sh_num = 0xffffffff;
-		else if (sh_num >= AMDGPU_GFX_MAX_SH_PER_SE)
-			return -EINVAL;
+		} else if (sh_num >= AMDGPU_GFX_MAX_SH_PER_SE) {
+			ret = -EINVAL;
+			goto out;
+		}
 
-		if (info->read_mmr_reg.count > 128)
-			return -EINVAL;
+		if (info->read_mmr_reg.count > 128) {
+			ret = -EINVAL;
+			goto out;
+		}
 
 		regs = kmalloc_array(info->read_mmr_reg.count, sizeof(*regs), GFP_KERNEL);
-		if (!regs)
-			return -ENOMEM;
+		if (!regs) {
+			ret = -ENOMEM;
+			goto out;
+		}
+
 		alloc_size = info->read_mmr_reg.count * sizeof(*regs);
 
 		amdgpu_gfx_off_ctrl(adev, false);
@@ -757,19 +883,22 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 					      info->read_mmr_reg.dword_offset + i);
 				kfree(regs);
 				amdgpu_gfx_off_ctrl(adev, true);
-				return -EFAULT;
+				ret = -EFAULT;
+				goto out;
 			}
 		}
 		amdgpu_gfx_off_ctrl(adev, true);
 		n = copy_to_user(out, regs, min(size, alloc_size));
 		kfree(regs);
-		return n ? -EFAULT : 0;
+		ret = (n ? -EFAULT : 0);
+out:
+		up_read(&adev->reset_domain->sem);
+		return ret;
 	}
 	case AMDGPU_INFO_DEV_INFO: {
 		struct drm_amdgpu_info_device *dev_info;
 		uint64_t vm_size;
-		uint32_t pcie_gen_mask;
-		int ret;
+		uint32_t pcie_gen_mask, pcie_width_mask;
 
 		dev_info = kzalloc(sizeof(*dev_info), GFP_KERNEL);
 		if (!dev_info)
@@ -804,22 +933,35 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		dev_info->ids_flags = 0;
 		if (adev->flags & AMD_IS_APU)
 			dev_info->ids_flags |= AMDGPU_IDS_FLAGS_FUSION;
-		if (amdgpu_mcbp)
+		if (adev->gfx.mcbp)
 			dev_info->ids_flags |= AMDGPU_IDS_FLAGS_PREEMPTION;
 		if (amdgpu_is_tmz(adev))
 			dev_info->ids_flags |= AMDGPU_IDS_FLAGS_TMZ;
 		if (adev->gfx.config.ta_cntl2_truncate_coord_mode)
 			dev_info->ids_flags |= AMDGPU_IDS_FLAGS_CONFORMANT_TRUNC_COORD;
 
+		/* Gang submit is not supported under SRIOV currently */
+		if (!amdgpu_sriov_vf(adev))
+			dev_info->ids_flags |= AMDGPU_IDS_FLAGS_GANG_SUBMIT;
+
+		if (amdgpu_passthrough(adev))
+			dev_info->ids_flags |= (AMDGPU_IDS_FLAGS_MODE_PT <<
+						AMDGPU_IDS_FLAGS_MODE_SHIFT) &
+						AMDGPU_IDS_FLAGS_MODE_MASK;
+		else if (amdgpu_sriov_vf(adev))
+			dev_info->ids_flags |= (AMDGPU_IDS_FLAGS_MODE_VF <<
+						AMDGPU_IDS_FLAGS_MODE_SHIFT) &
+						AMDGPU_IDS_FLAGS_MODE_MASK;
+
 		vm_size = adev->vm_manager.max_pfn * AMDGPU_GPU_PAGE_SIZE;
-		vm_size -= AMDGPU_VA_RESERVED_SIZE;
+		vm_size -= AMDGPU_VA_RESERVED_TOP;
 
 		/* Older VCE FW versions are buggy and can handle only 40bits */
 		if (adev->vce.fw_version &&
 		    adev->vce.fw_version < AMDGPU_VCE_FW_53_45)
 			vm_size = min(vm_size, 1ULL << 40);
 
-		dev_info->virtual_address_offset = AMDGPU_VA_RESERVED_SIZE;
+		dev_info->virtual_address_offset = AMDGPU_VA_RESERVED_BOTTOM;
 		dev_info->virtual_address_max =
 			min(vm_size, AMDGPU_GMC_HOLE_START);
 
@@ -836,7 +978,7 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		memcpy(&dev_info->cu_ao_bitmap[0], &adev->gfx.cu_info.ao_cu_bitmap[0],
 		       sizeof(adev->gfx.cu_info.ao_cu_bitmap));
 		memcpy(&dev_info->cu_bitmap[0], &adev->gfx.cu_info.bitmap[0],
-		       sizeof(adev->gfx.cu_info.bitmap));
+		       sizeof(dev_info->cu_bitmap));
 		dev_info->vram_type = adev->gmc.vram_type;
 		dev_info->vram_bit_width = adev->gmc.vram_width;
 		dev_info->vce_harvest_config = adev->vce.harvest_config;
@@ -857,15 +999,18 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		dev_info->tcc_disabled_mask = adev->gfx.config.tcc_disabled_mask;
 
 		/* Combine the chip gen mask with the platform (CPU/mobo) mask. */
-		pcie_gen_mask = adev->pm.pcie_gen_mask & (adev->pm.pcie_gen_mask >> 16);
+		pcie_gen_mask = adev->pm.pcie_gen_mask &
+			(adev->pm.pcie_gen_mask >> CAIL_PCIE_LINK_SPEED_SUPPORT_SHIFT);
+		pcie_width_mask = adev->pm.pcie_mlw_mask &
+			(adev->pm.pcie_mlw_mask >> CAIL_PCIE_LINK_WIDTH_SUPPORT_SHIFT);
 		dev_info->pcie_gen = fls(pcie_gen_mask);
 		dev_info->pcie_num_lanes =
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X32 ? 32 :
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 ? 16 :
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 ? 12 :
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 ? 8 :
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 ? 4 :
-			adev->pm.pcie_mlw_mask & CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 ? 2 : 1;
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X32 ? 32 :
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X16 ? 16 :
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X12 ? 12 :
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X8 ? 8 :
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X4 ? 4 :
+			pcie_width_mask & CAIL_ASIC_PCIE_LINK_WIDTH_SUPPORT_X2 ? 2 : 1;
 
 		dev_info->tcp_cache_size = adev->gfx.config.gc_tcp_l1_size;
 		dev_info->num_sqc_per_wgp = adev->gfx.config.gc_num_sqc_per_wgp;
@@ -876,13 +1021,28 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		dev_info->gl2c_cache_size = adev->gfx.config.gc_gl2c_per_gpu;
 		dev_info->mall_size = adev->gmc.mall_size;
 
+
+		if (adev->gfx.funcs->get_gfx_shadow_info) {
+			struct amdgpu_gfx_shadow_info shadow_info;
+
+			ret = amdgpu_gfx_get_gfx_shadow_info(adev, &shadow_info);
+			if (!ret) {
+				dev_info->shadow_size = shadow_info.shadow_size;
+				dev_info->shadow_alignment = shadow_info.shadow_alignment;
+				dev_info->csa_size = shadow_info.csa_size;
+				dev_info->csa_alignment = shadow_info.csa_alignment;
+			}
+		}
+
+		dev_info->userq_ip_mask = amdgpu_userq_get_supported_ip_mask(adev);
+
 		ret = copy_to_user(out, dev_info,
 				   min((size_t)size, sizeof(*dev_info))) ? -EFAULT : 0;
 		kfree(dev_info);
 		return ret;
 	}
 	case AMDGPU_INFO_VCE_CLOCK_TABLE: {
-		unsigned i;
+		unsigned int i;
 		struct drm_amdgpu_info_vce_clock_table vce_clk_table = {};
 		struct amd_vce_state *vce_state;
 
@@ -924,12 +1084,17 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 			struct atom_context *atom_context;
 
 			atom_context = adev->mode_info.atom_context;
-			memcpy(vbios_info.name, atom_context->name, sizeof(atom_context->name));
-			memcpy(vbios_info.vbios_pn, atom_context->vbios_pn, sizeof(atom_context->vbios_pn));
-			vbios_info.version = atom_context->version;
-			memcpy(vbios_info.vbios_ver_str, atom_context->vbios_ver_str,
-						sizeof(atom_context->vbios_ver_str));
-			memcpy(vbios_info.date, atom_context->date, sizeof(atom_context->date));
+			if (atom_context) {
+				memcpy(vbios_info.name, atom_context->name,
+				       sizeof(atom_context->name));
+				memcpy(vbios_info.vbios_pn, atom_context->vbios_pn,
+				       sizeof(atom_context->vbios_pn));
+				vbios_info.version = atom_context->version;
+				memcpy(vbios_info.vbios_ver_str, atom_context->vbios_ver_str,
+				       sizeof(atom_context->vbios_ver_str));
+				memcpy(vbios_info.date, atom_context->date,
+				       sizeof(atom_context->date));
+			}
 
 			return copy_to_user(out, &vbios_info,
 						min((size_t)size, sizeof(vbios_info))) ? -EFAULT : 0;
@@ -1003,7 +1168,21 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		case AMDGPU_INFO_SENSOR_GPU_AVG_POWER:
 			/* get average GPU power */
 			if (amdgpu_dpm_read_sensor(adev,
-						   AMDGPU_PP_SENSOR_GPU_POWER,
+						   AMDGPU_PP_SENSOR_GPU_AVG_POWER,
+						   (void *)&ui32, &ui32_size)) {
+				/* fall back to input power for backwards compat */
+				if (amdgpu_dpm_read_sensor(adev,
+							   AMDGPU_PP_SENSOR_GPU_INPUT_POWER,
+							   (void *)&ui32, &ui32_size)) {
+					return -EINVAL;
+				}
+			}
+			ui32 >>= 8;
+			break;
+		case AMDGPU_INFO_SENSOR_GPU_INPUT_POWER:
+			/* get input GPU power */
+			if (amdgpu_dpm_read_sensor(adev,
+						   AMDGPU_PP_SENSOR_GPU_INPUT_POWER,
 						   (void *)&ui32, &ui32_size)) {
 				return -EINVAL;
 			}
@@ -1088,6 +1267,9 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		struct drm_amdgpu_info_video_caps *caps;
 		int r;
 
+		if (!adev->asic_funcs->query_video_codecs)
+			return -EINVAL;
+
 		switch (info->video_cap.type) {
 		case AMDGPU_INFO_VIDEO_CAPS_DECODE:
 			r = amdgpu_asic_query_video_codecs(adev, false, &codecs);
@@ -1140,6 +1322,51 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		kfree(caps);
 		return r;
 	}
+	case AMDGPU_INFO_MAX_IBS: {
+		uint32_t max_ibs[AMDGPU_HW_IP_NUM];
+
+		for (i = 0; i < AMDGPU_HW_IP_NUM; ++i)
+			max_ibs[i] = amdgpu_ring_max_ibs(i);
+
+		return copy_to_user(out, max_ibs,
+				    min((size_t)size, sizeof(max_ibs))) ? -EFAULT : 0;
+	}
+	case AMDGPU_INFO_GPUVM_FAULT: {
+		struct amdgpu_fpriv *fpriv = filp->driver_priv;
+		struct amdgpu_vm *vm = &fpriv->vm;
+		struct drm_amdgpu_info_gpuvm_fault gpuvm_fault;
+		unsigned long flags;
+
+		if (!vm)
+			return -EINVAL;
+
+		memset(&gpuvm_fault, 0, sizeof(gpuvm_fault));
+
+		xa_lock_irqsave(&adev->vm_manager.pasids, flags);
+		gpuvm_fault.addr = vm->fault_info.addr;
+		gpuvm_fault.status = vm->fault_info.status;
+		gpuvm_fault.vmhub = vm->fault_info.vmhub;
+		xa_unlock_irqrestore(&adev->vm_manager.pasids, flags);
+
+		return copy_to_user(out, &gpuvm_fault,
+				    min((size_t)size, sizeof(gpuvm_fault))) ? -EFAULT : 0;
+	}
+	case AMDGPU_INFO_UQ_FW_AREAS: {
+		struct drm_amdgpu_info_uq_metadata meta_info = {};
+
+		switch (info->query_hw_ip.type) {
+		case AMDGPU_HW_IP_GFX:
+			ret = amdgpu_userq_metadata_info_gfx(adev, info, &meta_info.gfx);
+			if (ret)
+				return ret;
+
+			ret = copy_to_user(out, &meta_info,
+						min((size_t)size, sizeof(meta_info))) ? -EFAULT : 0;
+			return 0;
+		default:
+			return -EINVAL;
+		}
+	}
 	default:
 		DRM_DEBUG_KMS("Invalid request %d\n", info->query);
 		return -EINVAL;
@@ -1147,23 +1374,6 @@ int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	return 0;
 }
 
-
-/*
- * Outdated mess for old drm with Xorg being in charge (void function now).
- */
-/**
- * amdgpu_driver_lastclose_kms - drm callback for last close
- *
- * @dev: drm dev pointer
- *
- * Switch vga_switcheroo state after last close (all asics).
- */
-void amdgpu_driver_lastclose_kms(struct drm_device *dev)
-{
-	drm_fb_helper_lastclose(dev);
-	vga_switcheroo_process_delayed_switch();
-}
-
 /**
  * amdgpu_driver_open_kms - drm callback for open
  *
@@ -1206,13 +1416,15 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv)
 		pasid = 0;
 	}
 
-	r = amdgpu_vm_init(adev, &fpriv->vm);
+	r = amdgpu_xcp_open_device(adev, fpriv, file_priv);
 	if (r)
 		goto error_pasid;
 
-	r = amdgpu_vm_set_pasid(adev, &fpriv->vm, pasid);
+	amdgpu_debugfs_vm_init(file_priv);
+
+	r = amdgpu_vm_init(adev, &fpriv->vm, fpriv->xcp_id, pasid);
 	if (r)
-		goto error_vm;
+		goto error_pasid;
 
 	fpriv->prt_va = amdgpu_vm_bo_add(adev, &fpriv->vm, NULL);
 	if (!fpriv->prt_va) {
@@ -1220,7 +1432,7 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv)
 		goto error_vm;
 	}
 
-	if (amdgpu_mcbp) {
+	if (adev->gfx.mcbp) {
 		uint64_t csa_addr = amdgpu_csa_vaddr(adev) & AMDGPU_GMC_HOLE_MASK;
 
 		r = amdgpu_map_static_csa(adev, &fpriv->vm, adev->virt.csa_obj,
@@ -1229,9 +1441,21 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv)
 			goto error_vm;
 	}
 
+	r = amdgpu_seq64_map(adev, &fpriv->vm, &fpriv->seq64_va);
+	if (r)
+		goto error_vm;
+
 	mutex_init(&fpriv->bo_list_lock);
 	idr_init_base(&fpriv->bo_list_handles, 1);
 
+	r = amdgpu_userq_mgr_init(&fpriv->userq_mgr, file_priv, adev);
+	if (r)
+		DRM_WARN("Can't setup usermode queues, use legacy workload submission only\n");
+
+	r = amdgpu_eviction_fence_init(&fpriv->evf_mgr);
+	if (r)
+		goto error_vm;
+
 	amdgpu_ctx_mgr_init(&fpriv->ctx_mgr, adev);
 
 	file_priv->driver_priv = fpriv;
@@ -1241,15 +1465,12 @@ error_vm:
 	amdgpu_vm_fini(adev, &fpriv->vm);
 
 error_pasid:
-	if (pasid) {
+	if (pasid)
 		amdgpu_pasid_free(pasid);
-		amdgpu_vm_set_pasid(adev, &fpriv->vm, 0);
-	}
 
 	kfree(fpriv);
 
 out_suspend:
-	pm_runtime_mark_last_busy(dev->dev);
 pm_put:
 	pm_runtime_put_autosuspend(dev->dev);
 
@@ -1284,14 +1505,16 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev,
 	if (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_VCE) != NULL)
 		amdgpu_vce_free_handles(adev, file_priv);
 
-	if (amdgpu_mcbp) {
-		/* TODO: how to handle reserve failure */
-		BUG_ON(amdgpu_bo_reserve(adev->virt.csa_obj, true));
-		amdgpu_vm_bo_del(adev, fpriv->csa_va);
+	if (fpriv->csa_va) {
+		uint64_t csa_addr = amdgpu_csa_vaddr(adev) & AMDGPU_GMC_HOLE_MASK;
+
+		WARN_ON(amdgpu_unmap_static_csa(adev, &fpriv->vm, adev->virt.csa_obj,
+						fpriv->csa_va, csa_addr));
 		fpriv->csa_va = NULL;
-		amdgpu_bo_unreserve(adev->virt.csa_obj);
 	}
 
+	amdgpu_seq64_unmap(adev, fpriv);
+
 	pasid = fpriv->vm.pasid;
 	pd = amdgpu_bo_ref(fpriv->vm.root.bo);
 	if (!WARN_ON(amdgpu_bo_reserve(pd, true))) {
@@ -1315,7 +1538,6 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev,
 	kfree(fpriv);
 	file_priv->driver_priv = NULL;
 
-	pm_runtime_mark_last_busy(dev->dev);
 	pm_runtime_put_autosuspend(dev->dev);
 }
 
@@ -1441,7 +1663,7 @@ void amdgpu_disable_vblank_kms(struct drm_crtc *crtc)
 
 static int amdgpu_debugfs_firmware_info_show(struct seq_file *m, void *unused)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct amdgpu_device *adev = m->private;
 	struct drm_amdgpu_info_firmware fw_info;
 	struct drm_amdgpu_query_fw query_fw;
 	struct atom_context *ctx = adev->mode_info.atom_context;
@@ -1449,7 +1671,7 @@ static int amdgpu_debugfs_firmware_info_show(struct seq_file *m, void *unused)
 	int ret, i;
 
 	static const char *ta_fw_name[TA_FW_TYPE_MAX_INDEX] = {
-#define TA_FW_NAME(type) [TA_FW_TYPE_PSP_##type] = #type
+#define TA_FW_NAME(type)[TA_FW_TYPE_PSP_##type] = #type
 		TA_FW_NAME(XGMI),
 		TA_FW_NAME(RAS),
 		TA_FW_NAME(HDCP),
@@ -1548,7 +1770,7 @@ static int amdgpu_debugfs_firmware_info_show(struct seq_file *m, void *unused)
 		   fw_info.feature, fw_info.ver);
 
 	/* RLCV */
-        query_fw.fw_type = AMDGPU_INFO_FW_GFX_RLCV;
+	query_fw.fw_type = AMDGPU_INFO_FW_GFX_RLCV;
 	ret = amdgpu_firmware_info(&fw_info, &query_fw, adev);
 	if (ret)
 		return ret;
@@ -1692,7 +1914,15 @@ static int amdgpu_debugfs_firmware_info_show(struct seq_file *m, void *unused)
 	seq_printf(m, "MES feature version: %u, firmware version: 0x%08x\n",
 		   fw_info.feature, fw_info.ver);
 
-	seq_printf(m, "VBIOS version: %s\n", ctx->vbios_version);
+	/* VPE */
+	query_fw.fw_type = AMDGPU_INFO_FW_VPE;
+	ret = amdgpu_firmware_info(&fw_info, &query_fw, adev);
+	if (ret)
+		return ret;
+	seq_printf(m, "VPE feature version: %u, firmware version: 0x%08x\n",
+		   fw_info.feature, fw_info.ver);
+
+	seq_printf(m, "VBIOS version: %s\n", ctx->vbios_pn);
 
 	return 0;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c
index 8d9ff9e151de..3ca03b5e0f91 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.c
@@ -27,6 +27,16 @@
 #include "umc/umc_6_7_0_offset.h"
 #include "umc/umc_6_7_0_sh_mask.h"
 
+static bool amdgpu_mca_is_deferred_error(struct amdgpu_device *adev,
+					uint64_t mc_status)
+{
+	if (adev->umc.ras->check_ecc_err_status)
+		return adev->umc.ras->check_ecc_err_status(adev,
+				AMDGPU_MCA_ERROR_TYPE_DE, &mc_status);
+
+	return false;
+}
+
 void amdgpu_mca_query_correctable_error_count(struct amdgpu_device *adev,
 					      uint64_t mc_status_addr,
 					      unsigned long *error_count)
@@ -142,3 +152,479 @@ int amdgpu_mca_mpio_ras_sw_init(struct amdgpu_device *adev)
 
 	return 0;
 }
+
+static void amdgpu_mca_bank_set_init(struct mca_bank_set *mca_set)
+{
+	if (!mca_set)
+		return;
+
+	memset(mca_set, 0, sizeof(*mca_set));
+	INIT_LIST_HEAD(&mca_set->list);
+}
+
+static int amdgpu_mca_bank_set_add_entry(struct mca_bank_set *mca_set, struct mca_bank_entry *entry)
+{
+	struct mca_bank_node *node;
+
+	if (!entry)
+		return -EINVAL;
+
+	node = kvzalloc(sizeof(*node), GFP_KERNEL);
+	if (!node)
+		return -ENOMEM;
+
+	memcpy(&node->entry, entry, sizeof(*entry));
+
+	INIT_LIST_HEAD(&node->node);
+	list_add_tail(&node->node, &mca_set->list);
+
+	mca_set->nr_entries++;
+
+	return 0;
+}
+
+static int amdgpu_mca_bank_set_merge(struct mca_bank_set *mca_set, struct mca_bank_set *new)
+{
+	struct mca_bank_node *node;
+
+	list_for_each_entry(node, &new->list, node)
+		amdgpu_mca_bank_set_add_entry(mca_set, &node->entry);
+
+	return 0;
+}
+
+static void amdgpu_mca_bank_set_remove_node(struct mca_bank_set *mca_set, struct mca_bank_node *node)
+{
+	if (!node)
+		return;
+
+	list_del(&node->node);
+	kvfree(node);
+
+	mca_set->nr_entries--;
+}
+
+static void amdgpu_mca_bank_set_release(struct mca_bank_set *mca_set)
+{
+	struct mca_bank_node *node, *tmp;
+
+	if (list_empty(&mca_set->list))
+		return;
+
+	list_for_each_entry_safe(node, tmp, &mca_set->list, node)
+		amdgpu_mca_bank_set_remove_node(mca_set, node);
+}
+
+void amdgpu_mca_smu_init_funcs(struct amdgpu_device *adev, const struct amdgpu_mca_smu_funcs *mca_funcs)
+{
+	struct amdgpu_mca *mca = &adev->mca;
+
+	mca->mca_funcs = mca_funcs;
+}
+
+int amdgpu_mca_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_mca *mca = &adev->mca;
+	struct mca_bank_cache *mca_cache;
+	int i;
+
+	atomic_set(&mca->ue_update_flag, 0);
+
+	for (i = 0; i < ARRAY_SIZE(mca->mca_caches); i++) {
+		mca_cache = &mca->mca_caches[i];
+		mutex_init(&mca_cache->lock);
+		amdgpu_mca_bank_set_init(&mca_cache->mca_set);
+	}
+
+	return 0;
+}
+
+void amdgpu_mca_fini(struct amdgpu_device *adev)
+{
+	struct amdgpu_mca *mca = &adev->mca;
+	struct mca_bank_cache *mca_cache;
+	int i;
+
+	atomic_set(&mca->ue_update_flag, 0);
+
+	for (i = 0; i < ARRAY_SIZE(mca->mca_caches); i++) {
+		mca_cache = &mca->mca_caches[i];
+		amdgpu_mca_bank_set_release(&mca_cache->mca_set);
+		mutex_destroy(&mca_cache->lock);
+	}
+}
+
+int amdgpu_mca_reset(struct amdgpu_device *adev)
+{
+	amdgpu_mca_fini(adev);
+
+	return amdgpu_mca_init(adev);
+}
+
+int amdgpu_mca_smu_set_debug_mode(struct amdgpu_device *adev, bool enable)
+{
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+
+	if (mca_funcs && mca_funcs->mca_set_debug_mode)
+		return mca_funcs->mca_set_debug_mode(adev, enable);
+
+	return -EOPNOTSUPP;
+}
+
+static void amdgpu_mca_smu_mca_bank_dump(struct amdgpu_device *adev, int idx, struct mca_bank_entry *entry,
+					 struct ras_query_context *qctx)
+{
+	u64 event_id = qctx ? qctx->evid.event_id : RAS_EVENT_INVALID_ID;
+
+	RAS_EVENT_LOG(adev, event_id, HW_ERR "Accelerator Check Architecture events logged\n");
+	RAS_EVENT_LOG(adev, event_id, HW_ERR "aca entry[%02d].STATUS=0x%016llx\n",
+		      idx, entry->regs[MCA_REG_IDX_STATUS]);
+	RAS_EVENT_LOG(adev, event_id, HW_ERR "aca entry[%02d].ADDR=0x%016llx\n",
+		      idx, entry->regs[MCA_REG_IDX_ADDR]);
+	RAS_EVENT_LOG(adev, event_id, HW_ERR "aca entry[%02d].MISC0=0x%016llx\n",
+		      idx, entry->regs[MCA_REG_IDX_MISC0]);
+	RAS_EVENT_LOG(adev, event_id, HW_ERR "aca entry[%02d].IPID=0x%016llx\n",
+		      idx, entry->regs[MCA_REG_IDX_IPID]);
+	RAS_EVENT_LOG(adev, event_id, HW_ERR "aca entry[%02d].SYND=0x%016llx\n",
+		      idx, entry->regs[MCA_REG_IDX_SYND]);
+}
+
+static int amdgpu_mca_smu_get_valid_mca_count(struct amdgpu_device *adev, enum amdgpu_mca_error_type type, uint32_t *count)
+{
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+
+	if (!count)
+		return -EINVAL;
+
+	if (mca_funcs && mca_funcs->mca_get_valid_mca_count)
+		return mca_funcs->mca_get_valid_mca_count(adev, type, count);
+
+	return -EOPNOTSUPP;
+}
+
+static int amdgpu_mca_smu_get_mca_entry(struct amdgpu_device *adev, enum amdgpu_mca_error_type type,
+					int idx, struct mca_bank_entry *entry)
+{
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+	int count;
+
+	if (!mca_funcs || !mca_funcs->mca_get_mca_entry)
+		return -EOPNOTSUPP;
+
+	switch (type) {
+	case AMDGPU_MCA_ERROR_TYPE_UE:
+		count = mca_funcs->max_ue_count;
+		break;
+	case AMDGPU_MCA_ERROR_TYPE_CE:
+		count = mca_funcs->max_ce_count;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (idx >= count)
+		return -EINVAL;
+
+	return mca_funcs->mca_get_mca_entry(adev, type, idx, entry);
+}
+
+static bool amdgpu_mca_bank_should_update(struct amdgpu_device *adev, enum amdgpu_mca_error_type type)
+{
+	struct amdgpu_mca *mca = &adev->mca;
+	bool ret = true;
+
+	/*
+	 * Because the UE Valid MCA count will only be cleared after reset,
+	 * in order to avoid repeated counting of the error count,
+	 * the aca bank is only updated once during the gpu recovery stage.
+	 */
+	if (type == AMDGPU_MCA_ERROR_TYPE_UE) {
+		if (amdgpu_ras_intr_triggered())
+			ret = atomic_cmpxchg(&mca->ue_update_flag, 0, 1) == 0;
+		else
+			atomic_set(&mca->ue_update_flag, 0);
+	}
+
+	return ret;
+}
+
+static bool amdgpu_mca_bank_should_dump(struct amdgpu_device *adev, enum amdgpu_mca_error_type type,
+					struct mca_bank_entry *entry)
+{
+	bool ret;
+
+	switch (type) {
+	case AMDGPU_MCA_ERROR_TYPE_CE:
+		ret = amdgpu_mca_is_deferred_error(adev, entry->regs[MCA_REG_IDX_STATUS]);
+		break;
+	case AMDGPU_MCA_ERROR_TYPE_UE:
+	default:
+		ret = true;
+		break;
+	}
+
+	return ret;
+}
+
+static int amdgpu_mca_smu_get_mca_set(struct amdgpu_device *adev, enum amdgpu_mca_error_type type, struct mca_bank_set *mca_set,
+				      struct ras_query_context *qctx)
+{
+	struct mca_bank_entry entry;
+	uint32_t count = 0, i;
+	int ret;
+
+	if (!mca_set)
+		return -EINVAL;
+
+	if (!amdgpu_mca_bank_should_update(adev, type))
+		return 0;
+
+	ret = amdgpu_mca_smu_get_valid_mca_count(adev, type, &count);
+	if (ret)
+		return ret;
+
+	for (i = 0; i < count; i++) {
+		memset(&entry, 0, sizeof(entry));
+		ret = amdgpu_mca_smu_get_mca_entry(adev, type, i, &entry);
+		if (ret)
+			return ret;
+
+		amdgpu_mca_bank_set_add_entry(mca_set, &entry);
+
+		if (amdgpu_mca_bank_should_dump(adev, type, &entry))
+			amdgpu_mca_smu_mca_bank_dump(adev, i, &entry, qctx);
+	}
+
+	return 0;
+}
+
+static int amdgpu_mca_smu_parse_mca_error_count(struct amdgpu_device *adev, enum amdgpu_ras_block blk,
+						enum amdgpu_mca_error_type type, struct mca_bank_entry *entry, uint32_t *count)
+{
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+
+	if (!count || !entry)
+		return -EINVAL;
+
+	if (!mca_funcs || !mca_funcs->mca_parse_mca_error_count)
+		return -EOPNOTSUPP;
+
+	return mca_funcs->mca_parse_mca_error_count(adev, blk, type, entry, count);
+}
+
+static int amdgpu_mca_dispatch_mca_set(struct amdgpu_device *adev, enum amdgpu_ras_block blk, enum amdgpu_mca_error_type type,
+				       struct mca_bank_set *mca_set, struct ras_err_data *err_data)
+{
+	struct amdgpu_smuio_mcm_config_info mcm_info;
+	struct mca_bank_node *node, *tmp;
+	struct mca_bank_entry *entry;
+	uint32_t count;
+	int ret;
+
+	if (!mca_set)
+		return -EINVAL;
+
+	if (!mca_set->nr_entries)
+		return 0;
+
+	list_for_each_entry_safe(node, tmp, &mca_set->list, node) {
+		entry = &node->entry;
+
+		count = 0;
+		ret = amdgpu_mca_smu_parse_mca_error_count(adev, blk, type, entry, &count);
+		if (ret && ret != -EOPNOTSUPP)
+			return ret;
+
+		if (!count)
+			continue;
+
+		memset(&mcm_info, 0, sizeof(mcm_info));
+
+		mcm_info.socket_id = entry->info.socket_id;
+		mcm_info.die_id = entry->info.aid;
+
+		if (type == AMDGPU_MCA_ERROR_TYPE_UE) {
+			amdgpu_ras_error_statistic_ue_count(err_data,
+							    &mcm_info, (uint64_t)count);
+		} else {
+			if (amdgpu_mca_is_deferred_error(adev, entry->regs[MCA_REG_IDX_STATUS]))
+				amdgpu_ras_error_statistic_de_count(err_data,
+								    &mcm_info, (uint64_t)count);
+			else
+				amdgpu_ras_error_statistic_ce_count(err_data,
+								    &mcm_info, (uint64_t)count);
+		}
+
+		amdgpu_mca_bank_set_remove_node(mca_set, node);
+	}
+
+	return 0;
+}
+
+static int amdgpu_mca_add_mca_set_to_cache(struct amdgpu_device *adev, enum amdgpu_mca_error_type type, struct mca_bank_set *new)
+{
+	struct mca_bank_cache *mca_cache = &adev->mca.mca_caches[type];
+	int ret;
+
+	mutex_lock(&mca_cache->lock);
+	ret = amdgpu_mca_bank_set_merge(&mca_cache->mca_set, new);
+	mutex_unlock(&mca_cache->lock);
+
+	return ret;
+}
+
+int amdgpu_mca_smu_log_ras_error(struct amdgpu_device *adev, enum amdgpu_ras_block blk, enum amdgpu_mca_error_type type,
+				 struct ras_err_data *err_data, struct ras_query_context *qctx)
+{
+	struct mca_bank_set mca_set;
+	struct mca_bank_cache *mca_cache = &adev->mca.mca_caches[type];
+	int ret;
+
+	amdgpu_mca_bank_set_init(&mca_set);
+
+	ret = amdgpu_mca_smu_get_mca_set(adev, type, &mca_set, qctx);
+	if (ret)
+		goto out_mca_release;
+
+	ret = amdgpu_mca_dispatch_mca_set(adev, blk, type, &mca_set, err_data);
+	if (ret)
+		goto out_mca_release;
+
+	/* add remain mca bank to mca cache */
+	if (mca_set.nr_entries) {
+		ret = amdgpu_mca_add_mca_set_to_cache(adev, type, &mca_set);
+		if (ret)
+			goto out_mca_release;
+	}
+
+	/* dispatch mca set again if mca cache has valid data */
+	mutex_lock(&mca_cache->lock);
+	if (mca_cache->mca_set.nr_entries)
+		ret = amdgpu_mca_dispatch_mca_set(adev, blk, type, &mca_cache->mca_set, err_data);
+	mutex_unlock(&mca_cache->lock);
+
+out_mca_release:
+	amdgpu_mca_bank_set_release(&mca_set);
+
+	return ret;
+}
+
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_mca_smu_debug_mode_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	int ret;
+
+	ret = amdgpu_ras_set_mca_debug_mode(adev, val ? true : false);
+	if (ret)
+		return ret;
+
+	dev_info(adev->dev, "amdgpu set smu mca debug mode %s success\n", val ? "on" : "off");
+
+	return 0;
+}
+
+static void mca_dump_entry(struct seq_file *m, struct mca_bank_entry *entry)
+{
+	int i, idx = entry->idx;
+	int reg_idx_array[] = {
+		MCA_REG_IDX_STATUS,
+		MCA_REG_IDX_ADDR,
+		MCA_REG_IDX_MISC0,
+		MCA_REG_IDX_IPID,
+		MCA_REG_IDX_SYND,
+	};
+
+	seq_printf(m, "mca entry[%d].type: %s\n", idx, entry->type == AMDGPU_MCA_ERROR_TYPE_UE ? "UE" : "CE");
+	seq_printf(m, "mca entry[%d].ip: %d\n", idx, entry->ip);
+	seq_printf(m, "mca entry[%d].info: socketid:%d aid:%d hwid:0x%03x mcatype:0x%04x\n",
+		   idx, entry->info.socket_id, entry->info.aid, entry->info.hwid, entry->info.mcatype);
+
+	for (i = 0; i < ARRAY_SIZE(reg_idx_array); i++)
+		seq_printf(m, "mca entry[%d].regs[%d]: 0x%016llx\n", idx, reg_idx_array[i], entry->regs[reg_idx_array[i]]);
+}
+
+static int mca_dump_show(struct seq_file *m, enum amdgpu_mca_error_type type)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct mca_bank_node *node;
+	struct mca_bank_set mca_set;
+	struct ras_query_context qctx;
+	int ret;
+
+	amdgpu_mca_bank_set_init(&mca_set);
+
+	qctx.evid.event_id = RAS_EVENT_INVALID_ID;
+	ret = amdgpu_mca_smu_get_mca_set(adev, type, &mca_set, &qctx);
+	if (ret)
+		goto err_free_mca_set;
+
+	seq_printf(m, "amdgpu smu %s valid mca count: %d\n",
+		   type == AMDGPU_MCA_ERROR_TYPE_UE ? "UE" : "CE", mca_set.nr_entries);
+
+	if (!mca_set.nr_entries)
+		goto err_free_mca_set;
+
+	list_for_each_entry(node, &mca_set.list, node)
+		mca_dump_entry(m, &node->entry);
+
+	/* add mca bank to mca bank cache */
+	ret = amdgpu_mca_add_mca_set_to_cache(adev, type, &mca_set);
+
+err_free_mca_set:
+	amdgpu_mca_bank_set_release(&mca_set);
+
+	return ret;
+}
+
+static int mca_dump_ce_show(struct seq_file *m, void *unused)
+{
+	return mca_dump_show(m, AMDGPU_MCA_ERROR_TYPE_CE);
+}
+
+static int mca_dump_ce_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, mca_dump_ce_show, inode->i_private);
+}
+
+static const struct file_operations mca_ce_dump_debug_fops = {
+	.owner = THIS_MODULE,
+	.open = mca_dump_ce_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = single_release,
+};
+
+static int mca_dump_ue_show(struct seq_file *m, void *unused)
+{
+	return mca_dump_show(m, AMDGPU_MCA_ERROR_TYPE_UE);
+}
+
+static int mca_dump_ue_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, mca_dump_ue_show, inode->i_private);
+}
+
+static const struct file_operations mca_ue_dump_debug_fops = {
+	.owner = THIS_MODULE,
+	.open = mca_dump_ue_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = single_release,
+};
+
+DEFINE_DEBUGFS_ATTRIBUTE(mca_debug_mode_fops, NULL, amdgpu_mca_smu_debug_mode_set, "%llu\n");
+#endif
+
+void amdgpu_mca_smu_debugfs_init(struct amdgpu_device *adev, struct dentry *root)
+{
+#if defined(CONFIG_DEBUG_FS)
+	if (!root)
+		return;
+
+	debugfs_create_file("mca_debug_mode", 0200, root, adev, &mca_debug_mode_fops);
+	debugfs_create_file("mca_ue_dump", 0400, root, adev, &mca_ue_dump_debug_fops);
+	debugfs_create_file("mca_ce_dump", 0400, root, adev, &mca_ce_dump_debug_fops);
+#endif
+}
+
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h
index 997a073e2409..e80323ff90c1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mca.h
@@ -21,6 +21,53 @@
 #ifndef __AMDGPU_MCA_H__
 #define __AMDGPU_MCA_H__
 
+#include "amdgpu_ras.h"
+
+#define MCA_MAX_REGS_COUNT	(16)
+
+#define MCA_REG_FIELD(x, h, l)			(((x) & GENMASK_ULL(h, l)) >> l)
+#define MCA_REG__STATUS__VAL(x)			MCA_REG_FIELD(x, 63, 63)
+#define MCA_REG__STATUS__OVERFLOW(x)		MCA_REG_FIELD(x, 62, 62)
+#define MCA_REG__STATUS__UC(x)			MCA_REG_FIELD(x, 61, 61)
+#define MCA_REG__STATUS__EN(x)			MCA_REG_FIELD(x, 60, 60)
+#define MCA_REG__STATUS__MISCV(x)		MCA_REG_FIELD(x, 59, 59)
+#define MCA_REG__STATUS__ADDRV(x)		MCA_REG_FIELD(x, 58, 58)
+#define MCA_REG__STATUS__PCC(x)			MCA_REG_FIELD(x, 57, 57)
+#define MCA_REG__STATUS__ERRCOREIDVAL(x)	MCA_REG_FIELD(x, 56, 56)
+#define MCA_REG__STATUS__TCC(x)			MCA_REG_FIELD(x, 55, 55)
+#define MCA_REG__STATUS__SYNDV(x)		MCA_REG_FIELD(x, 53, 53)
+#define MCA_REG__STATUS__CECC(x)		MCA_REG_FIELD(x, 46, 46)
+#define MCA_REG__STATUS__UECC(x)		MCA_REG_FIELD(x, 45, 45)
+#define MCA_REG__STATUS__DEFERRED(x)		MCA_REG_FIELD(x, 44, 44)
+#define MCA_REG__STATUS__POISON(x)		MCA_REG_FIELD(x, 43, 43)
+#define MCA_REG__STATUS__SCRUB(x)		MCA_REG_FIELD(x, 40, 40)
+#define MCA_REG__STATUS__ERRCOREID(x)		MCA_REG_FIELD(x, 37, 32)
+#define MCA_REG__STATUS__ADDRLSB(x)		MCA_REG_FIELD(x, 29, 24)
+#define MCA_REG__STATUS__ERRORCODEEXT(x)	MCA_REG_FIELD(x, 21, 16)
+#define MCA_REG__STATUS__ERRORCODE(x)		MCA_REG_FIELD(x, 15, 0)
+
+#define MCA_REG__MISC0__ERRCNT(x)		MCA_REG_FIELD(x, 43, 32)
+
+#define MCA_REG__SYND__ERRORINFORMATION(x)	MCA_REG_FIELD(x, 17, 0)
+
+enum amdgpu_mca_ip {
+	AMDGPU_MCA_IP_UNKNOW = -1,
+	AMDGPU_MCA_IP_PSP = 0,
+	AMDGPU_MCA_IP_SDMA,
+	AMDGPU_MCA_IP_GC,
+	AMDGPU_MCA_IP_SMU,
+	AMDGPU_MCA_IP_MP5,
+	AMDGPU_MCA_IP_UMC,
+	AMDGPU_MCA_IP_PCS_XGMI,
+	AMDGPU_MCA_IP_COUNT,
+};
+
+enum amdgpu_mca_error_type {
+	AMDGPU_MCA_ERROR_TYPE_UE = 0,
+	AMDGPU_MCA_ERROR_TYPE_CE,
+	AMDGPU_MCA_ERROR_TYPE_DE,
+};
+
 struct amdgpu_mca_ras_block {
 	struct amdgpu_ras_block_object ras_block;
 };
@@ -30,10 +77,64 @@ struct amdgpu_mca_ras {
 	struct amdgpu_mca_ras_block *ras;
 };
 
+struct mca_bank_set {
+	int nr_entries;
+	struct list_head list;
+};
+
+struct mca_bank_cache {
+	struct mca_bank_set mca_set;
+	struct mutex lock;
+};
+
 struct amdgpu_mca {
 	struct amdgpu_mca_ras mp0;
 	struct amdgpu_mca_ras mp1;
 	struct amdgpu_mca_ras mpio;
+	const struct amdgpu_mca_smu_funcs *mca_funcs;
+	struct mca_bank_cache mca_caches[AMDGPU_MCA_ERROR_TYPE_DE];
+	atomic_t ue_update_flag;
+};
+
+enum mca_reg_idx {
+	MCA_REG_IDX_STATUS		= 1,
+	MCA_REG_IDX_ADDR		= 2,
+	MCA_REG_IDX_MISC0		= 3,
+	MCA_REG_IDX_IPID		= 5,
+	MCA_REG_IDX_SYND		= 6,
+	MCA_REG_IDX_COUNT		= 16,
+};
+
+struct mca_bank_info {
+	int socket_id;
+	int aid;
+	int hwid;
+	int mcatype;
+};
+
+struct mca_bank_entry {
+	int idx;
+	enum amdgpu_mca_error_type type;
+	enum amdgpu_mca_ip ip;
+	struct mca_bank_info info;
+	uint64_t regs[MCA_MAX_REGS_COUNT];
+};
+
+struct mca_bank_node {
+	struct mca_bank_entry entry;
+	struct list_head node;
+};
+
+struct amdgpu_mca_smu_funcs {
+	int max_ue_count;
+	int max_ce_count;
+	int (*mca_set_debug_mode)(struct amdgpu_device *adev, bool enable);
+	int (*mca_parse_mca_error_count)(struct amdgpu_device *adev, enum amdgpu_ras_block blk, enum amdgpu_mca_error_type type,
+					 struct mca_bank_entry *entry, uint32_t *count);
+	int (*mca_get_valid_mca_count)(struct amdgpu_device *adev, enum amdgpu_mca_error_type type,
+				       uint32_t *count);
+	int (*mca_get_mca_entry)(struct amdgpu_device *adev, enum amdgpu_mca_error_type type,
+				 int idx, struct mca_bank_entry *entry);
 };
 
 void amdgpu_mca_query_correctable_error_count(struct amdgpu_device *adev,
@@ -53,4 +154,16 @@ void amdgpu_mca_query_ras_error_count(struct amdgpu_device *adev,
 int amdgpu_mca_mp0_ras_sw_init(struct amdgpu_device *adev);
 int amdgpu_mca_mp1_ras_sw_init(struct amdgpu_device *adev);
 int amdgpu_mca_mpio_ras_sw_init(struct amdgpu_device *adev);
+
+void amdgpu_mca_smu_init_funcs(struct amdgpu_device *adev, const struct amdgpu_mca_smu_funcs *mca_funcs);
+int amdgpu_mca_init(struct amdgpu_device *adev);
+void amdgpu_mca_fini(struct amdgpu_device *adev);
+int amdgpu_mca_reset(struct amdgpu_device *adev);
+int amdgpu_mca_smu_set_debug_mode(struct amdgpu_device *adev, bool enable);
+int amdgpu_mca_smu_get_mca_set_error_count(struct amdgpu_device *adev, enum amdgpu_ras_block blk,
+					   enum amdgpu_mca_error_type type, uint32_t *total);
+void amdgpu_mca_smu_debugfs_init(struct amdgpu_device *adev, struct dentry *root);
+int amdgpu_mca_smu_log_ras_error(struct amdgpu_device *adev, enum amdgpu_ras_block blk, enum amdgpu_mca_error_type type,
+				 struct ras_err_data *err_data, struct ras_query_context *qctx);
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
index 0e55823ef6ca..9c182ce501af 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
@@ -22,6 +22,7 @@
  */
 
 #include <linux/firmware.h>
+#include <drm/drm_exec.h>
 
 #include "amdgpu_mes.h"
 #include "amdgpu.h"
@@ -38,123 +39,58 @@ int amdgpu_mes_doorbell_process_slice(struct amdgpu_device *adev)
 		       PAGE_SIZE);
 }
 
-int amdgpu_mes_alloc_process_doorbells(struct amdgpu_device *adev,
-				      unsigned int *doorbell_index)
-{
-	int r = ida_simple_get(&adev->mes.doorbell_ida, 2,
-			       adev->mes.max_doorbell_slices,
-			       GFP_KERNEL);
-	if (r > 0)
-		*doorbell_index = r;
-
-	return r;
-}
-
-void amdgpu_mes_free_process_doorbells(struct amdgpu_device *adev,
-				      unsigned int doorbell_index)
-{
-	if (doorbell_index)
-		ida_simple_remove(&adev->mes.doorbell_ida, doorbell_index);
-}
-
-unsigned int amdgpu_mes_get_doorbell_dw_offset_in_bar(
-					struct amdgpu_device *adev,
-					uint32_t doorbell_index,
-					unsigned int doorbell_id)
-{
-	return ((doorbell_index *
-		amdgpu_mes_doorbell_process_slice(adev)) / sizeof(u32) +
-		doorbell_id * 2);
-}
-
-static int amdgpu_mes_queue_doorbell_get(struct amdgpu_device *adev,
-					 struct amdgpu_mes_process *process,
-					 int ip_type, uint64_t *doorbell_index)
+static int amdgpu_mes_doorbell_init(struct amdgpu_device *adev)
 {
-	unsigned int offset, found;
+	int i;
+	struct amdgpu_mes *mes = &adev->mes;
 
-	if (ip_type == AMDGPU_RING_TYPE_SDMA) {
-		offset = adev->doorbell_index.sdma_engine[0];
-		found = find_next_zero_bit(process->doorbell_bitmap,
-					   AMDGPU_MES_MAX_NUM_OF_QUEUES_PER_PROCESS,
-					   offset);
-	} else {
-		found = find_first_zero_bit(process->doorbell_bitmap,
-					    AMDGPU_MES_MAX_NUM_OF_QUEUES_PER_PROCESS);
+	/* Bitmap for dynamic allocation of kernel doorbells */
+	mes->doorbell_bitmap = bitmap_zalloc(PAGE_SIZE / sizeof(u32), GFP_KERNEL);
+	if (!mes->doorbell_bitmap) {
+		dev_err(adev->dev, "Failed to allocate MES doorbell bitmap\n");
+		return -ENOMEM;
 	}
 
-	if (found >= AMDGPU_MES_MAX_NUM_OF_QUEUES_PER_PROCESS) {
-		DRM_WARN("No doorbell available\n");
-		return -ENOSPC;
+	mes->num_mes_dbs = PAGE_SIZE / AMDGPU_ONE_DOORBELL_SIZE;
+	for (i = 0; i < AMDGPU_MES_PRIORITY_NUM_LEVELS; i++) {
+		adev->mes.aggregated_doorbells[i] = mes->db_start_dw_offset + i * 2;
+		set_bit(i, mes->doorbell_bitmap);
 	}
 
-	set_bit(found, process->doorbell_bitmap);
-
-	*doorbell_index = amdgpu_mes_get_doorbell_dw_offset_in_bar(adev,
-				process->doorbell_index, found);
-
 	return 0;
 }
 
-static void amdgpu_mes_queue_doorbell_free(struct amdgpu_device *adev,
-					   struct amdgpu_mes_process *process,
-					   uint32_t doorbell_index)
-{
-	unsigned int old, doorbell_id;
-
-	doorbell_id = doorbell_index -
-		(process->doorbell_index *
-		 amdgpu_mes_doorbell_process_slice(adev)) / sizeof(u32);
-	doorbell_id /= 2;
-
-	old = test_and_clear_bit(doorbell_id, process->doorbell_bitmap);
-	WARN_ON(!old);
-}
-
-static int amdgpu_mes_doorbell_init(struct amdgpu_device *adev)
+static int amdgpu_mes_event_log_init(struct amdgpu_device *adev)
 {
-	size_t doorbell_start_offset;
-	size_t doorbell_aperture_size;
-	size_t doorbell_process_limit;
-	size_t aggregated_doorbell_start;
-	int i;
-
-	aggregated_doorbell_start = (adev->doorbell_index.max_assignment + 1) * sizeof(u32);
-	aggregated_doorbell_start =
-		roundup(aggregated_doorbell_start, PAGE_SIZE);
+	int r;
 
-	doorbell_start_offset = aggregated_doorbell_start + PAGE_SIZE;
-	doorbell_start_offset =
-		roundup(doorbell_start_offset,
-			amdgpu_mes_doorbell_process_slice(adev));
+	if (!amdgpu_mes_log_enable)
+		return 0;
 
-	doorbell_aperture_size = adev->doorbell.size;
-	doorbell_aperture_size =
-			rounddown(doorbell_aperture_size,
-				  amdgpu_mes_doorbell_process_slice(adev));
+	r = amdgpu_bo_create_kernel(adev, adev->mes.event_log_size, PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_VRAM,
+				    &adev->mes.event_log_gpu_obj,
+				    &adev->mes.event_log_gpu_addr,
+				    &adev->mes.event_log_cpu_addr);
+	if (r) {
+		dev_warn(adev->dev, "failed to create MES event log buffer (%d)", r);
+		return r;
+	}
 
-	if (doorbell_aperture_size > doorbell_start_offset)
-		doorbell_process_limit =
-			(doorbell_aperture_size - doorbell_start_offset) /
-			amdgpu_mes_doorbell_process_slice(adev);
-	else
-		return -ENOSPC;
+	memset(adev->mes.event_log_cpu_addr, 0, adev->mes.event_log_size);
 
-	adev->mes.doorbell_id_offset = doorbell_start_offset / sizeof(u32);
-	adev->mes.max_doorbell_slices = doorbell_process_limit;
+	return  0;
 
-	/* allocate Qword range for aggregated doorbell */
-	for (i = 0; i < AMDGPU_MES_PRIORITY_NUM_LEVELS; i++)
-		adev->mes.aggregated_doorbells[i] =
-			aggregated_doorbell_start / sizeof(u32) + i * 2;
+}
 
-	DRM_INFO("max_doorbell_slices=%zu\n", doorbell_process_limit);
-	return 0;
+static void amdgpu_mes_doorbell_free(struct amdgpu_device *adev)
+{
+	bitmap_free(adev->mes.doorbell_bitmap);
 }
 
 int amdgpu_mes_init(struct amdgpu_device *adev)
 {
-	int i, r;
+	int i, r, num_pipes;
 
 	adev->mes.adev = adev;
 
@@ -163,80 +99,125 @@ int amdgpu_mes_init(struct amdgpu_device *adev)
 	idr_init(&adev->mes.queue_id_idr);
 	ida_init(&adev->mes.doorbell_ida);
 	spin_lock_init(&adev->mes.queue_id_lock);
-	spin_lock_init(&adev->mes.ring_lock);
 	mutex_init(&adev->mes.mutex_hidden);
 
-	adev->mes.total_max_queue = AMDGPU_FENCE_MES_QUEUE_ID_MASK;
-	adev->mes.vmid_mask_mmhub = 0xffffff00;
-	adev->mes.vmid_mask_gfxhub = 0xffffff00;
+	for (i = 0; i < AMDGPU_MAX_MES_PIPES; i++)
+		spin_lock_init(&adev->mes.ring_lock[i]);
 
-	for (i = 0; i < AMDGPU_MES_MAX_COMPUTE_PIPES; i++) {
-		/* use only 1st MEC pipes */
-		if (i >= 4)
-			continue;
-		adev->mes.compute_hqd_mask[i] = 0xc;
-	}
+	adev->mes.total_max_queue = AMDGPU_FENCE_MES_QUEUE_ID_MASK;
+	adev->mes.vmid_mask_mmhub = 0xFF00;
+	adev->mes.vmid_mask_gfxhub = adev->gfx.disable_kq ? 0xFFFE : 0xFF00;
 
-	for (i = 0; i < AMDGPU_MES_MAX_GFX_PIPES; i++)
-		adev->mes.gfx_hqd_mask[i] = i ? 0 : 0xfffffffe;
+	num_pipes = adev->gfx.me.num_pipe_per_me * adev->gfx.me.num_me;
+	if (num_pipes > AMDGPU_MES_MAX_GFX_PIPES)
+		dev_warn(adev->dev, "more gfx pipes than supported by MES! (%d vs %d)\n",
+			 num_pipes, AMDGPU_MES_MAX_GFX_PIPES);
 
-	for (i = 0; i < AMDGPU_MES_MAX_SDMA_PIPES; i++) {
-		if (adev->ip_versions[SDMA0_HWIP][0] < IP_VERSION(6, 0, 0))
-			adev->mes.sdma_hqd_mask[i] = i ? 0 : 0x3fc;
-		/* zero sdma_hqd_mask for non-existent engine */
-		else if (adev->sdma.num_instances == 1)
-			adev->mes.sdma_hqd_mask[i] = i ? 0 : 0xfc;
+	for (i = 0; i < AMDGPU_MES_MAX_GFX_PIPES; i++) {
+		if (i >= num_pipes)
+			break;
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) >=
+		    IP_VERSION(12, 0, 0))
+			/*
+			 * GFX V12 has only one GFX pipe, but 8 queues in it.
+			 * GFX pipe 0 queue 0 is being used by Kernel queue.
+			 * Set GFX pipe 0 queue 1-7 for MES scheduling
+			 * mask = 1111 1110b
+			 */
+			adev->mes.gfx_hqd_mask[i] = adev->gfx.disable_kq ? 0xFF : 0xFE;
 		else
-			adev->mes.sdma_hqd_mask[i] = 0xfc;
+			/*
+			 * GFX pipe 0 queue 0 is being used by Kernel queue.
+			 * Set GFX pipe 0 queue 1 for MES scheduling
+			 * mask = 10b
+			 */
+			adev->mes.gfx_hqd_mask[i] = adev->gfx.disable_kq ? 0x3 : 0x2;
 	}
 
-	r = amdgpu_device_wb_get(adev, &adev->mes.sch_ctx_offs);
-	if (r) {
-		dev_err(adev->dev,
-			"(%d) ring trail_fence_offs wb alloc failed\n", r);
-		goto error_ids;
+	num_pipes = adev->gfx.mec.num_pipe_per_mec * adev->gfx.mec.num_mec;
+	if (num_pipes > AMDGPU_MES_MAX_COMPUTE_PIPES)
+		dev_warn(adev->dev, "more compute pipes than supported by MES! (%d vs %d)\n",
+			 num_pipes, AMDGPU_MES_MAX_COMPUTE_PIPES);
+
+	for (i = 0; i < AMDGPU_MES_MAX_COMPUTE_PIPES; i++) {
+		if (i >= num_pipes)
+			break;
+		adev->mes.compute_hqd_mask[i] = adev->gfx.disable_kq ? 0xF : 0xC;
 	}
-	adev->mes.sch_ctx_gpu_addr =
-		adev->wb.gpu_addr + (adev->mes.sch_ctx_offs * 4);
-	adev->mes.sch_ctx_ptr =
-		(uint64_t *)&adev->wb.wb[adev->mes.sch_ctx_offs];
 
-	r = amdgpu_device_wb_get(adev, &adev->mes.query_status_fence_offs);
-	if (r) {
-		amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs);
-		dev_err(adev->dev,
-			"(%d) query_status_fence_offs wb alloc failed\n", r);
-		goto error_ids;
+	num_pipes = adev->sdma.num_instances;
+	if (num_pipes > AMDGPU_MES_MAX_SDMA_PIPES)
+		dev_warn(adev->dev, "more SDMA pipes than supported by MES! (%d vs %d)\n",
+			 num_pipes, AMDGPU_MES_MAX_SDMA_PIPES);
+
+	for (i = 0; i < AMDGPU_MES_MAX_SDMA_PIPES; i++) {
+		if (i >= num_pipes)
+			break;
+		adev->mes.sdma_hqd_mask[i] = 0xfc;
 	}
-	adev->mes.query_status_fence_gpu_addr =
-		adev->wb.gpu_addr + (adev->mes.query_status_fence_offs * 4);
-	adev->mes.query_status_fence_ptr =
-		(uint64_t *)&adev->wb.wb[adev->mes.query_status_fence_offs];
 
-	r = amdgpu_device_wb_get(adev, &adev->mes.read_val_offs);
-	if (r) {
-		amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs);
-		amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs);
-		dev_err(adev->dev,
-			"(%d) read_val_offs alloc failed\n", r);
-		goto error_ids;
+	for (i = 0; i < AMDGPU_MAX_MES_PIPES; i++) {
+		r = amdgpu_device_wb_get(adev, &adev->mes.sch_ctx_offs[i]);
+		if (r) {
+			dev_err(adev->dev,
+				"(%d) ring trail_fence_offs wb alloc failed\n",
+				r);
+			goto error;
+		}
+		adev->mes.sch_ctx_gpu_addr[i] =
+			adev->wb.gpu_addr + (adev->mes.sch_ctx_offs[i] * 4);
+		adev->mes.sch_ctx_ptr[i] =
+			(uint64_t *)&adev->wb.wb[adev->mes.sch_ctx_offs[i]];
+
+		r = amdgpu_device_wb_get(adev,
+				 &adev->mes.query_status_fence_offs[i]);
+		if (r) {
+			dev_err(adev->dev,
+			      "(%d) query_status_fence_offs wb alloc failed\n",
+			      r);
+			goto error;
+		}
+		adev->mes.query_status_fence_gpu_addr[i] = adev->wb.gpu_addr +
+			(adev->mes.query_status_fence_offs[i] * 4);
+		adev->mes.query_status_fence_ptr[i] =
+			(uint64_t *)&adev->wb.wb[adev->mes.query_status_fence_offs[i]];
 	}
-	adev->mes.read_val_gpu_addr =
-		adev->wb.gpu_addr + (adev->mes.read_val_offs * 4);
-	adev->mes.read_val_ptr =
-		(uint32_t *)&adev->wb.wb[adev->mes.read_val_offs];
 
 	r = amdgpu_mes_doorbell_init(adev);
 	if (r)
 		goto error;
 
+	r = amdgpu_mes_event_log_init(adev);
+	if (r)
+		goto error_doorbell;
+
+	if (adev->mes.hung_queue_db_array_size) {
+		r = amdgpu_bo_create_kernel(adev,
+					    adev->mes.hung_queue_db_array_size * sizeof(u32),
+					    PAGE_SIZE,
+					    AMDGPU_GEM_DOMAIN_GTT,
+					    &adev->mes.hung_queue_db_array_gpu_obj,
+					    &adev->mes.hung_queue_db_array_gpu_addr,
+					    &adev->mes.hung_queue_db_array_cpu_addr);
+		if (r) {
+			dev_warn(adev->dev, "failed to create MES hung db array buffer (%d)", r);
+			goto error_doorbell;
+		}
+	}
+
 	return 0;
 
+error_doorbell:
+	amdgpu_mes_doorbell_free(adev);
 error:
-	amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs);
-	amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs);
-	amdgpu_device_wb_free(adev, adev->mes.read_val_offs);
-error_ids:
+	for (i = 0; i < AMDGPU_MAX_MES_PIPES; i++) {
+		if (adev->mes.sch_ctx_ptr[i])
+			amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs[i]);
+		if (adev->mes.query_status_fence_ptr[i])
+			amdgpu_device_wb_free(adev,
+				      adev->mes.query_status_fence_offs[i]);
+	}
+
 	idr_destroy(&adev->mes.pasid_idr);
 	idr_destroy(&adev->mes.gang_id_idr);
 	idr_destroy(&adev->mes.queue_id_idr);
@@ -247,9 +228,25 @@ error_ids:
 
 void amdgpu_mes_fini(struct amdgpu_device *adev)
 {
-	amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs);
-	amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs);
-	amdgpu_device_wb_free(adev, adev->mes.read_val_offs);
+	int i;
+
+	amdgpu_bo_free_kernel(&adev->mes.hung_queue_db_array_gpu_obj,
+			      &adev->mes.hung_queue_db_array_gpu_addr,
+			      &adev->mes.hung_queue_db_array_cpu_addr);
+
+	amdgpu_bo_free_kernel(&adev->mes.event_log_gpu_obj,
+			      &adev->mes.event_log_gpu_addr,
+			      &adev->mes.event_log_cpu_addr);
+
+	for (i = 0; i < AMDGPU_MAX_MES_PIPES; i++) {
+		if (adev->mes.sch_ctx_ptr[i])
+			amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs[i]);
+		if (adev->mes.query_status_fence_ptr[i])
+			amdgpu_device_wb_free(adev,
+				      adev->mes.query_status_fence_offs[i]);
+	}
+
+	amdgpu_mes_doorbell_free(adev);
 
 	idr_destroy(&adev->mes.pasid_idr);
 	idr_destroy(&adev->mes.gang_id_idr);
@@ -258,568 +255,183 @@ void amdgpu_mes_fini(struct amdgpu_device *adev)
 	mutex_destroy(&adev->mes.mutex_hidden);
 }
 
-static void amdgpu_mes_queue_free_mqd(struct amdgpu_mes_queue *q)
-{
-	amdgpu_bo_free_kernel(&q->mqd_obj,
-			      &q->mqd_gpu_addr,
-			      &q->mqd_cpu_ptr);
-}
-
-int amdgpu_mes_create_process(struct amdgpu_device *adev, int pasid,
-			      struct amdgpu_vm *vm)
+int amdgpu_mes_suspend(struct amdgpu_device *adev)
 {
-	struct amdgpu_mes_process *process;
+	struct mes_suspend_gang_input input;
 	int r;
 
-	/* allocate the mes process buffer */
-	process = kzalloc(sizeof(struct amdgpu_mes_process), GFP_KERNEL);
-	if (!process) {
-		DRM_ERROR("no more memory to create mes process\n");
-		return -ENOMEM;
-	}
+	if (!amdgpu_mes_suspend_resume_all_supported(adev))
+		return 0;
 
-	process->doorbell_bitmap =
-		kzalloc(DIV_ROUND_UP(AMDGPU_MES_MAX_NUM_OF_QUEUES_PER_PROCESS,
-				     BITS_PER_BYTE), GFP_KERNEL);
-	if (!process->doorbell_bitmap) {
-		DRM_ERROR("failed to allocate doorbell bitmap\n");
-		kfree(process);
-		return -ENOMEM;
-	}
-
-	/* allocate the process context bo and map it */
-	r = amdgpu_bo_create_kernel(adev, AMDGPU_MES_PROC_CTX_SIZE, PAGE_SIZE,
-				    AMDGPU_GEM_DOMAIN_GTT,
-				    &process->proc_ctx_bo,
-				    &process->proc_ctx_gpu_addr,
-				    &process->proc_ctx_cpu_ptr);
-	if (r) {
-		DRM_ERROR("failed to allocate process context bo\n");
-		goto clean_up_memory;
-	}
-	memset(process->proc_ctx_cpu_ptr, 0, AMDGPU_MES_PROC_CTX_SIZE);
+	memset(&input, 0x0, sizeof(struct mes_suspend_gang_input));
+	input.suspend_all_gangs = 1;
 
 	/*
 	 * Avoid taking any other locks under MES lock to avoid circular
 	 * lock dependencies.
 	 */
 	amdgpu_mes_lock(&adev->mes);
-
-	/* add the mes process to idr list */
-	r = idr_alloc(&adev->mes.pasid_idr, process, pasid, pasid + 1,
-		      GFP_KERNEL);
-	if (r < 0) {
-		DRM_ERROR("failed to lock pasid=%d\n", pasid);
-		goto clean_up_ctx;
-	}
-
-	/* allocate the starting doorbell index of the process */
-	r = amdgpu_mes_alloc_process_doorbells(adev, &process->doorbell_index);
-	if (r < 0) {
-		DRM_ERROR("failed to allocate doorbell for process\n");
-		goto clean_up_pasid;
-	}
-
-	DRM_DEBUG("process doorbell index = %d\n", process->doorbell_index);
-
-	INIT_LIST_HEAD(&process->gang_list);
-	process->vm = vm;
-	process->pasid = pasid;
-	process->process_quantum = adev->mes.default_process_quantum;
-	process->pd_gpu_addr = amdgpu_bo_gpu_offset(vm->root.bo);
-
+	r = adev->mes.funcs->suspend_gang(&adev->mes, &input);
 	amdgpu_mes_unlock(&adev->mes);
-	return 0;
+	if (r)
+		dev_err(adev->dev, "failed to suspend all gangs");
 
-clean_up_pasid:
-	idr_remove(&adev->mes.pasid_idr, pasid);
-	amdgpu_mes_unlock(&adev->mes);
-clean_up_ctx:
-	amdgpu_bo_free_kernel(&process->proc_ctx_bo,
-			      &process->proc_ctx_gpu_addr,
-			      &process->proc_ctx_cpu_ptr);
-clean_up_memory:
-	kfree(process->doorbell_bitmap);
-	kfree(process);
 	return r;
 }
 
-void amdgpu_mes_destroy_process(struct amdgpu_device *adev, int pasid)
-{
-	struct amdgpu_mes_process *process;
-	struct amdgpu_mes_gang *gang, *tmp1;
-	struct amdgpu_mes_queue *queue, *tmp2;
-	struct mes_remove_queue_input queue_input;
-	unsigned long flags;
-	int r;
-
-	/*
-	 * Avoid taking any other locks under MES lock to avoid circular
-	 * lock dependencies.
-	 */
-	amdgpu_mes_lock(&adev->mes);
-
-	process = idr_find(&adev->mes.pasid_idr, pasid);
-	if (!process) {
-		DRM_WARN("pasid %d doesn't exist\n", pasid);
-		amdgpu_mes_unlock(&adev->mes);
-		return;
-	}
-
-	/* Remove all queues from hardware */
-	list_for_each_entry_safe(gang, tmp1, &process->gang_list, list) {
-		list_for_each_entry_safe(queue, tmp2, &gang->queue_list, list) {
-			spin_lock_irqsave(&adev->mes.queue_id_lock, flags);
-			idr_remove(&adev->mes.queue_id_idr, queue->queue_id);
-			spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags);
-
-			queue_input.doorbell_offset = queue->doorbell_off;
-			queue_input.gang_context_addr = gang->gang_ctx_gpu_addr;
-
-			r = adev->mes.funcs->remove_hw_queue(&adev->mes,
-							     &queue_input);
-			if (r)
-				DRM_WARN("failed to remove hardware queue\n");
-		}
-
-		idr_remove(&adev->mes.gang_id_idr, gang->gang_id);
-	}
-
-	amdgpu_mes_free_process_doorbells(adev, process->doorbell_index);
-	idr_remove(&adev->mes.pasid_idr, pasid);
-	amdgpu_mes_unlock(&adev->mes);
-
-	/* free all memory allocated by the process */
-	list_for_each_entry_safe(gang, tmp1, &process->gang_list, list) {
-		/* free all queues in the gang */
-		list_for_each_entry_safe(queue, tmp2, &gang->queue_list, list) {
-			amdgpu_mes_queue_free_mqd(queue);
-			list_del(&queue->list);
-			kfree(queue);
-		}
-		amdgpu_bo_free_kernel(&gang->gang_ctx_bo,
-				      &gang->gang_ctx_gpu_addr,
-				      &gang->gang_ctx_cpu_ptr);
-		list_del(&gang->list);
-		kfree(gang);
-
-	}
-	amdgpu_bo_free_kernel(&process->proc_ctx_bo,
-			      &process->proc_ctx_gpu_addr,
-			      &process->proc_ctx_cpu_ptr);
-	kfree(process->doorbell_bitmap);
-	kfree(process);
-}
-
-int amdgpu_mes_add_gang(struct amdgpu_device *adev, int pasid,
-			struct amdgpu_mes_gang_properties *gprops,
-			int *gang_id)
+int amdgpu_mes_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_mes_process *process;
-	struct amdgpu_mes_gang *gang;
+	struct mes_resume_gang_input input;
 	int r;
 
-	/* allocate the mes gang buffer */
-	gang = kzalloc(sizeof(struct amdgpu_mes_gang), GFP_KERNEL);
-	if (!gang) {
-		return -ENOMEM;
-	}
+	if (!amdgpu_mes_suspend_resume_all_supported(adev))
+		return 0;
 
-	/* allocate the gang context bo and map it to cpu space */
-	r = amdgpu_bo_create_kernel(adev, AMDGPU_MES_GANG_CTX_SIZE, PAGE_SIZE,
-				    AMDGPU_GEM_DOMAIN_GTT,
-				    &gang->gang_ctx_bo,
-				    &gang->gang_ctx_gpu_addr,
-				    &gang->gang_ctx_cpu_ptr);
-	if (r) {
-		DRM_ERROR("failed to allocate process context bo\n");
-		goto clean_up_mem;
-	}
-	memset(gang->gang_ctx_cpu_ptr, 0, AMDGPU_MES_GANG_CTX_SIZE);
+	memset(&input, 0x0, sizeof(struct mes_resume_gang_input));
+	input.resume_all_gangs = 1;
 
 	/*
 	 * Avoid taking any other locks under MES lock to avoid circular
 	 * lock dependencies.
 	 */
 	amdgpu_mes_lock(&adev->mes);
-
-	process = idr_find(&adev->mes.pasid_idr, pasid);
-	if (!process) {
-		DRM_ERROR("pasid %d doesn't exist\n", pasid);
-		r = -EINVAL;
-		goto clean_up_ctx;
-	}
-
-	/* add the mes gang to idr list */
-	r = idr_alloc(&adev->mes.gang_id_idr, gang, 1, 0,
-		      GFP_KERNEL);
-	if (r < 0) {
-		DRM_ERROR("failed to allocate idr for gang\n");
-		goto clean_up_ctx;
-	}
-
-	gang->gang_id = r;
-	*gang_id = r;
-
-	INIT_LIST_HEAD(&gang->queue_list);
-	gang->process = process;
-	gang->priority = gprops->priority;
-	gang->gang_quantum = gprops->gang_quantum ?
-		gprops->gang_quantum : adev->mes.default_gang_quantum;
-	gang->global_priority_level = gprops->global_priority_level;
-	gang->inprocess_gang_priority = gprops->inprocess_gang_priority;
-	list_add_tail(&gang->list, &process->gang_list);
-
+	r = adev->mes.funcs->resume_gang(&adev->mes, &input);
 	amdgpu_mes_unlock(&adev->mes);
-	return 0;
+	if (r)
+		dev_err(adev->dev, "failed to resume all gangs");
 
-clean_up_ctx:
-	amdgpu_mes_unlock(&adev->mes);
-	amdgpu_bo_free_kernel(&gang->gang_ctx_bo,
-			      &gang->gang_ctx_gpu_addr,
-			      &gang->gang_ctx_cpu_ptr);
-clean_up_mem:
-	kfree(gang);
 	return r;
 }
 
-int amdgpu_mes_remove_gang(struct amdgpu_device *adev, int gang_id)
+int amdgpu_mes_map_legacy_queue(struct amdgpu_device *adev,
+				struct amdgpu_ring *ring)
 {
-	struct amdgpu_mes_gang *gang;
-
-	/*
-	 * Avoid taking any other locks under MES lock to avoid circular
-	 * lock dependencies.
-	 */
-	amdgpu_mes_lock(&adev->mes);
-
-	gang = idr_find(&adev->mes.gang_id_idr, gang_id);
-	if (!gang) {
-		DRM_ERROR("gang id %d doesn't exist\n", gang_id);
-		amdgpu_mes_unlock(&adev->mes);
-		return -EINVAL;
-	}
-
-	if (!list_empty(&gang->queue_list)) {
-		DRM_ERROR("queue list is not empty\n");
-		amdgpu_mes_unlock(&adev->mes);
-		return -EBUSY;
-	}
-
-	idr_remove(&adev->mes.gang_id_idr, gang->gang_id);
-	list_del(&gang->list);
-	amdgpu_mes_unlock(&adev->mes);
-
-	amdgpu_bo_free_kernel(&gang->gang_ctx_bo,
-			      &gang->gang_ctx_gpu_addr,
-			      &gang->gang_ctx_cpu_ptr);
+	struct mes_map_legacy_queue_input queue_input;
+	int r;
 
-	kfree(gang);
+	memset(&queue_input, 0, sizeof(queue_input));
 
-	return 0;
-}
-
-int amdgpu_mes_suspend(struct amdgpu_device *adev)
-{
-	struct idr *idp;
-	struct amdgpu_mes_process *process;
-	struct amdgpu_mes_gang *gang;
-	struct mes_suspend_gang_input input;
-	int r, pasid;
+	queue_input.queue_type = ring->funcs->type;
+	queue_input.doorbell_offset = ring->doorbell_index;
+	queue_input.pipe_id = ring->pipe;
+	queue_input.queue_id = ring->queue;
+	queue_input.mqd_addr = amdgpu_bo_gpu_offset(ring->mqd_obj);
+	queue_input.wptr_addr = ring->wptr_gpu_addr;
 
-	/*
-	 * Avoid taking any other locks under MES lock to avoid circular
-	 * lock dependencies.
-	 */
 	amdgpu_mes_lock(&adev->mes);
-
-	idp = &adev->mes.pasid_idr;
-
-	idr_for_each_entry(idp, process, pasid) {
-		list_for_each_entry(gang, &process->gang_list, list) {
-			r = adev->mes.funcs->suspend_gang(&adev->mes, &input);
-			if (r)
-				DRM_ERROR("failed to suspend pasid %d gangid %d",
-					 pasid, gang->gang_id);
-		}
-	}
-
+	r = adev->mes.funcs->map_legacy_queue(&adev->mes, &queue_input);
 	amdgpu_mes_unlock(&adev->mes);
-	return 0;
-}
-
-int amdgpu_mes_resume(struct amdgpu_device *adev)
-{
-	struct idr *idp;
-	struct amdgpu_mes_process *process;
-	struct amdgpu_mes_gang *gang;
-	struct mes_resume_gang_input input;
-	int r, pasid;
-
-	/*
-	 * Avoid taking any other locks under MES lock to avoid circular
-	 * lock dependencies.
-	 */
-	amdgpu_mes_lock(&adev->mes);
-
-	idp = &adev->mes.pasid_idr;
-
-	idr_for_each_entry(idp, process, pasid) {
-		list_for_each_entry(gang, &process->gang_list, list) {
-			r = adev->mes.funcs->resume_gang(&adev->mes, &input);
-			if (r)
-				DRM_ERROR("failed to resume pasid %d gangid %d",
-					 pasid, gang->gang_id);
-		}
-	}
+	if (r)
+		dev_err(adev->dev, "failed to map legacy queue\n");
 
-	amdgpu_mes_unlock(&adev->mes);
-	return 0;
+	return r;
 }
 
-static int amdgpu_mes_queue_alloc_mqd(struct amdgpu_device *adev,
-				     struct amdgpu_mes_queue *q,
-				     struct amdgpu_mes_queue_properties *p)
+int amdgpu_mes_unmap_legacy_queue(struct amdgpu_device *adev,
+				  struct amdgpu_ring *ring,
+				  enum amdgpu_unmap_queues_action action,
+				  u64 gpu_addr, u64 seq)
 {
-	struct amdgpu_mqd *mqd_mgr = &adev->mqds[p->queue_type];
-	u32 mqd_size = mqd_mgr->mqd_size;
+	struct mes_unmap_legacy_queue_input queue_input;
 	int r;
 
-	r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE,
-				    AMDGPU_GEM_DOMAIN_GTT,
-				    &q->mqd_obj,
-				    &q->mqd_gpu_addr, &q->mqd_cpu_ptr);
-	if (r) {
-		dev_warn(adev->dev, "failed to create queue mqd bo (%d)", r);
-		return r;
-	}
-	memset(q->mqd_cpu_ptr, 0, mqd_size);
-
-	r = amdgpu_bo_reserve(q->mqd_obj, false);
-	if (unlikely(r != 0))
-		goto clean_up;
+	queue_input.action = action;
+	queue_input.queue_type = ring->funcs->type;
+	queue_input.doorbell_offset = ring->doorbell_index;
+	queue_input.pipe_id = ring->pipe;
+	queue_input.queue_id = ring->queue;
+	queue_input.trail_fence_addr = gpu_addr;
+	queue_input.trail_fence_data = seq;
 
-	return 0;
+	amdgpu_mes_lock(&adev->mes);
+	r = adev->mes.funcs->unmap_legacy_queue(&adev->mes, &queue_input);
+	amdgpu_mes_unlock(&adev->mes);
+	if (r)
+		dev_err(adev->dev, "failed to unmap legacy queue\n");
 
-clean_up:
-	amdgpu_bo_free_kernel(&q->mqd_obj,
-			      &q->mqd_gpu_addr,
-			      &q->mqd_cpu_ptr);
 	return r;
 }
 
-static void amdgpu_mes_queue_init_mqd(struct amdgpu_device *adev,
-				     struct amdgpu_mes_queue *q,
-				     struct amdgpu_mes_queue_properties *p)
-{
-	struct amdgpu_mqd *mqd_mgr = &adev->mqds[p->queue_type];
-	struct amdgpu_mqd_prop mqd_prop = {0};
-
-	mqd_prop.mqd_gpu_addr = q->mqd_gpu_addr;
-	mqd_prop.hqd_base_gpu_addr = p->hqd_base_gpu_addr;
-	mqd_prop.rptr_gpu_addr = p->rptr_gpu_addr;
-	mqd_prop.wptr_gpu_addr = p->wptr_gpu_addr;
-	mqd_prop.queue_size = p->queue_size;
-	mqd_prop.use_doorbell = true;
-	mqd_prop.doorbell_index = p->doorbell_off;
-	mqd_prop.eop_gpu_addr = p->eop_gpu_addr;
-	mqd_prop.hqd_pipe_priority = p->hqd_pipe_priority;
-	mqd_prop.hqd_queue_priority = p->hqd_queue_priority;
-	mqd_prop.hqd_active = false;
-
-	mqd_mgr->init_mqd(adev, q->mqd_cpu_ptr, &mqd_prop);
-
-	amdgpu_bo_unreserve(q->mqd_obj);
-}
-
-int amdgpu_mes_add_hw_queue(struct amdgpu_device *adev, int gang_id,
-			    struct amdgpu_mes_queue_properties *qprops,
-			    int *queue_id)
+int amdgpu_mes_reset_legacy_queue(struct amdgpu_device *adev,
+				  struct amdgpu_ring *ring,
+				  unsigned int vmid,
+				  bool use_mmio)
 {
-	struct amdgpu_mes_queue *queue;
-	struct amdgpu_mes_gang *gang;
-	struct mes_add_queue_input queue_input;
-	unsigned long flags;
+	struct mes_reset_queue_input queue_input;
 	int r;
 
-	/* allocate the mes queue buffer */
-	queue = kzalloc(sizeof(struct amdgpu_mes_queue), GFP_KERNEL);
-	if (!queue) {
-		DRM_ERROR("Failed to allocate memory for queue\n");
-		return -ENOMEM;
-	}
+	memset(&queue_input, 0, sizeof(queue_input));
 
-	/* Allocate the queue mqd */
-	r = amdgpu_mes_queue_alloc_mqd(adev, queue, qprops);
-	if (r)
-		goto clean_up_memory;
+	queue_input.queue_type = ring->funcs->type;
+	queue_input.doorbell_offset = ring->doorbell_index;
+	queue_input.me_id = ring->me;
+	queue_input.pipe_id = ring->pipe;
+	queue_input.queue_id = ring->queue;
+	queue_input.mqd_addr = ring->mqd_obj ? amdgpu_bo_gpu_offset(ring->mqd_obj) : 0;
+	queue_input.wptr_addr = ring->wptr_gpu_addr;
+	queue_input.vmid = vmid;
+	queue_input.use_mmio = use_mmio;
+	queue_input.is_kq = true;
+	if (ring->funcs->type == AMDGPU_RING_TYPE_GFX)
+		queue_input.legacy_gfx = true;
 
-	/*
-	 * Avoid taking any other locks under MES lock to avoid circular
-	 * lock dependencies.
-	 */
 	amdgpu_mes_lock(&adev->mes);
-
-	gang = idr_find(&adev->mes.gang_id_idr, gang_id);
-	if (!gang) {
-		DRM_ERROR("gang id %d doesn't exist\n", gang_id);
-		r = -EINVAL;
-		goto clean_up_mqd;
-	}
-
-	/* add the mes gang to idr list */
-	spin_lock_irqsave(&adev->mes.queue_id_lock, flags);
-	r = idr_alloc(&adev->mes.queue_id_idr, queue, 1, 0,
-		      GFP_ATOMIC);
-	if (r < 0) {
-		spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags);
-		goto clean_up_mqd;
-	}
-	spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags);
-	*queue_id = queue->queue_id = r;
-
-	/* allocate a doorbell index for the queue */
-	r = amdgpu_mes_queue_doorbell_get(adev, gang->process,
-					  qprops->queue_type,
-					  &qprops->doorbell_off);
-	if (r)
-		goto clean_up_queue_id;
-
-	/* initialize the queue mqd */
-	amdgpu_mes_queue_init_mqd(adev, queue, qprops);
-
-	/* add hw queue to mes */
-	queue_input.process_id = gang->process->pasid;
-
-	queue_input.page_table_base_addr =
-		adev->vm_manager.vram_base_offset + gang->process->pd_gpu_addr -
-		adev->gmc.vram_start;
-
-	queue_input.process_va_start = 0;
-	queue_input.process_va_end =
-		(adev->vm_manager.max_pfn - 1) << AMDGPU_GPU_PAGE_SHIFT;
-	queue_input.process_quantum = gang->process->process_quantum;
-	queue_input.process_context_addr = gang->process->proc_ctx_gpu_addr;
-	queue_input.gang_quantum = gang->gang_quantum;
-	queue_input.gang_context_addr = gang->gang_ctx_gpu_addr;
-	queue_input.inprocess_gang_priority = gang->inprocess_gang_priority;
-	queue_input.gang_global_priority_level = gang->global_priority_level;
-	queue_input.doorbell_offset = qprops->doorbell_off;
-	queue_input.mqd_addr = queue->mqd_gpu_addr;
-	queue_input.wptr_addr = qprops->wptr_gpu_addr;
-	queue_input.wptr_mc_addr = qprops->wptr_mc_addr;
-	queue_input.queue_type = qprops->queue_type;
-	queue_input.paging = qprops->paging;
-	queue_input.is_kfd_process = 0;
-
-	r = adev->mes.funcs->add_hw_queue(&adev->mes, &queue_input);
-	if (r) {
-		DRM_ERROR("failed to add hardware queue to MES, doorbell=0x%llx\n",
-			  qprops->doorbell_off);
-		goto clean_up_doorbell;
-	}
-
-	DRM_DEBUG("MES hw queue was added, pasid=%d, gang id=%d, "
-		  "queue type=%d, doorbell=0x%llx\n",
-		  gang->process->pasid, gang_id, qprops->queue_type,
-		  qprops->doorbell_off);
-
-	queue->ring = qprops->ring;
-	queue->doorbell_off = qprops->doorbell_off;
-	queue->wptr_gpu_addr = qprops->wptr_gpu_addr;
-	queue->queue_type = qprops->queue_type;
-	queue->paging = qprops->paging;
-	queue->gang = gang;
-	queue->ring->mqd_ptr = queue->mqd_cpu_ptr;
-	list_add_tail(&queue->list, &gang->queue_list);
-
+	r = adev->mes.funcs->reset_hw_queue(&adev->mes, &queue_input);
 	amdgpu_mes_unlock(&adev->mes);
-	return 0;
+	if (r)
+		dev_err(adev->dev, "failed to reset legacy queue\n");
 
-clean_up_doorbell:
-	amdgpu_mes_queue_doorbell_free(adev, gang->process,
-				       qprops->doorbell_off);
-clean_up_queue_id:
-	spin_lock_irqsave(&adev->mes.queue_id_lock, flags);
-	idr_remove(&adev->mes.queue_id_idr, queue->queue_id);
-	spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags);
-clean_up_mqd:
-	amdgpu_mes_unlock(&adev->mes);
-	amdgpu_mes_queue_free_mqd(queue);
-clean_up_memory:
-	kfree(queue);
 	return r;
 }
 
-int amdgpu_mes_remove_hw_queue(struct amdgpu_device *adev, int queue_id)
+int amdgpu_mes_get_hung_queue_db_array_size(struct amdgpu_device *adev)
 {
-	unsigned long flags;
-	struct amdgpu_mes_queue *queue;
-	struct amdgpu_mes_gang *gang;
-	struct mes_remove_queue_input queue_input;
-	int r;
+	return adev->mes.hung_queue_db_array_size;
+}
 
-	/*
-	 * Avoid taking any other locks under MES lock to avoid circular
-	 * lock dependencies.
-	 */
-	amdgpu_mes_lock(&adev->mes);
+int amdgpu_mes_detect_and_reset_hung_queues(struct amdgpu_device *adev,
+					    int queue_type,
+					    bool detect_only,
+					    unsigned int *hung_db_num,
+					    u32 *hung_db_array)
 
-	/* remove the mes gang from idr list */
-	spin_lock_irqsave(&adev->mes.queue_id_lock, flags);
+{
+	struct mes_detect_and_reset_queue_input input;
+	u32 *db_array = adev->mes.hung_queue_db_array_cpu_addr;
+	int r, i;
 
-	queue = idr_find(&adev->mes.queue_id_idr, queue_id);
-	if (!queue) {
-		spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags);
-		amdgpu_mes_unlock(&adev->mes);
-		DRM_ERROR("queue id %d doesn't exist\n", queue_id);
+	if (!hung_db_num || !hung_db_array)
 		return -EINVAL;
-	}
-
-	idr_remove(&adev->mes.queue_id_idr, queue_id);
-	spin_unlock_irqrestore(&adev->mes.queue_id_lock, flags);
-
-	DRM_DEBUG("try to remove queue, doorbell off = 0x%llx\n",
-		  queue->doorbell_off);
 
-	gang = queue->gang;
-	queue_input.doorbell_offset = queue->doorbell_off;
-	queue_input.gang_context_addr = gang->gang_ctx_gpu_addr;
-
-	r = adev->mes.funcs->remove_hw_queue(&adev->mes, &queue_input);
-	if (r)
-		DRM_ERROR("failed to remove hardware queue, queue id = %d\n",
-			  queue_id);
-
-	list_del(&queue->list);
-	amdgpu_mes_queue_doorbell_free(adev, gang->process,
-				       queue->doorbell_off);
-	amdgpu_mes_unlock(&adev->mes);
-
-	amdgpu_mes_queue_free_mqd(queue);
-	kfree(queue);
-	return 0;
-}
+	if ((queue_type != AMDGPU_RING_TYPE_GFX) &&
+	    (queue_type != AMDGPU_RING_TYPE_COMPUTE) &&
+	    (queue_type != AMDGPU_RING_TYPE_SDMA))
+		return -EINVAL;
 
-int amdgpu_mes_unmap_legacy_queue(struct amdgpu_device *adev,
-				  struct amdgpu_ring *ring,
-				  enum amdgpu_unmap_queues_action action,
-				  u64 gpu_addr, u64 seq)
-{
-	struct mes_unmap_legacy_queue_input queue_input;
-	int r;
+	/* Clear the doorbell array before detection */
+	memset(adev->mes.hung_queue_db_array_cpu_addr, AMDGPU_MES_INVALID_DB_OFFSET,
+		adev->mes.hung_queue_db_array_size * sizeof(u32));
+	input.queue_type = queue_type;
+	input.detect_only = detect_only;
 
-	queue_input.action = action;
-	queue_input.queue_type = ring->funcs->type;
-	queue_input.doorbell_offset = ring->doorbell_index;
-	queue_input.pipe_id = ring->pipe;
-	queue_input.queue_id = ring->queue;
-	queue_input.trail_fence_addr = gpu_addr;
-	queue_input.trail_fence_data = seq;
+	r = adev->mes.funcs->detect_and_reset_hung_queues(&adev->mes,
+							  &input);
+	if (r) {
+		dev_err(adev->dev, "failed to detect and reset\n");
+	} else {
+		*hung_db_num = 0;
+		for (i = 0; i < adev->mes.hung_queue_hqd_info_offset; i++) {
+			if (db_array[i] != AMDGPU_MES_INVALID_DB_OFFSET) {
+				hung_db_array[i] = db_array[i];
+				*hung_db_num += 1;
+			}
+		}
 
-	r = adev->mes.funcs->unmap_legacy_queue(&adev->mes, &queue_input);
-	if (r)
-		DRM_ERROR("failed to unmap legacy queue\n");
+		/*
+		 * TODO: return HQD info for MES scheduled user compute queue reset cases
+		 * stored in hung_db_array hqd info offset to full array size
+		 */
+	}
 
 	return r;
 }
@@ -828,23 +440,36 @@ uint32_t amdgpu_mes_rreg(struct amdgpu_device *adev, uint32_t reg)
 {
 	struct mes_misc_op_input op_input;
 	int r, val = 0;
+	uint32_t addr_offset = 0;
+	uint64_t read_val_gpu_addr;
+	uint32_t *read_val_ptr;
 
+	if (amdgpu_device_wb_get(adev, &addr_offset)) {
+		dev_err(adev->dev, "critical bug! too many mes readers\n");
+		goto error;
+	}
+	read_val_gpu_addr = adev->wb.gpu_addr + (addr_offset * 4);
+	read_val_ptr = (uint32_t *)&adev->wb.wb[addr_offset];
 	op_input.op = MES_MISC_OP_READ_REG;
 	op_input.read_reg.reg_offset = reg;
-	op_input.read_reg.buffer_addr = adev->mes.read_val_gpu_addr;
+	op_input.read_reg.buffer_addr = read_val_gpu_addr;
 
 	if (!adev->mes.funcs->misc_op) {
-		DRM_ERROR("mes rreg is not supported!\n");
+		dev_err(adev->dev, "mes rreg is not supported!\n");
 		goto error;
 	}
 
+	amdgpu_mes_lock(&adev->mes);
 	r = adev->mes.funcs->misc_op(&adev->mes, &op_input);
+	amdgpu_mes_unlock(&adev->mes);
 	if (r)
-		DRM_ERROR("failed to read reg (0x%x)\n", reg);
+		dev_err(adev->dev, "failed to read reg (0x%x)\n", reg);
 	else
-		val = *(adev->mes.read_val_ptr);
+		val = *(read_val_ptr);
 
 error:
+	if (addr_offset)
+		amdgpu_device_wb_free(adev, addr_offset);
 	return val;
 }
 
@@ -859,14 +484,16 @@ int amdgpu_mes_wreg(struct amdgpu_device *adev,
 	op_input.write_reg.reg_value = val;
 
 	if (!adev->mes.funcs->misc_op) {
-		DRM_ERROR("mes wreg is not supported!\n");
+		dev_err(adev->dev, "mes wreg is not supported!\n");
 		r = -EINVAL;
 		goto error;
 	}
 
+	amdgpu_mes_lock(&adev->mes);
 	r = adev->mes.funcs->misc_op(&adev->mes, &op_input);
+	amdgpu_mes_unlock(&adev->mes);
 	if (r)
-		DRM_ERROR("failed to write reg (0x%x)\n", reg);
+		dev_err(adev->dev, "failed to write reg (0x%x)\n", reg);
 
 error:
 	return r;
@@ -886,546 +513,107 @@ int amdgpu_mes_reg_write_reg_wait(struct amdgpu_device *adev,
 	op_input.wrm_reg.mask = mask;
 
 	if (!adev->mes.funcs->misc_op) {
-		DRM_ERROR("mes reg_write_reg_wait is not supported!\n");
+		dev_err(adev->dev, "mes reg_write_reg_wait is not supported!\n");
 		r = -EINVAL;
 		goto error;
 	}
 
+	amdgpu_mes_lock(&adev->mes);
 	r = adev->mes.funcs->misc_op(&adev->mes, &op_input);
+	amdgpu_mes_unlock(&adev->mes);
 	if (r)
-		DRM_ERROR("failed to reg_write_reg_wait\n");
+		dev_err(adev->dev, "failed to reg_write_reg_wait\n");
 
 error:
 	return r;
 }
 
-int amdgpu_mes_reg_wait(struct amdgpu_device *adev, uint32_t reg,
-			uint32_t val, uint32_t mask)
+int amdgpu_mes_hdp_flush(struct amdgpu_device *adev)
 {
-	struct mes_misc_op_input op_input;
-	int r;
-
-	op_input.op = MES_MISC_OP_WRM_REG_WAIT;
-	op_input.wrm_reg.reg0 = reg;
-	op_input.wrm_reg.ref = val;
-	op_input.wrm_reg.mask = mask;
-
-	if (!adev->mes.funcs->misc_op) {
-		DRM_ERROR("mes reg wait is not supported!\n");
-		r = -EINVAL;
-		goto error;
-	}
+	uint32_t hdp_flush_req_offset, hdp_flush_done_offset, ref_and_mask;
 
-	r = adev->mes.funcs->misc_op(&adev->mes, &op_input);
-	if (r)
-		DRM_ERROR("failed to reg_write_reg_wait\n");
+	hdp_flush_req_offset = adev->nbio.funcs->get_hdp_flush_req_offset(adev);
+	hdp_flush_done_offset = adev->nbio.funcs->get_hdp_flush_done_offset(adev);
+	ref_and_mask = adev->nbio.hdp_flush_reg->ref_and_mask_cp0;
 
-error:
-	return r;
+	return amdgpu_mes_reg_write_reg_wait(adev, hdp_flush_req_offset, hdp_flush_done_offset,
+					     ref_and_mask, ref_and_mask);
 }
 
-static void
-amdgpu_mes_ring_to_queue_props(struct amdgpu_device *adev,
-			       struct amdgpu_ring *ring,
-			       struct amdgpu_mes_queue_properties *props)
+int amdgpu_mes_set_shader_debugger(struct amdgpu_device *adev,
+				uint64_t process_context_addr,
+				uint32_t spi_gdbg_per_vmid_cntl,
+				const uint32_t *tcp_watch_cntl,
+				uint32_t flags,
+				bool trap_en)
 {
-	props->queue_type = ring->funcs->type;
-	props->hqd_base_gpu_addr = ring->gpu_addr;
-	props->rptr_gpu_addr = ring->rptr_gpu_addr;
-	props->wptr_gpu_addr = ring->wptr_gpu_addr;
-	props->wptr_mc_addr =
-		ring->mes_ctx->meta_data_mc_addr + ring->wptr_offs;
-	props->queue_size = ring->ring_size;
-	props->eop_gpu_addr = ring->eop_gpu_addr;
-	props->hqd_pipe_priority = AMDGPU_GFX_PIPE_PRIO_NORMAL;
-	props->hqd_queue_priority = AMDGPU_GFX_QUEUE_PRIORITY_MINIMUM;
-	props->paging = false;
-	props->ring = ring;
-}
+	struct mes_misc_op_input op_input = {0};
+	int r;
 
-#define DEFINE_AMDGPU_MES_CTX_GET_OFFS_ENG(_eng)			\
-do {									\
-       if (id_offs < AMDGPU_MES_CTX_MAX_OFFS)				\
-		return offsetof(struct amdgpu_mes_ctx_meta_data,	\
-				_eng[ring->idx].slots[id_offs]);        \
-       else if (id_offs == AMDGPU_MES_CTX_RING_OFFS)			\
-		return offsetof(struct amdgpu_mes_ctx_meta_data,        \
-				_eng[ring->idx].ring);                  \
-       else if (id_offs == AMDGPU_MES_CTX_IB_OFFS)			\
-		return offsetof(struct amdgpu_mes_ctx_meta_data,        \
-				_eng[ring->idx].ib);                    \
-       else if (id_offs == AMDGPU_MES_CTX_PADDING_OFFS)			\
-		return offsetof(struct amdgpu_mes_ctx_meta_data,        \
-				_eng[ring->idx].padding);               \
-} while(0)
-
-int amdgpu_mes_ctx_get_offs(struct amdgpu_ring *ring, unsigned int id_offs)
-{
-	switch (ring->funcs->type) {
-	case AMDGPU_RING_TYPE_GFX:
-		DEFINE_AMDGPU_MES_CTX_GET_OFFS_ENG(gfx);
-		break;
-	case AMDGPU_RING_TYPE_COMPUTE:
-		DEFINE_AMDGPU_MES_CTX_GET_OFFS_ENG(compute);
-		break;
-	case AMDGPU_RING_TYPE_SDMA:
-		DEFINE_AMDGPU_MES_CTX_GET_OFFS_ENG(sdma);
-		break;
-	default:
-		break;
+	if (!adev->mes.funcs->misc_op) {
+		dev_err(adev->dev,
+			"mes set shader debugger is not supported!\n");
+		return -EINVAL;
 	}
 
-	WARN_ON(1);
-	return -EINVAL;
-}
-
-int amdgpu_mes_add_ring(struct amdgpu_device *adev, int gang_id,
-			int queue_type, int idx,
-			struct amdgpu_mes_ctx_data *ctx_data,
-			struct amdgpu_ring **out)
-{
-	struct amdgpu_ring *ring;
-	struct amdgpu_mes_gang *gang;
-	struct amdgpu_mes_queue_properties qprops = {0};
-	int r, queue_id, pasid;
+	op_input.op = MES_MISC_OP_SET_SHADER_DEBUGGER;
+	op_input.set_shader_debugger.process_context_addr = process_context_addr;
+	op_input.set_shader_debugger.flags.u32all = flags;
 
-	/*
-	 * Avoid taking any other locks under MES lock to avoid circular
-	 * lock dependencies.
-	 */
-	amdgpu_mes_lock(&adev->mes);
-	gang = idr_find(&adev->mes.gang_id_idr, gang_id);
-	if (!gang) {
-		DRM_ERROR("gang id %d doesn't exist\n", gang_id);
-		amdgpu_mes_unlock(&adev->mes);
+	/* use amdgpu mes_flush_shader_debugger instead */
+	if (op_input.set_shader_debugger.flags.process_ctx_flush)
 		return -EINVAL;
-	}
-	pasid = gang->process->pasid;
 
-	ring = kzalloc(sizeof(struct amdgpu_ring), GFP_KERNEL);
-	if (!ring) {
-		amdgpu_mes_unlock(&adev->mes);
-		return -ENOMEM;
-	}
+	op_input.set_shader_debugger.spi_gdbg_per_vmid_cntl = spi_gdbg_per_vmid_cntl;
+	memcpy(op_input.set_shader_debugger.tcp_watch_cntl, tcp_watch_cntl,
+			sizeof(op_input.set_shader_debugger.tcp_watch_cntl));
 
-	ring->ring_obj = NULL;
-	ring->use_doorbell = true;
-	ring->is_mes_queue = true;
-	ring->mes_ctx = ctx_data;
-	ring->idx = idx;
-	ring->no_scheduler = true;
-
-	if (queue_type == AMDGPU_RING_TYPE_COMPUTE) {
-		int offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				      compute[ring->idx].mec_hpd);
-		ring->eop_gpu_addr =
-			amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-	}
+	if (((adev->mes.sched_version & AMDGPU_MES_API_VERSION_MASK) >>
+			AMDGPU_MES_API_VERSION_SHIFT) >= 14)
+		op_input.set_shader_debugger.trap_en = trap_en;
 
-	switch (queue_type) {
-	case AMDGPU_RING_TYPE_GFX:
-		ring->funcs = adev->gfx.gfx_ring[0].funcs;
-		break;
-	case AMDGPU_RING_TYPE_COMPUTE:
-		ring->funcs = adev->gfx.compute_ring[0].funcs;
-		break;
-	case AMDGPU_RING_TYPE_SDMA:
-		ring->funcs = adev->sdma.instance[0].ring.funcs;
-		break;
-	default:
-		BUG();
-	}
+	amdgpu_mes_lock(&adev->mes);
 
-	r = amdgpu_ring_init(adev, ring, 1024, NULL, 0,
-			     AMDGPU_RING_PRIO_DEFAULT, NULL);
+	r = adev->mes.funcs->misc_op(&adev->mes, &op_input);
 	if (r)
-		goto clean_up_memory;
+		dev_err(adev->dev, "failed to set_shader_debugger\n");
 
-	amdgpu_mes_ring_to_queue_props(adev, ring, &qprops);
-
-	dma_fence_wait(gang->process->vm->last_update, false);
-	dma_fence_wait(ctx_data->meta_data_va->last_pt_update, false);
 	amdgpu_mes_unlock(&adev->mes);
 
-	r = amdgpu_mes_add_hw_queue(adev, gang_id, &qprops, &queue_id);
-	if (r)
-		goto clean_up_ring;
-
-	ring->hw_queue_id = queue_id;
-	ring->doorbell_index = qprops.doorbell_off;
-
-	if (queue_type == AMDGPU_RING_TYPE_GFX)
-		sprintf(ring->name, "gfx_%d.%d.%d", pasid, gang_id, queue_id);
-	else if (queue_type == AMDGPU_RING_TYPE_COMPUTE)
-		sprintf(ring->name, "compute_%d.%d.%d", pasid, gang_id,
-			queue_id);
-	else if (queue_type == AMDGPU_RING_TYPE_SDMA)
-		sprintf(ring->name, "sdma_%d.%d.%d", pasid, gang_id,
-			queue_id);
-	else
-		BUG();
-
-	*out = ring;
-	return 0;
-
-clean_up_ring:
-	amdgpu_ring_fini(ring);
-clean_up_memory:
-	kfree(ring);
-	amdgpu_mes_unlock(&adev->mes);
 	return r;
 }
 
-void amdgpu_mes_remove_ring(struct amdgpu_device *adev,
-			    struct amdgpu_ring *ring)
-{
-	if (!ring)
-		return;
-
-	amdgpu_mes_remove_hw_queue(adev, ring->hw_queue_id);
-	amdgpu_ring_fini(ring);
-	kfree(ring);
-}
-
-uint32_t amdgpu_mes_get_aggregated_doorbell_index(struct amdgpu_device *adev,
-						   enum amdgpu_mes_priority_level prio)
-{
-	return adev->mes.aggregated_doorbells[prio];
-}
-
-int amdgpu_mes_ctx_alloc_meta_data(struct amdgpu_device *adev,
-				   struct amdgpu_mes_ctx_data *ctx_data)
-{
-	int r;
-
-	r = amdgpu_bo_create_kernel(adev,
-			    sizeof(struct amdgpu_mes_ctx_meta_data),
-			    PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT,
-			    &ctx_data->meta_data_obj,
-			    &ctx_data->meta_data_mc_addr,
-			    &ctx_data->meta_data_ptr);
-	if (r) {
-		dev_warn(adev->dev, "(%d) create CTX bo failed\n", r);
-		return r;
-	}
-
-	if (!ctx_data->meta_data_obj)
-		return -ENOMEM;
-
-	memset(ctx_data->meta_data_ptr, 0,
-	       sizeof(struct amdgpu_mes_ctx_meta_data));
-
-	return 0;
-}
-
-void amdgpu_mes_ctx_free_meta_data(struct amdgpu_mes_ctx_data *ctx_data)
-{
-	if (ctx_data->meta_data_obj)
-		amdgpu_bo_free_kernel(&ctx_data->meta_data_obj,
-				      &ctx_data->meta_data_mc_addr,
-				      &ctx_data->meta_data_ptr);
-}
-
-int amdgpu_mes_ctx_map_meta_data(struct amdgpu_device *adev,
-				 struct amdgpu_vm *vm,
-				 struct amdgpu_mes_ctx_data *ctx_data)
+int amdgpu_mes_flush_shader_debugger(struct amdgpu_device *adev,
+				     uint64_t process_context_addr)
 {
-	struct amdgpu_bo_va *bo_va;
-	struct ww_acquire_ctx ticket;
-	struct list_head list;
-	struct amdgpu_bo_list_entry pd;
-	struct ttm_validate_buffer csa_tv;
-	struct amdgpu_sync sync;
+	struct mes_misc_op_input op_input = {0};
 	int r;
 
-	amdgpu_sync_create(&sync);
-	INIT_LIST_HEAD(&list);
-	INIT_LIST_HEAD(&csa_tv.head);
-
-	csa_tv.bo = &ctx_data->meta_data_obj->tbo;
-	csa_tv.num_shared = 1;
-
-	list_add(&csa_tv.head, &list);
-	amdgpu_vm_get_pd_bo(vm, &list, &pd);
-
-	r = ttm_eu_reserve_buffers(&ticket, &list, true, NULL);
-	if (r) {
-		DRM_ERROR("failed to reserve meta data BO: err=%d\n", r);
-		return r;
-	}
-
-	bo_va = amdgpu_vm_bo_add(adev, vm, ctx_data->meta_data_obj);
-	if (!bo_va) {
-		ttm_eu_backoff_reservation(&ticket, &list);
-		DRM_ERROR("failed to create bo_va for meta data BO\n");
-		return -ENOMEM;
-	}
-
-	r = amdgpu_vm_bo_map(adev, bo_va, ctx_data->meta_data_gpu_addr, 0,
-			     sizeof(struct amdgpu_mes_ctx_meta_data),
-			     AMDGPU_PTE_READABLE | AMDGPU_PTE_WRITEABLE |
-			     AMDGPU_PTE_EXECUTABLE);
-
-	if (r) {
-		DRM_ERROR("failed to do bo_map on meta data, err=%d\n", r);
-		goto error;
-	}
-
-	r = amdgpu_vm_bo_update(adev, bo_va, false);
-	if (r) {
-		DRM_ERROR("failed to do vm_bo_update on meta data\n");
-		goto error;
-	}
-	amdgpu_sync_fence(&sync, bo_va->last_pt_update);
-
-	r = amdgpu_vm_update_pdes(adev, vm, false);
-	if (r) {
-		DRM_ERROR("failed to update pdes on meta data\n");
-		goto error;
+	if (!adev->mes.funcs->misc_op) {
+		dev_err(adev->dev,
+			"mes flush shader debugger is not supported!\n");
+		return -EINVAL;
 	}
-	amdgpu_sync_fence(&sync, vm->last_update);
-
-	amdgpu_sync_wait(&sync, false);
-	ttm_eu_backoff_reservation(&ticket, &list);
-
-	amdgpu_sync_free(&sync);
-	ctx_data->meta_data_va = bo_va;
-	return 0;
-
-error:
-	amdgpu_vm_bo_del(adev, bo_va);
-	ttm_eu_backoff_reservation(&ticket, &list);
-	amdgpu_sync_free(&sync);
-	return r;
-}
 
-int amdgpu_mes_ctx_unmap_meta_data(struct amdgpu_device *adev,
-				   struct amdgpu_mes_ctx_data *ctx_data)
-{
-	struct amdgpu_bo_va *bo_va = ctx_data->meta_data_va;
-	struct amdgpu_bo *bo = ctx_data->meta_data_obj;
-	struct amdgpu_vm *vm = bo_va->base.vm;
-	struct amdgpu_bo_list_entry vm_pd;
-	struct list_head list, duplicates;
-	struct dma_fence *fence = NULL;
-	struct ttm_validate_buffer tv;
-	struct ww_acquire_ctx ticket;
-	long r = 0;
-
-	INIT_LIST_HEAD(&list);
-	INIT_LIST_HEAD(&duplicates);
-
-	tv.bo = &bo->tbo;
-	tv.num_shared = 2;
-	list_add(&tv.head, &list);
-
-	amdgpu_vm_get_pd_bo(vm, &list, &vm_pd);
-
-	r = ttm_eu_reserve_buffers(&ticket, &list, false, &duplicates);
-	if (r) {
-		dev_err(adev->dev, "leaking bo va because "
-			"we fail to reserve bo (%ld)\n", r);
-		return r;
-	}
+	op_input.op = MES_MISC_OP_SET_SHADER_DEBUGGER;
+	op_input.set_shader_debugger.process_context_addr = process_context_addr;
+	op_input.set_shader_debugger.flags.process_ctx_flush = true;
 
-	amdgpu_vm_bo_del(adev, bo_va);
-	if (!amdgpu_vm_ready(vm))
-		goto out_unlock;
+	amdgpu_mes_lock(&adev->mes);
 
-	r = dma_resv_get_singleton(bo->tbo.base.resv, DMA_RESV_USAGE_BOOKKEEP, &fence);
+	r = adev->mes.funcs->misc_op(&adev->mes, &op_input);
 	if (r)
-		goto out_unlock;
-	if (fence) {
-		amdgpu_bo_fence(bo, fence, true);
-		fence = NULL;
-	}
-
-	r = amdgpu_vm_clear_freed(adev, vm, &fence);
-	if (r || !fence)
-		goto out_unlock;
+		dev_err(adev->dev, "failed to set_shader_debugger\n");
 
-	dma_fence_wait(fence, false);
-	amdgpu_bo_fence(bo, fence, true);
-	dma_fence_put(fence);
-
-out_unlock:
-	if (unlikely(r < 0))
-		dev_err(adev->dev, "failed to clear page tables (%ld)\n", r);
-	ttm_eu_backoff_reservation(&ticket, &list);
+	amdgpu_mes_unlock(&adev->mes);
 
 	return r;
 }
 
-static int amdgpu_mes_test_create_gang_and_queues(struct amdgpu_device *adev,
-					  int pasid, int *gang_id,
-					  int queue_type, int num_queue,
-					  struct amdgpu_ring **added_rings,
-					  struct amdgpu_mes_ctx_data *ctx_data)
-{
-	struct amdgpu_ring *ring;
-	struct amdgpu_mes_gang_properties gprops = {0};
-	int r, j;
-
-	/* create a gang for the process */
-	gprops.priority = AMDGPU_MES_PRIORITY_LEVEL_NORMAL;
-	gprops.gang_quantum = adev->mes.default_gang_quantum;
-	gprops.inprocess_gang_priority = AMDGPU_MES_PRIORITY_LEVEL_NORMAL;
-	gprops.priority_level = AMDGPU_MES_PRIORITY_LEVEL_NORMAL;
-	gprops.global_priority_level = AMDGPU_MES_PRIORITY_LEVEL_NORMAL;
-
-	r = amdgpu_mes_add_gang(adev, pasid, &gprops, gang_id);
-	if (r) {
-		DRM_ERROR("failed to add gang\n");
-		return r;
-	}
-
-	/* create queues for the gang */
-	for (j = 0; j < num_queue; j++) {
-		r = amdgpu_mes_add_ring(adev, *gang_id, queue_type, j,
-					ctx_data, &ring);
-		if (r) {
-			DRM_ERROR("failed to add ring\n");
-			break;
-		}
-
-		DRM_INFO("ring %s was added\n", ring->name);
-		added_rings[j] = ring;
-	}
-
-	return 0;
-}
-
-static int amdgpu_mes_test_queues(struct amdgpu_ring **added_rings)
-{
-	struct amdgpu_ring *ring;
-	int i, r;
-
-	for (i = 0; i < AMDGPU_MES_CTX_MAX_RINGS; i++) {
-		ring = added_rings[i];
-		if (!ring)
-			continue;
-
-		r = amdgpu_ring_test_ring(ring);
-		if (r) {
-			DRM_DEV_ERROR(ring->adev->dev,
-				      "ring %s test failed (%d)\n",
-				      ring->name, r);
-			return r;
-		} else
-			DRM_INFO("ring %s test pass\n", ring->name);
-
-		r = amdgpu_ring_test_ib(ring, 1000 * 10);
-		if (r) {
-			DRM_DEV_ERROR(ring->adev->dev,
-				      "ring %s ib test failed (%d)\n",
-				      ring->name, r);
-			return r;
-		} else
-			DRM_INFO("ring %s ib test pass\n", ring->name);
-	}
-
-	return 0;
-}
-
-int amdgpu_mes_self_test(struct amdgpu_device *adev)
+uint32_t amdgpu_mes_get_aggregated_doorbell_index(struct amdgpu_device *adev,
+						   enum amdgpu_mes_priority_level prio)
 {
-	struct amdgpu_vm *vm = NULL;
-	struct amdgpu_mes_ctx_data ctx_data = {0};
-	struct amdgpu_ring *added_rings[AMDGPU_MES_CTX_MAX_RINGS] = { NULL };
-	int gang_ids[3] = {0};
-	int queue_types[][2] = { { AMDGPU_RING_TYPE_GFX, 1 },
-				 { AMDGPU_RING_TYPE_COMPUTE, 1 },
-				 { AMDGPU_RING_TYPE_SDMA, 1} };
-	int i, r, pasid, k = 0;
-
-	pasid = amdgpu_pasid_alloc(16);
-	if (pasid < 0) {
-		dev_warn(adev->dev, "No more PASIDs available!");
-		pasid = 0;
-	}
-
-	vm = kzalloc(sizeof(*vm), GFP_KERNEL);
-	if (!vm) {
-		r = -ENOMEM;
-		goto error_pasid;
-	}
-
-	r = amdgpu_vm_init(adev, vm);
-	if (r) {
-		DRM_ERROR("failed to initialize vm\n");
-		goto error_pasid;
-	}
-
-	r = amdgpu_mes_ctx_alloc_meta_data(adev, &ctx_data);
-	if (r) {
-		DRM_ERROR("failed to alloc ctx meta data\n");
-		goto error_fini;
-	}
-
-	ctx_data.meta_data_gpu_addr = AMDGPU_VA_RESERVED_SIZE;
-	r = amdgpu_mes_ctx_map_meta_data(adev, vm, &ctx_data);
-	if (r) {
-		DRM_ERROR("failed to map ctx meta data\n");
-		goto error_vm;
-	}
-
-	r = amdgpu_mes_create_process(adev, pasid, vm);
-	if (r) {
-		DRM_ERROR("failed to create MES process\n");
-		goto error_vm;
-	}
-
-	for (i = 0; i < ARRAY_SIZE(queue_types); i++) {
-		/* On GFX v10.3, fw hasn't supported to map sdma queue. */
-		if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0) &&
-		    adev->ip_versions[GC_HWIP][0] < IP_VERSION(11, 0, 0) &&
-		    queue_types[i][0] == AMDGPU_RING_TYPE_SDMA)
-			continue;
-
-		r = amdgpu_mes_test_create_gang_and_queues(adev, pasid,
-							   &gang_ids[i],
-							   queue_types[i][0],
-							   queue_types[i][1],
-							   &added_rings[k],
-							   &ctx_data);
-		if (r)
-			goto error_queues;
-
-		k += queue_types[i][1];
-	}
-
-	/* start ring test and ib test for MES queues */
-	amdgpu_mes_test_queues(added_rings);
-
-error_queues:
-	/* remove all queues */
-	for (i = 0; i < ARRAY_SIZE(added_rings); i++) {
-		if (!added_rings[i])
-			continue;
-		amdgpu_mes_remove_ring(adev, added_rings[i]);
-	}
-
-	for (i = 0; i < ARRAY_SIZE(gang_ids); i++) {
-		if (!gang_ids[i])
-			continue;
-		amdgpu_mes_remove_gang(adev, gang_ids[i]);
-	}
-
-	amdgpu_mes_destroy_process(adev, pasid);
-
-error_vm:
-	amdgpu_mes_ctx_unmap_meta_data(adev, &ctx_data);
-
-error_fini:
-	amdgpu_vm_fini(adev, vm);
-
-error_pasid:
-	if (pasid)
-		amdgpu_pasid_free(pasid);
-
-	amdgpu_mes_ctx_free_meta_data(&ctx_data);
-	kfree(vm);
-	return 0;
+	return adev->mes.aggregated_doorbells[prio];
 }
 
 int amdgpu_mes_init_microcode(struct amdgpu_device *adev, int pipe)
@@ -1433,14 +621,37 @@ int amdgpu_mes_init_microcode(struct amdgpu_device *adev, int pipe)
 	const struct mes_firmware_header_v1_0 *mes_hdr;
 	struct amdgpu_firmware_info *info;
 	char ucode_prefix[30];
-	char fw_name[40];
+	char fw_name[50];
+	bool need_retry = false;
+	u32 *ucode_ptr;
 	int r;
 
-	amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix));
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mes%s.bin",
-		ucode_prefix,
-		pipe == AMDGPU_MES_SCHED_PIPE ? "" : "1");
-	r = amdgpu_ucode_request(adev, &adev->mes.fw[pipe], fw_name);
+	amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix,
+				       sizeof(ucode_prefix));
+	if (adev->enable_uni_mes) {
+		snprintf(fw_name, sizeof(fw_name),
+			 "amdgpu/%s_uni_mes.bin", ucode_prefix);
+	} else if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(11, 0, 0) &&
+	    amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(12, 0, 0)) {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mes%s.bin",
+			 ucode_prefix,
+			 pipe == AMDGPU_MES_SCHED_PIPE ? "_2" : "1");
+		need_retry = true;
+	} else {
+		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mes%s.bin",
+			 ucode_prefix,
+			 pipe == AMDGPU_MES_SCHED_PIPE ? "" : "1");
+	}
+
+	r = amdgpu_ucode_request(adev, &adev->mes.fw[pipe], AMDGPU_UCODE_REQUIRED,
+				 "%s", fw_name);
+	if (r && need_retry && pipe == AMDGPU_MES_SCHED_PIPE) {
+		dev_info(adev->dev, "try to fall back to %s_mes.bin\n", ucode_prefix);
+		r = amdgpu_ucode_request(adev, &adev->mes.fw[pipe],
+					 AMDGPU_UCODE_REQUIRED,
+					 "amdgpu/%s_mes.bin", ucode_prefix);
+	}
+
 	if (r)
 		goto out;
 
@@ -1452,6 +663,10 @@ int amdgpu_mes_init_microcode(struct amdgpu_device *adev, int pipe)
 	adev->mes.data_start_addr[pipe] =
 		le32_to_cpu(mes_hdr->mes_data_start_addr_lo) |
 		((uint64_t)(le32_to_cpu(mes_hdr->mes_data_start_addr_hi)) << 32);
+	ucode_ptr = (u32 *)(adev->mes.fw[pipe]->data +
+			  sizeof(union amdgpu_firmware_header));
+	adev->mes.fw_version[pipe] =
+		le32_to_cpu(ucode_ptr[24]) & AMDGPU_MES_VERSION_MASK;
 
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 		int ucode, ucode_data;
@@ -1484,3 +699,86 @@ out:
 	amdgpu_ucode_release(&adev->mes.fw[pipe]);
 	return r;
 }
+
+bool amdgpu_mes_suspend_resume_all_supported(struct amdgpu_device *adev)
+{
+	uint32_t mes_rev = adev->mes.sched_version & AMDGPU_MES_VERSION_MASK;
+
+	return ((amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(11, 0, 0) &&
+		 amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(12, 0, 0) &&
+		 mes_rev >= 0x63) ||
+		amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(12, 0, 0));
+}
+
+/* Fix me -- node_id is used to identify the correct MES instances in the future */
+static int amdgpu_mes_set_enforce_isolation(struct amdgpu_device *adev,
+					    uint32_t node_id, bool enable)
+{
+	struct mes_misc_op_input op_input = {0};
+	int r;
+
+	op_input.op = MES_MISC_OP_CHANGE_CONFIG;
+	op_input.change_config.option.limit_single_process = enable ? 1 : 0;
+
+	if (!adev->mes.funcs->misc_op) {
+		dev_err(adev->dev, "mes change config is not supported!\n");
+		r = -EINVAL;
+		goto error;
+	}
+
+	amdgpu_mes_lock(&adev->mes);
+	r = adev->mes.funcs->misc_op(&adev->mes, &op_input);
+	amdgpu_mes_unlock(&adev->mes);
+	if (r)
+		dev_err(adev->dev, "failed to change_config.\n");
+
+error:
+	return r;
+}
+
+int amdgpu_mes_update_enforce_isolation(struct amdgpu_device *adev)
+{
+	int i, r = 0;
+
+	if (adev->enable_mes && adev->gfx.enable_cleaner_shader) {
+		mutex_lock(&adev->enforce_isolation_mutex);
+		for (i = 0; i < (adev->xcp_mgr ? adev->xcp_mgr->num_xcps : 1); i++) {
+			if (adev->enforce_isolation[i] == AMDGPU_ENFORCE_ISOLATION_ENABLE)
+				r |= amdgpu_mes_set_enforce_isolation(adev, i, true);
+			else
+				r |= amdgpu_mes_set_enforce_isolation(adev, i, false);
+		}
+		mutex_unlock(&adev->enforce_isolation_mutex);
+	}
+	return r;
+}
+
+#if defined(CONFIG_DEBUG_FS)
+
+static int amdgpu_debugfs_mes_event_log_show(struct seq_file *m, void *unused)
+{
+	struct amdgpu_device *adev = m->private;
+	uint32_t *mem = (uint32_t *)(adev->mes.event_log_cpu_addr);
+
+	seq_hex_dump(m, "", DUMP_PREFIX_OFFSET, 32, 4,
+		     mem, adev->mes.event_log_size, false);
+
+	return 0;
+}
+
+DEFINE_SHOW_ATTRIBUTE(amdgpu_debugfs_mes_event_log);
+
+#endif
+
+void amdgpu_debugfs_mes_event_log_init(struct amdgpu_device *adev)
+{
+
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+	if (adev->enable_mes && amdgpu_mes_log_enable)
+		debugfs_create_file("amdgpu_mes_event_log", 0444, root,
+				    adev, &amdgpu_debugfs_mes_event_log_fops);
+
+#endif
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
index 547ec35691fa..e989225b354b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
@@ -27,6 +27,7 @@
 #include "amdgpu_irq.h"
 #include "kgd_kfd_interface.h"
 #include "amdgpu_gfx.h"
+#include "amdgpu_doorbell.h"
 #include <linux/sched/mm.h>
 
 #define AMDGPU_MES_MAX_COMPUTE_PIPES        8
@@ -39,6 +40,8 @@
 #define AMDGPU_MES_VERSION_MASK		0x00000fff
 #define AMDGPU_MES_API_VERSION_MASK	0x00fff000
 #define AMDGPU_MES_FEAT_VERSION_MASK	0xff000000
+#define AMDGPU_MES_MSCRATCH_SIZE	0x40000
+#define AMDGPU_MES_INVALID_DB_OFFSET	0xffffffff
 
 enum amdgpu_mes_priority_level {
 	AMDGPU_MES_PRIORITY_LEVEL_LOW       = 0,
@@ -54,7 +57,7 @@ enum amdgpu_mes_priority_level {
 
 struct amdgpu_mes_funcs;
 
-enum admgpu_mes_pipe {
+enum amdgpu_mes_pipe {
 	AMDGPU_MES_SCHED_PIPE = 0,
 	AMDGPU_MES_KIQ_PIPE,
 	AMDGPU_MAX_MES_PIPES = 2,
@@ -74,16 +77,17 @@ struct amdgpu_mes {
 
 	uint32_t			sched_version;
 	uint32_t			kiq_version;
+	uint32_t			fw_version[AMDGPU_MAX_MES_PIPES];
+	bool                            enable_legacy_queue_map;
 
 	uint32_t                        total_max_queue;
-	uint32_t                        doorbell_id_offset;
 	uint32_t                        max_doorbell_slices;
 
 	uint64_t                        default_process_quantum;
 	uint64_t                        default_gang_quantum;
 
-	struct amdgpu_ring              ring;
-	spinlock_t                      ring_lock;
+	struct amdgpu_ring              ring[AMDGPU_MAX_MES_PIPES];
+	spinlock_t                      ring_lock[AMDGPU_MAX_MES_PIPES];
 
 	const struct firmware           *fw[AMDGPU_MAX_MES_PIPES];
 
@@ -108,19 +112,16 @@ struct amdgpu_mes {
 
 	uint32_t                        vmid_mask_gfxhub;
 	uint32_t                        vmid_mask_mmhub;
-	uint32_t                        compute_hqd_mask[AMDGPU_MES_MAX_COMPUTE_PIPES];
 	uint32_t                        gfx_hqd_mask[AMDGPU_MES_MAX_GFX_PIPES];
+	uint32_t                        compute_hqd_mask[AMDGPU_MES_MAX_COMPUTE_PIPES];
 	uint32_t                        sdma_hqd_mask[AMDGPU_MES_MAX_SDMA_PIPES];
 	uint32_t                        aggregated_doorbells[AMDGPU_MES_PRIORITY_NUM_LEVELS];
-	uint32_t                        sch_ctx_offs;
-	uint64_t			sch_ctx_gpu_addr;
-	uint64_t			*sch_ctx_ptr;
-	uint32_t			query_status_fence_offs;
-	uint64_t			query_status_fence_gpu_addr;
-	uint64_t			*query_status_fence_ptr;
-	uint32_t                        read_val_offs;
-	uint64_t			read_val_gpu_addr;
-	uint32_t			*read_val_ptr;
+	uint32_t                        sch_ctx_offs[AMDGPU_MAX_MES_PIPES];
+	uint64_t			sch_ctx_gpu_addr[AMDGPU_MAX_MES_PIPES];
+	uint64_t			*sch_ctx_ptr[AMDGPU_MAX_MES_PIPES];
+	uint32_t			query_status_fence_offs[AMDGPU_MAX_MES_PIPES];
+	uint64_t			query_status_fence_gpu_addr[AMDGPU_MAX_MES_PIPES];
+	uint64_t			*query_status_fence_ptr[AMDGPU_MAX_MES_PIPES];
 
 	uint32_t			saved_flags;
 
@@ -128,22 +129,30 @@ struct amdgpu_mes {
 	int                             (*kiq_hw_init)(struct amdgpu_device *adev);
 	int                             (*kiq_hw_fini)(struct amdgpu_device *adev);
 
+	/* MES doorbells */
+	uint32_t			db_start_dw_offset;
+	uint32_t			num_mes_dbs;
+	unsigned long			*doorbell_bitmap;
+
+	/* MES event log buffer */
+	uint32_t			event_log_size;
+	struct amdgpu_bo	*event_log_gpu_obj;
+	uint64_t			event_log_gpu_addr;
+	void				*event_log_cpu_addr;
+
 	/* ip specific functions */
 	const struct amdgpu_mes_funcs   *funcs;
-};
 
-struct amdgpu_mes_process {
-	int			pasid;
-	struct			amdgpu_vm *vm;
-	uint64_t		pd_gpu_addr;
-	struct amdgpu_bo 	*proc_ctx_bo;
-	uint64_t 		proc_ctx_gpu_addr;
-	void 			*proc_ctx_cpu_ptr;
-	uint64_t 		process_quantum;
-	struct 			list_head gang_list;
-	uint32_t 		doorbell_index;
-	unsigned long 		*doorbell_bitmap;
-	struct mutex		doorbell_lock;
+	/* mes resource_1 bo*/
+	struct amdgpu_bo    *resource_1[AMDGPU_MAX_MES_PIPES];
+	uint64_t            resource_1_gpu_addr[AMDGPU_MAX_MES_PIPES];
+	void                *resource_1_addr[AMDGPU_MAX_MES_PIPES];
+
+	int				hung_queue_db_array_size;
+	int				hung_queue_hqd_info_offset;
+	struct amdgpu_bo		*hung_queue_db_array_gpu_obj;
+	uint64_t			hung_queue_db_array_gpu_addr;
+	void				*hung_queue_db_array_cpu_addr;
 };
 
 struct amdgpu_mes_gang {
@@ -219,14 +228,27 @@ struct mes_add_queue_input {
 	uint32_t        gws_size;
 	uint64_t	tba_addr;
 	uint64_t	tma_addr;
+	uint32_t	trap_en;
+	uint32_t	skip_process_ctx_clear;
 	uint32_t	is_kfd_process;
 	uint32_t	is_aql_queue;
 	uint32_t	queue_size;
+	uint32_t	exclusively_scheduled;
 };
 
 struct mes_remove_queue_input {
 	uint32_t	doorbell_offset;
 	uint64_t	gang_context_addr;
+	bool		remove_queue_after_reset;
+};
+
+struct mes_map_legacy_queue_input {
+	uint32_t                           queue_type;
+	uint32_t                           doorbell_offset;
+	uint32_t                           pipe_id;
+	uint32_t                           queue_id;
+	uint64_t                           mqd_addr;
+	uint64_t                           wptr_addr;
 };
 
 struct mes_unmap_legacy_queue_input {
@@ -251,11 +273,39 @@ struct mes_resume_gang_input {
 	uint64_t	gang_context_addr;
 };
 
+struct mes_reset_queue_input {
+	uint32_t                           queue_type;
+	uint32_t                           doorbell_offset;
+	bool                               use_mmio;
+	uint32_t                           me_id;
+	uint32_t                           pipe_id;
+	uint32_t                           queue_id;
+	uint64_t                           mqd_addr;
+	uint64_t                           wptr_addr;
+	uint32_t                           vmid;
+	bool                               legacy_gfx;
+	bool                               is_kq;
+};
+
+struct mes_detect_and_reset_queue_input {
+	uint32_t                           queue_type;
+	bool                               detect_only;
+};
+
+struct mes_inv_tlbs_pasid_input {
+	uint32_t        xcc_id;
+	uint16_t        pasid;
+	uint8_t         hub_id;
+	uint8_t         flush_type;
+};
+
 enum mes_misc_opcode {
 	MES_MISC_OP_WRITE_REG,
 	MES_MISC_OP_READ_REG,
 	MES_MISC_OP_WRM_REG_WAIT,
 	MES_MISC_OP_WRM_REG_WR_WAIT,
+	MES_MISC_OP_SET_SHADER_DEBUGGER,
+	MES_MISC_OP_CHANGE_CONFIG,
 };
 
 struct mes_misc_op_input {
@@ -278,6 +328,37 @@ struct mes_misc_op_input {
 			uint32_t                   reg0;
 			uint32_t                   reg1;
 		} wrm_reg;
+
+		struct {
+			uint64_t process_context_addr;
+			union {
+				struct {
+					uint32_t single_memop : 1;
+					uint32_t single_alu_op : 1;
+					uint32_t reserved: 29;
+					uint32_t process_ctx_flush: 1;
+				};
+				uint32_t u32all;
+			} flags;
+			uint32_t spi_gdbg_per_vmid_cntl;
+			uint32_t tcp_watch_cntl[4];
+			uint32_t trap_en;
+		} set_shader_debugger;
+
+		struct {
+			union {
+				struct {
+					uint32_t limit_single_process : 1;
+					uint32_t enable_hws_logging_buffer : 1;
+					uint32_t reserved : 30;
+				};
+				uint32_t all;
+			} option;
+			struct {
+				uint32_t tdr_level;
+				uint32_t tdr_delay;
+			} tdr_config;
+		} change_config;
 	};
 };
 
@@ -288,6 +369,9 @@ struct amdgpu_mes_funcs {
 	int (*remove_hw_queue)(struct amdgpu_mes *mes,
 			       struct mes_remove_queue_input *input);
 
+	int (*map_legacy_queue)(struct amdgpu_mes *mes,
+				struct mes_map_legacy_queue_input *input);
+
 	int (*unmap_legacy_queue)(struct amdgpu_mes *mes,
 				  struct mes_unmap_legacy_queue_input *input);
 
@@ -299,77 +383,65 @@ struct amdgpu_mes_funcs {
 
 	int (*misc_op)(struct amdgpu_mes *mes,
 		       struct mes_misc_op_input *input);
+
+	int (*reset_hw_queue)(struct amdgpu_mes *mes,
+			      struct mes_reset_queue_input *input);
+
+	int (*detect_and_reset_hung_queues)(struct amdgpu_mes *mes,
+			      struct mes_detect_and_reset_queue_input *input);
+
+
+	int (*invalidate_tlbs_pasid)(struct amdgpu_mes *mes,
+			      struct mes_inv_tlbs_pasid_input *input);
 };
 
 #define amdgpu_mes_kiq_hw_init(adev) (adev)->mes.kiq_hw_init((adev))
 #define amdgpu_mes_kiq_hw_fini(adev) (adev)->mes.kiq_hw_fini((adev))
 
-int amdgpu_mes_ctx_get_offs(struct amdgpu_ring *ring, unsigned int id_offs);
-
 int amdgpu_mes_init_microcode(struct amdgpu_device *adev, int pipe);
 int amdgpu_mes_init(struct amdgpu_device *adev);
 void amdgpu_mes_fini(struct amdgpu_device *adev);
 
-int amdgpu_mes_create_process(struct amdgpu_device *adev, int pasid,
-			      struct amdgpu_vm *vm);
-void amdgpu_mes_destroy_process(struct amdgpu_device *adev, int pasid);
-
-int amdgpu_mes_add_gang(struct amdgpu_device *adev, int pasid,
-			struct amdgpu_mes_gang_properties *gprops,
-			int *gang_id);
-int amdgpu_mes_remove_gang(struct amdgpu_device *adev, int gang_id);
-
 int amdgpu_mes_suspend(struct amdgpu_device *adev);
 int amdgpu_mes_resume(struct amdgpu_device *adev);
 
-int amdgpu_mes_add_hw_queue(struct amdgpu_device *adev, int gang_id,
-			    struct amdgpu_mes_queue_properties *qprops,
-			    int *queue_id);
-int amdgpu_mes_remove_hw_queue(struct amdgpu_device *adev, int queue_id);
-
+int amdgpu_mes_map_legacy_queue(struct amdgpu_device *adev,
+				struct amdgpu_ring *ring);
 int amdgpu_mes_unmap_legacy_queue(struct amdgpu_device *adev,
 				  struct amdgpu_ring *ring,
 				  enum amdgpu_unmap_queues_action action,
 				  u64 gpu_addr, u64 seq);
+int amdgpu_mes_reset_legacy_queue(struct amdgpu_device *adev,
+				  struct amdgpu_ring *ring,
+				  unsigned int vmid,
+				  bool use_mmio);
+
+int amdgpu_mes_get_hung_queue_db_array_size(struct amdgpu_device *adev);
+int amdgpu_mes_detect_and_reset_hung_queues(struct amdgpu_device *adev,
+					    int queue_type,
+					    bool detect_only,
+					    unsigned int *hung_db_num,
+					    u32 *hung_db_array);
 
 uint32_t amdgpu_mes_rreg(struct amdgpu_device *adev, uint32_t reg);
 int amdgpu_mes_wreg(struct amdgpu_device *adev,
 		    uint32_t reg, uint32_t val);
-int amdgpu_mes_reg_wait(struct amdgpu_device *adev, uint32_t reg,
-			uint32_t val, uint32_t mask);
 int amdgpu_mes_reg_write_reg_wait(struct amdgpu_device *adev,
 				  uint32_t reg0, uint32_t reg1,
 				  uint32_t ref, uint32_t mask);
-
-int amdgpu_mes_add_ring(struct amdgpu_device *adev, int gang_id,
-			int queue_type, int idx,
-			struct amdgpu_mes_ctx_data *ctx_data,
-			struct amdgpu_ring **out);
-void amdgpu_mes_remove_ring(struct amdgpu_device *adev,
-			    struct amdgpu_ring *ring);
+int amdgpu_mes_hdp_flush(struct amdgpu_device *adev);
+int amdgpu_mes_set_shader_debugger(struct amdgpu_device *adev,
+				uint64_t process_context_addr,
+				uint32_t spi_gdbg_per_vmid_cntl,
+				const uint32_t *tcp_watch_cntl,
+				uint32_t flags,
+				bool trap_en);
+int amdgpu_mes_flush_shader_debugger(struct amdgpu_device *adev,
+				uint64_t process_context_addr);
 
 uint32_t amdgpu_mes_get_aggregated_doorbell_index(struct amdgpu_device *adev,
 						   enum amdgpu_mes_priority_level prio);
 
-int amdgpu_mes_ctx_alloc_meta_data(struct amdgpu_device *adev,
-				   struct amdgpu_mes_ctx_data *ctx_data);
-void amdgpu_mes_ctx_free_meta_data(struct amdgpu_mes_ctx_data *ctx_data);
-int amdgpu_mes_ctx_map_meta_data(struct amdgpu_device *adev,
-				 struct amdgpu_vm *vm,
-				 struct amdgpu_mes_ctx_data *ctx_data);
-int amdgpu_mes_ctx_unmap_meta_data(struct amdgpu_device *adev,
-				   struct amdgpu_mes_ctx_data *ctx_data);
-
-int amdgpu_mes_self_test(struct amdgpu_device *adev);
-
-int amdgpu_mes_alloc_process_doorbells(struct amdgpu_device *adev,
-					unsigned int *doorbell_index);
-void amdgpu_mes_free_process_doorbells(struct amdgpu_device *adev,
-					unsigned int doorbell_index);
-unsigned int amdgpu_mes_get_doorbell_dw_offset_in_bar(
-					struct amdgpu_device *adev,
-					uint32_t doorbell_index,
-					unsigned int doorbell_id);
 int amdgpu_mes_doorbell_process_slice(struct amdgpu_device *adev);
 
 /*
@@ -429,4 +501,9 @@ static inline void amdgpu_mes_unlock(struct amdgpu_mes *mes)
 	memalloc_noreclaim_restore(mes->saved_flags);
 	mutex_unlock(&mes->mutex_hidden);
 }
+
+bool amdgpu_mes_suspend_resume_all_supported(struct amdgpu_device *adev);
+
+int amdgpu_mes_update_enforce_isolation(struct amdgpu_device *adev);
+
 #endif /* __AMDGPU_MES_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h
index d21bb6dae56e..1ca9d4ed8063 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mmhub.h
@@ -21,6 +21,29 @@
 #ifndef __AMDGPU_MMHUB_H__
 #define __AMDGPU_MMHUB_H__
 
+enum amdgpu_mmhub_ras_memory_id {
+	AMDGPU_MMHUB_WGMI_PAGEMEM = 0,
+	AMDGPU_MMHUB_RGMI_PAGEMEM = 1,
+	AMDGPU_MMHUB_WDRAM_PAGEMEM = 2,
+	AMDGPU_MMHUB_RDRAM_PAGEMEM = 3,
+	AMDGPU_MMHUB_WIO_CMDMEM = 4,
+	AMDGPU_MMHUB_RIO_CMDMEM = 5,
+	AMDGPU_MMHUB_WGMI_CMDMEM = 6,
+	AMDGPU_MMHUB_RGMI_CMDMEM = 7,
+	AMDGPU_MMHUB_WDRAM_CMDMEM = 8,
+	AMDGPU_MMHUB_RDRAM_CMDMEM = 9,
+	AMDGPU_MMHUB_MAM_DMEM0 = 10,
+	AMDGPU_MMHUB_MAM_DMEM1 = 11,
+	AMDGPU_MMHUB_MAM_DMEM2 = 12,
+	AMDGPU_MMHUB_MAM_DMEM3 = 13,
+	AMDGPU_MMHUB_WRET_TAGMEM = 19,
+	AMDGPU_MMHUB_RRET_TAGMEM = 20,
+	AMDGPU_MMHUB_WIO_DATAMEM = 21,
+	AMDGPU_MMHUB_WGMI_DATAMEM = 22,
+	AMDGPU_MMHUB_WDRAM_DATAMEM = 23,
+	AMDGPU_MMHUB_MEMORY_BLOCK_LAST,
+};
+
 struct amdgpu_mmhub_ras {
 	struct amdgpu_ras_block_object ras_block;
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h
index 32fe05c810c6..dc8d2f52c7d6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h
@@ -32,7 +32,6 @@
 
 #include <drm/display/drm_dp_helper.h>
 #include <drm/drm_crtc.h>
-#include <drm/drm_edid.h>
 #include <drm/drm_encoder.h>
 #include <drm/drm_fixed.h>
 #include <drm/drm_framebuffer.h>
@@ -51,6 +50,8 @@ struct amdgpu_device;
 struct amdgpu_encoder;
 struct amdgpu_router;
 struct amdgpu_hpd;
+struct edid;
+struct drm_edid;
 
 #define to_amdgpu_crtc(x) container_of(x, struct amdgpu_crtc, base)
 #define to_amdgpu_connector(x) container_of(x, struct amdgpu_connector, base)
@@ -300,6 +301,7 @@ struct amdgpu_framebuffer {
 
 	uint64_t tiling_flags;
 	bool tmz_surface;
+	bool gfx12_dcc;
 
 	/* caching for later use */
 	uint64_t address;
@@ -327,8 +329,7 @@ struct amdgpu_mode_info {
 	/* Adaptive Backlight Modulation (power feature) */
 	struct drm_property *abm_level_property;
 	/* hardcoded DFP edid from BIOS */
-	struct edid *bios_hardcoded_edid;
-	int bios_hardcoded_edid_size;
+	const struct drm_edid *bios_hardcoded_edid;
 
 	/* firmware flags */
 	u32 firmware_flags;
@@ -343,6 +344,97 @@ struct amdgpu_mode_info {
 	int			disp_priority;
 	const struct amdgpu_display_funcs *funcs;
 	const enum drm_plane_type *plane_type;
+
+	/* Driver-private color mgmt props */
+
+	/* @plane_degamma_lut_property: Plane property to set a degamma LUT to
+	 * convert encoded values to light linear values before sampling or
+	 * blending.
+	 */
+	struct drm_property *plane_degamma_lut_property;
+	/* @plane_degamma_lut_size_property: Plane property to define the max
+	 * size of degamma LUT as supported by the driver (read-only).
+	 */
+	struct drm_property *plane_degamma_lut_size_property;
+	/**
+	 * @plane_degamma_tf_property: Plane pre-defined transfer function to
+	 * to go from scanout/encoded values to linear values.
+	 */
+	struct drm_property *plane_degamma_tf_property;
+	/**
+	 * @plane_hdr_mult_property:
+	 */
+	struct drm_property *plane_hdr_mult_property;
+
+	struct drm_property *plane_ctm_property;
+	/**
+	 * @shaper_lut_property: Plane property to set pre-blending shaper LUT
+	 * that converts color content before 3D LUT. If
+	 * plane_shaper_tf_property != Identity TF, AMD color module will
+	 * combine the user LUT values with pre-defined TF into the LUT
+	 * parameters to be programmed.
+	 */
+	struct drm_property *plane_shaper_lut_property;
+	/**
+	 * @shaper_lut_size_property: Plane property for the size of
+	 * pre-blending shaper LUT as supported by the driver (read-only).
+	 */
+	struct drm_property *plane_shaper_lut_size_property;
+	/**
+	 * @plane_shaper_tf_property: Plane property to set a predefined
+	 * transfer function for pre-blending shaper (before applying 3D LUT)
+	 * with or without LUT. There is no shaper ROM, but we can use AMD
+	 * color modules to program LUT parameters from predefined TF (or
+	 * from a combination of pre-defined TF and the custom 1D LUT).
+	 */
+	struct drm_property *plane_shaper_tf_property;
+	/**
+	 * @plane_lut3d_property: Plane property for color transformation using
+	 * a 3D LUT (pre-blending), a three-dimensional array where each
+	 * element is an RGB triplet. Each dimension has the size of
+	 * lut3d_size. The array contains samples from the approximated
+	 * function. On AMD, values between samples are estimated by
+	 * tetrahedral interpolation. The array is accessed with three indices,
+	 * one for each input dimension (color channel), blue being the
+	 * outermost dimension, red the innermost.
+	 */
+	struct drm_property *plane_lut3d_property;
+	/**
+	 * @plane_degamma_lut_size_property: Plane property to define the max
+	 * size of 3D LUT as supported by the driver (read-only). The max size
+	 * is the max size of one dimension and, therefore, the max number of
+	 * entries for 3D LUT array is the 3D LUT size cubed;
+	 */
+	struct drm_property *plane_lut3d_size_property;
+	/**
+	 * @plane_blend_lut_property: Plane property for output gamma before
+	 * blending. Userspace set a blend LUT to convert colors after 3D LUT
+	 * conversion. It works as a post-3DLUT 1D LUT. With shaper LUT, they
+	 * are sandwiching 3D LUT with two 1D LUT. If plane_blend_tf_property
+	 * != Identity TF, AMD color module will combine the user LUT values
+	 * with pre-defined TF into the LUT parameters to be programmed.
+	 */
+	struct drm_property *plane_blend_lut_property;
+	/**
+	 * @plane_blend_lut_size_property: Plane property to define the max
+	 * size of blend LUT as supported by the driver (read-only).
+	 */
+	struct drm_property *plane_blend_lut_size_property;
+	/**
+	 * @plane_blend_tf_property: Plane property to set a predefined
+	 * transfer function for pre-blending blend/out_gamma (after applying
+	 * 3D LUT) with or without LUT. There is no blend ROM, but we can use
+	 * AMD color modules to program LUT parameters from predefined TF (or
+	 * from a combination of pre-defined TF and the custom 1D LUT).
+	 */
+	struct drm_property *plane_blend_tf_property;
+	/* @regamma_tf_property: Transfer function for CRTC regamma
+	 * (post-blending). Possible values are defined by `enum
+	 * amdgpu_transfer_function`. There is no regamma ROM, but we can use
+	 * AMD color modules to program LUT parameters from predefined TF (or
+	 * from a combination of pre-defined TF and the custom 1D LUT).
+	 */
+	struct drm_property *regamma_tf_property;
 };
 
 #define AMDGPU_MAX_BL_LEVEL 0xFF
@@ -406,8 +498,6 @@ struct amdgpu_crtc {
 	struct drm_connector *connector;
 	/* for dpm */
 	u32 line_time;
-	u32 wm_low;
-	u32 wm_high;
 	u32 lb_vblank_lead_lines;
 	struct drm_display_mode hw_mode;
 	/* for virtual dce */
@@ -416,6 +506,10 @@ struct amdgpu_crtc {
 
 	int otg_inst;
 	struct drm_pending_vblank_event *event;
+
+	bool wb_pending;
+	bool wb_enabled;
+	struct drm_writeback_connector *wb_conn;
 };
 
 struct amdgpu_encoder_atom_dig {
@@ -515,6 +609,7 @@ struct amdgpu_i2c_adapter {
 	struct i2c_adapter base;
 
 	struct ddc_service *ddc_service;
+	bool oem;
 };
 
 #define TO_DM_AUX(x) container_of((x), struct amdgpu_dm_dp_aux, aux)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c
index a3bc00577a7c..a974265837f0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.c
@@ -45,6 +45,24 @@ int amdgpu_nbio_ras_sw_init(struct amdgpu_device *adev)
 	return 0;
 }
 
+u64 amdgpu_nbio_get_pcie_replay_count(struct amdgpu_device *adev)
+{
+	if (adev->nbio.funcs && adev->nbio.funcs->get_pcie_replay_count)
+		return adev->nbio.funcs->get_pcie_replay_count(adev);
+
+	return 0;
+}
+
+bool amdgpu_nbio_is_replay_cnt_supported(struct amdgpu_device *adev)
+{
+	if (amdgpu_sriov_vf(adev) || !adev->asic_funcs ||
+	    !adev->asic_funcs->get_pcie_replay_count ||
+	    (!adev->nbio.funcs || !adev->nbio.funcs->get_pcie_replay_count))
+		return false;
+
+	return true;
+}
+
 int amdgpu_nbio_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
 {
 	int r;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h
index c686ff4bcc39..b528de6a01f6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_nbio.h
@@ -61,6 +61,7 @@ struct amdgpu_nbio_funcs {
 	u32 (*get_hdp_flush_done_offset)(struct amdgpu_device *adev);
 	u32 (*get_pcie_index_offset)(struct amdgpu_device *adev);
 	u32 (*get_pcie_data_offset)(struct amdgpu_device *adev);
+	u32 (*get_pcie_index_hi_offset)(struct amdgpu_device *adev);
 	u32 (*get_pcie_port_index_offset)(struct amdgpu_device *adev);
 	u32 (*get_pcie_port_data_offset)(struct amdgpu_device *adev);
 	u32 (*get_rev_id)(struct amdgpu_device *adev);
@@ -68,6 +69,8 @@ struct amdgpu_nbio_funcs {
 	u32 (*get_memsize)(struct amdgpu_device *adev);
 	void (*sdma_doorbell_range)(struct amdgpu_device *adev, int instance,
 			bool use_doorbell, int doorbell_index, int doorbell_size);
+	void (*vpe_doorbell_range)(struct amdgpu_device *adev, int instance,
+			bool use_doorbell, int doorbell_index, int doorbell_size);
 	void (*vcn_doorbell_range)(struct amdgpu_device *adev, bool use_doorbell,
 				   int doorbell_index, int instance);
 	void (*gc_doorbell_init)(struct amdgpu_device *adev);
@@ -95,6 +98,12 @@ struct amdgpu_nbio_funcs {
 	void (*apply_l1_link_width_reconfig_wa)(struct amdgpu_device *adev);
 	void (*clear_doorbell_interrupt)(struct amdgpu_device *adev);
 	u32 (*get_rom_offset)(struct amdgpu_device *adev);
+	int (*get_compute_partition_mode)(struct amdgpu_device *adev);
+	u32 (*get_memory_partition_mode)(struct amdgpu_device *adev,
+					 u32 *supp_modes);
+	bool (*is_nps_switch_requested)(struct amdgpu_device *adev);
+	u64 (*get_pcie_replay_count)(struct amdgpu_device *adev);
+	void (*set_reg_remap)(struct amdgpu_device *adev);
 };
 
 struct amdgpu_nbio {
@@ -108,4 +117,8 @@ struct amdgpu_nbio {
 
 int amdgpu_nbio_ras_sw_init(struct amdgpu_device *adev);
 int amdgpu_nbio_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block);
+u64 amdgpu_nbio_get_pcie_replay_count(struct amdgpu_device *adev);
+
+bool amdgpu_nbio_is_replay_cnt_supported(struct amdgpu_device *adev);
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
index 2bd1a54ee866..e08f58de4b17 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
@@ -32,6 +32,7 @@
 #include <linux/list.h>
 #include <linux/slab.h>
 #include <linux/dma-buf.h>
+#include <linux/export.h>
 
 #include <drm/drm_drv.h>
 #include <drm/amdgpu_drm.h>
@@ -39,6 +40,9 @@
 #include "amdgpu.h"
 #include "amdgpu_trace.h"
 #include "amdgpu_amdkfd.h"
+#include "amdgpu_vram_mgr.h"
+#include "amdgpu_vm.h"
+#include "amdgpu_dma_buf.h"
 
 /**
  * DOC: amdgpu_object
@@ -59,7 +63,7 @@ static void amdgpu_bo_destroy(struct ttm_buffer_object *tbo)
 
 	amdgpu_bo_kunmap(bo);
 
-	if (bo->tbo.base.import_attach)
+	if (drm_gem_is_imported(&bo->tbo.base))
 		drm_prime_gem_destroy(&bo->tbo.base, bo->tbo.sg);
 	drm_gem_object_release(&bo->tbo.base);
 	amdgpu_bo_unref(&bo->parent);
@@ -76,23 +80,6 @@ static void amdgpu_bo_user_destroy(struct ttm_buffer_object *tbo)
 	amdgpu_bo_destroy(tbo);
 }
 
-static void amdgpu_bo_vm_destroy(struct ttm_buffer_object *tbo)
-{
-	struct amdgpu_device *adev = amdgpu_ttm_adev(tbo->bdev);
-	struct amdgpu_bo *bo = ttm_to_amdgpu_bo(tbo);
-	struct amdgpu_bo_vm *vmbo;
-
-	vmbo = to_amdgpu_bo_vm(bo);
-	/* in case amdgpu_device_recover_vram got NULL of bo->parent */
-	if (!list_empty(&vmbo->shadow_list)) {
-		mutex_lock(&adev->shadow_list_lock);
-		list_del_init(&vmbo->shadow_list);
-		mutex_unlock(&adev->shadow_list_lock);
-	}
-
-	amdgpu_bo_destroy(tbo);
-}
-
 /**
  * amdgpu_bo_is_amdgpu_bo - check if the buffer object is an &amdgpu_bo
  * @bo: buffer object to be checked
@@ -106,8 +93,7 @@ static void amdgpu_bo_vm_destroy(struct ttm_buffer_object *tbo)
 bool amdgpu_bo_is_amdgpu_bo(struct ttm_buffer_object *bo)
 {
 	if (bo->destroy == &amdgpu_bo_destroy ||
-	    bo->destroy == &amdgpu_bo_user_destroy ||
-	    bo->destroy == &amdgpu_bo_vm_destroy)
+	    bo->destroy == &amdgpu_bo_user_destroy)
 		return true;
 
 	return false;
@@ -130,20 +116,48 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 	u32 c = 0;
 
 	if (domain & AMDGPU_GEM_DOMAIN_VRAM) {
-		unsigned visible_pfn = adev->gmc.visible_vram_size >> PAGE_SHIFT;
-
-		places[c].fpfn = 0;
-		places[c].lpfn = 0;
+		unsigned int visible_pfn = adev->gmc.visible_vram_size >> PAGE_SHIFT;
+		int8_t mem_id = KFD_XCP_MEM_ID(adev, abo->xcp_id);
+
+		if (adev->gmc.mem_partitions && mem_id >= 0) {
+			places[c].fpfn = adev->gmc.mem_partitions[mem_id].range.fpfn;
+			/*
+			 * memory partition range lpfn is inclusive start + size - 1
+			 * TTM place lpfn is exclusive start + size
+			 */
+			places[c].lpfn = adev->gmc.mem_partitions[mem_id].range.lpfn + 1;
+		} else {
+			places[c].fpfn = 0;
+			places[c].lpfn = 0;
+		}
 		places[c].mem_type = TTM_PL_VRAM;
 		places[c].flags = 0;
 
 		if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)
-			places[c].lpfn = visible_pfn;
-		else if (adev->gmc.real_vram_size != adev->gmc.visible_vram_size)
+			places[c].lpfn = min_not_zero(places[c].lpfn, visible_pfn);
+		else
 			places[c].flags |= TTM_PL_FLAG_TOPDOWN;
 
-		if (flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)
+		if (abo->tbo.type == ttm_bo_type_kernel &&
+		    flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)
 			places[c].flags |= TTM_PL_FLAG_CONTIGUOUS;
+
+		c++;
+	}
+
+	if (domain & AMDGPU_GEM_DOMAIN_DOORBELL) {
+		places[c].fpfn = 0;
+		places[c].lpfn = 0;
+		places[c].mem_type = AMDGPU_PL_DOORBELL;
+		places[c].flags = 0;
+		c++;
+	}
+
+	if (domain & AMDGPU_GEM_DOMAIN_MMIO_REMAP) {
+		places[c].fpfn = 0;
+		places[c].lpfn = 0;
+		places[c].mem_type = AMDGPU_PL_MMIO_REMAP;
+		places[c].flags = 0;
 		c++;
 	}
 
@@ -154,6 +168,13 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 			abo->flags & AMDGPU_GEM_CREATE_PREEMPTIBLE ?
 			AMDGPU_PL_PREEMPT : TTM_PL_TT;
 		places[c].flags = 0;
+		/*
+		 * When GTT is just an alternative to VRAM make sure that we
+		 * only use it as fallback and still try to fill up VRAM first.
+		 */
+		if (abo->tbo.resource && !(adev->flags & AMD_IS_APU) &&
+		    domain & abo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM)
+			places[c].flags |= TTM_PL_FLAG_FALLBACK;
 		c++;
 	}
 
@@ -201,9 +222,6 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 
 	placement->num_placement = c;
 	placement->placement = places;
-
-	placement->num_busy_placement = c;
-	placement->busy_placement = places;
 }
 
 /**
@@ -316,6 +334,9 @@ error_free:
  *
  * Allocates and pins a BO for kernel internal use.
  *
+ * This function is exported to allow the V4L2 isp device
+ * external to drm device to create and access the kernel BO.
+ *
  * Note: For bo_ptr new BO is only created if bo_ptr points to NULL.
  *
  * Returns:
@@ -341,6 +362,74 @@ int amdgpu_bo_create_kernel(struct amdgpu_device *adev,
 }
 
 /**
+ * amdgpu_bo_create_isp_user - create user BO for isp
+ *
+ * @adev: amdgpu device object
+ * @dma_buf: DMABUF handle for isp buffer
+ * @domain: where to place it
+ * @bo:  used to initialize BOs in structures
+ * @gpu_addr: GPU addr of the pinned BO
+ *
+ * Imports isp DMABUF to allocate and pin a user BO for isp internal use. It does
+ * GART alloc to generate gpu_addr for BO to make it accessible through the
+ * GART aperture for ISP HW.
+ *
+ * This function is exported to allow the V4L2 isp device external to drm device
+ * to create and access the isp user BO.
+ *
+ * Returns:
+ * 0 on success, negative error code otherwise.
+ */
+int amdgpu_bo_create_isp_user(struct amdgpu_device *adev,
+			   struct dma_buf *dma_buf, u32 domain, struct amdgpu_bo **bo,
+			   u64 *gpu_addr)
+
+{
+	struct drm_gem_object *gem_obj;
+	int r;
+
+	gem_obj = amdgpu_gem_prime_import(&adev->ddev, dma_buf);
+	*bo = gem_to_amdgpu_bo(gem_obj);
+	if (!(*bo)) {
+		dev_err(adev->dev, "failed to get valid isp user bo\n");
+		return -EINVAL;
+	}
+
+	r = amdgpu_bo_reserve(*bo, false);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to reserve isp user bo\n", r);
+		return r;
+	}
+
+	r = amdgpu_bo_pin(*bo, domain);
+	if (r) {
+		dev_err(adev->dev, "(%d) isp user bo pin failed\n", r);
+		goto error_unreserve;
+	}
+
+	r = amdgpu_ttm_alloc_gart(&(*bo)->tbo);
+	if (r) {
+		dev_err(adev->dev, "%p bind failed\n", *bo);
+		goto error_unpin;
+	}
+
+	if (!WARN_ON(!gpu_addr))
+		*gpu_addr = amdgpu_bo_gpu_offset(*bo);
+
+	amdgpu_bo_unreserve(*bo);
+
+	return 0;
+
+error_unpin:
+	amdgpu_bo_unpin(*bo);
+error_unreserve:
+	amdgpu_bo_unreserve(*bo);
+	amdgpu_bo_unref(bo);
+
+	return r;
+}
+
+/**
  * amdgpu_bo_create_kernel_at - create BO for kernel use at specific location
  *
  * @adev: amdgpu device object
@@ -415,6 +504,9 @@ error:
  * @cpu_addr: pointer to where the BO's CPU memory space address was stored
  *
  * unmaps and unpin a BO for kernel internal use.
+ *
+ * This function is exported to allow the V4L2 isp device
+ * external to drm device to free the kernel BO.
  */
 void amdgpu_bo_free_kernel(struct amdgpu_bo **bo, u64 *gpu_addr,
 			   void **cpu_addr)
@@ -440,7 +532,29 @@ void amdgpu_bo_free_kernel(struct amdgpu_bo **bo, u64 *gpu_addr,
 		*cpu_addr = NULL;
 }
 
-/* Validate bo size is bit bigger then the request domain */
+/**
+ * amdgpu_bo_free_isp_user - free BO for isp use
+ *
+ * @bo: amdgpu isp user BO to free
+ *
+ * unpin and unref BO for isp internal use.
+ *
+ * This function is exported to allow the V4L2 isp device
+ * external to drm device to free the isp user BO.
+ */
+void amdgpu_bo_free_isp_user(struct amdgpu_bo *bo)
+{
+	if (bo == NULL)
+		return;
+
+	if (amdgpu_bo_reserve(bo, true) == 0) {
+		amdgpu_bo_unpin(bo);
+		amdgpu_bo_unreserve(bo);
+	}
+	amdgpu_bo_unref(&bo);
+}
+
+/* Validate bo size is bit bigger than the request domain */
 static bool amdgpu_bo_validate_size(struct amdgpu_device *adev,
 					  unsigned long size, u32 domain)
 {
@@ -450,29 +564,24 @@ static bool amdgpu_bo_validate_size(struct amdgpu_device *adev,
 	 * If GTT is part of requested domains the check must succeed to
 	 * allow fall back to GTT.
 	 */
-	if (domain & AMDGPU_GEM_DOMAIN_GTT) {
+	if (domain & AMDGPU_GEM_DOMAIN_GTT)
 		man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT);
-
-		if (man && size < man->size)
-			return true;
-		else if (!man)
-			WARN_ON_ONCE("GTT domain requested but GTT mem manager uninitialized");
-		goto fail;
-	} else if (domain & AMDGPU_GEM_DOMAIN_VRAM) {
+	else if (domain & AMDGPU_GEM_DOMAIN_VRAM)
 		man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM);
+	else
+		return true;
 
-		if (man && size < man->size)
-			return true;
-		goto fail;
+	if (!man) {
+		if (domain & AMDGPU_GEM_DOMAIN_GTT)
+			WARN_ON_ONCE("GTT domain requested but GTT mem manager uninitialized");
+		return false;
 	}
 
-	/* TODO add more domains checks, such as AMDGPU_GEM_DOMAIN_CPU */
-	return true;
+	/* TODO add more domains checks, such as AMDGPU_GEM_DOMAIN_CPU, _DOMAIN_DOORBELL */
+	if (size < man->size)
+		return true;
 
-fail:
-	if (man)
-		DRM_DEBUG("BO size %lu > total memory in domain: %llu\n", size,
-			  man->size);
+	DRM_DEBUG("BO size %lu > total memory in domain: %llu\n", size, man->size);
 	return false;
 }
 
@@ -563,6 +672,7 @@ int amdgpu_bo_create(struct amdgpu_device *adev,
 	if (bo == NULL)
 		return -ENOMEM;
 	drm_gem_private_object_init(adev_to_drm(adev), &bo->tbo.base, size);
+	bo->tbo.base.funcs = &amdgpu_gem_object_funcs;
 	bo->vm_bo = NULL;
 	bo->preferred_domains = bp->preferred_domain ? bp->preferred_domain :
 		bp->domain;
@@ -574,12 +684,16 @@ int amdgpu_bo_create(struct amdgpu_device *adev,
 
 	bo->flags = bp->flags;
 
+	if (adev->gmc.mem_partitions)
+		/* For GPUs with spatial partitioning, bo->xcp_id=-1 means any partition */
+		bo->xcp_id = bp->xcp_id_plus1 - 1;
+	else
+		/* For GPUs without spatial partitioning */
+		bo->xcp_id = 0;
+
 	if (!amdgpu_bo_support_uswc(bo->flags))
 		bo->flags &= ~AMDGPU_GEM_CREATE_CPU_GTT_USWC;
 
-	if (adev->ras_enabled)
-		bo->flags |= AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE;
-
 	bo->tbo.bdev = &adev->mman.bdev;
 	if (bp->domain & (AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA |
 			  AMDGPU_GEM_DOMAIN_GDS))
@@ -587,6 +701,8 @@ int amdgpu_bo_create(struct amdgpu_device *adev,
 	else
 		amdgpu_bo_placement_from_domain(bo, bp->domain);
 	if (bp->type == ttm_bo_type_kernel)
+		bo->tbo.priority = 2;
+	else if (!(bp->flags & AMDGPU_GEM_CREATE_DISCARDABLE))
 		bo->tbo.priority = 1;
 
 	if (!bp->destroy)
@@ -599,8 +715,7 @@ int amdgpu_bo_create(struct amdgpu_device *adev,
 		return r;
 
 	if (!amdgpu_gmc_vram_full_visible(&adev->gmc) &&
-	    bo->tbo.resource->mem_type == TTM_PL_VRAM &&
-	    amdgpu_bo_in_cpu_visible_vram(bo))
+	    amdgpu_res_cpu_visible(adev, bo->tbo.resource))
 		amdgpu_cs_report_moved_bytes(adev, ctx.bytes_moved,
 					     ctx.bytes_moved);
 	else
@@ -610,7 +725,7 @@ int amdgpu_bo_create(struct amdgpu_device *adev,
 	    bo->tbo.resource->mem_type == TTM_PL_VRAM) {
 		struct dma_fence *fence;
 
-		r = amdgpu_fill_buffer(bo, 0, bo->tbo.base.resv, &fence);
+		r = amdgpu_ttm_clear_buffer(bo, bo->tbo.base.resv, &fence);
 		if (unlikely(r))
 			goto fail_unreserve;
 
@@ -694,59 +809,10 @@ int amdgpu_bo_create_vm(struct amdgpu_device *adev,
 		return r;
 
 	*vmbo_ptr = to_amdgpu_bo_vm(bo_ptr);
-	INIT_LIST_HEAD(&(*vmbo_ptr)->shadow_list);
-	/* Set destroy callback to amdgpu_bo_vm_destroy after vmbo->shadow_list
-	 * is initialized.
-	 */
-	bo_ptr->tbo.destroy = &amdgpu_bo_vm_destroy;
 	return r;
 }
 
 /**
- * amdgpu_bo_add_to_shadow_list - add a BO to the shadow list
- *
- * @vmbo: BO that will be inserted into the shadow list
- *
- * Insert a BO to the shadow list.
- */
-void amdgpu_bo_add_to_shadow_list(struct amdgpu_bo_vm *vmbo)
-{
-	struct amdgpu_device *adev = amdgpu_ttm_adev(vmbo->bo.tbo.bdev);
-
-	mutex_lock(&adev->shadow_list_lock);
-	list_add_tail(&vmbo->shadow_list, &adev->shadow_list);
-	mutex_unlock(&adev->shadow_list_lock);
-}
-
-/**
- * amdgpu_bo_restore_shadow - restore an &amdgpu_bo shadow
- *
- * @shadow: &amdgpu_bo shadow to be restored
- * @fence: dma_fence associated with the operation
- *
- * Copies a buffer object's shadow content back to the object.
- * This is used for recovering a buffer from its shadow in case of a gpu
- * reset where vram context may be lost.
- *
- * Returns:
- * 0 for success or a negative error code on failure.
- */
-int amdgpu_bo_restore_shadow(struct amdgpu_bo *shadow, struct dma_fence **fence)
-
-{
-	struct amdgpu_device *adev = amdgpu_ttm_adev(shadow->tbo.bdev);
-	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
-	uint64_t shadow_addr, parent_addr;
-
-	shadow_addr = amdgpu_bo_gpu_offset(shadow);
-	parent_addr = amdgpu_bo_gpu_offset(shadow->parent);
-
-	return amdgpu_copy_buffer(ring, shadow_addr, parent_addr,
-				  amdgpu_bo_size(shadow), NULL, fence,
-				  true, false, false);
-}
-
-/**
  * amdgpu_bo_kmap - map an &amdgpu_bo buffer object
  * @bo: &amdgpu_bo buffer object to be mapped
  * @ptr: kernel virtual address to be returned
@@ -829,7 +895,7 @@ struct amdgpu_bo *amdgpu_bo_ref(struct amdgpu_bo *bo)
 	if (bo == NULL)
 		return NULL;
 
-	ttm_bo_get(&bo->tbo);
+	drm_gem_object_get(&bo->tbo.base);
 	return bo;
 }
 
@@ -841,40 +907,30 @@ struct amdgpu_bo *amdgpu_bo_ref(struct amdgpu_bo *bo)
  */
 void amdgpu_bo_unref(struct amdgpu_bo **bo)
 {
-	struct ttm_buffer_object *tbo;
-
 	if ((*bo) == NULL)
 		return;
 
-	tbo = &((*bo)->tbo);
-	ttm_bo_put(tbo);
+	drm_gem_object_put(&(*bo)->tbo.base);
 	*bo = NULL;
 }
 
 /**
- * amdgpu_bo_pin_restricted - pin an &amdgpu_bo buffer object
+ * amdgpu_bo_pin - pin an &amdgpu_bo buffer object
  * @bo: &amdgpu_bo buffer object to be pinned
  * @domain: domain to be pinned to
- * @min_offset: the start of requested address range
- * @max_offset: the end of requested address range
  *
- * Pins the buffer object according to requested domain and address range. If
- * the memory is unbound gart memory, binds the pages into gart table. Adjusts
- * pin_count and pin_size accordingly.
+ * Pins the buffer object according to requested domain. If the memory is
+ * unbound gart memory, binds the pages into gart table. Adjusts pin_count and
+ * pin_size accordingly.
  *
  * Pinning means to lock pages in memory along with keeping them at a fixed
  * offset. It is required when a buffer can not be moved, for example, when
  * a display buffer is being scanned out.
  *
- * Compared with amdgpu_bo_pin(), this function gives more flexibility on
- * where to pin a buffer if there are specific restrictions on where a buffer
- * must be located.
- *
  * Returns:
  * 0 for success or a negative error code on failure.
  */
-int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
-			     u64 min_offset, u64 max_offset)
+int amdgpu_bo_pin(struct amdgpu_bo *bo, u32 domain)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
 	struct ttm_operation_ctx ctx = { false, false };
@@ -883,15 +939,12 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
 	if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm))
 		return -EPERM;
 
-	if (WARN_ON_ONCE(min_offset > max_offset))
-		return -EINVAL;
-
 	/* Check domain to be pinned to against preferred domains */
 	if (bo->preferred_domains & domain)
 		domain = bo->preferred_domains & domain;
 
 	/* A shared bo cannot be migrated to VRAM */
-	if (bo->tbo.base.import_attach) {
+	if (drm_gem_is_imported(&bo->tbo.base)) {
 		if (domain & AMDGPU_GEM_DOMAIN_GTT)
 			domain = AMDGPU_GEM_DOMAIN_GTT;
 		else
@@ -911,14 +964,6 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
 			return -EINVAL;
 
 		ttm_bo_pin(&bo->tbo);
-
-		if (max_offset != 0) {
-			u64 domain_start = amdgpu_ttm_domain_start(adev,
-								   mem_type);
-			WARN_ON_ONCE(max_offset <
-				     (amdgpu_bo_gpu_offset(bo) - domain_start));
-		}
-
 		return 0;
 	}
 
@@ -927,7 +972,7 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
 	 */
 	domain = amdgpu_bo_get_preferred_domain(adev, domain);
 
-	if (bo->tbo.base.import_attach)
+	if (drm_gem_is_imported(&bo->tbo.base))
 		dma_buf_pin(bo->tbo.base.import_attach);
 
 	/* force to pin into visible video ram */
@@ -935,16 +980,9 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
 		bo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
 	amdgpu_bo_placement_from_domain(bo, domain);
 	for (i = 0; i < bo->placement.num_placement; i++) {
-		unsigned fpfn, lpfn;
-
-		fpfn = min_offset >> PAGE_SHIFT;
-		lpfn = max_offset >> PAGE_SHIFT;
-
-		if (fpfn > bo->placements[i].fpfn)
-			bo->placements[i].fpfn = fpfn;
-		if (!bo->placements[i].lpfn ||
-		    (lpfn && lpfn < bo->placements[i].lpfn))
-			bo->placements[i].lpfn = lpfn;
+		if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS &&
+		    bo->placements[i].mem_type == TTM_PL_VRAM)
+			bo->placements[i].flags |= TTM_PL_FLAG_CONTIGUOUS;
 	}
 
 	r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
@@ -955,12 +993,11 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
 
 	ttm_bo_pin(&bo->tbo);
 
-	domain = amdgpu_mem_type_to_domain(bo->tbo.resource->mem_type);
-	if (domain == AMDGPU_GEM_DOMAIN_VRAM) {
+	if (bo->tbo.resource->mem_type == TTM_PL_VRAM) {
 		atomic64_add(amdgpu_bo_size(bo), &adev->vram_pin_size);
 		atomic64_add(amdgpu_vram_mgr_bo_visible_size(bo),
 			     &adev->visible_pin_size);
-	} else if (domain == AMDGPU_GEM_DOMAIN_GTT) {
+	} else if (bo->tbo.resource->mem_type == TTM_PL_TT) {
 		atomic64_add(amdgpu_bo_size(bo), &adev->gart_pin_size);
 	}
 
@@ -969,24 +1006,6 @@ error:
 }
 
 /**
- * amdgpu_bo_pin - pin an &amdgpu_bo buffer object
- * @bo: &amdgpu_bo buffer object to be pinned
- * @domain: domain to be pinned to
- *
- * A simple wrapper to amdgpu_bo_pin_restricted().
- * Provides a simpler API for buffers that do not have any strict restrictions
- * on where a buffer must be located.
- *
- * Returns:
- * 0 for success or a negative error code on failure.
- */
-int amdgpu_bo_pin(struct amdgpu_bo *bo, u32 domain)
-{
-	bo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
-	return amdgpu_bo_pin_restricted(bo, domain, 0, 0);
-}
-
-/**
  * amdgpu_bo_unpin - unpin an &amdgpu_bo buffer object
  * @bo: &amdgpu_bo buffer object to be unpinned
  *
@@ -1004,7 +1023,7 @@ void amdgpu_bo_unpin(struct amdgpu_bo *bo)
 	if (bo->tbo.pin_count)
 		return;
 
-	if (bo->tbo.base.import_attach)
+	if (drm_gem_is_imported(&bo->tbo.base))
 		dma_buf_unpin(bo->tbo.base.import_attach);
 
 	if (bo->tbo.resource->mem_type == TTM_PL_VRAM) {
@@ -1014,9 +1033,10 @@ void amdgpu_bo_unpin(struct amdgpu_bo *bo)
 	} else if (bo->tbo.resource->mem_type == TTM_PL_TT) {
 		atomic64_sub(amdgpu_bo_size(bo), &adev->gart_pin_size);
 	}
+
 }
 
-static const char *amdgpu_vram_names[] = {
+static const char * const amdgpu_vram_names[] = {
 	"UNKNOWN",
 	"GDDR1",
 	"DDR2",
@@ -1029,7 +1049,8 @@ static const char *amdgpu_vram_names[] = {
 	"GDDR6",
 	"DDR5",
 	"LPDDR4",
-	"LPDDR5"
+	"LPDDR5",
+	"HBM3E"
 };
 
 /**
@@ -1044,7 +1065,7 @@ static const char *amdgpu_vram_names[] = {
 int amdgpu_bo_init(struct amdgpu_device *adev)
 {
 	/* On A+A platform, VRAM can be mapped as WB */
-	if (!adev->gmc.xgmi.connected_to_cpu) {
+	if (!adev->gmc.xgmi.connected_to_cpu && !adev->gmc.is_app_apu) {
 		/* reserve PAT memory space to WC for VRAM */
 		int r = arch_io_reserve_memtype_wc(adev->gmc.aper_base,
 				adev->gmc.aper_size);
@@ -1080,8 +1101,7 @@ void amdgpu_bo_fini(struct amdgpu_device *adev)
 	amdgpu_ttm_fini(adev);
 
 	if (drm_dev_enter(adev_to_drm(adev), &idx)) {
-
-		if (!adev->gmc.xgmi.connected_to_cpu) {
+		if (!adev->gmc.xgmi.connected_to_cpu && !adev->gmc.is_app_apu) {
 			arch_phys_wc_del(adev->gmc.vram_mtrr);
 			arch_io_free_memtype_wc(adev->gmc.aper_base, adev->gmc.aper_size);
 		}
@@ -1148,8 +1168,8 @@ void amdgpu_bo_get_tiling_flags(struct amdgpu_bo *bo, u64 *tiling_flags)
  * Returns:
  * 0 for success or a negative error code on failure.
  */
-int amdgpu_bo_set_metadata (struct amdgpu_bo *bo, void *metadata,
-			    uint32_t metadata_size, uint64_t flags)
+int amdgpu_bo_set_metadata(struct amdgpu_bo *bo, void *metadata,
+			   u32 metadata_size, uint64_t flags)
 {
 	struct amdgpu_bo_user *ubo;
 	void *buffer;
@@ -1227,7 +1247,7 @@ int amdgpu_bo_get_metadata(struct amdgpu_bo *bo, void *buffer,
  * amdgpu_bo_move_notify - notification about a memory move
  * @bo: pointer to a buffer object
  * @evict: if this move is evicting the buffer from the graphics address space
- * @new_mem: new information of the bufer object
+ * @new_mem: new resource for backing the BO
  *
  * Marks the corresponding &amdgpu_bo buffer object as invalid, also performs
  * bookkeeping.
@@ -1237,69 +1257,24 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo,
 			   bool evict,
 			   struct ttm_resource *new_mem)
 {
-	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);
-	struct amdgpu_bo *abo;
 	struct ttm_resource *old_mem = bo->resource;
+	struct amdgpu_bo *abo;
 
 	if (!amdgpu_bo_is_amdgpu_bo(bo))
 		return;
 
 	abo = ttm_to_amdgpu_bo(bo);
-	amdgpu_vm_bo_invalidate(adev, abo, evict);
+	amdgpu_vm_bo_move(abo, new_mem, evict);
 
 	amdgpu_bo_kunmap(abo);
 
-	if (abo->tbo.base.dma_buf && !abo->tbo.base.import_attach &&
-	    bo->resource->mem_type != TTM_PL_SYSTEM)
+	if (abo->tbo.base.dma_buf && !drm_gem_is_imported(&abo->tbo.base) &&
+	    old_mem && old_mem->mem_type != TTM_PL_SYSTEM)
 		dma_buf_move_notify(abo->tbo.base.dma_buf);
 
-	/* remember the eviction */
-	if (evict)
-		atomic64_inc(&adev->num_evictions);
-
-	/* update statistics */
-	if (!new_mem)
-		return;
-
 	/* move_notify is called before move happens */
-	trace_amdgpu_bo_move(abo, new_mem->mem_type, old_mem->mem_type);
-}
-
-void amdgpu_bo_get_memory(struct amdgpu_bo *bo,
-			  struct amdgpu_mem_stats *stats)
-{
-	unsigned int domain;
-	uint64_t size = amdgpu_bo_size(bo);
-
-	domain = amdgpu_mem_type_to_domain(bo->tbo.resource->mem_type);
-	switch (domain) {
-	case AMDGPU_GEM_DOMAIN_VRAM:
-		stats->vram += size;
-		if (amdgpu_bo_in_cpu_visible_vram(bo))
-			stats->visible_vram += size;
-		break;
-	case AMDGPU_GEM_DOMAIN_GTT:
-		stats->gtt += size;
-		break;
-	case AMDGPU_GEM_DOMAIN_CPU:
-	default:
-		stats->cpu += size;
-		break;
-	}
-
-	if (bo->preferred_domains & AMDGPU_GEM_DOMAIN_VRAM) {
-		stats->requested_vram += size;
-		if (bo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)
-			stats->requested_visible_vram += size;
-
-		if (domain != AMDGPU_GEM_DOMAIN_VRAM) {
-			stats->evicted_vram += size;
-			if (bo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)
-				stats->evicted_visible_vram += size;
-		}
-	} else if (bo->preferred_domains & AMDGPU_GEM_DOMAIN_GTT) {
-		stats->requested_gtt += size;
-	}
+	trace_amdgpu_bo_move(abo, new_mem ? new_mem->mem_type : -1,
+			     old_mem ? old_mem->mem_type : -1);
 }
 
 /**
@@ -1321,30 +1296,42 @@ void amdgpu_bo_release_notify(struct ttm_buffer_object *bo)
 
 	abo = ttm_to_amdgpu_bo(bo);
 
+	WARN_ON(abo->vm_bo);
+
 	if (abo->kfd_bo)
 		amdgpu_amdkfd_release_notify(abo);
 
-	/* We only remove the fence if the resv has individualized. */
-	WARN_ON_ONCE(bo->type == ttm_bo_type_kernel
-			&& bo->base.resv != &bo->base._resv);
-	if (bo->base.resv == &bo->base._resv)
-		amdgpu_amdkfd_remove_fence_on_pt_pd_bos(abo);
+	/*
+	 * We lock the private dma_resv object here and since the BO is about to
+	 * be released nobody else should have a pointer to it.
+	 * So when this locking here fails something is wrong with the reference
+	 * counting.
+	 */
+	if (WARN_ON_ONCE(!dma_resv_trylock(&bo->base._resv)))
+		return;
+
+	amdgpu_amdkfd_remove_all_eviction_fences(abo);
 
 	if (!bo->resource || bo->resource->mem_type != TTM_PL_VRAM ||
 	    !(abo->flags & AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE) ||
 	    adev->in_suspend || drm_dev_is_unplugged(adev_to_drm(adev)))
-		return;
+		goto out;
 
-	if (WARN_ON_ONCE(!dma_resv_trylock(bo->base.resv)))
-		return;
+	r = dma_resv_reserve_fences(&bo->base._resv, 1);
+	if (r)
+		goto out;
 
-	r = amdgpu_fill_buffer(abo, AMDGPU_POISON, bo->base.resv, &fence);
-	if (!WARN_ON(r)) {
-		amdgpu_bo_fence(abo, fence, false);
-		dma_fence_put(fence);
-	}
+	r = amdgpu_fill_buffer(abo, 0, &bo->base._resv, &fence, true,
+			       AMDGPU_KERNEL_JOB_ID_CLEAR_ON_RELEASE);
+	if (WARN_ON(r))
+		goto out;
+
+	amdgpu_vram_mgr_set_cleared(bo->resource);
+	dma_resv_add_fence(&bo->base._resv, fence, DMA_RESV_USAGE_KERNEL);
+	dma_fence_put(fence);
 
-	dma_resv_unlock(bo->base.resv);
+out:
+	dma_resv_unlock(&bo->base._resv);
 }
 
 /**
@@ -1368,10 +1355,7 @@ vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
 	/* Remember that this BO was accessed by the CPU */
 	abo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
 
-	if (bo->resource->mem_type != TTM_PL_VRAM)
-		return 0;
-
-	if (amdgpu_bo_in_cpu_visible_vram(abo))
+	if (amdgpu_res_cpu_visible(adev, bo->resource))
 		return 0;
 
 	/* Can't move a pinned BO to visible VRAM */
@@ -1384,8 +1368,7 @@ vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
 					AMDGPU_GEM_DOMAIN_GTT);
 
 	/* Avoid costly evictions; only set GTT as a busy placement */
-	abo->placement.num_busy_placement = 1;
-	abo->placement.busy_placement = &abo->placements[1];
+	abo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
 
 	r = ttm_bo_validate(bo, &abo->placement, &ctx);
 	if (unlikely(r == -EBUSY || r == -ERESTARTSYS))
@@ -1395,7 +1378,7 @@ vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo)
 
 	/* this should never happen */
 	if (bo->resource->mem_type == TTM_PL_VRAM &&
-	    !amdgpu_bo_in_cpu_visible_vram(abo))
+	    !amdgpu_res_cpu_visible(adev, bo->resource))
 		return VM_FAULT_SIGBUS;
 
 	ttm_bo_move_to_lru_tail_unlocked(bo);
@@ -1496,6 +1479,26 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
 }
 
 /**
+ * amdgpu_bo_fb_aper_addr - return FB aperture GPU offset of the VRAM bo
+ * @bo:	amdgpu VRAM buffer object for which we query the offset
+ *
+ * Returns:
+ * current FB aperture GPU offset of the object.
+ */
+u64 amdgpu_bo_fb_aper_addr(struct amdgpu_bo *bo)
+{
+	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
+	uint64_t offset, fb_base;
+
+	WARN_ON_ONCE(bo->tbo.resource->mem_type != TTM_PL_VRAM);
+
+	fb_base = adev->gmc.fb_start;
+	fb_base += adev->gmc.xgmi.physical_node_id * adev->gmc.xgmi.node_segment_size;
+	offset = (bo->tbo.resource->start << PAGE_SHIFT) + fb_base;
+	return amdgpu_gmc_sign_extend(offset);
+}
+
+/**
  * amdgpu_bo_gpu_offset_no_check - return GPU offset of bo
  * @bo:	amdgpu object for which we query the offset
  *
@@ -1505,15 +1508,60 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
 u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
-	uint64_t offset;
+	uint64_t offset = AMDGPU_BO_INVALID_OFFSET;
 
-	offset = (bo->tbo.resource->start << PAGE_SHIFT) +
-		 amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
+	if (bo->tbo.resource->mem_type == TTM_PL_TT)
+		offset = amdgpu_gmc_agp_addr(&bo->tbo);
+
+	if (offset == AMDGPU_BO_INVALID_OFFSET)
+		offset = (bo->tbo.resource->start << PAGE_SHIFT) +
+			amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type);
 
 	return amdgpu_gmc_sign_extend(offset);
 }
 
 /**
+ * amdgpu_bo_mem_stats_placement - bo placement for memory accounting
+ * @bo:	the buffer object we should look at
+ *
+ * BO can have multiple preferred placements, to avoid double counting we want
+ * to file it under a single placement for memory stats.
+ * Luckily, if we take the highest set bit in preferred_domains the result is
+ * quite sensible.
+ *
+ * Returns:
+ * Which of the placements should the BO be accounted under.
+ */
+uint32_t amdgpu_bo_mem_stats_placement(struct amdgpu_bo *bo)
+{
+	uint32_t domain = bo->preferred_domains & AMDGPU_GEM_DOMAIN_MASK;
+
+	if (!domain)
+		return TTM_PL_SYSTEM;
+
+	switch (rounddown_pow_of_two(domain)) {
+	case AMDGPU_GEM_DOMAIN_CPU:
+		return TTM_PL_SYSTEM;
+	case AMDGPU_GEM_DOMAIN_GTT:
+		return TTM_PL_TT;
+	case AMDGPU_GEM_DOMAIN_VRAM:
+		return TTM_PL_VRAM;
+	case AMDGPU_GEM_DOMAIN_GDS:
+		return AMDGPU_PL_GDS;
+	case AMDGPU_GEM_DOMAIN_GWS:
+		return AMDGPU_PL_GWS;
+	case AMDGPU_GEM_DOMAIN_OA:
+		return AMDGPU_PL_OA;
+	case AMDGPU_GEM_DOMAIN_DOORBELL:
+		return AMDGPU_PL_DOORBELL;
+	case AMDGPU_GEM_DOMAIN_MMIO_REMAP:
+		return AMDGPU_PL_MMIO_REMAP;
+	default:
+		return TTM_PL_SYSTEM;
+	}
+}
+
+/**
  * amdgpu_bo_get_preferred_domain - get preferred domain
  * @adev: amdgpu device object
  * @domain: allowed :ref:`memory domains <amdgpu_memory_domains>`
@@ -1555,25 +1603,54 @@ uint32_t amdgpu_bo_get_preferred_domain(struct amdgpu_device *adev,
  */
 u64 amdgpu_bo_print_info(int id, struct amdgpu_bo *bo, struct seq_file *m)
 {
+	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
 	struct dma_buf_attachment *attachment;
 	struct dma_buf *dma_buf;
-	unsigned int domain;
 	const char *placement;
 	unsigned int pin_count;
 	u64 size;
 
-	domain = amdgpu_mem_type_to_domain(bo->tbo.resource->mem_type);
-	switch (domain) {
-	case AMDGPU_GEM_DOMAIN_VRAM:
-		placement = "VRAM";
-		break;
-	case AMDGPU_GEM_DOMAIN_GTT:
-		placement = " GTT";
-		break;
-	case AMDGPU_GEM_DOMAIN_CPU:
-	default:
-		placement = " CPU";
-		break;
+	if (dma_resv_trylock(bo->tbo.base.resv)) {
+		if (!bo->tbo.resource) {
+			placement = "NONE";
+		} else {
+			switch (bo->tbo.resource->mem_type) {
+			case TTM_PL_VRAM:
+				if (amdgpu_res_cpu_visible(adev, bo->tbo.resource))
+					placement = "VRAM VISIBLE";
+				else
+					placement = "VRAM";
+				break;
+			case TTM_PL_TT:
+				placement = "GTT";
+				break;
+			case AMDGPU_PL_GDS:
+				placement = "GDS";
+				break;
+			case AMDGPU_PL_GWS:
+				placement = "GWS";
+				break;
+			case AMDGPU_PL_OA:
+				placement = "OA";
+				break;
+			case AMDGPU_PL_PREEMPT:
+				placement = "PREEMPTIBLE";
+				break;
+			case AMDGPU_PL_DOORBELL:
+				placement = "DOORBELL";
+				break;
+			case AMDGPU_PL_MMIO_REMAP:
+				placement = "MMIO REMAP";
+				break;
+			case TTM_PL_SYSTEM:
+			default:
+				placement = "CPU";
+				break;
+			}
+		}
+		dma_resv_unlock(bo->tbo.base.resv);
+	} else {
+		placement = "UNKNOWN";
 	}
 
 	size = amdgpu_bo_size(bo);
@@ -1599,7 +1676,11 @@ u64 amdgpu_bo_print_info(int id, struct amdgpu_bo *bo, struct seq_file *m)
 	amdgpu_bo_print_flag(m, bo, VRAM_CONTIGUOUS);
 	amdgpu_bo_print_flag(m, bo, VM_ALWAYS_VALID);
 	amdgpu_bo_print_flag(m, bo, EXPLICIT_SYNC);
-
+	/* Add the gem obj resv fence dump*/
+	if (dma_resv_trylock(bo->tbo.base.resv)) {
+		dma_resv_describe(bo->tbo.base.resv, m);
+		dma_resv_unlock(bo->tbo.base.resv);
+	}
 	seq_puts(m, "\n");
 
 	return size;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
index 35b8106816a1..52c2d1731aab 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
@@ -56,6 +56,8 @@ struct amdgpu_bo_param {
 	bool				no_wait_gpu;
 	struct dma_resv			*resv;
 	void				(*destroy)(struct ttm_buffer_object *bo);
+	/* xcp partition number plus 1, 0 means any partition */
+	int8_t				xcp_id_plus1;
 };
 
 /* bo virtual addresses in a vm */
@@ -67,7 +69,7 @@ struct amdgpu_bo_va_mapping {
 	uint64_t			last;
 	uint64_t			__subtree_last;
 	uint64_t			offset;
-	uint64_t			flags;
+	uint32_t			flags;
 };
 
 /* User space allocated BO in a VM */
@@ -88,6 +90,13 @@ struct amdgpu_bo_va {
 	bool				cleared;
 
 	bool				is_xgmi;
+
+	/*
+	 * protected by vm reservation lock
+	 * if non-zero, cannot unmap from GPU because user queues may still access it
+	 */
+	unsigned int			queue_refcount;
+	atomic_t			userq_va_mapped;
 };
 
 struct amdgpu_bo {
@@ -108,6 +117,13 @@ struct amdgpu_bo {
 	struct mmu_interval_notifier	notifier;
 #endif
 	struct kgd_mem                  *kfd_bo;
+
+	/*
+	 * For GPUs with spatial partitioning, xcp partition number, -1 means
+	 * any partition. For other ASICs without spatial partition, always 0
+	 * for memory accounting.
+	 */
+	int8_t				xcp_id;
 };
 
 struct amdgpu_bo_user {
@@ -121,32 +137,9 @@ struct amdgpu_bo_user {
 
 struct amdgpu_bo_vm {
 	struct amdgpu_bo		bo;
-	struct amdgpu_bo		*shadow;
-	struct list_head		shadow_list;
 	struct amdgpu_vm_bo_base        entries[];
 };
 
-struct amdgpu_mem_stats {
-	/* current VRAM usage, includes visible VRAM */
-	uint64_t vram;
-	/* current visible VRAM usage */
-	uint64_t visible_vram;
-	/* current GTT usage */
-	uint64_t gtt;
-	/* current system memory usage */
-	uint64_t cpu;
-	/* sum of evicted buffers, includes visible VRAM */
-	uint64_t evicted_vram;
-	/* sum of evicted buffers due to CPU access */
-	uint64_t evicted_visible_vram;
-	/* how much userspace asked for, includes vis.VRAM */
-	uint64_t requested_vram;
-	/* how much userspace asked for */
-	uint64_t requested_visible_vram;
-	/* how much userspace asked for */
-	uint64_t requested_gtt;
-};
-
 static inline struct amdgpu_bo *ttm_to_amdgpu_bo(struct ttm_buffer_object *tbo)
 {
 	return container_of(tbo, struct amdgpu_bo, tbo);
@@ -173,6 +166,10 @@ static inline unsigned amdgpu_mem_type_to_domain(u32 mem_type)
 		return AMDGPU_GEM_DOMAIN_GWS;
 	case AMDGPU_PL_OA:
 		return AMDGPU_GEM_DOMAIN_OA;
+	case AMDGPU_PL_DOORBELL:
+		return AMDGPU_GEM_DOMAIN_DOORBELL;
+	case AMDGPU_PL_MMIO_REMAP:
+		return AMDGPU_GEM_DOMAIN_MMIO_REMAP;
 	default:
 		break;
 	}
@@ -234,28 +231,6 @@ static inline u64 amdgpu_bo_mmap_offset(struct amdgpu_bo *bo)
 }
 
 /**
- * amdgpu_bo_in_cpu_visible_vram - check if BO is (partly) in visible VRAM
- */
-static inline bool amdgpu_bo_in_cpu_visible_vram(struct amdgpu_bo *bo)
-{
-	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
-	struct amdgpu_res_cursor cursor;
-
-	if (bo->tbo.resource->mem_type != TTM_PL_VRAM)
-		return false;
-
-	amdgpu_res_first(bo->tbo.resource, 0, amdgpu_bo_size(bo), &cursor);
-	while (cursor.remaining) {
-		if (cursor.start < adev->gmc.visible_vram_size)
-			return true;
-
-		amdgpu_res_next(&cursor, cursor.size);
-	}
-
-	return false;
-}
-
-/**
  * amdgpu_bo_explicit_sync - return whether the bo is explicitly synced
  */
 static inline bool amdgpu_bo_explicit_sync(struct amdgpu_bo *bo)
@@ -274,22 +249,6 @@ static inline bool amdgpu_bo_encrypted(struct amdgpu_bo *bo)
 	return bo->flags & AMDGPU_GEM_CREATE_ENCRYPTED;
 }
 
-/**
- * amdgpu_bo_shadowed - check if the BO is shadowed
- *
- * @bo: BO to be tested.
- *
- * Returns:
- * NULL if not shadowed or else return a BO pointer.
- */
-static inline struct amdgpu_bo *amdgpu_bo_shadowed(struct amdgpu_bo *bo)
-{
-	if (bo->tbo.type == ttm_bo_type_kernel)
-		return to_amdgpu_bo_vm(bo)->shadow;
-
-	return NULL;
-}
-
 bool amdgpu_bo_is_amdgpu_bo(struct ttm_buffer_object *bo);
 void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain);
 
@@ -304,6 +263,10 @@ int amdgpu_bo_create_kernel(struct amdgpu_device *adev,
 			    unsigned long size, int align,
 			    u32 domain, struct amdgpu_bo **bo_ptr,
 			    u64 *gpu_addr, void **cpu_addr);
+int amdgpu_bo_create_isp_user(struct amdgpu_device *adev,
+			   struct dma_buf *dbuf, u32 domain,
+			   struct amdgpu_bo **bo,
+			   u64 *gpu_addr);
 int amdgpu_bo_create_kernel_at(struct amdgpu_device *adev,
 			       uint64_t offset, uint64_t size,
 			       struct amdgpu_bo **bo_ptr, void **cpu_addr);
@@ -315,14 +278,13 @@ int amdgpu_bo_create_vm(struct amdgpu_device *adev,
 			struct amdgpu_bo_vm **ubo_ptr);
 void amdgpu_bo_free_kernel(struct amdgpu_bo **bo, u64 *gpu_addr,
 			   void **cpu_addr);
+void amdgpu_bo_free_isp_user(struct amdgpu_bo *bo);
 int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr);
 void *amdgpu_bo_kptr(struct amdgpu_bo *bo);
 void amdgpu_bo_kunmap(struct amdgpu_bo *bo);
 struct amdgpu_bo *amdgpu_bo_ref(struct amdgpu_bo *bo);
 void amdgpu_bo_unref(struct amdgpu_bo **bo);
 int amdgpu_bo_pin(struct amdgpu_bo *bo, u32 domain);
-int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
-			     u64 min_offset, u64 max_offset);
 void amdgpu_bo_unpin(struct amdgpu_bo *bo);
 int amdgpu_bo_init(struct amdgpu_device *adev);
 void amdgpu_bo_fini(struct amdgpu_device *adev);
@@ -345,12 +307,9 @@ int amdgpu_bo_sync_wait_resv(struct amdgpu_device *adev, struct dma_resv *resv,
 			     bool intr);
 int amdgpu_bo_sync_wait(struct amdgpu_bo *bo, void *owner, bool intr);
 u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo);
+u64 amdgpu_bo_fb_aper_addr(struct amdgpu_bo *bo);
 u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo);
-void amdgpu_bo_get_memory(struct amdgpu_bo *bo,
-			  struct amdgpu_mem_stats *stats);
-void amdgpu_bo_add_to_shadow_list(struct amdgpu_bo_vm *vmbo);
-int amdgpu_bo_restore_shadow(struct amdgpu_bo *shadow,
-			     struct dma_fence **fence);
+uint32_t amdgpu_bo_mem_stats_placement(struct amdgpu_bo *bo);
 uint32_t amdgpu_bo_get_preferred_domain(struct amdgpu_device *adev,
 					    uint32_t domain);
 
@@ -385,8 +344,7 @@ int amdgpu_sa_bo_manager_start(struct amdgpu_device *adev,
 int amdgpu_sa_bo_new(struct amdgpu_sa_manager *sa_manager,
 		     struct drm_suballoc **sa_bo,
 		     unsigned int size);
-void amdgpu_sa_bo_free(struct amdgpu_device *adev,
-		       struct drm_suballoc **sa_bo,
+void amdgpu_sa_bo_free(struct drm_suballoc **sa_bo,
 		       struct dma_fence *fence);
 #if defined(CONFIG_DEBUG_FS)
 void amdgpu_sa_bo_dump_debug_info(struct amdgpu_sa_manager *sa_manager,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pll.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pll.c
index 0bb2466d539a..675aa138ea11 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pll.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pll.c
@@ -94,7 +94,7 @@ static void amdgpu_pll_get_fb_ref_div(struct amdgpu_device *adev, unsigned int n
 		ref_div_max = min(128 / post_div, ref_div_max);
 
 	/* get matching reference and feedback divider */
-	*ref_div = min(max(DIV_ROUND_CLOSEST(den, post_div), 1u), ref_div_max);
+	*ref_div = clamp(DIV_ROUND_CLOSEST(den, post_div), 1u, ref_div_max);
 	*fb_div = DIV_ROUND_CLOSEST(nom * *ref_div * post_div, den);
 
 	/* limit fb divider to its maximum */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c
index 71ee361d0972..6e91ea1de5aa 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pmu.c
@@ -276,9 +276,8 @@ static void amdgpu_perf_read(struct perf_event *event)
 	    (!pe->adev->df.funcs->pmc_get_count))
 		return;
 
+	prev = local64_read(&hwc->prev_count);
 	do {
-		prev = local64_read(&hwc->prev_count);
-
 		switch (hwc->config_base) {
 		case AMDGPU_PMU_EVENT_CONFIG_TYPE_DF:
 		case AMDGPU_PMU_EVENT_CONFIG_TYPE_XGMI:
@@ -289,7 +288,7 @@ static void amdgpu_perf_read(struct perf_event *event)
 			count = 0;
 			break;
 		}
-	} while (local64_cmpxchg(&hwc->prev_count, prev, count) != prev);
+	} while (!local64_try_cmpxchg(&hwc->prev_count, &prev, count));
 
 	local64_add(count - prev, &event->count);
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_preempt_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_preempt_mgr.c
index e8adfd0a570a..34b5e22b44e5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_preempt_mgr.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_preempt_mgr.c
@@ -137,7 +137,8 @@ void amdgpu_preempt_mgr_fini(struct amdgpu_device *adev)
 	if (ret)
 		return;
 
-	device_remove_file(adev->dev, &dev_attr_mem_info_preempt_used);
+	if (adev->dev->kobj.sd)
+		device_remove_file(adev->dev, &dev_attr_mem_info_preempt_used);
 
 	ttm_resource_manager_cleanup(man);
 	ttm_set_driver_manager(&adev->mman.bdev, AMDGPU_PL_PREEMPT, NULL);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
index 9d7e6e0e73ed..0b10497d487c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
@@ -38,15 +38,13 @@
 #include "psp_v12_0.h"
 #include "psp_v13_0.h"
 #include "psp_v13_0_4.h"
+#include "psp_v14_0.h"
 
 #include "amdgpu_ras.h"
 #include "amdgpu_securedisplay.h"
 #include "amdgpu_atomfirmware.h"
 
-#define AMD_VBIOS_FILE_MAX_SIZE_B      (1024*1024*3)
-
-static int psp_sysfs_init(struct amdgpu_device *adev);
-static void psp_sysfs_fini(struct amdgpu_device *adev);
+#define AMD_VBIOS_FILE_MAX_SIZE_B      (1024*1024*16)
 
 static int psp_load_smu_fw(struct psp_context *psp);
 static int psp_rap_terminate(struct psp_context *psp);
@@ -103,7 +101,7 @@ static void psp_check_pmfw_centralized_cstate_management(struct psp_context *psp
 		return;
 	}
 
-	switch (adev->ip_versions[MP0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
 	case IP_VERSION(11, 0, 4):
 	case IP_VERSION(11, 0, 5):
@@ -131,7 +129,7 @@ static int psp_init_sriov_microcode(struct psp_context *psp)
 
 	amdgpu_ucode_ip_version_decode(adev, MP0_HWIP, ucode_prefix, sizeof(ucode_prefix));
 
-	switch (adev->ip_versions[MP0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 	case IP_VERSION(9, 0, 0):
 	case IP_VERSION(11, 0, 7):
 	case IP_VERSION(11, 0, 9):
@@ -146,78 +144,120 @@ static int psp_init_sriov_microcode(struct psp_context *psp)
 	case IP_VERSION(13, 0, 0):
 		adev->virt.autoload_ucode_id = 0;
 		break;
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 14):
+		ret = psp_init_cap_microcode(psp, ucode_prefix);
+		ret &= psp_init_ta_microcode(psp, ucode_prefix);
+		break;
 	case IP_VERSION(13, 0, 10):
 		adev->virt.autoload_ucode_id = AMDGPU_UCODE_ID_CP_MES1_DATA;
 		ret = psp_init_cap_microcode(psp, ucode_prefix);
 		break;
+	case IP_VERSION(13, 0, 12):
+		ret = psp_init_ta_microcode(psp, ucode_prefix);
+		break;
 	default:
 		return -EINVAL;
 	}
 	return ret;
 }
 
-static int psp_early_init(void *handle)
+static int psp_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct psp_context *psp = &adev->psp;
 
-	switch (adev->ip_versions[MP0_HWIP][0]) {
+	psp->autoload_supported = true;
+	psp->boot_time_tmr = true;
+
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 	case IP_VERSION(9, 0, 0):
 		psp_v3_1_set_psp_funcs(psp);
 		psp->autoload_supported = false;
+		psp->boot_time_tmr = false;
 		break;
 	case IP_VERSION(10, 0, 0):
 	case IP_VERSION(10, 0, 1):
 		psp_v10_0_set_psp_funcs(psp);
 		psp->autoload_supported = false;
+		psp->boot_time_tmr = false;
 		break;
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 4):
 		psp_v11_0_set_psp_funcs(psp);
 		psp->autoload_supported = false;
+		psp->boot_time_tmr = false;
 		break;
 	case IP_VERSION(11, 0, 0):
+	case IP_VERSION(11, 0, 7):
+		adev->psp.sup_pd_fw_up = !amdgpu_sriov_vf(adev);
+		fallthrough;
 	case IP_VERSION(11, 0, 5):
 	case IP_VERSION(11, 0, 9):
-	case IP_VERSION(11, 0, 7):
 	case IP_VERSION(11, 0, 11):
 	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 2):
 	case IP_VERSION(11, 0, 12):
 	case IP_VERSION(11, 0, 13):
 		psp_v11_0_set_psp_funcs(psp);
-		psp->autoload_supported = true;
+		psp->boot_time_tmr = false;
 		break;
 	case IP_VERSION(11, 0, 3):
 	case IP_VERSION(12, 0, 1):
 		psp_v12_0_set_psp_funcs(psp);
+		psp->autoload_supported = false;
+		psp->boot_time_tmr = false;
 		break;
 	case IP_VERSION(13, 0, 2):
+		psp->boot_time_tmr = false;
+		fallthrough;
 	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 14):
+		psp_v13_0_set_psp_funcs(psp);
+		psp->autoload_supported = false;
+		break;
+	case IP_VERSION(13, 0, 12):
 		psp_v13_0_set_psp_funcs(psp);
+		psp->autoload_supported = false;
+		adev->psp.sup_ifwi_up = !amdgpu_sriov_vf(adev);
 		break;
 	case IP_VERSION(13, 0, 1):
 	case IP_VERSION(13, 0, 3):
 	case IP_VERSION(13, 0, 5):
 	case IP_VERSION(13, 0, 8):
-	case IP_VERSION(13, 0, 10):
 	case IP_VERSION(13, 0, 11):
+	case IP_VERSION(14, 0, 0):
+	case IP_VERSION(14, 0, 1):
+	case IP_VERSION(14, 0, 4):
 		psp_v13_0_set_psp_funcs(psp);
-		psp->autoload_supported = true;
+		psp->boot_time_tmr = false;
 		break;
 	case IP_VERSION(11, 0, 8):
 		if (adev->apu_flags & AMD_APU_IS_CYAN_SKILLFISH2) {
 			psp_v11_0_8_set_psp_funcs(psp);
-			psp->autoload_supported = false;
 		}
+		psp->autoload_supported = false;
+		psp->boot_time_tmr = false;
 		break;
 	case IP_VERSION(13, 0, 0):
 	case IP_VERSION(13, 0, 7):
+	case IP_VERSION(13, 0, 10):
 		psp_v13_0_set_psp_funcs(psp);
-		psp->autoload_supported = true;
+		adev->psp.sup_ifwi_up = !amdgpu_sriov_vf(adev);
+		psp->boot_time_tmr = false;
 		break;
 	case IP_VERSION(13, 0, 4):
 		psp_v13_0_4_set_psp_funcs(psp);
-		psp->autoload_supported = true;
+		psp->boot_time_tmr = false;
+		break;
+	case IP_VERSION(14, 0, 2):
+	case IP_VERSION(14, 0, 3):
+		adev->psp.sup_ifwi_up = !amdgpu_sriov_vf(adev);
+		psp_v14_0_set_psp_funcs(psp);
+		break;
+	case IP_VERSION(14, 0, 5):
+		psp_v14_0_set_psp_funcs(psp);
+		psp->boot_time_tmr = false;
 		break;
 	default:
 		return -EINVAL;
@@ -225,6 +265,8 @@ static int psp_early_init(void *handle)
 
 	psp->adev = adev;
 
+	adev->psp_timeout = 20000;
+
 	psp_check_pmfw_centralized_cstate_management(psp);
 
 	if (amdgpu_sriov_vf(adev))
@@ -286,21 +328,22 @@ static int psp_memory_training_init(struct psp_context *psp)
 	struct psp_memory_training_context *ctx = &psp->mem_train_ctx;
 
 	if (ctx->init != PSP_MEM_TRAIN_RESERVE_SUCCESS) {
-		DRM_DEBUG("memory training is not supported!\n");
+		dev_dbg(psp->adev->dev, "memory training is not supported!\n");
 		return 0;
 	}
 
 	ctx->sys_cache = kzalloc(ctx->train_data_size, GFP_KERNEL);
 	if (ctx->sys_cache == NULL) {
-		DRM_ERROR("alloc mem_train_ctx.sys_cache failed!\n");
+		dev_err(psp->adev->dev, "alloc mem_train_ctx.sys_cache failed!\n");
 		ret = -ENOMEM;
 		goto Err_out;
 	}
 
-	DRM_DEBUG("train_data_size:%llx,p2c_train_data_offset:%llx,c2p_train_data_offset:%llx.\n",
-		  ctx->train_data_size,
-		  ctx->p2c_train_data_offset,
-		  ctx->c2p_train_data_offset);
+	dev_dbg(psp->adev->dev,
+		"train_data_size:%llx,p2c_train_data_offset:%llx,c2p_train_data_offset:%llx.\n",
+		ctx->train_data_size,
+		ctx->p2c_train_data_offset,
+		ctx->c2p_train_data_offset);
 	ctx->init = PSP_MEM_TRAIN_INIT_SUCCESS;
 	return 0;
 
@@ -329,6 +372,11 @@ static bool psp_get_runtime_db_entry(struct amdgpu_device *adev,
 	bool ret = false;
 	int i;
 
+	if (amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 6) ||
+	    amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 12) ||
+	    amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 14))
+		return false;
+
 	db_header_pos = adev->gmc.mc_vram_size - PSP_RUNTIME_DB_OFFSET;
 	db_dir_pos = db_header_pos + sizeof(struct psp_runtime_data_header);
 
@@ -388,9 +436,9 @@ static bool psp_get_runtime_db_entry(struct amdgpu_device *adev,
 	return ret;
 }
 
-static int psp_sw_init(void *handle)
+static int psp_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct psp_context *psp = &adev->psp;
 	int ret;
 	struct psp_runtime_boot_cfg_entry boot_cfg_entry;
@@ -399,19 +447,19 @@ static int psp_sw_init(void *handle)
 
 	psp->cmd = kzalloc(sizeof(struct psp_gfx_cmd_resp), GFP_KERNEL);
 	if (!psp->cmd) {
-		DRM_ERROR("Failed to allocate memory to command buffer!\n");
-		ret = -ENOMEM;
+		dev_err(adev->dev, "Failed to allocate memory to command buffer!\n");
+		return -ENOMEM;
 	}
 
 	adev->psp.xgmi_context.supports_extended_data =
 		!adev->gmc.xgmi.connected_to_cpu &&
-			adev->ip_versions[MP0_HWIP][0] == IP_VERSION(13, 0, 2);
+		amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 2);
 
 	memset(&scpm_entry, 0, sizeof(scpm_entry));
 	if ((psp_get_runtime_db_entry(adev,
 				PSP_RUNTIME_ENTRY_TYPE_PPTABLE_ERR_STATUS,
 				&scpm_entry)) &&
-	    (SCPM_DISABLE != scpm_entry.scpm_status)) {
+	    (scpm_entry.scpm_status != SCPM_DISABLE)) {
 		adev->scpm_enabled = true;
 		adev->scpm_status = scpm_entry.scpm_status;
 	} else {
@@ -431,41 +479,34 @@ static int psp_sw_init(void *handle)
 			/* If psp runtime database exists, then
 			 * only enable two stage memory training
 			 * when TWO_STAGE_DRAM_TRAINING bit is set
-			 * in runtime database */
+			 * in runtime database
+			 */
 			mem_training_ctx->enable_mem_training = true;
 		}
 
 	} else {
-		/* If psp runtime database doesn't exist or
-		 * is invalid, force enable two stage memory
-		 * training */
+		/* If psp runtime database doesn't exist or is
+		 * invalid, force enable two stage memory training
+		 */
 		mem_training_ctx->enable_mem_training = true;
 	}
 
 	if (mem_training_ctx->enable_mem_training) {
 		ret = psp_memory_training_init(psp);
 		if (ret) {
-			DRM_ERROR("Failed to initialize memory training!\n");
+			dev_err(adev->dev, "Failed to initialize memory training!\n");
 			return ret;
 		}
 
 		ret = psp_mem_training(psp, PSP_MEM_TRAIN_COLD_BOOT);
 		if (ret) {
-			DRM_ERROR("Failed to process memory training!\n");
-			return ret;
-		}
-	}
-
-	if (adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 0) ||
-	    adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 7)) {
-		ret= psp_sysfs_init(adev);
-		if (ret) {
+			dev_err(adev->dev, "Failed to process memory training!\n");
 			return ret;
 		}
 	}
 
 	ret = amdgpu_bo_create_kernel(adev, PSP_1_MEG, PSP_1_MEG,
-				      amdgpu_sriov_vf(adev) ?
+				      (amdgpu_sriov_vf(adev) || adev->debug_use_vram_fw_buf) ?
 				      AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT,
 				      &psp->fw_pri_bo,
 				      &psp->fw_pri_mc_addr,
@@ -474,7 +515,8 @@ static int psp_sw_init(void *handle)
 		return ret;
 
 	ret = amdgpu_bo_create_kernel(adev, PSP_FENCE_BUFFER_SIZE, PAGE_SIZE,
-				      AMDGPU_GEM_DOMAIN_VRAM,
+				      AMDGPU_GEM_DOMAIN_VRAM |
+				      AMDGPU_GEM_DOMAIN_GTT,
 				      &psp->fence_buf_bo,
 				      &psp->fence_buf_mc_addr,
 				      &psp->fence_buf);
@@ -482,7 +524,8 @@ static int psp_sw_init(void *handle)
 		goto failed1;
 
 	ret = amdgpu_bo_create_kernel(adev, PSP_CMD_BUFFER_SIZE, PAGE_SIZE,
-				      AMDGPU_GEM_DOMAIN_VRAM,
+				      AMDGPU_GEM_DOMAIN_VRAM |
+				      AMDGPU_GEM_DOMAIN_GTT,
 				      &psp->cmd_buf_bo, &psp->cmd_buf_mc_addr,
 				      (void **)&psp->cmd_buf_mem);
 	if (ret)
@@ -491,19 +534,18 @@ static int psp_sw_init(void *handle)
 	return 0;
 
 failed2:
-	amdgpu_bo_free_kernel(&psp->fw_pri_bo,
-			      &psp->fw_pri_mc_addr, &psp->fw_pri_buf);
-failed1:
 	amdgpu_bo_free_kernel(&psp->fence_buf_bo,
 			      &psp->fence_buf_mc_addr, &psp->fence_buf);
+failed1:
+	amdgpu_bo_free_kernel(&psp->fw_pri_bo,
+			      &psp->fw_pri_mc_addr, &psp->fw_pri_buf);
 	return ret;
 }
 
-static int psp_sw_fini(void *handle)
+static int psp_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct psp_context *psp = &adev->psp;
-	struct psp_gfx_cmd_resp *cmd = psp->cmd;
 
 	psp_memory_training_fini(psp);
 
@@ -513,12 +555,10 @@ static int psp_sw_fini(void *handle)
 	amdgpu_ucode_release(&psp->cap_fw);
 	amdgpu_ucode_release(&psp->toc_fw);
 
-	if (adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 0) ||
-	    adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 7))
-		psp_sysfs_fini(adev);
+	kfree(psp->cmd);
+	psp->cmd = NULL;
 
-	kfree(cmd);
-	cmd = NULL;
+	psp_free_shared_bufs(psp);
 
 	if (psp->km_ring.ring_mem)
 		amdgpu_bo_free_kernel(&adev->firmware.rbuf,
@@ -535,9 +575,11 @@ static int psp_sw_fini(void *handle)
 	return 0;
 }
 
-int psp_wait_for(struct psp_context *psp, uint32_t reg_index,
-		 uint32_t reg_val, uint32_t mask, bool check_changed)
+int psp_wait_for(struct psp_context *psp, uint32_t reg_index, uint32_t reg_val,
+		 uint32_t mask, uint32_t flags)
 {
+	bool check_changed = flags & PSP_WAITREG_CHANGED;
+	bool verbose = !(flags & PSP_WAITREG_NOVERBOSE);
 	uint32_t val;
 	int i;
 	struct amdgpu_device *adev = psp->adev;
@@ -557,6 +599,31 @@ int psp_wait_for(struct psp_context *psp, uint32_t reg_index,
 		udelay(1);
 	}
 
+	if (verbose)
+		dev_err(adev->dev,
+			"psp reg (0x%x) wait timed out, mask: %x, read: %x exp: %x",
+			reg_index, mask, val, reg_val);
+
+	return -ETIME;
+}
+
+int psp_wait_for_spirom_update(struct psp_context *psp, uint32_t reg_index,
+			       uint32_t reg_val, uint32_t mask, uint32_t msec_timeout)
+{
+	uint32_t val;
+	int i;
+	struct amdgpu_device *adev = psp->adev;
+
+	if (psp->adev->no_hw_access)
+		return 0;
+
+	for (i = 0; i < msec_timeout; i++) {
+		val = RREG32(reg_index);
+		if ((val & mask) == reg_val)
+			return 0;
+		msleep(1);
+	}
+
 	return -ETIME;
 }
 
@@ -593,11 +660,35 @@ static const char *psp_gfx_cmd_name(enum psp_gfx_cmd_id cmd_id)
 		return "AUTOLOAD_RLC";
 	case GFX_CMD_ID_BOOT_CFG:
 		return "BOOT_CFG";
+	case GFX_CMD_ID_CONFIG_SQ_PERFMON:
+		return "CONFIG_SQ_PERFMON";
+	case GFX_CMD_ID_FB_FW_RESERV_ADDR:
+		return "FB_FW_RESERV_ADDR";
+	case GFX_CMD_ID_FB_FW_RESERV_EXT_ADDR:
+		return "FB_FW_RESERV_EXT_ADDR";
+	case GFX_CMD_ID_SRIOV_SPATIAL_PART:
+		return "SPATIAL_PARTITION";
+	case GFX_CMD_ID_FB_NPS_MODE:
+		return "NPS_MODE_CHANGE";
 	default:
 		return "UNKNOWN CMD";
 	}
 }
 
+static bool psp_err_warn(struct psp_context *psp)
+{
+	struct psp_gfx_cmd_resp *cmd = psp->cmd_buf_mem;
+
+	/* This response indicates reg list is already loaded */
+	if (amdgpu_ip_version(psp->adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 2) &&
+	    cmd->cmd_id == GFX_CMD_ID_LOAD_IP_FW &&
+	    cmd->cmd.cmd_load_ip_fw.fw_type == GFX_FW_TYPE_REG_LIST &&
+	    cmd->resp.status == TEE_ERROR_CANCEL)
+		return false;
+
+	return true;
+}
+
 static int
 psp_cmd_submit_buf(struct psp_context *psp,
 		   struct amdgpu_firmware_info *ucode,
@@ -605,7 +696,7 @@ psp_cmd_submit_buf(struct psp_context *psp,
 {
 	int ret;
 	int index;
-	int timeout = 20000;
+	int timeout = psp->adev->psp_timeout;
 	bool ras_intr = false;
 	bool skip_unsupport = false;
 
@@ -643,7 +734,7 @@ psp_cmd_submit_buf(struct psp_context *psp,
 	skip_unsupport = (psp->cmd_buf_mem->resp.status == TEE_ERROR_NOT_SUPPORTED ||
 		psp->cmd_buf_mem->resp.status == PSP_ERR_UNKNOWN_COMMAND) && amdgpu_sriov_vf(psp->adev);
 
-	memcpy((void*)&cmd->resp, (void*)&psp->cmd_buf_mem->resp, sizeof(struct psp_gfx_resp));
+	memcpy(&cmd->resp, &psp->cmd_buf_mem->resp, sizeof(struct psp_gfx_resp));
 
 	/* In some cases, psp response status is not 0 even there is no
 	 * problem while the command is submitted. Some version of PSP FW
@@ -654,11 +745,16 @@ psp_cmd_submit_buf(struct psp_context *psp,
 	 */
 	if (!skip_unsupport && (psp->cmd_buf_mem->resp.status || !timeout) && !ras_intr) {
 		if (ucode)
-			DRM_WARN("failed to load ucode %s(0x%X) ",
-				  amdgpu_ucode_name(ucode->ucode_id), ucode->ucode_id);
-		DRM_WARN("psp gfx command %s(0x%X) failed and response status is (0x%X)\n",
-			 psp_gfx_cmd_name(psp->cmd_buf_mem->cmd_id), psp->cmd_buf_mem->cmd_id,
-			 psp->cmd_buf_mem->resp.status);
+			dev_warn(psp->adev->dev,
+				 "failed to load ucode %s(0x%X) ",
+				 amdgpu_ucode_name(ucode->ucode_id), ucode->ucode_id);
+		if (psp_err_warn(psp))
+			dev_warn(
+				psp->adev->dev,
+				"psp gfx command %s(0x%X) failed and response status is (0x%X)\n",
+				psp_gfx_cmd_name(psp->cmd_buf_mem->cmd_id),
+				psp->cmd_buf_mem->cmd_id,
+				psp->cmd_buf_mem->resp.status);
 		/* If any firmware (including CAP) load fails under SRIOV, it should
 		 * return failure to stop the VF from initializing.
 		 * Also return failure in case of timeout
@@ -699,8 +795,13 @@ static void psp_prep_tmr_cmd_buf(struct psp_context *psp,
 				 uint64_t tmr_mc, struct amdgpu_bo *tmr_bo)
 {
 	struct amdgpu_device *adev = psp->adev;
-	uint32_t size = amdgpu_bo_size(tmr_bo);
-	uint64_t tmr_pa = amdgpu_gmc_vram_pa(adev, tmr_bo);
+	uint32_t size = 0;
+	uint64_t tmr_pa = 0;
+
+	if (tmr_bo) {
+		size = amdgpu_bo_size(tmr_bo);
+		tmr_pa = amdgpu_gmc_vram_pa(adev, tmr_bo);
+	}
 
 	if (amdgpu_sriov_vf(psp->adev))
 		cmd->cmd_id = GFX_CMD_ID_SETUP_VMR;
@@ -763,39 +864,43 @@ static int psp_tmr_init(struct psp_context *psp)
 	tmr_size = PSP_TMR_SIZE(psp->adev);
 
 	/* For ASICs support RLC autoload, psp will parse the toc
-	 * and calculate the total size of TMR needed */
+	 * and calculate the total size of TMR needed
+	 */
 	if (!amdgpu_sriov_vf(psp->adev) &&
 	    psp->toc.start_addr &&
 	    psp->toc.size_bytes &&
 	    psp->fw_pri_buf) {
 		ret = psp_load_toc(psp, &tmr_size);
 		if (ret) {
-			DRM_ERROR("Failed to load toc\n");
+			dev_err(psp->adev->dev, "Failed to load toc\n");
 			return ret;
 		}
 	}
 
-	if (!psp->tmr_bo) {
+	if (!psp->tmr_bo && !psp->boot_time_tmr) {
 		pptr = amdgpu_sriov_vf(psp->adev) ? &tmr_buf : NULL;
 		ret = amdgpu_bo_create_kernel(psp->adev, tmr_size,
 					      PSP_TMR_ALIGNMENT,
-					      AMDGPU_HAS_VRAM(psp->adev) ?
-					      AMDGPU_GEM_DOMAIN_VRAM :
-					      AMDGPU_GEM_DOMAIN_GTT,
+					      AMDGPU_GEM_DOMAIN_GTT | AMDGPU_GEM_DOMAIN_VRAM,
 					      &psp->tmr_bo, &psp->tmr_mc_addr,
 					      pptr);
 	}
+	if (amdgpu_virt_xgmi_migrate_enabled(psp->adev) && psp->tmr_bo)
+		psp->tmr_mc_addr = amdgpu_bo_fb_aper_addr(psp->tmr_bo);
 
 	return ret;
 }
 
 static bool psp_skip_tmr(struct psp_context *psp)
 {
-	switch (psp->adev->ip_versions[MP0_HWIP][0]) {
+	switch (amdgpu_ip_version(psp->adev, MP0_HWIP, 0)) {
 	case IP_VERSION(11, 0, 9):
 	case IP_VERSION(11, 0, 7):
 	case IP_VERSION(13, 0, 2):
+	case IP_VERSION(13, 0, 6):
 	case IP_VERSION(13, 0, 10):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
 		return true;
 	default:
 		return false;
@@ -816,8 +921,9 @@ static int psp_tmr_load(struct psp_context *psp)
 	cmd = acquire_psp_cmd_buf(psp);
 
 	psp_prep_tmr_cmd_buf(psp, cmd, psp->tmr_mc_addr, psp->tmr_bo);
-	DRM_INFO("reserve 0x%lx from 0x%llx for PSP TMR\n",
-		 amdgpu_bo_size(psp->tmr_bo), psp->tmr_mc_addr);
+	if (psp->tmr_bo)
+		dev_info(psp->adev->dev, "reserve 0x%lx from 0x%llx for PSP TMR\n",
+			 amdgpu_bo_size(psp->tmr_bo), psp->tmr_mc_addr);
 
 	ret = psp_cmd_submit_buf(psp, NULL, cmd,
 				 psp->fence_buf_mc_addr);
@@ -828,7 +934,7 @@ static int psp_tmr_load(struct psp_context *psp)
 }
 
 static void psp_prep_tmr_unload_cmd_buf(struct psp_context *psp,
-				        struct psp_gfx_cmd_resp *cmd)
+					struct psp_gfx_cmd_resp *cmd)
 {
 	if (amdgpu_sriov_vf(psp->adev))
 		cmd->cmd_id = GFX_CMD_ID_DESTROY_VMR;
@@ -894,6 +1000,106 @@ int psp_get_fw_attestation_records_addr(struct psp_context *psp,
 	return ret;
 }
 
+static int psp_get_fw_reservation_info(struct psp_context *psp,
+						   uint32_t cmd_id,
+						   uint64_t *addr,
+						   uint32_t *size)
+{
+	int ret;
+	uint32_t status;
+	struct psp_gfx_cmd_resp *cmd;
+
+	cmd = acquire_psp_cmd_buf(psp);
+
+	cmd->cmd_id = cmd_id;
+
+	ret = psp_cmd_submit_buf(psp, NULL, cmd,
+				 psp->fence_buf_mc_addr);
+	if (ret) {
+		release_psp_cmd_buf(psp);
+		return ret;
+	}
+
+	status = cmd->resp.status;
+	if (status == PSP_ERR_UNKNOWN_COMMAND) {
+		release_psp_cmd_buf(psp);
+		*addr = 0;
+		*size = 0;
+		return 0;
+	}
+
+	*addr = (uint64_t)cmd->resp.uresp.fw_reserve_info.reserve_base_address_hi << 32 |
+		cmd->resp.uresp.fw_reserve_info.reserve_base_address_lo;
+	*size = cmd->resp.uresp.fw_reserve_info.reserve_size;
+
+	release_psp_cmd_buf(psp);
+
+	return 0;
+}
+
+int psp_update_fw_reservation(struct psp_context *psp)
+{
+	int ret;
+	uint64_t reserv_addr, reserv_addr_ext;
+	uint32_t reserv_size, reserv_size_ext, mp0_ip_ver;
+	struct amdgpu_device *adev = psp->adev;
+
+	mp0_ip_ver = amdgpu_ip_version(adev, MP0_HWIP, 0);
+
+	if (amdgpu_sriov_vf(psp->adev))
+		return 0;
+
+	switch (mp0_ip_ver) {
+	case IP_VERSION(14, 0, 2):
+		if (adev->psp.sos.fw_version < 0x3b0e0d)
+			return 0;
+		break;
+
+	case IP_VERSION(14, 0, 3):
+		if (adev->psp.sos.fw_version < 0x3a0e14)
+			return 0;
+		break;
+
+	default:
+		return 0;
+	}
+
+	ret = psp_get_fw_reservation_info(psp, GFX_CMD_ID_FB_FW_RESERV_ADDR, &reserv_addr, &reserv_size);
+	if (ret)
+		return ret;
+	ret = psp_get_fw_reservation_info(psp, GFX_CMD_ID_FB_FW_RESERV_EXT_ADDR, &reserv_addr_ext, &reserv_size_ext);
+	if (ret)
+		return ret;
+
+	if (reserv_addr != adev->gmc.real_vram_size - reserv_size) {
+		dev_warn(adev->dev, "reserve fw region is not valid!\n");
+		return 0;
+	}
+
+	amdgpu_bo_free_kernel(&adev->mman.fw_reserved_memory, NULL, NULL);
+
+	reserv_size = roundup(reserv_size, SZ_1M);
+
+	ret = amdgpu_bo_create_kernel_at(adev, reserv_addr, reserv_size, &adev->mman.fw_reserved_memory, NULL);
+	if (ret) {
+		dev_err(adev->dev, "reserve fw region failed(%d)!\n", ret);
+		amdgpu_bo_free_kernel(&adev->mman.fw_reserved_memory, NULL, NULL);
+		return ret;
+	}
+
+	reserv_size_ext = roundup(reserv_size_ext, SZ_1M);
+
+	ret = amdgpu_bo_create_kernel_at(adev, reserv_addr_ext, reserv_size_ext,
+					 &adev->mman.fw_reserved_memory_extend, NULL);
+	if (ret) {
+		dev_err(adev->dev, "reserve extend fw region failed(%d)!\n", ret);
+		amdgpu_bo_free_kernel(&adev->mman.fw_reserved_memory_extend, NULL, NULL);
+		return ret;
+	}
+
+	return 0;
+}
+
 static int psp_boot_config_get(struct amdgpu_device *adev, uint32_t *boot_cfg)
 {
 	struct psp_context *psp = &adev->psp;
@@ -969,6 +1175,52 @@ static int psp_rl_load(struct amdgpu_device *adev)
 	return ret;
 }
 
+int psp_memory_partition(struct psp_context *psp, int mode)
+{
+	struct psp_gfx_cmd_resp *cmd;
+	int ret;
+
+	if (amdgpu_sriov_vf(psp->adev))
+		return 0;
+
+	cmd = acquire_psp_cmd_buf(psp);
+
+	cmd->cmd_id = GFX_CMD_ID_FB_NPS_MODE;
+	cmd->cmd.cmd_memory_part.mode = mode;
+
+	dev_info(psp->adev->dev,
+		 "Requesting %d memory partition change through PSP", mode);
+	ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr);
+	if (ret)
+		dev_err(psp->adev->dev,
+			"PSP request failed to change to NPS%d mode\n", mode);
+
+	release_psp_cmd_buf(psp);
+
+	return ret;
+}
+
+int psp_spatial_partition(struct psp_context *psp, int mode)
+{
+	struct psp_gfx_cmd_resp *cmd;
+	int ret;
+
+	if (amdgpu_sriov_vf(psp->adev))
+		return 0;
+
+	cmd = acquire_psp_cmd_buf(psp);
+
+	cmd->cmd_id = GFX_CMD_ID_SRIOV_SPATIAL_PART;
+	cmd->cmd.cmd_spatial_part.mode = mode;
+
+	dev_info(psp->adev->dev, "Requesting %d partitions through PSP", mode);
+	ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr);
+
+	release_psp_cmd_buf(psp);
+
+	return ret;
+}
+
 static int psp_asd_initialize(struct psp_context *psp)
 {
 	int ret;
@@ -980,6 +1232,11 @@ static int psp_asd_initialize(struct psp_context *psp)
 	if (amdgpu_sriov_vf(psp->adev) || !psp->asd_context.bin_desc.size_bytes)
 		return 0;
 
+	/* bypass asd if display hardware is not available */
+	if (!amdgpu_device_has_display_hardware(psp->adev) &&
+	    amdgpu_ip_version(psp->adev, MP0_HWIP, 0) >= IP_VERSION(13, 0, 10))
+		return 0;
+
 	psp->asd_context.mem_context.shared_mc_addr  = 0;
 	psp->asd_context.mem_context.shared_mem_size = PSP_ASD_SHARED_MEM_SIZE;
 	psp->asd_context.ta_load_type                = GFX_CMD_ID_LOAD_ASD;
@@ -1053,7 +1310,7 @@ int psp_reg_program(struct psp_context *psp, enum psp_reg_prog_id reg,
 	psp_prep_reg_prog_cmd_buf(cmd, reg, value);
 	ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr);
 	if (ret)
-		DRM_ERROR("PSP failed to program reg id %d", reg);
+		dev_err(psp->adev->dev, "PSP failed to program reg id %d\n", reg);
 
 	release_psp_cmd_buf(psp);
 
@@ -1065,7 +1322,7 @@ static void psp_prep_ta_load_cmd_buf(struct psp_gfx_cmd_resp *cmd,
 				     struct ta_context *context)
 {
 	cmd->cmd_id				= context->ta_load_type;
-	cmd->cmd.cmd_load_ta.app_phy_addr_lo 	= lower_32_bits(ta_bin_mc);
+	cmd->cmd.cmd_load_ta.app_phy_addr_lo	= lower_32_bits(ta_bin_mc);
 	cmd->cmd.cmd_load_ta.app_phy_addr_hi	= upper_32_bits(ta_bin_mc);
 	cmd->cmd.cmd_load_ta.app_len		= context->bin_desc.size_bytes;
 
@@ -1080,9 +1337,9 @@ int psp_ta_init_shared_buf(struct psp_context *psp,
 				  struct ta_mem_context *mem_ctx)
 {
 	/*
-	* Allocate 16k memory aligned to 4k from Frame Buffer (local
-	* physical) for ta to host memory
-	*/
+	 * Allocate 16k memory aligned to 4k from Frame Buffer (local
+	 * physical) for ta to host memory
+	 */
 	return amdgpu_bo_create_kernel(psp->adev, mem_ctx->shared_mem_size,
 				      PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM |
 				      AMDGPU_GEM_DOMAIN_GTT,
@@ -1129,6 +1386,11 @@ int psp_ta_load(struct psp_context *psp, struct ta_context *context)
 	psp_copy_fw(psp, context->bin_desc.start_addr,
 		    context->bin_desc.size_bytes);
 
+	if (amdgpu_virt_xgmi_migrate_enabled(psp->adev) &&
+		context->mem_context.shared_bo)
+		context->mem_context.shared_mc_addr =
+			amdgpu_bo_fb_aper_addr(context->mem_context.shared_bo);
+
 	psp_prep_ta_load_cmd_buf(cmd, psp->fw_pri_mc_addr, context);
 
 	ret = psp_cmd_submit_buf(psp, NULL, cmd,
@@ -1136,9 +1398,8 @@ int psp_ta_load(struct psp_context *psp, struct ta_context *context)
 
 	context->resp_status = cmd->resp.status;
 
-	if (!ret) {
+	if (!ret)
 		context->session_id = cmd->resp.session_id;
-	}
 
 	release_psp_cmd_buf(psp);
 
@@ -1156,8 +1417,8 @@ int psp_xgmi_terminate(struct psp_context *psp)
 	struct amdgpu_device *adev = psp->adev;
 
 	/* XGMI TA unload currently is not supported on Arcturus/Aldebaran A+A */
-	if (adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 4) ||
-	    (adev->ip_versions[MP0_HWIP][0] == IP_VERSION(13, 0, 2) &&
+	if (amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(11, 0, 4) ||
+	    (amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 2) &&
 	     adev->gmc.xgmi.connected_to_cpu))
 		return 0;
 
@@ -1208,6 +1469,8 @@ invoke:
 	xgmi_cmd->cmd_id = TA_COMMAND_XGMI__INITIALIZE;
 
 	ret = psp_xgmi_invoke(psp, xgmi_cmd->cmd_id);
+	/* note down the capbility flag for XGMI TA */
+	psp->xgmi_context.xgmi_ta_caps = xgmi_cmd->caps_flag;
 
 	return ret;
 }
@@ -1254,8 +1517,11 @@ int psp_xgmi_get_node_id(struct psp_context *psp, uint64_t *node_id)
 
 static bool psp_xgmi_peer_link_info_supported(struct psp_context *psp)
 {
-	return psp->adev->ip_versions[MP0_HWIP][0] == IP_VERSION(13, 0, 2) &&
-		psp->xgmi_context.context.bin_desc.fw_version >= 0x2000000b;
+	return (amdgpu_ip_version(psp->adev, MP0_HWIP, 0) ==
+			IP_VERSION(13, 0, 2) &&
+		psp->xgmi_context.context.bin_desc.fw_version >= 0x2000000b) ||
+	       amdgpu_ip_version(psp->adev, MP0_HWIP, 0) >=
+		       IP_VERSION(13, 0, 6);
 }
 
 /*
@@ -1273,9 +1539,13 @@ static void psp_xgmi_reflect_topology_info(struct psp_context *psp,
 	uint64_t src_node_id = psp->adev->gmc.xgmi.node_id;
 	uint64_t dst_node_id = node_info.node_id;
 	uint8_t dst_num_hops = node_info.num_hops;
+	uint8_t dst_is_sharing_enabled = node_info.is_sharing_enabled;
 	uint8_t dst_num_links = node_info.num_links;
 
 	hive = amdgpu_get_xgmi_hive(psp->adev);
+	if (WARN_ON(!hive))
+		return;
+
 	list_for_each_entry(mirror_adev, &hive->device_list, gmc.xgmi.head) {
 		struct psp_xgmi_topology_info *mirror_top_info;
 		int j;
@@ -1289,13 +1559,20 @@ static void psp_xgmi_reflect_topology_info(struct psp_context *psp,
 				continue;
 
 			mirror_top_info->nodes[j].num_hops = dst_num_hops;
-			/*
-			 * prevent 0 num_links value re-reflection since reflection
+			mirror_top_info->nodes[j].is_sharing_enabled = dst_is_sharing_enabled;
+			/* prevent 0 num_links value re-reflection since reflection
 			 * criteria is based on num_hops (direct or indirect).
-			 *
 			 */
-			if (dst_num_links)
+			if (dst_num_links) {
 				mirror_top_info->nodes[j].num_links = dst_num_links;
+				/* swap src and dst due to frame of reference */
+				for (int k = 0; k < dst_num_links; k++) {
+					mirror_top_info->nodes[j].port_num[k].src_xgmi_port_num =
+						node_info.port_num[k].dst_xgmi_port_num;
+					mirror_top_info->nodes[j].port_num[k].dst_xgmi_port_num =
+						node_info.port_num[k].src_xgmi_port_num;
+				}
+			}
 
 			break;
 		}
@@ -1326,7 +1603,7 @@ int psp_xgmi_get_topology_info(struct psp_context *psp,
 
 	/* Fill in the shared memory with topology information as input */
 	topology_info_input = &xgmi_cmd->xgmi_in_message.get_topology_info;
-	xgmi_cmd->cmd_id = TA_COMMAND_XGMI__GET_GET_TOPOLOGY_INFO;
+	xgmi_cmd->cmd_id = TA_COMMAND_XGMI__GET_TOPOLOGY_INFO;
 	topology_info_input->num_nodes = number_devices;
 
 	for (i = 0; i < topology_info_input->num_nodes; i++) {
@@ -1337,7 +1614,7 @@ int psp_xgmi_get_topology_info(struct psp_context *psp,
 	}
 
 	/* Invoke xgmi ta to get the topology information */
-	ret = psp_xgmi_invoke(psp, TA_COMMAND_XGMI__GET_GET_TOPOLOGY_INFO);
+	ret = psp_xgmi_invoke(psp, TA_COMMAND_XGMI__GET_TOPOLOGY_INFO);
 	if (ret)
 		return ret;
 
@@ -1362,26 +1639,64 @@ int psp_xgmi_get_topology_info(struct psp_context *psp,
 
 	/* Invoke xgmi ta again to get the link information */
 	if (psp_xgmi_peer_link_info_supported(psp)) {
-		struct ta_xgmi_cmd_get_peer_link_info_output *link_info_output;
+		struct ta_xgmi_cmd_get_peer_link_info *link_info_output;
+		struct ta_xgmi_cmd_get_extend_peer_link_info *link_extend_info_output;
+		bool requires_reflection =
+			(psp->xgmi_context.supports_extended_data &&
+			 get_extended_data) ||
+			amdgpu_ip_version(psp->adev, MP0_HWIP, 0) ==
+				IP_VERSION(13, 0, 6) ||
+			amdgpu_ip_version(psp->adev, MP0_HWIP, 0) ==
+				IP_VERSION(13, 0, 14) ||
+			amdgpu_sriov_vf(psp->adev);
+		bool ta_port_num_support = psp->xgmi_context.xgmi_ta_caps & EXTEND_PEER_LINK_INFO_CMD_FLAG ||
+			amdgpu_sriov_xgmi_ta_ext_peer_link_en(psp->adev);
+
+		/* popluate the shared output buffer rather than the cmd input buffer
+		 * with node_ids as the input for GET_PEER_LINKS command execution.
+		 * This is required for GET_PEER_LINKS per xgmi ta implementation.
+		 * The same requirement for GET_EXTEND_PEER_LINKS command.
+		 */
+		if (ta_port_num_support) {
+			link_extend_info_output = &xgmi_cmd->xgmi_out_message.get_extend_link_info;
+
+			for (i = 0; i < topology->num_nodes; i++)
+				link_extend_info_output->nodes[i].node_id = topology->nodes[i].node_id;
 
-		xgmi_cmd->cmd_id = TA_COMMAND_XGMI__GET_PEER_LINKS;
+			link_extend_info_output->num_nodes = topology->num_nodes;
+			xgmi_cmd->cmd_id = TA_COMMAND_XGMI__GET_EXTEND_PEER_LINKS;
+		} else {
+			link_info_output = &xgmi_cmd->xgmi_out_message.get_link_info;
+
+			for (i = 0; i < topology->num_nodes; i++)
+				link_info_output->nodes[i].node_id = topology->nodes[i].node_id;
 
-		ret = psp_xgmi_invoke(psp, TA_COMMAND_XGMI__GET_PEER_LINKS);
+			link_info_output->num_nodes = topology->num_nodes;
+			xgmi_cmd->cmd_id = TA_COMMAND_XGMI__GET_PEER_LINKS;
+		}
 
+		ret = psp_xgmi_invoke(psp, xgmi_cmd->cmd_id);
 		if (ret)
 			return ret;
 
-		link_info_output = &xgmi_cmd->xgmi_out_message.get_link_info;
 		for (i = 0; i < topology->num_nodes; i++) {
+			uint8_t node_num_links = ta_port_num_support ?
+				link_extend_info_output->nodes[i].num_links : link_info_output->nodes[i].num_links;
 			/* accumulate num_links on extended data */
-			topology->nodes[i].num_links = get_extended_data ?
-					topology->nodes[i].num_links +
-							link_info_output->nodes[i].num_links :
-					link_info_output->nodes[i].num_links;
+			if (get_extended_data) {
+				topology->nodes[i].num_links = topology->nodes[i].num_links + node_num_links;
+			} else {
+				topology->nodes[i].num_links = (requires_reflection && topology->nodes[i].num_links) ?
+								topology->nodes[i].num_links : node_num_links;
+			}
+			/* popluate the connected port num info if supported and available */
+			if (ta_port_num_support && topology->nodes[i].num_links) {
+				memcpy(topology->nodes[i].port_num, link_extend_info_output->nodes[i].port_num,
+				       sizeof(struct xgmi_connected_port_num) * TA_XGMI__MAX_PORT_NUM);
+			}
 
 			/* reflect the topology information for bi-directionality */
-			if (psp->xgmi_context.supports_extended_data &&
-					get_extended_data && topology->nodes[i].num_hops)
+			if (requires_reflection && topology->nodes[i].num_hops)
 				psp_xgmi_reflect_topology_info(psp, topology->nodes[i]);
 		}
 	}
@@ -1427,24 +1742,84 @@ static void psp_ras_ta_check_status(struct psp_context *psp)
 	switch (ras_cmd->ras_status) {
 	case TA_RAS_STATUS__ERROR_UNSUPPORTED_IP:
 		dev_warn(psp->adev->dev,
-				"RAS WARNING: cmd failed due to unsupported ip\n");
+			 "RAS WARNING: cmd failed due to unsupported ip\n");
 		break;
 	case TA_RAS_STATUS__ERROR_UNSUPPORTED_ERROR_INJ:
 		dev_warn(psp->adev->dev,
-				"RAS WARNING: cmd failed due to unsupported error injection\n");
+			 "RAS WARNING: cmd failed due to unsupported error injection\n");
 		break;
 	case TA_RAS_STATUS__SUCCESS:
 		break;
 	case TA_RAS_STATUS__TEE_ERROR_ACCESS_DENIED:
 		if (ras_cmd->cmd_id == TA_RAS_COMMAND__TRIGGER_ERROR)
 			dev_warn(psp->adev->dev,
-					"RAS WARNING: Inject error to critical region is not allowed\n");
+				 "RAS WARNING: Inject error to critical region is not allowed\n");
 		break;
 	default:
 		dev_warn(psp->adev->dev,
-				"RAS WARNING: ras status = 0x%X\n", ras_cmd->ras_status);
+			 "RAS WARNING: ras status = 0x%X\n", ras_cmd->ras_status);
+		break;
+	}
+}
+
+static int psp_ras_send_cmd(struct psp_context *psp,
+		enum ras_command cmd_id, void *in, void *out)
+{
+	struct ta_ras_shared_memory *ras_cmd;
+	uint32_t cmd = cmd_id;
+	int ret = 0;
+
+	if (!in)
+		return -EINVAL;
+
+	mutex_lock(&psp->ras_context.mutex);
+	ras_cmd = (struct ta_ras_shared_memory *)psp->ras_context.context.mem_context.shared_buf;
+	memset(ras_cmd, 0, sizeof(struct ta_ras_shared_memory));
+
+	switch (cmd) {
+	case TA_RAS_COMMAND__ENABLE_FEATURES:
+	case TA_RAS_COMMAND__DISABLE_FEATURES:
+		memcpy(&ras_cmd->ras_in_message,
+			in, sizeof(ras_cmd->ras_in_message));
+		break;
+	case TA_RAS_COMMAND__TRIGGER_ERROR:
+		memcpy(&ras_cmd->ras_in_message.trigger_error,
+			in, sizeof(ras_cmd->ras_in_message.trigger_error));
 		break;
+	case TA_RAS_COMMAND__QUERY_ADDRESS:
+		memcpy(&ras_cmd->ras_in_message.address,
+			in, sizeof(ras_cmd->ras_in_message.address));
+		break;
+	default:
+		dev_err(psp->adev->dev, "Invalid ras cmd id: %u\n", cmd);
+		ret = -EINVAL;
+		goto err_out;
 	}
+
+	ras_cmd->cmd_id = cmd;
+	ret = psp_ras_invoke(psp, ras_cmd->cmd_id);
+
+	switch (cmd) {
+	case TA_RAS_COMMAND__TRIGGER_ERROR:
+		if (!ret && out)
+			memcpy(out, &ras_cmd->ras_status, sizeof(ras_cmd->ras_status));
+		break;
+	case TA_RAS_COMMAND__QUERY_ADDRESS:
+		if (ret || ras_cmd->ras_status || psp->cmd_buf_mem->resp.status)
+			ret = -EINVAL;
+		else if (out)
+			memcpy(out,
+				&ras_cmd->ras_out_message.address,
+				sizeof(ras_cmd->ras_out_message.address));
+		break;
+	default:
+		break;
+	}
+
+err_out:
+	mutex_unlock(&psp->ras_context.mutex);
+
+	return ret;
 }
 
 int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
@@ -1465,9 +1840,8 @@ int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 	if (amdgpu_ras_intr_triggered())
 		return ret;
 
-	if (ras_cmd->if_version > RAS_TA_HOST_IF_VER)
-	{
-		DRM_WARN("RAS: Unsupported Interface");
+	if (ras_cmd->if_version > RAS_TA_HOST_IF_VER) {
+		dev_warn(psp->adev->dev, "RAS: Unsupported Interface\n");
 		return -EINVAL;
 	}
 
@@ -1476,8 +1850,7 @@ int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 			dev_warn(psp->adev->dev, "ECC switch disabled\n");
 
 			ras_cmd->ras_status = TA_RAS_STATUS__ERROR_RAS_NOT_AVAILABLE;
-		}
-		else if (ras_cmd->ras_out_message.flags.reg_access_failure_flag)
+		} else if (ras_cmd->ras_out_message.flags.reg_access_failure_flag)
 			dev_warn(psp->adev->dev,
 				 "RAS internal register access blocked\n");
 
@@ -1490,23 +1863,15 @@ int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 int psp_ras_enable_features(struct psp_context *psp,
 		union ta_ras_cmd_input *info, bool enable)
 {
-	struct ta_ras_shared_memory *ras_cmd;
+	enum ras_command cmd_id;
 	int ret;
 
-	if (!psp->ras_context.context.initialized)
+	if (!psp->ras_context.context.initialized || !info)
 		return -EINVAL;
 
-	ras_cmd = (struct ta_ras_shared_memory *)psp->ras_context.context.mem_context.shared_buf;
-	memset(ras_cmd, 0, sizeof(struct ta_ras_shared_memory));
-
-	if (enable)
-		ras_cmd->cmd_id = TA_RAS_COMMAND__ENABLE_FEATURES;
-	else
-		ras_cmd->cmd_id = TA_RAS_COMMAND__DISABLE_FEATURES;
-
-	ras_cmd->ras_in_message = *info;
-
-	ret = psp_ras_invoke(psp, ras_cmd->cmd_id);
+	cmd_id = enable ?
+		TA_RAS_COMMAND__ENABLE_FEATURES : TA_RAS_COMMAND__DISABLE_FEATURES;
+	ret = psp_ras_send_cmd(psp, cmd_id, info, NULL);
 	if (ret)
 		return -EINVAL;
 
@@ -1530,6 +1895,8 @@ int psp_ras_terminate(struct psp_context *psp)
 
 	psp->ras_context.context.initialized = false;
 
+	mutex_destroy(&psp->ras_context.mutex);
+
 	return ret;
 }
 
@@ -1560,35 +1927,47 @@ int psp_ras_initialize(struct psp_context *psp)
 		if (ret)
 			dev_warn(adev->dev, "PSP get boot config failed\n");
 
-		if (!amdgpu_ras_is_supported(psp->adev, AMDGPU_RAS_BLOCK__UMC)) {
-			if (!boot_cfg) {
-				dev_info(adev->dev, "GECC is disabled\n");
-			} else {
-				/* disable GECC in next boot cycle if ras is
-				 * disabled by module parameter amdgpu_ras_enable
-				 * and/or amdgpu_ras_mask, or boot_config_get call
-				 * is failed
-				 */
-				ret = psp_boot_config_set(adev, 0);
-				if (ret)
-					dev_warn(adev->dev, "PSP set boot config failed\n");
-				else
-					dev_warn(adev->dev, "GECC will be disabled in next boot cycle "
-						 "if set amdgpu_ras_enable and/or amdgpu_ras_mask to 0x0\n");
-			}
+		if (boot_cfg == 1 && !adev->ras_default_ecc_enabled &&
+		    amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC)) {
+			dev_warn(adev->dev, "GECC is currently enabled, which may affect performance\n");
+			dev_warn(adev->dev,
+				"To disable GECC, please reboot the system and load the amdgpu driver with the parameter amdgpu_ras_enable=0\n");
 		} else {
-			if (1 == boot_cfg) {
-				dev_info(adev->dev, "GECC is enabled\n");
+			if ((adev->ras_default_ecc_enabled || amdgpu_ras_enable == 1) &&
+				amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC)) {
+				if (boot_cfg == 1) {
+					dev_info(adev->dev, "GECC is enabled\n");
+				} else {
+					/* enable GECC in next boot cycle if it is disabled
+					 * in boot config, or force enable GECC if failed to
+					 * get boot configuration
+					 */
+					ret = psp_boot_config_set(adev, BOOT_CONFIG_GECC);
+					if (ret)
+						dev_warn(adev->dev, "PSP set boot config failed\n");
+					else
+						dev_warn(adev->dev, "GECC will be enabled in next boot cycle\n");
+				}
 			} else {
-				/* enable GECC in next boot cycle if it is disabled
-				 * in boot config, or force enable GECC if failed to
-				 * get boot configuration
-				 */
-				ret = psp_boot_config_set(adev, BOOT_CONFIG_GECC);
-				if (ret)
-					dev_warn(adev->dev, "PSP set boot config failed\n");
-				else
-					dev_warn(adev->dev, "GECC will be enabled in next boot cycle\n");
+				if (!boot_cfg) {
+					if (!adev->ras_default_ecc_enabled &&
+					    amdgpu_ras_enable != 1 &&
+					    amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC))
+						dev_warn(adev->dev, "GECC is disabled, set amdgpu_ras_enable=1 to enable GECC in next boot cycle if needed\n");
+					else
+						dev_info(adev->dev, "GECC is disabled\n");
+				} else {
+					/* disable GECC in next boot cycle if ras is
+					 * disabled by module parameter amdgpu_ras_enable
+					 * and/or amdgpu_ras_mask, or boot_config_get call
+					 * is failed
+					 */
+					ret = psp_boot_config_set(adev, 0);
+					if (ret)
+						dev_warn(adev->dev, "PSP set boot config failed\n");
+					else
+						dev_warn(adev->dev, "GECC will be disabled in next boot cycle if set amdgpu_ras_enable and/or amdgpu_ras_mask to 0x0\n");
+				}
 			}
 		}
 	}
@@ -1607,16 +1986,24 @@ int psp_ras_initialize(struct psp_context *psp)
 
 	if (amdgpu_ras_is_poison_mode_supported(adev))
 		ras_cmd->ras_in_message.init_flags.poison_mode_en = 1;
-	if (!adev->gmc.xgmi.connected_to_cpu)
+	if (!adev->gmc.xgmi.connected_to_cpu && !adev->gmc.is_app_apu)
 		ras_cmd->ras_in_message.init_flags.dgpu_mode = 1;
+	ras_cmd->ras_in_message.init_flags.xcc_mask =
+		adev->gfx.xcc_mask;
+	ras_cmd->ras_in_message.init_flags.channel_dis_num = hweight32(adev->gmc.m_half_use) * 2;
+	if (adev->gmc.gmc_funcs->query_mem_partition_mode)
+		ras_cmd->ras_in_message.init_flags.nps_mode =
+			adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+	ras_cmd->ras_in_message.init_flags.active_umc_mask = adev->umc.active_mask;
 
 	ret = psp_ta_load(psp, &psp->ras_context.context);
 
-	if (!ret && !ras_cmd->ras_status)
+	if (!ret && !ras_cmd->ras_status) {
 		psp->ras_context.context.initialized = true;
-	else {
+		mutex_init(&psp->ras_context.mutex);
+	} else {
 		if (ras_cmd->ras_status)
-			dev_warn(psp->adev->dev, "RAS Init Status: 0x%X\n", ras_cmd->ras_status);
+			dev_warn(adev->dev, "RAS Init Status: 0x%X\n", ras_cmd->ras_status);
 
 		/* fail to load RAS TA */
 		psp->ras_context.context.initialized = false;
@@ -1626,36 +2013,71 @@ int psp_ras_initialize(struct psp_context *psp)
 }
 
 int psp_ras_trigger_error(struct psp_context *psp,
-			  struct ta_ras_trigger_error_input *info)
+			  struct ta_ras_trigger_error_input *info, uint32_t instance_mask)
 {
-	struct ta_ras_shared_memory *ras_cmd;
+	struct amdgpu_device *adev = psp->adev;
 	int ret;
+	uint32_t dev_mask;
+	uint32_t ras_status = 0;
 
-	if (!psp->ras_context.context.initialized)
+	if (!psp->ras_context.context.initialized || !info)
 		return -EINVAL;
 
-	ras_cmd = (struct ta_ras_shared_memory *)psp->ras_context.context.mem_context.shared_buf;
-	memset(ras_cmd, 0, sizeof(struct ta_ras_shared_memory));
+	switch (info->block_id) {
+	case TA_RAS_BLOCK__GFX:
+		dev_mask = GET_MASK(GC, instance_mask);
+		break;
+	case TA_RAS_BLOCK__SDMA:
+		dev_mask = GET_MASK(SDMA0, instance_mask);
+		break;
+	case TA_RAS_BLOCK__VCN:
+	case TA_RAS_BLOCK__JPEG:
+		dev_mask = GET_MASK(VCN, instance_mask);
+		break;
+	default:
+		dev_mask = instance_mask;
+		break;
+	}
 
-	ras_cmd->cmd_id = TA_RAS_COMMAND__TRIGGER_ERROR;
-	ras_cmd->ras_in_message.trigger_error = *info;
+	/* reuse sub_block_index for backward compatibility */
+	dev_mask <<= AMDGPU_RAS_INST_SHIFT;
+	dev_mask &= AMDGPU_RAS_INST_MASK;
+	info->sub_block_index |= dev_mask;
 
-	ret = psp_ras_invoke(psp, ras_cmd->cmd_id);
+	ret = psp_ras_send_cmd(psp,
+			TA_RAS_COMMAND__TRIGGER_ERROR, info, &ras_status);
 	if (ret)
 		return -EINVAL;
 
 	/* If err_event_athub occurs error inject was successful, however
-	   return status from TA is no long reliable */
+	 *  return status from TA is no long reliable
+	 */
 	if (amdgpu_ras_intr_triggered())
 		return 0;
 
-	if (ras_cmd->ras_status == TA_RAS_STATUS__TEE_ERROR_ACCESS_DENIED)
+	if (ras_status == TA_RAS_STATUS__TEE_ERROR_ACCESS_DENIED)
 		return -EACCES;
-	else if (ras_cmd->ras_status)
+	else if (ras_status)
 		return -EINVAL;
 
 	return 0;
 }
+
+int psp_ras_query_address(struct psp_context *psp,
+			  struct ta_ras_query_address_input *addr_in,
+			  struct ta_ras_query_address_output *addr_out)
+{
+	int ret;
+
+	if (!psp->ras_context.context.initialized ||
+		!addr_in || !addr_out)
+		return -EINVAL;
+
+	ret = psp_ras_send_cmd(psp,
+			TA_RAS_COMMAND__QUERY_ADDRESS, addr_in, addr_out);
+
+	return ret;
+}
 // ras end
 
 // HDCP start
@@ -1669,6 +2091,10 @@ static int psp_hdcp_initialize(struct psp_context *psp)
 	if (amdgpu_sriov_vf(psp->adev))
 		return 0;
 
+	/* bypass hdcp initialization if dmu is harvested */
+	if (!amdgpu_device_has_display_hardware(psp->adev))
+		return 0;
+
 	if (!psp->hdcp_context.context.bin_desc.size_bytes ||
 	    !psp->hdcp_context.context.bin_desc.start_addr) {
 		dev_info(psp->adev->dev, "HDCP: optional hdcp ta ucode is not available\n");
@@ -1701,6 +2127,9 @@ int psp_hdcp_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 	if (amdgpu_sriov_vf(psp->adev))
 		return 0;
 
+	if (!psp->hdcp_context.context.initialized)
+		return 0;
+
 	return psp_ta_invoke(psp, ta_cmd_id, &psp->hdcp_context.context);
 }
 
@@ -1736,6 +2165,10 @@ static int psp_dtm_initialize(struct psp_context *psp)
 	if (amdgpu_sriov_vf(psp->adev))
 		return 0;
 
+	/* bypass dtm initialization if dmu is harvested */
+	if (!amdgpu_device_has_display_hardware(psp->adev))
+		return 0;
+
 	if (!psp->dtm_context.context.bin_desc.size_bytes ||
 	    !psp->dtm_context.context.bin_desc.start_addr) {
 		dev_info(psp->adev->dev, "DTM: optional dtm ta ucode is not available\n");
@@ -1768,6 +2201,9 @@ int psp_dtm_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 	if (amdgpu_sriov_vf(psp->adev))
 		return 0;
 
+	if (!psp->dtm_context.context.initialized)
+		return 0;
+
 	return psp_ta_invoke(psp, ta_cmd_id, &psp->dtm_context.context);
 }
 
@@ -1902,9 +2338,14 @@ static int psp_securedisplay_initialize(struct psp_context *psp)
 	if (amdgpu_sriov_vf(psp->adev))
 		return 0;
 
+	/* bypass securedisplay initialization if dmu is harvested */
+	if (!amdgpu_device_has_display_hardware(psp->adev))
+		return 0;
+
 	if (!psp->securedisplay_context.context.bin_desc.size_bytes ||
 	    !psp->securedisplay_context.context.bin_desc.start_addr) {
-		dev_info(psp->adev->dev, "SECUREDISPLAY: securedisplay ta ucode is not available\n");
+		dev_info(psp->adev->dev,
+			 "SECUREDISPLAY: optional securedisplay ta ucode is not available\n");
 		return 0;
 	}
 
@@ -1920,11 +2361,14 @@ static int psp_securedisplay_initialize(struct psp_context *psp)
 	}
 
 	ret = psp_ta_load(psp, &psp->securedisplay_context.context);
-	if (!ret) {
+	if (!ret && !psp->securedisplay_context.context.resp_status) {
 		psp->securedisplay_context.context.initialized = true;
 		mutex_init(&psp->securedisplay_context.mutex);
-	} else
+	} else {
+		/* don't try again */
+		psp->securedisplay_context.context.bin_desc.size_bytes = 0;
 		return ret;
+	}
 
 	mutex_lock(&psp->securedisplay_context.mutex);
 
@@ -1947,6 +2391,8 @@ static int psp_securedisplay_initialize(struct psp_context *psp)
 		psp_securedisplay_parse_resp_status(psp, securedisplay_cmd->status);
 		dev_err(psp->adev->dev, "SECUREDISPLAY: query securedisplay TA failed. ret 0x%x\n",
 			securedisplay_cmd->securedisplay_out_message.query_ta.query_cmd_ret);
+		/* don't try again */
+		psp->securedisplay_context.context.bin_desc.size_bytes = 0;
 	}
 
 	return 0;
@@ -1980,7 +2426,8 @@ int psp_securedisplay_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 		return -EINVAL;
 
 	if (ta_cmd_id != TA_SECUREDISPLAY_COMMAND__QUERY_TA &&
-	    ta_cmd_id != TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC)
+	    ta_cmd_id != TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC &&
+	    ta_cmd_id != TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC_V2)
 		return -EINVAL;
 
 	ret = psp_ta_invoke(psp, ta_cmd_id, &psp->securedisplay_context.context);
@@ -1989,17 +2436,67 @@ int psp_securedisplay_invoke(struct psp_context *psp, uint32_t ta_cmd_id)
 }
 /* SECUREDISPLAY end */
 
+int amdgpu_psp_wait_for_bootloader(struct amdgpu_device *adev)
+{
+	struct psp_context *psp = &adev->psp;
+	int ret = 0;
+
+	if (!amdgpu_sriov_vf(adev) && psp->funcs && psp->funcs->wait_for_bootloader != NULL)
+		ret = psp->funcs->wait_for_bootloader(psp);
+
+	return ret;
+}
+
+bool amdgpu_psp_get_ras_capability(struct psp_context *psp)
+{
+	if (psp->funcs &&
+	    psp->funcs->get_ras_capability) {
+		return psp->funcs->get_ras_capability(psp);
+	} else {
+		return false;
+	}
+}
+
+bool amdgpu_psp_tos_reload_needed(struct amdgpu_device *adev)
+{
+	struct psp_context *psp = &adev->psp;
+
+	if (amdgpu_sriov_vf(adev) || (adev->flags & AMD_IS_APU))
+		return false;
+
+	if (psp->funcs && psp->funcs->is_reload_needed)
+		return psp->funcs->is_reload_needed(psp);
+
+	return false;
+}
+
+static void psp_update_gpu_addresses(struct amdgpu_device *adev)
+{
+	struct psp_context *psp = &adev->psp;
+
+	if (psp->cmd_buf_bo && psp->cmd_buf_mem) {
+		psp->fw_pri_mc_addr = amdgpu_bo_fb_aper_addr(psp->fw_pri_bo);
+		psp->fence_buf_mc_addr = amdgpu_bo_fb_aper_addr(psp->fence_buf_bo);
+		psp->cmd_buf_mc_addr = amdgpu_bo_fb_aper_addr(psp->cmd_buf_bo);
+	}
+	if (adev->firmware.rbuf && psp->km_ring.ring_mem)
+		psp->km_ring.ring_mem_mc_addr = amdgpu_bo_fb_aper_addr(adev->firmware.rbuf);
+}
+
 static int psp_hw_start(struct psp_context *psp)
 {
 	struct amdgpu_device *adev = psp->adev;
 	int ret;
 
+	if (amdgpu_virt_xgmi_migrate_enabled(adev))
+		psp_update_gpu_addresses(adev);
+
 	if (!amdgpu_sriov_vf(adev)) {
 		if ((is_psp_fw_valid(psp->kdb)) &&
 		    (psp->funcs->bootloader_load_kdb != NULL)) {
 			ret = psp_bootloader_load_kdb(psp);
 			if (ret) {
-				DRM_ERROR("PSP load kdb failed!\n");
+				dev_err(adev->dev, "PSP load kdb failed!\n");
 				return ret;
 			}
 		}
@@ -2008,7 +2505,7 @@ static int psp_hw_start(struct psp_context *psp)
 		    (psp->funcs->bootloader_load_spl != NULL)) {
 			ret = psp_bootloader_load_spl(psp);
 			if (ret) {
-				DRM_ERROR("PSP load spl failed!\n");
+				dev_err(adev->dev, "PSP load spl failed!\n");
 				return ret;
 			}
 		}
@@ -2017,7 +2514,7 @@ static int psp_hw_start(struct psp_context *psp)
 		    (psp->funcs->bootloader_load_sysdrv != NULL)) {
 			ret = psp_bootloader_load_sysdrv(psp);
 			if (ret) {
-				DRM_ERROR("PSP load sys drv failed!\n");
+				dev_err(adev->dev, "PSP load sys drv failed!\n");
 				return ret;
 			}
 		}
@@ -2026,7 +2523,7 @@ static int psp_hw_start(struct psp_context *psp)
 		    (psp->funcs->bootloader_load_soc_drv != NULL)) {
 			ret = psp_bootloader_load_soc_drv(psp);
 			if (ret) {
-				DRM_ERROR("PSP load soc drv failed!\n");
+				dev_err(adev->dev, "PSP load soc drv failed!\n");
 				return ret;
 			}
 		}
@@ -2035,7 +2532,7 @@ static int psp_hw_start(struct psp_context *psp)
 		    (psp->funcs->bootloader_load_intf_drv != NULL)) {
 			ret = psp_bootloader_load_intf_drv(psp);
 			if (ret) {
-				DRM_ERROR("PSP load intf drv failed!\n");
+				dev_err(adev->dev, "PSP load intf drv failed!\n");
 				return ret;
 			}
 		}
@@ -2044,7 +2541,7 @@ static int psp_hw_start(struct psp_context *psp)
 		    (psp->funcs->bootloader_load_dbg_drv != NULL)) {
 			ret = psp_bootloader_load_dbg_drv(psp);
 			if (ret) {
-				DRM_ERROR("PSP load dbg drv failed!\n");
+				dev_err(adev->dev, "PSP load dbg drv failed!\n");
 				return ret;
 			}
 		}
@@ -2053,7 +2550,25 @@ static int psp_hw_start(struct psp_context *psp)
 		    (psp->funcs->bootloader_load_ras_drv != NULL)) {
 			ret = psp_bootloader_load_ras_drv(psp);
 			if (ret) {
-				DRM_ERROR("PSP load ras_drv failed!\n");
+				dev_err(adev->dev, "PSP load ras_drv failed!\n");
+				return ret;
+			}
+		}
+
+		if ((is_psp_fw_valid(psp->ipkeymgr_drv)) &&
+		    (psp->funcs->bootloader_load_ipkeymgr_drv != NULL)) {
+			ret = psp_bootloader_load_ipkeymgr_drv(psp);
+			if (ret) {
+				dev_err(adev->dev, "PSP load ipkeymgr_drv failed!\n");
+				return ret;
+			}
+		}
+
+		if ((is_psp_fw_valid(psp->spdm_drv)) &&
+		    (psp->funcs->bootloader_load_spdm_drv != NULL)) {
+			ret = psp_bootloader_load_spdm_drv(psp);
+			if (ret) {
+				dev_err(adev->dev, "PSP load spdm_drv failed!\n");
 				return ret;
 			}
 		}
@@ -2062,7 +2577,7 @@ static int psp_hw_start(struct psp_context *psp)
 		    (psp->funcs->bootloader_load_sos != NULL)) {
 			ret = psp_bootloader_load_sos(psp);
 			if (ret) {
-				DRM_ERROR("PSP load sos failed!\n");
+				dev_err(adev->dev, "PSP load sos failed!\n");
 				return ret;
 			}
 		}
@@ -2070,17 +2585,27 @@ static int psp_hw_start(struct psp_context *psp)
 
 	ret = psp_ring_create(psp, PSP_RING_TYPE__KM);
 	if (ret) {
-		DRM_ERROR("PSP create ring failed!\n");
+		dev_err(adev->dev, "PSP create ring failed!\n");
 		return ret;
 	}
 
+	if (!amdgpu_in_reset(adev) && !adev->in_suspend) {
+		ret = psp_update_fw_reservation(psp);
+		if (ret) {
+			dev_err(adev->dev, "update fw reservation failed!\n");
+			return ret;
+		}
+	}
+
 	if (amdgpu_sriov_vf(adev) && amdgpu_in_reset(adev))
 		goto skip_pin_bo;
 
-	ret = psp_tmr_init(psp);
-	if (ret) {
-		DRM_ERROR("PSP tmr init failed!\n");
-		return ret;
+	if (!psp->boot_time_tmr || psp->autoload_supported) {
+		ret = psp_tmr_init(psp);
+		if (ret) {
+			dev_err(adev->dev, "PSP tmr init failed!\n");
+			return ret;
+		}
 	}
 
 skip_pin_bo:
@@ -2095,10 +2620,12 @@ skip_pin_bo:
 			return ret;
 	}
 
-	ret = psp_tmr_load(psp);
-	if (ret) {
-		DRM_ERROR("PSP load tmr failed!\n");
-		return ret;
+	if (!psp->boot_time_tmr || !psp->autoload_supported) {
+		ret = psp_tmr_load(psp);
+		if (ret) {
+			dev_err(adev->dev, "PSP load tmr failed!\n");
+			return ret;
+		}
 	}
 
 	return 0;
@@ -2244,6 +2771,7 @@ static int psp_get_fw_type(struct amdgpu_firmware_info *ucode,
 		*type = GFX_FW_TYPE_DMUB;
 		break;
 	case AMDGPU_UCODE_ID_SDMA_UCODE_TH0:
+	case AMDGPU_UCODE_ID_SDMA_RS64:
 		*type = GFX_FW_TYPE_SDMA_UCODE_TH0;
 		break;
 	case AMDGPU_UCODE_ID_SDMA_UCODE_TH1:
@@ -2288,6 +2816,33 @@ static int psp_get_fw_type(struct amdgpu_firmware_info *ucode,
 	case AMDGPU_UCODE_ID_CP_RS64_MEC_P3_STACK:
 		*type = GFX_FW_TYPE_RS64_MEC_P3_STACK;
 		break;
+	case AMDGPU_UCODE_ID_VPE_CTX:
+		*type = GFX_FW_TYPE_VPEC_FW1;
+		break;
+	case AMDGPU_UCODE_ID_VPE_CTL:
+		*type = GFX_FW_TYPE_VPEC_FW2;
+		break;
+	case AMDGPU_UCODE_ID_VPE:
+		*type = GFX_FW_TYPE_VPE;
+		break;
+	case AMDGPU_UCODE_ID_UMSCH_MM_UCODE:
+		*type = GFX_FW_TYPE_UMSCH_UCODE;
+		break;
+	case AMDGPU_UCODE_ID_UMSCH_MM_DATA:
+		*type = GFX_FW_TYPE_UMSCH_DATA;
+		break;
+	case AMDGPU_UCODE_ID_UMSCH_MM_CMD_BUFFER:
+		*type = GFX_FW_TYPE_UMSCH_CMD_BUFFER;
+		break;
+	case AMDGPU_UCODE_ID_P2S_TABLE:
+		*type = GFX_FW_TYPE_P2S_TABLE;
+		break;
+	case AMDGPU_UCODE_ID_JPEG_RAM:
+		*type = GFX_FW_TYPE_JPEG_RAM;
+		break;
+	case AMDGPU_UCODE_ID_ISP:
+		*type = GFX_FW_TYPE_ISP;
+		break;
 	case AMDGPU_UCODE_ID_MAXIMUM:
 	default:
 		return -EINVAL;
@@ -2344,7 +2899,8 @@ static void psp_print_fw_hdr(struct psp_context *psp,
 	}
 }
 
-static int psp_prep_load_ip_fw_cmd_buf(struct amdgpu_firmware_info *ucode,
+static int psp_prep_load_ip_fw_cmd_buf(struct psp_context *psp,
+				       struct amdgpu_firmware_info *ucode,
 				       struct psp_gfx_cmd_resp *cmd)
 {
 	int ret;
@@ -2357,18 +2913,18 @@ static int psp_prep_load_ip_fw_cmd_buf(struct amdgpu_firmware_info *ucode,
 
 	ret = psp_get_fw_type(ucode, &cmd->cmd.cmd_load_ip_fw.fw_type);
 	if (ret)
-		DRM_ERROR("Unknown firmware type\n");
+		dev_err(psp->adev->dev, "Unknown firmware type\n");
 
 	return ret;
 }
 
-static int psp_execute_non_psp_fw_load(struct psp_context *psp,
-			          struct amdgpu_firmware_info *ucode)
+int psp_execute_ip_fw_load(struct psp_context *psp,
+			   struct amdgpu_firmware_info *ucode)
 {
 	int ret = 0;
 	struct psp_gfx_cmd_resp *cmd = acquire_psp_cmd_buf(psp);
 
-	ret = psp_prep_load_ip_fw_cmd_buf(ucode, cmd);
+	ret = psp_prep_load_ip_fw_cmd_buf(psp, ucode, cmd);
 	if (!ret) {
 		ret = psp_cmd_submit_buf(psp, ucode, cmd,
 					 psp->fence_buf_mc_addr);
@@ -2379,6 +2935,33 @@ static int psp_execute_non_psp_fw_load(struct psp_context *psp,
 	return ret;
 }
 
+static int psp_load_p2s_table(struct psp_context *psp)
+{
+	int ret;
+	struct amdgpu_device *adev = psp->adev;
+	struct amdgpu_firmware_info *ucode =
+		&adev->firmware.ucode[AMDGPU_UCODE_ID_P2S_TABLE];
+
+	if (adev->in_runpm && ((adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) ||
+				(adev->pm.rpm_mode == AMDGPU_RUNPM_BAMACO)))
+		return 0;
+
+	if (amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 6) ||
+	    amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 14)) {
+		uint32_t supp_vers = adev->flags & AMD_IS_APU ? 0x0036013D :
+								0x0036003C;
+		if (psp->sos.fw_version < supp_vers)
+			return 0;
+	}
+
+	if (!ucode->fw || amdgpu_sriov_vf(psp->adev))
+		return 0;
+
+	ret = psp_execute_ip_fw_load(psp, ucode);
+
+	return ret;
+}
+
 static int psp_load_smu_fw(struct psp_context *psp)
 {
 	int ret;
@@ -2391,26 +2974,25 @@ static int psp_load_smu_fw(struct psp_context *psp)
 	 * Skip SMU FW reloading in case of using BACO for runpm only,
 	 * as SMU is always alive.
 	 */
-	if (adev->in_runpm && (adev->pm.rpm_mode == AMDGPU_RUNPM_BACO))
+	if (adev->in_runpm && ((adev->pm.rpm_mode == AMDGPU_RUNPM_BACO) ||
+				(adev->pm.rpm_mode == AMDGPU_RUNPM_BAMACO)))
 		return 0;
 
 	if (!ucode->fw || amdgpu_sriov_vf(psp->adev))
 		return 0;
 
-	if ((amdgpu_in_reset(adev) &&
-	     ras && adev->ras_enabled &&
-	     (adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 4) ||
-	      adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 2)))) {
+	if ((amdgpu_in_reset(adev) && ras && adev->ras_enabled &&
+	     (amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(11, 0, 4) ||
+	      amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(11, 0, 2)))) {
 		ret = amdgpu_dpm_set_mp1_state(adev, PP_MP1_STATE_UNLOAD);
-		if (ret) {
-			DRM_WARN("Failed to set MP1 state prepare for reload\n");
-		}
+		if (ret)
+			dev_err(adev->dev, "Failed to set MP1 state prepare for reload\n");
 	}
 
-	ret = psp_execute_non_psp_fw_load(psp, ucode);
+	ret = psp_execute_ip_fw_load(psp, ucode);
 
 	if (ret)
-		DRM_ERROR("PSP load smu failed!\n");
+		dev_err(adev->dev, "PSP load smu failed!\n");
 
 	return ret;
 }
@@ -2421,6 +3003,9 @@ static bool fw_load_skip_check(struct psp_context *psp,
 	if (!ucode->fw || !ucode->ucode_size)
 		return true;
 
+	if (ucode->ucode_id == AMDGPU_UCODE_ID_P2S_TABLE)
+		return true;
+
 	if (ucode->ucode_id == AMDGPU_UCODE_ID_SMC &&
 	    (psp_smu_reload_quirk(psp) ||
 	     psp->autoload_supported ||
@@ -2449,7 +3034,7 @@ int psp_load_fw_list(struct psp_context *psp,
 	for (i = 0; i < ucode_count; ++i) {
 		ucode = ucode_list[i];
 		psp_print_fw_hdr(psp, ucode);
-		ret = psp_execute_non_psp_fw_load(psp, ucode);
+		ret = psp_execute_ip_fw_load(psp, ucode);
 		if (ret)
 			return ret;
 	}
@@ -2469,6 +3054,9 @@ static int psp_load_non_psp_fw(struct psp_context *psp)
 			return ret;
 	}
 
+	/* Load P2S table first if it's available */
+	psp_load_p2s_table(psp);
+
 	for (i = 0; i < adev->firmware.max_ucodes; i++) {
 		ucode = &adev->firmware.ucode[i];
 
@@ -2484,28 +3072,32 @@ static int psp_load_non_psp_fw(struct psp_context *psp)
 			continue;
 
 		if (psp->autoload_supported &&
-		    (adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 7) ||
-		     adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 11) ||
-		     adev->ip_versions[MP0_HWIP][0] == IP_VERSION(11, 0, 12)) &&
+		    (amdgpu_ip_version(adev, MP0_HWIP, 0) ==
+			     IP_VERSION(11, 0, 7) ||
+		     amdgpu_ip_version(adev, MP0_HWIP, 0) ==
+			     IP_VERSION(11, 0, 11) ||
+		     amdgpu_ip_version(adev, MP0_HWIP, 0) ==
+			     IP_VERSION(11, 0, 12)) &&
 		    (ucode->ucode_id == AMDGPU_UCODE_ID_SDMA1 ||
 		     ucode->ucode_id == AMDGPU_UCODE_ID_SDMA2 ||
 		     ucode->ucode_id == AMDGPU_UCODE_ID_SDMA3))
 			/* PSP only receive one SDMA fw for sienna_cichlid,
-			 * as all four sdma fw are same */
+			 * as all four sdma fw are same
+			 */
 			continue;
 
 		psp_print_fw_hdr(psp, ucode);
 
-		ret = psp_execute_non_psp_fw_load(psp, ucode);
+		ret = psp_execute_ip_fw_load(psp, ucode);
 		if (ret)
 			return ret;
 
-		/* Start rlc autoload after psp recieved all the gfx firmware */
+		/* Start rlc autoload after psp received all the gfx firmware */
 		if (psp->autoload_supported && ucode->ucode_id == (amdgpu_sriov_vf(adev) ?
 		    adev->virt.autoload_ucode_id : AMDGPU_UCODE_ID_RLC_G)) {
 			ret = psp_rlc_autoload_start(psp);
 			if (ret) {
-				DRM_ERROR("Failed to start rlc autoload\n");
+				dev_err(adev->dev, "Failed to start rlc autoload\n");
 				return ret;
 			}
 		}
@@ -2527,7 +3119,7 @@ static int psp_load_fw(struct amdgpu_device *adev)
 
 		ret = psp_ring_init(psp, PSP_RING_TYPE__KM);
 		if (ret) {
-			DRM_ERROR("PSP ring init failed!\n");
+			dev_err(adev->dev, "PSP ring init failed!\n");
 			goto failed;
 		}
 	}
@@ -2542,13 +3134,13 @@ static int psp_load_fw(struct amdgpu_device *adev)
 
 	ret = psp_asd_initialize(psp);
 	if (ret) {
-		DRM_ERROR("PSP load asd failed!\n");
+		dev_err(adev->dev, "PSP load asd failed!\n");
 		goto failed1;
 	}
 
 	ret = psp_rl_load(adev);
 	if (ret) {
-		DRM_ERROR("PSP load RL failed!\n");
+		dev_err(adev->dev, "PSP load RL failed!\n");
 		goto failed1;
 	}
 
@@ -2556,8 +3148,8 @@ static int psp_load_fw(struct amdgpu_device *adev)
 		if (adev->gmc.xgmi.num_physical_nodes > 1) {
 			ret = psp_xgmi_initialize(psp, false, true);
 			/* Warning the XGMI seesion initialize failure
-			* Instead of stop driver initialization
-			*/
+			 * Instead of stop driver initialization
+			 */
 			if (ret)
 				dev_err(psp->adev->dev,
 					"XGMI: Failed to initialize XGMI session\n");
@@ -2568,7 +3160,7 @@ static int psp_load_fw(struct amdgpu_device *adev)
 		ret = psp_ras_initialize(psp);
 		if (ret)
 			dev_err(psp->adev->dev,
-					"RAS: Failed to initialize RAS\n");
+				"RAS: Failed to initialize RAS\n");
 
 		ret = psp_hdcp_initialize(psp);
 		if (ret)
@@ -2605,23 +3197,20 @@ failed:
 	return ret;
 }
 
-static int psp_hw_init(void *handle)
+static int psp_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int ret;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	mutex_lock(&adev->firmware.mutex);
-	/*
-	 * This sequence is just used on hw_init only once, no need on
-	 * resume.
-	 */
+
 	ret = amdgpu_ucode_init_bo(adev);
 	if (ret)
 		goto failed;
 
 	ret = psp_load_fw(adev);
 	if (ret) {
-		DRM_ERROR("PSP firmware loading failed\n");
+		dev_err(adev->dev, "PSP firmware loading failed\n");
 		goto failed;
 	}
 
@@ -2634,9 +3223,9 @@ failed:
 	return -EINVAL;
 }
 
-static int psp_hw_fini(void *handle)
+static int psp_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct psp_context *psp = &adev->psp;
 
 	if (psp->ta_fw) {
@@ -2655,22 +3244,20 @@ static int psp_hw_fini(void *handle)
 
 	psp_ring_destroy(psp, PSP_RING_TYPE__KM);
 
-	psp_free_shared_bufs(psp);
-
 	return 0;
 }
 
-static int psp_suspend(void *handle)
+static int psp_suspend(struct amdgpu_ip_block *ip_block)
 {
 	int ret = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct psp_context *psp = &adev->psp;
 
 	if (adev->gmc.xgmi.num_physical_nodes > 1 &&
 	    psp->xgmi_context.context.initialized) {
 		ret = psp_xgmi_terminate(psp);
 		if (ret) {
-			DRM_ERROR("Failed to terminate xgmi ta\n");
+			dev_err(adev->dev, "Failed to terminate xgmi ta\n");
 			goto out;
 		}
 	}
@@ -2678,70 +3265,73 @@ static int psp_suspend(void *handle)
 	if (psp->ta_fw) {
 		ret = psp_ras_terminate(psp);
 		if (ret) {
-			DRM_ERROR("Failed to terminate ras ta\n");
+			dev_err(adev->dev, "Failed to terminate ras ta\n");
 			goto out;
 		}
 		ret = psp_hdcp_terminate(psp);
 		if (ret) {
-			DRM_ERROR("Failed to terminate hdcp ta\n");
+			dev_err(adev->dev, "Failed to terminate hdcp ta\n");
 			goto out;
 		}
 		ret = psp_dtm_terminate(psp);
 		if (ret) {
-			DRM_ERROR("Failed to terminate dtm ta\n");
+			dev_err(adev->dev, "Failed to terminate dtm ta\n");
 			goto out;
 		}
 		ret = psp_rap_terminate(psp);
 		if (ret) {
-			DRM_ERROR("Failed to terminate rap ta\n");
+			dev_err(adev->dev, "Failed to terminate rap ta\n");
 			goto out;
 		}
 		ret = psp_securedisplay_terminate(psp);
 		if (ret) {
-			DRM_ERROR("Failed to terminate securedisplay ta\n");
+			dev_err(adev->dev, "Failed to terminate securedisplay ta\n");
 			goto out;
 		}
 	}
 
 	ret = psp_asd_terminate(psp);
 	if (ret) {
-		DRM_ERROR("Failed to terminate asd\n");
+		dev_err(adev->dev, "Failed to terminate asd\n");
 		goto out;
 	}
 
 	ret = psp_tmr_terminate(psp);
 	if (ret) {
-		DRM_ERROR("Failed to terminate tmr\n");
+		dev_err(adev->dev, "Failed to terminate tmr\n");
 		goto out;
 	}
 
 	ret = psp_ring_stop(psp, PSP_RING_TYPE__KM);
-	if (ret) {
-		DRM_ERROR("PSP ring stop failed\n");
-	}
+	if (ret)
+		dev_err(adev->dev, "PSP ring stop failed\n");
 
 out:
 	return ret;
 }
 
-static int psp_resume(void *handle)
+static int psp_resume(struct amdgpu_ip_block *ip_block)
 {
 	int ret;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct psp_context *psp = &adev->psp;
 
-	DRM_INFO("PSP is resuming...\n");
+	dev_info(adev->dev, "PSP is resuming...\n");
 
 	if (psp->mem_train_ctx.enable_mem_training) {
 		ret = psp_mem_training(psp, PSP_MEM_TRAIN_RESUME);
 		if (ret) {
-			DRM_ERROR("Failed to process memory training!\n");
+			dev_err(adev->dev, "Failed to process memory training!\n");
 			return ret;
 		}
 	}
 
 	mutex_lock(&adev->firmware.mutex);
 
+	ret = amdgpu_ucode_init_bo(adev);
+	if (ret)
+		goto failed;
+
 	ret = psp_hw_start(psp);
 	if (ret)
 		goto failed;
@@ -2752,7 +3342,7 @@ static int psp_resume(void *handle)
 
 	ret = psp_asd_initialize(psp);
 	if (ret) {
-		DRM_ERROR("PSP load asd failed!\n");
+		dev_err(adev->dev, "PSP load asd failed!\n");
 		goto failed;
 	}
 
@@ -2776,7 +3366,7 @@ static int psp_resume(void *handle)
 		ret = psp_ras_initialize(psp);
 		if (ret)
 			dev_err(psp->adev->dev,
-					"RAS: Failed to initialize RAS\n");
+				"RAS: Failed to initialize RAS\n");
 
 		ret = psp_hdcp_initialize(psp);
 		if (ret)
@@ -2804,7 +3394,7 @@ static int psp_resume(void *handle)
 	return 0;
 
 failed:
-	DRM_ERROR("PSP resume failed\n");
+	dev_err(adev->dev, "PSP resume failed\n");
 	mutex_unlock(&adev->firmware.mutex);
 	return ret;
 }
@@ -2838,19 +3428,6 @@ int psp_rlc_autoload_start(struct psp_context *psp)
 	return ret;
 }
 
-int psp_update_vcn_sram(struct amdgpu_device *adev, int inst_idx,
-			uint64_t cmd_gpu_addr, int cmd_size)
-{
-	struct amdgpu_firmware_info ucode = {0};
-
-	ucode.ucode_id = inst_idx ? AMDGPU_UCODE_ID_VCN1_RAM :
-		AMDGPU_UCODE_ID_VCN0_RAM;
-	ucode.mc_addr = cmd_gpu_addr;
-	ucode.ucode_size = cmd_size;
-
-	return psp_execute_non_psp_fw_load(&adev->psp, &ucode);
-}
-
 int psp_ring_cmd_submit(struct psp_context *psp,
 			uint64_t cmd_buf_mc_addr,
 			uint64_t fence_mc_addr,
@@ -2878,9 +3455,11 @@ int psp_ring_cmd_submit(struct psp_context *psp,
 		write_frame = ring_buffer_start + (psp_write_ptr_reg / rb_frame_size_dw);
 	/* Check invalid write_frame ptr address */
 	if ((write_frame < ring_buffer_start) || (ring_buffer_end < write_frame)) {
-		DRM_ERROR("ring_buffer_start = %p; ring_buffer_end = %p; write_frame = %p\n",
-			  ring_buffer_start, ring_buffer_end, write_frame);
-		DRM_ERROR("write_frame is pointing to address out of bounds\n");
+		dev_err(adev->dev,
+			"ring_buffer_start = %p; ring_buffer_end = %p; write_frame = %p\n",
+			ring_buffer_start, ring_buffer_end, write_frame);
+		dev_err(adev->dev,
+			"write_frame is pointing to address out of bounds\n");
 		return -EINVAL;
 	}
 
@@ -2904,12 +3483,11 @@ int psp_ring_cmd_submit(struct psp_context *psp,
 int psp_init_asd_microcode(struct psp_context *psp, const char *chip_name)
 {
 	struct amdgpu_device *adev = psp->adev;
-	char fw_name[PSP_FW_NAME_LEN];
 	const struct psp_firmware_header_v1_0 *asd_hdr;
 	int err = 0;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_asd.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->psp.asd_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->psp.asd_fw, AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_asd.bin", chip_name);
 	if (err)
 		goto out;
 
@@ -2928,12 +3506,11 @@ out:
 int psp_init_toc_microcode(struct psp_context *psp, const char *chip_name)
 {
 	struct amdgpu_device *adev = psp->adev;
-	char fw_name[PSP_FW_NAME_LEN];
 	const struct psp_firmware_header_v1_0 *toc_hdr;
 	int err = 0;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_toc.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->psp.toc_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->psp.toc_fw, AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_toc.bin", chip_name);
 	if (err)
 		goto out;
 
@@ -2967,7 +3544,7 @@ static int parse_sos_bin_descriptor(struct psp_context *psp,
 		psp->sos.fw_version        = le32_to_cpu(desc->fw_version);
 		psp->sos.feature_version   = le32_to_cpu(desc->fw_version);
 		psp->sos.size_bytes        = le32_to_cpu(desc->size_bytes);
-		psp->sos.start_addr 	   = ucode_start_addr;
+		psp->sos.start_addr	   = ucode_start_addr;
 		break;
 	case PSP_FW_TYPE_PSP_SYS_DRV:
 		psp->sys.fw_version        = le32_to_cpu(desc->fw_version);
@@ -3023,6 +3600,18 @@ static int parse_sos_bin_descriptor(struct psp_context *psp,
 		psp->ras_drv.size_bytes         = le32_to_cpu(desc->size_bytes);
 		psp->ras_drv.start_addr         = ucode_start_addr;
 		break;
+	case PSP_FW_TYPE_PSP_IPKEYMGR_DRV:
+		psp->ipkeymgr_drv.fw_version         = le32_to_cpu(desc->fw_version);
+		psp->ipkeymgr_drv.feature_version    = le32_to_cpu(desc->fw_version);
+		psp->ipkeymgr_drv.size_bytes         = le32_to_cpu(desc->size_bytes);
+		psp->ipkeymgr_drv.start_addr         = ucode_start_addr;
+		break;
+	case PSP_FW_TYPE_PSP_SPDM_DRV:
+		psp->spdm_drv.fw_version	= le32_to_cpu(desc->fw_version);
+		psp->spdm_drv.feature_version	= le32_to_cpu(desc->fw_version);
+		psp->spdm_drv.size_bytes	= le32_to_cpu(desc->size_bytes);
+		psp->spdm_drv.start_addr	= ucode_start_addr;
+		break;
 	default:
 		dev_warn(psp->adev->dev, "Unsupported PSP FW type: %d\n", desc->fw_type);
 		break;
@@ -3042,7 +3631,7 @@ static int psp_init_sos_base_fw(struct amdgpu_device *adev)
 		le32_to_cpu(sos_hdr->header.ucode_array_offset_bytes);
 
 	if (adev->gmc.xgmi.connected_to_cpu ||
-	    (adev->ip_versions[MP0_HWIP][0] != IP_VERSION(13, 0, 2))) {
+	    (amdgpu_ip_version(adev, MP0_HWIP, 0) != IP_VERSION(13, 0, 2))) {
 		adev->psp.sos.fw_version = le32_to_cpu(sos_hdr->header.ucode_version);
 		adev->psp.sos.feature_version = le32_to_cpu(sos_hdr->sos.fw_version);
 
@@ -3079,18 +3668,23 @@ static int psp_init_sos_base_fw(struct amdgpu_device *adev)
 int psp_init_sos_microcode(struct psp_context *psp, const char *chip_name)
 {
 	struct amdgpu_device *adev = psp->adev;
-	char fw_name[PSP_FW_NAME_LEN];
 	const struct psp_firmware_header_v1_0 *sos_hdr;
 	const struct psp_firmware_header_v1_1 *sos_hdr_v1_1;
 	const struct psp_firmware_header_v1_2 *sos_hdr_v1_2;
 	const struct psp_firmware_header_v1_3 *sos_hdr_v1_3;
 	const struct psp_firmware_header_v2_0 *sos_hdr_v2_0;
-	int err = 0;
+	const struct psp_firmware_header_v2_1 *sos_hdr_v2_1;
+	int fw_index, fw_bin_count, start_index = 0;
+	const struct psp_fw_bin_desc *fw_bin;
 	uint8_t *ucode_array_start_addr;
-	int fw_index = 0;
+	int err = 0;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sos.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->psp.sos_fw, fw_name);
+	if (amdgpu_is_kicker_fw(adev))
+		err = amdgpu_ucode_request(adev, &adev->psp.sos_fw, AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_sos_kicker.bin", chip_name);
+	else
+		err = amdgpu_ucode_request(adev, &adev->psp.sos_fw, AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_sos.bin", chip_name);
 	if (err)
 		goto out;
 
@@ -3139,15 +3733,30 @@ int psp_init_sos_microcode(struct psp_context *psp, const char *chip_name)
 	case 2:
 		sos_hdr_v2_0 = (const struct psp_firmware_header_v2_0 *)adev->psp.sos_fw->data;
 
-		if (le32_to_cpu(sos_hdr_v2_0->psp_fw_bin_count) >= UCODE_MAX_PSP_PACKAGING) {
+		fw_bin_count = le32_to_cpu(sos_hdr_v2_0->psp_fw_bin_count);
+
+		if (fw_bin_count >= UCODE_MAX_PSP_PACKAGING) {
 			dev_err(adev->dev, "packed SOS count exceeds maximum limit\n");
 			err = -EINVAL;
 			goto out;
 		}
 
-		for (fw_index = 0; fw_index < le32_to_cpu(sos_hdr_v2_0->psp_fw_bin_count); fw_index++) {
-			err = parse_sos_bin_descriptor(psp,
-						       &sos_hdr_v2_0->psp_fw_bin[fw_index],
+		if (sos_hdr_v2_0->header.header_version_minor == 1) {
+			sos_hdr_v2_1 = (const struct psp_firmware_header_v2_1 *)adev->psp.sos_fw->data;
+
+			fw_bin = sos_hdr_v2_1->psp_fw_bin;
+
+			if (psp_is_aux_sos_load_required(psp))
+				start_index = le32_to_cpu(sos_hdr_v2_1->psp_aux_fw_bin_index);
+			else
+				fw_bin_count -= le32_to_cpu(sos_hdr_v2_1->psp_aux_fw_bin_index);
+
+		} else {
+			fw_bin = sos_hdr_v2_0->psp_fw_bin;
+		}
+
+		for (fw_index = start_index; fw_index < fw_bin_count; fw_index++) {
+			err = parse_sos_bin_descriptor(psp, fw_bin + fw_index,
 						       sos_hdr_v2_0);
 			if (err)
 				goto out;
@@ -3167,6 +3776,36 @@ out:
 	return err;
 }
 
+static bool is_ta_fw_applicable(struct psp_context *psp,
+			     const struct psp_fw_bin_desc *desc)
+{
+	struct amdgpu_device *adev = psp->adev;
+	uint32_t fw_version;
+
+	switch (desc->fw_type) {
+	case TA_FW_TYPE_PSP_XGMI:
+	case TA_FW_TYPE_PSP_XGMI_AUX:
+		/* for now, AUX TA only exists on 13.0.6 ta bin,
+		 * from v20.00.0x.14
+		 */
+		if (amdgpu_ip_version(adev, MP0_HWIP, 0) ==
+		    IP_VERSION(13, 0, 6)) {
+			fw_version = le32_to_cpu(desc->fw_version);
+
+			if (adev->flags & AMD_IS_APU &&
+			    (fw_version & 0xff) >= 0x14)
+				return desc->fw_type == TA_FW_TYPE_PSP_XGMI_AUX;
+			else
+				return desc->fw_type == TA_FW_TYPE_PSP_XGMI;
+		}
+		break;
+	default:
+		break;
+	}
+
+	return true;
+}
+
 static int parse_ta_bin_descriptor(struct psp_context *psp,
 				   const struct psp_fw_bin_desc *desc,
 				   const struct ta_firmware_header_v2_0 *ta_hdr)
@@ -3176,6 +3815,9 @@ static int parse_ta_bin_descriptor(struct psp_context *psp,
 	if (!psp || !desc || !ta_hdr)
 		return -EINVAL;
 
+	if (!is_ta_fw_applicable(psp, desc))
+		return 0;
+
 	ucode_start_addr  = (uint8_t *)ta_hdr +
 			    le32_to_cpu(desc->offset_bytes) +
 			    le32_to_cpu(ta_hdr->header.ucode_array_offset_bytes);
@@ -3188,6 +3830,7 @@ static int parse_ta_bin_descriptor(struct psp_context *psp,
 		psp->asd_context.bin_desc.start_addr        = ucode_start_addr;
 		break;
 	case TA_FW_TYPE_PSP_XGMI:
+	case TA_FW_TYPE_PSP_XGMI_AUX:
 		psp->xgmi_context.context.bin_desc.fw_version       = le32_to_cpu(desc->fw_version);
 		psp->xgmi_context.context.bin_desc.size_bytes       = le32_to_cpu(desc->size_bytes);
 		psp->xgmi_context.context.bin_desc.start_addr       = ucode_start_addr;
@@ -3315,11 +3958,14 @@ int psp_init_ta_microcode(struct psp_context *psp, const char *chip_name)
 {
 	const struct common_firmware_header *hdr;
 	struct amdgpu_device *adev = psp->adev;
-	char fw_name[PSP_FW_NAME_LEN];
 	int err;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ta.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->psp.ta_fw, fw_name);
+	if (amdgpu_is_kicker_fw(adev))
+		err = amdgpu_ucode_request(adev, &adev->psp.ta_fw, AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_ta_kicker.bin", chip_name);
+	else
+		err = amdgpu_ucode_request(adev, &adev->psp.ta_fw, AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_ta.bin", chip_name);
 	if (err)
 		return err;
 
@@ -3345,7 +3991,6 @@ int psp_init_ta_microcode(struct psp_context *psp, const char *chip_name)
 int psp_init_cap_microcode(struct psp_context *psp, const char *chip_name)
 {
 	struct amdgpu_device *adev = psp->adev;
-	char fw_name[PSP_FW_NAME_LEN];
 	const struct psp_firmware_header_v1_0 *cap_hdr_v1_0;
 	struct amdgpu_firmware_info *info = NULL;
 	int err = 0;
@@ -3355,15 +4000,16 @@ int psp_init_cap_microcode(struct psp_context *psp, const char *chip_name)
 		return -EINVAL;
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_cap.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->psp.cap_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->psp.cap_fw, AMDGPU_UCODE_OPTIONAL,
+				   "amdgpu/%s_cap.bin", chip_name);
 	if (err) {
 		if (err == -ENODEV) {
 			dev_warn(adev->dev, "cap microcode does not exist, skip\n");
 			err = 0;
-			goto out;
+		} else {
+			dev_err(adev->dev, "fail to initialize cap microcode\n");
 		}
-		dev_err(adev->dev, "fail to initialize cap microcode\n");
+		goto out;
 	}
 
 	info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CAP];
@@ -3384,13 +4030,49 @@ out:
 	return err;
 }
 
-static int psp_set_clockgating_state(void *handle,
-				     enum amd_clockgating_state state)
+int psp_config_sq_perfmon(struct psp_context *psp,
+		uint32_t xcp_id, bool core_override_enable,
+		bool reg_override_enable, bool perfmon_override_enable)
+{
+	int ret;
+
+	if (amdgpu_sriov_vf(psp->adev))
+		return 0;
+
+	if (xcp_id > MAX_XCP) {
+		dev_err(psp->adev->dev, "invalid xcp_id %d\n", xcp_id);
+		return -EINVAL;
+	}
+
+	if (amdgpu_ip_version(psp->adev, MP0_HWIP, 0) != IP_VERSION(13, 0, 6)) {
+		dev_err(psp->adev->dev, "Unsupported MP0 version 0x%x for CONFIG_SQ_PERFMON command\n",
+			amdgpu_ip_version(psp->adev, MP0_HWIP, 0));
+		return -EINVAL;
+	}
+	struct psp_gfx_cmd_resp *cmd = acquire_psp_cmd_buf(psp);
+
+	cmd->cmd_id	=	GFX_CMD_ID_CONFIG_SQ_PERFMON;
+	cmd->cmd.config_sq_perfmon.gfx_xcp_mask	=	BIT_MASK(xcp_id);
+	cmd->cmd.config_sq_perfmon.core_override	=	core_override_enable;
+	cmd->cmd.config_sq_perfmon.reg_override	=	reg_override_enable;
+	cmd->cmd.config_sq_perfmon.perfmon_override = perfmon_override_enable;
+
+	ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr);
+	if (ret)
+		dev_warn(psp->adev->dev, "PSP failed to config sq: xcp%d core%d reg%d perfmon%d\n",
+			xcp_id, core_override_enable, reg_override_enable, perfmon_override_enable);
+
+	release_psp_cmd_buf(psp);
+	return ret;
+}
+
+static int psp_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int psp_set_powergating_state(void *handle,
+static int psp_set_powergating_state(struct amdgpu_ip_block *ip_block,
 				     enum amd_powergating_state state)
 {
 	return 0;
@@ -3402,11 +4084,13 @@ static ssize_t psp_usbc_pd_fw_sysfs_read(struct device *dev,
 {
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(ddev);
+	struct amdgpu_ip_block *ip_block;
 	uint32_t fw_ver;
 	int ret;
 
-	if (!adev->ip_blocks[AMD_IP_BLOCK_TYPE_PSP].status.late_initialized) {
-		DRM_INFO("PSP block is not ready yet.");
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_PSP);
+	if (!ip_block || !ip_block->status.late_initialized) {
+		dev_info(adev->dev, "PSP block is not ready yet\n.");
 		return -EBUSY;
 	}
 
@@ -3415,7 +4099,7 @@ static ssize_t psp_usbc_pd_fw_sysfs_read(struct device *dev,
 	mutex_unlock(&adev->psp.mutex);
 
 	if (ret) {
-		DRM_ERROR("Failed to read USBC PD FW, err = %d", ret);
+		dev_err(adev->dev, "Failed to read USBC PD FW, err = %d\n", ret);
 		return ret;
 	}
 
@@ -3430,22 +4114,23 @@ static ssize_t psp_usbc_pd_fw_sysfs_write(struct device *dev,
 	struct drm_device *ddev = dev_get_drvdata(dev);
 	struct amdgpu_device *adev = drm_to_adev(ddev);
 	int ret, idx;
-	char fw_name[100];
 	const struct firmware *usbc_pd_fw;
 	struct amdgpu_bo *fw_buf_bo = NULL;
 	uint64_t fw_pri_mc_addr;
 	void *fw_pri_cpu_addr;
+	struct amdgpu_ip_block *ip_block;
 
-	if (!adev->ip_blocks[AMD_IP_BLOCK_TYPE_PSP].status.late_initialized) {
-		DRM_INFO("PSP block is not ready yet.");
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_PSP);
+	if (!ip_block || !ip_block->status.late_initialized) {
+		dev_err(adev->dev, "PSP block is not ready yet.");
 		return -EBUSY;
 	}
 
 	if (!drm_dev_enter(ddev, &idx))
 		return -ENODEV;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s", buf);
-	ret = request_firmware(&usbc_pd_fw, fw_name, adev->dev);
+	ret = amdgpu_ucode_request(adev, &usbc_pd_fw, AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s", buf);
 	if (ret)
 		goto fail;
 
@@ -3467,10 +4152,10 @@ static ssize_t psp_usbc_pd_fw_sysfs_write(struct device *dev,
 	amdgpu_bo_free_kernel(&fw_buf_bo, &fw_pri_mc_addr, &fw_pri_cpu_addr);
 
 rel_buf:
-	release_firmware(usbc_pd_fw);
+	amdgpu_ucode_release(&usbc_pd_fw);
 fail:
 	if (ret) {
-		DRM_ERROR("Failed to load USBC PD FW, err = %d", ret);
+		dev_err(adev->dev, "Failed to load USBC PD FW, err = %d", ret);
 		count = ret;
 	}
 
@@ -3491,7 +4176,12 @@ void psp_copy_fw(struct psp_context *psp, uint8_t *start_addr, uint32_t bin_size
 	drm_dev_exit(idx);
 }
 
-static DEVICE_ATTR(usbc_pd_fw, S_IRUGO | S_IWUSR,
+/**
+ * DOC: usbc_pd_fw
+ * Reading from this file will retrieve the USB-C PD firmware version. Writing to
+ * this file will trigger the update process.
+ */
+static DEVICE_ATTR(usbc_pd_fw, 0644,
 		   psp_usbc_pd_fw_sysfs_read,
 		   psp_usbc_pd_fw_sysfs_write);
 
@@ -3501,7 +4191,7 @@ int is_psp_fw_valid(struct psp_bin_desc bin)
 }
 
 static ssize_t amdgpu_psp_vbflash_write(struct file *filp, struct kobject *kobj,
-					struct bin_attribute *bin_attr,
+					const struct bin_attribute *bin_attr,
 					char *buffer, loff_t pos, size_t count)
 {
 	struct device *dev = kobj_to_dev(kobj);
@@ -3512,7 +4202,7 @@ static ssize_t amdgpu_psp_vbflash_write(struct file *filp, struct kobject *kobj,
 
 	/* Safeguard against memory drain */
 	if (adev->psp.vbflash_image_size > AMD_VBIOS_FILE_MAX_SIZE_B) {
-		dev_err(adev->dev, "File size cannot exceed %u", AMD_VBIOS_FILE_MAX_SIZE_B);
+		dev_err(adev->dev, "File size cannot exceed %u\n", AMD_VBIOS_FILE_MAX_SIZE_B);
 		kvfree(adev->psp.vbflash_tmp_buf);
 		adev->psp.vbflash_tmp_buf = NULL;
 		adev->psp.vbflash_image_size = 0;
@@ -3531,13 +4221,13 @@ static ssize_t amdgpu_psp_vbflash_write(struct file *filp, struct kobject *kobj,
 	adev->psp.vbflash_image_size += count;
 	mutex_unlock(&adev->psp.mutex);
 
-	dev_info(adev->dev, "VBIOS flash write PSP done");
+	dev_dbg(adev->dev, "IFWI staged for update\n");
 
 	return count;
 }
 
 static ssize_t amdgpu_psp_vbflash_read(struct file *filp, struct kobject *kobj,
-				       struct bin_attribute *bin_attr, char *buffer,
+				       const struct bin_attribute *bin_attr, char *buffer,
 				       loff_t pos, size_t count)
 {
 	struct device *dev = kobj_to_dev(kobj);
@@ -3548,7 +4238,10 @@ static ssize_t amdgpu_psp_vbflash_read(struct file *filp, struct kobject *kobj,
 	void *fw_pri_cpu_addr;
 	int ret;
 
-	dev_info(adev->dev, "VBIOS flash to PSP started");
+	if (adev->psp.vbflash_image_size == 0)
+		return -EINVAL;
+
+	dev_dbg(adev->dev, "PSP IFWI flash process initiated\n");
 
 	ret = amdgpu_bo_create_kernel(adev, adev->psp.vbflash_image_size,
 					AMDGPU_GPU_PAGE_SIZE,
@@ -3573,14 +4266,32 @@ rel_buf:
 	adev->psp.vbflash_image_size = 0;
 
 	if (ret) {
-		dev_err(adev->dev, "Failed to load VBIOS FW, err = %d", ret);
+		dev_err(adev->dev, "Failed to load IFWI, err = %d\n", ret);
 		return ret;
 	}
 
-	dev_info(adev->dev, "VBIOS flash to PSP done");
+	dev_dbg(adev->dev, "PSP IFWI flash process done\n");
 	return 0;
 }
 
+/**
+ * DOC: psp_vbflash
+ * Writing to this file will stage an IFWI for update. Reading from this file
+ * will trigger the update process.
+ */
+static const struct bin_attribute psp_vbflash_bin_attr = {
+	.attr = {.name = "psp_vbflash", .mode = 0660},
+	.size = 0,
+	.write = amdgpu_psp_vbflash_write,
+	.read = amdgpu_psp_vbflash_read,
+};
+
+/**
+ * DOC: psp_vbflash_status
+ * The status of the flash process.
+ * 0: IFWI flash not complete.
+ * 1: IFWI flash complete.
+ */
 static ssize_t amdgpu_psp_vbflash_status(struct device *dev,
 					 struct device_attribute *attr,
 					 char *buf)
@@ -3597,84 +4308,167 @@ static ssize_t amdgpu_psp_vbflash_status(struct device *dev,
 
 	return sysfs_emit(buf, "0x%x\n", vbflash_status);
 }
+static DEVICE_ATTR(psp_vbflash_status, 0440, amdgpu_psp_vbflash_status, NULL);
 
-static const struct bin_attribute psp_vbflash_bin_attr = {
-	.attr = {.name = "psp_vbflash", .mode = 0664},
-	.size = 0,
-	.write = amdgpu_psp_vbflash_write,
-	.read = amdgpu_psp_vbflash_read,
+static const struct bin_attribute *const bin_flash_attrs[] = {
+	&psp_vbflash_bin_attr,
+	NULL
 };
 
-static DEVICE_ATTR(psp_vbflash_status, 0444, amdgpu_psp_vbflash_status, NULL);
+static struct attribute *flash_attrs[] = {
+	&dev_attr_psp_vbflash_status.attr,
+	&dev_attr_usbc_pd_fw.attr,
+	NULL
+};
 
-int amdgpu_psp_sysfs_init(struct amdgpu_device *adev)
+static umode_t amdgpu_flash_attr_is_visible(struct kobject *kobj, struct attribute *attr, int idx)
 {
-	int ret = 0;
-	struct psp_context *psp = &adev->psp;
+	struct device *dev = kobj_to_dev(kobj);
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
 
-	if (amdgpu_sriov_vf(adev))
-		return -EINVAL;
+	if (attr == &dev_attr_usbc_pd_fw.attr)
+		return adev->psp.sup_pd_fw_up ? 0660 : 0;
 
-	switch (adev->ip_versions[MP0_HWIP][0]) {
-	case IP_VERSION(13, 0, 0):
-	case IP_VERSION(13, 0, 7):
-		if (!psp->adev) {
-			psp->adev = adev;
-			psp_v13_0_set_psp_funcs(psp);
-		}
-		ret = sysfs_create_bin_file(&adev->dev->kobj, &psp_vbflash_bin_attr);
-		if (ret)
-			dev_err(adev->dev, "Failed to create device file psp_vbflash");
-		ret = device_create_file(adev->dev, &dev_attr_psp_vbflash_status);
-		if (ret)
-			dev_err(adev->dev, "Failed to create device file psp_vbflash_status");
-		return ret;
-	default:
-		return 0;
-	}
+	return adev->psp.sup_ifwi_up ? 0440 : 0;
 }
 
-const struct amd_ip_funcs psp_ip_funcs = {
-	.name = "psp",
-	.early_init = psp_early_init,
-	.late_init = NULL,
-	.sw_init = psp_sw_init,
-	.sw_fini = psp_sw_fini,
-	.hw_init = psp_hw_init,
-	.hw_fini = psp_hw_fini,
-	.suspend = psp_suspend,
-	.resume = psp_resume,
-	.is_idle = NULL,
-	.check_soft_reset = NULL,
-	.wait_for_idle = NULL,
-	.soft_reset = NULL,
-	.set_clockgating_state = psp_set_clockgating_state,
-	.set_powergating_state = psp_set_powergating_state,
+static umode_t amdgpu_bin_flash_attr_is_visible(struct kobject *kobj,
+						const struct bin_attribute *attr,
+						int idx)
+{
+	struct device *dev = kobj_to_dev(kobj);
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	return adev->psp.sup_ifwi_up ? 0660 : 0;
+}
+
+const struct attribute_group amdgpu_flash_attr_group = {
+	.attrs = flash_attrs,
+	.bin_attrs = bin_flash_attrs,
+	.is_bin_visible = amdgpu_bin_flash_attr_is_visible,
+	.is_visible = amdgpu_flash_attr_is_visible,
 };
 
-static int psp_sysfs_init(struct amdgpu_device *adev)
+#if defined(CONFIG_DEBUG_FS)
+static int psp_read_spirom_debugfs_open(struct inode *inode, struct file *filp)
 {
-	int ret = device_create_file(adev->dev, &dev_attr_usbc_pd_fw);
+	struct amdgpu_device *adev = filp->f_inode->i_private;
+	struct spirom_bo *bo_triplet;
+	int ret;
 
+	/* serialize the open() file calling */
+	if (!mutex_trylock(&adev->psp.mutex))
+		return -EBUSY;
+
+	/*
+	 * make sure only one userpace process is alive for dumping so that
+	 * only one memory buffer of AMD_VBIOS_FILE_MAX_SIZE * 2 is consumed.
+	 * let's say the case where one process try opening the file while
+	 * another one has proceeded to read or release. In this way, eliminate
+	 * the use of mutex for read() or release() callback as well.
+	 */
+	if (adev->psp.spirom_dump_trip) {
+		mutex_unlock(&adev->psp.mutex);
+		return -EBUSY;
+	}
+
+	bo_triplet = kzalloc(sizeof(struct spirom_bo), GFP_KERNEL);
+	if (!bo_triplet) {
+		mutex_unlock(&adev->psp.mutex);
+		return -ENOMEM;
+	}
+
+	ret = amdgpu_bo_create_kernel(adev, AMD_VBIOS_FILE_MAX_SIZE_B * 2,
+				      AMDGPU_GPU_PAGE_SIZE,
+				      AMDGPU_GEM_DOMAIN_GTT,
+				      &bo_triplet->bo,
+				      &bo_triplet->mc_addr,
+				      &bo_triplet->cpu_addr);
+	if (ret)
+		goto rel_trip;
+
+	ret = psp_dump_spirom(&adev->psp, bo_triplet->mc_addr);
 	if (ret)
-		DRM_ERROR("Failed to create USBC PD FW control file!");
+		goto rel_bo;
 
+	adev->psp.spirom_dump_trip = bo_triplet;
+	mutex_unlock(&adev->psp.mutex);
+	return 0;
+rel_bo:
+	amdgpu_bo_free_kernel(&bo_triplet->bo, &bo_triplet->mc_addr,
+			      &bo_triplet->cpu_addr);
+rel_trip:
+	kfree(bo_triplet);
+	mutex_unlock(&adev->psp.mutex);
+	dev_err(adev->dev, "Trying IFWI dump fails, err = %d\n", ret);
 	return ret;
 }
 
-void amdgpu_psp_sysfs_fini(struct amdgpu_device *adev)
+static ssize_t psp_read_spirom_debugfs_read(struct file *filp, char __user *buf, size_t size,
+					    loff_t *pos)
 {
-	sysfs_remove_bin_file(&adev->dev->kobj, &psp_vbflash_bin_attr);
-	device_remove_file(adev->dev, &dev_attr_psp_vbflash_status);
+	struct amdgpu_device *adev = filp->f_inode->i_private;
+	struct spirom_bo *bo_triplet = adev->psp.spirom_dump_trip;
+
+	if (!bo_triplet)
+		return -EINVAL;
+
+	return simple_read_from_buffer(buf,
+				       size,
+				       pos, bo_triplet->cpu_addr,
+				       AMD_VBIOS_FILE_MAX_SIZE_B * 2);
 }
 
-static void psp_sysfs_fini(struct amdgpu_device *adev)
+static int psp_read_spirom_debugfs_release(struct inode *inode, struct file *filp)
 {
-	device_remove_file(adev->dev, &dev_attr_usbc_pd_fw);
+	struct amdgpu_device *adev = filp->f_inode->i_private;
+	struct spirom_bo *bo_triplet = adev->psp.spirom_dump_trip;
+
+	if (bo_triplet) {
+		amdgpu_bo_free_kernel(&bo_triplet->bo, &bo_triplet->mc_addr,
+				      &bo_triplet->cpu_addr);
+		kfree(bo_triplet);
+	}
+
+	adev->psp.spirom_dump_trip = NULL;
+	return 0;
 }
 
-const struct amdgpu_ip_block_version psp_v3_1_ip_block =
+static const struct file_operations psp_dump_spirom_debugfs_ops = {
+	.owner = THIS_MODULE,
+	.open = psp_read_spirom_debugfs_open,
+	.read = psp_read_spirom_debugfs_read,
+	.release = psp_read_spirom_debugfs_release,
+	.llseek = default_llseek,
+};
+#endif
+
+void amdgpu_psp_debugfs_init(struct amdgpu_device *adev)
 {
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+
+	debugfs_create_file_size("psp_spirom_dump", 0444, minor->debugfs_root,
+				 adev, &psp_dump_spirom_debugfs_ops, AMD_VBIOS_FILE_MAX_SIZE_B * 2);
+#endif
+}
+
+const struct amd_ip_funcs psp_ip_funcs = {
+	.name = "psp",
+	.early_init = psp_early_init,
+	.sw_init = psp_sw_init,
+	.sw_fini = psp_sw_fini,
+	.hw_init = psp_hw_init,
+	.hw_fini = psp_hw_fini,
+	.suspend = psp_suspend,
+	.resume = psp_resume,
+	.set_clockgating_state = psp_set_clockgating_state,
+	.set_powergating_state = psp_set_powergating_state,
+};
+
+const struct amdgpu_ip_block_version psp_v3_1_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_PSP,
 	.major = 3,
 	.minor = 1,
@@ -3682,8 +4476,7 @@ const struct amdgpu_ip_block_version psp_v3_1_ip_block =
 	.funcs = &psp_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version psp_v10_0_ip_block =
-{
+const struct amdgpu_ip_block_version psp_v10_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_PSP,
 	.major = 10,
 	.minor = 0,
@@ -3691,8 +4484,7 @@ const struct amdgpu_ip_block_version psp_v10_0_ip_block =
 	.funcs = &psp_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version psp_v11_0_ip_block =
-{
+const struct amdgpu_ip_block_version psp_v11_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_PSP,
 	.major = 11,
 	.minor = 0,
@@ -3708,8 +4500,7 @@ const struct amdgpu_ip_block_version psp_v11_0_8_ip_block = {
 	.funcs = &psp_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version psp_v12_0_ip_block =
-{
+const struct amdgpu_ip_block_version psp_v12_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_PSP,
 	.major = 12,
 	.minor = 0,
@@ -3732,3 +4523,11 @@ const struct amdgpu_ip_block_version psp_v13_0_4_ip_block = {
 	.rev = 4,
 	.funcs = &psp_ip_funcs,
 };
+
+const struct amdgpu_ip_block_version psp_v14_0_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_PSP,
+	.major = 14,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &psp_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
index cf4f60c66122..237b624aa51c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h
@@ -39,6 +39,31 @@
 #define PSP_TMR_ALIGNMENT	0x100000
 #define PSP_FW_NAME_LEN		0x24
 
+/* VBIOS gfl defines */
+#define MBOX_READY_MASK 0x80000000
+#define MBOX_STATUS_MASK 0x0000FFFF
+#define MBOX_COMMAND_MASK 0x00FF0000
+#define MBOX_READY_FLAG 0x80000000
+#define C2PMSG_CMD_SPI_UPDATE_ROM_IMAGE_ADDR_LO 0x2
+#define C2PMSG_CMD_SPI_UPDATE_ROM_IMAGE_ADDR_HI 0x3
+#define C2PMSG_CMD_SPI_UPDATE_FLASH_IMAGE 0x4
+#define C2PMSG_CMD_SPI_GET_ROM_IMAGE_ADDR_LO 0xf
+#define C2PMSG_CMD_SPI_GET_ROM_IMAGE_ADDR_HI 0x10
+#define C2PMSG_CMD_SPI_GET_FLASH_IMAGE 0x11
+
+/* Command register bit 31 set to indicate readiness */
+#define MBOX_TOS_READY_FLAG (GFX_FLAG_RESPONSE)
+#define MBOX_TOS_READY_MASK (GFX_CMD_RESPONSE_MASK | GFX_CMD_STATUS_MASK)
+
+/* Values to check for a successful GFX_CMD response wait. Check against
+ * both status bits and response state - helps to detect a command failure
+ * or other unexpected cases like a device drop reading all 0xFFs
+ */
+#define MBOX_TOS_RESP_FLAG (GFX_FLAG_RESPONSE)
+#define MBOX_TOS_RESP_MASK (GFX_CMD_RESPONSE_MASK | GFX_CMD_STATUS_MASK)
+
+extern const struct attribute_group amdgpu_flash_attr_group;
+
 enum psp_shared_mem_size {
 	PSP_ASD_SHARED_MEM_SIZE				= 0x0,
 	PSP_XGMI_SHARED_MEM_SIZE			= 0x4000,
@@ -71,15 +96,17 @@ enum psp_bootloader_cmd {
 	PSP_BL__LOAD_KEY_DATABASE	= 0x80000,
 	PSP_BL__LOAD_SOCDRV             = 0xB0000,
 	PSP_BL__LOAD_DBGDRV             = 0xC0000,
+	PSP_BL__LOAD_HADDRV		= PSP_BL__LOAD_DBGDRV,
 	PSP_BL__LOAD_INTFDRV		= 0xD0000,
-	PSP_BL__LOAD_RASDRV		    = 0xE0000,
+	PSP_BL__LOAD_RASDRV		= 0xE0000,
+	PSP_BL__LOAD_IPKEYMGRDRV	= 0xF0000,
 	PSP_BL__DRAM_LONG_TRAIN		= 0x100000,
 	PSP_BL__DRAM_SHORT_TRAIN	= 0x200000,
 	PSP_BL__LOAD_TOS_SPL_TABLE	= 0x10000000,
+	PSP_BL__LOAD_SPDMDRV		= 0x20000000,
 };
 
-enum psp_ring_type
-{
+enum psp_ring_type {
 	PSP_RING_TYPE__INVALID = 0,
 	/*
 	 * These values map to the way the PSP kernel identifies the
@@ -89,8 +116,7 @@ enum psp_ring_type
 	PSP_RING_TYPE__KM = 2  /* Kernel mode ring (formerly called GPCOM) */
 };
 
-struct psp_ring
-{
+struct psp_ring {
 	enum psp_ring_type		ring_type;
 	struct psp_gfx_rb_frame		*ring_mem;
 	uint64_t			ring_mem_mc_addr;
@@ -104,12 +130,16 @@ enum psp_reg_prog_id {
 	PSP_REG_IH_RB_CNTL        = 0,  /* register IH_RB_CNTL */
 	PSP_REG_IH_RB_CNTL_RING1  = 1,  /* register IH_RB_CNTL_RING1 */
 	PSP_REG_IH_RB_CNTL_RING2  = 2,  /* register IH_RB_CNTL_RING2 */
+	PSP_REG_MMHUB_L1_TLB_CNTL = 25,
 	PSP_REG_LAST
 };
 
-struct psp_funcs
-{
+#define PSP_WAITREG_CHANGED BIT(0) /* check if the value has changed */
+#define PSP_WAITREG_NOVERBOSE BIT(1) /* No error verbose */
+
+struct psp_funcs {
 	int (*init_microcode)(struct psp_context *psp);
+	int (*wait_for_bootloader)(struct psp_context *psp);
 	int (*bootloader_load_kdb)(struct psp_context *psp);
 	int (*bootloader_load_spl)(struct psp_context *psp);
 	int (*bootloader_load_sysdrv)(struct psp_context *psp);
@@ -117,6 +147,8 @@ struct psp_funcs
 	int (*bootloader_load_intf_drv)(struct psp_context *psp);
 	int (*bootloader_load_dbg_drv)(struct psp_context *psp);
 	int (*bootloader_load_ras_drv)(struct psp_context *psp);
+	int (*bootloader_load_ipkeymgr_drv)(struct psp_context *psp);
+	int (*bootloader_load_spdm_drv)(struct psp_context *psp);
 	int (*bootloader_load_sos)(struct psp_context *psp);
 	int (*ring_create)(struct psp_context *psp,
 			   enum psp_ring_type ring_type);
@@ -132,7 +164,14 @@ struct psp_funcs
 	int (*load_usbc_pd_fw)(struct psp_context *psp, uint64_t fw_pri_mc_addr);
 	int (*read_usbc_pd_fw)(struct psp_context *psp, uint32_t *fw_ver);
 	int (*update_spirom)(struct psp_context *psp, uint64_t fw_pri_mc_addr);
+	int (*dump_spirom)(struct psp_context *psp, uint64_t fw_pri_mc_addr);
 	int (*vbflash_stat)(struct psp_context *psp);
+	int (*fatal_error_recovery_quirk)(struct psp_context *psp);
+	bool (*get_ras_capability)(struct psp_context *psp);
+	bool (*is_aux_sos_load_required)(struct psp_context *psp);
+	bool (*is_reload_needed)(struct psp_context *psp);
+	int (*reg_program_no_ring)(struct psp_context *psp, uint32_t val,
+				   enum psp_reg_prog_id id);
 };
 
 struct ta_funcs {
@@ -148,6 +187,7 @@ struct psp_xgmi_node_info {
 	uint8_t					is_sharing_enabled;
 	enum ta_xgmi_assigned_sdma_engine	sdma_engine;
 	uint8_t					num_links;
+	struct xgmi_connected_port_num		port_num[TA_XGMI__MAX_PORT_NUM];
 };
 
 struct psp_xgmi_topology_info {
@@ -188,18 +228,20 @@ struct psp_xgmi_context {
 	struct ta_context		context;
 	struct psp_xgmi_topology_info	top_info;
 	bool				supports_extended_data;
+	uint8_t				xgmi_ta_caps;
 };
 
 struct psp_ras_context {
 	struct ta_context		context;
 	struct amdgpu_ras		*ras;
+	struct mutex			mutex;
 };
 
 #define MEM_TRAIN_SYSTEM_SIGNATURE		0x54534942
 #define GDDR6_MEM_TRAINING_DATA_SIZE_IN_BYTES	0x1000
 #define GDDR6_MEM_TRAINING_OFFSET		0x8000
 /*Define the VRAM size that will be encroached by BIST training.*/
-#define GDDR6_MEM_TRAINING_ENCROACHED_SIZE	0x2000000
+#define BIST_MEM_TRAINING_ENCROACHED_SIZE	0x2000000
 
 enum psp_memory_training_init_flag {
 	PSP_MEM_TRAIN_NOT_SUPPORT	= 0x0,
@@ -307,10 +349,17 @@ struct psp_runtime_scpm_entry {
 	enum psp_runtime_scpm_authentication scpm_status;
 };
 
-struct psp_context
-{
-	struct amdgpu_device            *adev;
-	struct psp_ring                 km_ring;
+#if defined(CONFIG_DEBUG_FS)
+struct spirom_bo {
+	struct amdgpu_bo *bo;
+	uint64_t mc_addr;
+	void *cpu_addr;
+};
+#endif
+
+struct psp_context {
+	struct amdgpu_device		*adev;
+	struct psp_ring			km_ring;
 	struct psp_gfx_cmd_resp		*cmd;
 
 	const struct psp_funcs		*funcs;
@@ -333,13 +382,15 @@ struct psp_context
 	struct psp_bin_desc		intf_drv;
 	struct psp_bin_desc		dbg_drv;
 	struct psp_bin_desc		ras_drv;
+	struct psp_bin_desc		ipkeymgr_drv;
+	struct psp_bin_desc		spdm_drv;
 
 	/* tmr buffer */
 	struct amdgpu_bo		*tmr_bo;
 	uint64_t			tmr_mc_addr;
 
 	/* asd firmware */
-	const struct firmware	*asd_fw;
+	const struct firmware		*asd_fw;
 
 	/* toc firmware */
 	const struct firmware		*toc_fw;
@@ -361,6 +412,8 @@ struct psp_context
 	atomic_t			fence_value;
 	/* flag to mark whether gfx fw autoload is supported or not */
 	bool				autoload_supported;
+	/* flag to mark whether psp use runtime TMR or boottime TMR */
+	bool				boot_time_tmr;
 	/* flag to mark whether df cstate management centralized to PMFW */
 	bool				pmfw_centralized_cstate_management;
 
@@ -384,9 +437,16 @@ struct psp_context
 
 	uint32_t			boot_cfg_bitmask;
 
-	char *vbflash_tmp_buf;
-	size_t vbflash_image_size;
-	bool vbflash_done;
+	/* firmware upgrades supported */
+	bool				sup_pd_fw_up;
+	bool				sup_ifwi_up;
+
+	char				*vbflash_tmp_buf;
+	size_t				vbflash_image_size;
+	bool				vbflash_done;
+#if defined(CONFIG_DEBUG_FS)
+	struct spirom_bo *spirom_dump_trip;
+#endif
 };
 
 struct amdgpu_psp_funcs {
@@ -415,6 +475,12 @@ struct amdgpu_psp_funcs {
 #define psp_bootloader_load_ras_drv(psp) \
 		((psp)->funcs->bootloader_load_ras_drv ? \
 		(psp)->funcs->bootloader_load_ras_drv((psp)) : 0)
+#define psp_bootloader_load_ipkeymgr_drv(psp) \
+		((psp)->funcs->bootloader_load_ipkeymgr_drv ? \
+		 (psp)->funcs->bootloader_load_ipkeymgr_drv((psp)) : 0)
+#define psp_bootloader_load_spdm_drv(psp) \
+		((psp)->funcs->bootloader_load_spdm_drv ? \
+		 (psp)->funcs->bootloader_load_spdm_drv((psp)) : 0)
 #define psp_bootloader_load_sos(psp) \
 		((psp)->funcs->bootloader_load_sos ? (psp)->funcs->bootloader_load_sos((psp)) : 0)
 #define psp_smu_reload_quirk(psp) \
@@ -439,10 +505,25 @@ struct amdgpu_psp_funcs {
 	((psp)->funcs->update_spirom ? \
 	(psp)->funcs->update_spirom((psp), fw_pri_mc_addr) : -EINVAL)
 
+#define psp_dump_spirom(psp, fw_pri_mc_addr) \
+	((psp)->funcs->dump_spirom ? \
+	(psp)->funcs->dump_spirom((psp), fw_pri_mc_addr) : -EINVAL)
+
 #define psp_vbflash_status(psp) \
 	((psp)->funcs->vbflash_stat ? \
 	(psp)->funcs->vbflash_stat((psp)) : -EINVAL)
 
+#define psp_fatal_error_recovery_quirk(psp) \
+	((psp)->funcs->fatal_error_recovery_quirk ? \
+	(psp)->funcs->fatal_error_recovery_quirk((psp)) : 0)
+
+#define psp_is_aux_sos_load_required(psp) \
+	((psp)->funcs->is_aux_sos_load_required ? (psp)->funcs->is_aux_sos_load_required((psp)) : 0)
+
+#define psp_reg_program_no_ring(psp, val, id) \
+	((psp)->funcs->reg_program_no_ring ? \
+	(psp)->funcs->reg_program_no_ring((psp), val, id) : -EINVAL)
+
 extern const struct amd_ip_funcs psp_ip_funcs;
 
 extern const struct amdgpu_ip_block_version psp_v3_1_ip_block;
@@ -452,13 +533,17 @@ extern const struct amdgpu_ip_block_version psp_v11_0_8_ip_block;
 extern const struct amdgpu_ip_block_version psp_v12_0_ip_block;
 extern const struct amdgpu_ip_block_version psp_v13_0_ip_block;
 extern const struct amdgpu_ip_block_version psp_v13_0_4_ip_block;
+extern const struct amdgpu_ip_block_version psp_v14_0_ip_block;
+
+int psp_wait_for(struct psp_context *psp, uint32_t reg_index,
+		 uint32_t field_val, uint32_t mask, uint32_t flags);
+extern int psp_wait_for_spirom_update(struct psp_context *psp, uint32_t reg_index,
+			uint32_t field_val, uint32_t mask, uint32_t msec_timeout);
 
-extern int psp_wait_for(struct psp_context *psp, uint32_t reg_index,
-			uint32_t field_val, uint32_t mask, bool check_changed);
+int psp_execute_ip_fw_load(struct psp_context *psp,
+			   struct amdgpu_firmware_info *ucode);
 
 int psp_gpu_reset(struct amdgpu_device *adev);
-int psp_update_vcn_sram(struct amdgpu_device *adev, int inst_idx,
-			uint64_t cmd_gpu_addr, int cmd_size);
 
 int psp_ta_init_shared_buf(struct psp_context *psp,
 				  struct ta_mem_context *mem_ctx);
@@ -486,8 +571,11 @@ int psp_ras_invoke(struct psp_context *psp, uint32_t ta_cmd_id);
 int psp_ras_enable_features(struct psp_context *psp,
 		union ta_ras_cmd_input *info, bool enable);
 int psp_ras_trigger_error(struct psp_context *psp,
-			  struct ta_ras_trigger_error_input *info);
+			  struct ta_ras_trigger_error_input *info, uint32_t instance_mask);
 int psp_ras_terminate(struct psp_context *psp);
+int psp_ras_query_address(struct psp_context *psp,
+			  struct ta_ras_query_address_input *addr_in,
+			  struct ta_ras_query_address_output *addr_out);
 
 int psp_hdcp_invoke(struct psp_context *psp, uint32_t ta_cmd_id);
 int psp_dtm_invoke(struct psp_context *psp, uint32_t ta_cmd_id);
@@ -514,13 +602,25 @@ int psp_init_cap_microcode(struct psp_context *psp,
 			  const char *chip_name);
 int psp_get_fw_attestation_records_addr(struct psp_context *psp,
 					uint64_t *output_ptr);
-
+int psp_update_fw_reservation(struct psp_context *psp);
 int psp_load_fw_list(struct psp_context *psp,
 		     struct amdgpu_firmware_info **ucode_list, int ucode_count);
 void psp_copy_fw(struct psp_context *psp, uint8_t *start_addr, uint32_t bin_size);
 
+int psp_spatial_partition(struct psp_context *psp, int mode);
+int psp_memory_partition(struct psp_context *psp, int mode);
+
 int is_psp_fw_valid(struct psp_bin_desc bin);
 
-int amdgpu_psp_sysfs_init(struct amdgpu_device *adev);
-void amdgpu_psp_sysfs_fini(struct amdgpu_device *adev);
+int amdgpu_psp_wait_for_bootloader(struct amdgpu_device *adev);
+bool amdgpu_psp_get_ras_capability(struct psp_context *psp);
+
+int psp_config_sq_perfmon(struct psp_context *psp, uint32_t xcp_id,
+	bool core_override_enable, bool reg_override_enable, bool perfmon_override_enable);
+bool amdgpu_psp_tos_reload_needed(struct amdgpu_device *adev);
+int amdgpu_psp_reg_program_no_ring(struct psp_context *psp, uint32_t val,
+				   enum psp_reg_prog_id id);
+void amdgpu_psp_debugfs_init(struct amdgpu_device *adev);
+
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c
index 468a67b302d4..6e8aad91bcd3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c
@@ -166,15 +166,14 @@ static ssize_t ta_if_load_debugfs_write(struct file *fp, const char *buf, size_t
 	if (ret)
 		return -EFAULT;
 
+	if (ta_bin_len > PSP_1_MEG)
+		return -EINVAL;
+
 	copy_pos += sizeof(uint32_t);
 
-	ta_bin = kzalloc(ta_bin_len, GFP_KERNEL);
-	if (!ta_bin)
-		return -ENOMEM;
-	if (copy_from_user((void *)ta_bin, &buf[copy_pos], ta_bin_len)) {
-		ret = -EFAULT;
-		goto err_free_bin;
-	}
+	ta_bin = memdup_user(&buf[copy_pos], ta_bin_len);
+	if (IS_ERR(ta_bin))
+		return PTR_ERR(ta_bin);
 
 	/* Set TA context and functions */
 	set_ta_context_funcs(psp, ta_type, &context);
@@ -324,17 +323,13 @@ static ssize_t ta_if_invoke_debugfs_write(struct file *fp, const char *buf, size
 		return -EFAULT;
 	copy_pos += sizeof(uint32_t);
 
-	shared_buf = kzalloc(shared_buf_len, GFP_KERNEL);
-	if (!shared_buf)
-		return -ENOMEM;
-	if (copy_from_user((void *)shared_buf, &buf[copy_pos], shared_buf_len)) {
-		ret = -EFAULT;
-		goto err_free_shared_buf;
-	}
+	shared_buf = memdup_user(&buf[copy_pos], shared_buf_len);
+	if (IS_ERR(shared_buf))
+		return PTR_ERR(shared_buf);
 
 	set_ta_context_funcs(psp, ta_type, &context);
 
-	if (!context->initialized) {
+	if (!context || !context->initialized) {
 		dev_err(adev->dev, "TA is not initialized\n");
 		ret = -EINVAL;
 		goto err_free_shared_buf;
@@ -348,6 +343,7 @@ static ssize_t ta_if_invoke_debugfs_write(struct file *fp, const char *buf, size
 
 	context->session_id = ta_id;
 
+	mutex_lock(&psp->ras_context.mutex);
 	ret = prep_ta_mem_context(&context->mem_context, shared_buf, shared_buf_len);
 	if (ret)
 		goto err_free_shared_buf;
@@ -362,10 +358,11 @@ static ssize_t ta_if_invoke_debugfs_write(struct file *fp, const char *buf, size
 		}
 	}
 
-	if (copy_to_user((char *)buf, context->mem_context.shared_buf, shared_buf_len))
+	if (copy_to_user((char *)&buf[copy_pos], context->mem_context.shared_buf, shared_buf_len))
 		ret = -EFAULT;
 
 err_free_shared_buf:
+	mutex_unlock(&psp->ras_context.mutex);
 	kfree(shared_buf);
 
 	return ret;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_rap.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_rap.c
index 12010c988c8b..bacf888735db 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_rap.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_rap.c
@@ -101,7 +101,6 @@ static ssize_t amdgpu_rap_debugfs_write(struct file *f, const char __user *buf,
 	}
 
 	amdgpu_gfx_off_ctrl(adev, true);
-	pm_runtime_mark_last_busy(dev->dev);
 	pm_runtime_put_autosuspend(dev->dev);
 
 	return size;
@@ -116,7 +115,6 @@ static const struct file_operations amdgpu_rap_debugfs_ops = {
 
 void amdgpu_rap_debugfs_init(struct amdgpu_device *adev)
 {
-#if defined(CONFIG_DEBUG_FS)
 	struct drm_minor *minor = adev_to_drm(adev)->primary;
 
 	if (!adev->psp.rap_context.context.initialized)
@@ -124,5 +122,4 @@ void amdgpu_rap_debugfs_init(struct amdgpu_device *adev)
 
 	debugfs_create_file("rap_test", S_IWUSR, minor->debugfs_root,
 				adev, &amdgpu_rap_debugfs_ops);
-#endif
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
index 4069bce9479f..2a6cf7963dde 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
@@ -28,6 +28,7 @@
 #include <linux/reboot.h>
 #include <linux/syscalls.h>
 #include <linux/pm_runtime.h>
+#include <linux/list_sort.h>
 
 #include "amdgpu.h"
 #include "amdgpu_ras.h"
@@ -35,8 +36,12 @@
 #include "amdgpu_xgmi.h"
 #include "ivsrcid/nbio/irqsrcs_nbif_7_4.h"
 #include "nbio_v4_3.h"
+#include "nbif_v6_3_1.h"
+#include "nbio_v7_9.h"
 #include "atom.h"
 #include "amdgpu_reset.h"
+#include "amdgpu_psp.h"
+#include "amdgpu_ras_mgr.h"
 
 #ifdef CONFIG_X86_MCE_AMD
 #include <asm/mce.h>
@@ -71,6 +76,9 @@ const char *ras_block_string[] = {
 	"mca",
 	"vcn",
 	"jpeg",
+	"ih",
+	"mpio",
+	"mmsch",
 };
 
 const char *ras_mca_block_string[] = {
@@ -92,7 +100,8 @@ const char *get_ras_block_str(struct ras_common_if *ras_block)
 	if (!ras_block)
 		return "NULL";
 
-	if (ras_block->block >= AMDGPU_RAS_BLOCK_COUNT)
+	if (ras_block->block >= AMDGPU_RAS_BLOCK_COUNT ||
+	    ras_block->block >= ARRAY_SIZE(ras_block_string))
 		return "OUT OF RANGE";
 
 	if (ras_block->block == AMDGPU_RAS_BLOCK__MCA)
@@ -114,6 +123,15 @@ const char *get_ras_block_str(struct ras_common_if *ras_block)
 /* typical ECC bad page rate is 1 bad page per 100MB VRAM */
 #define RAS_BAD_PAGE_COVER              (100 * 1024 * 1024ULL)
 
+#define MAX_UMC_POISON_POLLING_TIME_ASYNC  10
+
+#define AMDGPU_RAS_RETIRE_PAGE_INTERVAL 100  //ms
+
+#define MAX_FLUSH_RETIRE_DWORK_TIMES  100
+
+#define BYPASS_ALLOCATED_ADDRESS        0x0
+#define BYPASS_INITIALIZATION_ADDRESS   0x1
+
 enum amdgpu_ras_retire_page_reservation {
 	AMDGPU_RAS_RETIRE_PAGE_RESERVED,
 	AMDGPU_RAS_RETIRE_PAGE_PENDING,
@@ -122,12 +140,18 @@ enum amdgpu_ras_retire_page_reservation {
 
 atomic_t amdgpu_ras_in_intr = ATOMIC_INIT(0);
 
-static bool amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con,
+static int amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con,
 				uint64_t addr);
-static bool amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
+static int amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
 				uint64_t addr);
+
+static void amdgpu_ras_critical_region_init(struct amdgpu_device *adev);
+static void amdgpu_ras_critical_region_fini(struct amdgpu_device *adev);
+
 #ifdef CONFIG_X86_MCE_AMD
 static void amdgpu_register_bad_pages_mca_notifier(struct amdgpu_device *adev);
+static void
+amdgpu_unregister_bad_pages_mca_notifier(struct amdgpu_device *adev);
 struct mce_notifier_adev_list {
 	struct amdgpu_device *devs[MAX_GPU_INSTANCE];
 	int num_gpu;
@@ -151,35 +175,39 @@ static bool amdgpu_ras_get_error_query_ready(struct amdgpu_device *adev)
 
 static int amdgpu_reserve_page_direct(struct amdgpu_device *adev, uint64_t address)
 {
-	struct ras_err_data err_data = {0, 0, 0, NULL};
+	struct ras_err_data err_data;
 	struct eeprom_table_record err_rec;
+	int ret;
 
-	if ((address >= adev->gmc.mc_vram_size) ||
-	    (address >= RAS_UMC_INJECT_ADDR_LIMIT)) {
+	ret = amdgpu_ras_check_bad_page(adev, address);
+	if (ret == -EINVAL) {
 		dev_warn(adev->dev,
-		         "RAS WARN: input address 0x%llx is invalid.\n",
-		         address);
+			"RAS WARN: input address 0x%llx is invalid.\n",
+			address);
 		return -EINVAL;
-	}
-
-	if (amdgpu_ras_check_bad_page(adev, address)) {
+	} else if (ret == 1) {
 		dev_warn(adev->dev,
-			 "RAS WARN: 0x%llx has already been marked as bad page!\n",
-			 address);
+			"RAS WARN: 0x%llx has already been marked as bad page!\n",
+			address);
 		return 0;
 	}
 
+	ret = amdgpu_ras_error_data_init(&err_data);
+	if (ret)
+		return ret;
+
 	memset(&err_rec, 0x0, sizeof(struct eeprom_table_record));
 	err_data.err_addr = &err_rec;
-	amdgpu_umc_fill_error_record(&err_data, address,
-			(address >> AMDGPU_GPU_PAGE_SHIFT), 0, 0);
+	amdgpu_umc_fill_error_record(&err_data, address, address, 0, 0);
 
 	if (amdgpu_bad_page_threshold != 0) {
 		amdgpu_ras_add_bad_pages(adev, err_data.err_addr,
-					 err_data.err_addr_cnt);
+					 err_data.err_addr_cnt, false);
 		amdgpu_ras_save_bad_pages(adev, NULL);
 	}
 
+	amdgpu_ras_error_data_fini(&err_data);
+
 	dev_warn(adev->dev, "WARNING: THIS IS ONLY FOR TEST PURPOSES AND WILL CORRUPT RAS EEPROM\n");
 	dev_warn(adev->dev, "Clear EEPROM:\n");
 	dev_warn(adev->dev, "    echo 1 > /sys/kernel/debug/dri/0/ras/ras_eeprom_reset\n");
@@ -187,6 +215,56 @@ static int amdgpu_reserve_page_direct(struct amdgpu_device *adev, uint64_t addre
 	return 0;
 }
 
+static int amdgpu_check_address_validity(struct amdgpu_device *adev,
+			uint64_t address, uint64_t flags)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct amdgpu_vram_block_info blk_info;
+	uint64_t page_pfns[32] = {0};
+	int i, ret, count;
+	bool hit = false;
+
+	if (amdgpu_ip_version(adev, UMC_HWIP, 0) < IP_VERSION(12, 0, 0))
+		return 0;
+
+	if (amdgpu_sriov_vf(adev)) {
+		if (amdgpu_virt_check_vf_critical_region(adev, address, &hit))
+			return -EPERM;
+		return hit ? -EACCES : 0;
+	}
+
+	if ((address >= adev->gmc.mc_vram_size) ||
+	    (address >= RAS_UMC_INJECT_ADDR_LIMIT))
+		return -EFAULT;
+
+	count = amdgpu_umc_lookup_bad_pages_in_a_row(adev,
+				address, page_pfns, ARRAY_SIZE(page_pfns));
+	if (count <= 0)
+		return -EPERM;
+
+	for (i = 0; i < count; i++) {
+		memset(&blk_info, 0, sizeof(blk_info));
+		ret = amdgpu_vram_mgr_query_address_block_info(&adev->mman.vram_mgr,
+					page_pfns[i] << AMDGPU_GPU_PAGE_SHIFT, &blk_info);
+		if (!ret) {
+			/* The input address that needs to be checked is allocated by
+			 * current calling process, so it is necessary to exclude
+			 * the calling process.
+			 */
+			if ((flags == BYPASS_ALLOCATED_ADDRESS) &&
+			    ((blk_info.task.pid != task_pid_nr(current)) ||
+				strncmp(blk_info.task.comm, current->comm, TASK_COMM_LEN)))
+				return -EACCES;
+			else if ((flags == BYPASS_INITIALIZATION_ADDRESS) &&
+				(blk_info.task.pid == con->init_task_pid) &&
+				!strncmp(blk_info.task.comm, con->init_task_comm, TASK_COMM_LEN))
+				return -EACCES;
+		}
+	}
+
+	return 0;
+}
+
 static ssize_t amdgpu_ras_debugfs_read(struct file *f, char __user *buf,
 					size_t size, loff_t *pos)
 {
@@ -201,8 +279,8 @@ static ssize_t amdgpu_ras_debugfs_read(struct file *f, char __user *buf,
 		return -EINVAL;
 
 	/* Hardware counter will be reset automatically after the query on Vega20 and Arcturus */
-	if (obj->adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 2) &&
-	    obj->adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 4)) {
+	if (amdgpu_ip_version(obj->adev, MP0_HWIP, 0) != IP_VERSION(11, 0, 2) &&
+	    amdgpu_ip_version(obj->adev, MP0_HWIP, 0) != IP_VERSION(11, 0, 4)) {
 		if (amdgpu_ras_reset_error_status(obj->adev, info.head.block))
 			dev_warn(obj->adev->dev, "Failed to reset error counter and error status");
 	}
@@ -256,6 +334,8 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
 	int block_id;
 	uint32_t sub_block;
 	u64 address, value;
+	/* default value is 0 if the mask is not set by user */
+	u32 instance_mask = 0;
 
 	if (*pos)
 		return -EINVAL;
@@ -275,6 +355,8 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
 		op = 2;
 	else if (strstr(str, "retire_page") != NULL)
 		op = 3;
+	else if (strstr(str, "check_address") != NULL)
+		op = 4;
 	else if (str[0] && str[1] && str[2] && str[3])
 		/* ascii string, but commands are not matched. */
 		return -EINVAL;
@@ -289,24 +371,39 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
 			data->inject.address = address;
 
 			return 0;
+		} else if (op == 4) {
+			if (sscanf(str, "%*s 0x%llx 0x%llx", &address, &value) != 2 &&
+			    sscanf(str, "%*s %llu %llu", &address, &value) != 2)
+				return -EINVAL;
+
+			data->op = op;
+			data->inject.address = address;
+			data->inject.value = value;
+			return 0;
 		}
 
 		if (amdgpu_ras_find_block_id_by_name(block_name, &block_id))
 			return -EINVAL;
 
 		data->head.block = block_id;
-		/* only ue and ce errors are supported */
+		/* only ue, ce and poison errors are supported */
 		if (!memcmp("ue", err, 2))
 			data->head.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE;
 		else if (!memcmp("ce", err, 2))
 			data->head.type = AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE;
+		else if (!memcmp("poison", err, 6))
+			data->head.type = AMDGPU_RAS_ERROR__POISON;
 		else
 			return -EINVAL;
 
 		data->op = op;
 
 		if (op == 2) {
-			if (sscanf(str, "%*s %*s %*s 0x%x 0x%llx 0x%llx",
+			if (sscanf(str, "%*s %*s %*s 0x%x 0x%llx 0x%llx 0x%x",
+				   &sub_block, &address, &value, &instance_mask) != 4 &&
+			    sscanf(str, "%*s %*s %*s %u %llu %llu %u",
+				   &sub_block, &address, &value, &instance_mask) != 4 &&
+				sscanf(str, "%*s %*s %*s 0x%x 0x%llx 0x%llx",
 				   &sub_block, &address, &value) != 3 &&
 			    sscanf(str, "%*s %*s %*s %u %llu %llu",
 				   &sub_block, &address, &value) != 3)
@@ -314,6 +411,7 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
 			data->head.sub_block_index = sub_block;
 			data->inject.address = address;
 			data->inject.value = value;
+			data->inject.instance_mask = instance_mask;
 		}
 	} else {
 		if (size < sizeof(*data))
@@ -326,6 +424,46 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
 	return 0;
 }
 
+static void amdgpu_ras_instance_mask_check(struct amdgpu_device *adev,
+				struct ras_debug_if *data)
+{
+	int num_xcc = adev->gfx.xcc_mask ? NUM_XCC(adev->gfx.xcc_mask) : 1;
+	uint32_t mask, inst_mask = data->inject.instance_mask;
+
+	/* no need to set instance mask if there is only one instance */
+	if (num_xcc <= 1 && inst_mask) {
+		data->inject.instance_mask = 0;
+		dev_dbg(adev->dev,
+			"RAS inject mask(0x%x) isn't supported and force it to 0.\n",
+			inst_mask);
+
+		return;
+	}
+
+	switch (data->head.block) {
+	case AMDGPU_RAS_BLOCK__GFX:
+		mask = GENMASK(num_xcc - 1, 0);
+		break;
+	case AMDGPU_RAS_BLOCK__SDMA:
+		mask = GENMASK(adev->sdma.num_instances - 1, 0);
+		break;
+	case AMDGPU_RAS_BLOCK__VCN:
+	case AMDGPU_RAS_BLOCK__JPEG:
+		mask = GENMASK(adev->vcn.num_vcn_inst - 1, 0);
+		break;
+	default:
+		mask = inst_mask;
+		break;
+	}
+
+	/* remove invalid bits in instance mask */
+	data->inject.instance_mask &= mask;
+	if (inst_mask != data->inject.instance_mask)
+		dev_dbg(adev->dev,
+			"Adjust RAS inject mask 0x%x to 0x%x\n",
+			inst_mask, data->inject.instance_mask);
+}
+
 /**
  * DOC: AMDGPU RAS debugfs control interface
  *
@@ -341,7 +479,7 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
  * sub_block_index: some IPs have subcomponets. say, GFX, sDMA.
  * name: the name of IP.
  *
- * inject has two more members than head, they are address, value.
+ * inject has three more members than head, they are address, value and mask.
  * As their names indicate, inject operation will write the
  * value to the address.
  *
@@ -365,7 +503,7 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
  *
  *	echo "disable <block>" > /sys/kernel/debug/dri/<N>/ras/ras_ctrl
  *	echo "enable  <block> <error>" > /sys/kernel/debug/dri/<N>/ras/ras_ctrl
- *	echo "inject  <block> <error> <sub-block> <address> <value> > /sys/kernel/debug/dri/<N>/ras/ras_ctrl
+ *	echo "inject  <block> <error> <sub-block> <address> <value> <mask>" > /sys/kernel/debug/dri/<N>/ras/ras_ctrl
  *
  * Where N, is the card which you want to affect.
  *
@@ -376,19 +514,21 @@ static int amdgpu_ras_debugfs_ctrl_parse_data(struct file *f,
  * The block is one of: umc, sdma, gfx, etc.
  *	see ras_block_string[] for details
  *
- * The error type is one of: ue, ce, where,
+ * The error type is one of: ue, ce and poison where,
  *	ue is multi-uncorrectable
  *	ce is single-correctable
+ *	poison is poison
  *
  * The sub-block is a the sub-block index, pass 0 if there is no sub-block.
  * The address and value are hexadecimal numbers, leading 0x is optional.
+ * The mask means instance mask, is optional, default value is 0x1.
  *
  * For instance,
  *
  * .. code-block:: bash
  *
  *	echo inject umc ue 0x0 0x0 0x0 > /sys/kernel/debug/dri/0/ras/ras_ctrl
- *	echo inject umc ce 0 0 0 > /sys/kernel/debug/dri/0/ras/ras_ctrl
+ *	echo inject umc ce 0 0 0 3 > /sys/kernel/debug/dri/0/ras/ras_ctrl
  *	echo disable umc > /sys/kernel/debug/dri/0/ras/ras_ctrl
  *
  * How to check the result of the operation?
@@ -429,6 +569,9 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f,
 			return size;
 		else
 			return ret;
+	} else if (data.op == 4) {
+		ret = amdgpu_check_address_validity(adev, data.inject.address, data.inject.value);
+		return ret ? ret : size;
 	}
 
 	if (!amdgpu_ras_is_supported(adev, data.head.block))
@@ -442,24 +585,21 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f,
 		ret = amdgpu_ras_feature_enable(adev, &data.head, 1);
 		break;
 	case 2:
-		if ((data.inject.address >= adev->gmc.mc_vram_size) ||
-		    (data.inject.address >= RAS_UMC_INJECT_ADDR_LIMIT)) {
-			dev_warn(adev->dev, "RAS WARN: input address "
-					"0x%llx is invalid.",
+		/* umc ce/ue error injection for a bad page is not allowed */
+		if (data.head.block == AMDGPU_RAS_BLOCK__UMC)
+			ret = amdgpu_ras_check_bad_page(adev, data.inject.address);
+		if (ret == -EINVAL) {
+			dev_warn(adev->dev, "RAS WARN: input address 0x%llx is invalid.",
 					data.inject.address);
-			ret = -EINVAL;
 			break;
-		}
-
-		/* umc ce/ue error injection for a bad page is not allowed */
-		if ((data.head.block == AMDGPU_RAS_BLOCK__UMC) &&
-		    amdgpu_ras_check_bad_page(adev, data.inject.address)) {
-			dev_warn(adev->dev, "RAS WARN: inject: 0x%llx has "
-				 "already been marked as bad!\n",
-				 data.inject.address);
+		} else if (ret == 1) {
+			dev_warn(adev->dev, "RAS WARN: inject: 0x%llx has already been marked as bad!\n",
+					data.inject.address);
 			break;
 		}
 
+		amdgpu_ras_instance_mask_check(adev, &data);
+
 		/* data.inject.address is offset instead of absolute gpu address */
 		ret = amdgpu_ras_error_inject(adev, &data.inject);
 		break;
@@ -474,6 +614,8 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f,
 	return size;
 }
 
+static int amdgpu_uniras_clear_badpages_info(struct amdgpu_device *adev);
+
 /**
  * DOC: AMDGPU RAS debugfs EEPROM table reset interface
  *
@@ -498,6 +640,11 @@ static ssize_t amdgpu_ras_debugfs_eeprom_write(struct file *f,
 		(struct amdgpu_device *)file_inode(f)->i_private;
 	int ret;
 
+	if (amdgpu_uniras_enabled(adev)) {
+		ret = amdgpu_uniras_clear_badpages_info(adev);
+		return ret ? ret : size;
+	}
+
 	ret = amdgpu_ras_eeprom_reset_table(
 		&(amdgpu_ras_get_context(adev)->eeprom_control));
 
@@ -560,14 +707,18 @@ static ssize_t amdgpu_ras_sysfs_read(struct device *dev,
 	if (amdgpu_ras_query_error_status(obj->adev, &info))
 		return -EINVAL;
 
-	if (obj->adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 2) &&
-	    obj->adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 4)) {
+	if (amdgpu_ip_version(obj->adev, MP0_HWIP, 0) != IP_VERSION(11, 0, 2) &&
+	    amdgpu_ip_version(obj->adev, MP0_HWIP, 0) != IP_VERSION(11, 0, 4)) {
 		if (amdgpu_ras_reset_error_status(obj->adev, info.head.block))
 			dev_warn(obj->adev->dev, "Failed to reset error counter and error status");
 	}
 
-	return sysfs_emit(buf, "%s: %lu\n%s: %lu\n", "ue", info.ue_count,
-			  "ce", info.ce_count);
+	if (info.head.block == AMDGPU_RAS_BLOCK__UMC)
+		return sysfs_emit(buf, "%s: %lu\n%s: %lu\n%s: %lu\n", "ue", info.ue_count,
+				"ce", info.ce_count, "de", info.de_count);
+	else
+		return sysfs_emit(buf, "%s: %lu\n%s: %lu\n", "ue", info.ue_count,
+				"ce", info.ce_count);
 }
 
 /* obj begin */
@@ -577,8 +728,11 @@ static ssize_t amdgpu_ras_sysfs_read(struct device *dev,
 
 static inline void put_obj(struct ras_manager *obj)
 {
-	if (obj && (--obj->use == 0))
+	if (obj && (--obj->use == 0)) {
 		list_del(&obj->node);
+		amdgpu_ras_error_data_fini(&obj->err_data);
+	}
+
 	if (obj && (obj->use < 0))
 		DRM_ERROR("RAS ERROR: Unbalance obj(%s) use\n", get_ras_block_str(&obj->head));
 }
@@ -608,6 +762,9 @@ static struct ras_manager *amdgpu_ras_create_obj(struct amdgpu_device *adev,
 	if (alive_obj(obj))
 		return NULL;
 
+	if (amdgpu_ras_error_data_init(&obj->err_data))
+		return NULL;
+
 	obj->head = *head;
 	obj->adev = adev;
 	list_add(&obj->node, &con->head);
@@ -707,28 +864,28 @@ static int __amdgpu_ras_feature_enable(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int amdgpu_ras_check_feature_allowed(struct amdgpu_device *adev,
-		struct ras_common_if *head)
-{
-	if (amdgpu_ras_is_feature_allowed(adev, head) ||
-		amdgpu_ras_is_poison_mode_supported(adev))
-		return 1;
-	else
-		return 0;
-}
-
 /* wrapper of psp_ras_enable_features */
 int amdgpu_ras_feature_enable(struct amdgpu_device *adev,
 		struct ras_common_if *head, bool enable)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 	union ta_ras_cmd_input *info;
-	int ret = 0;
+	int ret;
 
 	if (!con)
 		return -EINVAL;
 
-	if (head->block == AMDGPU_RAS_BLOCK__GFX) {
+	/* For non-gfx ip, do not enable ras feature if it is not allowed */
+	/* For gfx ip, regardless of feature support status, */
+	/* Force issue enable or disable ras feature commands */
+	if (head->block != AMDGPU_RAS_BLOCK__GFX &&
+	    !amdgpu_ras_is_feature_allowed(adev, head))
+		return 0;
+
+	/* Only enable gfx ras feature from host side */
+	if (head->block == AMDGPU_RAS_BLOCK__GFX &&
+	    !amdgpu_sriov_vf(adev) &&
+	    !amdgpu_ras_intr_triggered()) {
 		info = kzalloc(sizeof(union ta_ras_cmd_input), GFP_KERNEL);
 		if (!info)
 			return -ENOMEM;
@@ -744,32 +901,24 @@ int amdgpu_ras_feature_enable(struct amdgpu_device *adev,
 				.error_type = amdgpu_ras_error_to_ta(head->type),
 			};
 		}
-	}
 
-	/* Do not enable if it is not allowed. */
-	if (enable && !amdgpu_ras_check_feature_allowed(adev, head))
-		goto out;
-
-	/* Only enable ras feature operation handle on host side */
-	if (head->block == AMDGPU_RAS_BLOCK__GFX &&
-		!amdgpu_sriov_vf(adev) &&
-		!amdgpu_ras_intr_triggered()) {
 		ret = psp_ras_enable_features(&adev->psp, info, enable);
 		if (ret) {
 			dev_err(adev->dev, "ras %s %s failed poison:%d ret:%d\n",
 				enable ? "enable":"disable",
 				get_ras_block_str(head),
 				amdgpu_ras_is_poison_mode_supported(adev), ret);
-			goto out;
+			kfree(info);
+			return ret;
 		}
+
+		kfree(info);
 	}
 
 	/* setup the obj */
 	__amdgpu_ras_feature_enable(adev, head, enable);
-out:
-	if (head->block == AMDGPU_RAS_BLOCK__GFX)
-		kfree(info);
-	return ret;
+
+	return 0;
 }
 
 /* Only used in device probe stage and called only once. */
@@ -808,7 +957,7 @@ int amdgpu_ras_feature_enable_on_boot(struct amdgpu_device *adev,
 			if (ret)
 				return ret;
 
-			/* gfx block ras dsiable cmd must send to ras-ta */
+			/* gfx block ras disable cmd must send to ras-ta */
 			if (head->block == AMDGPU_RAS_BLOCK__GFX)
 				con->features |= BIT(head->block);
 
@@ -972,103 +1121,524 @@ static void amdgpu_ras_get_ecc_info(struct amdgpu_device *adev, struct ras_err_d
 	}
 }
 
-/* query/inject/cure begin */
-int amdgpu_ras_query_error_status(struct amdgpu_device *adev,
-				  struct ras_query_if *info)
+static void amdgpu_ras_error_print_error_data(struct amdgpu_device *adev,
+					      struct ras_manager *ras_mgr,
+					      struct ras_err_data *err_data,
+					      struct ras_query_context *qctx,
+					      const char *blk_name,
+					      bool is_ue,
+					      bool is_de)
 {
-	struct amdgpu_ras_block_object *block_obj = NULL;
-	struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head);
-	struct ras_err_data err_data = {0, 0, 0, NULL};
+	struct amdgpu_smuio_mcm_config_info *mcm_info;
+	struct ras_err_node *err_node;
+	struct ras_err_info *err_info;
+	u64 event_id = qctx->evid.event_id;
+
+	if (is_ue) {
+		for_each_ras_error(err_node, err_data) {
+			err_info = &err_node->err_info;
+			mcm_info = &err_info->mcm_info;
+			if (err_info->ue_count) {
+				RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d, "
+					      "%lld new uncorrectable hardware errors detected in %s block\n",
+					      mcm_info->socket_id,
+					      mcm_info->die_id,
+					      err_info->ue_count,
+					      blk_name);
+			}
+		}
 
-	if (!obj)
-		return -EINVAL;
+		for_each_ras_error(err_node, &ras_mgr->err_data) {
+			err_info = &err_node->err_info;
+			mcm_info = &err_info->mcm_info;
+			RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d, "
+				      "%lld uncorrectable hardware errors detected in total in %s block\n",
+				      mcm_info->socket_id, mcm_info->die_id, err_info->ue_count, blk_name);
+		}
 
-	if (info->head.block == AMDGPU_RAS_BLOCK__UMC) {
-		amdgpu_ras_get_ecc_info(adev, &err_data);
 	} else {
-		block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, 0);
-		if (!block_obj || !block_obj->hw_ops)   {
-			dev_dbg_once(adev->dev, "%s doesn't config RAS function\n",
-				     get_ras_block_str(&info->head));
-			return -EINVAL;
+		if (is_de) {
+			for_each_ras_error(err_node, err_data) {
+				err_info = &err_node->err_info;
+				mcm_info = &err_info->mcm_info;
+				if (err_info->de_count) {
+					RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d, "
+						      "%lld new deferred hardware errors detected in %s block\n",
+						      mcm_info->socket_id,
+						      mcm_info->die_id,
+						      err_info->de_count,
+						      blk_name);
+				}
+			}
+
+			for_each_ras_error(err_node, &ras_mgr->err_data) {
+				err_info = &err_node->err_info;
+				mcm_info = &err_info->mcm_info;
+				RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d, "
+					      "%lld deferred hardware errors detected in total in %s block\n",
+					      mcm_info->socket_id, mcm_info->die_id,
+					      err_info->de_count, blk_name);
+			}
+		} else {
+			if (adev->debug_disable_ce_logs)
+				return;
+
+			for_each_ras_error(err_node, err_data) {
+				err_info = &err_node->err_info;
+				mcm_info = &err_info->mcm_info;
+				if (err_info->ce_count) {
+					RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d, "
+						      "%lld new correctable hardware errors detected in %s block\n",
+						      mcm_info->socket_id,
+						      mcm_info->die_id,
+						      err_info->ce_count,
+						      blk_name);
+				}
+			}
+
+			for_each_ras_error(err_node, &ras_mgr->err_data) {
+				err_info = &err_node->err_info;
+				mcm_info = &err_info->mcm_info;
+				RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d, "
+					      "%lld correctable hardware errors detected in total in %s block\n",
+					      mcm_info->socket_id, mcm_info->die_id,
+					      err_info->ce_count, blk_name);
+			}
+		}
+	}
+}
+
+static inline bool err_data_has_source_info(struct ras_err_data *data)
+{
+	return !list_empty(&data->err_node_list);
+}
+
+static void amdgpu_ras_error_generate_report(struct amdgpu_device *adev,
+					     struct ras_query_if *query_if,
+					     struct ras_err_data *err_data,
+					     struct ras_query_context *qctx)
+{
+	struct ras_manager *ras_mgr = amdgpu_ras_find_obj(adev, &query_if->head);
+	const char *blk_name = get_ras_block_str(&query_if->head);
+	u64 event_id = qctx->evid.event_id;
+
+	if (err_data->ce_count) {
+		if (err_data_has_source_info(err_data)) {
+			amdgpu_ras_error_print_error_data(adev, ras_mgr, err_data, qctx,
+							  blk_name, false, false);
+		} else if (!adev->aid_mask &&
+			   adev->smuio.funcs &&
+			   adev->smuio.funcs->get_socket_id &&
+			   adev->smuio.funcs->get_die_id) {
+			RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d "
+				      "%ld correctable hardware errors "
+				      "detected in %s block\n",
+				      adev->smuio.funcs->get_socket_id(adev),
+				      adev->smuio.funcs->get_die_id(adev),
+				      ras_mgr->err_data.ce_count,
+				      blk_name);
+		} else {
+			RAS_EVENT_LOG(adev, event_id, "%ld correctable hardware errors "
+				      "detected in %s block\n",
+				      ras_mgr->err_data.ce_count,
+				      blk_name);
+		}
+	}
+
+	if (err_data->ue_count) {
+		if (err_data_has_source_info(err_data)) {
+			amdgpu_ras_error_print_error_data(adev, ras_mgr, err_data, qctx,
+							  blk_name, true, false);
+		} else if (!adev->aid_mask &&
+			   adev->smuio.funcs &&
+			   adev->smuio.funcs->get_socket_id &&
+			   adev->smuio.funcs->get_die_id) {
+			RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d "
+				      "%ld uncorrectable hardware errors "
+				      "detected in %s block\n",
+				      adev->smuio.funcs->get_socket_id(adev),
+				      adev->smuio.funcs->get_die_id(adev),
+				      ras_mgr->err_data.ue_count,
+				      blk_name);
+		} else {
+			RAS_EVENT_LOG(adev, event_id, "%ld uncorrectable hardware errors "
+				      "detected in %s block\n",
+				      ras_mgr->err_data.ue_count,
+				      blk_name);
+		}
+	}
+
+	if (err_data->de_count) {
+		if (err_data_has_source_info(err_data)) {
+			amdgpu_ras_error_print_error_data(adev, ras_mgr, err_data, qctx,
+							  blk_name, false, true);
+		} else if (!adev->aid_mask &&
+			   adev->smuio.funcs &&
+			   adev->smuio.funcs->get_socket_id &&
+			   adev->smuio.funcs->get_die_id) {
+			RAS_EVENT_LOG(adev, event_id, "socket: %d, die: %d "
+				      "%ld deferred hardware errors "
+				      "detected in %s block\n",
+				      adev->smuio.funcs->get_socket_id(adev),
+				      adev->smuio.funcs->get_die_id(adev),
+				      ras_mgr->err_data.de_count,
+				      blk_name);
+		} else {
+			RAS_EVENT_LOG(adev, event_id, "%ld deferred hardware errors "
+				      "detected in %s block\n",
+				      ras_mgr->err_data.de_count,
+				      blk_name);
+		}
+	}
+}
+
+static void amdgpu_ras_virt_error_generate_report(struct amdgpu_device *adev,
+						  struct ras_query_if *query_if,
+						  struct ras_err_data *err_data,
+						  struct ras_query_context *qctx)
+{
+	unsigned long new_ue, new_ce, new_de;
+	struct ras_manager *obj = amdgpu_ras_find_obj(adev, &query_if->head);
+	const char *blk_name = get_ras_block_str(&query_if->head);
+	u64 event_id = qctx->evid.event_id;
+
+	new_ce = err_data->ce_count - obj->err_data.ce_count;
+	new_ue = err_data->ue_count - obj->err_data.ue_count;
+	new_de = err_data->de_count - obj->err_data.de_count;
+
+	if (new_ce) {
+		RAS_EVENT_LOG(adev, event_id, "%lu correctable hardware errors "
+			      "detected in %s block\n",
+			      new_ce,
+			      blk_name);
+	}
+
+	if (new_ue) {
+		RAS_EVENT_LOG(adev, event_id, "%lu uncorrectable hardware errors "
+			      "detected in %s block\n",
+			      new_ue,
+			      blk_name);
+	}
+
+	if (new_de) {
+		RAS_EVENT_LOG(adev, event_id, "%lu deferred hardware errors "
+			      "detected in %s block\n",
+			      new_de,
+			      blk_name);
+	}
+}
+
+static void amdgpu_rasmgr_error_data_statistic_update(struct ras_manager *obj, struct ras_err_data *err_data)
+{
+	struct ras_err_node *err_node;
+	struct ras_err_info *err_info;
+
+	if (err_data_has_source_info(err_data)) {
+		for_each_ras_error(err_node, err_data) {
+			err_info = &err_node->err_info;
+			amdgpu_ras_error_statistic_de_count(&obj->err_data,
+					&err_info->mcm_info, err_info->de_count);
+			amdgpu_ras_error_statistic_ce_count(&obj->err_data,
+					&err_info->mcm_info, err_info->ce_count);
+			amdgpu_ras_error_statistic_ue_count(&obj->err_data,
+					&err_info->mcm_info, err_info->ue_count);
 		}
+	} else {
+		/* for legacy asic path which doesn't has error source info */
+		obj->err_data.ue_count += err_data->ue_count;
+		obj->err_data.ce_count += err_data->ce_count;
+		obj->err_data.de_count += err_data->de_count;
+	}
+}
+
+static void amdgpu_ras_mgr_virt_error_data_statistics_update(struct ras_manager *obj,
+							     struct ras_err_data *err_data)
+{
+	/* Host reports absolute counts */
+	obj->err_data.ue_count = err_data->ue_count;
+	obj->err_data.ce_count = err_data->ce_count;
+	obj->err_data.de_count = err_data->de_count;
+}
 
-		if (block_obj->hw_ops->query_ras_error_count)
-			block_obj->hw_ops->query_ras_error_count(adev, &err_data);
+static struct ras_manager *get_ras_manager(struct amdgpu_device *adev, enum amdgpu_ras_block blk)
+{
+	struct ras_common_if head;
 
-		if ((info->head.block == AMDGPU_RAS_BLOCK__SDMA) ||
-		    (info->head.block == AMDGPU_RAS_BLOCK__GFX) ||
-		    (info->head.block == AMDGPU_RAS_BLOCK__MMHUB)) {
+	memset(&head, 0, sizeof(head));
+	head.block = blk;
+
+	return amdgpu_ras_find_obj(adev, &head);
+}
+
+int amdgpu_ras_bind_aca(struct amdgpu_device *adev, enum amdgpu_ras_block blk,
+			const struct aca_info *aca_info, void *data)
+{
+	struct ras_manager *obj;
+
+	/* in resume phase, no need to create aca fs node */
+	if (adev->in_suspend || amdgpu_reset_in_recovery(adev))
+		return 0;
+
+	obj = get_ras_manager(adev, blk);
+	if (!obj)
+		return -EINVAL;
+
+	return amdgpu_aca_add_handle(adev, &obj->aca_handle, ras_block_str(blk), aca_info, data);
+}
+
+int amdgpu_ras_unbind_aca(struct amdgpu_device *adev, enum amdgpu_ras_block blk)
+{
+	struct ras_manager *obj;
+
+	obj = get_ras_manager(adev, blk);
+	if (!obj)
+		return -EINVAL;
+
+	amdgpu_aca_remove_handle(&obj->aca_handle);
+
+	return 0;
+}
+
+static int amdgpu_aca_log_ras_error_data(struct amdgpu_device *adev, enum amdgpu_ras_block blk,
+					 enum aca_error_type type, struct ras_err_data *err_data,
+					 struct ras_query_context *qctx)
+{
+	struct ras_manager *obj;
+
+	obj = get_ras_manager(adev, blk);
+	if (!obj)
+		return -EINVAL;
+
+	return amdgpu_aca_get_error_data(adev, &obj->aca_handle, type, err_data, qctx);
+}
+
+ssize_t amdgpu_ras_aca_sysfs_read(struct device *dev, struct device_attribute *attr,
+				  struct aca_handle *handle, char *buf, void *data)
+{
+	struct ras_manager *obj = container_of(handle, struct ras_manager, aca_handle);
+	struct ras_query_if info = {
+		.head = obj->head,
+	};
+
+	if (!amdgpu_ras_get_error_query_ready(obj->adev))
+		return sysfs_emit(buf, "Query currently inaccessible\n");
+
+	if (amdgpu_ras_query_error_status(obj->adev, &info))
+		return -EINVAL;
+
+	return sysfs_emit(buf, "%s: %lu\n%s: %lu\n%s: %lu\n", "ue", info.ue_count,
+			  "ce", info.ce_count, "de", info.de_count);
+}
+
+static int amdgpu_ras_query_error_status_helper(struct amdgpu_device *adev,
+						struct ras_query_if *info,
+						struct ras_err_data *err_data,
+						struct ras_query_context *qctx,
+						unsigned int error_query_mode)
+{
+	enum amdgpu_ras_block blk = info ? info->head.block : AMDGPU_RAS_BLOCK_COUNT;
+	struct amdgpu_ras_block_object *block_obj = NULL;
+	int ret;
+
+	if (blk == AMDGPU_RAS_BLOCK_COUNT)
+		return -EINVAL;
+
+	if (error_query_mode == AMDGPU_RAS_INVALID_ERROR_QUERY)
+		return -EINVAL;
+
+	if (error_query_mode == AMDGPU_RAS_VIRT_ERROR_COUNT_QUERY) {
+		return amdgpu_virt_req_ras_err_count(adev, blk, err_data);
+	} else if (error_query_mode == AMDGPU_RAS_DIRECT_ERROR_QUERY) {
+		if (info->head.block == AMDGPU_RAS_BLOCK__UMC) {
+			amdgpu_ras_get_ecc_info(adev, err_data);
+		} else {
+			block_obj = amdgpu_ras_get_ras_block(adev, info->head.block, 0);
+			if (!block_obj || !block_obj->hw_ops) {
+				dev_dbg_once(adev->dev, "%s doesn't config RAS function\n",
+					     get_ras_block_str(&info->head));
+				return -EINVAL;
+			}
+
+			if (block_obj->hw_ops->query_ras_error_count)
+				block_obj->hw_ops->query_ras_error_count(adev, err_data);
+
+			if ((info->head.block == AMDGPU_RAS_BLOCK__SDMA) ||
+			    (info->head.block == AMDGPU_RAS_BLOCK__GFX) ||
+			    (info->head.block == AMDGPU_RAS_BLOCK__MMHUB)) {
 				if (block_obj->hw_ops->query_ras_error_status)
 					block_obj->hw_ops->query_ras_error_status(adev);
 			}
+		}
+	} else {
+		if (amdgpu_aca_is_enabled(adev)) {
+			ret = amdgpu_aca_log_ras_error_data(adev, blk, ACA_ERROR_TYPE_UE, err_data, qctx);
+			if (ret)
+				return ret;
+
+			ret = amdgpu_aca_log_ras_error_data(adev, blk, ACA_ERROR_TYPE_CE, err_data, qctx);
+			if (ret)
+				return ret;
+
+			ret = amdgpu_aca_log_ras_error_data(adev, blk, ACA_ERROR_TYPE_DEFERRED, err_data, qctx);
+			if (ret)
+				return ret;
+		} else {
+			/* FIXME: add code to check return value later */
+			amdgpu_mca_smu_log_ras_error(adev, blk, AMDGPU_MCA_ERROR_TYPE_UE, err_data, qctx);
+			amdgpu_mca_smu_log_ras_error(adev, blk, AMDGPU_MCA_ERROR_TYPE_CE, err_data, qctx);
+		}
 	}
 
-	obj->err_data.ue_count += err_data.ue_count;
-	obj->err_data.ce_count += err_data.ce_count;
+	return 0;
+}
+
+/* query/inject/cure begin */
+static int amdgpu_ras_query_error_status_with_event(struct amdgpu_device *adev,
+						    struct ras_query_if *info,
+						    enum ras_event_type type)
+{
+	struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head);
+	struct ras_err_data err_data;
+	struct ras_query_context qctx;
+	unsigned int error_query_mode;
+	int ret;
+
+	if (!obj)
+		return -EINVAL;
+
+	ret = amdgpu_ras_error_data_init(&err_data);
+	if (ret)
+		return ret;
+
+	if (!amdgpu_ras_get_error_query_mode(adev, &error_query_mode))
+		return -EINVAL;
+
+	memset(&qctx, 0, sizeof(qctx));
+	qctx.evid.type = type;
+	qctx.evid.event_id = amdgpu_ras_acquire_event_id(adev, type);
+
+	if (!down_read_trylock(&adev->reset_domain->sem)) {
+		ret = -EIO;
+		goto out_fini_err_data;
+	}
+
+	ret = amdgpu_ras_query_error_status_helper(adev, info,
+						   &err_data,
+						   &qctx,
+						   error_query_mode);
+	up_read(&adev->reset_domain->sem);
+	if (ret)
+		goto out_fini_err_data;
+
+	if (error_query_mode != AMDGPU_RAS_VIRT_ERROR_COUNT_QUERY) {
+		amdgpu_rasmgr_error_data_statistic_update(obj, &err_data);
+		amdgpu_ras_error_generate_report(adev, info, &err_data, &qctx);
+	} else {
+		/* Host provides absolute error counts. First generate the report
+		 * using the previous VF internal count against new host count.
+		 * Then Update VF internal count.
+		 */
+		amdgpu_ras_virt_error_generate_report(adev, info, &err_data, &qctx);
+		amdgpu_ras_mgr_virt_error_data_statistics_update(obj, &err_data);
+	}
 
 	info->ue_count = obj->err_data.ue_count;
 	info->ce_count = obj->err_data.ce_count;
+	info->de_count = obj->err_data.de_count;
 
-	if (err_data.ce_count) {
-		if (adev->smuio.funcs &&
-		    adev->smuio.funcs->get_socket_id &&
-		    adev->smuio.funcs->get_die_id) {
-			dev_info(adev->dev, "socket: %d, die: %d "
-					"%ld correctable hardware errors "
-					"detected in %s block, no user "
-					"action is needed.\n",
-					adev->smuio.funcs->get_socket_id(adev),
-					adev->smuio.funcs->get_die_id(adev),
-					obj->err_data.ce_count,
-					get_ras_block_str(&info->head));
-		} else {
-			dev_info(adev->dev, "%ld correctable hardware errors "
-					"detected in %s block, no user "
-					"action is needed.\n",
-					obj->err_data.ce_count,
-					get_ras_block_str(&info->head));
-		}
-	}
-	if (err_data.ue_count) {
-		if (adev->smuio.funcs &&
-		    adev->smuio.funcs->get_socket_id &&
-		    adev->smuio.funcs->get_die_id) {
-			dev_info(adev->dev, "socket: %d, die: %d "
-					"%ld uncorrectable hardware errors "
-					"detected in %s block\n",
-					adev->smuio.funcs->get_socket_id(adev),
-					adev->smuio.funcs->get_die_id(adev),
-					obj->err_data.ue_count,
-					get_ras_block_str(&info->head));
-		} else {
-			dev_info(adev->dev, "%ld uncorrectable hardware errors "
-					"detected in %s block\n",
-					obj->err_data.ue_count,
-					get_ras_block_str(&info->head));
-		}
+out_fini_err_data:
+	amdgpu_ras_error_data_fini(&err_data);
+
+	return ret;
+}
+
+static int amdgpu_uniras_clear_badpages_info(struct amdgpu_device *adev)
+{
+	struct ras_cmd_dev_handle req = {0};
+	int ret;
+
+	ret = amdgpu_ras_mgr_handle_ras_cmd(adev, RAS_CMD__CLEAR_BAD_PAGE_INFO,
+				&req, sizeof(req), NULL, 0);
+	if (ret) {
+		dev_err(adev->dev, "Failed to clear bad pages info, ret: %d\n", ret);
+		return ret;
 	}
 
 	return 0;
 }
 
-int amdgpu_ras_reset_error_status(struct amdgpu_device *adev,
-		enum amdgpu_ras_block block)
+static int amdgpu_uniras_query_block_ecc(struct amdgpu_device *adev,
+			struct ras_query_if *info)
 {
-	struct amdgpu_ras_block_object *block_obj = amdgpu_ras_get_ras_block(adev, block, 0);
+	struct ras_cmd_block_ecc_info_req req = {0};
+	struct ras_cmd_block_ecc_info_rsp rsp = {0};
+	int ret;
 
-	if (!amdgpu_ras_is_supported(adev, block))
+	if (!info)
 		return -EINVAL;
 
-	if (!block_obj || !block_obj->hw_ops)   {
+	req.block_id = info->head.block;
+	req.subblock_id = info->head.sub_block_index;
+
+	ret = amdgpu_ras_mgr_handle_ras_cmd(adev, RAS_CMD__GET_BLOCK_ECC_STATUS,
+				&req, sizeof(req), &rsp, sizeof(rsp));
+	if (!ret) {
+		info->ce_count = rsp.ce_count;
+		info->ue_count = rsp.ue_count;
+		info->de_count = rsp.de_count;
+	}
+
+	return ret;
+}
+
+int amdgpu_ras_query_error_status(struct amdgpu_device *adev, struct ras_query_if *info)
+{
+	if (amdgpu_uniras_enabled(adev))
+		return amdgpu_uniras_query_block_ecc(adev, info);
+	else
+		return amdgpu_ras_query_error_status_with_event(adev, info, RAS_EVENT_TYPE_INVALID);
+}
+
+int amdgpu_ras_reset_error_count(struct amdgpu_device *adev,
+		enum amdgpu_ras_block block)
+{
+	struct amdgpu_ras_block_object *block_obj = amdgpu_ras_get_ras_block(adev, block, 0);
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+	const struct aca_smu_funcs *smu_funcs = adev->aca.smu_funcs;
+
+	if (!block_obj || !block_obj->hw_ops) {
 		dev_dbg_once(adev->dev, "%s doesn't config RAS function\n",
-			     ras_block_str(block));
-		return -EINVAL;
+				ras_block_str(block));
+		return -EOPNOTSUPP;
 	}
 
+	if (!amdgpu_ras_is_supported(adev, block) ||
+	    !amdgpu_ras_get_aca_debug_mode(adev))
+		return -EOPNOTSUPP;
+
+	if (amdgpu_sriov_vf(adev))
+		return -EOPNOTSUPP;
+
+	/* skip ras error reset in gpu reset */
+	if ((amdgpu_in_reset(adev) || amdgpu_ras_in_recovery(adev)) &&
+	    ((smu_funcs && smu_funcs->set_debug_mode) ||
+	     (mca_funcs && mca_funcs->mca_set_debug_mode)))
+		return -EOPNOTSUPP;
+
 	if (block_obj->hw_ops->reset_ras_error_count)
 		block_obj->hw_ops->reset_ras_error_count(adev);
 
+	return 0;
+}
+
+int amdgpu_ras_reset_error_status(struct amdgpu_device *adev,
+		enum amdgpu_ras_block block)
+{
+	struct amdgpu_ras_block_object *block_obj = amdgpu_ras_get_ras_block(adev, block, 0);
+
+	if (amdgpu_ras_reset_error_count(adev, block) == -EOPNOTSUPP)
+		return 0;
+
 	if ((block == AMDGPU_RAS_BLOCK__GFX) ||
 	    (block == AMDGPU_RAS_BLOCK__MMHUB)) {
 		if (block_obj->hw_ops->reset_ras_error_status)
@@ -1078,6 +1648,27 @@ int amdgpu_ras_reset_error_status(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int amdgpu_uniras_error_inject(struct amdgpu_device *adev,
+		struct ras_inject_if *info)
+{
+	struct ras_cmd_inject_error_req inject_req;
+	struct ras_cmd_inject_error_rsp rsp;
+
+	if (!info)
+		return -EINVAL;
+
+	memset(&inject_req, 0, sizeof(inject_req));
+	inject_req.block_id = info->head.block;
+	inject_req.subblock_id = info->head.sub_block_index;
+	inject_req.address = info->address;
+	inject_req.error_type = info->head.type;
+	inject_req.instance_mask = info->instance_mask;
+	inject_req.method = info->value;
+
+	return amdgpu_ras_mgr_handle_ras_cmd(adev, RAS_CMD__INJECT_ERROR,
+			&inject_req, sizeof(inject_req), &rsp, sizeof(rsp));
+}
+
 /* wrapper of psp_ras_trigger_error */
 int amdgpu_ras_error_inject(struct amdgpu_device *adev,
 		struct ras_inject_if *info)
@@ -1095,6 +1686,9 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev,
 							info->head.block,
 							info->head.sub_block_index);
 
+	if (amdgpu_uniras_enabled(adev))
+		return amdgpu_uniras_error_inject(adev, info);
+
 	/* inject on guest isn't allowed, return success directly */
 	if (amdgpu_sriov_vf(adev))
 		return 0;
@@ -1109,21 +1703,22 @@ int amdgpu_ras_error_inject(struct amdgpu_device *adev,
 	}
 
 	/* Calculate XGMI relative offset */
-	if (adev->gmc.xgmi.num_physical_nodes > 1) {
+	if (adev->gmc.xgmi.num_physical_nodes > 1 &&
+	    info->head.block != AMDGPU_RAS_BLOCK__GFX) {
 		block_info.address =
 			amdgpu_xgmi_get_relative_phy_addr(adev,
 							  block_info.address);
 	}
 
-	if (info->head.block == AMDGPU_RAS_BLOCK__GFX) {
-		if (block_obj->hw_ops->ras_error_inject)
-			ret = block_obj->hw_ops->ras_error_inject(adev, info);
+	if (block_obj->hw_ops->ras_error_inject) {
+		if (info->head.block == AMDGPU_RAS_BLOCK__GFX)
+			ret = block_obj->hw_ops->ras_error_inject(adev, info, info->instance_mask);
+		else /* Special ras_error_inject is defined (e.g: xgmi) */
+			ret = block_obj->hw_ops->ras_error_inject(adev, &block_info,
+						info->instance_mask);
 	} else {
-		/* If defined special ras_error_inject(e.g: xgmi), implement special ras_error_inject */
-		if (block_obj->hw_ops->ras_error_inject)
-			ret = block_obj->hw_ops->ras_error_inject(adev, &block_info);
-		else  /*If not defined .ras_error_inject, use default ras_error_inject*/
-			ret = psp_ras_trigger_error(&adev->psp, &block_info);
+		/* default path */
+		ret = psp_ras_trigger_error(&adev->psp, &block_info, info->instance_mask);
 	}
 
 	if (ret)
@@ -1163,8 +1758,8 @@ static int amdgpu_ras_query_error_count_helper(struct amdgpu_device *adev,
 
 	/* some hardware/IP supports read to clear
 	 * no need to explictly reset the err status after the query call */
-	if (adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 2) &&
-	    adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 4)) {
+	if (amdgpu_ip_version(adev, MP0_HWIP, 0) != IP_VERSION(11, 0, 2) &&
+	    amdgpu_ip_version(adev, MP0_HWIP, 0) != IP_VERSION(11, 0, 4)) {
 		if (amdgpu_ras_reset_error_status(adev, query_info->head.block))
 			dev_warn(adev->dev,
 				 "Failed to reset error counter and error status\n");
@@ -1238,7 +1833,9 @@ int amdgpu_ras_query_error_count(struct amdgpu_device *adev,
 /* sysfs begin */
 
 static int amdgpu_ras_badpages_read(struct amdgpu_device *adev,
-		struct ras_badpage **bps, unsigned int *count);
+		struct ras_badpage *bps, uint32_t count, uint32_t start);
+static int amdgpu_uniras_badpages_read(struct amdgpu_device *adev,
+		struct ras_badpage *bps, uint32_t count, uint32_t start);
 
 static char *amdgpu_ras_badpage_flags_str(unsigned int flags)
 {
@@ -1284,7 +1881,7 @@ static char *amdgpu_ras_badpage_flags_str(unsigned int flags)
  */
 
 static ssize_t amdgpu_ras_sysfs_badpages_read(struct file *f,
-		struct kobject *kobj, struct bin_attribute *attr,
+		struct kobject *kobj, const struct bin_attribute *attr,
 		char *buf, loff_t ppos, size_t count)
 {
 	struct amdgpu_ras *con =
@@ -1296,19 +1893,50 @@ static ssize_t amdgpu_ras_sysfs_badpages_read(struct file *f,
 	unsigned int end = div64_ul(ppos + count - 1, element_size);
 	ssize_t s = 0;
 	struct ras_badpage *bps = NULL;
-	unsigned int bps_count = 0;
+	int bps_count = 0, i, status;
+	uint64_t address;
 
 	memset(buf, 0, count);
 
-	if (amdgpu_ras_badpages_read(adev, &bps, &bps_count))
+	bps_count = end - start;
+	bps = kmalloc_array(bps_count, sizeof(*bps), GFP_KERNEL);
+	if (!bps)
 		return 0;
 
-	for (; start < end && start < bps_count; start++)
+	memset(bps, 0, sizeof(*bps) * bps_count);
+
+	if (amdgpu_uniras_enabled(adev))
+		bps_count = amdgpu_uniras_badpages_read(adev, bps, bps_count, start);
+	else
+		bps_count = amdgpu_ras_badpages_read(adev, bps, bps_count, start);
+
+	if (bps_count <= 0) {
+		kfree(bps);
+		return 0;
+	}
+
+	for (i = 0; i < bps_count; i++) {
+		address = ((uint64_t)bps[i].bp) << AMDGPU_GPU_PAGE_SHIFT;
+		if (amdgpu_ras_check_critical_address(adev, address))
+			continue;
+
+		bps[i].size = AMDGPU_GPU_PAGE_SIZE;
+
+		status = amdgpu_vram_mgr_query_page_status(&adev->mman.vram_mgr,
+					address);
+		if (status == -EBUSY)
+			bps[i].flags = AMDGPU_RAS_RETIRE_PAGE_PENDING;
+		else if (status == -ENOENT)
+			bps[i].flags = AMDGPU_RAS_RETIRE_PAGE_FAULT;
+		else
+			bps[i].flags = AMDGPU_RAS_RETIRE_PAGE_RESERVED;
+
 		s += scnprintf(&buf[s], element_size + 1,
 				"0x%08x : 0x%08x : %1s\n",
-				bps[start].bp,
-				bps[start].size,
-				amdgpu_ras_badpage_flags_str(bps[start].flags));
+				bps[i].bp,
+				bps[i].size,
+				amdgpu_ras_badpage_flags_str(bps[i].flags));
+	}
 
 	kfree(bps);
 
@@ -1324,20 +1952,103 @@ static ssize_t amdgpu_ras_sysfs_features_read(struct device *dev,
 	return sysfs_emit(buf, "feature mask: 0x%x\n", con->features);
 }
 
+static bool amdgpu_ras_get_version_info(struct amdgpu_device *adev, u32 *major,
+			u32 *minor, u32 *rev)
+{
+	int i;
+
+	if (!adev || !major || !minor || !rev || !amdgpu_uniras_enabled(adev))
+		return false;
+
+	for (i = 0; i < adev->num_ip_blocks; i++) {
+		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_RAS) {
+			*major = adev->ip_blocks[i].version->major;
+			*minor = adev->ip_blocks[i].version->minor;
+			*rev = adev->ip_blocks[i].version->rev;
+			return true;
+		}
+	}
+
+	return false;
+}
+
+static ssize_t amdgpu_ras_sysfs_version_show(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct amdgpu_ras *con =
+		container_of(attr, struct amdgpu_ras, version_attr);
+	u32 major, minor, rev;
+	ssize_t size = 0;
+
+	size += sysfs_emit_at(buf, size, "table version: 0x%x\n",
+			con->eeprom_control.tbl_hdr.version);
+
+	if (amdgpu_ras_get_version_info(con->adev, &major, &minor, &rev))
+		size += sysfs_emit_at(buf, size, "ras version: %u.%u.%u\n",
+			major, minor, rev);
+
+	return size;
+}
+
+static ssize_t amdgpu_ras_sysfs_schema_show(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct amdgpu_ras *con =
+		container_of(attr, struct amdgpu_ras, schema_attr);
+	return sysfs_emit(buf, "schema: 0x%x\n", con->schema);
+}
+
+static struct {
+	enum ras_event_type type;
+	const char *name;
+} dump_event[] = {
+	{RAS_EVENT_TYPE_FATAL, "Fatal Error"},
+	{RAS_EVENT_TYPE_POISON_CREATION, "Poison Creation"},
+	{RAS_EVENT_TYPE_POISON_CONSUMPTION, "Poison Consumption"},
+};
+
+static ssize_t amdgpu_ras_sysfs_event_state_show(struct device *dev,
+						 struct device_attribute *attr, char *buf)
+{
+	struct amdgpu_ras *con =
+		container_of(attr, struct amdgpu_ras, event_state_attr);
+	struct ras_event_manager *event_mgr = con->event_mgr;
+	struct ras_event_state *event_state;
+	int i, size = 0;
+
+	if (!event_mgr)
+		return -EINVAL;
+
+	size += sysfs_emit_at(buf, size, "current seqno: %llu\n", atomic64_read(&event_mgr->seqno));
+	for (i = 0; i < ARRAY_SIZE(dump_event); i++) {
+		event_state = &event_mgr->event_state[dump_event[i].type];
+		size += sysfs_emit_at(buf, size, "%s: count:%llu, last_seqno:%llu\n",
+				      dump_event[i].name,
+				      atomic64_read(&event_state->count),
+				      event_state->last_seqno);
+	}
+
+	return (ssize_t)size;
+}
+
 static void amdgpu_ras_sysfs_remove_bad_page_node(struct amdgpu_device *adev)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 
-	sysfs_remove_file_from_group(&adev->dev->kobj,
+	if (adev->dev->kobj.sd)
+		sysfs_remove_file_from_group(&adev->dev->kobj,
 				&con->badpages_attr.attr,
 				RAS_FS_NAME);
 }
 
-static int amdgpu_ras_sysfs_remove_feature_node(struct amdgpu_device *adev)
+static int amdgpu_ras_sysfs_remove_dev_attr_node(struct amdgpu_device *adev)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 	struct attribute *attrs[] = {
 		&con->features_attr.attr,
+		&con->version_attr.attr,
+		&con->schema_attr.attr,
+		&con->event_state_attr.attr,
 		NULL
 	};
 	struct attribute_group group = {
@@ -1345,7 +2056,8 @@ static int amdgpu_ras_sysfs_remove_feature_node(struct amdgpu_device *adev)
 		.attrs = attrs,
 	};
 
-	sysfs_remove_group(&adev->dev->kobj, &group);
+	if (adev->dev->kobj.sd)
+		sysfs_remove_group(&adev->dev->kobj, &group);
 
 	return 0;
 }
@@ -1355,9 +2067,15 @@ int amdgpu_ras_sysfs_create(struct amdgpu_device *adev,
 {
 	struct ras_manager *obj = amdgpu_ras_find_obj(adev, head);
 
+	if (amdgpu_aca_is_enabled(adev))
+		return 0;
+
 	if (!obj || obj->attr_inuse)
 		return -EINVAL;
 
+	if (amdgpu_sriov_vf(adev) && !amdgpu_virt_ras_telemetry_block_en(adev, head->block))
+		return 0;
+
 	get_obj(obj);
 
 	snprintf(obj->fs_data.sysfs_name, sizeof(obj->fs_data.sysfs_name),
@@ -1389,10 +2107,14 @@ int amdgpu_ras_sysfs_remove(struct amdgpu_device *adev,
 {
 	struct ras_manager *obj = amdgpu_ras_find_obj(adev, head);
 
+	if (amdgpu_aca_is_enabled(adev))
+		return 0;
+
 	if (!obj || !obj->attr_inuse)
 		return -EINVAL;
 
-	sysfs_remove_file_from_group(&adev->dev->kobj,
+	if (adev->dev->kobj.sd)
+		sysfs_remove_file_from_group(&adev->dev->kobj,
 				&obj->sysfs_attr.attr,
 				RAS_FS_NAME);
 	obj->attr_inuse = 0;
@@ -1413,7 +2135,7 @@ static int amdgpu_ras_sysfs_remove_all(struct amdgpu_device *adev)
 	if (amdgpu_bad_page_threshold != 0)
 		amdgpu_ras_sysfs_remove_bad_page_node(adev);
 
-	amdgpu_ras_sysfs_remove_feature_node(adev);
+	amdgpu_ras_sysfs_remove_dev_attr_node(adev);
 
 	return 0;
 }
@@ -1441,6 +2163,7 @@ static int amdgpu_ras_sysfs_remove_all(struct amdgpu_device *adev)
 static struct dentry *amdgpu_ras_debugfs_create_ctrl_node(struct amdgpu_device *adev)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct amdgpu_ras_eeprom_control *eeprom = &con->eeprom_control;
 	struct drm_minor  *minor = adev_to_drm(adev)->primary;
 	struct dentry     *dir;
 
@@ -1451,6 +2174,7 @@ static struct dentry *amdgpu_ras_debugfs_create_ctrl_node(struct amdgpu_device *
 			    &amdgpu_ras_debugfs_eeprom_ops);
 	debugfs_create_u32("bad_page_cnt_threshold", 0444, dir,
 			   &con->bad_page_cnt_threshold);
+	debugfs_create_u32("ras_num_recs", 0444, dir, &eeprom->ras_num_recs);
 	debugfs_create_x32("ras_hw_enabled", 0444, dir, &adev->ras_hw_enabled);
 	debugfs_create_x32("ras_enabled", 0444, dir, &adev->ras_enabled);
 	debugfs_create_file("ras_eeprom_size", S_IRUGO, dir, adev,
@@ -1498,6 +2222,24 @@ static void amdgpu_ras_debugfs_create(struct amdgpu_device *adev,
 			    obj, &amdgpu_ras_debugfs_ops);
 }
 
+static bool amdgpu_ras_aca_is_supported(struct amdgpu_device *adev)
+{
+	bool ret;
+
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
+		ret = true;
+		break;
+	default:
+		ret = false;
+		break;
+	}
+
+	return ret;
+}
+
 void amdgpu_ras_debugfs_create_all(struct amdgpu_device *adev)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
@@ -1523,15 +2265,28 @@ void amdgpu_ras_debugfs_create_all(struct amdgpu_device *adev)
 			amdgpu_ras_debugfs_create(adev, &fs_info, dir);
 		}
 	}
+
+	if (amdgpu_ras_aca_is_supported(adev)) {
+		if (amdgpu_aca_is_enabled(adev))
+			amdgpu_aca_smu_debugfs_init(adev, dir);
+		else
+			amdgpu_mca_smu_debugfs_init(adev, dir);
+	}
 }
 
 /* debugfs end */
 
 /* ras fs */
-static BIN_ATTR(gpu_vram_bad_pages, S_IRUGO,
-		amdgpu_ras_sysfs_badpages_read, NULL, 0);
+static const BIN_ATTR(gpu_vram_bad_pages, S_IRUGO,
+		      amdgpu_ras_sysfs_badpages_read, NULL, 0);
 static DEVICE_ATTR(features, S_IRUGO,
 		amdgpu_ras_sysfs_features_read, NULL);
+static DEVICE_ATTR(version, 0444,
+		amdgpu_ras_sysfs_version_show, NULL);
+static DEVICE_ATTR(schema, 0444,
+		amdgpu_ras_sysfs_schema_show, NULL);
+static DEVICE_ATTR(event_state, 0444,
+		   amdgpu_ras_sysfs_event_state_show, NULL);
 static int amdgpu_ras_fs_init(struct amdgpu_device *adev)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
@@ -1540,26 +2295,41 @@ static int amdgpu_ras_fs_init(struct amdgpu_device *adev)
 	};
 	struct attribute *attrs[] = {
 		&con->features_attr.attr,
+		&con->version_attr.attr,
+		&con->schema_attr.attr,
+		&con->event_state_attr.attr,
 		NULL
 	};
-	struct bin_attribute *bin_attrs[] = {
+	const struct bin_attribute *bin_attrs[] = {
 		NULL,
 		NULL,
 	};
 	int r;
 
+	group.attrs = attrs;
+
 	/* add features entry */
 	con->features_attr = dev_attr_features;
-	group.attrs = attrs;
 	sysfs_attr_init(attrs[0]);
 
+	/* add version entry */
+	con->version_attr = dev_attr_version;
+	sysfs_attr_init(attrs[1]);
+
+	/* add schema entry */
+	con->schema_attr = dev_attr_schema;
+	sysfs_attr_init(attrs[2]);
+
+	/* add event_state entry */
+	con->event_state_attr = dev_attr_event_state;
+	sysfs_attr_init(attrs[3]);
+
 	if (amdgpu_bad_page_threshold != 0) {
 		/* add bad_page_features entry */
-		bin_attr_gpu_vram_bad_pages.private = NULL;
 		con->badpages_attr = bin_attr_gpu_vram_bad_pages;
+		sysfs_bin_attr_init(&con->badpages_attr);
 		bin_attrs[0] = &con->badpages_attr;
 		group.bin_attrs = bin_attrs;
-		sysfs_bin_attr_init(bin_attrs[0]);
 	}
 
 	r = sysfs_create_group(&adev->dev->kobj, &group);
@@ -1597,10 +2367,24 @@ static int amdgpu_ras_fs_fini(struct amdgpu_device *adev)
 void amdgpu_ras_interrupt_fatal_error_handler(struct amdgpu_device *adev)
 {
 	/* Fatal error events are handled on host side */
-	if (amdgpu_sriov_vf(adev) ||
-		!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__PCIE_BIF))
+	if (amdgpu_sriov_vf(adev))
+		return;
+	/*
+	 * If the current interrupt is caused by a non-fatal RAS error, skip
+	 * check for fatal error. For fatal errors, FED status of all devices
+	 * in XGMI hive gets set when the first device gets fatal error
+	 * interrupt. The error gets propagated to other devices as well, so
+	 * make sure to ack the interrupt regardless of FED status.
+	 */
+	if (!amdgpu_ras_get_fed_status(adev) &&
+	    amdgpu_ras_is_err_state(adev, AMDGPU_RAS_BLOCK__ANY))
 		return;
 
+	if (amdgpu_uniras_enabled(adev)) {
+		amdgpu_ras_mgr_handle_fatal_interrupt(adev, NULL);
+		return;
+	}
+
 	if (adev->nbio.ras &&
 	    adev->nbio.ras->handle_ras_controller_intr_no_bifring)
 		adev->nbio.ras->handle_ras_controller_intr_no_bifring(adev);
@@ -1617,10 +2401,19 @@ static void amdgpu_ras_interrupt_poison_consumption_handler(struct ras_manager *
 	struct amdgpu_device *adev = obj->adev;
 	struct amdgpu_ras_block_object *block_obj =
 		amdgpu_ras_get_ras_block(adev, obj->head.block, 0);
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	enum ras_event_type type = RAS_EVENT_TYPE_POISON_CONSUMPTION;
+	u64 event_id;
+	int ret;
 
-	if (!block_obj)
+	if (!block_obj || !con)
 		return;
 
+	ret = amdgpu_ras_mark_ras_event(adev, type);
+	if (ret)
+		return;
+
+	amdgpu_ras_set_err_poison(adev, block_obj->ras_comm.block);
 	/* both query_poison_status and handle_poison_consumption are optional,
 	 * but at least one of them should be implemented if we need poison
 	 * consumption handler
@@ -1636,42 +2429,69 @@ static void amdgpu_ras_interrupt_poison_consumption_handler(struct ras_manager *
 		}
 	}
 
-	if (!adev->gmc.xgmi.connected_to_cpu)
-		amdgpu_umc_poison_handler(adev, false);
+	amdgpu_umc_poison_handler(adev, obj->head.block, 0);
 
 	if (block_obj->hw_ops && block_obj->hw_ops->handle_poison_consumption)
 		poison_stat = block_obj->hw_ops->handle_poison_consumption(adev);
 
-	/* gpu reset is fallback for failed and default cases */
-	if (poison_stat) {
-		dev_info(adev->dev, "GPU reset for %s RAS poison consumption is issued!\n",
-				block_obj->ras_comm.name);
+	/* gpu reset is fallback for failed and default cases.
+	 * For RMA case, amdgpu_umc_poison_handler will handle gpu reset.
+	 */
+	if (poison_stat && !amdgpu_ras_is_rma(adev)) {
+		event_id = amdgpu_ras_acquire_event_id(adev, type);
+		RAS_EVENT_LOG(adev, event_id,
+			      "GPU reset for %s RAS poison consumption is issued!\n",
+			      block_obj->ras_comm.name);
 		amdgpu_ras_reset_gpu(adev);
-	} else {
-		amdgpu_gfx_poison_consumption_handler(adev, entry);
 	}
+
+	if (!poison_stat)
+		amdgpu_gfx_poison_consumption_handler(adev, entry);
 }
 
 static void amdgpu_ras_interrupt_poison_creation_handler(struct ras_manager *obj,
 				struct amdgpu_iv_entry *entry)
 {
-	dev_info(obj->adev->dev,
-		"Poison is created, no user action is needed.\n");
+	struct amdgpu_device *adev = obj->adev;
+	enum ras_event_type type = RAS_EVENT_TYPE_POISON_CREATION;
+	u64 event_id;
+	int ret;
+
+	ret = amdgpu_ras_mark_ras_event(adev, type);
+	if (ret)
+		return;
+
+	event_id = amdgpu_ras_acquire_event_id(adev, type);
+	RAS_EVENT_LOG(adev, event_id, "Poison is created\n");
+
+	if (amdgpu_ip_version(obj->adev, UMC_HWIP, 0) >= IP_VERSION(12, 0, 0)) {
+		struct amdgpu_ras *con = amdgpu_ras_get_context(obj->adev);
+
+		atomic_inc(&con->page_retirement_req_cnt);
+		atomic_inc(&con->poison_creation_count);
+
+		wake_up(&con->page_retirement_wq);
+	}
 }
 
 static void amdgpu_ras_interrupt_umc_handler(struct ras_manager *obj,
 				struct amdgpu_iv_entry *entry)
 {
 	struct ras_ih_data *data = &obj->ih_data;
-	struct ras_err_data err_data = {0, 0, 0, NULL};
+	struct ras_err_data err_data;
 	int ret;
 
 	if (!data->cb)
 		return;
 
+	ret = amdgpu_ras_error_data_init(&err_data);
+	if (ret)
+		return;
+
 	/* Let IP handle its data, maybe we need get the output
 	 * from the callback to update the error type/count, etc
 	 */
+	amdgpu_ras_set_fed(obj->adev, true);
 	ret = data->cb(obj->adev, &err_data, entry);
 	/* ue will trigger an interrupt, and in that case
 	 * we need do a reset to recovery the whole system.
@@ -1684,7 +2504,10 @@ static void amdgpu_ras_interrupt_umc_handler(struct ras_manager *obj,
 		 */
 		obj->err_data.ue_count += err_data.ue_count;
 		obj->err_data.ce_count += err_data.ce_count;
+		obj->err_data.de_count += err_data.de_count;
 	}
+
+	amdgpu_ras_error_data_fini(&err_data);
 }
 
 static void amdgpu_ras_interrupt_handler(struct ras_manager *obj)
@@ -1729,12 +2552,25 @@ static void amdgpu_ras_interrupt_process_handler(struct work_struct *work)
 int amdgpu_ras_interrupt_dispatch(struct amdgpu_device *adev,
 		struct ras_dispatch_if *info)
 {
-	struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head);
-	struct ras_ih_data *data = &obj->ih_data;
+	struct ras_manager *obj;
+	struct ras_ih_data *data;
+
+	if (amdgpu_uniras_enabled(adev)) {
+		struct ras_ih_info ih_info;
+
+		memset(&ih_info, 0, sizeof(ih_info));
+		ih_info.block = info->head.block;
+		memcpy(&ih_info.iv_entry, info->entry, sizeof(struct amdgpu_iv_entry));
+
+		return amdgpu_ras_mgr_handle_controller_interrupt(adev, &ih_info);
+	}
 
+	obj = amdgpu_ras_find_obj(adev, &info->head);
 	if (!obj)
 		return -EINVAL;
 
+	data = &obj->ih_data;
+
 	if (data->inuse == 0)
 		return 0;
 
@@ -1831,7 +2667,7 @@ static int amdgpu_ras_interrupt_remove_all(struct amdgpu_device *adev)
 /* ih end */
 
 /* traversal all IPs except NBIO to query error counter */
-static void amdgpu_ras_log_on_err_counter(struct amdgpu_device *adev)
+static void amdgpu_ras_log_on_err_counter(struct amdgpu_device *adev, enum ras_event_type type)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 	struct ras_manager *obj;
@@ -1860,14 +2696,18 @@ static void amdgpu_ras_log_on_err_counter(struct amdgpu_device *adev)
 		 * should be removed until smu fix handle ecc_info table.
 		 */
 		if ((info.head.block == AMDGPU_RAS_BLOCK__UMC) &&
-			(adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2)))
+		    (amdgpu_ip_version(adev, MP1_HWIP, 0) ==
+		     IP_VERSION(13, 0, 2)))
 			continue;
 
-		amdgpu_ras_query_error_status(adev, &info);
+		amdgpu_ras_query_error_status_with_event(adev, &info, type);
 
-		if (adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 2) &&
-		    adev->ip_versions[MP0_HWIP][0] != IP_VERSION(11, 0, 4) &&
-		    adev->ip_versions[MP0_HWIP][0] != IP_VERSION(13, 0, 0)) {
+		if (amdgpu_ip_version(adev, MP0_HWIP, 0) !=
+			    IP_VERSION(11, 0, 2) &&
+		    amdgpu_ip_version(adev, MP0_HWIP, 0) !=
+			    IP_VERSION(11, 0, 4) &&
+		    amdgpu_ip_version(adev, MP0_HWIP, 0) !=
+			    IP_VERSION(13, 0, 0)) {
 			if (amdgpu_ras_reset_error_status(adev, info.head.block))
 				dev_warn(adev->dev, "Failed to reset error counter and error status");
 		}
@@ -1919,54 +2759,121 @@ static void amdgpu_ras_query_err_status(struct amdgpu_device *adev)
 	}
 }
 
-/* recovery begin */
-
-/* return 0 on success.
- * caller need free bps.
- */
 static int amdgpu_ras_badpages_read(struct amdgpu_device *adev,
-		struct ras_badpage **bps, unsigned int *count)
+		struct ras_badpage *bps, uint32_t count, uint32_t start)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 	struct ras_err_handler_data *data;
-	int i = 0;
-	int ret = 0, status;
+	int r = 0;
+	uint32_t i;
 
 	if (!con || !con->eh_data || !bps || !count)
 		return -EINVAL;
 
 	mutex_lock(&con->recovery_lock);
 	data = con->eh_data;
-	if (!data || data->count == 0) {
-		*bps = NULL;
-		ret = -EINVAL;
-		goto out;
+	if (start < data->count) {
+		for (i = start; i < data->count; i++) {
+			if (!data->bps[i].ts)
+				continue;
+
+			bps[r].bp = data->bps[i].retired_page;
+			r++;
+			if (r >= count)
+				break;
+		}
 	}
+	mutex_unlock(&con->recovery_lock);
 
-	*bps = kmalloc(sizeof(struct ras_badpage) * data->count, GFP_KERNEL);
-	if (!*bps) {
-		ret = -ENOMEM;
-		goto out;
-	}
+	return r;
+}
 
-	for (; i < data->count; i++) {
-		(*bps)[i] = (struct ras_badpage){
-			.bp = data->bps[i].retired_page,
-			.size = AMDGPU_GPU_PAGE_SIZE,
-			.flags = AMDGPU_RAS_RETIRE_PAGE_RESERVED,
-		};
-		status = amdgpu_vram_mgr_query_page_status(&adev->mman.vram_mgr,
-				data->bps[i].retired_page);
-		if (status == -EBUSY)
-			(*bps)[i].flags = AMDGPU_RAS_RETIRE_PAGE_PENDING;
-		else if (status == -ENOENT)
-			(*bps)[i].flags = AMDGPU_RAS_RETIRE_PAGE_FAULT;
+static int amdgpu_uniras_badpages_read(struct amdgpu_device *adev,
+		struct ras_badpage *bps, uint32_t count, uint32_t start)
+{
+	struct ras_cmd_bad_pages_info_req cmd_input;
+	struct ras_cmd_bad_pages_info_rsp *output;
+	uint32_t group, start_group, end_group;
+	uint32_t pos, pos_in_group;
+	int r = 0, i;
+
+	if (!bps || !count)
+		return -EINVAL;
+
+	output = kmalloc(sizeof(*output), GFP_KERNEL);
+	if (!output)
+		return -ENOMEM;
+
+	memset(&cmd_input, 0, sizeof(cmd_input));
+
+	start_group = start / RAS_CMD_MAX_BAD_PAGES_PER_GROUP;
+	end_group = (start + count + RAS_CMD_MAX_BAD_PAGES_PER_GROUP - 1) /
+				RAS_CMD_MAX_BAD_PAGES_PER_GROUP;
+
+	pos = start;
+	for (group = start_group; group < end_group; group++) {
+		memset(output, 0, sizeof(*output));
+		cmd_input.group_index = group;
+		if (amdgpu_ras_mgr_handle_ras_cmd(adev, RAS_CMD__GET_BAD_PAGES,
+			&cmd_input, sizeof(cmd_input), output, sizeof(*output)))
+			goto out;
+
+		if (pos >= output->bp_total_cnt)
+			goto out;
+
+		pos_in_group = pos - group * RAS_CMD_MAX_BAD_PAGES_PER_GROUP;
+		for (i = pos_in_group; i < output->bp_in_group; i++, pos++) {
+			if (!output->records[i].ts)
+				continue;
+
+			bps[r].bp = output->records[i].retired_page;
+			r++;
+			if (r >= count)
+				goto out;
+		}
 	}
 
-	*count = data->count;
 out:
-	mutex_unlock(&con->recovery_lock);
-	return ret;
+	kfree(output);
+	return r;
+}
+
+static void amdgpu_ras_set_fed_all(struct amdgpu_device *adev,
+				   struct amdgpu_hive_info *hive, bool status)
+{
+	struct amdgpu_device *tmp_adev;
+
+	if (hive) {
+		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head)
+			amdgpu_ras_set_fed(tmp_adev, status);
+	} else {
+		amdgpu_ras_set_fed(adev, status);
+	}
+}
+
+bool amdgpu_ras_in_recovery(struct amdgpu_device *adev)
+{
+	struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev);
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+	int hive_ras_recovery = 0;
+
+	if (hive) {
+		hive_ras_recovery = atomic_read(&hive->ras_recovery);
+		amdgpu_put_xgmi_hive(hive);
+	}
+
+	if (ras && (atomic_read(&ras->in_recovery) || hive_ras_recovery))
+		return true;
+
+	return false;
+}
+
+static enum ras_event_type amdgpu_ras_get_fatal_error_event(struct amdgpu_device *adev)
+{
+	if (amdgpu_ras_intr_triggered())
+		return RAS_EVENT_TYPE_FATAL;
+	else
+		return RAS_EVENT_TYPE_POISON_CONSUMPTION;
 }
 
 static void amdgpu_ras_do_recovery(struct work_struct *work)
@@ -1976,9 +2883,26 @@ static void amdgpu_ras_do_recovery(struct work_struct *work)
 	struct amdgpu_device *remote_adev = NULL;
 	struct amdgpu_device *adev = ras->adev;
 	struct list_head device_list, *device_list_handle =  NULL;
+	struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev);
+	unsigned int error_query_mode;
+	enum ras_event_type type;
 
+	if (hive) {
+		atomic_set(&hive->ras_recovery, 1);
+
+		/* If any device which is part of the hive received RAS fatal
+		 * error interrupt, set fatal error status on all. This
+		 * condition will need a recovery, and flag will be cleared
+		 * as part of recovery.
+		 */
+		list_for_each_entry(remote_adev, &hive->device_list,
+				    gmc.xgmi.head)
+			if (amdgpu_ras_get_fed_status(remote_adev)) {
+				amdgpu_ras_set_fed_all(adev, hive, true);
+				break;
+			}
+	}
 	if (!ras->disable_ras_err_cnt_harvest) {
-		struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev);
 
 		/* Build list of devices to query RAS related errors */
 		if  (hive && adev->gmc.xgmi.num_physical_nodes > 1) {
@@ -1989,13 +2913,24 @@ static void amdgpu_ras_do_recovery(struct work_struct *work)
 			device_list_handle = &device_list;
 		}
 
+		if (amdgpu_ras_get_error_query_mode(adev, &error_query_mode)) {
+			if (error_query_mode == AMDGPU_RAS_FIRMWARE_ERROR_QUERY) {
+				/* wait 500ms to ensure pmfw polling mca bank info done */
+				msleep(500);
+			}
+		}
+
+		type = amdgpu_ras_get_fatal_error_event(adev);
 		list_for_each_entry(remote_adev,
 				device_list_handle, gmc.xgmi.head) {
-			amdgpu_ras_query_err_status(remote_adev);
-			amdgpu_ras_log_on_err_counter(remote_adev);
+			if (amdgpu_uniras_enabled(remote_adev)) {
+				amdgpu_ras_mgr_update_ras_ecc(remote_adev);
+			} else {
+				amdgpu_ras_query_err_status(remote_adev);
+				amdgpu_ras_log_on_err_counter(remote_adev, type);
+			}
 		}
 
-		amdgpu_put_xgmi_hive(hive);
 	}
 
 	if (amdgpu_device_should_recover_gpu(ras->adev)) {
@@ -2004,16 +2939,38 @@ static void amdgpu_ras_do_recovery(struct work_struct *work)
 
 		reset_context.method = AMD_RESET_METHOD_NONE;
 		reset_context.reset_req_dev = adev;
+		reset_context.src = AMDGPU_RESET_SRC_RAS;
+		set_bit(AMDGPU_SKIP_COREDUMP, &reset_context.flags);
 
 		/* Perform full reset in fatal error mode */
 		if (!amdgpu_ras_is_poison_mode_supported(ras->adev))
 			set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
-		else
+		else {
 			clear_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
 
+			if (ras->gpu_reset_flags & AMDGPU_RAS_GPU_RESET_MODE2_RESET) {
+				ras->gpu_reset_flags &= ~AMDGPU_RAS_GPU_RESET_MODE2_RESET;
+				reset_context.method = AMD_RESET_METHOD_MODE2;
+			}
+
+			/* Fatal error occurs in poison mode, mode1 reset is used to
+			 * recover gpu.
+			 */
+			if (ras->gpu_reset_flags & AMDGPU_RAS_GPU_RESET_MODE1_RESET) {
+				ras->gpu_reset_flags &= ~AMDGPU_RAS_GPU_RESET_MODE1_RESET;
+				set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
+
+				psp_fatal_error_recovery_quirk(&adev->psp);
+			}
+		}
+
 		amdgpu_device_gpu_recover(ras->adev, NULL, &reset_context);
 	}
 	atomic_set(&ras->in_recovery, 0);
+	if (hive) {
+		atomic_set(&hive->ras_recovery, 0);
+		amdgpu_put_xgmi_hive(hive);
+	}
 }
 
 /* alloc/realloc bps array */
@@ -2023,7 +2980,7 @@ static int amdgpu_ras_realloc_eh_data_space(struct amdgpu_device *adev,
 	unsigned int old_space = data->count + data->space_left;
 	unsigned int new_space = old_space + pages;
 	unsigned int align_space = ALIGN(new_space, 512);
-	void *bps = kmalloc(align_space * sizeof(*data->bps), GFP_KERNEL);
+	void *bps = kmalloc_array(align_space, sizeof(*data->bps), GFP_KERNEL);
 
 	if (!bps) {
 		return -ENOMEM;
@@ -2040,43 +2997,283 @@ static int amdgpu_ras_realloc_eh_data_space(struct amdgpu_device *adev,
 	return 0;
 }
 
-/* it deal with vram only. */
-int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev,
-		struct eeprom_table_record *bps, int pages)
+static int amdgpu_ras_mca2pa_by_idx(struct amdgpu_device *adev,
+			struct eeprom_table_record *bps,
+			struct ras_err_data *err_data)
 {
-	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
-	struct ras_err_handler_data *data;
+	struct ta_ras_query_address_input addr_in;
+	uint32_t socket = 0;
 	int ret = 0;
-	uint32_t i;
 
-	if (!con || !con->eh_data || !bps || pages <= 0)
-		return 0;
+	if (adev->smuio.funcs && adev->smuio.funcs->get_socket_id)
+		socket = adev->smuio.funcs->get_socket_id(adev);
 
-	mutex_lock(&con->recovery_lock);
-	data = con->eh_data;
-	if (!data)
-		goto out;
+	/* reinit err_data */
+	err_data->err_addr_cnt = 0;
+	err_data->err_addr_len = adev->umc.retire_unit;
+
+	memset(&addr_in, 0, sizeof(addr_in));
+	addr_in.ma.err_addr = bps->address;
+	addr_in.ma.socket_id = socket;
+	addr_in.ma.ch_inst = bps->mem_channel;
+	if (!amdgpu_ras_smu_eeprom_supported(adev)) {
+		/* tell RAS TA the node instance is not used */
+		addr_in.ma.node_inst = TA_RAS_INV_NODE;
+	} else {
+		addr_in.ma.umc_inst = bps->mcumc_id;
+		addr_in.ma.node_inst = bps->cu;
+	}
+
+	if (adev->umc.ras && adev->umc.ras->convert_ras_err_addr)
+		ret = adev->umc.ras->convert_ras_err_addr(adev, err_data,
+				&addr_in, NULL, false);
+
+	return ret;
+}
+
+static int amdgpu_ras_mca2pa(struct amdgpu_device *adev,
+			struct eeprom_table_record *bps,
+			struct ras_err_data *err_data)
+{
+	struct ta_ras_query_address_input addr_in;
+	uint32_t die_id, socket = 0;
+
+	if (adev->smuio.funcs && adev->smuio.funcs->get_socket_id)
+		socket = adev->smuio.funcs->get_socket_id(adev);
+
+	/* although die id is gotten from PA in nps1 mode, the id is
+	 * fitable for any nps mode
+	 */
+	if (adev->umc.ras && adev->umc.ras->get_die_id_from_pa)
+		die_id = adev->umc.ras->get_die_id_from_pa(adev, bps->address,
+					bps->retired_page << AMDGPU_GPU_PAGE_SHIFT);
+	else
+		return -EINVAL;
 
-	for (i = 0; i < pages; i++) {
+	/* reinit err_data */
+	err_data->err_addr_cnt = 0;
+	err_data->err_addr_len = adev->umc.retire_unit;
+
+	memset(&addr_in, 0, sizeof(addr_in));
+	addr_in.ma.err_addr = bps->address;
+	addr_in.ma.ch_inst = bps->mem_channel;
+	addr_in.ma.umc_inst = bps->mcumc_id;
+	addr_in.ma.node_inst = die_id;
+	addr_in.ma.socket_id = socket;
+
+	if (adev->umc.ras && adev->umc.ras->convert_ras_err_addr)
+		return adev->umc.ras->convert_ras_err_addr(adev, err_data,
+					&addr_in, NULL, false);
+	else
+		return  -EINVAL;
+}
+
+static int __amdgpu_ras_restore_bad_pages(struct amdgpu_device *adev,
+					struct eeprom_table_record *bps, int count)
+{
+	int j;
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct ras_err_handler_data *data = con->eh_data;
+
+	for (j = 0; j < count; j++) {
 		if (amdgpu_ras_check_bad_page_unlock(con,
-			bps[i].retired_page << AMDGPU_GPU_PAGE_SHIFT))
+			bps[j].retired_page << AMDGPU_GPU_PAGE_SHIFT)) {
+			data->count++;
+			data->space_left--;
 			continue;
+		}
 
 		if (!data->space_left &&
-			amdgpu_ras_realloc_eh_data_space(adev, data, 256)) {
-			ret = -ENOMEM;
-			goto out;
+		    amdgpu_ras_realloc_eh_data_space(adev, data, 256)) {
+			return -ENOMEM;
 		}
 
-		amdgpu_vram_mgr_reserve_range(&adev->mman.vram_mgr,
-			bps[i].retired_page << AMDGPU_GPU_PAGE_SHIFT,
-			AMDGPU_GPU_PAGE_SIZE);
+		amdgpu_ras_reserve_page(adev, bps[j].retired_page);
 
-		memcpy(&data->bps[data->count], &bps[i], sizeof(*data->bps));
+		memcpy(&data->bps[data->count], &(bps[j]),
+				sizeof(struct eeprom_table_record));
 		data->count++;
 		data->space_left--;
+		con->bad_page_num++;
+	}
+
+	return 0;
+}
+
+static int __amdgpu_ras_convert_rec_array_from_rom(struct amdgpu_device *adev,
+				struct eeprom_table_record *bps, struct ras_err_data *err_data,
+				enum amdgpu_memory_partition nps)
+{
+	int i = 0;
+	enum amdgpu_memory_partition save_nps;
+
+	save_nps = (bps[0].retired_page >> UMC_NPS_SHIFT) & UMC_NPS_MASK;
+
+	/*old asics just have pa in eeprom*/
+	if (IP_VERSION_MAJ(amdgpu_ip_version(adev, UMC_HWIP, 0)) < 12) {
+		memcpy(err_data->err_addr, bps,
+			sizeof(struct eeprom_table_record) * adev->umc.retire_unit);
+		goto out;
 	}
+
+	for (i = 0; i < adev->umc.retire_unit; i++)
+		bps[i].retired_page &= ~(UMC_NPS_MASK << UMC_NPS_SHIFT);
+
+	if (save_nps) {
+		if (save_nps == nps) {
+			if (amdgpu_umc_pages_in_a_row(adev, err_data,
+					bps[0].retired_page << AMDGPU_GPU_PAGE_SHIFT))
+				return -EINVAL;
+			for (i = 0; i < adev->umc.retire_unit; i++) {
+				err_data->err_addr[i].address = bps[0].address;
+				err_data->err_addr[i].mem_channel = bps[0].mem_channel;
+				err_data->err_addr[i].bank = bps[0].bank;
+				err_data->err_addr[i].err_type = bps[0].err_type;
+				err_data->err_addr[i].mcumc_id = bps[0].mcumc_id;
+			}
+		} else {
+			if (amdgpu_ras_mca2pa_by_idx(adev, &bps[0], err_data))
+				return -EINVAL;
+		}
+	} else {
+		if (bps[0].address == 0) {
+			/* for specific old eeprom data, mca address is not stored,
+			 * calc it from pa
+			 */
+			if (amdgpu_umc_pa2mca(adev, bps[0].retired_page << AMDGPU_GPU_PAGE_SHIFT,
+				&(bps[0].address), AMDGPU_NPS1_PARTITION_MODE))
+				return -EINVAL;
+		}
+
+		if (amdgpu_ras_mca2pa(adev, &bps[0], err_data)) {
+			if (nps == AMDGPU_NPS1_PARTITION_MODE)
+				memcpy(err_data->err_addr, bps,
+					sizeof(struct eeprom_table_record) * adev->umc.retire_unit);
+			else
+				return -EOPNOTSUPP;
+		}
+	}
+
 out:
+	return __amdgpu_ras_restore_bad_pages(adev, err_data->err_addr, adev->umc.retire_unit);
+}
+
+static int __amdgpu_ras_convert_rec_from_rom(struct amdgpu_device *adev,
+				struct eeprom_table_record *bps, struct ras_err_data *err_data,
+				enum amdgpu_memory_partition nps)
+{
+	int i = 0;
+	enum amdgpu_memory_partition save_nps;
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev)) {
+		save_nps = (bps->retired_page >> UMC_NPS_SHIFT) & UMC_NPS_MASK;
+		bps->retired_page &= ~(UMC_NPS_MASK << UMC_NPS_SHIFT);
+	} else {
+		/* if pmfw manages eeprom, save_nps is not stored on eeprom,
+		 * we should always convert mca address into physical address,
+		 * make save_nps different from nps
+		 */
+		save_nps = nps + 1;
+	}
+
+	if (save_nps == nps) {
+		if (amdgpu_umc_pages_in_a_row(adev, err_data,
+				bps->retired_page << AMDGPU_GPU_PAGE_SHIFT))
+			return -EINVAL;
+		for (i = 0; i < adev->umc.retire_unit; i++) {
+			err_data->err_addr[i].address = bps->address;
+			err_data->err_addr[i].mem_channel = bps->mem_channel;
+			err_data->err_addr[i].bank = bps->bank;
+			err_data->err_addr[i].err_type = bps->err_type;
+			err_data->err_addr[i].mcumc_id = bps->mcumc_id;
+		}
+	} else {
+		if (bps->address) {
+			if (amdgpu_ras_mca2pa_by_idx(adev, bps, err_data))
+				return -EINVAL;
+		} else {
+			/* for specific old eeprom data, mca address is not stored,
+			 * calc it from pa
+			 */
+			if (amdgpu_umc_pa2mca(adev, bps->retired_page << AMDGPU_GPU_PAGE_SHIFT,
+				&(bps->address), AMDGPU_NPS1_PARTITION_MODE))
+				return -EINVAL;
+
+			if (amdgpu_ras_mca2pa(adev, bps, err_data))
+				return -EOPNOTSUPP;
+		}
+	}
+
+	return __amdgpu_ras_restore_bad_pages(adev, err_data->err_addr,
+									adev->umc.retire_unit);
+}
+
+/* it deal with vram only. */
+int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev,
+		struct eeprom_table_record *bps, int pages, bool from_rom)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct ras_err_data err_data;
+	struct amdgpu_ras_eeprom_control *control =
+			&adev->psp.ras_context.ras->eeprom_control;
+	enum amdgpu_memory_partition nps = AMDGPU_NPS1_PARTITION_MODE;
+	int ret = 0;
+	uint32_t i = 0;
+
+	if (!con || !con->eh_data || !bps || pages <= 0)
+		return 0;
+
+	if (from_rom) {
+		err_data.err_addr =
+			kcalloc(adev->umc.retire_unit,
+				sizeof(struct eeprom_table_record), GFP_KERNEL);
+		if (!err_data.err_addr) {
+			dev_warn(adev->dev, "Failed to alloc UMC error address record in mca2pa conversion!\n");
+			return -ENOMEM;
+		}
+
+		if (adev->gmc.gmc_funcs->query_mem_partition_mode)
+			nps = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+	}
+
+	mutex_lock(&con->recovery_lock);
+
+	if (from_rom) {
+		/* there is no pa recs in V3, so skip pa recs processing */
+		if ((control->tbl_hdr.version < RAS_TABLE_VER_V3) &&
+		    !amdgpu_ras_smu_eeprom_supported(adev)) {
+			for (i = 0; i < pages; i++) {
+				if (control->ras_num_recs - i >= adev->umc.retire_unit) {
+					if ((bps[i].address == bps[i + 1].address) &&
+						(bps[i].mem_channel == bps[i + 1].mem_channel)) {
+						/* deal with retire_unit records a time */
+						ret = __amdgpu_ras_convert_rec_array_from_rom(adev,
+										&bps[i], &err_data, nps);
+						if (ret)
+							con->bad_page_num -= adev->umc.retire_unit;
+						i += (adev->umc.retire_unit - 1);
+					} else {
+						break;
+					}
+				} else {
+					break;
+				}
+			}
+		}
+		for (; i < pages; i++) {
+			ret = __amdgpu_ras_convert_rec_from_rom(adev,
+				&bps[i], &err_data, nps);
+			if (ret)
+				con->bad_page_num -= adev->umc.retire_unit;
+		}
+
+		con->eh_data->count_saved = con->eh_data->count;
+	} else {
+		ret = __amdgpu_ras_restore_bad_pages(adev, bps, pages);
+	}
+
+	if (from_rom)
+		kfree(err_data.err_addr);
 	mutex_unlock(&con->recovery_lock);
 
 	return ret;
@@ -2093,7 +3290,7 @@ int amdgpu_ras_save_bad_pages(struct amdgpu_device *adev,
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 	struct ras_err_handler_data *data;
 	struct amdgpu_ras_eeprom_control *control;
-	int save_count;
+	int save_count, unit_num, i;
 
 	if (!con || !con->eh_data) {
 		if (new_cnt)
@@ -2102,25 +3299,53 @@ int amdgpu_ras_save_bad_pages(struct amdgpu_device *adev,
 		return 0;
 	}
 
+	if (!con->eeprom_control.is_eeprom_valid) {
+		dev_warn(adev->dev,
+			"Failed to save EEPROM table data because of EEPROM data corruption!");
+		if (new_cnt)
+			*new_cnt = 0;
+
+		return 0;
+	}
+
 	mutex_lock(&con->recovery_lock);
 	control = &con->eeprom_control;
 	data = con->eh_data;
-	save_count = data->count - control->ras_num_recs;
+	if (amdgpu_ras_smu_eeprom_supported(adev))
+		unit_num = control->ras_num_recs -
+			control->ras_num_recs_old;
+	else
+		unit_num = data->count / adev->umc.retire_unit -
+			control->ras_num_recs;
+
+	save_count = con->bad_page_num - control->ras_num_bad_pages;
 	mutex_unlock(&con->recovery_lock);
 
 	if (new_cnt)
-		*new_cnt = save_count / adev->umc.retire_unit;
+		*new_cnt = unit_num;
 
 	/* only new entries are saved */
-	if (save_count > 0) {
-		if (amdgpu_ras_eeprom_append(control,
-					     &data->bps[control->ras_num_recs],
-					     save_count)) {
-			dev_err(adev->dev, "Failed to save EEPROM table data!");
-			return -EIO;
+	if (unit_num && save_count) {
+		/*old asics only save pa to eeprom like before*/
+		if (IP_VERSION_MAJ(amdgpu_ip_version(adev, UMC_HWIP, 0)) < 12) {
+			if (amdgpu_ras_eeprom_append(control,
+					&data->bps[data->count_saved], unit_num)) {
+				dev_err(adev->dev, "Failed to save EEPROM table data!");
+				return -EIO;
+			}
+		} else {
+			for (i = 0; i < unit_num; i++) {
+				if (amdgpu_ras_eeprom_append(control,
+						&data->bps[data->count_saved +
+						i * adev->umc.retire_unit], 1)) {
+					dev_err(adev->dev, "Failed to save EEPROM table data!");
+					return -EIO;
+				}
+			}
 		}
 
 		dev_info(adev->dev, "Saved %d pages to EEPROM table.\n", save_count);
+		data->count_saved = data->count;
 	}
 
 	return 0;
@@ -2135,7 +3360,7 @@ static int amdgpu_ras_load_bad_pages(struct amdgpu_device *adev)
 	struct amdgpu_ras_eeprom_control *control =
 		&adev->psp.ras_context.ras->eeprom_control;
 	struct eeprom_table_record *bps;
-	int ret;
+	int ret, i = 0;
 
 	/* no bad page record, skip eeprom access */
 	if (control->ras_num_recs == 0 || amdgpu_bad_page_threshold == 0)
@@ -2146,27 +3371,72 @@ static int amdgpu_ras_load_bad_pages(struct amdgpu_device *adev)
 		return -ENOMEM;
 
 	ret = amdgpu_ras_eeprom_read(control, bps, control->ras_num_recs);
-	if (ret)
+	if (ret) {
 		dev_err(adev->dev, "Failed to load EEPROM table records!");
-	else
-		ret = amdgpu_ras_add_bad_pages(adev, bps, control->ras_num_recs);
+	} else {
+		if (adev->umc.ras && adev->umc.ras->convert_ras_err_addr) {
+			/*In V3, there is no pa recs, and some cases(when address==0) may be parsed
+			as pa recs, so add verion check to avoid it.
+			*/
+			if ((control->tbl_hdr.version < RAS_TABLE_VER_V3) &&
+			    !amdgpu_ras_smu_eeprom_supported(adev)) {
+				for (i = 0; i < control->ras_num_recs; i++) {
+					if ((control->ras_num_recs - i) >= adev->umc.retire_unit) {
+						if ((bps[i].address == bps[i + 1].address) &&
+							(bps[i].mem_channel == bps[i + 1].mem_channel)) {
+							control->ras_num_pa_recs += adev->umc.retire_unit;
+							i += (adev->umc.retire_unit - 1);
+						} else {
+							control->ras_num_mca_recs +=
+										(control->ras_num_recs - i);
+							break;
+						}
+					} else {
+						control->ras_num_mca_recs += (control->ras_num_recs - i);
+						break;
+					}
+				}
+			} else {
+				control->ras_num_mca_recs = control->ras_num_recs;
+			}
+		}
+
+		ret = amdgpu_ras_add_bad_pages(adev, bps, control->ras_num_recs, true);
+		if (ret)
+			goto out;
 
+		ret = amdgpu_ras_eeprom_check(control);
+		if (ret)
+			goto out;
+
+		/* HW not usable */
+		if (amdgpu_ras_is_rma(adev))
+			ret = -EHWPOISON;
+	}
+
+out:
 	kfree(bps);
 	return ret;
 }
 
-static bool amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con,
+static int amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con,
 				uint64_t addr)
 {
 	struct ras_err_handler_data *data = con->eh_data;
+	struct amdgpu_device *adev = con->adev;
 	int i;
 
+	if ((addr >= adev->gmc.mc_vram_size &&
+	    adev->gmc.mc_vram_size) ||
+	    (addr >= RAS_UMC_INJECT_ADDR_LIMIT))
+		return -EINVAL;
+
 	addr >>= AMDGPU_GPU_PAGE_SHIFT;
 	for (i = 0; i < data->count; i++)
 		if (addr == data->bps[i].retired_page)
-			return true;
+			return 1;
 
-	return false;
+	return 0;
 }
 
 /*
@@ -2174,11 +3444,11 @@ static bool amdgpu_ras_check_bad_page_unlock(struct amdgpu_ras *con,
  *
  * Note: this check is only for umc block
  */
-static bool amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
+static int amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
 				uint64_t addr)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
-	bool ret = false;
+	int ret = 0;
 
 	if (!con || !con->eh_data)
 		return ret;
@@ -2195,43 +3465,379 @@ static void amdgpu_ras_validate_threshold(struct amdgpu_device *adev,
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 
 	/*
-	 * Justification of value bad_page_cnt_threshold in ras structure
-	 *
-	 * Generally, 0 <= amdgpu_bad_page_threshold <= max record length
-	 * in eeprom or amdgpu_bad_page_threshold == -2, introduce two
-	 * scenarios accordingly.
-	 *
-	 * Bad page retirement enablement:
-	 *    - If amdgpu_bad_page_threshold = -2,
-	 *      bad_page_cnt_threshold = typical value by formula.
-	 *
-	 *    - When the value from user is 0 < amdgpu_bad_page_threshold <
-	 *      max record length in eeprom, use it directly.
-	 *
-	 * Bad page retirement disablement:
-	 *    - If amdgpu_bad_page_threshold = 0, bad page retirement
-	 *      functionality is disabled, and bad_page_cnt_threshold will
-	 *      take no effect.
+	 * amdgpu_bad_page_threshold is used to config
+	 * the threshold for the number of bad pages.
+	 * -1:  Threshold is set to default value
+	 *      Driver will issue a warning message when threshold is reached
+	 *      and continue runtime services.
+	 * 0:   Disable bad page retirement
+	 *      Driver will not retire bad pages
+	 *      which is intended for debugging purpose.
+	 * -2:  Threshold is determined by a formula
+	 *      that assumes 1 bad page per 100M of local memory.
+	 *      Driver will continue runtime services when threhold is reached.
+	 * 0 < threshold < max number of bad page records in EEPROM,
+	 *      A user-defined threshold is set
+	 *      Driver will halt runtime services when this custom threshold is reached.
 	 */
-
-	if (amdgpu_bad_page_threshold < 0) {
+	if (amdgpu_bad_page_threshold == -2) {
 		u64 val = adev->gmc.mc_vram_size;
 
 		do_div(val, RAS_BAD_PAGE_COVER);
 		con->bad_page_cnt_threshold = min(lower_32_bits(val),
 						  max_count);
+	} else if (amdgpu_bad_page_threshold == -1) {
+		con->bad_page_cnt_threshold = ((con->reserved_pages_in_bytes) >> 21) << 4;
 	} else {
 		con->bad_page_cnt_threshold = min_t(int, max_count,
 						    amdgpu_bad_page_threshold);
 	}
 }
 
-int amdgpu_ras_recovery_init(struct amdgpu_device *adev)
+int amdgpu_ras_put_poison_req(struct amdgpu_device *adev,
+		enum amdgpu_ras_block block, uint16_t pasid,
+		pasid_notify pasid_fn, void *data, uint32_t reset)
+{
+	int ret = 0;
+	struct ras_poison_msg poison_msg;
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+
+	memset(&poison_msg, 0, sizeof(poison_msg));
+	poison_msg.block = block;
+	poison_msg.pasid = pasid;
+	poison_msg.reset = reset;
+	poison_msg.pasid_fn = pasid_fn;
+	poison_msg.data = data;
+
+	ret = kfifo_put(&con->poison_fifo, poison_msg);
+	if (!ret) {
+		dev_err(adev->dev, "Poison message fifo is full!\n");
+		return -ENOSPC;
+	}
+
+	return 0;
+}
+
+static int amdgpu_ras_get_poison_req(struct amdgpu_device *adev,
+		struct ras_poison_msg *poison_msg)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+
+	return kfifo_get(&con->poison_fifo, poison_msg);
+}
+
+static void amdgpu_ras_ecc_log_init(struct ras_ecc_log_info *ecc_log)
+{
+	mutex_init(&ecc_log->lock);
+
+	INIT_RADIX_TREE(&ecc_log->de_page_tree, GFP_KERNEL);
+	ecc_log->de_queried_count = 0;
+	ecc_log->consumption_q_count = 0;
+}
+
+static void amdgpu_ras_ecc_log_fini(struct ras_ecc_log_info *ecc_log)
+{
+	struct radix_tree_iter iter;
+	void __rcu **slot;
+	struct ras_ecc_err *ecc_err;
+
+	mutex_lock(&ecc_log->lock);
+	radix_tree_for_each_slot(slot, &ecc_log->de_page_tree, &iter, 0) {
+		ecc_err = radix_tree_deref_slot(slot);
+		kfree(ecc_err->err_pages.pfn);
+		kfree(ecc_err);
+		radix_tree_iter_delete(&ecc_log->de_page_tree, &iter, slot);
+	}
+	mutex_unlock(&ecc_log->lock);
+
+	mutex_destroy(&ecc_log->lock);
+	ecc_log->de_queried_count = 0;
+	ecc_log->consumption_q_count = 0;
+}
+
+static bool amdgpu_ras_schedule_retirement_dwork(struct amdgpu_ras *con,
+				uint32_t delayed_ms)
+{
+	int ret;
+
+	mutex_lock(&con->umc_ecc_log.lock);
+	ret = radix_tree_tagged(&con->umc_ecc_log.de_page_tree,
+			UMC_ECC_NEW_DETECTED_TAG);
+	mutex_unlock(&con->umc_ecc_log.lock);
+
+	if (ret)
+		schedule_delayed_work(&con->page_retirement_dwork,
+			msecs_to_jiffies(delayed_ms));
+
+	return ret ? true : false;
+}
+
+static void amdgpu_ras_do_page_retirement(struct work_struct *work)
+{
+	struct amdgpu_ras *con = container_of(work, struct amdgpu_ras,
+					      page_retirement_dwork.work);
+	struct amdgpu_device *adev = con->adev;
+	struct ras_err_data err_data;
+
+	/* If gpu reset is ongoing, delay retiring the bad pages */
+	if (amdgpu_in_reset(adev) || amdgpu_ras_in_recovery(adev)) {
+		amdgpu_ras_schedule_retirement_dwork(con,
+				AMDGPU_RAS_RETIRE_PAGE_INTERVAL * 3);
+		return;
+	}
+
+	amdgpu_ras_error_data_init(&err_data);
+
+	amdgpu_umc_handle_bad_pages(adev, &err_data);
+
+	amdgpu_ras_error_data_fini(&err_data);
+
+	amdgpu_ras_schedule_retirement_dwork(con,
+			AMDGPU_RAS_RETIRE_PAGE_INTERVAL);
+}
+
+static int amdgpu_ras_poison_creation_handler(struct amdgpu_device *adev,
+				uint32_t poison_creation_count)
+{
+	int ret = 0;
+	struct ras_ecc_log_info *ecc_log;
+	struct ras_query_if info;
+	u32 timeout = MAX_UMC_POISON_POLLING_TIME_ASYNC;
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+	u64 de_queried_count;
+	u64 consumption_q_count;
+	enum ras_event_type type = RAS_EVENT_TYPE_POISON_CREATION;
+
+	memset(&info, 0, sizeof(info));
+	info.head.block = AMDGPU_RAS_BLOCK__UMC;
+
+	ecc_log = &ras->umc_ecc_log;
+	ecc_log->de_queried_count = 0;
+	ecc_log->consumption_q_count = 0;
+
+	do {
+		ret = amdgpu_ras_query_error_status_with_event(adev, &info, type);
+		if (ret)
+			return ret;
+
+		de_queried_count = ecc_log->de_queried_count;
+		consumption_q_count = ecc_log->consumption_q_count;
+
+		if (de_queried_count && consumption_q_count)
+			break;
+
+		msleep(100);
+	} while (--timeout);
+
+	if (de_queried_count)
+		schedule_delayed_work(&ras->page_retirement_dwork, 0);
+
+	if (amdgpu_ras_is_rma(adev) && atomic_cmpxchg(&ras->rma_in_recovery, 0, 1) == 0)
+		amdgpu_ras_reset_gpu(adev);
+
+	return 0;
+}
+
+static void amdgpu_ras_clear_poison_fifo(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct ras_poison_msg msg;
+	int ret;
+
+	do {
+		ret = kfifo_get(&con->poison_fifo, &msg);
+	} while (ret);
+}
+
+static int amdgpu_ras_poison_consumption_handler(struct amdgpu_device *adev,
+			uint32_t msg_count, uint32_t *gpu_reset)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	uint32_t reset_flags = 0, reset = 0;
+	struct ras_poison_msg msg;
+	int ret, i;
+
+	kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
+
+	for (i = 0; i < msg_count; i++) {
+		ret = amdgpu_ras_get_poison_req(adev, &msg);
+		if (!ret)
+			continue;
+
+		if (msg.pasid_fn)
+			msg.pasid_fn(adev, msg.pasid, msg.data);
+
+		reset_flags |= msg.reset;
+	}
+
+	/*
+	 * Try to ensure poison creation handler is completed first
+	 * to set rma if bad page exceed threshold.
+	 */
+	flush_delayed_work(&con->page_retirement_dwork);
+
+	/* for RMA, amdgpu_ras_poison_creation_handler will trigger gpu reset */
+	if (reset_flags && !amdgpu_ras_is_rma(adev)) {
+		if (reset_flags & AMDGPU_RAS_GPU_RESET_MODE1_RESET)
+			reset = AMDGPU_RAS_GPU_RESET_MODE1_RESET;
+		else if (reset_flags & AMDGPU_RAS_GPU_RESET_MODE2_RESET)
+			reset = AMDGPU_RAS_GPU_RESET_MODE2_RESET;
+		else
+			reset = reset_flags;
+
+		con->gpu_reset_flags |= reset;
+		amdgpu_ras_reset_gpu(adev);
+
+		*gpu_reset = reset;
+
+		/* Wait for gpu recovery to complete */
+		flush_work(&con->recovery_work);
+	}
+
+	return 0;
+}
+
+static int amdgpu_ras_page_retirement_thread(void *param)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)param;
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	uint32_t poison_creation_count, msg_count;
+	uint32_t gpu_reset;
+	int ret;
+
+	while (!kthread_should_stop()) {
+
+		wait_event_interruptible(con->page_retirement_wq,
+				kthread_should_stop() ||
+				atomic_read(&con->page_retirement_req_cnt));
+
+		if (kthread_should_stop())
+			break;
+
+		mutex_lock(&con->poison_lock);
+		gpu_reset = 0;
+
+		do {
+			poison_creation_count = atomic_read(&con->poison_creation_count);
+			ret = amdgpu_ras_poison_creation_handler(adev, poison_creation_count);
+			if (ret == -EIO)
+				break;
+
+			if (poison_creation_count) {
+				atomic_sub(poison_creation_count, &con->poison_creation_count);
+				atomic_sub(poison_creation_count, &con->page_retirement_req_cnt);
+			}
+		} while (atomic_read(&con->poison_creation_count) &&
+			!atomic_read(&con->poison_consumption_count));
+
+		if (ret != -EIO) {
+			msg_count = kfifo_len(&con->poison_fifo);
+			if (msg_count) {
+				ret = amdgpu_ras_poison_consumption_handler(adev,
+						msg_count, &gpu_reset);
+				if ((ret != -EIO) &&
+				    (gpu_reset != AMDGPU_RAS_GPU_RESET_MODE1_RESET))
+					atomic_sub(msg_count, &con->page_retirement_req_cnt);
+			}
+		}
+
+		if ((ret == -EIO) || (gpu_reset == AMDGPU_RAS_GPU_RESET_MODE1_RESET)) {
+			/* gpu mode-1 reset is ongoing or just completed ras mode-1 reset */
+			/* Clear poison creation request */
+			atomic_set(&con->poison_creation_count, 0);
+			atomic_set(&con->poison_consumption_count, 0);
+
+			/* Clear poison fifo */
+			amdgpu_ras_clear_poison_fifo(adev);
+
+			/* Clear all poison requests */
+			atomic_set(&con->page_retirement_req_cnt, 0);
+
+			if (ret == -EIO) {
+				/* Wait for mode-1 reset to complete */
+				down_read(&adev->reset_domain->sem);
+				up_read(&adev->reset_domain->sem);
+			}
+
+			/* Wake up work to save bad pages to eeprom */
+			schedule_delayed_work(&con->page_retirement_dwork, 0);
+		} else if (gpu_reset) {
+			/* gpu just completed mode-2 reset or other reset */
+			/* Clear poison consumption messages cached in fifo */
+			msg_count = kfifo_len(&con->poison_fifo);
+			if (msg_count) {
+				amdgpu_ras_clear_poison_fifo(adev);
+				atomic_sub(msg_count, &con->page_retirement_req_cnt);
+			}
+
+			atomic_set(&con->poison_consumption_count, 0);
+
+			/* Wake up work to save bad pages to eeprom */
+			schedule_delayed_work(&con->page_retirement_dwork, 0);
+		}
+		mutex_unlock(&con->poison_lock);
+	}
+
+	return 0;
+}
+
+int amdgpu_ras_init_badpage_info(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct amdgpu_ras_eeprom_control *control;
+	int ret;
+
+	if (!con || amdgpu_sriov_vf(adev))
+		return 0;
+
+	if (amdgpu_uniras_enabled(adev))
+		return 0;
+
+	control = &con->eeprom_control;
+	con->ras_smu_drv = amdgpu_dpm_get_ras_smu_driver(adev);
+
+	ret = amdgpu_ras_eeprom_init(control);
+	control->is_eeprom_valid = !ret;
+
+	if (!adev->umc.ras || !adev->umc.ras->convert_ras_err_addr)
+		control->ras_num_pa_recs = control->ras_num_recs;
+
+	if (adev->umc.ras &&
+	    adev->umc.ras->get_retire_flip_bits)
+		adev->umc.ras->get_retire_flip_bits(adev);
+
+	if (control->ras_num_recs && control->is_eeprom_valid) {
+		ret = amdgpu_ras_load_bad_pages(adev);
+		if (ret) {
+			control->is_eeprom_valid = false;
+			return 0;
+		}
+
+		amdgpu_dpm_send_hbm_bad_pages_num(
+			adev, control->ras_num_bad_pages);
+
+		if (con->update_channel_flag == true) {
+			amdgpu_dpm_send_hbm_bad_channel_flag(
+				adev, control->bad_channel_bitmap);
+			con->update_channel_flag = false;
+		}
+
+		/* The format action is only applied to new ASICs */
+		if (IP_VERSION_MAJ(amdgpu_ip_version(adev, UMC_HWIP, 0)) >= 12 &&
+		    control->tbl_hdr.version < RAS_TABLE_VER_V3)
+			if (!amdgpu_ras_eeprom_reset_table(control))
+				if (amdgpu_ras_save_bad_pages(adev, NULL))
+					dev_warn(adev->dev, "Failed to format RAS EEPROM data in V3 version!\n");
+	}
+
+	return 0;
+}
+
+int amdgpu_ras_recovery_init(struct amdgpu_device *adev, bool init_bp_info)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 	struct ras_err_handler_data **data;
 	u32  max_eeprom_records_count = 0;
-	bool exc_err_limit = false;
 	int ret;
 
 	if (!con || amdgpu_sriov_vf(adev))
@@ -2248,47 +3854,44 @@ int amdgpu_ras_recovery_init(struct amdgpu_device *adev)
 		return 0;
 
 	data = &con->eh_data;
-	*data = kmalloc(sizeof(**data), GFP_KERNEL | __GFP_ZERO);
+	*data = kzalloc(sizeof(**data), GFP_KERNEL);
 	if (!*data) {
 		ret = -ENOMEM;
 		goto out;
 	}
 
 	mutex_init(&con->recovery_lock);
+	mutex_init(&con->poison_lock);
 	INIT_WORK(&con->recovery_work, amdgpu_ras_do_recovery);
 	atomic_set(&con->in_recovery, 0);
+	atomic_set(&con->rma_in_recovery, 0);
 	con->eeprom_control.bad_channel_bitmap = 0;
 
-	max_eeprom_records_count = amdgpu_ras_eeprom_max_record_count();
+	max_eeprom_records_count = amdgpu_ras_eeprom_max_record_count(&con->eeprom_control);
 	amdgpu_ras_validate_threshold(adev, max_eeprom_records_count);
 
-	/* Todo: During test the SMU might fail to read the eeprom through I2C
-	 * when the GPU is pending on XGMI reset during probe time
-	 * (Mostly after second bus reset), skip it now
-	 */
-	if (adev->gmc.xgmi.pending_reset)
-		return 0;
-	ret = amdgpu_ras_eeprom_init(&con->eeprom_control, &exc_err_limit);
-	/*
-	 * This calling fails when exc_err_limit is true or
-	 * ret != 0.
-	 */
-	if (exc_err_limit || ret)
-		goto free;
-
-	if (con->eeprom_control.ras_num_recs) {
-		ret = amdgpu_ras_load_bad_pages(adev);
+	if (init_bp_info) {
+		ret = amdgpu_ras_init_badpage_info(adev);
 		if (ret)
 			goto free;
+	}
 
-		amdgpu_dpm_send_hbm_bad_pages_num(adev, con->eeprom_control.ras_num_recs);
-
-		if (con->update_channel_flag == true) {
-			amdgpu_dpm_send_hbm_bad_channel_flag(adev, con->eeprom_control.bad_channel_bitmap);
-			con->update_channel_flag = false;
-		}
+	mutex_init(&con->page_rsv_lock);
+	INIT_KFIFO(con->poison_fifo);
+	mutex_init(&con->page_retirement_lock);
+	init_waitqueue_head(&con->page_retirement_wq);
+	atomic_set(&con->page_retirement_req_cnt, 0);
+	atomic_set(&con->poison_creation_count, 0);
+	atomic_set(&con->poison_consumption_count, 0);
+	con->page_retirement_thread =
+		kthread_run(amdgpu_ras_page_retirement_thread, adev, "umc_page_retirement");
+	if (IS_ERR(con->page_retirement_thread)) {
+		con->page_retirement_thread = NULL;
+		dev_warn(adev->dev, "Failed to create umc_page_retirement thread!!!\n");
 	}
 
+	INIT_DELAYED_WORK(&con->page_retirement_dwork, amdgpu_ras_do_page_retirement);
+	amdgpu_ras_ecc_log_init(&con->umc_ecc_log);
 #ifdef CONFIG_X86_MCE_AMD
 	if ((adev->asic_type == CHIP_ALDEBARAN) &&
 	    (adev->gmc.xgmi.connected_to_cpu))
@@ -2307,7 +3910,7 @@ out:
 	 * Except error threshold exceeding case, other failure cases in this
 	 * function would not fail amdgpu driver init.
 	 */
-	if (!exc_err_limit)
+	if (!amdgpu_ras_is_rma(adev))
 		ret = 0;
 	else
 		ret = -EINVAL;
@@ -2319,19 +3922,43 @@ static int amdgpu_ras_recovery_fini(struct amdgpu_device *adev)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 	struct ras_err_handler_data *data = con->eh_data;
+	int max_flush_timeout = MAX_FLUSH_RETIRE_DWORK_TIMES;
+	bool ret;
 
 	/* recovery_init failed to init it, fini is useless */
 	if (!data)
 		return 0;
 
+	/* Save all cached bad pages to eeprom */
+	do {
+		flush_delayed_work(&con->page_retirement_dwork);
+		ret = amdgpu_ras_schedule_retirement_dwork(con, 0);
+	} while (ret && max_flush_timeout--);
+
+	if (con->page_retirement_thread)
+		kthread_stop(con->page_retirement_thread);
+
+	atomic_set(&con->page_retirement_req_cnt, 0);
+	atomic_set(&con->poison_creation_count, 0);
+
+	mutex_destroy(&con->page_rsv_lock);
+
 	cancel_work_sync(&con->recovery_work);
 
+	cancel_delayed_work_sync(&con->page_retirement_dwork);
+
+	amdgpu_ras_ecc_log_fini(&con->umc_ecc_log);
+
 	mutex_lock(&con->recovery_lock);
 	con->eh_data = NULL;
 	kfree(data->bps);
 	kfree(data);
 	mutex_unlock(&con->recovery_lock);
 
+	amdgpu_ras_critical_region_init(adev);
+#ifdef CONFIG_X86_MCE_AMD
+	amdgpu_unregister_bad_pages_mca_notifier(adev);
+#endif
 	return 0;
 }
 /* recovery end */
@@ -2339,9 +3966,11 @@ static int amdgpu_ras_recovery_fini(struct amdgpu_device *adev)
 static bool amdgpu_ras_asic_supported(struct amdgpu_device *adev)
 {
 	if (amdgpu_sriov_vf(adev)) {
-		switch (adev->ip_versions[MP0_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 		case IP_VERSION(13, 0, 2):
-		case IP_VERSION(13, 0, 10):
+		case IP_VERSION(13, 0, 6):
+		case IP_VERSION(13, 0, 12):
+		case IP_VERSION(13, 0, 14):
 			return true;
 		default:
 			return false;
@@ -2349,9 +3978,13 @@ static bool amdgpu_ras_asic_supported(struct amdgpu_device *adev)
 	}
 
 	if (adev->asic_type == CHIP_IP_DISCOVERY) {
-		switch (adev->ip_versions[MP0_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 		case IP_VERSION(13, 0, 0):
+		case IP_VERSION(13, 0, 6):
 		case IP_VERSION(13, 0, 10):
+		case IP_VERSION(13, 0, 12):
+		case IP_VERSION(13, 0, 14):
+		case IP_VERSION(14, 0, 3):
 			return true;
 		default:
 			return false;
@@ -2377,13 +4010,95 @@ static void amdgpu_ras_get_quirks(struct amdgpu_device *adev)
 	if (!ctx)
 		return;
 
-	if (strnstr(ctx->vbios_version, "D16406",
-		    sizeof(ctx->vbios_version)) ||
-		strnstr(ctx->vbios_version, "D36002",
-			sizeof(ctx->vbios_version)))
+	if (strnstr(ctx->vbios_pn, "D16406",
+		    sizeof(ctx->vbios_pn)) ||
+		strnstr(ctx->vbios_pn, "D36002",
+			sizeof(ctx->vbios_pn)))
 		adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__GFX);
 }
 
+/* Query ras capablity via atomfirmware interface */
+static void amdgpu_ras_query_ras_capablity_from_vbios(struct amdgpu_device *adev)
+{
+	/* mem_ecc cap */
+	if (amdgpu_atomfirmware_mem_ecc_supported(adev)) {
+		dev_info(adev->dev, "MEM ECC is active.\n");
+		adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__UMC |
+					 1 << AMDGPU_RAS_BLOCK__DF);
+	} else {
+		dev_info(adev->dev, "MEM ECC is not presented.\n");
+	}
+
+	/* sram_ecc cap */
+	if (amdgpu_atomfirmware_sram_ecc_supported(adev)) {
+		dev_info(adev->dev, "SRAM ECC is active.\n");
+		if (!amdgpu_sriov_vf(adev))
+			adev->ras_hw_enabled |= ~(1 << AMDGPU_RAS_BLOCK__UMC |
+						  1 << AMDGPU_RAS_BLOCK__DF);
+		else
+			adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__PCIE_BIF |
+						 1 << AMDGPU_RAS_BLOCK__SDMA |
+						 1 << AMDGPU_RAS_BLOCK__GFX);
+
+		/*
+		 * VCN/JPEG RAS can be supported on both bare metal and
+		 * SRIOV environment
+		 */
+		if (amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(2, 6, 0) ||
+		    amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(4, 0, 0) ||
+		    amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(4, 0, 3) ||
+		    amdgpu_ip_version(adev, VCN_HWIP, 0) == IP_VERSION(5, 0, 1))
+			adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__VCN |
+						 1 << AMDGPU_RAS_BLOCK__JPEG);
+		else
+			adev->ras_hw_enabled &= ~(1 << AMDGPU_RAS_BLOCK__VCN |
+						  1 << AMDGPU_RAS_BLOCK__JPEG);
+
+		/*
+		 * XGMI RAS is not supported if xgmi num physical nodes
+		 * is zero
+		 */
+		if (!adev->gmc.xgmi.num_physical_nodes)
+			adev->ras_hw_enabled &= ~(1 << AMDGPU_RAS_BLOCK__XGMI_WAFL);
+	} else {
+		dev_info(adev->dev, "SRAM ECC is not presented.\n");
+	}
+}
+
+/* Query poison mode from umc/df IP callbacks */
+static void amdgpu_ras_query_poison_mode(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	bool df_poison, umc_poison;
+
+	/* poison setting is useless on SRIOV guest */
+	if (amdgpu_sriov_vf(adev) || !con)
+		return;
+
+	/* Init poison supported flag, the default value is false */
+	if (adev->gmc.xgmi.connected_to_cpu ||
+	    adev->gmc.is_app_apu) {
+		/* enabled by default when GPU is connected to CPU */
+		con->poison_supported = true;
+	} else if (adev->df.funcs &&
+	    adev->df.funcs->query_ras_poison_mode &&
+	    adev->umc.ras &&
+	    adev->umc.ras->query_ras_poison_mode) {
+		df_poison =
+			adev->df.funcs->query_ras_poison_mode(adev);
+		umc_poison =
+			adev->umc.ras->query_ras_poison_mode(adev);
+
+		/* Only poison is set in both DF and UMC, we can support it */
+		if (df_poison && umc_poison)
+			con->poison_supported = true;
+		else if (df_poison != umc_poison)
+			dev_warn(adev->dev,
+				"Poison setting is inconsistent in DF/UMC(%d:%d)!\n",
+				df_poison, umc_poison);
+	}
+}
+
 /*
  * check hardware's ras ability which will be saved in hw_supported.
  * if hardware does not support ras, we can skip some ras initializtion and
@@ -2397,42 +4112,21 @@ static void amdgpu_ras_check_supported(struct amdgpu_device *adev)
 {
 	adev->ras_hw_enabled = adev->ras_enabled = 0;
 
-	if (!adev->is_atom_fw ||
-	    !amdgpu_ras_asic_supported(adev))
+	if (!amdgpu_ras_asic_supported(adev))
 		return;
 
-	if (!adev->gmc.xgmi.connected_to_cpu) {
-		if (amdgpu_atomfirmware_mem_ecc_supported(adev)) {
-			dev_info(adev->dev, "MEM ECC is active.\n");
-			adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__UMC |
-						   1 << AMDGPU_RAS_BLOCK__DF);
-		} else {
-			dev_info(adev->dev, "MEM ECC is not presented.\n");
-		}
+	if (amdgpu_sriov_vf(adev)) {
+		if (amdgpu_virt_get_ras_capability(adev))
+			goto init_ras_enabled_flag;
+	}
 
-		if (amdgpu_atomfirmware_sram_ecc_supported(adev)) {
-			dev_info(adev->dev, "SRAM ECC is active.\n");
-			if (!amdgpu_sriov_vf(adev))
-				adev->ras_hw_enabled |= ~(1 << AMDGPU_RAS_BLOCK__UMC |
-							    1 << AMDGPU_RAS_BLOCK__DF);
-			else
-				adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__PCIE_BIF |
-								1 << AMDGPU_RAS_BLOCK__SDMA |
-								1 << AMDGPU_RAS_BLOCK__GFX);
+	/* query ras capability from psp */
+	if (amdgpu_psp_get_ras_capability(&adev->psp))
+		goto init_ras_enabled_flag;
 
-			/* VCN/JPEG RAS can be supported on both bare metal and
-			 * SRIOV environment
-			 */
-			if (adev->ip_versions[VCN_HWIP][0] == IP_VERSION(2, 6, 0) ||
-			    adev->ip_versions[VCN_HWIP][0] == IP_VERSION(4, 0, 0))
-				adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__VCN |
-							1 << AMDGPU_RAS_BLOCK__JPEG);
-			else
-				adev->ras_hw_enabled &= ~(1 << AMDGPU_RAS_BLOCK__VCN |
-							1 << AMDGPU_RAS_BLOCK__JPEG);
-		} else {
-			dev_info(adev->dev, "SRAM ECC is not presented.\n");
-		}
+	/* query ras capablity from bios */
+	if (!adev->gmc.xgmi.connected_to_cpu && !adev->gmc.is_app_apu) {
+		amdgpu_ras_query_ras_capablity_from_vbios(adev);
 	} else {
 		/* driver only manages a few IP blocks RAS feature
 		 * when GPU is connected cpu through XGMI */
@@ -2441,13 +4135,31 @@ static void amdgpu_ras_check_supported(struct amdgpu_device *adev)
 					   1 << AMDGPU_RAS_BLOCK__MMHUB);
 	}
 
+	/* apply asic specific settings (vega20 only for now) */
 	amdgpu_ras_get_quirks(adev);
 
+	/* query poison mode from umc/df ip callback */
+	amdgpu_ras_query_poison_mode(adev);
+
+init_ras_enabled_flag:
 	/* hw_supported needs to be aligned with RAS block mask. */
 	adev->ras_hw_enabled &= AMDGPU_RAS_BLOCK_MASK;
 
 	adev->ras_enabled = amdgpu_ras_enable == 0 ? 0 :
 		adev->ras_hw_enabled & amdgpu_ras_mask;
+
+	/* aca is disabled by default except for psp v13_0_6/v13_0_12/v13_0_14 */
+	if (!amdgpu_sriov_vf(adev)) {
+		adev->aca.is_enabled =
+			(amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 6) ||
+			amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 12) ||
+			amdgpu_ip_version(adev, MP0_HWIP, 0) == IP_VERSION(13, 0, 14));
+	}
+
+	/* bad page feature is not applicable to specific app platform */
+	if (adev->gmc.is_app_apu &&
+	    amdgpu_ip_version(adev, UMC_HWIP, 0) == IP_VERSION(12, 0, 0))
+		amdgpu_bad_page_threshold = 0;
 }
 
 static void amdgpu_ras_counte_dw(struct work_struct *work)
@@ -2470,40 +4182,72 @@ static void amdgpu_ras_counte_dw(struct work_struct *work)
 		atomic_set(&con->ras_ue_count, ue_count);
 	}
 
-	pm_runtime_mark_last_busy(dev->dev);
 Out:
 	pm_runtime_put_autosuspend(dev->dev);
 }
 
-static void amdgpu_ras_query_poison_mode(struct amdgpu_device *adev)
+static int amdgpu_get_ras_schema(struct amdgpu_device *adev)
 {
-	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
-	bool df_poison, umc_poison;
+	return  amdgpu_ras_is_poison_mode_supported(adev) ? AMDGPU_RAS_ERROR__POISON : 0 |
+			AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE |
+			AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE |
+			AMDGPU_RAS_ERROR__PARITY;
+}
 
-	/* poison setting is useless on SRIOV guest */
-	if (amdgpu_sriov_vf(adev) || !con)
+static void ras_event_mgr_init(struct ras_event_manager *mgr)
+{
+	struct ras_event_state *event_state;
+	int i;
+
+	memset(mgr, 0, sizeof(*mgr));
+	atomic64_set(&mgr->seqno, 0);
+
+	for (i = 0; i < ARRAY_SIZE(mgr->event_state); i++) {
+		event_state = &mgr->event_state[i];
+		event_state->last_seqno = RAS_EVENT_INVALID_ID;
+		atomic64_set(&event_state->count, 0);
+	}
+}
+
+static void amdgpu_ras_event_mgr_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+	struct amdgpu_hive_info *hive;
+
+	if (!ras)
 		return;
 
-	/* Init poison supported flag, the default value is false */
-	if (adev->gmc.xgmi.connected_to_cpu) {
-		/* enabled by default when GPU is connected to CPU */
-		con->poison_supported = true;
-	} else if (adev->df.funcs &&
-	    adev->df.funcs->query_ras_poison_mode &&
-	    adev->umc.ras &&
-	    adev->umc.ras->query_ras_poison_mode) {
-		df_poison =
-			adev->df.funcs->query_ras_poison_mode(adev);
-		umc_poison =
-			adev->umc.ras->query_ras_poison_mode(adev);
+	hive = amdgpu_get_xgmi_hive(adev);
+	ras->event_mgr = hive ? &hive->event_mgr : &ras->__event_mgr;
 
-		/* Only poison is set in both DF and UMC, we can support it */
-		if (df_poison && umc_poison)
-			con->poison_supported = true;
-		else if (df_poison != umc_poison)
-			dev_warn(adev->dev,
-				"Poison setting is inconsistent in DF/UMC(%d:%d)!\n",
-				df_poison, umc_poison);
+	/* init event manager with node 0 on xgmi system */
+	if (!amdgpu_reset_in_recovery(adev)) {
+		if (!hive || adev->gmc.xgmi.node_id == 0)
+			ras_event_mgr_init(ras->event_mgr);
+	}
+
+	if (hive)
+		amdgpu_put_xgmi_hive(hive);
+}
+
+static void amdgpu_ras_init_reserved_vram_size(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+
+	if (!con || (adev->flags & AMD_IS_APU))
+		return;
+
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
+	case IP_VERSION(13, 0, 2):
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 12):
+		con->reserved_pages_in_bytes = AMDGPU_RAS_RESERVED_VRAM_SIZE_DEFAULT;
+		break;
+	case IP_VERSION(13, 0, 14):
+		con->reserved_pages_in_bytes = (AMDGPU_RAS_RESERVED_VRAM_SIZE_DEFAULT << 1);
+		break;
+	default:
+		break;
 	}
 }
 
@@ -2515,10 +4259,10 @@ int amdgpu_ras_init(struct amdgpu_device *adev)
 	if (con)
 		return 0;
 
-	con = kmalloc(sizeof(struct amdgpu_ras) +
+	con = kzalloc(sizeof(*con) +
 			sizeof(struct ras_manager) * AMDGPU_RAS_BLOCK_COUNT +
 			sizeof(struct ras_manager) * AMDGPU_RAS_MCA_BLOCK_COUNT,
-			GFP_KERNEL|__GFP_ZERO);
+			GFP_KERNEL);
 	if (!con)
 		return -ENOMEM;
 
@@ -2549,6 +4293,7 @@ int amdgpu_ras_init(struct amdgpu_device *adev)
 
 	con->update_channel_flag = false;
 	con->features = 0;
+	con->schema = 0;
 	INIT_LIST_HEAD(&con->head);
 	/* Might need get this flag from vbios. */
 	con->flags = RAS_DEFAULT_FLAGS;
@@ -2556,7 +4301,7 @@ int amdgpu_ras_init(struct amdgpu_device *adev)
 	/* initialize nbio ras function ahead of any other
 	 * ras functions so hardware fatal error interrupt
 	 * can be enabled as early as possible */
-	switch (adev->ip_versions[NBIO_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, NBIO_HWIP, 0)) {
 	case IP_VERSION(7, 4, 0):
 	case IP_VERSION(7, 4, 1):
 	case IP_VERSION(7, 4, 4):
@@ -2573,6 +4318,22 @@ int amdgpu_ras_init(struct amdgpu_device *adev)
 			 * check DF RAS */
 			adev->nbio.ras = &nbio_v4_3_ras;
 		break;
+	case IP_VERSION(6, 3, 1):
+		if (adev->ras_hw_enabled & (1 << AMDGPU_RAS_BLOCK__DF))
+			/* unlike other generation of nbio ras,
+			 * nbif v6_3_1 only support fatal error interrupt
+			 * to inform software that DF is freezed due to
+			 * system fatal error event. driver should not
+			 * enable nbio ras in such case. Instead,
+			 * check DF RAS
+			 */
+			adev->nbio.ras = &nbif_v6_3_1_ras;
+		break;
+	case IP_VERSION(7, 9, 0):
+	case IP_VERSION(7, 9, 1):
+		if (!adev->gmc.is_app_apu)
+			adev->nbio.ras = &nbio_v7_9_ras;
+		break;
 	default:
 		/* nbio ras is not available */
 		break;
@@ -2598,13 +4359,37 @@ int amdgpu_ras_init(struct amdgpu_device *adev)
 			goto release_con;
 	}
 
-	amdgpu_ras_query_poison_mode(adev);
+	/* Packed socket_id to ras feature mask bits[31:29] */
+	if (adev->smuio.funcs &&
+	    adev->smuio.funcs->get_socket_id)
+		con->features |= ((adev->smuio.funcs->get_socket_id(adev)) <<
+					AMDGPU_RAS_FEATURES_SOCKETID_SHIFT);
+
+	/* Get RAS schema for particular SOC */
+	con->schema = amdgpu_get_ras_schema(adev);
+
+	amdgpu_ras_init_reserved_vram_size(adev);
 
 	if (amdgpu_ras_fs_init(adev)) {
 		r = -EINVAL;
 		goto release_con;
 	}
 
+	if (amdgpu_ras_aca_is_supported(adev)) {
+		if (amdgpu_aca_is_enabled(adev))
+			r = amdgpu_aca_init(adev);
+		else
+			r = amdgpu_mca_init(adev);
+		if (r)
+			goto release_con;
+	}
+
+	con->init_task_pid = task_pid_nr(current);
+	get_task_comm(con->init_task_comm, current);
+
+	mutex_init(&con->critical_region_lock);
+	INIT_LIST_HEAD(&con->critical_region_head);
+
 	dev_info(adev->dev, "RAS INFO: ras initialized successfully, "
 		 "hardware ability[%x] ras_mask[%x]\n",
 		 adev->ras_hw_enabled, adev->ras_enabled);
@@ -2619,7 +4404,8 @@ release_con:
 
 int amdgpu_persistent_edc_harvesting_supported(struct amdgpu_device *adev)
 {
-	if (adev->gmc.xgmi.connected_to_cpu)
+	if (adev->gmc.xgmi.connected_to_cpu ||
+	    adev->gmc.is_app_apu)
 		return 1;
 	return 0;
 }
@@ -2671,7 +4457,7 @@ int amdgpu_ras_block_late_init(struct amdgpu_device *adev,
 
 	r = amdgpu_ras_feature_enable_on_boot(adev, ras_block, 1);
 	if (r) {
-		if (adev->in_suspend || amdgpu_in_reset(adev)) {
+		if (adev->in_suspend || amdgpu_reset_in_recovery(adev)) {
 			/* in resume phase, if fail to enable ras,
 			 * clean up all ras fs nodes, and disable ras */
 			goto cleanup;
@@ -2683,7 +4469,7 @@ int amdgpu_ras_block_late_init(struct amdgpu_device *adev,
 	amdgpu_persistent_edc_harvesting(adev, ras_block);
 
 	/* in resume phase, no need to create ras fs node */
-	if (adev->in_suspend || amdgpu_in_reset(adev))
+	if (adev->in_suspend || amdgpu_reset_in_recovery(adev))
 		return 0;
 
 	ras_obj = container_of(ras_block, struct amdgpu_ras_block_object, ras_comm);
@@ -2695,23 +4481,28 @@ int amdgpu_ras_block_late_init(struct amdgpu_device *adev,
 			goto cleanup;
 	}
 
-	r = amdgpu_ras_sysfs_create(adev, ras_block);
-	if (r)
-		goto interrupt;
+	if (ras_obj->hw_ops &&
+	    (ras_obj->hw_ops->query_ras_error_count ||
+	     ras_obj->hw_ops->query_ras_error_status)) {
+		r = amdgpu_ras_sysfs_create(adev, ras_block);
+		if (r)
+			goto interrupt;
 
-	/* Those are the cached values at init.
-	 */
-	query_info = kzalloc(sizeof(struct ras_query_if), GFP_KERNEL);
-	if (!query_info)
-		return -ENOMEM;
-	memcpy(&query_info->head, ras_block, sizeof(struct ras_common_if));
+		/* Those are the cached values at init.
+		 */
+		query_info = kzalloc(sizeof(*query_info), GFP_KERNEL);
+		if (!query_info)
+			return -ENOMEM;
+		memcpy(&query_info->head, ras_block, sizeof(struct ras_common_if));
 
-	if (amdgpu_ras_query_error_count(adev, &ce_count, &ue_count, query_info) == 0) {
-		atomic_set(&con->ras_ce_count, ce_count);
-		atomic_set(&con->ras_ue_count, ue_count);
+		if (amdgpu_ras_query_error_count(adev, &ce_count, &ue_count, query_info) == 0) {
+			atomic_set(&con->ras_ce_count, ce_count);
+			atomic_set(&con->ras_ue_count, ue_count);
+		}
+
+		kfree(query_info);
 	}
 
-	kfree(query_info);
 	return 0;
 
 interrupt:
@@ -2795,7 +4586,7 @@ void amdgpu_ras_suspend(struct amdgpu_device *adev)
 
 	amdgpu_ras_disable_all_features(adev, 0);
 	/* Make sure all ras objects are disabled. */
-	if (con->features)
+	if (AMDGPU_RAS_GET_FEATURES(con->features))
 		amdgpu_ras_disable_all_features(adev, 1);
 }
 
@@ -2805,17 +4596,40 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev)
 	struct amdgpu_ras_block_object *obj;
 	int r;
 
+	amdgpu_ras_event_mgr_init(adev);
+
+	if (amdgpu_ras_aca_is_supported(adev)) {
+		if (amdgpu_reset_in_recovery(adev)) {
+			if (amdgpu_aca_is_enabled(adev))
+				r = amdgpu_aca_reset(adev);
+			else
+				r = amdgpu_mca_reset(adev);
+			if (r)
+				return r;
+		}
+
+		if (!amdgpu_sriov_vf(adev)) {
+			if (amdgpu_aca_is_enabled(adev))
+				amdgpu_ras_set_aca_debug_mode(adev, false);
+			else
+				amdgpu_ras_set_mca_debug_mode(adev, false);
+		}
+	}
+
 	/* Guest side doesn't need init ras feature */
-	if (amdgpu_sriov_vf(adev))
+	if (amdgpu_sriov_vf(adev) && !amdgpu_sriov_ras_telemetry_en(adev))
 		return 0;
 
 	list_for_each_entry_safe(node, tmp, &adev->ras_list, node) {
-		if (!node->ras_obj) {
+		obj = node->ras_obj;
+		if (!obj) {
 			dev_warn(adev->dev, "Warning: abnormal ras list node.\n");
 			continue;
 		}
 
-		obj = node->ras_obj;
+		if (!amdgpu_ras_is_supported(adev, obj->ras_comm.block))
+			continue;
+
 		if (obj->ras_late_init) {
 			r = obj->ras_late_init(adev, &obj->ras_comm);
 			if (r) {
@@ -2840,7 +4654,7 @@ int amdgpu_ras_pre_fini(struct amdgpu_device *adev)
 
 
 	/* Need disable ras on all IPs here before ip [hw/sw]fini */
-	if (con->features)
+	if (AMDGPU_RAS_GET_FEATURES(con->features))
 		amdgpu_ras_disable_all_features(adev, 0);
 	amdgpu_ras_recovery_fini(adev);
 	return 0;
@@ -2855,6 +4669,9 @@ int amdgpu_ras_fini(struct amdgpu_device *adev)
 	if (!adev->ras_enabled || !con)
 		return 0;
 
+	amdgpu_ras_critical_region_fini(adev);
+	mutex_destroy(&con->critical_region_lock);
+
 	list_for_each_entry_safe(ras_node, tmp, &adev->ras_list, node) {
 		if (ras_node->ras_obj) {
 			obj = ras_node->ras_obj;
@@ -2873,10 +4690,17 @@ int amdgpu_ras_fini(struct amdgpu_device *adev)
 	amdgpu_ras_fs_fini(adev);
 	amdgpu_ras_interrupt_remove_all(adev);
 
-	WARN(con->features, "Feature mask is not cleared");
+	if (amdgpu_ras_aca_is_supported(adev)) {
+		if (amdgpu_aca_is_enabled(adev))
+			amdgpu_aca_fini(adev);
+		else
+			amdgpu_mca_fini(adev);
+	}
 
-	if (con->features)
-		amdgpu_ras_disable_all_features(adev, 1);
+	WARN(AMDGPU_RAS_GET_FEATURES(con->features), "Feature mask is not cleared");
+
+	if (AMDGPU_RAS_GET_FEATURES(con->features))
+		amdgpu_ras_disable_all_features(adev, 0);
 
 	cancel_delayed_work_sync(&con->ras_counte_delay_work);
 
@@ -2886,18 +4710,161 @@ int amdgpu_ras_fini(struct amdgpu_device *adev)
 	return 0;
 }
 
-void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev)
+bool amdgpu_ras_get_fed_status(struct amdgpu_device *adev)
 {
-	amdgpu_ras_check_supported(adev);
-	if (!adev->ras_hw_enabled)
-		return;
+	struct amdgpu_ras *ras;
+
+	ras = amdgpu_ras_get_context(adev);
+	if (!ras)
+		return false;
+
+	return test_bit(AMDGPU_RAS_BLOCK__LAST, &ras->ras_err_state);
+}
+
+void amdgpu_ras_set_fed(struct amdgpu_device *adev, bool status)
+{
+	struct amdgpu_ras *ras;
+
+	ras = amdgpu_ras_get_context(adev);
+	if (ras) {
+		if (status)
+			set_bit(AMDGPU_RAS_BLOCK__LAST, &ras->ras_err_state);
+		else
+			clear_bit(AMDGPU_RAS_BLOCK__LAST, &ras->ras_err_state);
+	}
+}
+
+void amdgpu_ras_clear_err_state(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *ras;
+
+	ras = amdgpu_ras_get_context(adev);
+	if (ras) {
+		ras->ras_err_state = 0;
+		ras->gpu_reset_flags = 0;
+	}
+}
 
+void amdgpu_ras_set_err_poison(struct amdgpu_device *adev,
+			       enum amdgpu_ras_block block)
+{
+	struct amdgpu_ras *ras;
+
+	ras = amdgpu_ras_get_context(adev);
+	if (ras)
+		set_bit(block, &ras->ras_err_state);
+}
+
+bool amdgpu_ras_is_err_state(struct amdgpu_device *adev, int block)
+{
+	struct amdgpu_ras *ras;
+
+	ras = amdgpu_ras_get_context(adev);
+	if (ras) {
+		if (block == AMDGPU_RAS_BLOCK__ANY)
+			return (ras->ras_err_state != 0);
+		else
+			return test_bit(block, &ras->ras_err_state) ||
+			       test_bit(AMDGPU_RAS_BLOCK__LAST,
+					&ras->ras_err_state);
+	}
+
+	return false;
+}
+
+static struct ras_event_manager *__get_ras_event_mgr(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *ras;
+
+	ras = amdgpu_ras_get_context(adev);
+	if (!ras)
+		return NULL;
+
+	return ras->event_mgr;
+}
+
+int amdgpu_ras_mark_ras_event_caller(struct amdgpu_device *adev, enum ras_event_type type,
+				     const void *caller)
+{
+	struct ras_event_manager *event_mgr;
+	struct ras_event_state *event_state;
+	int ret = 0;
+
+	if (amdgpu_uniras_enabled(adev))
+		return 0;
+
+	if (type >= RAS_EVENT_TYPE_COUNT) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	event_mgr = __get_ras_event_mgr(adev);
+	if (!event_mgr) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	event_state = &event_mgr->event_state[type];
+	event_state->last_seqno = atomic64_inc_return(&event_mgr->seqno);
+	atomic64_inc(&event_state->count);
+
+out:
+	if (ret && caller)
+		dev_warn(adev->dev, "failed mark ras event (%d) in %ps, ret:%d\n",
+			 (int)type, caller, ret);
+
+	return ret;
+}
+
+u64 amdgpu_ras_acquire_event_id(struct amdgpu_device *adev, enum ras_event_type type)
+{
+	struct ras_event_manager *event_mgr;
+	u64 id;
+
+	if (type >= RAS_EVENT_TYPE_COUNT)
+		return RAS_EVENT_INVALID_ID;
+
+	switch (type) {
+	case RAS_EVENT_TYPE_FATAL:
+	case RAS_EVENT_TYPE_POISON_CREATION:
+	case RAS_EVENT_TYPE_POISON_CONSUMPTION:
+		event_mgr = __get_ras_event_mgr(adev);
+		if (!event_mgr)
+			return RAS_EVENT_INVALID_ID;
+
+		id = event_mgr->event_state[type].last_seqno;
+		break;
+	case RAS_EVENT_TYPE_INVALID:
+	default:
+		id = RAS_EVENT_INVALID_ID;
+		break;
+	}
+
+	return id;
+}
+
+int amdgpu_ras_global_ras_isr(struct amdgpu_device *adev)
+{
 	if (atomic_cmpxchg(&amdgpu_ras_in_intr, 0, 1) == 0) {
-		dev_info(adev->dev, "uncorrectable hardware error"
-			"(ERREVENT_ATHUB_INTERRUPT) detected!\n");
+		struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+		enum ras_event_type type = RAS_EVENT_TYPE_FATAL;
+		u64 event_id = RAS_EVENT_INVALID_ID;
 
+		if (amdgpu_uniras_enabled(adev))
+			return 0;
+
+		if (!amdgpu_ras_mark_ras_event(adev, type))
+			event_id = amdgpu_ras_acquire_event_id(adev, type);
+
+		RAS_EVENT_LOG(adev, event_id, "uncorrectable hardware error"
+			      "(ERREVENT_ATHUB_INTERRUPT) detected!\n");
+
+		amdgpu_ras_set_fed(adev, true);
+		ras->gpu_reset_flags |= AMDGPU_RAS_GPU_RESET_MODE1_RESET;
 		amdgpu_ras_reset_gpu(adev);
 	}
+
+	return -EBUSY;
 }
 
 bool amdgpu_ras_need_emergency_restart(struct amdgpu_device *adev)
@@ -3025,6 +4992,28 @@ static void amdgpu_register_bad_pages_mca_notifier(struct amdgpu_device *adev)
 		notifier_registered = true;
 	}
 }
+static void amdgpu_unregister_bad_pages_mca_notifier(struct amdgpu_device *adev)
+{
+	int i, j;
+
+	if (!notifier_registered && !mce_adev_list.num_gpu)
+		return;
+	for (i = 0, j = 0; i < mce_adev_list.num_gpu; i++) {
+		if (mce_adev_list.devs[i] == adev)
+			mce_adev_list.devs[i] = NULL;
+		if (!mce_adev_list.devs[i])
+			++j;
+	}
+
+	if (j == mce_adev_list.num_gpu) {
+		mce_adev_list.num_gpu = 0;
+		/* Unregister x86 notifier with MCE subsystem. */
+		if (notifier_registered) {
+			mce_unregister_decode_chain(&amdgpu_bad_page_nb);
+			notifier_registered = false;
+		}
+	}
+}
 #endif
 
 struct amdgpu_ras *amdgpu_ras_get_context(struct amdgpu_device *adev)
@@ -3063,6 +5052,11 @@ int amdgpu_ras_is_supported(struct amdgpu_device *adev,
 	 * that the ras block supports ras function.
 	 */
 	if (!ret &&
+	    (block == AMDGPU_RAS_BLOCK__GFX ||
+	     block == AMDGPU_RAS_BLOCK__SDMA ||
+	     block == AMDGPU_RAS_BLOCK__VCN ||
+	     block == AMDGPU_RAS_BLOCK__JPEG) &&
+		(amdgpu_ras_mask & (1 << block)) &&
 	    amdgpu_ras_is_poison_mode_supported(adev) &&
 	    amdgpu_ras_get_ras_block(adev, block, 0))
 		ret = 1;
@@ -3074,11 +5068,106 @@ int amdgpu_ras_reset_gpu(struct amdgpu_device *adev)
 {
 	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 
-	if (atomic_cmpxchg(&ras->in_recovery, 0, 1) == 0)
+	/* mode1 is the only selection for RMA status */
+	if (amdgpu_ras_is_rma(adev)) {
+		ras->gpu_reset_flags = 0;
+		ras->gpu_reset_flags |= AMDGPU_RAS_GPU_RESET_MODE1_RESET;
+	}
+
+	if (atomic_cmpxchg(&ras->in_recovery, 0, 1) == 0) {
+		struct amdgpu_hive_info *hive = amdgpu_get_xgmi_hive(adev);
+		int hive_ras_recovery = 0;
+
+		if (hive) {
+			hive_ras_recovery = atomic_read(&hive->ras_recovery);
+			amdgpu_put_xgmi_hive(hive);
+		}
+		/* In the case of multiple GPUs, after a GPU has started
+		 * resetting all GPUs on hive, other GPUs do not need to
+		 * trigger GPU reset again.
+		 */
+		if (!hive_ras_recovery)
+			amdgpu_reset_domain_schedule(ras->adev->reset_domain, &ras->recovery_work);
+		else
+			atomic_set(&ras->in_recovery, 0);
+	} else {
+		flush_work(&ras->recovery_work);
 		amdgpu_reset_domain_schedule(ras->adev->reset_domain, &ras->recovery_work);
+	}
+
 	return 0;
 }
 
+int amdgpu_ras_set_mca_debug_mode(struct amdgpu_device *adev, bool enable)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	int ret = 0;
+
+	if (con) {
+		ret = amdgpu_mca_smu_set_debug_mode(adev, enable);
+		if (!ret)
+			con->is_aca_debug_mode = enable;
+	}
+
+	return ret;
+}
+
+int amdgpu_ras_set_aca_debug_mode(struct amdgpu_device *adev, bool enable)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	int ret = 0;
+
+	if (con) {
+		if (amdgpu_aca_is_enabled(adev))
+			ret = amdgpu_aca_smu_set_debug_mode(adev, enable);
+		else
+			ret = amdgpu_mca_smu_set_debug_mode(adev, enable);
+		if (!ret)
+			con->is_aca_debug_mode = enable;
+	}
+
+	return ret;
+}
+
+bool amdgpu_ras_get_aca_debug_mode(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	const struct aca_smu_funcs *smu_funcs = adev->aca.smu_funcs;
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+
+	if (!con)
+		return false;
+
+	if ((amdgpu_aca_is_enabled(adev) && smu_funcs && smu_funcs->set_debug_mode) ||
+	    (!amdgpu_aca_is_enabled(adev) && mca_funcs && mca_funcs->mca_set_debug_mode))
+		return con->is_aca_debug_mode;
+	else
+		return true;
+}
+
+bool amdgpu_ras_get_error_query_mode(struct amdgpu_device *adev,
+				     unsigned int *error_query_mode)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	const struct amdgpu_mca_smu_funcs *mca_funcs = adev->mca.mca_funcs;
+	const struct aca_smu_funcs *smu_funcs = adev->aca.smu_funcs;
+
+	if (!con) {
+		*error_query_mode = AMDGPU_RAS_INVALID_ERROR_QUERY;
+		return false;
+	}
+
+	if (amdgpu_sriov_vf(adev)) {
+		*error_query_mode = AMDGPU_RAS_VIRT_ERROR_COUNT_QUERY;
+	} else if ((smu_funcs && smu_funcs->set_debug_mode) || (mca_funcs && mca_funcs->mca_set_debug_mode)) {
+		*error_query_mode =
+			(con->is_aca_debug_mode) ? AMDGPU_RAS_DIRECT_ERROR_QUERY : AMDGPU_RAS_FIRMWARE_ERROR_QUERY;
+	} else {
+		*error_query_mode = AMDGPU_RAS_DIRECT_ERROR_QUERY;
+	}
+
+	return true;
+}
 
 /* Register each ip ras block into amdgpu ras */
 int amdgpu_ras_register_ras_block(struct amdgpu_device *adev,
@@ -3098,3 +5187,562 @@ int amdgpu_ras_register_ras_block(struct amdgpu_device *adev,
 
 	return 0;
 }
+
+void amdgpu_ras_get_error_type_name(uint32_t err_type, char *err_type_name)
+{
+	if (!err_type_name)
+		return;
+
+	switch (err_type) {
+	case AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE:
+		sprintf(err_type_name, "correctable");
+		break;
+	case AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE:
+		sprintf(err_type_name, "uncorrectable");
+		break;
+	default:
+		sprintf(err_type_name, "unknown");
+		break;
+	}
+}
+
+bool amdgpu_ras_inst_get_memory_id_field(struct amdgpu_device *adev,
+					 const struct amdgpu_ras_err_status_reg_entry *reg_entry,
+					 uint32_t instance,
+					 uint32_t *memory_id)
+{
+	uint32_t err_status_lo_data, err_status_lo_offset;
+
+	if (!reg_entry)
+		return false;
+
+	err_status_lo_offset =
+		AMDGPU_RAS_REG_ENTRY_OFFSET(reg_entry->hwip, instance,
+					    reg_entry->seg_lo, reg_entry->reg_lo);
+	err_status_lo_data = RREG32(err_status_lo_offset);
+
+	if ((reg_entry->flags & AMDGPU_RAS_ERR_STATUS_VALID) &&
+	    !REG_GET_FIELD(err_status_lo_data, ERR_STATUS_LO, ERR_STATUS_VALID_FLAG))
+		return false;
+
+	*memory_id = REG_GET_FIELD(err_status_lo_data, ERR_STATUS_LO, MEMORY_ID);
+
+	return true;
+}
+
+bool amdgpu_ras_inst_get_err_cnt_field(struct amdgpu_device *adev,
+				       const struct amdgpu_ras_err_status_reg_entry *reg_entry,
+				       uint32_t instance,
+				       unsigned long *err_cnt)
+{
+	uint32_t err_status_hi_data, err_status_hi_offset;
+
+	if (!reg_entry)
+		return false;
+
+	err_status_hi_offset =
+		AMDGPU_RAS_REG_ENTRY_OFFSET(reg_entry->hwip, instance,
+					    reg_entry->seg_hi, reg_entry->reg_hi);
+	err_status_hi_data = RREG32(err_status_hi_offset);
+
+	if ((reg_entry->flags & AMDGPU_RAS_ERR_INFO_VALID) &&
+	    !REG_GET_FIELD(err_status_hi_data, ERR_STATUS_HI, ERR_INFO_VALID_FLAG))
+		/* keep the check here in case we need to refer to the result later */
+		dev_dbg(adev->dev, "Invalid err_info field\n");
+
+	/* read err count */
+	*err_cnt = REG_GET_FIELD(err_status_hi_data, ERR_STATUS, ERR_CNT);
+
+	return true;
+}
+
+void amdgpu_ras_inst_query_ras_error_count(struct amdgpu_device *adev,
+					   const struct amdgpu_ras_err_status_reg_entry *reg_list,
+					   uint32_t reg_list_size,
+					   const struct amdgpu_ras_memory_id_entry *mem_list,
+					   uint32_t mem_list_size,
+					   uint32_t instance,
+					   uint32_t err_type,
+					   unsigned long *err_count)
+{
+	uint32_t memory_id;
+	unsigned long err_cnt;
+	char err_type_name[16];
+	uint32_t i, j;
+
+	for (i = 0; i < reg_list_size; i++) {
+		/* query memory_id from err_status_lo */
+		if (!amdgpu_ras_inst_get_memory_id_field(adev, &reg_list[i],
+							 instance, &memory_id))
+			continue;
+
+		/* query err_cnt from err_status_hi */
+		if (!amdgpu_ras_inst_get_err_cnt_field(adev, &reg_list[i],
+						       instance, &err_cnt) ||
+		    !err_cnt)
+			continue;
+
+		*err_count += err_cnt;
+
+		/* log the errors */
+		amdgpu_ras_get_error_type_name(err_type, err_type_name);
+		if (!mem_list) {
+			/* memory_list is not supported */
+			dev_info(adev->dev,
+				 "%ld %s hardware errors detected in %s, instance: %d, memory_id: %d\n",
+				 err_cnt, err_type_name,
+				 reg_list[i].block_name,
+				 instance, memory_id);
+		} else {
+			for (j = 0; j < mem_list_size; j++) {
+				if (memory_id == mem_list[j].memory_id) {
+					dev_info(adev->dev,
+						 "%ld %s hardware errors detected in %s, instance: %d, memory block: %s\n",
+						 err_cnt, err_type_name,
+						 reg_list[i].block_name,
+						 instance, mem_list[j].name);
+					break;
+				}
+			}
+		}
+	}
+}
+
+void amdgpu_ras_inst_reset_ras_error_count(struct amdgpu_device *adev,
+					   const struct amdgpu_ras_err_status_reg_entry *reg_list,
+					   uint32_t reg_list_size,
+					   uint32_t instance)
+{
+	uint32_t err_status_lo_offset, err_status_hi_offset;
+	uint32_t i;
+
+	for (i = 0; i < reg_list_size; i++) {
+		err_status_lo_offset =
+			AMDGPU_RAS_REG_ENTRY_OFFSET(reg_list[i].hwip, instance,
+						    reg_list[i].seg_lo, reg_list[i].reg_lo);
+		err_status_hi_offset =
+			AMDGPU_RAS_REG_ENTRY_OFFSET(reg_list[i].hwip, instance,
+						    reg_list[i].seg_hi, reg_list[i].reg_hi);
+		WREG32(err_status_lo_offset, 0);
+		WREG32(err_status_hi_offset, 0);
+	}
+}
+
+int amdgpu_ras_error_data_init(struct ras_err_data *err_data)
+{
+	memset(err_data, 0, sizeof(*err_data));
+
+	INIT_LIST_HEAD(&err_data->err_node_list);
+
+	return 0;
+}
+
+static void amdgpu_ras_error_node_release(struct ras_err_node *err_node)
+{
+	if (!err_node)
+		return;
+
+	list_del(&err_node->node);
+	kvfree(err_node);
+}
+
+void amdgpu_ras_error_data_fini(struct ras_err_data *err_data)
+{
+	struct ras_err_node *err_node, *tmp;
+
+	list_for_each_entry_safe(err_node, tmp, &err_data->err_node_list, node)
+		amdgpu_ras_error_node_release(err_node);
+}
+
+static struct ras_err_node *amdgpu_ras_error_find_node_by_id(struct ras_err_data *err_data,
+							     struct amdgpu_smuio_mcm_config_info *mcm_info)
+{
+	struct ras_err_node *err_node;
+	struct amdgpu_smuio_mcm_config_info *ref_id;
+
+	if (!err_data || !mcm_info)
+		return NULL;
+
+	for_each_ras_error(err_node, err_data) {
+		ref_id = &err_node->err_info.mcm_info;
+
+		if (mcm_info->socket_id == ref_id->socket_id &&
+		    mcm_info->die_id == ref_id->die_id)
+			return err_node;
+	}
+
+	return NULL;
+}
+
+static struct ras_err_node *amdgpu_ras_error_node_new(void)
+{
+	struct ras_err_node *err_node;
+
+	err_node = kvzalloc(sizeof(*err_node), GFP_KERNEL);
+	if (!err_node)
+		return NULL;
+
+	INIT_LIST_HEAD(&err_node->node);
+
+	return err_node;
+}
+
+static int ras_err_info_cmp(void *priv, const struct list_head *a, const struct list_head *b)
+{
+	struct ras_err_node *nodea = container_of(a, struct ras_err_node, node);
+	struct ras_err_node *nodeb = container_of(b, struct ras_err_node, node);
+	struct amdgpu_smuio_mcm_config_info *infoa = &nodea->err_info.mcm_info;
+	struct amdgpu_smuio_mcm_config_info *infob = &nodeb->err_info.mcm_info;
+
+	if (unlikely(infoa->socket_id != infob->socket_id))
+		return infoa->socket_id - infob->socket_id;
+	else
+		return infoa->die_id - infob->die_id;
+
+	return 0;
+}
+
+static struct ras_err_info *amdgpu_ras_error_get_info(struct ras_err_data *err_data,
+				struct amdgpu_smuio_mcm_config_info *mcm_info)
+{
+	struct ras_err_node *err_node;
+
+	err_node = amdgpu_ras_error_find_node_by_id(err_data, mcm_info);
+	if (err_node)
+		return &err_node->err_info;
+
+	err_node = amdgpu_ras_error_node_new();
+	if (!err_node)
+		return NULL;
+
+	memcpy(&err_node->err_info.mcm_info, mcm_info, sizeof(*mcm_info));
+
+	err_data->err_list_count++;
+	list_add_tail(&err_node->node, &err_data->err_node_list);
+	list_sort(NULL, &err_data->err_node_list, ras_err_info_cmp);
+
+	return &err_node->err_info;
+}
+
+int amdgpu_ras_error_statistic_ue_count(struct ras_err_data *err_data,
+					struct amdgpu_smuio_mcm_config_info *mcm_info,
+					u64 count)
+{
+	struct ras_err_info *err_info;
+
+	if (!err_data || !mcm_info)
+		return -EINVAL;
+
+	if (!count)
+		return 0;
+
+	err_info = amdgpu_ras_error_get_info(err_data, mcm_info);
+	if (!err_info)
+		return -EINVAL;
+
+	err_info->ue_count += count;
+	err_data->ue_count += count;
+
+	return 0;
+}
+
+int amdgpu_ras_error_statistic_ce_count(struct ras_err_data *err_data,
+					struct amdgpu_smuio_mcm_config_info *mcm_info,
+					u64 count)
+{
+	struct ras_err_info *err_info;
+
+	if (!err_data || !mcm_info)
+		return -EINVAL;
+
+	if (!count)
+		return 0;
+
+	err_info = amdgpu_ras_error_get_info(err_data, mcm_info);
+	if (!err_info)
+		return -EINVAL;
+
+	err_info->ce_count += count;
+	err_data->ce_count += count;
+
+	return 0;
+}
+
+int amdgpu_ras_error_statistic_de_count(struct ras_err_data *err_data,
+					struct amdgpu_smuio_mcm_config_info *mcm_info,
+					u64 count)
+{
+	struct ras_err_info *err_info;
+
+	if (!err_data || !mcm_info)
+		return -EINVAL;
+
+	if (!count)
+		return 0;
+
+	err_info = amdgpu_ras_error_get_info(err_data, mcm_info);
+	if (!err_info)
+		return -EINVAL;
+
+	err_info->de_count += count;
+	err_data->de_count += count;
+
+	return 0;
+}
+
+#define mmMP0_SMN_C2PMSG_92	0x1609C
+#define mmMP0_SMN_C2PMSG_126	0x160BE
+static void amdgpu_ras_boot_time_error_reporting(struct amdgpu_device *adev,
+						 u32 instance)
+{
+	u32 socket_id, aid_id, hbm_id;
+	u32 fw_status;
+	u32 boot_error;
+	u64 reg_addr;
+
+	/* The pattern for smn addressing in other SOC could be different from
+	 * the one for aqua_vanjaram. We should revisit the code if the pattern
+	 * is changed. In such case, replace the aqua_vanjaram implementation
+	 * with more common helper */
+	reg_addr = (mmMP0_SMN_C2PMSG_92 << 2) +
+		   aqua_vanjaram_encode_ext_smn_addressing(instance);
+	fw_status = amdgpu_device_indirect_rreg_ext(adev, reg_addr);
+
+	reg_addr = (mmMP0_SMN_C2PMSG_126 << 2) +
+		   aqua_vanjaram_encode_ext_smn_addressing(instance);
+	boot_error = amdgpu_device_indirect_rreg_ext(adev, reg_addr);
+
+	socket_id = AMDGPU_RAS_GPU_ERR_SOCKET_ID(boot_error);
+	aid_id = AMDGPU_RAS_GPU_ERR_AID_ID(boot_error);
+	hbm_id = ((1 == AMDGPU_RAS_GPU_ERR_HBM_ID(boot_error)) ? 0 : 1);
+
+	if (AMDGPU_RAS_GPU_ERR_MEM_TRAINING(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, hbm: %d, fw_status: 0x%x, memory training failed\n",
+			 socket_id, aid_id, hbm_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_FW_LOAD(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, fw_status: 0x%x, firmware load failed at boot time\n",
+			 socket_id, aid_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_WAFL_LINK_TRAINING(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, fw_status: 0x%x, wafl link training failed\n",
+			 socket_id, aid_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_XGMI_LINK_TRAINING(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, fw_status: 0x%x, xgmi link training failed\n",
+			 socket_id, aid_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_USR_CP_LINK_TRAINING(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, fw_status: 0x%x, usr cp link training failed\n",
+			 socket_id, aid_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_USR_DP_LINK_TRAINING(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, fw_status: 0x%x, usr dp link training failed\n",
+			 socket_id, aid_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_HBM_MEM_TEST(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, hbm: %d, fw_status: 0x%x, hbm memory test failed\n",
+			 socket_id, aid_id, hbm_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_HBM_BIST_TEST(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, hbm: %d, fw_status: 0x%x, hbm bist test failed\n",
+			 socket_id, aid_id, hbm_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_DATA_ABORT(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, fw_status: 0x%x, data abort exception\n",
+			 socket_id, aid_id, fw_status);
+
+	if (AMDGPU_RAS_GPU_ERR_GENERIC(boot_error))
+		dev_info(adev->dev,
+			 "socket: %d, aid: %d, fw_status: 0x%x, Boot Controller Generic Error\n",
+			 socket_id, aid_id, fw_status);
+}
+
+static bool amdgpu_ras_boot_error_detected(struct amdgpu_device *adev,
+					   u32 instance)
+{
+	u64 reg_addr;
+	u32 reg_data;
+	int retry_loop;
+
+	reg_addr = (mmMP0_SMN_C2PMSG_92 << 2) +
+		   aqua_vanjaram_encode_ext_smn_addressing(instance);
+
+	for (retry_loop = 0; retry_loop < AMDGPU_RAS_BOOT_STATUS_POLLING_LIMIT; retry_loop++) {
+		reg_data = amdgpu_device_indirect_rreg_ext(adev, reg_addr);
+		if ((reg_data & AMDGPU_RAS_BOOT_STATUS_MASK) == AMDGPU_RAS_BOOT_STEADY_STATUS)
+			return false;
+		else
+			msleep(1);
+	}
+
+	return true;
+}
+
+void amdgpu_ras_query_boot_status(struct amdgpu_device *adev, u32 num_instances)
+{
+	u32 i;
+
+	for (i = 0; i < num_instances; i++) {
+		if (amdgpu_ras_boot_error_detected(adev, i))
+			amdgpu_ras_boot_time_error_reporting(adev, i);
+	}
+}
+
+int amdgpu_ras_reserve_page(struct amdgpu_device *adev, uint64_t pfn)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr;
+	uint64_t start = pfn << AMDGPU_GPU_PAGE_SHIFT;
+	int ret = 0;
+
+	if (amdgpu_ras_check_critical_address(adev, start))
+		return 0;
+
+	mutex_lock(&con->page_rsv_lock);
+	ret = amdgpu_vram_mgr_query_page_status(mgr, start);
+	if (ret == -ENOENT)
+		ret = amdgpu_vram_mgr_reserve_range(mgr, start, AMDGPU_GPU_PAGE_SIZE);
+	mutex_unlock(&con->page_rsv_lock);
+
+	return ret;
+}
+
+void amdgpu_ras_event_log_print(struct amdgpu_device *adev, u64 event_id,
+				const char *fmt, ...)
+{
+	struct va_format vaf;
+	va_list args;
+
+	va_start(args, fmt);
+	vaf.fmt = fmt;
+	vaf.va = &args;
+
+	if (RAS_EVENT_ID_IS_VALID(event_id))
+		dev_printk(KERN_INFO, adev->dev, "{%llu}%pV", event_id, &vaf);
+	else
+		dev_printk(KERN_INFO, adev->dev, "%pV", &vaf);
+
+	va_end(args);
+}
+
+bool amdgpu_ras_is_rma(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+
+	if (amdgpu_uniras_enabled(adev))
+		return amdgpu_ras_mgr_is_rma(adev);
+
+	if (!con)
+		return false;
+
+	return con->is_rma;
+}
+
+int amdgpu_ras_add_critical_region(struct amdgpu_device *adev,
+			struct amdgpu_bo *bo)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct amdgpu_vram_mgr_resource *vres;
+	struct ras_critical_region *region;
+	struct drm_buddy_block *block;
+	int ret = 0;
+
+	if (!bo || !bo->tbo.resource)
+		return -EINVAL;
+
+	vres = to_amdgpu_vram_mgr_resource(bo->tbo.resource);
+
+	mutex_lock(&con->critical_region_lock);
+
+	/* Check if the bo had been recorded */
+	list_for_each_entry(region, &con->critical_region_head, node)
+		if (region->bo == bo)
+			goto out;
+
+	/* Record new critical amdgpu bo */
+	list_for_each_entry(block, &vres->blocks, link) {
+		region = kzalloc(sizeof(*region), GFP_KERNEL);
+		if (!region) {
+			ret = -ENOMEM;
+			goto out;
+		}
+		region->bo = bo;
+		region->start = amdgpu_vram_mgr_block_start(block);
+		region->size = amdgpu_vram_mgr_block_size(block);
+		list_add_tail(&region->node, &con->critical_region_head);
+	}
+
+out:
+	mutex_unlock(&con->critical_region_lock);
+
+	return ret;
+}
+
+static void amdgpu_ras_critical_region_init(struct amdgpu_device *adev)
+{
+	amdgpu_ras_add_critical_region(adev, adev->mman.fw_reserved_memory);
+}
+
+static void amdgpu_ras_critical_region_fini(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct ras_critical_region *region, *tmp;
+
+	mutex_lock(&con->critical_region_lock);
+	list_for_each_entry_safe(region, tmp, &con->critical_region_head, node) {
+		list_del(&region->node);
+		kfree(region);
+	}
+	mutex_unlock(&con->critical_region_lock);
+}
+
+bool amdgpu_ras_check_critical_address(struct amdgpu_device *adev, uint64_t addr)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct ras_critical_region *region;
+	bool ret = false;
+
+	mutex_lock(&con->critical_region_lock);
+	list_for_each_entry(region, &con->critical_region_head, node) {
+		if ((region->start <= addr) &&
+		    (addr < (region->start + region->size))) {
+			ret = true;
+			break;
+		}
+	}
+	mutex_unlock(&con->critical_region_lock);
+
+	return ret;
+}
+
+void amdgpu_ras_pre_reset(struct amdgpu_device *adev,
+					  struct list_head *device_list)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+
+	list_for_each_entry(tmp_adev, device_list, reset_list) {
+		if (amdgpu_uniras_enabled(tmp_adev))
+			amdgpu_ras_mgr_pre_reset(tmp_adev);
+	}
+}
+
+void amdgpu_ras_post_reset(struct amdgpu_device *adev,
+					  struct list_head *device_list)
+{
+	struct amdgpu_device *tmp_adev = NULL;
+
+	list_for_each_entry(tmp_adev, device_list, reset_list) {
+		if (amdgpu_uniras_enabled(tmp_adev))
+			amdgpu_ras_mgr_post_reset(tmp_adev);
+	}
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
index ef38f4c93df0..ff44190d7d98 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h
@@ -26,12 +26,57 @@
 
 #include <linux/debugfs.h>
 #include <linux/list.h>
+#include <linux/kfifo.h>
+#include <linux/radix-tree.h>
 #include "ta_ras_if.h"
 #include "amdgpu_ras_eeprom.h"
+#include "amdgpu_smuio.h"
+#include "amdgpu_aca.h"
 
 struct amdgpu_iv_entry;
 
+#define AMDGPU_RAS_GPU_ERR_MEM_TRAINING(x)		AMDGPU_GET_REG_FIELD(x, 0, 0)
+#define AMDGPU_RAS_GPU_ERR_FW_LOAD(x)			AMDGPU_GET_REG_FIELD(x, 1, 1)
+#define AMDGPU_RAS_GPU_ERR_WAFL_LINK_TRAINING(x)	AMDGPU_GET_REG_FIELD(x, 2, 2)
+#define AMDGPU_RAS_GPU_ERR_XGMI_LINK_TRAINING(x)	AMDGPU_GET_REG_FIELD(x, 3, 3)
+#define AMDGPU_RAS_GPU_ERR_USR_CP_LINK_TRAINING(x)	AMDGPU_GET_REG_FIELD(x, 4, 4)
+#define AMDGPU_RAS_GPU_ERR_USR_DP_LINK_TRAINING(x)	AMDGPU_GET_REG_FIELD(x, 5, 5)
+#define AMDGPU_RAS_GPU_ERR_HBM_MEM_TEST(x)		AMDGPU_GET_REG_FIELD(x, 6, 6)
+#define AMDGPU_RAS_GPU_ERR_HBM_BIST_TEST(x)		AMDGPU_GET_REG_FIELD(x, 7, 7)
+#define AMDGPU_RAS_GPU_ERR_SOCKET_ID(x)			AMDGPU_GET_REG_FIELD(x, 10, 8)
+#define AMDGPU_RAS_GPU_ERR_AID_ID(x)			AMDGPU_GET_REG_FIELD(x, 12, 11)
+#define AMDGPU_RAS_GPU_ERR_HBM_ID(x)			AMDGPU_GET_REG_FIELD(x, 14, 13)
+#define AMDGPU_RAS_GPU_ERR_DATA_ABORT(x)		AMDGPU_GET_REG_FIELD(x, 29, 29)
+#define AMDGPU_RAS_GPU_ERR_GENERIC(x)			AMDGPU_GET_REG_FIELD(x, 30, 30)
+
+#define AMDGPU_RAS_BOOT_STATUS_POLLING_LIMIT	100
+#define AMDGPU_RAS_BOOT_STEADY_STATUS		0xBA
+#define AMDGPU_RAS_BOOT_STATUS_MASK		0xFF
+
 #define AMDGPU_RAS_FLAG_INIT_BY_VBIOS		(0x1 << 0)
+/* position of instance value in sub_block_index of
+ * ta_ras_trigger_error_input, the sub block uses lower 12 bits
+ */
+#define AMDGPU_RAS_INST_MASK 0xfffff000
+#define AMDGPU_RAS_INST_SHIFT 0xc
+
+#define AMDGPU_RAS_FEATURES_SOCKETID_SHIFT 29
+#define AMDGPU_RAS_FEATURES_SOCKETID_MASK 0xe0000000
+
+/* Reserve 8 physical dram row for possible retirement.
+ * In worst cases, it will lose 8 * 2MB memory in vram domain */
+#define AMDGPU_RAS_RESERVED_VRAM_SIZE_DEFAULT	(16ULL << 20)
+/* The high three bits indicates socketid */
+#define AMDGPU_RAS_GET_FEATURES(val)  ((val) & ~AMDGPU_RAS_FEATURES_SOCKETID_MASK)
+
+#define RAS_EVENT_INVALID_ID		(BIT_ULL(63))
+#define RAS_EVENT_ID_IS_VALID(x)	(!((x) & BIT_ULL(63)))
+
+#define RAS_EVENT_LOG(adev, id, fmt, ...)	\
+	amdgpu_ras_event_log_print((adev), (id), (fmt), ##__VA_ARGS__)
+
+#define amdgpu_ras_mark_ras_event(adev, type)	\
+	(amdgpu_ras_mark_ras_event_caller((adev), (type), __builtin_return_address(0)))
 
 enum amdgpu_ras_block {
 	AMDGPU_RAS_BLOCK__UMC = 0,
@@ -51,8 +96,12 @@ enum amdgpu_ras_block {
 	AMDGPU_RAS_BLOCK__MCA,
 	AMDGPU_RAS_BLOCK__VCN,
 	AMDGPU_RAS_BLOCK__JPEG,
+	AMDGPU_RAS_BLOCK__IH,
+	AMDGPU_RAS_BLOCK__MPIO,
+	AMDGPU_RAS_BLOCK__MMSCH,
 
-	AMDGPU_RAS_BLOCK__LAST
+	AMDGPU_RAS_BLOCK__LAST,
+	AMDGPU_RAS_BLOCK__ANY = -1
 };
 
 enum amdgpu_ras_mca_block {
@@ -314,6 +363,53 @@ enum amdgpu_ras_ret {
 	AMDGPU_RAS_PT,
 };
 
+enum amdgpu_ras_error_query_mode {
+	AMDGPU_RAS_INVALID_ERROR_QUERY		= 0,
+	AMDGPU_RAS_DIRECT_ERROR_QUERY		= 1,
+	AMDGPU_RAS_FIRMWARE_ERROR_QUERY		= 2,
+	AMDGPU_RAS_VIRT_ERROR_COUNT_QUERY	= 3,
+};
+
+/* ras error status reisger fields */
+#define ERR_STATUS_LO__ERR_STATUS_VALID_FLAG__SHIFT	0x0
+#define ERR_STATUS_LO__ERR_STATUS_VALID_FLAG_MASK	0x00000001L
+#define ERR_STATUS_LO__MEMORY_ID__SHIFT			0x18
+#define ERR_STATUS_LO__MEMORY_ID_MASK			0xFF000000L
+#define ERR_STATUS_HI__ERR_INFO_VALID_FLAG__SHIFT	0x2
+#define ERR_STATUS_HI__ERR_INFO_VALID_FLAG_MASK		0x00000004L
+#define ERR_STATUS__ERR_CNT__SHIFT			0x17
+#define ERR_STATUS__ERR_CNT_MASK			0x03800000L
+
+#define AMDGPU_RAS_REG_ENTRY(ip, inst, reg_lo, reg_hi) \
+	ip##_HWIP, inst, reg_lo##_BASE_IDX, reg_lo, reg_hi##_BASE_IDX, reg_hi
+
+#define AMDGPU_RAS_REG_ENTRY_OFFSET(hwip, ip_inst, segment, reg) \
+	(adev->reg_offset[hwip][ip_inst][segment] + (reg))
+
+#define AMDGPU_RAS_ERR_INFO_VALID	(1 << 0)
+#define AMDGPU_RAS_ERR_STATUS_VALID	(1 << 1)
+#define AMDGPU_RAS_ERR_ADDRESS_VALID	(1 << 2)
+
+#define AMDGPU_RAS_GPU_RESET_MODE2_RESET  (0x1 << 0)
+#define AMDGPU_RAS_GPU_RESET_MODE1_RESET  (0x1 << 1)
+
+struct amdgpu_ras_err_status_reg_entry {
+	uint32_t hwip;
+	uint32_t ip_inst;
+	uint32_t seg_lo;
+	uint32_t reg_lo;
+	uint32_t seg_hi;
+	uint32_t reg_hi;
+	uint32_t reg_inst;
+	uint32_t flags;
+	const char *block_name;
+};
+
+struct amdgpu_ras_memory_id_entry {
+	uint32_t memory_id;
+	const char *name;
+};
+
 struct ras_common_if {
 	enum amdgpu_ras_block block;
 	enum amdgpu_ras_error_type type;
@@ -340,13 +436,111 @@ struct umc_ecc_info {
 	int record_ce_addr_supported;
 };
 
+enum ras_event_type {
+	RAS_EVENT_TYPE_INVALID = 0,
+	RAS_EVENT_TYPE_FATAL,
+	RAS_EVENT_TYPE_POISON_CREATION,
+	RAS_EVENT_TYPE_POISON_CONSUMPTION,
+	RAS_EVENT_TYPE_COUNT,
+};
+
+struct ras_event_state {
+	u64 last_seqno;
+	atomic64_t count;
+};
+
+struct ras_event_manager {
+	atomic64_t seqno;
+	struct ras_event_state event_state[RAS_EVENT_TYPE_COUNT];
+};
+
+struct ras_event_id {
+	enum ras_event_type type;
+	u64 event_id;
+};
+
+struct ras_query_context {
+	struct ras_event_id evid;
+};
+
+typedef int (*pasid_notify)(struct amdgpu_device *adev,
+		uint16_t pasid, void *data);
+
+struct ras_poison_msg {
+	enum amdgpu_ras_block block;
+	uint16_t pasid;
+	uint32_t reset;
+	pasid_notify pasid_fn;
+	void *data;
+};
+
+struct ras_err_pages {
+	uint32_t count;
+	uint64_t *pfn;
+};
+
+struct ras_ecc_err {
+	uint64_t status;
+	uint64_t ipid;
+	uint64_t addr;
+	uint64_t pa_pfn;
+	/* save global channel index across all UMC instances */
+	uint32_t channel_idx;
+	struct ras_err_pages err_pages;
+};
+
+struct ras_ecc_log_info {
+	struct mutex lock;
+	struct radix_tree_root de_page_tree;
+	uint64_t de_queried_count;
+	uint64_t consumption_q_count;
+};
+
+struct ras_critical_region {
+	struct list_head node;
+	struct amdgpu_bo *bo;
+	uint64_t start;
+	uint64_t size;
+};
+
+struct ras_eeprom_table_version {
+	uint32_t minor    : 16;
+	uint32_t major    : 16;
+};
+
+struct ras_eeprom_smu_funcs {
+	int (*get_ras_table_version)(struct amdgpu_device *adev,
+							uint32_t *table_version);
+	int (*get_badpage_count)(struct amdgpu_device *adev, uint32_t *count, uint32_t timeout);
+	int (*get_badpage_mca_addr)(struct amdgpu_device *adev, uint16_t index, uint64_t *mca_addr);
+	int (*set_timestamp)(struct amdgpu_device *adev, uint64_t timestamp);
+	int (*get_timestamp)(struct amdgpu_device *adev,
+							uint16_t index, uint64_t *timestamp);
+	int (*get_badpage_ipid)(struct amdgpu_device *adev, uint16_t index, uint64_t *ipid);
+	int (*erase_ras_table)(struct amdgpu_device *adev, uint32_t *result);
+};
+
+enum ras_smu_feature_flags {
+	RAS_SMU_FEATURE_BIT__RAS_EEPROM = BIT_ULL(0),
+};
+
+struct ras_smu_drv {
+	const struct ras_eeprom_smu_funcs *smu_eeprom_funcs;
+	void (*ras_smu_feature_flags)(struct amdgpu_device *adev, uint64_t *flags);
+};
+
 struct amdgpu_ras {
+	void *ras_mgr;
 	/* ras infrastructure */
 	/* for ras itself. */
 	uint32_t features;
+	uint32_t schema;
 	struct list_head head;
 	/* sysfs */
 	struct device_attribute features_attr;
+	struct device_attribute version_attr;
+	struct device_attribute schema_attr;
+	struct device_attribute event_state_attr;
 	struct bin_attribute badpages_attr;
 	struct dentry *de_ras_eeprom_table;
 	/* block array */
@@ -355,6 +549,7 @@ struct amdgpu_ras {
 	/* gpu recovery */
 	struct work_struct recovery_work;
 	atomic_t in_recovery;
+	atomic_t rma_in_recovery;
 	struct amdgpu_device *adev;
 	/* error handler data */
 	struct ras_err_handler_data *eh_data;
@@ -385,25 +580,86 @@ struct amdgpu_ras {
 
 	/* Indicates smu whether need update bad channel info */
 	bool update_channel_flag;
+	/* Record status of smu mca debug mode */
+	bool is_aca_debug_mode;
+	bool is_rma;
+
+	/* Record special requirements of gpu reset caller */
+	uint32_t  gpu_reset_flags;
+
+	struct task_struct *page_retirement_thread;
+	wait_queue_head_t page_retirement_wq;
+	struct mutex page_retirement_lock;
+	atomic_t page_retirement_req_cnt;
+	atomic_t poison_creation_count;
+	atomic_t poison_consumption_count;
+	struct mutex page_rsv_lock;
+	DECLARE_KFIFO(poison_fifo, struct ras_poison_msg, 128);
+	struct ras_ecc_log_info  umc_ecc_log;
+	struct delayed_work page_retirement_dwork;
+
+	/* ras errors detected */
+	unsigned long ras_err_state;
+
+	/* RAS event manager */
+	struct ras_event_manager __event_mgr;
+	struct ras_event_manager *event_mgr;
+
+	uint64_t reserved_pages_in_bytes;
+
+	pid_t init_task_pid;
+	char init_task_comm[TASK_COMM_LEN];
+
+	int bad_page_num;
+
+	struct list_head critical_region_head;
+	struct mutex critical_region_lock;
+
+	/* Protect poison injection */
+	struct mutex poison_lock;
+
+	/* Disable/Enable uniras switch */
+	bool uniras_enabled;
+	const struct ras_smu_drv *ras_smu_drv;
 };
 
 struct ras_fs_data {
-	char sysfs_name[32];
+	char sysfs_name[48];
 	char debugfs_name[32];
 };
 
+struct ras_err_info {
+	struct amdgpu_smuio_mcm_config_info mcm_info;
+	u64 ce_count;
+	u64 ue_count;
+	u64 de_count;
+};
+
+struct ras_err_node {
+	struct list_head node;
+	struct ras_err_info err_info;
+};
+
 struct ras_err_data {
 	unsigned long ue_count;
 	unsigned long ce_count;
+	unsigned long de_count;
 	unsigned long err_addr_cnt;
 	struct eeprom_table_record *err_addr;
+	unsigned long err_addr_len;
+	u32 err_list_count;
+	struct list_head err_node_list;
 };
 
+#define for_each_ras_error(err_node, err_data) \
+	list_for_each_entry(err_node, &(err_data)->err_node_list, node)
+
 struct ras_err_handler_data {
 	/* point to bad page records array */
 	struct eeprom_table_record *bps;
 	/* the count of entries */
 	int count;
+	int count_saved;
 	/* the space can place new entries */
 	int space_left;
 };
@@ -446,6 +702,8 @@ struct ras_manager {
 	struct ras_ih_data ih_data;
 
 	struct ras_err_data err_data;
+
+	struct aca_handle aca_handle;
 };
 
 struct ras_badpage {
@@ -465,12 +723,14 @@ struct ras_query_if {
 	struct ras_common_if head;
 	unsigned long ue_count;
 	unsigned long ce_count;
+	unsigned long de_count;
 };
 
 struct ras_inject_if {
 	struct ras_common_if head;
 	uint64_t address;
 	uint64_t value;
+	uint32_t instance_mask;
 };
 
 struct ras_cure_if {
@@ -508,7 +768,8 @@ struct amdgpu_ras_block_object {
 };
 
 struct amdgpu_ras_block_hw_ops {
-	int  (*ras_error_inject)(struct amdgpu_device *adev, void *inject_if);
+	int  (*ras_error_inject)(struct amdgpu_device *adev,
+			void *inject_if, uint32_t instance_mask);
 	void (*query_ras_error_count)(struct amdgpu_device *adev, void *ras_error_status);
 	void (*query_ras_error_status)(struct amdgpu_device *adev);
 	void (*query_ras_error_address)(struct amdgpu_device *adev, void *ras_error_status);
@@ -532,8 +793,8 @@ struct amdgpu_ras_block_hw_ops {
  * 8: feature disable
  */
 
-
-int amdgpu_ras_recovery_init(struct amdgpu_device *adev);
+int amdgpu_ras_init_badpage_info(struct amdgpu_device *adev);
+int amdgpu_ras_recovery_init(struct amdgpu_device *adev, bool init_bp_info);
 
 void amdgpu_ras_resume(struct amdgpu_device *adev);
 void amdgpu_ras_suspend(struct amdgpu_device *adev);
@@ -545,7 +806,7 @@ int amdgpu_ras_query_error_count(struct amdgpu_device *adev,
 
 /* error handling functions */
 int amdgpu_ras_add_bad_pages(struct amdgpu_device *adev,
-		struct eeprom_table_record *bps, int pages);
+		struct eeprom_table_record *bps, int pages, bool from_rom);
 
 int amdgpu_ras_save_bad_pages(struct amdgpu_device *adev,
 		unsigned long *new_cnt);
@@ -583,6 +844,16 @@ amdgpu_ras_block_to_ta(enum amdgpu_ras_block block) {
 		return TA_RAS_BLOCK__FUSE;
 	case AMDGPU_RAS_BLOCK__MCA:
 		return TA_RAS_BLOCK__MCA;
+	case AMDGPU_RAS_BLOCK__VCN:
+		return TA_RAS_BLOCK__VCN;
+	case AMDGPU_RAS_BLOCK__JPEG:
+		return TA_RAS_BLOCK__JPEG;
+	case AMDGPU_RAS_BLOCK__IH:
+		return TA_RAS_BLOCK__IH;
+	case AMDGPU_RAS_BLOCK__MPIO:
+		return TA_RAS_BLOCK__MPIO;
+	case AMDGPU_RAS_BLOCK__MMSCH:
+		return TA_RAS_BLOCK__MMSCH;
 	default:
 		WARN_ONCE(1, "RAS ERROR: unexpected block id %d\n", block);
 		return TA_RAS_BLOCK__UMC;
@@ -637,6 +908,8 @@ void amdgpu_ras_debugfs_create_all(struct amdgpu_device *adev);
 int amdgpu_ras_query_error_status(struct amdgpu_device *adev,
 		struct ras_query_if *info);
 
+int amdgpu_ras_reset_error_count(struct amdgpu_device *adev,
+		enum amdgpu_ras_block block);
 int amdgpu_ras_reset_error_status(struct amdgpu_device *adev,
 		enum amdgpu_ras_block block);
 
@@ -667,7 +940,7 @@ static inline void amdgpu_ras_intr_cleared(void)
 	atomic_set(&amdgpu_ras_in_intr, 0);
 }
 
-void amdgpu_ras_global_ras_isr(struct amdgpu_device *adev);
+int amdgpu_ras_global_ras_isr(struct amdgpu_device *adev);
 
 void amdgpu_ras_set_error_query_ready(struct amdgpu_device *adev, bool ready);
 
@@ -689,7 +962,86 @@ struct amdgpu_ras* amdgpu_ras_get_context(struct amdgpu_device *adev);
 
 int amdgpu_ras_set_context(struct amdgpu_device *adev, struct amdgpu_ras *ras_con);
 
+int amdgpu_ras_set_mca_debug_mode(struct amdgpu_device *adev, bool enable);
+int amdgpu_ras_set_aca_debug_mode(struct amdgpu_device *adev, bool enable);
+bool amdgpu_ras_get_aca_debug_mode(struct amdgpu_device *adev);
+bool amdgpu_ras_get_error_query_mode(struct amdgpu_device *adev,
+				     unsigned int *mode);
+
 int amdgpu_ras_register_ras_block(struct amdgpu_device *adev,
 				struct amdgpu_ras_block_object *ras_block_obj);
 void amdgpu_ras_interrupt_fatal_error_handler(struct amdgpu_device *adev);
+void amdgpu_ras_get_error_type_name(uint32_t err_type, char *err_type_name);
+bool amdgpu_ras_inst_get_memory_id_field(struct amdgpu_device *adev,
+					 const struct amdgpu_ras_err_status_reg_entry *reg_entry,
+					 uint32_t instance,
+					 uint32_t *memory_id);
+bool amdgpu_ras_inst_get_err_cnt_field(struct amdgpu_device *adev,
+				       const struct amdgpu_ras_err_status_reg_entry *reg_entry,
+				       uint32_t instance,
+				       unsigned long *err_cnt);
+void amdgpu_ras_inst_query_ras_error_count(struct amdgpu_device *adev,
+					   const struct amdgpu_ras_err_status_reg_entry *reg_list,
+					   uint32_t reg_list_size,
+					   const struct amdgpu_ras_memory_id_entry *mem_list,
+					   uint32_t mem_list_size,
+					   uint32_t instance,
+					   uint32_t err_type,
+					   unsigned long *err_count);
+void amdgpu_ras_inst_reset_ras_error_count(struct amdgpu_device *adev,
+					   const struct amdgpu_ras_err_status_reg_entry *reg_list,
+					   uint32_t reg_list_size,
+					   uint32_t instance);
+
+int amdgpu_ras_error_data_init(struct ras_err_data *err_data);
+void amdgpu_ras_error_data_fini(struct ras_err_data *err_data);
+int amdgpu_ras_error_statistic_ce_count(struct ras_err_data *err_data,
+					struct amdgpu_smuio_mcm_config_info *mcm_info,
+					u64 count);
+int amdgpu_ras_error_statistic_ue_count(struct ras_err_data *err_data,
+					struct amdgpu_smuio_mcm_config_info *mcm_info,
+					u64 count);
+int amdgpu_ras_error_statistic_de_count(struct ras_err_data *err_data,
+					struct amdgpu_smuio_mcm_config_info *mcm_info,
+					u64 count);
+void amdgpu_ras_query_boot_status(struct amdgpu_device *adev, u32 num_instances);
+int amdgpu_ras_bind_aca(struct amdgpu_device *adev, enum amdgpu_ras_block blk,
+			       const struct aca_info *aca_info, void *data);
+int amdgpu_ras_unbind_aca(struct amdgpu_device *adev, enum amdgpu_ras_block blk);
+
+ssize_t amdgpu_ras_aca_sysfs_read(struct device *dev, struct device_attribute *attr,
+				  struct aca_handle *handle, char *buf, void *data);
+
+void amdgpu_ras_set_fed(struct amdgpu_device *adev, bool status);
+bool amdgpu_ras_get_fed_status(struct amdgpu_device *adev);
+void amdgpu_ras_set_err_poison(struct amdgpu_device *adev,
+			       enum amdgpu_ras_block block);
+void amdgpu_ras_clear_err_state(struct amdgpu_device *adev);
+bool amdgpu_ras_is_err_state(struct amdgpu_device *adev, int block);
+
+u64 amdgpu_ras_acquire_event_id(struct amdgpu_device *adev, enum ras_event_type type);
+int amdgpu_ras_mark_ras_event_caller(struct amdgpu_device *adev, enum ras_event_type type,
+				     const void *caller);
+
+int amdgpu_ras_reserve_page(struct amdgpu_device *adev, uint64_t pfn);
+
+int amdgpu_ras_add_critical_region(struct amdgpu_device *adev, struct amdgpu_bo *bo);
+bool amdgpu_ras_check_critical_address(struct amdgpu_device *adev, uint64_t addr);
+
+int amdgpu_ras_put_poison_req(struct amdgpu_device *adev,
+		enum amdgpu_ras_block block, uint16_t pasid,
+		pasid_notify pasid_fn, void *data, uint32_t reset);
+
+bool amdgpu_ras_in_recovery(struct amdgpu_device *adev);
+
+__printf(3, 4)
+void amdgpu_ras_event_log_print(struct amdgpu_device *adev, u64 event_id,
+				const char *fmt, ...);
+
+bool amdgpu_ras_is_rma(struct amdgpu_device *adev);
+
+void amdgpu_ras_pre_reset(struct amdgpu_device *adev,
+					  struct list_head *device_list);
+void amdgpu_ras_post_reset(struct amdgpu_device *adev,
+					  struct list_head *device_list);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c
index 3106fa8a15ef..64dd7a81bff5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c
@@ -32,6 +32,7 @@
 #include <linux/uaccess.h>
 
 #include "amdgpu_reset.h"
+#include "amdgpu_ras_mgr.h"
 
 /* These are memory addresses as would be seen by one or more EEPROM
  * chips strung on the I2C bus, usually by manipulating pins 1-3 of a
@@ -58,7 +59,7 @@
 #define EEPROM_I2C_MADDR_4      0x40000
 
 /*
- * The 2 macros bellow represent the actual size in bytes that
+ * The 2 macros below represent the actual size in bytes that
  * those entities occupy in the EEPROM memory.
  * RAS_TABLE_RECORD_SIZE is different than sizeof(eeprom_table_record) which
  * uses uint64 to store 6b fields such as retired_page.
@@ -68,11 +69,24 @@
 
 /* Table hdr is 'AMDR' */
 #define RAS_TABLE_HDR_VAL       0x414d4452
-#define RAS_TABLE_VER           0x00010000
 
 /* Bad GPU tag ‘BADG’ */
 #define RAS_TABLE_HDR_BAD       0x42414447
 
+/*
+ * EEPROM Table structure v1
+ * ---------------------------------
+ * |                               |
+ * |     EEPROM TABLE HEADER       |
+ * |      ( size 20 Bytes )        |
+ * |                               |
+ * ---------------------------------
+ * |                               |
+ * |    BAD PAGE RECORD AREA       |
+ * |                               |
+ * ---------------------------------
+ */
+
 /* Assume 2-Mbit size EEPROM and take up the whole space. */
 #define RAS_TBL_SIZE_BYTES      (256 * 1024)
 #define RAS_TABLE_START         0
@@ -81,6 +95,37 @@
 #define RAS_MAX_RECORD_COUNT    ((RAS_TBL_SIZE_BYTES - RAS_TABLE_HEADER_SIZE) \
 				 / RAS_TABLE_RECORD_SIZE)
 
+/*
+ * EEPROM Table structrue v2.1
+ * ---------------------------------
+ * |                               |
+ * |     EEPROM TABLE HEADER       |
+ * |      ( size 20 Bytes )        |
+ * |                               |
+ * ---------------------------------
+ * |                               |
+ * |     EEPROM TABLE RAS INFO     |
+ * | (available info size 4 Bytes) |
+ * |  ( reserved size 252 Bytes )  |
+ * |                               |
+ * ---------------------------------
+ * |                               |
+ * |     BAD PAGE RECORD AREA      |
+ * |                               |
+ * ---------------------------------
+ */
+
+/* EEPROM Table V2_1 */
+#define RAS_TABLE_V2_1_INFO_SIZE       256
+#define RAS_TABLE_V2_1_INFO_START      RAS_TABLE_HEADER_SIZE
+#define RAS_RECORD_START_V2_1          (RAS_HDR_START + RAS_TABLE_HEADER_SIZE + \
+					RAS_TABLE_V2_1_INFO_SIZE)
+#define RAS_MAX_RECORD_COUNT_V2_1      ((RAS_TBL_SIZE_BYTES - RAS_TABLE_HEADER_SIZE - \
+					RAS_TABLE_V2_1_INFO_SIZE) \
+					/ RAS_TABLE_RECORD_SIZE)
+
+#define RAS_SMU_MESSAGE_TIMEOUT_MS 1000 /* 1s */
+
 /* Given a zero-based index of an EEPROM RAS record, yields the EEPROM
  * offset off of RAS_TABLE_START.  That is, this is something you can
  * add to control->i2c_address, and then tell I2C layer to read
@@ -103,52 +148,25 @@
 #define RAS_NUM_RECS(_tbl_hdr)  (((_tbl_hdr)->tbl_size - \
 				  RAS_TABLE_HEADER_SIZE) / RAS_TABLE_RECORD_SIZE)
 
-#define to_amdgpu_device(x) (container_of(x, struct amdgpu_ras, eeprom_control))->adev
-
-static bool __is_ras_eeprom_supported(struct amdgpu_device *adev)
-{
-	if (adev->asic_type == CHIP_IP_DISCOVERY) {
-		switch (adev->ip_versions[MP1_HWIP][0]) {
-		case IP_VERSION(13, 0, 0):
-		case IP_VERSION(13, 0, 10):
-			return true;
-		default:
-			return false;
-		}
-	}
-
-	return  adev->asic_type == CHIP_VEGA20 ||
-		adev->asic_type == CHIP_ARCTURUS ||
-		adev->asic_type == CHIP_SIENNA_CICHLID ||
-		adev->asic_type == CHIP_ALDEBARAN;
-}
-
-static bool __get_eeprom_i2c_addr_arct(struct amdgpu_device *adev,
-				       struct amdgpu_ras_eeprom_control *control)
-{
-	struct atom_context *atom_ctx = adev->mode_info.atom_context;
-
-	if (!control || !atom_ctx)
-		return false;
-
-	if (strnstr(atom_ctx->vbios_version,
-	            "D342",
-		    sizeof(atom_ctx->vbios_version)))
-		control->i2c_address = EEPROM_I2C_MADDR_0;
-	else
-		control->i2c_address = EEPROM_I2C_MADDR_4;
+#define RAS_NUM_RECS_V2_1(_tbl_hdr)  (((_tbl_hdr)->tbl_size - \
+				       RAS_TABLE_HEADER_SIZE - \
+				       RAS_TABLE_V2_1_INFO_SIZE) / RAS_TABLE_RECORD_SIZE)
 
-	return true;
-}
+#define to_amdgpu_device(x) ((container_of(x, struct amdgpu_ras, eeprom_control))->adev)
 
-static bool __get_eeprom_i2c_addr_ip_discovery(struct amdgpu_device *adev,
-				       struct amdgpu_ras_eeprom_control *control)
+static bool __is_ras_eeprom_supported(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[MP1_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
+	case IP_VERSION(11, 0, 2): /* VEGA20 and ARCTURUS */
+	case IP_VERSION(11, 0, 7): /* Sienna cichlid */
 	case IP_VERSION(13, 0, 0):
+	case IP_VERSION(13, 0, 2): /* Aldebaran */
 	case IP_VERSION(13, 0, 10):
-		control->i2c_address = EEPROM_I2C_MADDR_4;
 		return true;
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
+		return (adev->gmc.is_app_apu) ? false : true;
 	default:
 		return false;
 	}
@@ -163,7 +181,7 @@ static bool __get_eeprom_i2c_addr(struct amdgpu_device *adev,
 	if (!control)
 		return false;
 
-	if (amdgpu_atomfirmware_ras_rom_addr(adev, &i2c_addr)) {
+	if (adev->bios && amdgpu_atomfirmware_ras_rom_addr(adev, &i2c_addr)) {
 		/* The address given by VBIOS is an 8-bit, wire-format
 		 * address, i.e. the most significant byte.
 		 *
@@ -178,29 +196,41 @@ static bool __get_eeprom_i2c_addr(struct amdgpu_device *adev,
 		return true;
 	}
 
-	switch (adev->asic_type) {
-	case CHIP_VEGA20:
-		control->i2c_address = EEPROM_I2C_MADDR_0;
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
+	case IP_VERSION(11, 0, 2):
+		/* VEGA20 and ARCTURUS */
+		if (adev->asic_type == CHIP_VEGA20)
+			control->i2c_address = EEPROM_I2C_MADDR_0;
+		else if (strnstr(atom_ctx->vbios_pn,
+				 "D342",
+				 sizeof(atom_ctx->vbios_pn)))
+			control->i2c_address = EEPROM_I2C_MADDR_0;
+		else
+			control->i2c_address = EEPROM_I2C_MADDR_4;
 		return true;
-
-	case CHIP_ARCTURUS:
-		return __get_eeprom_i2c_addr_arct(adev, control);
-
-	case CHIP_SIENNA_CICHLID:
+	case IP_VERSION(11, 0, 7):
 		control->i2c_address = EEPROM_I2C_MADDR_0;
 		return true;
-
-	case CHIP_ALDEBARAN:
-		if (strnstr(atom_ctx->vbios_version, "D673",
-			    sizeof(atom_ctx->vbios_version)))
+	case IP_VERSION(13, 0, 2):
+		if (strnstr(atom_ctx->vbios_pn, "D673",
+			    sizeof(atom_ctx->vbios_pn)))
 			control->i2c_address = EEPROM_I2C_MADDR_4;
 		else
 			control->i2c_address = EEPROM_I2C_MADDR_0;
 		return true;
-
-	case CHIP_IP_DISCOVERY:
-		return __get_eeprom_i2c_addr_ip_discovery(adev, control);
-
+	case IP_VERSION(13, 0, 0):
+		if (strnstr(atom_ctx->vbios_pn, "D707",
+			    sizeof(atom_ctx->vbios_pn)))
+			control->i2c_address = EEPROM_I2C_MADDR_0;
+		else
+			control->i2c_address = EEPROM_I2C_MADDR_4;
+		return true;
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 10):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
+		control->i2c_address = EEPROM_I2C_MADDR_4;
+		return true;
 	default:
 		return false;
 	}
@@ -250,15 +280,81 @@ static int __write_table_header(struct amdgpu_ras_eeprom_control *control)
 	up_read(&adev->reset_domain->sem);
 
 	if (res < 0) {
-		DRM_ERROR("Failed to write EEPROM table header:%d", res);
+		dev_err(adev->dev, "Failed to write EEPROM table header:%d",
+			res);
 	} else if (res < RAS_TABLE_HEADER_SIZE) {
-		DRM_ERROR("Short write:%d out of %d\n",
-			  res, RAS_TABLE_HEADER_SIZE);
+		dev_err(adev->dev, "Short write:%d out of %d\n", res,
+			RAS_TABLE_HEADER_SIZE);
+		res = -EIO;
+	} else {
+		res = 0;
+	}
+
+	return res;
+}
+
+static void
+__encode_table_ras_info_to_buf(struct amdgpu_ras_eeprom_table_ras_info *rai,
+			       unsigned char *buf)
+{
+	u32 *pp = (uint32_t *)buf;
+	u32 tmp;
+
+	tmp = ((uint32_t)(rai->rma_status) & 0xFF) |
+	      (((uint32_t)(rai->health_percent) << 8) & 0xFF00) |
+	      (((uint32_t)(rai->ecc_page_threshold) << 16) & 0xFFFF0000);
+	pp[0] = cpu_to_le32(tmp);
+}
+
+static void
+__decode_table_ras_info_from_buf(struct amdgpu_ras_eeprom_table_ras_info *rai,
+				 unsigned char *buf)
+{
+	u32 *pp = (uint32_t *)buf;
+	u32 tmp;
+
+	tmp = le32_to_cpu(pp[0]);
+	rai->rma_status = tmp & 0xFF;
+	rai->health_percent = (tmp >> 8) & 0xFF;
+	rai->ecc_page_threshold = (tmp >> 16) & 0xFFFF;
+}
+
+static int __write_table_ras_info(struct amdgpu_ras_eeprom_control *control)
+{
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	u8 *buf;
+	int res;
+
+	buf = kzalloc(RAS_TABLE_V2_1_INFO_SIZE, GFP_KERNEL);
+	if (!buf) {
+		dev_err(adev->dev,
+			"Failed to alloc buf to write table ras info\n");
+		return -ENOMEM;
+	}
+
+	__encode_table_ras_info_to_buf(&control->tbl_rai, buf);
+
+	/* i2c may be unstable in gpu reset */
+	down_read(&adev->reset_domain->sem);
+	res = amdgpu_eeprom_write(adev->pm.ras_eeprom_i2c_bus,
+				  control->i2c_address +
+				  control->ras_info_offset,
+				  buf, RAS_TABLE_V2_1_INFO_SIZE);
+	up_read(&adev->reset_domain->sem);
+
+	if (res < 0) {
+		dev_err(adev->dev, "Failed to write EEPROM table ras info:%d",
+			res);
+	} else if (res < RAS_TABLE_V2_1_INFO_SIZE) {
+		dev_err(adev->dev, "Short write:%d out of %d\n", res,
+			RAS_TABLE_V2_1_INFO_SIZE);
 		res = -EIO;
 	} else {
 		res = 0;
 	}
 
+	kfree(buf);
+
 	return res;
 }
 
@@ -278,6 +374,21 @@ static u8 __calc_hdr_byte_sum(const struct amdgpu_ras_eeprom_control *control)
 	return csum;
 }
 
+static u8 __calc_ras_info_byte_sum(const struct amdgpu_ras_eeprom_control *control)
+{
+	int ii;
+	u8  *pp, csum;
+	size_t sz;
+
+	sz = sizeof(control->tbl_rai);
+	pp = (u8 *) &control->tbl_rai;
+	csum = 0;
+	for (ii = 0; ii < sz; ii++, pp++)
+		csum += *pp;
+
+	return csum;
+}
+
 static int amdgpu_ras_eeprom_correct_header_tag(
 	struct amdgpu_ras_eeprom_control *control,
 	uint32_t header)
@@ -303,6 +414,25 @@ static int amdgpu_ras_eeprom_correct_header_tag(
 	return res;
 }
 
+static void amdgpu_ras_set_eeprom_table_version(struct amdgpu_ras_eeprom_control *control)
+{
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	struct amdgpu_ras_eeprom_table_header *hdr = &control->tbl_hdr;
+
+	switch (amdgpu_ip_version(adev, UMC_HWIP, 0)) {
+	case IP_VERSION(8, 10, 0):
+		hdr->version = RAS_TABLE_VER_V2_1;
+		return;
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 5, 0):
+		hdr->version = RAS_TABLE_VER_V3;
+		return;
+	default:
+		hdr->version = RAS_TABLE_VER_V1;
+		return;
+	}
+}
+
 /**
  * amdgpu_ras_eeprom_reset_table -- Reset the RAS EEPROM table
  * @control: pointer to control structure
@@ -314,26 +444,66 @@ int amdgpu_ras_eeprom_reset_table(struct amdgpu_ras_eeprom_control *control)
 {
 	struct amdgpu_device *adev = to_amdgpu_device(control);
 	struct amdgpu_ras_eeprom_table_header *hdr = &control->tbl_hdr;
+	struct amdgpu_ras_eeprom_table_ras_info *rai = &control->tbl_rai;
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	u32 erase_res = 0;
 	u8 csum;
 	int res;
 
 	mutex_lock(&control->ras_tbl_mutex);
 
-	hdr->header = RAS_TABLE_HDR_VAL;
-	hdr->version = RAS_TABLE_VER;
-	hdr->first_rec_offset = RAS_RECORD_START;
-	hdr->tbl_size = RAS_TABLE_HEADER_SIZE;
+	if (!amdgpu_ras_smu_eeprom_supported(adev)) {
+		hdr->header = RAS_TABLE_HDR_VAL;
+		amdgpu_ras_set_eeprom_table_version(control);
+
+		if (hdr->version >= RAS_TABLE_VER_V2_1) {
+			hdr->first_rec_offset = RAS_RECORD_START_V2_1;
+			hdr->tbl_size = RAS_TABLE_HEADER_SIZE +
+					RAS_TABLE_V2_1_INFO_SIZE;
+			rai->rma_status = GPU_HEALTH_USABLE;
+
+			control->ras_record_offset = RAS_RECORD_START_V2_1;
+			control->ras_max_record_count = RAS_MAX_RECORD_COUNT_V2_1;
+			/**
+			 * GPU health represented as a percentage.
+			 * 0 means worst health, 100 means fully health.
+			 */
+			rai->health_percent = 100;
+			/* ecc_page_threshold = 0 means disable bad page retirement */
+			rai->ecc_page_threshold = con->bad_page_cnt_threshold;
+		} else {
+			hdr->first_rec_offset = RAS_RECORD_START;
+			hdr->tbl_size = RAS_TABLE_HEADER_SIZE;
 
-	csum = __calc_hdr_byte_sum(control);
-	csum = -csum;
-	hdr->checksum = csum;
-	res = __write_table_header(control);
+			control->ras_record_offset = RAS_RECORD_START;
+			control->ras_max_record_count = RAS_MAX_RECORD_COUNT;
+		}
+
+		csum = __calc_hdr_byte_sum(control);
+		if (hdr->version >= RAS_TABLE_VER_V2_1)
+			csum += __calc_ras_info_byte_sum(control);
+		csum = -csum;
+		hdr->checksum = csum;
+		res = __write_table_header(control);
+		if (!res && hdr->version > RAS_TABLE_VER_V1)
+			res = __write_table_ras_info(control);
+	} else {
+		res = amdgpu_ras_smu_erase_ras_table(adev, &erase_res);
+		if (res || erase_res) {
+			dev_warn(adev->dev, "RAS EEPROM reset failed, res:%d result:%d",
+										res, erase_res);
+			if (!res)
+				res = -EIO;
+		}
+	}
 
 	control->ras_num_recs = 0;
+	control->ras_num_bad_pages = 0;
+	control->ras_num_mca_recs = 0;
+	control->ras_num_pa_recs = 0;
 	control->ras_fri = 0;
 
-	amdgpu_dpm_send_hbm_bad_pages_num(adev, control->ras_num_recs);
+	amdgpu_dpm_send_hbm_bad_pages_num(adev, control->ras_num_bad_pages);
 
 	control->bad_channel_bitmap = 0;
 	amdgpu_dpm_send_hbm_bad_channel_flag(adev, control->bad_channel_bitmap);
@@ -406,6 +576,9 @@ bool amdgpu_ras_eeprom_check_err_threshold(struct amdgpu_device *adev)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 
+	if (amdgpu_uniras_enabled(adev))
+		return amdgpu_ras_mgr_check_eeprom_safety_watermark(adev);
+
 	if (!__is_ras_eeprom_supported(adev) ||
 	    !amdgpu_bad_page_threshold)
 		return false;
@@ -418,16 +591,17 @@ bool amdgpu_ras_eeprom_check_err_threshold(struct amdgpu_device *adev)
 			return false;
 
 	if (con->eeprom_control.tbl_hdr.header == RAS_TABLE_HDR_BAD) {
-		if (amdgpu_bad_page_threshold == -1) {
+		if (con->eeprom_control.ras_num_bad_pages > con->bad_page_cnt_threshold)
 			dev_warn(adev->dev, "RAS records:%d exceed threshold:%d",
-				con->eeprom_control.ras_num_recs, con->bad_page_cnt_threshold);
+				 con->eeprom_control.ras_num_bad_pages, con->bad_page_cnt_threshold);
+		if ((amdgpu_bad_page_threshold == -1) ||
+		    (amdgpu_bad_page_threshold == -2)) {
 			dev_warn(adev->dev,
-				"But GPU can be operated due to bad_page_threshold = -1.\n");
+				 "Please consult AMD Service Action Guide (SAG) for appropriate service procedures.\n");
 			return false;
 		} else {
-			dev_warn(adev->dev, "This GPU is in BAD status.");
-			dev_warn(adev->dev, "Please retire it or set a larger "
-				 "threshold value when reloading driver.\n");
+			dev_warn(adev->dev,
+				 "Please consider adjusting the customized threshold.\n");
 			return true;
 		}
 	}
@@ -461,13 +635,13 @@ static int __amdgpu_ras_eeprom_write(struct amdgpu_ras_eeprom_control *control,
 				  buf, buf_size);
 	up_read(&adev->reset_domain->sem);
 	if (res < 0) {
-		DRM_ERROR("Writing %d EEPROM table records error:%d",
-			  num, res);
+		dev_err(adev->dev, "Writing %d EEPROM table records error:%d",
+			num, res);
 	} else if (res < buf_size) {
 		/* Short write, return error.
 		 */
-		DRM_ERROR("Wrote %d records out of %d",
-			  res / RAS_TABLE_RECORD_SIZE, num);
+		dev_err(adev->dev, "Wrote %d records out of %d",
+			res / RAS_TABLE_RECORD_SIZE, num);
 		res = -EIO;
 	} else {
 		res = 0;
@@ -482,6 +656,7 @@ amdgpu_ras_eeprom_append_table(struct amdgpu_ras_eeprom_control *control,
 			       const u32 num)
 {
 	struct amdgpu_ras *con = amdgpu_ras_get_context(to_amdgpu_device(control));
+	struct amdgpu_device *adev = to_amdgpu_device(control);
 	u32 a, b, i;
 	u8 *buf, *pp;
 	int res;
@@ -497,7 +672,8 @@ amdgpu_ras_eeprom_append_table(struct amdgpu_ras_eeprom_control *control,
 		__encode_table_record_to_buf(control, &record[i], pp);
 
 		/* update bad channel bitmap */
-		if (!(control->bad_channel_bitmap & (1 << record[i].mem_channel))) {
+		if ((record[i].mem_channel < BITS_PER_TYPE(control->bad_channel_bitmap)) &&
+		    !(control->bad_channel_bitmap & (1 << record[i].mem_channel))) {
 			control->bad_channel_bitmap |= 1 << record[i].mem_channel;
 			con->update_channel_flag = true;
 		}
@@ -583,6 +759,14 @@ amdgpu_ras_eeprom_append_table(struct amdgpu_ras_eeprom_control *control,
 	control->ras_num_recs = 1 + (control->ras_max_record_count + b
 				     - control->ras_fri)
 		% control->ras_max_record_count;
+
+	/*old asics only save pa to eeprom like before*/
+	if (IP_VERSION_MAJ(amdgpu_ip_version(adev, UMC_HWIP, 0)) < 12)
+		control->ras_num_pa_recs += num;
+	else
+		control->ras_num_mca_recs += num;
+
+	control->ras_num_bad_pages = con->bad_page_num;
 Out:
 	kfree(buf);
 	return res;
@@ -600,23 +784,44 @@ amdgpu_ras_eeprom_update_header(struct amdgpu_ras_eeprom_control *control)
 	/* Modify the header if it exceeds.
 	 */
 	if (amdgpu_bad_page_threshold != 0 &&
-	    control->ras_num_recs >= ras->bad_page_cnt_threshold) {
+	    control->ras_num_bad_pages > ras->bad_page_cnt_threshold) {
 		dev_warn(adev->dev,
 			"Saved bad pages %d reaches threshold value %d\n",
-			control->ras_num_recs, ras->bad_page_cnt_threshold);
-		control->tbl_hdr.header = RAS_TABLE_HDR_BAD;
+			control->ras_num_bad_pages, ras->bad_page_cnt_threshold);
+
+		if (adev->cper.enabled && !amdgpu_uniras_enabled(adev) &&
+		    amdgpu_cper_generate_bp_threshold_record(adev))
+			dev_warn(adev->dev, "fail to generate bad page threshold cper records\n");
+
+		if ((amdgpu_bad_page_threshold != -1) &&
+		    (amdgpu_bad_page_threshold != -2)) {
+			control->tbl_hdr.header = RAS_TABLE_HDR_BAD;
+			if (control->tbl_hdr.version >= RAS_TABLE_VER_V2_1) {
+				control->tbl_rai.rma_status = GPU_RETIRED__ECC_REACH_THRESHOLD;
+				control->tbl_rai.health_percent = 0;
+			}
+			ras->is_rma = true;
+		}
+
+		/* ignore the -ENOTSUPP return value */
+		amdgpu_dpm_send_rma_reason(adev);
 	}
 
-	control->tbl_hdr.version = RAS_TABLE_VER;
-	control->tbl_hdr.first_rec_offset = RAS_INDEX_TO_OFFSET(control, control->ras_fri);
-	control->tbl_hdr.tbl_size = RAS_TABLE_HEADER_SIZE + control->ras_num_recs * RAS_TABLE_RECORD_SIZE;
+	if (control->tbl_hdr.version >= RAS_TABLE_VER_V2_1)
+		control->tbl_hdr.tbl_size = RAS_TABLE_HEADER_SIZE +
+					    RAS_TABLE_V2_1_INFO_SIZE +
+					    control->ras_num_recs * RAS_TABLE_RECORD_SIZE;
+	else
+		control->tbl_hdr.tbl_size = RAS_TABLE_HEADER_SIZE +
+					    control->ras_num_recs * RAS_TABLE_RECORD_SIZE;
 	control->tbl_hdr.checksum = 0;
 
 	buf_size = control->ras_num_recs * RAS_TABLE_RECORD_SIZE;
 	buf = kcalloc(control->ras_num_recs, RAS_TABLE_RECORD_SIZE, GFP_KERNEL);
 	if (!buf) {
-		DRM_ERROR("allocating memory for table of size %d bytes failed\n",
-			  control->tbl_hdr.tbl_size);
+		dev_err(adev->dev,
+			"allocating memory for table of size %d bytes failed\n",
+			control->tbl_hdr.tbl_size);
 		res = -ENOMEM;
 		goto Out;
 	}
@@ -628,16 +833,26 @@ amdgpu_ras_eeprom_update_header(struct amdgpu_ras_eeprom_control *control)
 				 buf, buf_size);
 	up_read(&adev->reset_domain->sem);
 	if (res < 0) {
-		DRM_ERROR("EEPROM failed reading records:%d\n",
-			  res);
+		dev_err(adev->dev, "EEPROM failed reading records:%d\n", res);
 		goto Out;
 	} else if (res < buf_size) {
-		DRM_ERROR("EEPROM read %d out of %d bytes\n",
-			  res, buf_size);
+		dev_err(adev->dev, "EEPROM read %d out of %d bytes\n", res,
+			buf_size);
 		res = -EIO;
 		goto Out;
 	}
 
+	/**
+	 * bad page records have been stored in eeprom,
+	 * now calculate gpu health percent
+	 */
+	if (amdgpu_bad_page_threshold != 0 &&
+	    control->tbl_hdr.version >= RAS_TABLE_VER_V2_1 &&
+	    control->ras_num_bad_pages <= ras->bad_page_cnt_threshold)
+		control->tbl_rai.health_percent = ((ras->bad_page_cnt_threshold -
+						   control->ras_num_bad_pages) * 100) /
+						   ras->bad_page_cnt_threshold;
+
 	/* Recalc the checksum.
 	 */
 	csum = 0;
@@ -645,15 +860,84 @@ amdgpu_ras_eeprom_update_header(struct amdgpu_ras_eeprom_control *control)
 		csum += *pp;
 
 	csum += __calc_hdr_byte_sum(control);
+	if (control->tbl_hdr.version >= RAS_TABLE_VER_V2_1)
+		csum += __calc_ras_info_byte_sum(control);
 	/* avoid sign extension when assigning to "checksum" */
 	csum = -csum;
 	control->tbl_hdr.checksum = csum;
 	res = __write_table_header(control);
+	if (!res && control->tbl_hdr.version > RAS_TABLE_VER_V1)
+		res = __write_table_ras_info(control);
 Out:
 	kfree(buf);
 	return res;
 }
 
+int amdgpu_ras_eeprom_update_record_num(struct amdgpu_ras_eeprom_control *control)
+{
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	int ret, retry = 20;
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return 0;
+
+	control->ras_num_recs_old = control->ras_num_recs;
+
+	do {
+		/* 1000ms timeout is long enough, smu_get_badpage_count won't
+		 * return -EBUSY before timeout.
+		 */
+		ret = amdgpu_ras_smu_get_badpage_count(adev,
+			&(control->ras_num_recs), RAS_SMU_MESSAGE_TIMEOUT_MS);
+		if (!ret &&
+		    (control->ras_num_recs_old == control->ras_num_recs)) {
+			/* record number update in PMFW needs some time,
+			 * smu_get_badpage_count may return immediately without
+			 * count update, sleep for a while and retry again.
+			 */
+			msleep(50);
+			retry--;
+		} else {
+			break;
+		}
+	} while (retry);
+
+	/* no update of record number is not a real failure,
+	 * don't print warning here
+	 */
+	if (!ret && (control->ras_num_recs_old == control->ras_num_recs))
+		ret = -EINVAL;
+
+	return ret;
+}
+
+static int amdgpu_ras_smu_eeprom_append(struct amdgpu_ras_eeprom_control *control)
+{
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev) || !con)
+		return 0;
+
+	control->ras_num_bad_pages = con->bad_page_num;
+
+	if (amdgpu_bad_page_threshold != 0 &&
+	    control->ras_num_bad_pages > con->bad_page_cnt_threshold) {
+		dev_warn(adev->dev,
+			"Saved bad pages %d reaches threshold value %d\n",
+			control->ras_num_bad_pages, con->bad_page_cnt_threshold);
+
+		if (adev->cper.enabled && amdgpu_cper_generate_bp_threshold_record(adev))
+			dev_warn(adev->dev, "fail to generate bad page threshold cper records\n");
+
+		if ((amdgpu_bad_page_threshold != -1) &&
+		    (amdgpu_bad_page_threshold != -2))
+			con->is_rma = true;
+	}
+
+	return 0;
+}
+
 /**
  * amdgpu_ras_eeprom_append -- append records to the EEPROM RAS table
  * @control: pointer to control structure
@@ -672,20 +956,32 @@ int amdgpu_ras_eeprom_append(struct amdgpu_ras_eeprom_control *control,
 			     const u32 num)
 {
 	struct amdgpu_device *adev = to_amdgpu_device(control);
-	int res;
+	int res, i;
+	uint64_t nps = AMDGPU_NPS1_PARTITION_MODE;
 
 	if (!__is_ras_eeprom_supported(adev))
 		return 0;
 
+	if (amdgpu_ras_smu_eeprom_supported(adev))
+		return amdgpu_ras_smu_eeprom_append(control);
+
 	if (num == 0) {
-		DRM_ERROR("will not append 0 records\n");
+		dev_err(adev->dev, "will not append 0 records\n");
 		return -EINVAL;
 	} else if (num > control->ras_max_record_count) {
-		DRM_ERROR("cannot append %d records than the size of table %d\n",
-			  num, control->ras_max_record_count);
+		dev_err(adev->dev,
+			"cannot append %d records than the size of table %d\n",
+			num, control->ras_max_record_count);
 		return -EINVAL;
 	}
 
+	if (adev->gmc.gmc_funcs->query_mem_partition_mode)
+		nps = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+
+	/* set the new channel index flag */
+	for (i = 0; i < num; i++)
+		record[i].retired_page |= (nps << UMC_NPS_SHIFT);
+
 	mutex_lock(&control->ras_tbl_mutex);
 
 	res = amdgpu_ras_eeprom_append_table(control, record, num);
@@ -695,6 +991,11 @@ int amdgpu_ras_eeprom_append(struct amdgpu_ras_eeprom_control *control,
 		amdgpu_ras_debugfs_set_ret_size(control);
 
 	mutex_unlock(&control->ras_tbl_mutex);
+
+	/* clear channel index flag, the flag is only saved on eeprom */
+	for (i = 0; i < num; i++)
+		record[i].retired_page &= ~(nps << UMC_NPS_SHIFT);
+
 	return res;
 }
 
@@ -724,13 +1025,13 @@ static int __amdgpu_ras_eeprom_read(struct amdgpu_ras_eeprom_control *control,
 				 buf, buf_size);
 	up_read(&adev->reset_domain->sem);
 	if (res < 0) {
-		DRM_ERROR("Reading %d EEPROM table records error:%d",
-			  num, res);
+		dev_err(adev->dev, "Reading %d EEPROM table records error:%d",
+			num, res);
 	} else if (res < buf_size) {
 		/* Short read, return error.
 		 */
-		DRM_ERROR("Read %d records out of %d",
-			  res / RAS_TABLE_RECORD_SIZE, num);
+		dev_err(adev->dev, "Read %d records out of %d",
+			res / RAS_TABLE_RECORD_SIZE, num);
 		res = -EIO;
 	} else {
 		res = 0;
@@ -739,6 +1040,50 @@ static int __amdgpu_ras_eeprom_read(struct amdgpu_ras_eeprom_control *control,
 	return res;
 }
 
+int amdgpu_ras_eeprom_read_idx(struct amdgpu_ras_eeprom_control *control,
+			struct eeprom_table_record *record, u32 rec_idx,
+			const u32 num)
+{
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	uint64_t ts, end_idx;
+	int i, ret;
+	u64 mca, ipid;
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return 0;
+
+	if (!adev->umc.ras || !adev->umc.ras->mca_ipid_parse)
+		return -EOPNOTSUPP;
+
+	end_idx = rec_idx + num;
+	for (i = rec_idx; i < end_idx; i++) {
+		ret = amdgpu_ras_smu_get_badpage_mca_addr(adev, i, &mca);
+		if (ret)
+			return ret;
+
+		ret = amdgpu_ras_smu_get_badpage_ipid(adev, i, &ipid);
+		if (ret)
+			return ret;
+
+		ret = amdgpu_ras_smu_get_timestamp(adev, i, &ts);
+		if (ret)
+			return ret;
+
+		record[i - rec_idx].address = mca;
+		/* retired_page (pa) is unused now */
+		record[i - rec_idx].retired_page = 0x1ULL;
+		record[i - rec_idx].ts = ts;
+		record[i - rec_idx].err_type = AMDGPU_RAS_EEPROM_ERR_NON_RECOVERABLE;
+
+		adev->umc.ras->mca_ipid_parse(adev, ipid,
+			(uint32_t *)&(record[i - rec_idx].cu),
+			(uint32_t *)&(record[i - rec_idx].mem_channel),
+			(uint32_t *)&(record[i - rec_idx].mcumc_id), NULL);
+	}
+
+	return 0;
+}
+
 /**
  * amdgpu_ras_eeprom_read -- read EEPROM
  * @control: pointer to control structure
@@ -760,15 +1105,18 @@ int amdgpu_ras_eeprom_read(struct amdgpu_ras_eeprom_control *control,
 	u8 *buf, *pp;
 	u32 g0, g1;
 
+	if (amdgpu_ras_smu_eeprom_supported(adev))
+		return amdgpu_ras_eeprom_read_idx(control, record, 0, num);
+
 	if (!__is_ras_eeprom_supported(adev))
 		return 0;
 
 	if (num == 0) {
-		DRM_ERROR("will not read 0 records\n");
+		dev_err(adev->dev, "will not read 0 records\n");
 		return -EINVAL;
 	} else if (num > control->ras_num_recs) {
-		DRM_ERROR("too many records to read:%d available:%d\n",
-			  num, control->ras_num_recs);
+		dev_err(adev->dev, "too many records to read:%d available:%d\n",
+			num, control->ras_num_recs);
 		return -EINVAL;
 	}
 
@@ -827,7 +1175,8 @@ int amdgpu_ras_eeprom_read(struct amdgpu_ras_eeprom_control *control,
 		__decode_table_record_from_buf(control, &record[i], pp);
 
 		/* update bad channel bitmap */
-		if (!(control->bad_channel_bitmap & (1 << record[i].mem_channel))) {
+		if ((record[i].mem_channel < BITS_PER_TYPE(control->bad_channel_bitmap)) &&
+		    !(control->bad_channel_bitmap & (1 << record[i].mem_channel))) {
 			control->bad_channel_bitmap |= 1 << record[i].mem_channel;
 			con->update_channel_flag = true;
 		}
@@ -839,9 +1188,15 @@ Out:
 	return res;
 }
 
-uint32_t amdgpu_ras_eeprom_max_record_count(void)
+uint32_t amdgpu_ras_eeprom_max_record_count(struct amdgpu_ras_eeprom_control *control)
 {
-	return RAS_MAX_RECORD_COUNT;
+	/* get available eeprom table version first before eeprom table init */
+	amdgpu_ras_set_eeprom_table_version(control);
+
+	if (control->tbl_hdr.version >= RAS_TABLE_VER_V2_1)
+		return RAS_MAX_RECORD_COUNT_V2_1;
+	else
+		return RAS_MAX_RECORD_COUNT;
 }
 
 static ssize_t
@@ -924,6 +1279,10 @@ static ssize_t amdgpu_ras_debugfs_table_read(struct file *f, char __user *buf,
 	int res = -EFAULT;
 	size_t data_len;
 
+	/* pmfw manages eeprom data by itself */
+	if (amdgpu_ras_smu_eeprom_supported(adev))
+		return 0;
+
 	mutex_lock(&control->ras_tbl_mutex);
 
 	/* We want *pos - data_len > 0, which means there's
@@ -1083,11 +1442,18 @@ static int __verify_ras_table_checksum(struct amdgpu_ras_eeprom_control *control
 	int buf_size, res;
 	u8  csum, *buf, *pp;
 
-	buf_size = RAS_TABLE_HEADER_SIZE +
-		control->ras_num_recs * RAS_TABLE_RECORD_SIZE;
+	if (control->tbl_hdr.version >= RAS_TABLE_VER_V2_1)
+		buf_size = RAS_TABLE_HEADER_SIZE +
+			   RAS_TABLE_V2_1_INFO_SIZE +
+			   control->ras_num_recs * RAS_TABLE_RECORD_SIZE;
+	else
+		buf_size = RAS_TABLE_HEADER_SIZE +
+			   control->ras_num_recs * RAS_TABLE_RECORD_SIZE;
+
 	buf = kzalloc(buf_size, GFP_KERNEL);
 	if (!buf) {
-		DRM_ERROR("Out of memory checking RAS table checksum.\n");
+		dev_err(adev->dev,
+			"Out of memory checking RAS table checksum.\n");
 		return -ENOMEM;
 	}
 
@@ -1096,7 +1462,7 @@ static int __verify_ras_table_checksum(struct amdgpu_ras_eeprom_control *control
 				 control->ras_header_offset,
 				 buf, buf_size);
 	if (res < buf_size) {
-		DRM_ERROR("Partial read for checksum, res:%d\n", res);
+		dev_err(adev->dev, "Partial read for checksum, res:%d\n", res);
 		/* On partial reads, return -EIO.
 		 */
 		if (res >= 0)
@@ -1112,8 +1478,78 @@ Out:
 	return res < 0 ? res : csum;
 }
 
-int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control,
-			   bool *exceed_err_limit)
+static int __read_table_ras_info(struct amdgpu_ras_eeprom_control *control)
+{
+	struct amdgpu_ras_eeprom_table_ras_info *rai = &control->tbl_rai;
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	unsigned char *buf;
+	int res;
+
+	buf = kzalloc(RAS_TABLE_V2_1_INFO_SIZE, GFP_KERNEL);
+	if (!buf) {
+		dev_err(adev->dev,
+			"Failed to alloc buf to read EEPROM table ras info\n");
+		return -ENOMEM;
+	}
+
+	/**
+	 * EEPROM table V2_1 supports ras info,
+	 * read EEPROM table ras info
+	 */
+	res = amdgpu_eeprom_read(adev->pm.ras_eeprom_i2c_bus,
+				 control->i2c_address + control->ras_info_offset,
+				 buf, RAS_TABLE_V2_1_INFO_SIZE);
+	if (res < RAS_TABLE_V2_1_INFO_SIZE) {
+		dev_err(adev->dev,
+			"Failed to read EEPROM table ras info, res:%d", res);
+		res = res >= 0 ? -EIO : res;
+		goto Out;
+	}
+
+	__decode_table_ras_info_from_buf(rai, buf);
+
+Out:
+	kfree(buf);
+	return res == RAS_TABLE_V2_1_INFO_SIZE ? 0 : res;
+}
+
+static int amdgpu_ras_smu_eeprom_init(struct amdgpu_ras_eeprom_control *control)
+{
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	struct amdgpu_ras_eeprom_table_header *hdr = &control->tbl_hdr;
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+	uint64_t local_time;
+	int res;
+
+	ras->is_rma = false;
+
+	if (!__is_ras_eeprom_supported(adev))
+		return 0;
+	mutex_init(&control->ras_tbl_mutex);
+
+	res = amdgpu_ras_smu_get_table_version(adev, &(hdr->version));
+	if (res)
+		return res;
+
+	res = amdgpu_ras_smu_get_badpage_count(adev,
+								&(control->ras_num_recs), 100);
+	if (res)
+		return res;
+
+	local_time = (uint64_t)ktime_get_real_seconds();
+	res = amdgpu_ras_smu_set_timestamp(adev, local_time);
+	if (res)
+		return res;
+
+	control->ras_max_record_count = 4000;
+
+	control->ras_num_mca_recs = 0;
+	control->ras_num_pa_recs = 0;
+
+	return 0;
+}
+
+int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control)
 {
 	struct amdgpu_device *adev = to_amdgpu_device(control);
 	unsigned char buf[RAS_TABLE_HEADER_SIZE] = { 0 };
@@ -1121,7 +1557,10 @@ int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control,
 	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
 	int res;
 
-	*exceed_err_limit = false;
+	if (amdgpu_ras_smu_eeprom_supported(adev))
+		return amdgpu_ras_smu_eeprom_init(control);
+
+	ras->is_rma = false;
 
 	if (!__is_ras_eeprom_supported(adev))
 		return 0;
@@ -1134,8 +1573,7 @@ int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control,
 		return -EINVAL;
 
 	control->ras_header_offset = RAS_HDR_START;
-	control->ras_record_offset = RAS_RECORD_START;
-	control->ras_max_record_count  = RAS_MAX_RECORD_COUNT;
+	control->ras_info_offset = RAS_TABLE_V2_1_INFO_START;
 	mutex_init(&control->ras_tbl_mutex);
 
 	/* Read the table header from EEPROM address */
@@ -1143,36 +1581,153 @@ int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control,
 				 control->i2c_address + control->ras_header_offset,
 				 buf, RAS_TABLE_HEADER_SIZE);
 	if (res < RAS_TABLE_HEADER_SIZE) {
-		DRM_ERROR("Failed to read EEPROM table header, res:%d", res);
+		dev_err(adev->dev, "Failed to read EEPROM table header, res:%d",
+			res);
 		return res >= 0 ? -EIO : res;
 	}
 
 	__decode_table_header_from_buf(hdr, buf);
 
-	control->ras_num_recs = RAS_NUM_RECS(hdr);
+	if (hdr->header != RAS_TABLE_HDR_VAL &&
+	    hdr->header != RAS_TABLE_HDR_BAD) {
+		dev_info(adev->dev, "Creating a new EEPROM table");
+		return amdgpu_ras_eeprom_reset_table(control);
+	}
+
+	switch (hdr->version) {
+	case RAS_TABLE_VER_V2_1:
+	case RAS_TABLE_VER_V3:
+		control->ras_num_recs = RAS_NUM_RECS_V2_1(hdr);
+		control->ras_record_offset = RAS_RECORD_START_V2_1;
+		control->ras_max_record_count = RAS_MAX_RECORD_COUNT_V2_1;
+		break;
+	case RAS_TABLE_VER_V1:
+		control->ras_num_recs = RAS_NUM_RECS(hdr);
+		control->ras_record_offset = RAS_RECORD_START;
+		control->ras_max_record_count = RAS_MAX_RECORD_COUNT;
+		break;
+	default:
+		dev_err(adev->dev,
+			"RAS header invalid, unsupported version: %u",
+			hdr->version);
+		return -EINVAL;
+	}
+
+	if (control->ras_num_recs > control->ras_max_record_count) {
+		dev_err(adev->dev,
+			"RAS header invalid, records in header: %u max allowed :%u",
+			control->ras_num_recs, control->ras_max_record_count);
+		return -EINVAL;
+	}
+
 	control->ras_fri = RAS_OFFSET_TO_INDEX(control, hdr->first_rec_offset);
+	control->ras_num_mca_recs = 0;
+	control->ras_num_pa_recs = 0;
+	return 0;
+}
+
+static int amdgpu_ras_smu_eeprom_check(struct amdgpu_ras_eeprom_control *control)
+{
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+
+	if (!__is_ras_eeprom_supported(adev))
+		return 0;
+
+	control->ras_num_bad_pages = ras->bad_page_num;
+
+	if ((ras->bad_page_cnt_threshold < control->ras_num_bad_pages) &&
+	    amdgpu_bad_page_threshold != 0) {
+		dev_warn(adev->dev,
+			"RAS records:%d exceed threshold:%d\n",
+			control->ras_num_bad_pages, ras->bad_page_cnt_threshold);
+		if ((amdgpu_bad_page_threshold == -1) ||
+			(amdgpu_bad_page_threshold == -2)) {
+			dev_warn(adev->dev,
+				 "Please consult AMD Service Action Guide (SAG) for appropriate service procedures\n");
+		} else {
+			ras->is_rma = true;
+			dev_warn(adev->dev,
+				 "User defined threshold is set, runtime service will be halt when threshold is reached\n");
+		}
+
+		return 0;
+	}
+
+	dev_dbg(adev->dev,
+		"Found existing EEPROM table with %d records",
+		control->ras_num_bad_pages);
+
+	/* Warn if we are at 90% of the threshold or above
+	 */
+	if (10 * control->ras_num_bad_pages >= 9 * ras->bad_page_cnt_threshold)
+		dev_warn(adev->dev, "RAS records:%u exceeds 90%% of threshold:%d",
+				control->ras_num_bad_pages,
+				ras->bad_page_cnt_threshold);
+	return 0;
+}
+
+int amdgpu_ras_eeprom_check(struct amdgpu_ras_eeprom_control *control)
+{
+	struct amdgpu_device *adev = to_amdgpu_device(control);
+	struct amdgpu_ras_eeprom_table_header *hdr = &control->tbl_hdr;
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+	int res = 0;
+
+	if (amdgpu_ras_smu_eeprom_supported(adev))
+		return amdgpu_ras_smu_eeprom_check(control);
+
+	if (!__is_ras_eeprom_supported(adev))
+		return 0;
+
+	/* Verify i2c adapter is initialized */
+	if (!adev->pm.ras_eeprom_i2c_bus || !adev->pm.ras_eeprom_i2c_bus->algo)
+		return -ENOENT;
+
+	if (!__get_eeprom_i2c_addr(adev, control))
+		return -EINVAL;
+
+	control->ras_num_bad_pages = ras->bad_page_num;
 
 	if (hdr->header == RAS_TABLE_HDR_VAL) {
-		DRM_DEBUG_DRIVER("Found existing EEPROM table with %d records",
-				 control->ras_num_recs);
+		dev_dbg(adev->dev,
+			"Found existing EEPROM table with %d records",
+			control->ras_num_bad_pages);
+
+		if (hdr->version >= RAS_TABLE_VER_V2_1) {
+			res = __read_table_ras_info(control);
+			if (res)
+				return res;
+		}
+
 		res = __verify_ras_table_checksum(control);
 		if (res)
-			DRM_ERROR("RAS table incorrect checksum or error:%d\n",
-				  res);
+			dev_err(adev->dev,
+				"RAS table incorrect checksum or error:%d\n",
+				res);
 
 		/* Warn if we are at 90% of the threshold or above
 		 */
-		if (10 * control->ras_num_recs >= 9 * ras->bad_page_cnt_threshold)
+		if (10 * control->ras_num_bad_pages >= 9 * ras->bad_page_cnt_threshold)
 			dev_warn(adev->dev, "RAS records:%u exceeds 90%% of threshold:%d",
-					control->ras_num_recs,
+					control->ras_num_bad_pages,
 					ras->bad_page_cnt_threshold);
 	} else if (hdr->header == RAS_TABLE_HDR_BAD &&
 		   amdgpu_bad_page_threshold != 0) {
+		if (hdr->version >= RAS_TABLE_VER_V2_1) {
+			res = __read_table_ras_info(control);
+			if (res)
+				return res;
+		}
+
 		res = __verify_ras_table_checksum(control);
-		if (res)
-			DRM_ERROR("RAS Table incorrect checksum or error:%d\n",
-				  res);
-		if (ras->bad_page_cnt_threshold > control->ras_num_recs) {
+		if (res) {
+			dev_err(adev->dev,
+				"RAS Table incorrect checksum or error:%d\n",
+				res);
+			return -EINVAL;
+		}
+		if (ras->bad_page_cnt_threshold >= control->ras_num_bad_pages) {
 			/* This means that, the threshold was increased since
 			 * the last time the system was booted, and now,
 			 * ras->bad_page_cnt_threshold - control->num_recs > 0,
@@ -1182,29 +1737,194 @@ int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control,
 			dev_info(adev->dev,
 				 "records:%d threshold:%d, resetting "
 				 "RAS table header signature",
-				 control->ras_num_recs,
+				 control->ras_num_bad_pages,
 				 ras->bad_page_cnt_threshold);
 			res = amdgpu_ras_eeprom_correct_header_tag(control,
 								   RAS_TABLE_HDR_VAL);
 		} else {
-			dev_err(adev->dev, "RAS records:%d exceed threshold:%d",
-				control->ras_num_recs, ras->bad_page_cnt_threshold);
-			if (amdgpu_bad_page_threshold == -1) {
-				dev_warn(adev->dev, "GPU will be initialized due to bad_page_threshold = -1.");
+			dev_warn(adev->dev,
+				"RAS records:%d exceed threshold:%d\n",
+				control->ras_num_bad_pages, ras->bad_page_cnt_threshold);
+			if ((amdgpu_bad_page_threshold == -1) ||
+			    (amdgpu_bad_page_threshold == -2)) {
 				res = 0;
+				dev_warn(adev->dev,
+					 "Please consult AMD Service Action Guide (SAG) for appropriate service procedures\n");
 			} else {
-				*exceed_err_limit = true;
-				dev_err(adev->dev,
-					"RAS records:%d exceed threshold:%d, "
-					"GPU will not be initialized. Replace this GPU or increase the threshold",
-					control->ras_num_recs, ras->bad_page_cnt_threshold);
+				ras->is_rma = true;
+				dev_warn(adev->dev,
+					 "User defined threshold is set, runtime service will be halt when threshold is reached\n");
 			}
 		}
-	} else {
-		DRM_INFO("Creating a new EEPROM table");
-
-		res = amdgpu_ras_eeprom_reset_table(control);
 	}
 
 	return res < 0 ? res : 0;
 }
+
+void amdgpu_ras_eeprom_check_and_recover(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+	struct amdgpu_ras_eeprom_control *control;
+	int res;
+
+	if (!__is_ras_eeprom_supported(adev) || !ras ||
+	    amdgpu_ras_smu_eeprom_supported(adev))
+		return;
+	control = &ras->eeprom_control;
+	if (!control->is_eeprom_valid)
+		return;
+	res = __verify_ras_table_checksum(control);
+	if (res) {
+		dev_warn(adev->dev,
+			"RAS table incorrect checksum or error:%d, try to recover\n",
+			res);
+		if (!amdgpu_ras_eeprom_reset_table(control))
+			if (!amdgpu_ras_save_bad_pages(adev, NULL))
+				if (!__verify_ras_table_checksum(control)) {
+					dev_info(adev->dev, "RAS table recovery succeed\n");
+					return;
+				}
+		dev_err(adev->dev, "RAS table recovery failed\n");
+		control->is_eeprom_valid = false;
+	}
+	return;
+}
+
+static const struct ras_smu_drv *amdgpu_ras_get_smu_ras_drv(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+
+	if (!ras)
+		return NULL;
+
+	return ras->ras_smu_drv;
+}
+
+static uint64_t amdgpu_ras_smu_get_feature_flags(struct amdgpu_device *adev)
+{
+	const struct ras_smu_drv *ras_smu_drv = amdgpu_ras_get_smu_ras_drv(adev);
+	uint64_t flags = 0ULL;
+
+	if (!ras_smu_drv)
+		goto out;
+
+	if (ras_smu_drv->ras_smu_feature_flags)
+		ras_smu_drv->ras_smu_feature_flags(adev, &flags);
+
+out:
+	return flags;
+}
+
+bool amdgpu_ras_smu_eeprom_supported(struct amdgpu_device *adev)
+{
+	const struct ras_smu_drv *smu_ras_drv = amdgpu_ras_get_smu_ras_drv(adev);
+	uint64_t flags = 0ULL;
+
+	if (!__is_ras_eeprom_supported(adev) || !smu_ras_drv)
+		return false;
+
+	if (!smu_ras_drv->smu_eeprom_funcs)
+		return false;
+
+	flags = amdgpu_ras_smu_get_feature_flags(adev);
+
+	return !!(flags & RAS_SMU_FEATURE_BIT__RAS_EEPROM);
+}
+
+int amdgpu_ras_smu_get_table_version(struct amdgpu_device *adev,
+				     uint32_t *table_version)
+{
+	const struct ras_smu_drv *smu_ras_drv = amdgpu_ras_get_smu_ras_drv(adev);
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return -EOPNOTSUPP;
+
+	if (smu_ras_drv->smu_eeprom_funcs->get_ras_table_version)
+		return smu_ras_drv->smu_eeprom_funcs->get_ras_table_version(adev,
+										 table_version);
+	return -EOPNOTSUPP;
+}
+
+int amdgpu_ras_smu_get_badpage_count(struct amdgpu_device *adev,
+				     uint32_t *count, uint32_t timeout)
+{
+	const struct ras_smu_drv *smu_ras_drv = amdgpu_ras_get_smu_ras_drv(adev);
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return -EOPNOTSUPP;
+
+	if (smu_ras_drv->smu_eeprom_funcs->get_badpage_count)
+		return smu_ras_drv->smu_eeprom_funcs->get_badpage_count(adev,
+									     count, timeout);
+	return -EOPNOTSUPP;
+}
+
+int amdgpu_ras_smu_get_badpage_mca_addr(struct amdgpu_device *adev,
+					uint16_t index, uint64_t *mca_addr)
+{
+	const struct ras_smu_drv *smu_ras_drv = amdgpu_ras_get_smu_ras_drv(adev);
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return -EOPNOTSUPP;
+
+	if (smu_ras_drv->smu_eeprom_funcs->get_badpage_mca_addr)
+		return smu_ras_drv->smu_eeprom_funcs->get_badpage_mca_addr(adev,
+										index, mca_addr);
+	return -EOPNOTSUPP;
+}
+
+int amdgpu_ras_smu_set_timestamp(struct amdgpu_device *adev,
+				 uint64_t timestamp)
+{
+	const struct ras_smu_drv *smu_ras_drv = amdgpu_ras_get_smu_ras_drv(adev);
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return -EOPNOTSUPP;
+
+	if (smu_ras_drv->smu_eeprom_funcs->set_timestamp)
+		return smu_ras_drv->smu_eeprom_funcs->set_timestamp(adev,
+									 timestamp);
+	return -EOPNOTSUPP;
+}
+
+int amdgpu_ras_smu_get_timestamp(struct amdgpu_device *adev,
+				 uint16_t index, uint64_t *timestamp)
+{
+	const struct ras_smu_drv *smu_ras_drv = amdgpu_ras_get_smu_ras_drv(adev);
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return -EOPNOTSUPP;
+
+	if (smu_ras_drv->smu_eeprom_funcs->get_timestamp)
+		return smu_ras_drv->smu_eeprom_funcs->get_timestamp(adev,
+									 index, timestamp);
+	return -EOPNOTSUPP;
+}
+
+int amdgpu_ras_smu_get_badpage_ipid(struct amdgpu_device *adev,
+				    uint16_t index, uint64_t *ipid)
+{
+	const struct ras_smu_drv *smu_ras_drv = amdgpu_ras_get_smu_ras_drv(adev);
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return -EOPNOTSUPP;
+
+	if (smu_ras_drv->smu_eeprom_funcs->get_badpage_ipid)
+		return smu_ras_drv->smu_eeprom_funcs->get_badpage_ipid(adev,
+									    index, ipid);
+	return -EOPNOTSUPP;
+}
+
+int amdgpu_ras_smu_erase_ras_table(struct amdgpu_device *adev,
+				   uint32_t *result)
+{
+	const struct ras_smu_drv *smu_ras_drv = amdgpu_ras_get_smu_ras_drv(adev);
+
+	if (!amdgpu_ras_smu_eeprom_supported(adev))
+		return -EOPNOTSUPP;
+
+	if (smu_ras_drv->smu_eeprom_funcs->erase_ras_table)
+		return smu_ras_drv->smu_eeprom_funcs->erase_ras_table(adev,
+									   result);
+	return -EOPNOTSUPP;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h
index 54d9bfe0881d..2e5d63957e71 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.h
@@ -26,8 +26,17 @@
 
 #include <linux/i2c.h>
 
+#define RAS_TABLE_VER_V1           0x00010000
+#define RAS_TABLE_VER_V2_1         0x00021000
+#define RAS_TABLE_VER_V3           0x00030000
+
 struct amdgpu_device;
 
+enum amdgpu_ras_gpu_health_status {
+	GPU_HEALTH_USABLE = 0,
+	GPU_RETIRED__ECC_REACH_THRESHOLD = 2,
+};
+
 enum amdgpu_ras_eeprom_err_type {
 	AMDGPU_RAS_EEPROM_ERR_NA,
 	AMDGPU_RAS_EEPROM_ERR_RECOVERABLE,
@@ -43,9 +52,18 @@ struct amdgpu_ras_eeprom_table_header {
 	uint32_t checksum;
 } __packed;
 
+struct amdgpu_ras_eeprom_table_ras_info {
+	u8  rma_status;
+	u8  health_percent;
+	u16 ecc_page_threshold;
+	u32 padding[64 - 1];
+} __packed;
+
 struct amdgpu_ras_eeprom_control {
 	struct amdgpu_ras_eeprom_table_header tbl_hdr;
 
+	struct amdgpu_ras_eeprom_table_ras_info tbl_rai;
+
 	/* Base I2C EEPPROM 19-bit memory address,
 	 * where the table is located. For more information,
 	 * see top of amdgpu_eeprom.c.
@@ -58,11 +76,24 @@ struct amdgpu_ras_eeprom_control {
 	 * right after the header.
 	 */
 	u32 ras_header_offset;
+	u32 ras_info_offset;
 	u32 ras_record_offset;
 
 	/* Number of records in the table.
 	 */
 	u32 ras_num_recs;
+	u32 ras_num_recs_old;
+
+	/* the bad page number is ras_num_recs or
+	 * ras_num_recs * umc.retire_unit
+	 */
+	u32 ras_num_bad_pages;
+
+	/* Number of records store mca address */
+	u32 ras_num_mca_recs;
+
+	/* Number of records store physical address */
+	u32 ras_num_pa_recs;
 
 	/* First record index to read, 0-based.
 	 * Range is [0, num_recs-1]. This is
@@ -84,6 +115,8 @@ struct amdgpu_ras_eeprom_control {
 	/* Record channel info which occurred bad pages
 	 */
 	u32 bad_channel_bitmap;
+
+	bool is_eeprom_valid;
 };
 
 /*
@@ -111,8 +144,7 @@ struct eeprom_table_record {
 	unsigned char mcumc_id;
 } __packed;
 
-int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control,
-			   bool *exceed_err_limit);
+int amdgpu_ras_eeprom_init(struct amdgpu_ras_eeprom_control *control);
 
 int amdgpu_ras_eeprom_reset_table(struct amdgpu_ras_eeprom_control *control);
 
@@ -124,10 +156,43 @@ int amdgpu_ras_eeprom_read(struct amdgpu_ras_eeprom_control *control,
 int amdgpu_ras_eeprom_append(struct amdgpu_ras_eeprom_control *control,
 			     struct eeprom_table_record *records, const u32 num);
 
-uint32_t amdgpu_ras_eeprom_max_record_count(void);
+uint32_t amdgpu_ras_eeprom_max_record_count(struct amdgpu_ras_eeprom_control *control);
 
 void amdgpu_ras_debugfs_set_ret_size(struct amdgpu_ras_eeprom_control *control);
 
+int amdgpu_ras_eeprom_check(struct amdgpu_ras_eeprom_control *control);
+
+void amdgpu_ras_eeprom_check_and_recover(struct amdgpu_device *adev);
+
+bool amdgpu_ras_smu_eeprom_supported(struct amdgpu_device *adev);
+
+int amdgpu_ras_smu_get_table_version(struct amdgpu_device *adev,
+							uint32_t *table_version);
+
+int amdgpu_ras_smu_get_badpage_count(struct amdgpu_device *adev,
+								uint32_t *count, uint32_t timeout);
+
+int amdgpu_ras_smu_get_badpage_mca_addr(struct amdgpu_device *adev,
+								uint16_t index, uint64_t *mca_addr);
+
+int amdgpu_ras_smu_set_timestamp(struct amdgpu_device *adev,
+										uint64_t timestamp);
+
+int amdgpu_ras_smu_get_timestamp(struct amdgpu_device *adev,
+							uint16_t index, uint64_t *timestamp);
+
+int amdgpu_ras_smu_get_badpage_ipid(struct amdgpu_device *adev,
+								uint16_t index, uint64_t *ipid);
+
+int amdgpu_ras_smu_erase_ras_table(struct amdgpu_device *adev,
+									uint32_t *result);
+
+int amdgpu_ras_eeprom_read_idx(struct amdgpu_ras_eeprom_control *control,
+			struct eeprom_table_record *record, u32 rec_idx,
+			const u32 num);
+
+int amdgpu_ras_eeprom_update_record_num(struct amdgpu_ras_eeprom_control *control);
+
 extern const struct file_operations amdgpu_ras_debugfs_eeprom_size_ops;
 extern const struct file_operations amdgpu_ras_debugfs_eeprom_table_ops;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_res_cursor.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_res_cursor.h
index 5c4f93ee0c57..be2e56ce1355 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_res_cursor.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_res_cursor.h
@@ -90,6 +90,8 @@ static inline void amdgpu_res_first(struct ttm_resource *res,
 		cur->node = block;
 		break;
 	case TTM_PL_TT:
+	case AMDGPU_PL_DOORBELL:
+	case AMDGPU_PL_MMIO_REMAP:
 		node = to_ttm_range_mgr_node(res)->mm_nodes;
 		while (start >= node->size << PAGE_SHIFT)
 			start -= node++->size << PAGE_SHIFT;
@@ -111,7 +113,6 @@ fallback:
 	cur->remaining = size;
 	cur->node = NULL;
 	WARN_ON(res && start + size > res->size);
-	return;
 }
 
 /**
@@ -152,6 +153,8 @@ static inline void amdgpu_res_next(struct amdgpu_res_cursor *cur, uint64_t size)
 		cur->size = min(amdgpu_vram_mgr_block_size(block), cur->remaining);
 		break;
 	case TTM_PL_TT:
+	case AMDGPU_PL_DOORBELL:
+	case AMDGPU_PL_MMIO_REMAP:
 		node = cur->node;
 
 		cur->node = ++node;
@@ -163,4 +166,29 @@ static inline void amdgpu_res_next(struct amdgpu_res_cursor *cur, uint64_t size)
 	}
 }
 
+/**
+ * amdgpu_res_cleared - check if blocks are cleared
+ *
+ * @cur: the cursor to extract the block
+ *
+ * Check if the @cur block is cleared
+ */
+static inline bool amdgpu_res_cleared(struct amdgpu_res_cursor *cur)
+{
+	struct drm_buddy_block *block;
+
+	switch (cur->mem_type) {
+	case TTM_PL_VRAM:
+		block = cur->node;
+
+		if (!amdgpu_vram_mgr_is_cleared(block))
+			return false;
+		break;
+	default:
+		return false;
+	}
+
+	return true;
+}
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c
index 6437ead87e5f..28c4ad62f50e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.c
@@ -26,20 +26,165 @@
 #include "sienna_cichlid.h"
 #include "smu_v13_0_10.h"
 
-int amdgpu_reset_add_handler(struct amdgpu_reset_control *reset_ctl,
-			     struct amdgpu_reset_handler *handler)
+static int amdgpu_reset_xgmi_reset_on_init_suspend(struct amdgpu_device *adev)
 {
-	/* TODO: Check if handler exists? */
-	list_add_tail(&handler->handler_list, &reset_ctl->reset_handlers);
+	int i;
+
+	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
+		if (!adev->ip_blocks[i].status.valid)
+			continue;
+		if (!adev->ip_blocks[i].status.hw)
+			continue;
+		/* displays are handled in phase1 */
+		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE)
+			continue;
+
+		/* XXX handle errors */
+		amdgpu_ip_block_suspend(&adev->ip_blocks[i]);
+		adev->ip_blocks[i].status.hw = false;
+	}
+
+	/* VCN FW shared region is in frambuffer, there are some flags
+	 * initialized in that region during sw_init. Make sure the region is
+	 * backed up.
+	 */
+	amdgpu_vcn_save_vcpu_bo(adev);
+
 	return 0;
 }
 
+static int amdgpu_reset_xgmi_reset_on_init_prep_hwctxt(
+	struct amdgpu_reset_control *reset_ctl,
+	struct amdgpu_reset_context *reset_context)
+{
+	struct list_head *reset_device_list = reset_context->reset_device_list;
+	struct amdgpu_device *tmp_adev;
+	int r;
+
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
+		amdgpu_unregister_gpu_instance(tmp_adev);
+		r = amdgpu_reset_xgmi_reset_on_init_suspend(tmp_adev);
+		if (r) {
+			dev_err(tmp_adev->dev,
+				"xgmi reset on init: prepare for reset failed");
+			return r;
+		}
+	}
+
+	return r;
+}
+
+static int amdgpu_reset_xgmi_reset_on_init_restore_hwctxt(
+	struct amdgpu_reset_control *reset_ctl,
+	struct amdgpu_reset_context *reset_context)
+{
+	struct list_head *reset_device_list = reset_context->reset_device_list;
+	struct amdgpu_device *tmp_adev = NULL;
+	int r;
+
+	r = amdgpu_device_reinit_after_reset(reset_context);
+	if (r)
+		return r;
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
+		if (!tmp_adev->kfd.init_complete) {
+			kgd2kfd_init_zone_device(tmp_adev);
+			amdgpu_amdkfd_device_init(tmp_adev);
+			amdgpu_amdkfd_drm_client_create(tmp_adev);
+		}
+	}
+
+	return r;
+}
+
+static int amdgpu_reset_xgmi_reset_on_init_perform_reset(
+	struct amdgpu_reset_control *reset_ctl,
+	struct amdgpu_reset_context *reset_context)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)reset_ctl->handle;
+	struct list_head *reset_device_list = reset_context->reset_device_list;
+	struct amdgpu_device *tmp_adev = NULL;
+	int r;
+
+	dev_dbg(adev->dev, "xgmi roi - hw reset\n");
+
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
+		mutex_lock(&tmp_adev->reset_cntl->reset_lock);
+		tmp_adev->reset_cntl->active_reset =
+			amdgpu_asic_reset_method(adev);
+	}
+	r = 0;
+	/* Mode1 reset needs to be triggered on all devices together */
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
+		/* For XGMI run all resets in parallel to speed up the process */
+		if (!queue_work(system_unbound_wq, &tmp_adev->xgmi_reset_work))
+			r = -EALREADY;
+		if (r) {
+			dev_err(tmp_adev->dev,
+				"xgmi reset on init: reset failed with error, %d",
+				r);
+			break;
+		}
+	}
+
+	/* For XGMI wait for all resets to complete before proceed */
+	if (!r) {
+		list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
+			flush_work(&tmp_adev->xgmi_reset_work);
+			r = tmp_adev->asic_reset_res;
+			if (r)
+				break;
+		}
+	}
+
+	list_for_each_entry(tmp_adev, reset_device_list, reset_list) {
+		mutex_unlock(&tmp_adev->reset_cntl->reset_lock);
+		tmp_adev->reset_cntl->active_reset = AMD_RESET_METHOD_NONE;
+	}
+
+	return r;
+}
+
+int amdgpu_reset_do_xgmi_reset_on_init(
+	struct amdgpu_reset_context *reset_context)
+{
+	struct list_head *reset_device_list = reset_context->reset_device_list;
+	struct amdgpu_device *adev;
+	int r;
+
+	if (!reset_device_list || list_empty(reset_device_list) ||
+	    list_is_singular(reset_device_list))
+		return -EINVAL;
+
+	adev = list_first_entry(reset_device_list, struct amdgpu_device,
+				reset_list);
+	r = amdgpu_reset_prepare_hwcontext(adev, reset_context);
+	if (r)
+		return r;
+
+	r = amdgpu_reset_perform_reset(adev, reset_context);
+
+	return r;
+}
+
+struct amdgpu_reset_handler xgmi_reset_on_init_handler = {
+	.reset_method = AMD_RESET_METHOD_ON_INIT,
+	.prepare_env = NULL,
+	.prepare_hwcontext = amdgpu_reset_xgmi_reset_on_init_prep_hwctxt,
+	.perform_reset = amdgpu_reset_xgmi_reset_on_init_perform_reset,
+	.restore_hwcontext = amdgpu_reset_xgmi_reset_on_init_restore_hwctxt,
+	.restore_env = NULL,
+	.do_reset = NULL,
+};
+
 int amdgpu_reset_init(struct amdgpu_device *adev)
 {
 	int ret = 0;
 
-	switch (adev->ip_versions[MP1_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
 	case IP_VERSION(13, 0, 2):
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
 		ret = aldebaran_reset_init(adev);
 		break;
 	case IP_VERSION(11, 0, 7):
@@ -59,8 +204,11 @@ int amdgpu_reset_fini(struct amdgpu_device *adev)
 {
 	int ret = 0;
 
-	switch (adev->ip_versions[MP1_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
 	case IP_VERSION(13, 0, 2):
+	case IP_VERSION(13, 0, 6):
+	case IP_VERSION(13, 0, 12):
+	case IP_VERSION(13, 0, 14):
 		ret = aldebaran_reset_fini(adev);
 		break;
 	case IP_VERSION(11, 0, 7):
@@ -85,7 +233,7 @@ int amdgpu_reset_prepare_hwcontext(struct amdgpu_device *adev,
 		reset_handler = adev->reset_cntl->get_reset_handler(
 			adev->reset_cntl, reset_context);
 	if (!reset_handler)
-		return -ENOSYS;
+		return -EOPNOTSUPP;
 
 	return reset_handler->prepare_hwcontext(adev->reset_cntl,
 						reset_context);
@@ -101,7 +249,7 @@ int amdgpu_reset_perform_reset(struct amdgpu_device *adev,
 		reset_handler = adev->reset_cntl->get_reset_handler(
 			adev->reset_cntl, reset_context);
 	if (!reset_handler)
-		return -ENOSYS;
+		return -EOPNOTSUPP;
 
 	ret = reset_handler->perform_reset(adev->reset_cntl, reset_context);
 	if (ret)
@@ -165,5 +313,42 @@ void amdgpu_device_unlock_reset_domain(struct amdgpu_reset_domain *reset_domain)
 	up_write(&reset_domain->sem);
 }
 
+void amdgpu_reset_get_desc(struct amdgpu_reset_context *rst_ctxt, char *buf,
+			   size_t len)
+{
+	if (!buf || !len)
+		return;
 
+	switch (rst_ctxt->src) {
+	case AMDGPU_RESET_SRC_JOB:
+		if (rst_ctxt->job) {
+			snprintf(buf, len, "job hang on ring:%s",
+				 rst_ctxt->job->base.sched->name);
+		} else {
+			strscpy(buf, "job hang", len);
+		}
+		break;
+	case AMDGPU_RESET_SRC_RAS:
+		strscpy(buf, "RAS error", len);
+		break;
+	case AMDGPU_RESET_SRC_MES:
+		strscpy(buf, "MES hang", len);
+		break;
+	case AMDGPU_RESET_SRC_HWS:
+		strscpy(buf, "HWS hang", len);
+		break;
+	case AMDGPU_RESET_SRC_USER:
+		strscpy(buf, "user trigger", len);
+		break;
+	case AMDGPU_RESET_SRC_USERQ:
+		strscpy(buf, "user queue trigger", len);
+		break;
+	default:
+		strscpy(buf, "unknown", len);
+	}
+}
 
+bool amdgpu_reset_in_recovery(struct amdgpu_device *adev)
+{
+	return (adev->init_lvl->level == AMDGPU_INIT_LEVEL_RESET_RECOVERY);
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
index f4a501ff87d9..07b4d37f1db6 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_reset.h
@@ -26,11 +26,24 @@
 
 #include "amdgpu.h"
 
+#define AMDGPU_RESET_MAX_HANDLERS 5
+
 enum AMDGPU_RESET_FLAGS {
 
 	AMDGPU_NEED_FULL_RESET = 0,
 	AMDGPU_SKIP_HW_RESET = 1,
-	AMDGPU_RESET_FOR_DEVICE_REMOVE = 2,
+	AMDGPU_SKIP_COREDUMP = 2,
+	AMDGPU_HOST_FLR = 3,
+};
+
+enum AMDGPU_RESET_SRCS {
+	AMDGPU_RESET_SRC_UNKNOWN,
+	AMDGPU_RESET_SRC_JOB,
+	AMDGPU_RESET_SRC_RAS,
+	AMDGPU_RESET_SRC_MES,
+	AMDGPU_RESET_SRC_HWS,
+	AMDGPU_RESET_SRC_USER,
+	AMDGPU_RESET_SRC_USERQ,
 };
 
 struct amdgpu_reset_context {
@@ -40,11 +53,11 @@ struct amdgpu_reset_context {
 	struct amdgpu_hive_info *hive;
 	struct list_head *reset_device_list;
 	unsigned long flags;
+	enum AMDGPU_RESET_SRCS src;
 };
 
 struct amdgpu_reset_handler {
 	enum amd_reset_method reset_method;
-	struct list_head handler_list;
 	int (*prepare_env)(struct amdgpu_reset_control *reset_ctl,
 			   struct amdgpu_reset_context *context);
 	int (*prepare_hwcontext)(struct amdgpu_reset_control *reset_ctl,
@@ -63,7 +76,8 @@ struct amdgpu_reset_control {
 	void *handle;
 	struct work_struct reset_work;
 	struct mutex reset_lock;
-	struct list_head reset_handlers;
+	struct amdgpu_reset_handler *(
+		*reset_handlers)[AMDGPU_RESET_MAX_HANDLERS];
 	atomic_t in_reset;
 	enum amd_reset_method active_reset;
 	struct amdgpu_reset_handler *(*get_reset_handler)(
@@ -87,7 +101,6 @@ struct amdgpu_reset_domain {
 	atomic_t reset_res;
 };
 
-
 int amdgpu_reset_init(struct amdgpu_device *adev);
 int amdgpu_reset_fini(struct amdgpu_device *adev);
 
@@ -97,8 +110,10 @@ int amdgpu_reset_prepare_hwcontext(struct amdgpu_device *adev,
 int amdgpu_reset_perform_reset(struct amdgpu_device *adev,
 			       struct amdgpu_reset_context *reset_context);
 
-int amdgpu_reset_add_handler(struct amdgpu_reset_control *reset_ctl,
-			     struct amdgpu_reset_handler *handler);
+int amdgpu_reset_prepare_env(struct amdgpu_device *adev,
+			     struct amdgpu_reset_context *reset_context);
+int amdgpu_reset_restore_env(struct amdgpu_device *adev,
+			     struct amdgpu_reset_context *reset_context);
 
 struct amdgpu_reset_domain *amdgpu_reset_create_reset_domain(enum amdgpu_reset_domain_type type,
 							     char *wq_name);
@@ -122,8 +137,40 @@ static inline bool amdgpu_reset_domain_schedule(struct amdgpu_reset_domain *doma
 	return queue_work(domain->wq, work);
 }
 
+static inline bool amdgpu_reset_pending(struct amdgpu_reset_domain *domain)
+{
+	lockdep_assert_held(&domain->sem);
+	return rwsem_is_contended(&domain->sem);
+}
+
 void amdgpu_device_lock_reset_domain(struct amdgpu_reset_domain *reset_domain);
 
 void amdgpu_device_unlock_reset_domain(struct amdgpu_reset_domain *reset_domain);
 
+void amdgpu_reset_get_desc(struct amdgpu_reset_context *rst_ctxt, char *buf,
+			   size_t len);
+
+#define for_each_handler(i, handler, reset_ctl)                  \
+	for (i = 0; (i < AMDGPU_RESET_MAX_HANDLERS) &&           \
+		    (handler = (*reset_ctl->reset_handlers)[i]); \
+	     ++i)
+
+extern struct amdgpu_reset_handler xgmi_reset_on_init_handler;
+int amdgpu_reset_do_xgmi_reset_on_init(
+	struct amdgpu_reset_context *reset_context);
+
+bool amdgpu_reset_in_recovery(struct amdgpu_device *adev);
+
+static inline void amdgpu_reset_set_dpc_status(struct amdgpu_device *adev,
+					       bool status)
+{
+	adev->pcie_reset_ctx.occurs_dpc = status;
+	adev->no_hw_access = status;
+}
+
+static inline bool amdgpu_reset_in_dpc(struct amdgpu_device *adev)
+{
+	return adev->pcie_reset_ctx.occurs_dpc;
+}
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c
index dc474b809604..c596b6df2e2d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c
@@ -33,6 +33,7 @@
 
 #include <drm/amdgpu_drm.h>
 #include "amdgpu.h"
+#include "amdgpu_ras_mgr.h"
 #include "atom.h"
 
 /*
@@ -50,6 +51,26 @@
  */
 
 /**
+ * amdgpu_ring_max_ibs - Return max IBs that fit in a single submission.
+ *
+ * @type: ring type for which to return the limit.
+ */
+unsigned int amdgpu_ring_max_ibs(enum amdgpu_ring_type type)
+{
+	switch (type) {
+	case AMDGPU_RING_TYPE_GFX:
+		/* Need to keep at least 192 on GFX7+ for old radv. */
+		return 192;
+	case AMDGPU_RING_TYPE_COMPUTE:
+		return 125;
+	case AMDGPU_RING_TYPE_VCN_JPEG:
+		return 16;
+	default:
+		return 49;
+	}
+}
+
+/**
  * amdgpu_ring_alloc - allocate space on the ring buffer
  *
  * @ring: amdgpu_ring structure holding ring information
@@ -58,7 +79,7 @@
  * Allocate @ndw dwords in the ring buffer (all asics).
  * Returns 0 on success, error on failure.
  */
-int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned ndw)
+int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned int ndw)
 {
 	/* Align requested size with padding so unlock_commit can
 	 * pad safely */
@@ -79,6 +100,29 @@ int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned ndw)
 	return 0;
 }
 
+/**
+ * amdgpu_ring_alloc_reemit - allocate space on the ring buffer for reemit
+ *
+ * @ring: amdgpu_ring structure holding ring information
+ * @ndw: number of dwords to allocate in the ring buffer
+ *
+ * Allocate @ndw dwords in the ring buffer (all asics).
+ * doesn't check the max_dw limit as we may be reemitting
+ * several submissions.
+ */
+static void amdgpu_ring_alloc_reemit(struct amdgpu_ring *ring, unsigned int ndw)
+{
+	/* Align requested size with padding so unlock_commit can
+	 * pad safely */
+	ndw = (ndw + ring->funcs->align_mask) & ~ring->funcs->align_mask;
+
+	ring->count_dw = ndw;
+	ring->wptr_old = ring->wptr;
+
+	if (ring->funcs->begin_use)
+		ring->funcs->begin_use(ring);
+}
+
 /** amdgpu_ring_insert_nop - insert NOP packets
  *
  * @ring: amdgpu_ring structure holding ring information
@@ -88,10 +132,22 @@ int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned ndw)
  */
 void amdgpu_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
 {
-	int i;
+	uint32_t occupied, chunk1, chunk2;
+
+	occupied = ring->wptr & ring->buf_mask;
+	chunk1 = ring->buf_mask + 1 - occupied;
+	chunk1 = (chunk1 >= count) ? count : chunk1;
+	chunk2 = count - chunk1;
+
+	if (chunk1)
+		memset32(&ring->ring[occupied], ring->funcs->nop, chunk1);
+
+	if (chunk2)
+		memset32(ring->ring, ring->funcs->nop, chunk2);
 
-	for (i = 0; i < count; i++)
-		amdgpu_ring_write(ring, ring->funcs->nop);
+	ring->wptr += count;
+	ring->wptr &= ring->ptr_mask;
+	ring->count_dw -= count;
 }
 
 /**
@@ -104,8 +160,16 @@ void amdgpu_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
  */
 void amdgpu_ring_generic_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib)
 {
-	while (ib->length_dw & ring->funcs->align_mask)
-		ib->ptr[ib->length_dw++] = ring->funcs->nop;
+	u32 align_mask = ring->funcs->align_mask;
+	u32 count = ib->length_dw & align_mask;
+
+	if (count) {
+		count = align_mask + 1 - count;
+
+		memset32(&ib->ptr[ib->length_dw], ring->funcs->nop, count);
+
+		ib->length_dw += count;
+	}
 }
 
 /**
@@ -121,11 +185,16 @@ void amdgpu_ring_commit(struct amdgpu_ring *ring)
 {
 	uint32_t count;
 
+	if (ring->count_dw < 0)
+		DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n");
+
 	/* We pad to match fetch size */
 	count = ring->funcs->align_mask + 1 -
 		(ring->wptr & ring->funcs->align_mask);
-	count %= ring->funcs->align_mask + 1;
-	ring->funcs->insert_nop(ring, count);
+	count &= ring->funcs->align_mask;
+
+	if (count != 0)
+		ring->funcs->insert_nop(ring, count);
 
 	mb();
 	amdgpu_ring_set_wptr(ring);
@@ -150,14 +219,10 @@ void amdgpu_ring_undo(struct amdgpu_ring *ring)
 }
 
 #define amdgpu_ring_get_gpu_addr(ring, offset)				\
-	(ring->is_mes_queue ?						\
-	 (ring->mes_ctx->meta_data_gpu_addr + offset) :			\
-	 (ring->adev->wb.gpu_addr + offset * 4))
+	 (ring->adev->wb.gpu_addr + offset * 4)
 
 #define amdgpu_ring_get_cpu_addr(ring, offset)				\
-	(ring->is_mes_queue ?						\
-	 (void *)((uint8_t *)(ring->mes_ctx->meta_data_ptr) + offset) : \
-	 (&ring->adev->wb.wb[offset]))
+	 (&ring->adev->wb.wb[offset])
 
 /**
  * amdgpu_ring_init - init driver ring struct.
@@ -182,6 +247,7 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 	int sched_hw_submission = amdgpu_sched_hw_submission;
 	u32 *num_sched;
 	u32 hw_ip;
+	unsigned int max_ibs_dw;
 
 	/* Set the hw submission limit higher for KIQ because
 	 * it's used for a number of gfx/compute tasks by both
@@ -191,6 +257,8 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 	 */
 	if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ)
 		sched_hw_submission = max(sched_hw_submission, 256);
+	if (ring->funcs->type == AMDGPU_RING_TYPE_MES)
+		sched_hw_submission = 8;
 	else if (ring == &adev->sdma.instance[0].page)
 		sched_hw_submission = 256;
 
@@ -203,57 +271,42 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 		ring->sched_score = sched_score;
 		ring->vmid_wait = dma_fence_get_stub();
 
-		if (!ring->is_mes_queue) {
-			ring->idx = adev->num_rings++;
-			adev->rings[ring->idx] = ring;
-		}
+		ring->idx = adev->num_rings++;
+		adev->rings[ring->idx] = ring;
 
 		r = amdgpu_fence_driver_init_ring(ring);
 		if (r)
 			return r;
 	}
 
-	if (ring->is_mes_queue) {
-		ring->rptr_offs = amdgpu_mes_ctx_get_offs(ring,
-				AMDGPU_MES_CTX_RPTR_OFFS);
-		ring->wptr_offs = amdgpu_mes_ctx_get_offs(ring,
-				AMDGPU_MES_CTX_WPTR_OFFS);
-		ring->fence_offs = amdgpu_mes_ctx_get_offs(ring,
-				AMDGPU_MES_CTX_FENCE_OFFS);
-		ring->trail_fence_offs = amdgpu_mes_ctx_get_offs(ring,
-				AMDGPU_MES_CTX_TRAIL_FENCE_OFFS);
-		ring->cond_exe_offs = amdgpu_mes_ctx_get_offs(ring,
-				AMDGPU_MES_CTX_COND_EXE_OFFS);
-	} else {
-		r = amdgpu_device_wb_get(adev, &ring->rptr_offs);
-		if (r) {
-			dev_err(adev->dev, "(%d) ring rptr_offs wb alloc failed\n", r);
-			return r;
-		}
+	r = amdgpu_device_wb_get(adev, &ring->rptr_offs);
+	if (r) {
+		dev_err(adev->dev, "(%d) ring rptr_offs wb alloc failed\n", r);
+		return r;
+	}
 
-		r = amdgpu_device_wb_get(adev, &ring->wptr_offs);
-		if (r) {
-			dev_err(adev->dev, "(%d) ring wptr_offs wb alloc failed\n", r);
-			return r;
-		}
+	r = amdgpu_device_wb_get(adev, &ring->wptr_offs);
+	if (r) {
+		dev_err(adev->dev, "(%d) ring wptr_offs wb alloc failed\n", r);
+		return r;
+	}
 
-		r = amdgpu_device_wb_get(adev, &ring->fence_offs);
-		if (r) {
-			dev_err(adev->dev, "(%d) ring fence_offs wb alloc failed\n", r);
-			return r;
-		}
+	r = amdgpu_device_wb_get(adev, &ring->fence_offs);
+	if (r) {
+		dev_err(adev->dev, "(%d) ring fence_offs wb alloc failed\n", r);
+		return r;
+	}
 
-		r = amdgpu_device_wb_get(adev, &ring->trail_fence_offs);
-		if (r) {
-			dev_err(adev->dev, "(%d) ring trail_fence_offs wb alloc failed\n", r);
-			return r;
-		}
+	r = amdgpu_device_wb_get(adev, &ring->trail_fence_offs);
+	if (r) {
+		dev_err(adev->dev, "(%d) ring trail_fence_offs wb alloc failed\n", r);
+		return r;
+	}
 
-		r = amdgpu_device_wb_get(adev, &ring->cond_exe_offs);
-		if (r) {
-			dev_err(adev->dev, "(%d) ring cond_exec_polling wb alloc failed\n", r);
-			return r;
-		}
+	r = amdgpu_device_wb_get(adev, &ring->cond_exe_offs);
+	if (r) {
+		dev_err(adev->dev, "(%d) ring cond_exec_polling wb alloc failed\n", r);
+		return r;
 	}
 
 	ring->fence_gpu_addr =
@@ -284,38 +337,51 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 	/* always set cond_exec_polling to CONTINUE */
 	*ring->cond_exe_cpu_addr = 1;
 
-	r = amdgpu_fence_driver_start_ring(ring, irq_src, irq_type);
-	if (r) {
-		dev_err(adev->dev, "failed initializing fences (%d).\n", r);
-		return r;
-	}
+	if (ring->funcs->type != AMDGPU_RING_TYPE_CPER) {
+		r = amdgpu_fence_driver_start_ring(ring, irq_src, irq_type);
+		if (r) {
+			dev_err(adev->dev, "failed initializing fences (%d).\n", r);
+			return r;
+		}
 
-	ring->ring_size = roundup_pow_of_two(max_dw * 4 * sched_hw_submission);
+		max_ibs_dw = ring->funcs->emit_frame_size +
+			     amdgpu_ring_max_ibs(ring->funcs->type) * ring->funcs->emit_ib_size;
+		max_ibs_dw = (max_ibs_dw + ring->funcs->align_mask) & ~ring->funcs->align_mask;
+
+		if (WARN_ON(max_ibs_dw > max_dw))
+			max_dw = max_ibs_dw;
+
+		ring->ring_size = roundup_pow_of_two(max_dw * 4 * sched_hw_submission);
+	} else {
+		ring->ring_size = roundup_pow_of_two(max_dw * 4);
+		ring->count_dw = (ring->ring_size - 4) >> 2;
+		/* ring buffer is empty now */
+		ring->wptr = *ring->rptr_cpu_addr = 0;
+	}
 
 	ring->buf_mask = (ring->ring_size / 4) - 1;
 	ring->ptr_mask = ring->funcs->support_64bit_ptrs ?
 		0xffffffffffffffff : ring->buf_mask;
+	/*  Initialize cached_rptr to 0 */
+	ring->cached_rptr = 0;
 
-	/* Allocate ring buffer */
-	if (ring->is_mes_queue) {
-		int offset = 0;
-
-		BUG_ON(ring->ring_size > PAGE_SIZE*4);
-
-		offset = amdgpu_mes_ctx_get_offs(ring,
-					 AMDGPU_MES_CTX_RING_OFFS);
-		ring->gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		ring->ring = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-		amdgpu_ring_clear_ring(ring);
+	if (!ring->ring_backup) {
+		ring->ring_backup = kvzalloc(ring->ring_size, GFP_KERNEL);
+		if (!ring->ring_backup)
+			return -ENOMEM;
+	}
 
-	} else if (ring->ring_obj == NULL) {
-		r = amdgpu_bo_create_kernel(adev, ring->ring_size + ring->funcs->extra_dw, PAGE_SIZE,
+	/* Allocate ring buffer */
+	if (ring->ring_obj == NULL) {
+		r = amdgpu_bo_create_kernel(adev, ring->ring_size + ring->funcs->extra_bytes,
+					    PAGE_SIZE,
 					    AMDGPU_GEM_DOMAIN_GTT,
 					    &ring->ring_obj,
 					    &ring->gpu_addr,
 					    (void **)&ring->ring);
 		if (r) {
 			dev_err(adev->dev, "(%d) ring create failed\n", r);
+			kvfree(ring->ring_backup);
 			return r;
 		}
 		amdgpu_ring_clear_ring(ring);
@@ -324,7 +390,7 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 	ring->max_dw = max_dw;
 	ring->hw_prio = hw_prio;
 
-	if (!ring->no_scheduler) {
+	if (!ring->no_scheduler && ring->funcs->type < AMDGPU_HW_IP_NUM) {
 		hw_ip = ring->funcs->type;
 		num_sched = &adev->gpu_sched[hw_ip][hw_prio].num_scheds;
 		adev->gpu_sched[hw_ip][hw_prio].sched[(*num_sched)++] =
@@ -345,30 +411,26 @@ void amdgpu_ring_fini(struct amdgpu_ring *ring)
 {
 
 	/* Not to finish a ring which is not initialized */
-	if (!(ring->adev) ||
-	    (!ring->is_mes_queue && !(ring->adev->rings[ring->idx])))
+	if (!(ring->adev) || !(ring->adev->rings[ring->idx]))
 		return;
 
 	ring->sched.ready = false;
 
-	if (!ring->is_mes_queue) {
-		amdgpu_device_wb_free(ring->adev, ring->rptr_offs);
-		amdgpu_device_wb_free(ring->adev, ring->wptr_offs);
+	amdgpu_device_wb_free(ring->adev, ring->rptr_offs);
+	amdgpu_device_wb_free(ring->adev, ring->wptr_offs);
 
-		amdgpu_device_wb_free(ring->adev, ring->cond_exe_offs);
-		amdgpu_device_wb_free(ring->adev, ring->fence_offs);
+	amdgpu_device_wb_free(ring->adev, ring->cond_exe_offs);
+	amdgpu_device_wb_free(ring->adev, ring->fence_offs);
 
-		amdgpu_bo_free_kernel(&ring->ring_obj,
-				      &ring->gpu_addr,
-				      (void **)&ring->ring);
-	}
+	amdgpu_bo_free_kernel(&ring->ring_obj,
+			      &ring->gpu_addr,
+			      (void **)&ring->ring);
+	kvfree(ring->ring_backup);
+	ring->ring_backup = NULL;
 
 	dma_fence_put(ring->vmid_wait);
 	ring->vmid_wait = NULL;
 	ring->me = 0;
-
-	if (!ring->is_mes_queue)
-		ring->adev->rings[ring->idx] = NULL;
 }
 
 /**
@@ -403,17 +465,30 @@ void amdgpu_ring_emit_reg_write_reg_wait_helper(struct amdgpu_ring *ring,
 bool amdgpu_ring_soft_recovery(struct amdgpu_ring *ring, unsigned int vmid,
 			       struct dma_fence *fence)
 {
-	ktime_t deadline = ktime_add_us(ktime_get(), 10000);
+	unsigned long flags;
+	ktime_t deadline;
+	bool ret;
+
+	deadline = ktime_add_us(ktime_get(), 10000);
 
 	if (amdgpu_sriov_vf(ring->adev) || !ring->funcs->soft_recovery || !fence)
 		return false;
 
-	atomic_inc(&ring->adev->gpu_reset_counter);
+	spin_lock_irqsave(fence->lock, flags);
+	if (!dma_fence_is_signaled_locked(fence))
+		dma_fence_set_error(fence, -ENODATA);
+	spin_unlock_irqrestore(fence->lock, flags);
+
 	while (!dma_fence_is_signaled(fence) &&
 	       ktime_to_ns(ktime_sub(deadline, ktime_get())) > 0)
 		ring->funcs->soft_recovery(ring, vmid);
 
-	return dma_fence_is_signaled(fence);
+	ret = dma_fence_is_signaled(fence);
+	/* increment the counter only if soft reset worked */
+	if (ret)
+		atomic_inc(&ring->adev->gpu_reset_counter);
+
+	return ret;
 }
 
 /*
@@ -421,6 +496,66 @@ bool amdgpu_ring_soft_recovery(struct amdgpu_ring *ring, unsigned int vmid,
  */
 #if defined(CONFIG_DEBUG_FS)
 
+static ssize_t amdgpu_ras_cper_debugfs_read(struct file *f, char __user *buf,
+					    size_t size, loff_t *offset)
+{
+	const uint8_t ring_header_size = 12;
+	struct amdgpu_ring *ring = file_inode(f)->i_private;
+	struct ras_cmd_cper_snapshot_req *snapshot_req __free(kfree) =
+		kzalloc(sizeof(struct ras_cmd_cper_snapshot_req), GFP_KERNEL);
+	struct ras_cmd_cper_snapshot_rsp *snapshot_rsp __free(kfree) =
+		kzalloc(sizeof(struct ras_cmd_cper_snapshot_rsp), GFP_KERNEL);
+	struct ras_cmd_cper_record_req *record_req __free(kfree) =
+		kzalloc(sizeof(struct ras_cmd_cper_record_req), GFP_KERNEL);
+	struct ras_cmd_cper_record_rsp *record_rsp __free(kfree) =
+		kzalloc(sizeof(struct ras_cmd_cper_record_rsp), GFP_KERNEL);
+	uint8_t *ring_header __free(kfree) =
+		kzalloc(ring_header_size, GFP_KERNEL);
+	uint32_t total_cper_num;
+	uint64_t start_cper_id;
+	int r;
+
+	if (!snapshot_req || !snapshot_rsp || !record_req || !record_rsp ||
+	    !ring_header)
+		return -ENOMEM;
+
+	if (!(*offset)) {
+		/* Need at least 12 bytes for the header on the first read */
+		if (size < ring_header_size)
+			return -EINVAL;
+
+		if (copy_to_user(buf, ring_header, ring_header_size))
+			return -EFAULT;
+		buf += ring_header_size;
+		size -= ring_header_size;
+	}
+
+	r = amdgpu_ras_mgr_handle_ras_cmd(ring->adev,
+					  RAS_CMD__GET_CPER_SNAPSHOT,
+					  snapshot_req, sizeof(struct ras_cmd_cper_snapshot_req),
+					  snapshot_rsp, sizeof(struct ras_cmd_cper_snapshot_rsp));
+	if (r || !snapshot_rsp->total_cper_num)
+		return r;
+
+	start_cper_id = snapshot_rsp->start_cper_id;
+	total_cper_num = snapshot_rsp->total_cper_num;
+
+	record_req->buf_ptr = (uint64_t)(uintptr_t)buf;
+	record_req->buf_size = size;
+	record_req->cper_start_id = start_cper_id + *offset;
+	record_req->cper_num = total_cper_num;
+	r = amdgpu_ras_mgr_handle_ras_cmd(ring->adev, RAS_CMD__GET_CPER_RECORD,
+					  record_req, sizeof(struct ras_cmd_cper_record_req),
+					  record_rsp, sizeof(struct ras_cmd_cper_record_rsp));
+	if (r)
+		return r;
+
+	r = *offset ? record_rsp->real_data_size : record_rsp->real_data_size + ring_header_size;
+	(*offset) += record_rsp->real_cper_num;
+
+	return r;
+}
+
 /* Layout of file is 12 bytes consisting of
  * - rptr
  * - wptr
@@ -432,8 +567,13 @@ static ssize_t amdgpu_debugfs_ring_read(struct file *f, char __user *buf,
 					size_t size, loff_t *pos)
 {
 	struct amdgpu_ring *ring = file_inode(f)->i_private;
-	int r, i;
 	uint32_t value, result, early[3];
+	uint64_t p;
+	loff_t i;
+	int r;
+
+	if (ring->funcs->type == AMDGPU_RING_TYPE_CPER && amdgpu_uniras_enabled(ring->adev))
+		return amdgpu_ras_cper_debugfs_read(f, buf, size, pos);
 
 	if (*pos & 3 || size & 3)
 		return -EINVAL;
@@ -441,13 +581,18 @@ static ssize_t amdgpu_debugfs_ring_read(struct file *f, char __user *buf,
 	result = 0;
 
 	if (*pos < 12) {
+		if (ring->funcs->type == AMDGPU_RING_TYPE_CPER)
+			mutex_lock(&ring->adev->cper.ring_lock);
+
 		early[0] = amdgpu_ring_get_rptr(ring) & ring->buf_mask;
 		early[1] = amdgpu_ring_get_wptr(ring) & ring->buf_mask;
 		early[2] = ring->wptr & ring->buf_mask;
 		for (i = *pos / 4; i < 3 && size; i++) {
 			r = put_user(early[i], (uint32_t *)buf);
-			if (r)
-				return r;
+			if (r) {
+				result = r;
+				goto out;
+			}
 			buf += 4;
 			result += 4;
 			size -= 4;
@@ -455,29 +600,113 @@ static ssize_t amdgpu_debugfs_ring_read(struct file *f, char __user *buf,
 		}
 	}
 
-	while (size) {
-		if (*pos >= (ring->ring_size + 12))
-			return result;
+	if (ring->funcs->type != AMDGPU_RING_TYPE_CPER) {
+		while (size) {
+			if (*pos >= (ring->ring_size + 12))
+				return result;
 
-		value = ring->ring[(*pos - 12)/4];
-		r = put_user(value, (uint32_t *)buf);
-		if (r)
-			return r;
-		buf += 4;
-		result += 4;
-		size -= 4;
-		*pos += 4;
+			value = ring->ring[(*pos - 12)/4];
+			r = put_user(value, (uint32_t *)buf);
+			if (r)
+				return r;
+			buf += 4;
+			result += 4;
+			size -= 4;
+			*pos += 4;
+		}
+	} else {
+		p = early[0];
+		if (early[0] <= early[1])
+			size = (early[1] - early[0]);
+		else
+			size = ring->ring_size - (early[0] - early[1]);
+
+		while (size) {
+			if (p == early[1])
+				goto out;
+
+			value = ring->ring[p];
+			r = put_user(value, (uint32_t *)buf);
+			if (r) {
+				result = r;
+				goto out;
+			}
+
+			buf += 4;
+			result += 4;
+			size--;
+			p++;
+			p &= ring->ptr_mask;
+		}
 	}
 
+out:
+	if (ring->funcs->type == AMDGPU_RING_TYPE_CPER)
+		mutex_unlock(&ring->adev->cper.ring_lock);
+
 	return result;
 }
 
+static ssize_t amdgpu_debugfs_virt_ring_read(struct file *f, char __user *buf,
+	size_t size, loff_t *pos)
+{
+	struct amdgpu_ring *ring = file_inode(f)->i_private;
+
+	if (*pos & 3 || size & 3)
+		return -EINVAL;
+
+	if (ring->funcs->type == AMDGPU_RING_TYPE_CPER)
+		amdgpu_virt_req_ras_cper_dump(ring->adev, false);
+
+	return amdgpu_debugfs_ring_read(f, buf, size, pos);
+}
+
 static const struct file_operations amdgpu_debugfs_ring_fops = {
 	.owner = THIS_MODULE,
 	.read = amdgpu_debugfs_ring_read,
 	.llseek = default_llseek
 };
 
+static const struct file_operations amdgpu_debugfs_virt_ring_fops = {
+	.owner = THIS_MODULE,
+	.read = amdgpu_debugfs_virt_ring_read,
+	.llseek = default_llseek
+};
+
+static ssize_t amdgpu_debugfs_mqd_read(struct file *f, char __user *buf,
+				       size_t size, loff_t *pos)
+{
+	struct amdgpu_ring *ring = file_inode(f)->i_private;
+	ssize_t bytes = min_t(ssize_t, ring->mqd_size - *pos, size);
+	void *from = ((u8 *)ring->mqd_ptr) + *pos;
+
+	if (*pos > ring->mqd_size)
+		return 0;
+
+	if (copy_to_user(buf, from, bytes))
+		return -EFAULT;
+
+	*pos += bytes;
+	return bytes;
+}
+
+static const struct file_operations amdgpu_debugfs_mqd_fops = {
+	.owner = THIS_MODULE,
+	.read = amdgpu_debugfs_mqd_read,
+	.llseek = default_llseek
+};
+
+static int amdgpu_debugfs_ring_error(void *data, u64 val)
+{
+	struct amdgpu_ring *ring = data;
+
+	amdgpu_fence_driver_set_error(ring, val);
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE_SIGNED(amdgpu_debugfs_error_fops, NULL,
+				amdgpu_debugfs_ring_error, "%lld\n");
+
 #endif
 
 void amdgpu_debugfs_ring_init(struct amdgpu_device *adev,
@@ -489,9 +718,25 @@ void amdgpu_debugfs_ring_init(struct amdgpu_device *adev,
 	char name[32];
 
 	sprintf(name, "amdgpu_ring_%s", ring->name);
-	debugfs_create_file_size(name, S_IFREG | S_IRUGO, root, ring,
-				 &amdgpu_debugfs_ring_fops,
-				 ring->ring_size + 12);
+	if (amdgpu_sriov_vf(adev))
+		debugfs_create_file_size(name, S_IFREG | 0444, root, ring,
+					 &amdgpu_debugfs_virt_ring_fops,
+					 ring->ring_size + 12);
+	else
+		debugfs_create_file_size(name, S_IFREG | 0444, root, ring,
+					 &amdgpu_debugfs_ring_fops,
+					 ring->ring_size + 12);
+
+	if (ring->mqd_obj) {
+		sprintf(name, "amdgpu_mqd_%s", ring->name);
+		debugfs_create_file_size(name, S_IFREG | 0444, root, ring,
+					 &amdgpu_debugfs_mqd_fops,
+					 ring->mqd_size);
+	}
+
+	sprintf(name, "amdgpu_error_%s", ring->name);
+	debugfs_create_file(name, 0200, root, ring,
+			    &amdgpu_debugfs_error_fops);
 
 #endif
 }
@@ -519,6 +764,7 @@ int amdgpu_ring_test_helper(struct amdgpu_ring *ring)
 			      ring->name);
 
 	ring->sched.ready = !r;
+
 	return r;
 }
 
@@ -526,6 +772,10 @@ static void amdgpu_ring_to_mqd_prop(struct amdgpu_ring *ring,
 				    struct amdgpu_mqd_prop *prop)
 {
 	struct amdgpu_device *adev = ring->adev;
+	bool is_high_prio_compute = ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE &&
+				    amdgpu_gfx_is_high_priority_compute_queue(adev, ring);
+	bool is_high_prio_gfx = ring->funcs->type == AMDGPU_RING_TYPE_GFX &&
+				amdgpu_gfx_is_high_priority_graphics_queue(adev, ring);
 
 	memset(prop, 0, sizeof(*prop));
 
@@ -537,16 +787,15 @@ static void amdgpu_ring_to_mqd_prop(struct amdgpu_ring *ring,
 	prop->eop_gpu_addr = ring->eop_gpu_addr;
 	prop->use_doorbell = ring->use_doorbell;
 	prop->doorbell_index = ring->doorbell_index;
+	prop->kernel_queue = true;
 
 	/* map_queues packet doesn't need activate the queue,
 	 * so only kiq need set this field.
 	 */
 	prop->hqd_active = ring->funcs->type == AMDGPU_RING_TYPE_KIQ;
 
-	if ((ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE &&
-	     amdgpu_gfx_is_high_priority_compute_queue(adev, ring)) ||
-	    (ring->funcs->type == AMDGPU_RING_TYPE_GFX &&
-	     amdgpu_gfx_is_high_priority_graphics_queue(adev, ring))) {
+	prop->allow_tunneling = is_high_prio_compute;
+	if (is_high_prio_compute || is_high_prio_gfx) {
 		prop->hqd_pipe_priority = AMDGPU_GFX_PIPE_PRIO_HIGH;
 		prop->hqd_queue_priority = AMDGPU_GFX_QUEUE_PRIORITY_MAXIMUM;
 	}
@@ -581,3 +830,98 @@ void amdgpu_ring_ib_end(struct amdgpu_ring *ring)
 	if (ring->is_sw_ring)
 		amdgpu_sw_ring_ib_end(ring);
 }
+
+void amdgpu_ring_ib_on_emit_cntl(struct amdgpu_ring *ring)
+{
+	if (ring->is_sw_ring)
+		amdgpu_sw_ring_ib_mark_offset(ring, AMDGPU_MUX_OFFSET_TYPE_CONTROL);
+}
+
+void amdgpu_ring_ib_on_emit_ce(struct amdgpu_ring *ring)
+{
+	if (ring->is_sw_ring)
+		amdgpu_sw_ring_ib_mark_offset(ring, AMDGPU_MUX_OFFSET_TYPE_CE);
+}
+
+void amdgpu_ring_ib_on_emit_de(struct amdgpu_ring *ring)
+{
+	if (ring->is_sw_ring)
+		amdgpu_sw_ring_ib_mark_offset(ring, AMDGPU_MUX_OFFSET_TYPE_DE);
+}
+
+bool amdgpu_ring_sched_ready(struct amdgpu_ring *ring)
+{
+	if (!ring)
+		return false;
+
+	if (ring->no_scheduler || !drm_sched_wqueue_ready(&ring->sched))
+		return false;
+
+	return true;
+}
+
+void amdgpu_ring_reset_helper_begin(struct amdgpu_ring *ring,
+				    struct amdgpu_fence *guilty_fence)
+{
+	/* Stop the scheduler to prevent anybody else from touching the ring buffer. */
+	drm_sched_wqueue_stop(&ring->sched);
+	/* back up the non-guilty commands */
+	amdgpu_ring_backup_unprocessed_commands(ring, guilty_fence);
+}
+
+int amdgpu_ring_reset_helper_end(struct amdgpu_ring *ring,
+				 struct amdgpu_fence *guilty_fence)
+{
+	unsigned int i;
+	int r;
+
+	/* verify that the ring is functional */
+	r = amdgpu_ring_test_ring(ring);
+	if (r)
+		return r;
+
+	/* signal the guilty fence and set an error on all fences from the context */
+	if (guilty_fence)
+		amdgpu_fence_driver_guilty_force_completion(guilty_fence);
+	/* Re-emit the non-guilty commands */
+	if (ring->ring_backup_entries_to_copy) {
+		amdgpu_ring_alloc_reemit(ring, ring->ring_backup_entries_to_copy);
+		for (i = 0; i < ring->ring_backup_entries_to_copy; i++)
+			amdgpu_ring_write(ring, ring->ring_backup[i]);
+		amdgpu_ring_commit(ring);
+	}
+	/* Start the scheduler again */
+	drm_sched_wqueue_start(&ring->sched);
+	return 0;
+}
+
+bool amdgpu_ring_is_reset_type_supported(struct amdgpu_ring *ring,
+					 u32 reset_type)
+{
+	switch (ring->funcs->type) {
+	case AMDGPU_RING_TYPE_GFX:
+		if (ring->adev->gfx.gfx_supported_reset & reset_type)
+			return true;
+		break;
+	case AMDGPU_RING_TYPE_COMPUTE:
+		if (ring->adev->gfx.compute_supported_reset & reset_type)
+			return true;
+		break;
+	case AMDGPU_RING_TYPE_SDMA:
+		if (ring->adev->sdma.supported_reset & reset_type)
+			return true;
+		break;
+	case AMDGPU_RING_TYPE_VCN_DEC:
+	case AMDGPU_RING_TYPE_VCN_ENC:
+		if (ring->adev->vcn.supported_reset & reset_type)
+			return true;
+		break;
+	case AMDGPU_RING_TYPE_VCN_JPEG:
+		if (ring->adev->jpeg.supported_reset & reset_type)
+			return true;
+		break;
+	default:
+		break;
+	}
+	return false;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
index 018f36b10de8..7a27c6c4bb44 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h
@@ -37,13 +37,14 @@ struct amdgpu_job;
 struct amdgpu_vm;
 
 /* max number of rings */
-#define AMDGPU_MAX_RINGS		28
-#define AMDGPU_MAX_HWIP_RINGS		8
+#define AMDGPU_MAX_RINGS		149
+#define AMDGPU_MAX_HWIP_RINGS		64
 #define AMDGPU_MAX_GFX_RINGS		2
 #define AMDGPU_MAX_SW_GFX_RINGS         2
 #define AMDGPU_MAX_COMPUTE_RINGS	8
 #define AMDGPU_MAX_VCE_RINGS		3
 #define AMDGPU_MAX_UVD_ENC_RINGS	2
+#define AMDGPU_MAX_VPE_RINGS		2
 
 enum amdgpu_ring_priority_level {
 	AMDGPU_RING_PRIO_0,
@@ -77,8 +78,12 @@ enum amdgpu_ring_type {
 	AMDGPU_RING_TYPE_VCN_DEC	= AMDGPU_HW_IP_VCN_DEC,
 	AMDGPU_RING_TYPE_VCN_ENC	= AMDGPU_HW_IP_VCN_ENC,
 	AMDGPU_RING_TYPE_VCN_JPEG	= AMDGPU_HW_IP_VCN_JPEG,
+	AMDGPU_RING_TYPE_VPE		= AMDGPU_HW_IP_VPE,
 	AMDGPU_RING_TYPE_KIQ,
-	AMDGPU_RING_TYPE_MES
+	AMDGPU_RING_TYPE_MES,
+	AMDGPU_RING_TYPE_UMSCH_MM,
+	AMDGPU_RING_TYPE_CPER,
+	AMDGPU_RING_TYPE_MAX,
 };
 
 enum amdgpu_ib_pool_type {
@@ -110,10 +115,11 @@ struct amdgpu_sched {
  */
 struct amdgpu_fence_driver {
 	uint64_t			gpu_addr;
-	volatile uint32_t		*cpu_addr;
+	uint32_t			*cpu_addr;
 	/* sync_seq is protected by ring emission lock */
 	uint32_t			sync_seq;
 	atomic_t			last_seq;
+	u64				signalled_wptr;
 	bool				initialized;
 	struct amdgpu_irq_src		*irq_src;
 	unsigned			irq_type;
@@ -123,10 +129,33 @@ struct amdgpu_fence_driver {
 	struct dma_fence		**fences;
 };
 
+/*
+ * Fences mark an event in the GPUs pipeline and are used
+ * for GPU/CPU synchronization.  When the fence is written,
+ * it is expected that all buffers associated with that fence
+ * are no longer in use by the associated ring on the GPU and
+ * that the relevant GPU caches have been flushed.
+ */
+
+struct amdgpu_fence {
+	struct dma_fence base;
+
+	/* RB, DMA, etc. */
+	struct amdgpu_ring		*ring;
+	ktime_t				start_timestamp;
+
+	/* wptr for the fence for resets */
+	u64				wptr;
+	/* fence context for resets */
+	u64				context;
+};
+
 extern const struct drm_sched_backend_ops amdgpu_sched_ops;
 
-void amdgpu_fence_driver_clear_job_fences(struct amdgpu_ring *ring);
+void amdgpu_fence_driver_set_error(struct amdgpu_ring *ring, int error);
 void amdgpu_fence_driver_force_completion(struct amdgpu_ring *ring);
+void amdgpu_fence_driver_guilty_force_completion(struct amdgpu_fence *af);
+void amdgpu_fence_save_wptr(struct amdgpu_fence *af);
 
 int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring);
 int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring,
@@ -136,8 +165,8 @@ void amdgpu_fence_driver_hw_init(struct amdgpu_device *adev);
 void amdgpu_fence_driver_hw_fini(struct amdgpu_device *adev);
 int amdgpu_fence_driver_sw_init(struct amdgpu_device *adev);
 void amdgpu_fence_driver_sw_fini(struct amdgpu_device *adev);
-int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **fence, struct amdgpu_job *job,
-		      unsigned flags);
+int amdgpu_fence_emit(struct amdgpu_ring *ring, struct amdgpu_fence *af,
+		      unsigned int flags);
 int amdgpu_fence_emit_polling(struct amdgpu_ring *ring, uint32_t *s,
 			      uint32_t timeout);
 bool amdgpu_fence_process(struct amdgpu_ring *ring);
@@ -159,14 +188,40 @@ void amdgpu_fence_update_start_timestamp(struct amdgpu_ring *ring, uint32_t seq,
 
 /* provided by hw blocks that expose a ring buffer for commands */
 struct amdgpu_ring_funcs {
+	/**
+	 * @type:
+	 *
+	 * GFX, Compute, SDMA, UVD, VCE, VCN, VPE, KIQ, MES, UMSCH, and CPER
+	 * use ring buffers. The type field just identifies which component the
+	 * ring buffer is associated with.
+	 */
 	enum amdgpu_ring_type	type;
 	uint32_t		align_mask;
+
+	/**
+	 * @nop:
+	 *
+	 * Every block in the amdgpu has no-op instructions (e.g., GFX 10
+	 * uses PACKET3(PACKET3_NOP, 0x3FFF), VCN 5 uses VCN_ENC_CMD_NO_OP,
+	 * etc). This field receives the specific no-op for the component
+	 * that initializes the ring.
+	 */
 	u32			nop;
 	bool			support_64bit_ptrs;
 	bool			no_user_fence;
 	bool			secure_submission_supported;
-	unsigned		vmhub;
-	unsigned		extra_dw;
+
+	/**
+	 * @extra_bytes:
+	 *
+	 * Optional extra space in bytes that is added to the ring size
+	 * when allocating the BO that holds the contents of the ring.
+	 * This space isn't used for command submission to the ring,
+	 * but is just there to satisfy some hardware requirements or
+	 * implement workarounds. It's up to the implementation of each
+	 * specific ring to initialize this space.
+	 */
+	unsigned		extra_bytes;
 
 	/* ring read/write ptr handling */
 	u64 (*get_rptr)(struct amdgpu_ring *ring);
@@ -206,13 +261,14 @@ struct amdgpu_ring_funcs {
 	void (*insert_end)(struct amdgpu_ring *ring);
 	/* pad the indirect buffer to the necessary number of dw */
 	void (*pad_ib)(struct amdgpu_ring *ring, struct amdgpu_ib *ib);
-	unsigned (*init_cond_exec)(struct amdgpu_ring *ring);
-	void (*patch_cond_exec)(struct amdgpu_ring *ring, unsigned offset);
+	unsigned (*init_cond_exec)(struct amdgpu_ring *ring, uint64_t addr);
 	/* note usage for clock and power gating */
 	void (*begin_use)(struct amdgpu_ring *ring);
 	void (*end_use)(struct amdgpu_ring *ring);
 	void (*emit_switch_buffer) (struct amdgpu_ring *ring);
 	void (*emit_cntxcntl) (struct amdgpu_ring *ring, uint32_t flags);
+	void (*emit_gfx_shadow)(struct amdgpu_ring *ring, u64 shadow_va, u64 csa_va,
+				u64 gds_va, bool init_shadow, int vmid);
 	void (*emit_rreg)(struct amdgpu_ring *ring, uint32_t reg,
 			  uint32_t reg_val_offs);
 	void (*emit_wreg)(struct amdgpu_ring *ring, uint32_t reg, uint32_t val);
@@ -228,8 +284,17 @@ struct amdgpu_ring_funcs {
 	int (*preempt_ib)(struct amdgpu_ring *ring);
 	void (*emit_mem_sync)(struct amdgpu_ring *ring);
 	void (*emit_wave_limit)(struct amdgpu_ring *ring, bool enable);
+	void (*patch_cntl)(struct amdgpu_ring *ring, unsigned offset);
+	void (*patch_ce)(struct amdgpu_ring *ring, unsigned offset);
+	void (*patch_de)(struct amdgpu_ring *ring, unsigned offset);
+	int (*reset)(struct amdgpu_ring *ring, unsigned int vmid,
+		     struct amdgpu_fence *timedout_fence);
+	void (*emit_cleaner_shader)(struct amdgpu_ring *ring);
 };
 
+/**
+ * amdgpu_ring - Holds ring information
+ */
 struct amdgpu_ring {
 	struct amdgpu_device		*adev;
 	const struct amdgpu_ring_funcs	*funcs;
@@ -237,60 +302,122 @@ struct amdgpu_ring {
 	struct drm_gpu_scheduler	sched;
 
 	struct amdgpu_bo	*ring_obj;
-	volatile uint32_t	*ring;
+	uint32_t		*ring;
+	/* backups for resets */
+	uint32_t		*ring_backup;
+	unsigned int		ring_backup_entries_to_copy;
 	unsigned		rptr_offs;
 	u64			rptr_gpu_addr;
-	volatile u32		*rptr_cpu_addr;
+	u32			*rptr_cpu_addr;
+
+	/**
+	 * @wptr:
+	 *
+	 * This is part of the Ring buffer implementation and represents the
+	 * write pointer. The wptr determines where the host has written.
+	 */
 	u64			wptr;
+
+	/**
+	 * @wptr_old:
+	 *
+	 * Before update wptr with the new value, usually the old value is
+	 * stored in the wptr_old.
+	 */
 	u64			wptr_old;
 	unsigned		ring_size;
+
+	/**
+	 * @max_dw:
+	 *
+	 * Maximum number of DWords for ring allocation. This information is
+	 * provided at the ring initialization time, and each IP block can
+	 * specify a specific value. Check places that invoke
+	 * amdgpu_ring_init() to see the maximum size per block.
+	 */
 	unsigned		max_dw;
+
+	/**
+	 * @count_dw:
+	 *
+	 * This value starts with the maximum amount of DWords supported by the
+	 * ring. This value is updated based on the ring manipulation.
+	 */
 	int			count_dw;
 	uint64_t		gpu_addr;
+
+	/**
+	 * @ptr_mask:
+	 *
+	 * Some IPs provide support for 64-bit pointers and others for 32-bit
+	 * only; this behavior is component-specific and defined by the field
+	 * support_64bit_ptr. If the IP block supports 64-bits, the mask
+	 * 0xffffffffffffffff is set; otherwise, this value assumes buf_mask.
+	 * Notice that this field is used to keep wptr under a valid range.
+	 */
 	uint64_t		ptr_mask;
+
+	/**
+	 * @buf_mask:
+	 *
+	 * Buffer mask is a value used to keep wptr count under its
+	 * thresholding. Buffer mask initialized during the ring buffer
+	 * initialization time, and it is defined as (ring_size / 4) -1.
+	 */
 	uint32_t		buf_mask;
 	u32			idx;
+	u32			xcc_id;
+	u32			xcp_id;
 	u32			me;
 	u32			pipe;
 	u32			queue;
 	struct amdgpu_bo	*mqd_obj;
 	uint64_t                mqd_gpu_addr;
 	void                    *mqd_ptr;
+	unsigned                mqd_size;
 	uint64_t                eop_gpu_addr;
 	u32			doorbell_index;
 	bool			use_doorbell;
 	bool			use_pollmem;
 	unsigned		wptr_offs;
 	u64			wptr_gpu_addr;
-	volatile u32		*wptr_cpu_addr;
+
+	/**
+	 * @wptr_cpu_addr:
+	 *
+	 * This is the CPU address pointer in the writeback slot. This is used
+	 * to commit changes to the GPU.
+	 */
+	u32			*wptr_cpu_addr;
 	unsigned		fence_offs;
 	u64			fence_gpu_addr;
-	volatile u32		*fence_cpu_addr;
+	u32			*fence_cpu_addr;
 	uint64_t		current_ctx;
 	char			name[16];
 	u32                     trail_seq;
 	unsigned		trail_fence_offs;
 	u64			trail_fence_gpu_addr;
-	volatile u32		*trail_fence_cpu_addr;
+	u32			*trail_fence_cpu_addr;
 	unsigned		cond_exe_offs;
 	u64			cond_exe_gpu_addr;
-	volatile u32		*cond_exe_cpu_addr;
+	u32			*cond_exe_cpu_addr;
+	unsigned int		set_q_mode_offs;
+	u32			*set_q_mode_ptr;
+	u64			set_q_mode_token;
+	unsigned		vm_hub;
 	unsigned		vm_inv_eng;
 	struct dma_fence	*vmid_wait;
 	bool			has_compute_vm_bug;
 	bool			no_scheduler;
+	bool			no_user_submission;
 	int			hw_prio;
 	unsigned 		num_hw_submission;
 	atomic_t		*sched_score;
 
-	/* used for mes */
-	bool			is_mes_queue;
-	uint32_t		hw_queue_id;
-	struct amdgpu_mes_ctx_data *mes_ctx;
-
 	bool            is_sw_ring;
 	unsigned int    entry_index;
-
+	/* store the cached rptr to restore after reset */
+	uint64_t cached_rptr;
 };
 
 #define amdgpu_ring_parse_cs(r, p, job, ib) ((r)->funcs->parse_cs((p), (job), (ib)))
@@ -308,19 +435,27 @@ struct amdgpu_ring {
 #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r))
 #define amdgpu_ring_emit_switch_buffer(r) (r)->funcs->emit_switch_buffer((r))
 #define amdgpu_ring_emit_cntxcntl(r, d) (r)->funcs->emit_cntxcntl((r), (d))
+#define amdgpu_ring_emit_gfx_shadow(r, s, c, g, i, v) ((r)->funcs->emit_gfx_shadow((r), (s), (c), (g), (i), (v)))
 #define amdgpu_ring_emit_rreg(r, d, o) (r)->funcs->emit_rreg((r), (d), (o))
 #define amdgpu_ring_emit_wreg(r, d, v) (r)->funcs->emit_wreg((r), (d), (v))
 #define amdgpu_ring_emit_reg_wait(r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m))
 #define amdgpu_ring_emit_reg_write_reg_wait(r, d0, d1, v, m) (r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m))
 #define amdgpu_ring_emit_frame_cntl(r, b, s) (r)->funcs->emit_frame_cntl((r), (b), (s))
 #define amdgpu_ring_pad_ib(r, ib) ((r)->funcs->pad_ib((r), (ib)))
-#define amdgpu_ring_init_cond_exec(r) (r)->funcs->init_cond_exec((r))
-#define amdgpu_ring_patch_cond_exec(r,o) (r)->funcs->patch_cond_exec((r),(o))
+#define amdgpu_ring_init_cond_exec(r, a) (r)->funcs->init_cond_exec((r), (a))
 #define amdgpu_ring_preempt_ib(r) (r)->funcs->preempt_ib(r)
+#define amdgpu_ring_patch_cntl(r, o) ((r)->funcs->patch_cntl((r), (o)))
+#define amdgpu_ring_patch_ce(r, o) ((r)->funcs->patch_ce((r), (o)))
+#define amdgpu_ring_patch_de(r, o) ((r)->funcs->patch_de((r), (o)))
+#define amdgpu_ring_reset(r, v, f) (r)->funcs->reset((r), (v), (f))
 
+unsigned int amdgpu_ring_max_ibs(enum amdgpu_ring_type type);
 int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned ndw);
 void amdgpu_ring_ib_begin(struct amdgpu_ring *ring);
 void amdgpu_ring_ib_end(struct amdgpu_ring *ring);
+void amdgpu_ring_ib_on_emit_cntl(struct amdgpu_ring *ring);
+void amdgpu_ring_ib_on_emit_ce(struct amdgpu_ring *ring);
+void amdgpu_ring_ib_on_emit_de(struct amdgpu_ring *ring);
 
 void amdgpu_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count);
 void amdgpu_ring_generic_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib);
@@ -345,16 +480,11 @@ static inline void amdgpu_ring_set_preempt_cond_exec(struct amdgpu_ring *ring,
 
 static inline void amdgpu_ring_clear_ring(struct amdgpu_ring *ring)
 {
-	int i = 0;
-	while (i <= ring->buf_mask)
-		ring->ring[i++] = ring->funcs->nop;
-
+	memset32(ring->ring, ring->funcs->nop, ring->buf_mask + 1);
 }
 
 static inline void amdgpu_ring_write(struct amdgpu_ring *ring, uint32_t v)
 {
-	if (ring->count_dw <= 0)
-		DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n");
 	ring->ring[ring->wptr++ & ring->buf_mask] = v;
 	ring->wptr &= ring->ptr_mask;
 	ring->count_dw--;
@@ -364,26 +494,20 @@ static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring,
 					      void *src, int count_dw)
 {
 	unsigned occupied, chunk1, chunk2;
-	void *dst;
-
-	if (unlikely(ring->count_dw < count_dw))
-		DRM_ERROR("amdgpu: writing more dwords to the ring than expected!\n");
 
 	occupied = ring->wptr & ring->buf_mask;
-	dst = (void *)&ring->ring[occupied];
 	chunk1 = ring->buf_mask + 1 - occupied;
-	chunk1 = (chunk1 >= count_dw) ? count_dw: chunk1;
+	chunk1 = (chunk1 >= count_dw) ? count_dw : chunk1;
 	chunk2 = count_dw - chunk1;
 	chunk1 <<= 2;
 	chunk2 <<= 2;
 
 	if (chunk1)
-		memcpy(dst, src, chunk1);
+		memcpy(&ring->ring[occupied], src, chunk1);
 
 	if (chunk2) {
 		src += chunk1;
-		dst = (void *)ring->ring;
-		memcpy(dst, src, chunk2);
+		memcpy(ring->ring, src, chunk2);
 	}
 
 	ring->wptr += count_dw;
@@ -391,14 +515,29 @@ static inline void amdgpu_ring_write_multiple(struct amdgpu_ring *ring,
 	ring->count_dw -= count_dw;
 }
 
-#define amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset)			\
-	(ring->is_mes_queue && ring->mes_ctx ?				\
-	 (ring->mes_ctx->meta_data_gpu_addr + offset) : 0)
+/**
+ * amdgpu_ring_patch_cond_exec - patch dw count of conditional execute
+ * @ring: amdgpu_ring structure
+ * @offset: offset returned by amdgpu_ring_init_cond_exec
+ *
+ * Calculate the dw count and patch it into a cond_exec command.
+ */
+static inline void amdgpu_ring_patch_cond_exec(struct amdgpu_ring *ring,
+					       unsigned int offset)
+{
+	unsigned cur;
+
+	if (!ring->funcs->init_cond_exec)
+		return;
 
-#define amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset)			\
-	(ring->is_mes_queue && ring->mes_ctx ?				\
-	 (void *)((uint8_t *)(ring->mes_ctx->meta_data_ptr) + offset) : \
-	 NULL)
+	WARN_ON(offset > ring->buf_mask);
+	WARN_ON(ring->ring[offset] != 0);
+
+	cur = (ring->wptr - 1) & ring->buf_mask;
+	if (cur < offset)
+		cur += ring->ring_size >> 2;
+	ring->ring[offset] = cur - offset;
+}
 
 int amdgpu_ring_test_helper(struct amdgpu_ring *ring);
 
@@ -422,13 +561,20 @@ int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		  unsigned size,
 		  enum amdgpu_ib_pool_type pool,
 		  struct amdgpu_ib *ib);
-void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib,
-		    struct dma_fence *f);
+void amdgpu_ib_free(struct amdgpu_ib *ib, struct dma_fence *f);
 int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
 		       struct amdgpu_ib *ibs, struct amdgpu_job *job,
 		       struct dma_fence **f);
 int amdgpu_ib_pool_init(struct amdgpu_device *adev);
 void amdgpu_ib_pool_fini(struct amdgpu_device *adev);
 int amdgpu_ib_ring_tests(struct amdgpu_device *adev);
-
+bool amdgpu_ring_sched_ready(struct amdgpu_ring *ring);
+void amdgpu_ring_backup_unprocessed_commands(struct amdgpu_ring *ring,
+					     struct amdgpu_fence *guilty_fence);
+void amdgpu_ring_reset_helper_begin(struct amdgpu_ring *ring,
+				    struct amdgpu_fence *guilty_fence);
+int amdgpu_ring_reset_helper_end(struct amdgpu_ring *ring,
+				 struct amdgpu_fence *guilty_fence);
+bool amdgpu_ring_is_reset_type_supported(struct amdgpu_ring *ring,
+					 u32 reset_type);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.c
index 62079f0e3ee8..7e7d6c3865bc 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.c
@@ -105,6 +105,16 @@ static void amdgpu_mux_resubmit_chunks(struct amdgpu_ring_mux *mux)
 				amdgpu_fence_update_start_timestamp(e->ring,
 								    chunk->sync_seq,
 								    ktime_get());
+				if (chunk->sync_seq ==
+					le32_to_cpu(*(e->ring->fence_drv.cpu_addr + 2))) {
+					if (chunk->cntl_offset <= e->ring->buf_mask)
+						amdgpu_ring_patch_cntl(e->ring,
+								       chunk->cntl_offset);
+					if (chunk->ce_offset <= e->ring->buf_mask)
+						amdgpu_ring_patch_ce(e->ring, chunk->ce_offset);
+					if (chunk->de_offset <= e->ring->buf_mask)
+						amdgpu_ring_patch_de(e->ring, chunk->de_offset);
+				}
 				amdgpu_ring_mux_copy_pkt_from_sw_ring(mux, e->ring,
 								      chunk->start,
 								      chunk->end);
@@ -114,7 +124,7 @@ static void amdgpu_mux_resubmit_chunks(struct amdgpu_ring_mux *mux)
 		}
 	}
 
-	del_timer(&mux->resubmit_timer);
+	timer_delete(&mux->resubmit_timer);
 	mux->s_resubmit = false;
 }
 
@@ -125,7 +135,8 @@ static void amdgpu_ring_mux_schedule_resubmit(struct amdgpu_ring_mux *mux)
 
 static void amdgpu_mux_resubmit_fallback(struct timer_list *t)
 {
-	struct amdgpu_ring_mux *mux = from_timer(mux, t, resubmit_timer);
+	struct amdgpu_ring_mux *mux = timer_container_of(mux, t,
+							 resubmit_timer);
 
 	if (!spin_trylock(&mux->lock)) {
 		amdgpu_ring_mux_schedule_resubmit(mux);
@@ -149,9 +160,7 @@ int amdgpu_ring_mux_init(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring,
 	mux->ring_entry_size = entry_size;
 	mux->s_resubmit = false;
 
-	amdgpu_mux_chunk_slab = kmem_cache_create("amdgpu_mux_chunk",
-						  sizeof(struct amdgpu_mux_chunk), 0,
-						  SLAB_HWCACHE_ALIGN, NULL);
+	amdgpu_mux_chunk_slab = KMEM_CACHE(amdgpu_mux_chunk, SLAB_HWCACHE_ALIGN);
 	if (!amdgpu_mux_chunk_slab) {
 		DRM_ERROR("create amdgpu_mux_chunk cache failed\n");
 		return -ENOMEM;
@@ -387,7 +396,7 @@ void amdgpu_sw_ring_ib_begin(struct amdgpu_ring *ring)
 	struct amdgpu_ring_mux *mux = &adev->gfx.muxer;
 
 	WARN_ON(!ring->is_sw_ring);
-	if (ring->hw_prio > AMDGPU_RING_PRIO_DEFAULT) {
+	if (adev->gfx.mcbp && ring->hw_prio > AMDGPU_RING_PRIO_DEFAULT) {
 		if (amdgpu_mcbp_scan(mux) > 0)
 			amdgpu_mcbp_trigger_preempt(mux);
 		return;
@@ -402,11 +411,25 @@ void amdgpu_sw_ring_ib_end(struct amdgpu_ring *ring)
 	struct amdgpu_ring_mux *mux = &adev->gfx.muxer;
 
 	WARN_ON(!ring->is_sw_ring);
-	if (ring->hw_prio > AMDGPU_RING_PRIO_DEFAULT)
+	if (adev->gfx.mcbp && ring->hw_prio > AMDGPU_RING_PRIO_DEFAULT)
 		return;
 	amdgpu_ring_mux_end_ib(mux, ring);
 }
 
+void amdgpu_sw_ring_ib_mark_offset(struct amdgpu_ring *ring, enum amdgpu_ring_mux_offset_type type)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_ring_mux *mux = &adev->gfx.muxer;
+	unsigned offset;
+
+	if (ring->hw_prio > AMDGPU_RING_PRIO_DEFAULT)
+		return;
+
+	offset = ring->wptr & ring->buf_mask;
+
+	amdgpu_ring_mux_ib_mark_offset(mux, ring, offset, type);
+}
+
 void amdgpu_ring_mux_start_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring)
 {
 	struct amdgpu_mux_entry *e;
@@ -429,6 +452,10 @@ void amdgpu_ring_mux_start_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *r
 	}
 
 	chunk->start = ring->wptr;
+	/* the initialized value used to check if they are set by the ib submission*/
+	chunk->cntl_offset = ring->buf_mask + 1;
+	chunk->de_offset = ring->buf_mask + 1;
+	chunk->ce_offset = ring->buf_mask + 1;
 	list_add_tail(&chunk->entry, &e->list);
 }
 
@@ -454,6 +481,41 @@ static void scan_and_remove_signaled_chunk(struct amdgpu_ring_mux *mux, struct a
 	}
 }
 
+void amdgpu_ring_mux_ib_mark_offset(struct amdgpu_ring_mux *mux,
+				    struct amdgpu_ring *ring, u64 offset,
+				    enum amdgpu_ring_mux_offset_type type)
+{
+	struct amdgpu_mux_entry *e;
+	struct amdgpu_mux_chunk *chunk;
+
+	e = amdgpu_ring_mux_sw_entry(mux, ring);
+	if (!e) {
+		DRM_ERROR("cannot find entry!\n");
+		return;
+	}
+
+	chunk = list_last_entry(&e->list, struct amdgpu_mux_chunk, entry);
+	if (!chunk) {
+		DRM_ERROR("cannot find chunk!\n");
+		return;
+	}
+
+	switch (type) {
+	case AMDGPU_MUX_OFFSET_TYPE_CONTROL:
+		chunk->cntl_offset = offset;
+		break;
+	case AMDGPU_MUX_OFFSET_TYPE_DE:
+		chunk->de_offset = offset;
+		break;
+	case AMDGPU_MUX_OFFSET_TYPE_CE:
+		chunk->ce_offset = offset;
+		break;
+	default:
+		DRM_ERROR("invalid type (%d)\n", type);
+		break;
+	}
+}
+
 void amdgpu_ring_mux_end_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring)
 {
 	struct amdgpu_mux_entry *e;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.h
index 4be45fc14954..d3186b570b82 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring_mux.h
@@ -50,6 +50,21 @@ struct amdgpu_mux_entry {
 	struct list_head        list;
 };
 
+enum amdgpu_ring_mux_offset_type {
+	AMDGPU_MUX_OFFSET_TYPE_CONTROL,
+	AMDGPU_MUX_OFFSET_TYPE_DE,
+	AMDGPU_MUX_OFFSET_TYPE_CE,
+};
+
+enum ib_complete_status {
+	/* IB not started/reset value, default value. */
+	IB_COMPLETION_STATUS_DEFAULT = 0,
+	/* IB preempted, started but not completed. */
+	IB_COMPLETION_STATUS_PREEMPTED = 1,
+	/* IB completed. */
+	IB_COMPLETION_STATUS_COMPLETED = 2,
+};
+
 struct amdgpu_ring_mux {
 	struct amdgpu_ring      *real_ring;
 
@@ -72,12 +87,18 @@ struct amdgpu_ring_mux {
  * @sync_seq: the fence seqno related with the saved IB.
  * @start:- start location on the software ring.
  * @end:- end location on the software ring.
+ * @control_offset:- the PRE_RESUME bit position used for resubmission.
+ * @de_offset:- the anchor in write_data for de meta of resubmission.
+ * @ce_offset:- the anchor in write_data for ce meta of resubmission.
  */
 struct amdgpu_mux_chunk {
 	struct list_head        entry;
 	uint32_t                sync_seq;
 	u64                     start;
 	u64                     end;
+	u64                     cntl_offset;
+	u64                     de_offset;
+	u64                     ce_offset;
 };
 
 int amdgpu_ring_mux_init(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring,
@@ -89,6 +110,8 @@ u64 amdgpu_ring_mux_get_wptr(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ri
 u64 amdgpu_ring_mux_get_rptr(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring);
 void amdgpu_ring_mux_start_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring);
 void amdgpu_ring_mux_end_ib(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring);
+void amdgpu_ring_mux_ib_mark_offset(struct amdgpu_ring_mux *mux, struct amdgpu_ring *ring,
+				    u64 offset, enum amdgpu_ring_mux_offset_type type);
 bool amdgpu_mcbp_handle_trailing_fence_irq(struct amdgpu_ring_mux *mux);
 
 u64 amdgpu_sw_ring_get_rptr_gfx(struct amdgpu_ring *ring);
@@ -97,6 +120,7 @@ void amdgpu_sw_ring_set_wptr_gfx(struct amdgpu_ring *ring);
 void amdgpu_sw_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count);
 void amdgpu_sw_ring_ib_begin(struct amdgpu_ring *ring);
 void amdgpu_sw_ring_ib_end(struct amdgpu_ring *ring);
+void amdgpu_sw_ring_ib_mark_offset(struct amdgpu_ring *ring, enum amdgpu_ring_mux_offset_type type);
 const char *amdgpu_sw_ring_name(int idx);
 unsigned int amdgpu_sw_ring_priority(int idx);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c
index 85fb730d9fc8..5aa830a02d80 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.c
@@ -31,12 +31,13 @@
  * amdgpu_gfx_rlc_enter_safe_mode - Set RLC into safe mode
  *
  * @adev: amdgpu_device pointer
+ * @xcc_id: xcc accelerated compute core id
  *
  * Set RLC enter into safe mode if RLC is enabled and haven't in safe mode.
  */
-void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev)
+void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
-	if (adev->gfx.rlc.in_safe_mode)
+	if (adev->gfx.rlc.in_safe_mode[xcc_id])
 		return;
 
 	/* if RLC is not enabled, do nothing */
@@ -46,8 +47,8 @@ void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev)
 	if (adev->cg_flags &
 	    (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG |
 	     AMD_CG_SUPPORT_GFX_3D_CGCG)) {
-		adev->gfx.rlc.funcs->set_safe_mode(adev);
-		adev->gfx.rlc.in_safe_mode = true;
+		adev->gfx.rlc.funcs->set_safe_mode(adev, xcc_id);
+		adev->gfx.rlc.in_safe_mode[xcc_id] = true;
 	}
 }
 
@@ -55,12 +56,13 @@ void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev)
  * amdgpu_gfx_rlc_exit_safe_mode - Set RLC out of safe mode
  *
  * @adev: amdgpu_device pointer
+ * @xcc_id: xcc accelerated compute core id
  *
  * Set RLC exit safe mode if RLC is enabled and have entered into safe mode.
  */
-void amdgpu_gfx_rlc_exit_safe_mode(struct amdgpu_device *adev)
+void amdgpu_gfx_rlc_exit_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
-	if (!(adev->gfx.rlc.in_safe_mode))
+	if (!(adev->gfx.rlc.in_safe_mode[xcc_id]))
 		return;
 
 	/* if RLC is not enabled, do nothing */
@@ -70,8 +72,8 @@ void amdgpu_gfx_rlc_exit_safe_mode(struct amdgpu_device *adev)
 	if (adev->cg_flags &
 	    (AMD_CG_SUPPORT_GFX_CGCG | AMD_CG_SUPPORT_GFX_MGCG |
 	     AMD_CG_SUPPORT_GFX_3D_CGCG)) {
-		adev->gfx.rlc.funcs->unset_safe_mode(adev);
-		adev->gfx.rlc.in_safe_mode = false;
+		adev->gfx.rlc.funcs->unset_safe_mode(adev, xcc_id);
+		adev->gfx.rlc.in_safe_mode[xcc_id] = false;
 	}
 }
 
@@ -87,7 +89,7 @@ void amdgpu_gfx_rlc_exit_safe_mode(struct amdgpu_device *adev)
 int amdgpu_gfx_rlc_init_sr(struct amdgpu_device *adev, u32 dws)
 {
 	const u32 *src_ptr;
-	volatile u32 *dst_ptr;
+	u32 *dst_ptr;
 	u32 i;
 	int r;
 
@@ -187,7 +189,7 @@ int amdgpu_gfx_rlc_init_cpt(struct amdgpu_device *adev)
 void amdgpu_gfx_rlc_setup_cp_table(struct amdgpu_device *adev)
 {
 	const __le32 *fw_data;
-	volatile u32 *dst_ptr;
+	u32 *dst_ptr;
 	int me, i, max_me;
 	u32 bo_offset = 0;
 	u32 table_offset, table_size;
@@ -239,7 +241,7 @@ void amdgpu_gfx_rlc_setup_cp_table(struct amdgpu_device *adev)
 			table_size = le32_to_cpu(hdr->jt_size);
 		}
 
-		for (i = 0; i < table_size; i ++) {
+		for (i = 0; i < table_size; i++) {
 			dst_ptr[bo_offset + i] =
 				cpu_to_le32(le32_to_cpu(fw_data[table_offset + i]));
 		}
@@ -529,13 +531,12 @@ int amdgpu_gfx_rlc_init_microcode(struct amdgpu_device *adev,
 	if (version_major == 2 && version_minor == 1)
 		adev->gfx.rlc.is_rlc_v2_1 = true;
 
-	if (version_minor >= 0) {
-		err = amdgpu_gfx_rlc_init_microcode_v2_0(adev);
-		if (err) {
-			dev_err(adev->dev, "fail to init rlc v2_0 microcode\n");
-			return err;
-		}
+	err = amdgpu_gfx_rlc_init_microcode_v2_0(adev);
+	if (err) {
+		dev_err(adev->dev, "fail to init rlc v2_0 microcode\n");
+		return err;
 	}
+
 	if (version_minor >= 1)
 		amdgpu_gfx_rlc_init_microcode_v2_1(adev);
 	if (version_minor >= 2)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.h
index 23f060db9255..2ce310b31942 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_rlc.h
@@ -26,6 +26,8 @@
 
 #include "clearstate_defs.h"
 
+#define AMDGPU_MAX_RLC_INSTANCES	8
+
 /* firmware ID used in rlc toc */
 typedef enum _FIRMWARE_ID_ {
 	FIRMWARE_ID_INVALID					= 0,
@@ -110,6 +112,53 @@ typedef enum _SOC21_FIRMWARE_ID_ {
     SOC21_FIRMWARE_ID_MAX                         = 37
 } SOC21_FIRMWARE_ID;
 
+typedef enum _SOC24_FIRMWARE_ID_ {
+    SOC24_FIRMWARE_ID_INVALID                     = 0,
+    SOC24_FIRMWARE_ID_RLC_G_UCODE                 = 1,
+    SOC24_FIRMWARE_ID_RLC_TOC                     = 2,
+    SOC24_FIRMWARE_ID_RLCG_SCRATCH                = 3,
+    SOC24_FIRMWARE_ID_RLC_SRM_ARAM                = 4,
+    SOC24_FIRMWARE_ID_RLC_P_UCODE                 = 5,
+    SOC24_FIRMWARE_ID_RLC_V_UCODE                 = 6,
+    SOC24_FIRMWARE_ID_RLX6_UCODE                  = 7,
+    SOC24_FIRMWARE_ID_RLX6_UCODE_CORE1            = 8,
+    SOC24_FIRMWARE_ID_RLX6_DRAM_BOOT              = 9,
+    SOC24_FIRMWARE_ID_RLX6_DRAM_BOOT_CORE1        = 10,
+    SOC24_FIRMWARE_ID_SDMA_UCODE_TH0              = 11,
+    SOC24_FIRMWARE_ID_SDMA_UCODE_TH1              = 12,
+    SOC24_FIRMWARE_ID_CP_PFP                      = 13,
+    SOC24_FIRMWARE_ID_CP_ME                       = 14,
+    SOC24_FIRMWARE_ID_CP_MEC                      = 15,
+    SOC24_FIRMWARE_ID_RS64_MES_P0                 = 16,
+    SOC24_FIRMWARE_ID_RS64_MES_P1                 = 17,
+    SOC24_FIRMWARE_ID_RS64_PFP                    = 18,
+    SOC24_FIRMWARE_ID_RS64_ME                     = 19,
+    SOC24_FIRMWARE_ID_RS64_MEC                    = 20,
+    SOC24_FIRMWARE_ID_RS64_MES_P0_STACK           = 21,
+    SOC24_FIRMWARE_ID_RS64_MES_P1_STACK           = 22,
+    SOC24_FIRMWARE_ID_RS64_PFP_P0_STACK           = 23,
+    SOC24_FIRMWARE_ID_RS64_PFP_P1_STACK           = 24,
+    SOC24_FIRMWARE_ID_RS64_ME_P0_STACK            = 25,
+    SOC24_FIRMWARE_ID_RS64_ME_P1_STACK            = 26,
+    SOC24_FIRMWARE_ID_RS64_MEC_P0_STACK           = 27,
+    SOC24_FIRMWARE_ID_RS64_MEC_P1_STACK           = 28,
+    SOC24_FIRMWARE_ID_RS64_MEC_P2_STACK           = 29,
+    SOC24_FIRMWARE_ID_RS64_MEC_P3_STACK           = 30,
+    SOC24_FIRMWARE_ID_RLC_SRM_DRAM_SR             = 31,
+    SOC24_FIRMWARE_ID_RLCG_SCRATCH_SR             = 32,
+    SOC24_FIRMWARE_ID_RLCP_SCRATCH_SR             = 33,
+    SOC24_FIRMWARE_ID_RLCV_SCRATCH_SR             = 34,
+    SOC24_FIRMWARE_ID_RLX6_DRAM_SR                = 35,
+    SOC24_FIRMWARE_ID_RLX6_DRAM_SR_CORE1          = 36,
+    SOC24_FIRMWARE_ID_RLCDEBUGLOG                 = 37,
+    SOC24_FIRMWARE_ID_SRIOV_DEBUG                 = 38,
+    SOC24_FIRMWARE_ID_SRIOV_CSA_RLC               = 39,
+    SOC24_FIRMWARE_ID_SRIOV_CSA_SDMA              = 40,
+    SOC24_FIRMWARE_ID_SRIOV_CSA_CP                = 41,
+    SOC24_FIRMWARE_ID_UMF_ZONE_PAD                = 42,
+    SOC24_FIRMWARE_ID_MAX                         = 43
+} SOC24_FIRMWARE_ID;
+
 typedef struct _RLC_TABLE_OF_CONTENT {
 	union {
 		unsigned int	DW0;
@@ -153,21 +202,62 @@ typedef struct _RLC_TABLE_OF_CONTENT {
 	};
 } RLC_TABLE_OF_CONTENT;
 
+typedef struct _RLC_TABLE_OF_CONTENT_V2 {
+	union {
+		unsigned int    DW0;
+		struct {
+			uint32_t offset         : 25;
+			uint32_t id             : 7;
+		};
+	};
+
+	union {
+		unsigned int    DW1;
+		struct {
+			uint32_t reserved0              : 1;
+			uint32_t reserved1              : 1;
+			uint32_t reserved2              : 1;
+			uint32_t memory_destination     : 2;
+			uint32_t vfflr_image_code       : 4;
+			uint32_t reserved9              : 1;
+			uint32_t reserved10             : 1;
+			uint32_t reserved11             : 1;
+			uint32_t size_x16               : 1;
+			uint32_t reserved13             : 1;
+			uint32_t size                   : 18;
+		};
+	};
+} RLC_TABLE_OF_CONTENT_V2;
+
 #define RLC_TOC_MAX_SIZE		64
 
 struct amdgpu_rlc_funcs {
 	bool (*is_rlc_enabled)(struct amdgpu_device *adev);
-	void (*set_safe_mode)(struct amdgpu_device *adev);
-	void (*unset_safe_mode)(struct amdgpu_device *adev);
+	void (*set_safe_mode)(struct amdgpu_device *adev, int xcc_id);
+	void (*unset_safe_mode)(struct amdgpu_device *adev, int xcc_id);
 	int  (*init)(struct amdgpu_device *adev);
 	u32  (*get_csb_size)(struct amdgpu_device *adev);
-	void (*get_csb_buffer)(struct amdgpu_device *adev, volatile u32 *buffer);
+
+	/**
+	 * @get_csb_buffer: Get the clear state to be put into the hardware.
+	 *
+	 * The parameter adev is used to get the CS data and other gfx info,
+	 * and buffer is the RLC CS pointer
+	 *
+	 * Sometimes, the user space puts a request to clear the state in the
+	 * command buffer; this function provides the clear state that gets put
+	 * into the hardware. Note that the driver programs Clear State
+	 * Indirect Buffer (CSB) explicitly when it sets up the kernel rings,
+	 * and it also provides a pointer to it which is used by the firmware
+	 * to load the clear state in some cases.
+	 */
+	void (*get_csb_buffer)(struct amdgpu_device *adev, u32 *buffer);
 	int  (*get_cp_table_num)(struct amdgpu_device *adev);
 	int  (*resume)(struct amdgpu_device *adev);
 	void (*stop)(struct amdgpu_device *adev);
 	void (*reset)(struct amdgpu_device *adev);
 	void (*start)(struct amdgpu_device *adev);
-	void (*update_spm_vmid)(struct amdgpu_device *adev, unsigned vmid);
+	void (*update_spm_vmid)(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned vmid);
 	bool (*is_rlcg_access_range)(struct amdgpu_device *adev, uint32_t reg);
 };
 
@@ -185,23 +275,23 @@ struct amdgpu_rlc {
 	/* for power gating */
 	struct amdgpu_bo        *save_restore_obj;
 	uint64_t                save_restore_gpu_addr;
-	volatile uint32_t       *sr_ptr;
+	uint32_t		*sr_ptr;
 	const u32               *reg_list;
 	u32                     reg_list_size;
 	/* for clear state */
 	struct amdgpu_bo        *clear_state_obj;
 	uint64_t                clear_state_gpu_addr;
-	volatile uint32_t       *cs_ptr;
+	uint32_t		*cs_ptr;
 	const struct cs_section_def   *cs_data;
 	u32                     clear_state_size;
 	/* for cp tables */
 	struct amdgpu_bo        *cp_table_obj;
 	uint64_t                cp_table_gpu_addr;
-	volatile uint32_t       *cp_table_ptr;
+	uint32_t		*cp_table_ptr;
 	u32                     cp_table_size;
 
 	/* safe mode for updating CG/PG state */
-	bool in_safe_mode;
+	bool in_safe_mode[AMDGPU_MAX_RLC_INSTANCES];
 	const struct amdgpu_rlc_funcs *funcs;
 
 	/* for firmware data */
@@ -257,11 +347,11 @@ struct amdgpu_rlc {
 
 	bool rlcg_reg_access_supported;
 	/* registers for rlcg indirect reg access */
-	struct amdgpu_rlcg_reg_access_ctrl reg_access_ctrl;
+	struct amdgpu_rlcg_reg_access_ctrl reg_access_ctrl[AMDGPU_MAX_RLC_INSTANCES];
 };
 
-void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev);
-void amdgpu_gfx_rlc_exit_safe_mode(struct amdgpu_device *adev);
+void amdgpu_gfx_rlc_enter_safe_mode(struct amdgpu_device *adev, int xcc_id);
+void amdgpu_gfx_rlc_exit_safe_mode(struct amdgpu_device *adev, int xcc_id);
 int amdgpu_gfx_rlc_init_sr(struct amdgpu_device *adev, u32 dws);
 int amdgpu_gfx_rlc_init_csb(struct amdgpu_device *adev);
 int amdgpu_gfx_rlc_init_cpt(struct amdgpu_device *adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c
index c6b4337eb20c..39070b2a4c04 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c
@@ -81,7 +81,7 @@ int amdgpu_sa_bo_new(struct amdgpu_sa_manager *sa_manager,
 		     unsigned int size)
 {
 	struct drm_suballoc *sa = drm_suballoc_new(&sa_manager->base, size,
-						   GFP_KERNEL, true, 0);
+						   GFP_KERNEL, false, 0);
 
 	if (IS_ERR(sa)) {
 		*sa_bo = NULL;
@@ -93,8 +93,7 @@ int amdgpu_sa_bo_new(struct amdgpu_sa_manager *sa_manager,
 	return 0;
 }
 
-void amdgpu_sa_bo_free(struct amdgpu_device *adev, struct drm_suballoc **sa_bo,
-		       struct dma_fence *fence)
+void amdgpu_sa_bo_free(struct drm_suballoc **sa_bo, struct dma_fence *fence)
 {
 	if (sa_bo == NULL || *sa_bo == NULL) {
 		return;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c
index e9b45089a28a..341beec59537 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c
@@ -22,7 +22,6 @@
  * Authors: Andres Rodriguez <andresx7@gmail.com>
  */
 
-#include <linux/fdtable.h>
 #include <linux/file.h>
 #include <linux/pid.h>
 
@@ -36,25 +35,26 @@ static int amdgpu_sched_process_priority_override(struct amdgpu_device *adev,
 						  int fd,
 						  int32_t priority)
 {
-	struct fd f = fdget(fd);
+	CLASS(fd, f)(fd);
 	struct amdgpu_fpriv *fpriv;
+	struct amdgpu_ctx_mgr *mgr;
 	struct amdgpu_ctx *ctx;
 	uint32_t id;
 	int r;
 
-	if (!f.file)
+	if (fd_empty(f))
 		return -EINVAL;
 
-	r = amdgpu_file_to_fpriv(f.file, &fpriv);
-	if (r) {
-		fdput(f);
+	r = amdgpu_file_to_fpriv(fd_file(f), &fpriv);
+	if (r)
 		return r;
-	}
 
-	idr_for_each_entry(&fpriv->ctx_mgr.ctx_handles, ctx, id)
+	mgr = &fpriv->ctx_mgr;
+	mutex_lock(&mgr->lock);
+	idr_for_each_entry(&mgr->ctx_handles, ctx, id)
 		amdgpu_ctx_priority_override(ctx, priority);
+	mutex_unlock(&mgr->lock);
 
-	fdput(f);
 	return 0;
 }
 
@@ -63,31 +63,25 @@ static int amdgpu_sched_context_priority_override(struct amdgpu_device *adev,
 						  unsigned ctx_id,
 						  int32_t priority)
 {
-	struct fd f = fdget(fd);
+	CLASS(fd, f)(fd);
 	struct amdgpu_fpriv *fpriv;
 	struct amdgpu_ctx *ctx;
 	int r;
 
-	if (!f.file)
+	if (fd_empty(f))
 		return -EINVAL;
 
-	r = amdgpu_file_to_fpriv(f.file, &fpriv);
-	if (r) {
-		fdput(f);
+	r = amdgpu_file_to_fpriv(fd_file(f), &fpriv);
+	if (r)
 		return r;
-	}
 
 	ctx = amdgpu_ctx_get(fpriv, ctx_id);
 
-	if (!ctx) {
-		fdput(f);
+	if (!ctx)
 		return -EINVAL;
-	}
 
 	amdgpu_ctx_priority_override(ctx, priority);
 	amdgpu_ctx_put(ctx);
-	fdput(f);
-
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c
index 231ca06bc9c7..8b8a04138711 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c
@@ -25,6 +25,9 @@
 #include "amdgpu.h"
 #include "amdgpu_sdma.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_reset.h"
+#include "gc/gc_10_1_0_offset.h"
+#include "gc/gc_10_3_0_sh_mask.h"
 
 #define AMDGPU_CSA_SDMA_SIZE 64
 /* SDMA CSA reside in the 3rd page of CSA */
@@ -64,7 +67,7 @@ int amdgpu_sdma_get_index_from_ring(struct amdgpu_ring *ring, uint32_t *index)
 }
 
 uint64_t amdgpu_sdma_get_csa_mc_addr(struct amdgpu_ring *ring,
-				     unsigned vmid)
+				     unsigned int vmid)
 {
 	struct amdgpu_device *adev = ring->adev;
 	uint64_t csa_mc_addr;
@@ -72,25 +75,17 @@ uint64_t amdgpu_sdma_get_csa_mc_addr(struct amdgpu_ring *ring,
 	int r;
 
 	/* don't enable OS preemption on SDMA under SRIOV */
-	if (amdgpu_sriov_vf(adev) || vmid == 0 || !amdgpu_mcbp)
+	if (amdgpu_sriov_vf(adev) || vmid == 0 || !adev->gfx.mcbp)
 		return 0;
 
-	if (ring->is_mes_queue) {
-		uint32_t offset = 0;
+	r = amdgpu_sdma_get_index_from_ring(ring, &index);
 
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  sdma[ring->idx].sdma_meta_data);
-		csa_mc_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-	} else {
-		r = amdgpu_sdma_get_index_from_ring(ring, &index);
-
-		if (r || index > 31)
-			csa_mc_addr = 0;
-		else
-			csa_mc_addr = amdgpu_csa_vaddr(adev) +
-				AMDGPU_CSA_SDMA_OFFSET +
-				index * AMDGPU_CSA_SDMA_SIZE;
-	}
+	if (r || index > 31)
+		csa_mc_addr = 0;
+	else
+		csa_mc_addr = amdgpu_csa_vaddr(adev) +
+			AMDGPU_CSA_SDMA_OFFSET +
+			index * AMDGPU_CSA_SDMA_SIZE;
 
 	return csa_mc_addr;
 }
@@ -158,6 +153,7 @@ static int amdgpu_sdma_init_inst_ctx(struct amdgpu_sdma_instance *sdma_inst)
 	const struct common_firmware_header *header = NULL;
 	const struct sdma_firmware_header_v1_0 *hdr;
 	const struct sdma_firmware_header_v2_0 *hdr_v2;
+	const struct sdma_firmware_header_v3_0 *hdr_v3;
 
 	header = (const struct common_firmware_header *)
 		sdma_inst->fw->data;
@@ -174,6 +170,11 @@ static int amdgpu_sdma_init_inst_ctx(struct amdgpu_sdma_instance *sdma_inst)
 		sdma_inst->fw_version = le32_to_cpu(hdr_v2->header.ucode_version);
 		sdma_inst->feature_version = le32_to_cpu(hdr_v2->ucode_feature_version);
 		break;
+	case 3:
+		hdr_v3 = (const struct sdma_firmware_header_v3_0 *)sdma_inst->fw->data;
+		sdma_inst->fw_version = le32_to_cpu(hdr_v3->header.ucode_version);
+		sdma_inst->feature_version = le32_to_cpu(hdr_v3->ucode_feature_version);
+		break;
 	default:
 		return -EINVAL;
 	}
@@ -206,16 +207,19 @@ int amdgpu_sdma_init_microcode(struct amdgpu_device *adev,
 	const struct common_firmware_header *header = NULL;
 	int err, i;
 	const struct sdma_firmware_header_v2_0 *sdma_hdr;
+	const struct sdma_firmware_header_v3_0 *sdma_hv3;
 	uint16_t version_major;
 	char ucode_prefix[30];
-	char fw_name[40];
 
 	amdgpu_ucode_ip_version_decode(adev, SDMA0_HWIP, ucode_prefix, sizeof(ucode_prefix));
 	if (instance == 0)
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s.bin", ucode_prefix);
+		err = amdgpu_ucode_request(adev, &adev->sdma.instance[instance].fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s.bin", ucode_prefix);
 	else
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s%d.bin", ucode_prefix, instance);
-	err = amdgpu_ucode_request(adev, &adev->sdma.instance[instance].fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->sdma.instance[instance].fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s%d.bin", ucode_prefix, instance);
 	if (err)
 		goto out;
 
@@ -239,9 +243,6 @@ int amdgpu_sdma_init_microcode(struct amdgpu_device *adev,
 			       sizeof(struct amdgpu_sdma_instance));
 	}
 
-	if (amdgpu_sriov_vf(adev))
-		return 0;
-
 	DRM_DEBUG("psp_load == '%s'\n",
 		  adev->firmware.load_type == AMDGPU_FW_LOAD_PSP ? "true" : "false");
 
@@ -252,6 +253,19 @@ int amdgpu_sdma_init_microcode(struct amdgpu_device *adev,
 				if (!duplicate && (instance != i))
 					continue;
 				else {
+					/* Use a single copy per SDMA firmware type. PSP uses the same instance for all
+					 * groups of SDMAs */
+					if ((amdgpu_ip_version(adev, SDMA0_HWIP, 0) ==
+						IP_VERSION(4, 4, 2) ||
+					     amdgpu_ip_version(adev, SDMA0_HWIP, 0) ==
+						IP_VERSION(4, 4, 4) ||
+					     amdgpu_ip_version(adev, SDMA0_HWIP, 0) ==
+						IP_VERSION(4, 4, 5)) &&
+					    adev->firmware.load_type ==
+						AMDGPU_FW_LOAD_PSP &&
+					    adev->sdma.num_inst_per_aid == i) {
+						break;
+					}
 					info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SDMA0 + i];
 					info->ucode_id = AMDGPU_UCODE_ID_SDMA0 + i;
 					info->fw = adev->sdma.instance[i].fw;
@@ -274,6 +288,15 @@ int amdgpu_sdma_init_microcode(struct amdgpu_device *adev,
 			adev->firmware.fw_size +=
 				ALIGN(le32_to_cpu(sdma_hdr->ctl_ucode_size_bytes), PAGE_SIZE);
 			break;
+		case 3:
+			sdma_hv3 = (const struct sdma_firmware_header_v3_0 *)
+				adev->sdma.instance[0].fw->data;
+			info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SDMA_RS64];
+			info->ucode_id = AMDGPU_UCODE_ID_SDMA_RS64;
+			info->fw = adev->sdma.instance[0].fw;
+			adev->firmware.fw_size +=
+				ALIGN(le32_to_cpu(sdma_hv3->ucode_size_bytes), PAGE_SIZE);
+			break;
 		default:
 			err = -EINVAL;
 		}
@@ -285,27 +308,6 @@ out:
 	return err;
 }
 
-void amdgpu_sdma_unset_buffer_funcs_helper(struct amdgpu_device *adev)
-{
-	struct amdgpu_ring *sdma;
-	int i;
-
-	for (i = 0; i < adev->sdma.num_instances; i++) {
-		if (adev->sdma.has_page_queue) {
-			sdma = &adev->sdma.instance[i].page;
-			if (adev->mman.buffer_funcs_ring == sdma) {
-				amdgpu_ttm_set_buffer_funcs_status(adev, false);
-				break;
-			}
-		}
-		sdma = &adev->sdma.instance[i].ring;
-		if (adev->mman.buffer_funcs_ring == sdma) {
-			amdgpu_ttm_set_buffer_funcs_status(adev, false);
-			break;
-		}
-	}
-}
-
 int amdgpu_sdma_ras_sw_init(struct amdgpu_device *adev)
 {
 	int err = 0;
@@ -340,3 +342,270 @@ int amdgpu_sdma_ras_sw_init(struct amdgpu_device *adev)
 
 	return 0;
 }
+
+/*
+ * debugfs for to enable/disable sdma job submission to specific core.
+ */
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_debugfs_sdma_sched_mask_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u64 i, num_ring;
+	u64 mask = 0;
+	struct amdgpu_ring *ring, *page = NULL;
+
+	if (!adev)
+		return -ENODEV;
+
+	/* Determine the number of rings per SDMA instance
+	 * (1 for sdma gfx ring, 2 if page queue exists)
+	 */
+	if (adev->sdma.has_page_queue)
+		num_ring = 2;
+	else
+		num_ring = 1;
+
+	/* Calculate the maximum possible mask value
+	 * based on the number of SDMA instances and rings
+	*/
+	mask = BIT_ULL(adev->sdma.num_instances * num_ring) - 1;
+
+	if ((val & mask) == 0)
+		return -EINVAL;
+
+	for (i = 0; i < adev->sdma.num_instances; ++i) {
+		ring = &adev->sdma.instance[i].ring;
+		if (adev->sdma.has_page_queue)
+			page = &adev->sdma.instance[i].page;
+		if (val & BIT_ULL(i * num_ring))
+			ring->sched.ready = true;
+		else
+			ring->sched.ready = false;
+
+		if (page) {
+			if (val & BIT_ULL(i * num_ring + 1))
+				page->sched.ready = true;
+			else
+				page->sched.ready = false;
+		}
+	}
+	/* publish sched.ready flag update effective immediately across smp */
+	smp_rmb();
+	return 0;
+}
+
+static int amdgpu_debugfs_sdma_sched_mask_get(void *data, u64 *val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u64 i, num_ring;
+	u64 mask = 0;
+	struct amdgpu_ring *ring, *page = NULL;
+
+	if (!adev)
+		return -ENODEV;
+
+	/* Determine the number of rings per SDMA instance
+	 * (1 for sdma gfx ring, 2 if page queue exists)
+	 */
+	if (adev->sdma.has_page_queue)
+		num_ring = 2;
+	else
+		num_ring = 1;
+
+	for (i = 0; i < adev->sdma.num_instances; ++i) {
+		ring = &adev->sdma.instance[i].ring;
+		if (adev->sdma.has_page_queue)
+			page = &adev->sdma.instance[i].page;
+
+		if (ring->sched.ready)
+			mask |= BIT_ULL(i * num_ring);
+		else
+			mask &= ~BIT_ULL(i * num_ring);
+
+		if (page) {
+			if (page->sched.ready)
+				mask |= BIT_ULL(i * num_ring + 1);
+			else
+				mask &= ~BIT_ULL(i * num_ring + 1);
+		}
+	}
+
+	*val = mask;
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_debugfs_sdma_sched_mask_fops,
+			 amdgpu_debugfs_sdma_sched_mask_get,
+			 amdgpu_debugfs_sdma_sched_mask_set, "%llx\n");
+
+#endif
+
+void amdgpu_debugfs_sdma_sched_mask_init(struct amdgpu_device *adev)
+{
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+	char name[32];
+
+	if (!(adev->sdma.num_instances > 1))
+		return;
+	sprintf(name, "amdgpu_sdma_sched_mask");
+	debugfs_create_file(name, 0600, root, adev,
+			    &amdgpu_debugfs_sdma_sched_mask_fops);
+#endif
+}
+
+static ssize_t amdgpu_get_sdma_reset_mask(struct device *dev,
+						struct device_attribute *attr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (!adev)
+		return -ENODEV;
+
+	return amdgpu_show_reset_mask(buf, adev->sdma.supported_reset);
+}
+
+static DEVICE_ATTR(sdma_reset_mask, 0444,
+		   amdgpu_get_sdma_reset_mask, NULL);
+
+int amdgpu_sdma_sysfs_reset_mask_init(struct amdgpu_device *adev)
+{
+	int r = 0;
+
+	if (!amdgpu_gpu_recovery)
+		return r;
+
+	if (adev->sdma.num_instances) {
+		r = device_create_file(adev->dev, &dev_attr_sdma_reset_mask);
+		if (r)
+			return r;
+	}
+
+	return r;
+}
+
+void amdgpu_sdma_sysfs_reset_mask_fini(struct amdgpu_device *adev)
+{
+	if (!amdgpu_gpu_recovery)
+		return;
+
+	if (adev->dev->kobj.sd) {
+		if (adev->sdma.num_instances)
+			device_remove_file(adev->dev, &dev_attr_sdma_reset_mask);
+	}
+}
+
+struct amdgpu_ring *amdgpu_sdma_get_shared_ring(struct amdgpu_device *adev, struct amdgpu_ring *ring)
+{
+	if (adev->sdma.has_page_queue &&
+	    (ring->me < adev->sdma.num_instances) &&
+	    (ring == &adev->sdma.instance[ring->me].ring))
+		return &adev->sdma.instance[ring->me].page;
+	else
+		return NULL;
+}
+
+/**
+* amdgpu_sdma_is_shared_inv_eng - Check if a ring is an SDMA ring that shares a VM invalidation engine
+* @adev: Pointer to the AMDGPU device structure
+* @ring: Pointer to the ring structure to check
+*
+* This function checks if the given ring is an SDMA ring that shares a VM invalidation engine.
+* It returns true if the ring is such an SDMA ring, false otherwise.
+*/
+bool amdgpu_sdma_is_shared_inv_eng(struct amdgpu_device *adev, struct amdgpu_ring *ring)
+{
+	int i = ring->me;
+
+	if (!adev->sdma.has_page_queue || i >= adev->sdma.num_instances)
+		return false;
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 4) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 5, 0))
+		return (ring == &adev->sdma.instance[i].page);
+	else
+		return false;
+}
+
+static int amdgpu_sdma_soft_reset(struct amdgpu_device *adev, u32 instance_id)
+{
+	struct amdgpu_sdma_instance *sdma_instance = &adev->sdma.instance[instance_id];
+
+	if (sdma_instance->funcs->soft_reset_kernel_queue)
+		return sdma_instance->funcs->soft_reset_kernel_queue(adev, instance_id);
+
+	return -EOPNOTSUPP;
+}
+
+/**
+ * amdgpu_sdma_reset_engine - Reset a specific SDMA engine
+ * @adev: Pointer to the AMDGPU device
+ * @instance_id: Logical ID of the SDMA engine instance to reset
+ * @caller_handles_kernel_queues: Skip kernel queue processing. Caller
+ * will handle it.
+ *
+ * Returns: 0 on success, or a negative error code on failure.
+ */
+int amdgpu_sdma_reset_engine(struct amdgpu_device *adev, uint32_t instance_id,
+			     bool caller_handles_kernel_queues)
+{
+	int ret = 0;
+	struct amdgpu_sdma_instance *sdma_instance = &adev->sdma.instance[instance_id];
+	struct amdgpu_ring *gfx_ring = &sdma_instance->ring;
+	struct amdgpu_ring *page_ring = &sdma_instance->page;
+
+	mutex_lock(&sdma_instance->engine_reset_mutex);
+
+	if (!caller_handles_kernel_queues) {
+		/* Stop the scheduler's work queue for the GFX and page rings if they are running.
+		 * This ensures that no new tasks are submitted to the queues while
+		 * the reset is in progress.
+		 */
+		drm_sched_wqueue_stop(&gfx_ring->sched);
+
+		if (adev->sdma.has_page_queue)
+			drm_sched_wqueue_stop(&page_ring->sched);
+	}
+
+	if (sdma_instance->funcs->stop_kernel_queue) {
+		sdma_instance->funcs->stop_kernel_queue(gfx_ring);
+		if (adev->sdma.has_page_queue)
+			sdma_instance->funcs->stop_kernel_queue(page_ring);
+	}
+
+	/* Perform the SDMA reset for the specified instance */
+	ret = amdgpu_sdma_soft_reset(adev, instance_id);
+	if (ret) {
+		dev_err(adev->dev, "Failed to reset SDMA logical instance %u\n", instance_id);
+		goto exit;
+	}
+
+	if (sdma_instance->funcs->start_kernel_queue) {
+		sdma_instance->funcs->start_kernel_queue(gfx_ring);
+		if (adev->sdma.has_page_queue)
+			sdma_instance->funcs->start_kernel_queue(page_ring);
+	}
+
+exit:
+	if (!caller_handles_kernel_queues) {
+		/* Restart the scheduler's work queue for the GFX and page rings
+		 * if they were stopped by this function. This allows new tasks
+		 * to be submitted to the queues after the reset is complete.
+		 */
+		if (!ret) {
+			amdgpu_fence_driver_force_completion(gfx_ring);
+			drm_sched_wqueue_start(&gfx_ring->sched);
+			if (adev->sdma.has_page_queue) {
+				amdgpu_fence_driver_force_completion(page_ring);
+				drm_sched_wqueue_start(&page_ring->sched);
+			}
+		}
+	}
+	mutex_unlock(&sdma_instance->engine_reset_mutex);
+
+	return ret;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h
index fc8528812598..34311f32be4c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h
@@ -26,7 +26,7 @@
 #include "amdgpu_ras.h"
 
 /* max number of IP instances */
-#define AMDGPU_MAX_SDMA_INSTANCES		8
+#define AMDGPU_MAX_SDMA_INSTANCES		16
 
 enum amdgpu_sdma_irq {
 	AMDGPU_SDMA_IRQ_INSTANCE0  = 0,
@@ -37,9 +37,25 @@ enum amdgpu_sdma_irq {
 	AMDGPU_SDMA_IRQ_INSTANCE5,
 	AMDGPU_SDMA_IRQ_INSTANCE6,
 	AMDGPU_SDMA_IRQ_INSTANCE7,
+	AMDGPU_SDMA_IRQ_INSTANCE8,
+	AMDGPU_SDMA_IRQ_INSTANCE9,
+	AMDGPU_SDMA_IRQ_INSTANCE10,
+	AMDGPU_SDMA_IRQ_INSTANCE11,
+	AMDGPU_SDMA_IRQ_INSTANCE12,
+	AMDGPU_SDMA_IRQ_INSTANCE13,
+	AMDGPU_SDMA_IRQ_INSTANCE14,
+	AMDGPU_SDMA_IRQ_INSTANCE15,
 	AMDGPU_SDMA_IRQ_LAST
 };
 
+#define NUM_SDMA(x) hweight32(x)
+
+struct amdgpu_sdma_funcs {
+	int (*stop_kernel_queue)(struct amdgpu_ring *ring);
+	int (*start_kernel_queue)(struct amdgpu_ring *ring);
+	int (*soft_reset_kernel_queue)(struct amdgpu_device *adev, u32 instance_id);
+};
+
 struct amdgpu_sdma_instance {
 	/* SDMA firmware */
 	const struct firmware	*fw;
@@ -49,6 +65,44 @@ struct amdgpu_sdma_instance {
 	struct amdgpu_ring	ring;
 	struct amdgpu_ring	page;
 	bool			burst_nop;
+	uint32_t		aid_id;
+
+	struct amdgpu_bo	*sdma_fw_obj;
+	uint64_t		sdma_fw_gpu_addr;
+	uint32_t		*sdma_fw_ptr;
+	struct mutex		engine_reset_mutex;
+	/* track guilty state of GFX and PAGE queues */
+	bool			gfx_guilty;
+	bool			page_guilty;
+	const struct amdgpu_sdma_funcs   *funcs;
+};
+
+enum amdgpu_sdma_ras_memory_id {
+	AMDGPU_SDMA_MBANK_DATA_BUF0 = 1,
+	AMDGPU_SDMA_MBANK_DATA_BUF1 = 2,
+	AMDGPU_SDMA_MBANK_DATA_BUF2 = 3,
+	AMDGPU_SDMA_MBANK_DATA_BUF3 = 4,
+	AMDGPU_SDMA_MBANK_DATA_BUF4 = 5,
+	AMDGPU_SDMA_MBANK_DATA_BUF5 = 6,
+	AMDGPU_SDMA_MBANK_DATA_BUF6 = 7,
+	AMDGPU_SDMA_MBANK_DATA_BUF7 = 8,
+	AMDGPU_SDMA_MBANK_DATA_BUF8 = 9,
+	AMDGPU_SDMA_MBANK_DATA_BUF9 = 10,
+	AMDGPU_SDMA_MBANK_DATA_BUF10 = 11,
+	AMDGPU_SDMA_MBANK_DATA_BUF11 = 12,
+	AMDGPU_SDMA_MBANK_DATA_BUF12 = 13,
+	AMDGPU_SDMA_MBANK_DATA_BUF13 = 14,
+	AMDGPU_SDMA_MBANK_DATA_BUF14 = 15,
+	AMDGPU_SDMA_MBANK_DATA_BUF15 = 16,
+	AMDGPU_SDMA_UCODE_BUF = 17,
+	AMDGPU_SDMA_RB_CMD_BUF = 18,
+	AMDGPU_SDMA_IB_CMD_BUF = 19,
+	AMDGPU_SDMA_UTCL1_RD_FIFO = 20,
+	AMDGPU_SDMA_UTCL1_RDBST_FIFO = 21,
+	AMDGPU_SDMA_UTCL1_WR_FIFO = 22,
+	AMDGPU_SDMA_DATA_LUT_FIFO = 23,
+	AMDGPU_SDMA_SPLIT_DAT_BUF = 24,
+	AMDGPU_SDMA_MEMORY_BLOCK_LAST,
 };
 
 struct amdgpu_sdma_ras {
@@ -59,17 +113,26 @@ struct amdgpu_sdma {
 	struct amdgpu_sdma_instance instance[AMDGPU_MAX_SDMA_INSTANCES];
 	struct amdgpu_irq_src	trap_irq;
 	struct amdgpu_irq_src	illegal_inst_irq;
+	struct amdgpu_irq_src	fence_irq;
 	struct amdgpu_irq_src	ecc_irq;
 	struct amdgpu_irq_src	vm_hole_irq;
 	struct amdgpu_irq_src	doorbell_invalid_irq;
 	struct amdgpu_irq_src	pool_timeout_irq;
 	struct amdgpu_irq_src	srbm_write_irq;
+	struct amdgpu_irq_src	ctxt_empty_irq;
 
 	int			num_instances;
+	uint32_t 		sdma_mask;
+	int			num_inst_per_aid;
 	uint32_t                    srbm_soft_reset;
 	bool			has_page_queue;
 	struct ras_common_if	*ras_if;
 	struct amdgpu_sdma_ras	*ras;
+	uint32_t		*ip_dump;
+	uint32_t 		supported_reset;
+	struct list_head	reset_callback_list;
+	bool			no_user_submission;
+	bool			disable_uq;
 };
 
 /*
@@ -91,7 +154,7 @@ struct amdgpu_buffer_funcs {
 				 uint64_t dst_offset,
 				 /* number of byte to transfer */
 				 uint32_t byte_count,
-				 bool tmz);
+				 uint32_t copy_flags);
 
 	/* maximum bytes in a single operation */
 	uint32_t	fill_max_bytes;
@@ -109,6 +172,9 @@ struct amdgpu_buffer_funcs {
 				 uint32_t byte_count);
 };
 
+int amdgpu_sdma_reset_engine(struct amdgpu_device *adev, uint32_t instance_id,
+			     bool caller_handles_kernel_queues);
+
 #define amdgpu_emit_copy_buffer(adev, ib, s, d, b, t) (adev)->mman.buffer_funcs->emit_copy_buffer((ib),  (s), (d), (b), (t))
 #define amdgpu_emit_fill_buffer(adev, ib, s, d, b) (adev)->mman.buffer_funcs->emit_fill_buffer((ib), (s), (d), (b))
 
@@ -128,7 +194,11 @@ int amdgpu_sdma_init_microcode(struct amdgpu_device *adev, u32 instance,
 			       bool duplicate);
 void amdgpu_sdma_destroy_inst_ctx(struct amdgpu_device *adev,
         bool duplicate);
-void amdgpu_sdma_unset_buffer_funcs_helper(struct amdgpu_device *adev);
 int amdgpu_sdma_ras_sw_init(struct amdgpu_device *adev);
-
+void amdgpu_debugfs_sdma_sched_mask_init(struct amdgpu_device *adev);
+int amdgpu_sdma_sysfs_reset_mask_init(struct amdgpu_device *adev);
+void amdgpu_sdma_sysfs_reset_mask_fini(struct amdgpu_device *adev);
+bool amdgpu_sdma_is_shared_inv_eng(struct amdgpu_device *adev, struct amdgpu_ring *ring);
+struct amdgpu_ring *amdgpu_sdma_get_shared_ring(struct amdgpu_device *adev,
+	struct amdgpu_ring *ring);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_securedisplay.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_securedisplay.c
index 8ed0e073656f..3739be1b71e0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_securedisplay.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_securedisplay.c
@@ -135,6 +135,10 @@ static ssize_t amdgpu_securedisplay_debugfs_write(struct file *f, const char __u
 		mutex_unlock(&psp->securedisplay_context.mutex);
 		break;
 	case 2:
+		if (size < 3 || phy_id >= TA_SECUREDISPLAY_MAX_PHY) {
+			dev_err(adev->dev, "Invalid input: %s\n", str);
+			return -EINVAL;
+		}
 		mutex_lock(&psp->securedisplay_context.mutex);
 		psp_prep_securedisplay_cmd_buf(psp, &securedisplay_cmd,
 			TA_SECUREDISPLAY_COMMAND__SEND_ROI_CRC);
@@ -155,7 +159,6 @@ static ssize_t amdgpu_securedisplay_debugfs_write(struct file *f, const char __u
 		dev_err(adev->dev, "Invalid input: %s\n", str);
 	}
 
-	pm_runtime_mark_last_busy(dev->dev);
 	pm_runtime_put_autosuspend(dev->dev);
 
 	return size;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c
new file mode 100644
index 000000000000..a0b479d5fff1
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.c
@@ -0,0 +1,262 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+#include "amdgpu_seq64.h"
+
+#include <drm/drm_exec.h>
+
+/**
+ * DOC: amdgpu_seq64
+ *
+ * amdgpu_seq64 allocates a 64bit memory on each request in sequence order.
+ * seq64 driver is required for user queue fence memory allocation, TLB
+ * counters and VM updates. It has maximum count of 32768 64 bit slots.
+ */
+
+/**
+ * amdgpu_seq64_get_va_base - Get the seq64 va base address
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Returns:
+ * va base address on success
+ */
+static inline u64 amdgpu_seq64_get_va_base(struct amdgpu_device *adev)
+{
+	u64 addr = AMDGPU_VA_RESERVED_SEQ64_START(adev);
+
+	addr = amdgpu_gmc_sign_extend(addr);
+
+	return addr;
+}
+
+/**
+ * amdgpu_seq64_map - Map the seq64 memory to VM
+ *
+ * @adev: amdgpu_device pointer
+ * @vm: vm pointer
+ * @bo_va: bo_va pointer
+ *
+ * Map the seq64 memory to the given VM.
+ *
+ * Returns:
+ * 0 on success or a negative error code on failure
+ */
+int amdgpu_seq64_map(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+		     struct amdgpu_bo_va **bo_va)
+{
+	struct amdgpu_bo *bo;
+	struct drm_exec exec;
+	u64 seq64_addr;
+	int r;
+
+	bo = adev->seq64.sbo;
+	if (!bo)
+		return -EINVAL;
+
+	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT, 0);
+	drm_exec_until_all_locked(&exec) {
+		r = amdgpu_vm_lock_pd(vm, &exec, 0);
+		if (likely(!r))
+			r = drm_exec_lock_obj(&exec, &bo->tbo.base);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(r))
+			goto error;
+	}
+
+	*bo_va = amdgpu_vm_bo_add(adev, vm, bo);
+	if (!*bo_va) {
+		r = -ENOMEM;
+		goto error;
+	}
+
+	seq64_addr = amdgpu_seq64_get_va_base(adev) & AMDGPU_GMC_HOLE_MASK;
+
+	r = amdgpu_vm_bo_map(adev, *bo_va, seq64_addr, 0,
+			     AMDGPU_VA_RESERVED_SEQ64_SIZE,
+			     AMDGPU_VM_PAGE_READABLE | AMDGPU_VM_MTYPE_UC);
+	if (r) {
+		DRM_ERROR("failed to do bo_map on userq sem, err=%d\n", r);
+		amdgpu_vm_bo_del(adev, *bo_va);
+		goto error;
+	}
+
+	r = amdgpu_vm_bo_update(adev, *bo_va, false);
+	if (r) {
+		DRM_ERROR("failed to do vm_bo_update on userq sem\n");
+		amdgpu_vm_bo_del(adev, *bo_va);
+		goto error;
+	}
+
+error:
+	drm_exec_fini(&exec);
+	return r;
+}
+
+/**
+ * amdgpu_seq64_unmap - Unmap the seq64 memory
+ *
+ * @adev: amdgpu_device pointer
+ * @fpriv: DRM file private
+ *
+ * Unmap the seq64 memory from the given VM.
+ */
+void amdgpu_seq64_unmap(struct amdgpu_device *adev, struct amdgpu_fpriv *fpriv)
+{
+	struct amdgpu_vm *vm;
+	struct amdgpu_bo *bo;
+	struct drm_exec exec;
+	int r;
+
+	if (!fpriv->seq64_va)
+		return;
+
+	bo = adev->seq64.sbo;
+	if (!bo)
+		return;
+
+	vm = &fpriv->vm;
+
+	drm_exec_init(&exec, 0, 0);
+	drm_exec_until_all_locked(&exec) {
+		r = amdgpu_vm_lock_pd(vm, &exec, 0);
+		if (likely(!r))
+			r = drm_exec_lock_obj(&exec, &bo->tbo.base);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(r))
+			goto error;
+	}
+
+	amdgpu_vm_bo_del(adev, fpriv->seq64_va);
+
+	fpriv->seq64_va = NULL;
+
+error:
+	drm_exec_fini(&exec);
+}
+
+/**
+ * amdgpu_seq64_alloc - Allocate a 64 bit memory
+ *
+ * @adev: amdgpu_device pointer
+ * @va: VA to access the seq in process address space
+ * @gpu_addr: GPU address to access the seq
+ * @cpu_addr: CPU address to access the seq
+ *
+ * Alloc a 64 bit memory from seq64 pool.
+ *
+ * Returns:
+ * 0 on success or a negative error code on failure
+ */
+int amdgpu_seq64_alloc(struct amdgpu_device *adev, u64 *va,
+		       u64 *gpu_addr, u64 **cpu_addr)
+{
+	unsigned long bit_pos;
+
+	bit_pos = find_first_zero_bit(adev->seq64.used, adev->seq64.num_sem);
+	if (bit_pos >= adev->seq64.num_sem)
+		return -ENOSPC;
+
+	__set_bit(bit_pos, adev->seq64.used);
+
+	*va = bit_pos * sizeof(u64) + amdgpu_seq64_get_va_base(adev);
+
+	if (gpu_addr)
+		*gpu_addr = bit_pos * sizeof(u64) + adev->seq64.gpu_addr;
+
+	*cpu_addr = bit_pos + adev->seq64.cpu_base_addr;
+
+	return 0;
+}
+
+/**
+ * amdgpu_seq64_free - Free the given 64 bit memory
+ *
+ * @adev: amdgpu_device pointer
+ * @va: gpu start address to be freed
+ *
+ * Free the given 64 bit memory from seq64 pool.
+ */
+void amdgpu_seq64_free(struct amdgpu_device *adev, u64 va)
+{
+	unsigned long bit_pos;
+
+	bit_pos = (va - amdgpu_seq64_get_va_base(adev)) / sizeof(u64);
+	if (bit_pos < adev->seq64.num_sem)
+		__clear_bit(bit_pos, adev->seq64.used);
+}
+
+/**
+ * amdgpu_seq64_fini - Cleanup seq64 driver
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Free the memory space allocated for seq64.
+ *
+ */
+void amdgpu_seq64_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_free_kernel(&adev->seq64.sbo,
+			      NULL,
+			      (void **)&adev->seq64.cpu_base_addr);
+}
+
+/**
+ * amdgpu_seq64_init - Initialize seq64 driver
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Allocate the required memory space for seq64.
+ *
+ * Returns:
+ * 0 on success or a negative error code on failure
+ */
+int amdgpu_seq64_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (adev->seq64.sbo)
+		return 0;
+
+	/*
+	 * AMDGPU_MAX_SEQ64_SLOTS * sizeof(u64) * 8 = AMDGPU_MAX_SEQ64_SLOTS
+	 * 64bit slots
+	 */
+	r = amdgpu_bo_create_kernel(adev, AMDGPU_VA_RESERVED_SEQ64_SIZE,
+				    PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT,
+				    &adev->seq64.sbo, &adev->seq64.gpu_addr,
+				    (void **)&adev->seq64.cpu_base_addr);
+	if (r) {
+		dev_warn(adev->dev, "(%d) create seq64 failed\n", r);
+		return r;
+	}
+
+	memset(adev->seq64.cpu_base_addr, 0, AMDGPU_VA_RESERVED_SEQ64_SIZE);
+
+	adev->seq64.num_sem = AMDGPU_MAX_SEQ64_SLOTS;
+	memset(&adev->seq64.used, 0, sizeof(adev->seq64.used));
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.h
new file mode 100644
index 000000000000..26a249aaaee1
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_seq64.h
@@ -0,0 +1,49 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __AMDGPU_SEQ64_H__
+#define __AMDGPU_SEQ64_H__
+
+#include "amdgpu_vm.h"
+
+#define AMDGPU_MAX_SEQ64_SLOTS         (AMDGPU_VA_RESERVED_SEQ64_SIZE / sizeof(u64))
+
+struct amdgpu_seq64 {
+	struct amdgpu_bo *sbo;
+	u32 num_sem;
+	u64 gpu_addr;
+	u64 *cpu_base_addr;
+	DECLARE_BITMAP(used, AMDGPU_MAX_SEQ64_SLOTS);
+};
+
+void amdgpu_seq64_fini(struct amdgpu_device *adev);
+int amdgpu_seq64_init(struct amdgpu_device *adev);
+int amdgpu_seq64_alloc(struct amdgpu_device *adev, u64 *va, u64 *gpu_addr, u64 **cpu_addr);
+void amdgpu_seq64_free(struct amdgpu_device *adev, u64 gpu_addr);
+int amdgpu_seq64_map(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+		     struct amdgpu_bo_va **bo_va);
+void amdgpu_seq64_unmap(struct amdgpu_device *adev, struct amdgpu_fpriv *fpriv);
+
+#endif
+
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h
index c7a823f3f2c5..ec9d12f85f39 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_smuio.h
@@ -23,6 +23,18 @@
 #ifndef __AMDGPU_SMUIO_H__
 #define __AMDGPU_SMUIO_H__
 
+enum amdgpu_pkg_type {
+	AMDGPU_PKG_TYPE_APU = 2,
+	AMDGPU_PKG_TYPE_CEM = 3,
+	AMDGPU_PKG_TYPE_OAM = 4,
+	AMDGPU_PKG_TYPE_UNKNOWN,
+};
+
+struct amdgpu_smuio_mcm_config_info {
+	int socket_id;
+	int die_id;
+};
+
 struct amdgpu_smuio_funcs {
 	u32 (*get_rom_index_offset)(struct amdgpu_device *adev);
 	u32 (*get_rom_data_offset)(struct amdgpu_device *adev);
@@ -30,7 +42,9 @@ struct amdgpu_smuio_funcs {
 	void (*get_clock_gating_state)(struct amdgpu_device *adev, u64 *flags);
 	u32 (*get_die_id)(struct amdgpu_device *adev);
 	u32 (*get_socket_id)(struct amdgpu_device *adev);
+	enum amdgpu_pkg_type (*get_pkg_type)(struct amdgpu_device *adev);
 	bool (*is_host_gpu_xgmi_supported)(struct amdgpu_device *adev);
+	u64 (*get_gpu_clock_counter)(struct amdgpu_device *adev);
 };
 
 struct amdgpu_smuio {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
index dcd8c066bc1f..d6ae9974c952 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
@@ -135,11 +135,16 @@ static bool amdgpu_sync_add_later(struct amdgpu_sync *sync, struct dma_fence *f)
 	struct amdgpu_sync_entry *e;
 
 	hash_for_each_possible(sync->fences, e, node, f->context) {
-		if (unlikely(e->fence->context != f->context))
-			continue;
+		if (dma_fence_is_signaled(e->fence)) {
+			dma_fence_put(e->fence);
+			e->fence = dma_fence_get(f);
+			return true;
+		}
 
-		amdgpu_sync_keep_later(&e->fence, f);
-		return true;
+		if (likely(e->fence->context == f->context)) {
+			amdgpu_sync_keep_later(&e->fence, f);
+			return true;
+		}
 	}
 	return false;
 }
@@ -149,10 +154,12 @@ static bool amdgpu_sync_add_later(struct amdgpu_sync *sync, struct dma_fence *f)
  *
  * @sync: sync object to add fence to
  * @f: fence to sync to
+ * @flags: memory allocation flags to use when allocating sync entry
  *
  * Add the fence to the sync object.
  */
-int amdgpu_sync_fence(struct amdgpu_sync *sync, struct dma_fence *f)
+int amdgpu_sync_fence(struct amdgpu_sync *sync, struct dma_fence *f,
+		      gfp_t flags)
 {
 	struct amdgpu_sync_entry *e;
 
@@ -162,7 +169,7 @@ int amdgpu_sync_fence(struct amdgpu_sync *sync, struct dma_fence *f)
 	if (amdgpu_sync_add_later(sync, f))
 		return 0;
 
-	e = kmem_cache_alloc(amdgpu_sync_slab, GFP_KERNEL);
+	e = kmem_cache_alloc(amdgpu_sync_slab, flags);
 	if (!e)
 		return -ENOMEM;
 
@@ -191,7 +198,8 @@ static bool amdgpu_sync_test_fence(struct amdgpu_device *adev,
 
 	/* Never sync to VM updates either. */
 	if (fence_owner == AMDGPU_FENCE_OWNER_VM &&
-	    owner != AMDGPU_FENCE_OWNER_UNDEFINED)
+	    owner != AMDGPU_FENCE_OWNER_UNDEFINED &&
+	    owner != AMDGPU_FENCE_OWNER_KFD)
 		return false;
 
 	/* Ignore fences depending on the sync mode */
@@ -241,14 +249,13 @@ int amdgpu_sync_resv(struct amdgpu_device *adev, struct amdgpu_sync *sync,
 
 	if (resv == NULL)
 		return -EINVAL;
-
-	/* TODO: Use DMA_RESV_USAGE_READ here */
-	dma_resv_for_each_fence(&cursor, resv, DMA_RESV_USAGE_BOOKKEEP, f) {
+	/* Implicitly sync only to KERNEL, WRITE and READ */
+	dma_resv_for_each_fence(&cursor, resv, DMA_RESV_USAGE_READ, f) {
 		dma_fence_chain_for_each(f, f) {
 			struct dma_fence *tmp = dma_fence_chain_contained(f);
 
 			if (amdgpu_sync_test_fence(adev, mode, owner, tmp)) {
-				r = amdgpu_sync_fence(sync, f);
+				r = amdgpu_sync_fence(sync, f, GFP_KERNEL);
 				dma_fence_put(f);
 				if (r)
 					return r;
@@ -259,6 +266,36 @@ int amdgpu_sync_resv(struct amdgpu_device *adev, struct amdgpu_sync *sync,
 	return 0;
 }
 
+/**
+ * amdgpu_sync_kfd - sync to KFD fences
+ *
+ * @sync: sync object to add KFD fences to
+ * @resv: reservation object with KFD fences
+ *
+ * Extract all KFD fences and add them to the sync object.
+ */
+int amdgpu_sync_kfd(struct amdgpu_sync *sync, struct dma_resv *resv)
+{
+	struct dma_resv_iter cursor;
+	struct dma_fence *f;
+	int r = 0;
+
+	dma_resv_iter_begin(&cursor, resv, DMA_RESV_USAGE_BOOKKEEP);
+	dma_resv_for_each_fence_unlocked(&cursor, f) {
+		void *fence_owner = amdgpu_sync_get_owner(f);
+
+		if (fence_owner != AMDGPU_FENCE_OWNER_KFD)
+			continue;
+
+		r = amdgpu_sync_fence(sync, f, GFP_KERNEL);
+		if (r)
+			break;
+	}
+	dma_resv_iter_end(&cursor);
+
+	return r;
+}
+
 /* Free the entry back to the slab */
 static void amdgpu_sync_entry_free(struct amdgpu_sync_entry *e)
 {
@@ -357,7 +394,7 @@ int amdgpu_sync_clone(struct amdgpu_sync *source, struct amdgpu_sync *clone)
 	hash_for_each_safe(source->fences, i, tmp, e, node) {
 		f = e->fence;
 		if (!dma_fence_is_signaled(f)) {
-			r = amdgpu_sync_fence(clone, f);
+			r = amdgpu_sync_fence(clone, f, GFP_KERNEL);
 			if (r)
 				return r;
 		} else {
@@ -369,6 +406,25 @@ int amdgpu_sync_clone(struct amdgpu_sync *source, struct amdgpu_sync *clone)
 }
 
 /**
+ * amdgpu_sync_move - move all fences from src to dst
+ *
+ * @src: source of the fences, empty after function
+ * @dst: destination for the fences
+ *
+ * Moves all fences from source to destination. All fences in destination are
+ * freed and source is empty after the function call.
+ */
+void amdgpu_sync_move(struct amdgpu_sync *src, struct amdgpu_sync *dst)
+{
+	unsigned int i;
+
+	amdgpu_sync_free(dst);
+
+	for (i = 0; i < HASH_SIZE(src->fences); ++i)
+		hlist_move_list(&src->fences[i], &dst->fences[i]);
+}
+
+/**
  * amdgpu_sync_push_to_job - push fences into job
  * @sync: sync object to get the fences from
  * @job: job to push the fences into
@@ -440,9 +496,7 @@ void amdgpu_sync_free(struct amdgpu_sync *sync)
  */
 int amdgpu_sync_init(void)
 {
-	amdgpu_sync_slab = kmem_cache_create(
-		"amdgpu_sync", sizeof(struct amdgpu_sync_entry), 0,
-		SLAB_HWCACHE_ALIGN, NULL);
+	amdgpu_sync_slab = KMEM_CACHE(amdgpu_sync_entry, SLAB_HWCACHE_ALIGN);
 	if (!amdgpu_sync_slab)
 		return -ENOMEM;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h
index cf1e9e858efd..51eb4382c91e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.h
@@ -47,14 +47,17 @@ struct amdgpu_sync {
 };
 
 void amdgpu_sync_create(struct amdgpu_sync *sync);
-int amdgpu_sync_fence(struct amdgpu_sync *sync, struct dma_fence *f);
+int amdgpu_sync_fence(struct amdgpu_sync *sync, struct dma_fence *f,
+		      gfp_t flags);
 int amdgpu_sync_resv(struct amdgpu_device *adev, struct amdgpu_sync *sync,
 		     struct dma_resv *resv, enum amdgpu_sync_mode mode,
 		     void *owner);
+int amdgpu_sync_kfd(struct amdgpu_sync *sync, struct dma_resv *resv);
 struct dma_fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync,
 				     struct amdgpu_ring *ring);
 struct dma_fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync);
 int amdgpu_sync_clone(struct amdgpu_sync *source, struct amdgpu_sync *clone);
+void amdgpu_sync_move(struct amdgpu_sync *src, struct amdgpu_sync *dst);
 int amdgpu_sync_push_to_job(struct amdgpu_sync *sync, struct amdgpu_job *job);
 int amdgpu_sync_wait(struct amdgpu_sync *sync, bool intr);
 void amdgpu_sync_free(struct amdgpu_sync *sync);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h
index 98d91ebf5c26..d13e64a69e25 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h
@@ -167,25 +167,23 @@ TRACE_EVENT(amdgpu_cs_ioctl,
 	    TP_PROTO(struct amdgpu_job *job),
 	    TP_ARGS(job),
 	    TP_STRUCT__entry(
-			     __field(uint64_t, sched_job_id)
 			     __string(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job))
-			     __field(unsigned int, context)
-			     __field(unsigned int, seqno)
+			     __field(u64, context)
+			     __field(u64, seqno)
 			     __field(struct dma_fence *, fence)
 			     __string(ring, to_amdgpu_ring(job->base.sched)->name)
 			     __field(u32, num_ibs)
 			     ),
 
 	    TP_fast_assign(
-			   __entry->sched_job_id = job->base.id;
-			   __assign_str(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job));
+			   __assign_str(timeline);
 			   __entry->context = job->base.s_fence->finished.context;
 			   __entry->seqno = job->base.s_fence->finished.seqno;
-			   __assign_str(ring, to_amdgpu_ring(job->base.sched)->name);
+			   __assign_str(ring);
 			   __entry->num_ibs = job->num_ibs;
 			   ),
-	    TP_printk("sched_job=%llu, timeline=%s, context=%u, seqno=%u, ring_name=%s, num_ibs=%u",
-		      __entry->sched_job_id, __get_str(timeline), __entry->context,
+	    TP_printk("timeline=%s, fence=%llu:%llu, ring_name=%s, num_ibs=%u",
+		      __get_str(timeline), __entry->context,
 		      __entry->seqno, __get_str(ring), __entry->num_ibs)
 );
 
@@ -193,24 +191,22 @@ TRACE_EVENT(amdgpu_sched_run_job,
 	    TP_PROTO(struct amdgpu_job *job),
 	    TP_ARGS(job),
 	    TP_STRUCT__entry(
-			     __field(uint64_t, sched_job_id)
 			     __string(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job))
-			     __field(unsigned int, context)
-			     __field(unsigned int, seqno)
+			     __field(u64, context)
+			     __field(u64, seqno)
 			     __string(ring, to_amdgpu_ring(job->base.sched)->name)
 			     __field(u32, num_ibs)
 			     ),
 
 	    TP_fast_assign(
-			   __entry->sched_job_id = job->base.id;
-			   __assign_str(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job));
+			   __assign_str(timeline);
 			   __entry->context = job->base.s_fence->finished.context;
 			   __entry->seqno = job->base.s_fence->finished.seqno;
-			   __assign_str(ring, to_amdgpu_ring(job->base.sched)->name);
+			   __assign_str(ring);
 			   __entry->num_ibs = job->num_ibs;
 			   ),
-	    TP_printk("sched_job=%llu, timeline=%s, context=%u, seqno=%u, ring_name=%s, num_ibs=%u",
-		      __entry->sched_job_id, __get_str(timeline), __entry->context,
+	    TP_printk("timeline=%s, fence=%llu:%llu, ring_name=%s, num_ibs=%u",
+		      __get_str(timeline), __entry->context,
 		      __entry->seqno, __get_str(ring), __entry->num_ibs)
 );
 
@@ -231,9 +227,9 @@ TRACE_EVENT(amdgpu_vm_grab_id,
 
 	    TP_fast_assign(
 			   __entry->pasid = vm->pasid;
-			   __assign_str(ring, ring->name);
+			   __assign_str(ring);
 			   __entry->vmid = job->vmid;
-			   __entry->vm_hub = ring->funcs->vmhub,
+			   __entry->vm_hub = ring->vm_hub,
 			   __entry->pd_addr = job->vm_pd_addr;
 			   __entry->needs_flush = job->vm_needs_flush;
 			   ),
@@ -425,14 +421,14 @@ TRACE_EVENT(amdgpu_vm_flush,
 			     ),
 
 	    TP_fast_assign(
-			   __assign_str(ring, ring->name);
+			   __assign_str(ring);
 			   __entry->vmid = vmid;
-			   __entry->vm_hub = ring->funcs->vmhub;
+			   __entry->vm_hub = ring->vm_hub;
 			   __entry->pd_addr = pd_addr;
 			   ),
 	    TP_printk("ring=%s, id=%u, hub=%u, pd_addr=%010Lx",
 		      __get_str(ring), __entry->vmid,
-		      __entry->vm_hub,__entry->pd_addr)
+		      __entry->vm_hub, __entry->pd_addr)
 );
 
 DECLARE_EVENT_CLASS(amdgpu_pasid,
@@ -457,6 +453,38 @@ DEFINE_EVENT(amdgpu_pasid, amdgpu_pasid_freed,
 	    TP_ARGS(pasid)
 );
 
+TRACE_EVENT(amdgpu_isolation,
+	    TP_PROTO(void *prev, void *next),
+	    TP_ARGS(prev, next),
+	    TP_STRUCT__entry(
+			     __field(void *, prev)
+			     __field(void *, next)
+			     ),
+
+	    TP_fast_assign(
+			   __entry->prev = prev;
+			   __entry->next = next;
+			   ),
+	    TP_printk("prev=%p, next=%p",
+		      __entry->prev,
+		      __entry->next)
+);
+
+TRACE_EVENT(amdgpu_cleaner_shader,
+	    TP_PROTO(struct amdgpu_ring *ring, struct dma_fence *fence),
+	    TP_ARGS(ring, fence),
+	    TP_STRUCT__entry(
+			     __string(ring, ring->name)
+			     __field(u64, seqno)
+			     ),
+
+	    TP_fast_assign(
+			   __assign_str(ring);
+			   __entry->seqno = fence->seqno;
+			   ),
+	    TP_printk("ring=%s, seqno=%Lu", __get_str(ring), __entry->seqno)
+);
+
 TRACE_EVENT(amdgpu_bo_list_set,
 	    TP_PROTO(struct amdgpu_bo_list *list, struct amdgpu_bo *bo),
 	    TP_ARGS(list, bo),
@@ -494,7 +522,7 @@ TRACE_EVENT(amdgpu_cs_bo_status,
 );
 
 TRACE_EVENT(amdgpu_bo_move,
-	    TP_PROTO(struct amdgpu_bo* bo, uint32_t new_placement, uint32_t old_placement),
+	    TP_PROTO(struct amdgpu_bo *bo, uint32_t new_placement, uint32_t old_placement),
 	    TP_ARGS(bo, new_placement, old_placement),
 	    TP_STRUCT__entry(
 			__field(struct amdgpu_bo *, bo)
@@ -519,23 +547,19 @@ TRACE_EVENT(amdgpu_ib_pipe_sync,
 	    TP_ARGS(sched_job, fence),
 	    TP_STRUCT__entry(
 			     __string(ring, sched_job->base.sched->name)
-			     __field(uint64_t, id)
 			     __field(struct dma_fence *, fence)
-			     __field(uint64_t, ctx)
-			     __field(unsigned, seqno)
+			     __field(u64, ctx)
+			     __field(u64, seqno)
 			     ),
 
 	    TP_fast_assign(
-			   __assign_str(ring, sched_job->base.sched->name);
-			   __entry->id = sched_job->base.id;
+			   __assign_str(ring);
 			   __entry->fence = fence;
 			   __entry->ctx = fence->context;
 			   __entry->seqno = fence->seqno;
 			   ),
-	    TP_printk("job ring=%s, id=%llu, need pipe sync to fence=%p, context=%llu, seq=%u",
-		      __get_str(ring), __entry->id,
-		      __entry->fence, __entry->ctx,
-		      __entry->seqno)
+	    TP_printk("job ring=%s need pipe sync to fence=%llu:%llu",
+		      __get_str(ring), __entry->ctx, __entry->seqno)
 );
 
 TRACE_EVENT(amdgpu_reset_reg_dumps,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 2cd081cbf706..2b931e855abd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -38,7 +38,6 @@
 #include <linux/seq_file.h>
 #include <linux/slab.h>
 #include <linux/swap.h>
-#include <linux/swiotlb.h>
 #include <linux/dma-buf.h>
 #include <linux/sizes.h>
 #include <linux/module.h>
@@ -50,7 +49,6 @@
 #include <drm/ttm/ttm_tt.h>
 
 #include <drm/amdgpu_drm.h>
-#include <drm/drm_drv.h>
 
 #include "amdgpu.h"
 #include "amdgpu_object.h"
@@ -63,9 +61,9 @@
 #include "amdgpu_res_cursor.h"
 #include "bif/bif_4_1_d.h"
 
-MODULE_IMPORT_NS(DMA_BUF);
+MODULE_IMPORT_NS("DMA_BUF");
 
-#define AMDGPU_TTM_VRAM_MAX_DW_READ	(size_t)128
+#define AMDGPU_TTM_VRAM_MAX_DW_READ	((size_t)128)
 
 static int amdgpu_ttm_backend_bind(struct ttm_device *bdev,
 				   struct ttm_tt *ttm,
@@ -104,23 +102,19 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
 	/* Don't handle scatter gather BOs */
 	if (bo->type == ttm_bo_type_sg) {
 		placement->num_placement = 0;
-		placement->num_busy_placement = 0;
 		return;
 	}
 
 	/* Object isn't an AMDGPU object so ignore */
 	if (!amdgpu_bo_is_amdgpu_bo(bo)) {
 		placement->placement = &placements;
-		placement->busy_placement = &placements;
 		placement->num_placement = 1;
-		placement->num_busy_placement = 1;
 		return;
 	}
 
 	abo = ttm_to_amdgpu_bo(bo);
 	if (abo->flags & AMDGPU_GEM_CREATE_DISCARDABLE) {
 		placement->num_placement = 0;
-		placement->num_busy_placement = 0;
 		return;
 	}
 
@@ -128,17 +122,19 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
 	case AMDGPU_PL_GDS:
 	case AMDGPU_PL_GWS:
 	case AMDGPU_PL_OA:
+	case AMDGPU_PL_DOORBELL:
+	case AMDGPU_PL_MMIO_REMAP:
 		placement->num_placement = 0;
-		placement->num_busy_placement = 0;
 		return;
 
 	case TTM_PL_VRAM:
 		if (!adev->mman.buffer_funcs_enabled) {
 			/* Move to system memory */
 			amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_CPU);
+
 		} else if (!amdgpu_gmc_vram_full_visible(&adev->gmc) &&
 			   !(abo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) &&
-			   amdgpu_bo_in_cpu_visible_vram(abo)) {
+			   amdgpu_res_cpu_visible(adev, bo->resource)) {
 
 			/* Try evicting to the CPU inaccessible part of VRAM
 			 * first, but only set GTT as busy placement, so this
@@ -150,8 +146,7 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
 							AMDGPU_GEM_DOMAIN_CPU);
 			abo->placements[0].fpfn = adev->gmc.visible_vram_size >> PAGE_SHIFT;
 			abo->placements[0].lpfn = 0;
-			abo->placement.busy_placement = &abo->placements[1];
-			abo->placement.num_busy_placement = 1;
+			abo->placements[0].flags |= TTM_PL_FLAG_DESIRED;
 		} else {
 			/* Move to GTT memory */
 			amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_GTT |
@@ -184,16 +179,15 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
 static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo,
 				 struct ttm_resource *mem,
 				 struct amdgpu_res_cursor *mm_cur,
-				 unsigned window, struct amdgpu_ring *ring,
+				 unsigned int window, struct amdgpu_ring *ring,
 				 bool tmz, uint64_t *size, uint64_t *addr)
 {
 	struct amdgpu_device *adev = ring->adev;
-	unsigned offset, num_pages, num_dw, num_bytes;
+	unsigned int offset, num_pages, num_dw, num_bytes;
 	uint64_t src_addr, dst_addr;
 	struct amdgpu_job *job;
 	void *cpu_addr;
 	uint64_t flags;
-	unsigned int i;
 	int r;
 
 	BUG_ON(adev->mman.buffer_funcs->copy_max_bytes <
@@ -229,10 +223,11 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo,
 	num_dw = ALIGN(adev->mman.buffer_funcs->copy_num_dw, 8);
 	num_bytes = num_pages * 8 * AMDGPU_GPU_PAGES_IN_CPU_PAGE;
 
-	r = amdgpu_job_alloc_with_ib(adev, &adev->mman.entity,
+	r = amdgpu_job_alloc_with_ib(adev, &adev->mman.high_pr,
 				     AMDGPU_FENCE_OWNER_UNDEFINED,
 				     num_dw * 4 + num_bytes,
-				     AMDGPU_IB_POOL_DELAYED, &job);
+				     AMDGPU_IB_POOL_DELAYED, &job,
+				     AMDGPU_KERNEL_JOB_ID_TTM_MAP_BUFFER);
 	if (r)
 		return r;
 
@@ -242,7 +237,7 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo,
 	dst_addr = amdgpu_bo_gpu_offset(adev->gart.bo);
 	dst_addr += window * AMDGPU_GTT_MAX_TRANSFER_SIZE * 8;
 	amdgpu_emit_copy_buffer(adev, &job->ibs[0], src_addr,
-				dst_addr, num_bytes, false);
+				dst_addr, num_bytes, 0);
 
 	amdgpu_ring_pad_ib(ring, &job->ibs[0]);
 	WARN_ON(job->ibs[0].length_dw > num_dw);
@@ -259,16 +254,9 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo,
 		dma_addr = &bo->ttm->dma_address[mm_cur->start >> PAGE_SHIFT];
 		amdgpu_gart_map(adev, 0, num_pages, dma_addr, flags, cpu_addr);
 	} else {
-		dma_addr_t dma_address;
-
-		dma_address = mm_cur->start;
-		dma_address += adev->vm_manager.vram_base_offset;
+		u64 pa = mm_cur->start + adev->vm_manager.vram_base_offset;
 
-		for (i = 0; i < num_pages; ++i) {
-			amdgpu_gart_map(adev, i << PAGE_SHIFT, 1, &dma_address,
-					flags, cpu_addr);
-			dma_address += PAGE_SIZE;
-		}
+		amdgpu_gart_map_vram_range(adev, pa, 0, num_pages, flags, cpu_addr);
 	}
 
 	dma_fence_put(amdgpu_job_submit(job));
@@ -290,20 +278,24 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo,
  * move and different for a BO to BO copy.
  *
  */
-int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
-			       const struct amdgpu_copy_mem *src,
-			       const struct amdgpu_copy_mem *dst,
-			       uint64_t size, bool tmz,
-			       struct dma_resv *resv,
-			       struct dma_fence **f)
+__attribute__((nonnull))
+static int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
+				      const struct amdgpu_copy_mem *src,
+				      const struct amdgpu_copy_mem *dst,
+				      uint64_t size, bool tmz,
+				      struct dma_resv *resv,
+				      struct dma_fence **f)
 {
 	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
 	struct amdgpu_res_cursor src_mm, dst_mm;
 	struct dma_fence *fence = NULL;
 	int r = 0;
+	uint32_t copy_flags = 0;
+	struct amdgpu_bo *abo_src, *abo_dst;
 
 	if (!adev->mman.buffer_funcs_enabled) {
-		DRM_ERROR("Trying to move memory with ring turned off.\n");
+		dev_err(adev->dev,
+			"Trying to move memory with ring turned off.\n");
 		return -EINVAL;
 	}
 
@@ -312,7 +304,8 @@ int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
 
 	mutex_lock(&adev->mman.gtt_window_lock);
 	while (src_mm.remaining) {
-		uint64_t from, to, cur_size;
+		uint64_t from, to, cur_size, tiling_flags;
+		uint32_t num_type, data_format, max_com, write_compress_disable;
 		struct dma_fence *next;
 
 		/* Never copy more than 256MiB at once to avoid a timeout */
@@ -329,8 +322,31 @@ int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
 		if (r)
 			goto error;
 
-		r = amdgpu_copy_buffer(ring, from, to, cur_size,
-				       resv, &next, false, true, tmz);
+		abo_src = ttm_to_amdgpu_bo(src->bo);
+		abo_dst = ttm_to_amdgpu_bo(dst->bo);
+		if (tmz)
+			copy_flags |= AMDGPU_COPY_FLAGS_TMZ;
+		if ((abo_src->flags & AMDGPU_GEM_CREATE_GFX12_DCC) &&
+		    (abo_src->tbo.resource->mem_type == TTM_PL_VRAM))
+			copy_flags |= AMDGPU_COPY_FLAGS_READ_DECOMPRESSED;
+		if ((abo_dst->flags & AMDGPU_GEM_CREATE_GFX12_DCC) &&
+		    (dst->mem->mem_type == TTM_PL_VRAM)) {
+			copy_flags |= AMDGPU_COPY_FLAGS_WRITE_COMPRESSED;
+			amdgpu_bo_get_tiling_flags(abo_dst, &tiling_flags);
+			max_com = AMDGPU_TILING_GET(tiling_flags, GFX12_DCC_MAX_COMPRESSED_BLOCK);
+			num_type = AMDGPU_TILING_GET(tiling_flags, GFX12_DCC_NUMBER_TYPE);
+			data_format = AMDGPU_TILING_GET(tiling_flags, GFX12_DCC_DATA_FORMAT);
+			write_compress_disable =
+				AMDGPU_TILING_GET(tiling_flags, GFX12_DCC_WRITE_COMPRESS_DISABLE);
+			copy_flags |= (AMDGPU_COPY_FLAGS_SET(MAX_COMPRESSED, max_com) |
+				       AMDGPU_COPY_FLAGS_SET(NUMBER_TYPE, num_type) |
+				       AMDGPU_COPY_FLAGS_SET(DATA_FORMAT, data_format) |
+				       AMDGPU_COPY_FLAGS_SET(WRITE_COMPRESS_DISABLE,
+							     write_compress_disable));
+		}
+
+		r = amdgpu_copy_buffer(ring, from, to, cur_size, resv,
+				       &next, false, true, copy_flags);
 		if (r)
 			goto error;
 
@@ -342,9 +358,7 @@ int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
 	}
 error:
 	mutex_unlock(&adev->mman.gtt_window_lock);
-	if (f)
-		*f = dma_fence_get(fence);
-	dma_fence_put(fence);
+	*f = fence;
 	return r;
 }
 
@@ -384,10 +398,12 @@ static int amdgpu_move_blit(struct ttm_buffer_object *bo,
 	    (abo->flags & AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE)) {
 		struct dma_fence *wipe_fence = NULL;
 
-		r = amdgpu_fill_buffer(abo, AMDGPU_POISON, NULL, &wipe_fence);
+		r = amdgpu_fill_buffer(abo, 0, NULL, &wipe_fence,
+				       false, AMDGPU_KERNEL_JOB_ID_MOVE_BLIT);
 		if (r) {
 			goto error;
 		} else if (wipe_fence) {
+			amdgpu_vram_mgr_set_cleared(bo->resource);
 			dma_fence_put(fence);
 			fence = wipe_fence;
 		}
@@ -408,40 +424,56 @@ error:
 	return r;
 }
 
-/*
- * amdgpu_mem_visible - Check that memory can be accessed by ttm_bo_move_memcpy
+/**
+ * amdgpu_res_cpu_visible - Check that resource can be accessed by CPU
+ * @adev: amdgpu device
+ * @res: the resource to check
  *
- * Called by amdgpu_bo_move()
+ * Returns: true if the full resource is CPU visible, false otherwise.
  */
-static bool amdgpu_mem_visible(struct amdgpu_device *adev,
-			       struct ttm_resource *mem)
+bool amdgpu_res_cpu_visible(struct amdgpu_device *adev,
+			    struct ttm_resource *res)
 {
-	u64 mem_size = (u64)mem->size;
 	struct amdgpu_res_cursor cursor;
-	u64 end;
 
-	if (mem->mem_type == TTM_PL_SYSTEM ||
-	    mem->mem_type == TTM_PL_TT)
+	if (!res)
+		return false;
+
+	if (res->mem_type == TTM_PL_SYSTEM || res->mem_type == TTM_PL_TT ||
+	    res->mem_type == AMDGPU_PL_PREEMPT || res->mem_type == AMDGPU_PL_DOORBELL ||
+	    res->mem_type == AMDGPU_PL_MMIO_REMAP)
 		return true;
-	if (mem->mem_type != TTM_PL_VRAM)
+
+	if (res->mem_type != TTM_PL_VRAM)
 		return false;
 
-	amdgpu_res_first(mem, 0, mem_size, &cursor);
-	end = cursor.start + cursor.size;
+	amdgpu_res_first(res, 0, res->size, &cursor);
 	while (cursor.remaining) {
+		if ((cursor.start + cursor.size) > adev->gmc.visible_vram_size)
+			return false;
 		amdgpu_res_next(&cursor, cursor.size);
+	}
 
-		if (!cursor.remaining)
-			break;
+	return true;
+}
 
-		/* ttm_resource_ioremap only supports contiguous memory */
-		if (end != cursor.start)
-			return false;
+/*
+ * amdgpu_res_copyable - Check that memory can be accessed by ttm_bo_move_memcpy
+ *
+ * Called by amdgpu_bo_move()
+ */
+static bool amdgpu_res_copyable(struct amdgpu_device *adev,
+				struct ttm_resource *mem)
+{
+	if (!amdgpu_res_cpu_visible(adev, mem))
+		return false;
 
-		end = cursor.start + cursor.size;
-	}
+	/* ttm_resource_ioremap only supports contiguous memory */
+	if (mem->mem_type == TTM_PL_VRAM &&
+	    !(mem->placement & TTM_PL_FLAG_CONTIGUOUS))
+		return false;
 
-	return end <= adev->gmc.visible_vram_size;
+	return true;
 }
 
 /*
@@ -471,14 +503,16 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict,
 
 	if (!old_mem || (old_mem->mem_type == TTM_PL_SYSTEM &&
 			 bo->ttm == NULL)) {
+		amdgpu_bo_move_notify(bo, evict, new_mem);
 		ttm_bo_move_null(bo, new_mem);
-		goto out;
+		return 0;
 	}
 	if (old_mem->mem_type == TTM_PL_SYSTEM &&
 	    (new_mem->mem_type == TTM_PL_TT ||
 	     new_mem->mem_type == AMDGPU_PL_PREEMPT)) {
+		amdgpu_bo_move_notify(bo, evict, new_mem);
 		ttm_bo_move_null(bo, new_mem);
-		goto out;
+		return 0;
 	}
 	if ((old_mem->mem_type == TTM_PL_TT ||
 	     old_mem->mem_type == AMDGPU_PL_PREEMPT) &&
@@ -488,20 +522,26 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict,
 			return r;
 
 		amdgpu_ttm_backend_unbind(bo->bdev, bo->ttm);
+		amdgpu_bo_move_notify(bo, evict, new_mem);
 		ttm_resource_free(bo, &bo->resource);
 		ttm_bo_assign_mem(bo, new_mem);
-		goto out;
+		return 0;
 	}
 
 	if (old_mem->mem_type == AMDGPU_PL_GDS ||
 	    old_mem->mem_type == AMDGPU_PL_GWS ||
 	    old_mem->mem_type == AMDGPU_PL_OA ||
+	    old_mem->mem_type == AMDGPU_PL_DOORBELL ||
+	    old_mem->mem_type == AMDGPU_PL_MMIO_REMAP ||
 	    new_mem->mem_type == AMDGPU_PL_GDS ||
 	    new_mem->mem_type == AMDGPU_PL_GWS ||
-	    new_mem->mem_type == AMDGPU_PL_OA) {
+	    new_mem->mem_type == AMDGPU_PL_OA ||
+	    new_mem->mem_type == AMDGPU_PL_DOORBELL ||
+	    new_mem->mem_type == AMDGPU_PL_MMIO_REMAP) {
 		/* Nothing to save here */
+		amdgpu_bo_move_notify(bo, evict, new_mem);
 		ttm_bo_move_null(bo, new_mem);
-		goto out;
+		return 0;
 	}
 
 	if (bo->type == ttm_bo_type_device &&
@@ -513,27 +553,28 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict,
 		abo->flags &= ~AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
 	}
 
-	if (adev->mman.buffer_funcs_enabled) {
-		if (((old_mem->mem_type == TTM_PL_SYSTEM &&
-		      new_mem->mem_type == TTM_PL_VRAM) ||
-		     (old_mem->mem_type == TTM_PL_VRAM &&
-		      new_mem->mem_type == TTM_PL_SYSTEM))) {
-			hop->fpfn = 0;
-			hop->lpfn = 0;
-			hop->mem_type = TTM_PL_TT;
-			hop->flags = TTM_PL_FLAG_TEMPORARY;
-			return -EMULTIHOP;
-		}
+	if (adev->mman.buffer_funcs_enabled &&
+	    ((old_mem->mem_type == TTM_PL_SYSTEM &&
+	      new_mem->mem_type == TTM_PL_VRAM) ||
+	     (old_mem->mem_type == TTM_PL_VRAM &&
+	      new_mem->mem_type == TTM_PL_SYSTEM))) {
+		hop->fpfn = 0;
+		hop->lpfn = 0;
+		hop->mem_type = TTM_PL_TT;
+		hop->flags = TTM_PL_FLAG_TEMPORARY;
+		return -EMULTIHOP;
+	}
 
+	amdgpu_bo_move_notify(bo, evict, new_mem);
+	if (adev->mman.buffer_funcs_enabled)
 		r = amdgpu_move_blit(bo, evict, new_mem, old_mem);
-	} else {
+	else
 		r = -ENODEV;
-	}
 
 	if (r) {
 		/* Check that all memory is CPU accessible */
-		if (!amdgpu_mem_visible(adev, old_mem) ||
-		    !amdgpu_mem_visible(adev, new_mem)) {
+		if (!amdgpu_res_copyable(adev, old_mem) ||
+		    !amdgpu_res_copyable(adev, new_mem)) {
 			pr_err("Move buffer fallback to memcpy unavailable\n");
 			return r;
 		}
@@ -543,10 +584,10 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict,
 			return r;
 	}
 
-out:
-	/* update statistics */
+	/* update statistics after the move */
+	if (evict)
+		atomic64_inc(&adev->num_evictions);
 	atomic64_add(bo->base.size, &adev->num_bytes_moved);
-	amdgpu_bo_move_notify(bo, evict, new_mem);
 	return 0;
 }
 
@@ -559,7 +600,6 @@ static int amdgpu_ttm_io_mem_reserve(struct ttm_device *bdev,
 				     struct ttm_resource *mem)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bdev);
-	size_t bus_size = (size_t)mem->size;
 
 	switch (mem->mem_type) {
 	case TTM_PL_SYSTEM:
@@ -570,9 +610,6 @@ static int amdgpu_ttm_io_mem_reserve(struct ttm_device *bdev,
 		break;
 	case TTM_PL_VRAM:
 		mem->bus.offset = mem->start << PAGE_SHIFT;
-		/* check if it's visible */
-		if ((mem->bus.offset + bus_size) > adev->gmc.visible_vram_size)
-			return -EINVAL;
 
 		if (adev->mman.aper_base_kaddr &&
 		    mem->placement & TTM_PL_FLAG_CONTIGUOUS)
@@ -582,6 +619,18 @@ static int amdgpu_ttm_io_mem_reserve(struct ttm_device *bdev,
 		mem->bus.offset += adev->gmc.aper_base;
 		mem->bus.is_iomem = true;
 		break;
+	case AMDGPU_PL_DOORBELL:
+		mem->bus.offset = mem->start << PAGE_SHIFT;
+		mem->bus.offset += adev->doorbell.base;
+		mem->bus.is_iomem = true;
+		mem->bus.caching = ttm_uncached;
+		break;
+	case AMDGPU_PL_MMIO_REMAP:
+		mem->bus.offset = mem->start << PAGE_SHIFT;
+		mem->bus.offset += adev->rmmio_remap.bus_addr;
+		mem->bus.is_iomem = true;
+		mem->bus.caching = ttm_uncached;
+		break;
 	default:
 		return -EINVAL;
 	}
@@ -596,6 +645,12 @@ static unsigned long amdgpu_ttm_io_mem_pfn(struct ttm_buffer_object *bo,
 
 	amdgpu_res_first(bo->resource, (u64)page_offset << PAGE_SHIFT, 0,
 			 &cursor);
+
+	if (bo->resource->mem_type == AMDGPU_PL_DOORBELL)
+		return ((uint64_t)(adev->doorbell.base + cursor.start)) >> PAGE_SHIFT;
+	else if (bo->resource->mem_type == AMDGPU_PL_MMIO_REMAP)
+		return ((uint64_t)(adev->rmmio_remap.bus_addr + cursor.start)) >> PAGE_SHIFT;
+
 	return (adev->gmc.aper_base + cursor.start) >> PAGE_SHIFT;
 }
 
@@ -631,6 +686,7 @@ struct amdgpu_ttm_tt {
 	struct task_struct	*usertask;
 	uint32_t		userflags;
 	bool			bound;
+	int32_t			pool_id;
 };
 
 #define ttm_to_amdgpu_ttm_tt(ptr)	container_of(ptr, struct amdgpu_ttm_tt, ttm)
@@ -641,10 +697,11 @@ struct amdgpu_ttm_tt {
  * memory and start HMM tracking CPU page table update
  *
  * Calling function must call amdgpu_ttm_tt_userptr_range_done() once and only
- * once afterwards to stop HMM tracking
+ * once afterwards to stop HMM tracking. Its the caller responsibility to ensure
+ * that range is a valid memory and it is freed too.
  */
-int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages,
-				 struct hmm_range **range)
+int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo,
+				 struct amdgpu_hmm_range *range)
 {
 	struct ttm_tt *ttm = bo->tbo.ttm;
 	struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm);
@@ -654,9 +711,6 @@ int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages,
 	bool readonly;
 	int r = 0;
 
-	/* Make sure get_user_pages_done() can cleanup gracefully */
-	*range = NULL;
-
 	mm = bo->notifier.mm;
 	if (unlikely(!mm)) {
 		DRM_DEBUG_DRIVER("BO is not registered?\n");
@@ -680,7 +734,7 @@ int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages,
 
 	readonly = amdgpu_ttm_tt_is_readonly(ttm);
 	r = amdgpu_hmm_range_get_pages(&bo->notifier, start, ttm->num_pages,
-				       readonly, NULL, pages, range);
+				       readonly, NULL, range);
 out_unlock:
 	mmap_read_unlock(mm);
 	if (r)
@@ -691,38 +745,6 @@ out_unlock:
 	return r;
 }
 
-/* amdgpu_ttm_tt_discard_user_pages - Discard range and pfn array allocations
- */
-void amdgpu_ttm_tt_discard_user_pages(struct ttm_tt *ttm,
-				      struct hmm_range *range)
-{
-	struct amdgpu_ttm_tt *gtt = (void *)ttm;
-
-	if (gtt && gtt->userptr && range)
-		amdgpu_hmm_range_get_pages_done(range);
-}
-
-/*
- * amdgpu_ttm_tt_get_user_pages_done - stop HMM track the CPU page table change
- * Check if the pages backing this ttm range have been invalidated
- *
- * Returns: true if pages are still valid
- */
-bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm,
-				       struct hmm_range *range)
-{
-	struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm);
-
-	if (!gtt || !gtt->userptr || !range)
-		return false;
-
-	DRM_DEBUG_DRIVER("user_pages_done 0x%llx pages 0x%x\n",
-		gtt->userptr, ttm->num_pages);
-
-	WARN_ONCE(!range->hmm_pfns, "No user pages to check\n");
-
-	return !amdgpu_hmm_range_get_pages_done(range);
-}
 #endif
 
 /*
@@ -732,12 +754,12 @@ bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm,
  * that backs user memory and will ultimately be mapped into the device
  * address space.
  */
-void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages)
+void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct amdgpu_hmm_range *range)
 {
 	unsigned long i;
 
 	for (i = 0; i < ttm->num_pages; ++i)
-		ttm->pages[i] = pages ? pages[i] : NULL;
+		ttm->pages[i] = range ? hmm_pfn_to_page(range->hmm_range.hmm_pfns[i]) : NULL;
 }
 
 /*
@@ -765,7 +787,7 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_device *bdev,
 	/* Map SG to device */
 	r = dma_map_sgtable(adev->dev, ttm->sg, direction, 0);
 	if (r)
-		goto release_sg;
+		goto release_sg_table;
 
 	/* convert SG to linear array of pages and dma addresses */
 	drm_prime_sg_to_dma_addr_array(ttm->sg, gtt->ttm.dma_address,
@@ -773,6 +795,8 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_device *bdev,
 
 	return 0;
 
+release_sg_table:
+	sg_free_table(ttm->sg);
 release_sg:
 	kfree(ttm->sg);
 	ttm->sg = NULL;
@@ -800,6 +824,43 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_device *bdev,
 	sg_free_table(ttm->sg);
 }
 
+/*
+ * total_pages is constructed as MQD0+CtrlStack0 + MQD1+CtrlStack1 + ...
+ * MQDn+CtrlStackn where n is the number of XCCs per partition.
+ * pages_per_xcc is the size of one MQD+CtrlStack. The first page is MQD
+ * and uses memory type default, UC. The rest of pages_per_xcc are
+ * Ctrl stack and modify their memory type to NC.
+ */
+static void amdgpu_ttm_gart_bind_gfx9_mqd(struct amdgpu_device *adev,
+				struct ttm_tt *ttm, uint64_t flags)
+{
+	struct amdgpu_ttm_tt *gtt = (void *)ttm;
+	uint64_t total_pages = ttm->num_pages;
+	int num_xcc = max(1U, adev->gfx.num_xcc_per_xcp);
+	uint64_t page_idx, pages_per_xcc;
+	int i;
+	uint64_t ctrl_flags = AMDGPU_PTE_MTYPE_VG10(flags, AMDGPU_MTYPE_NC);
+
+	pages_per_xcc = total_pages;
+	do_div(pages_per_xcc, num_xcc);
+
+	for (i = 0, page_idx = 0; i < num_xcc; i++, page_idx += pages_per_xcc) {
+		/* MQD page: use default flags */
+		amdgpu_gart_bind(adev,
+				gtt->offset + (page_idx << PAGE_SHIFT),
+				1, &gtt->ttm.dma_address[page_idx], flags);
+		/*
+		 * Ctrl pages - modify the memory type to NC (ctrl_flags) from
+		 * the second page of the BO onward.
+		 */
+		amdgpu_gart_bind(adev,
+				gtt->offset + ((page_idx + 1) << PAGE_SHIFT),
+				pages_per_xcc - 1,
+				&gtt->ttm.dma_address[page_idx + 1],
+				ctrl_flags);
+	}
+}
+
 static void amdgpu_ttm_gart_bind(struct amdgpu_device *adev,
 				 struct ttm_buffer_object *tbo,
 				 uint64_t flags)
@@ -812,25 +873,12 @@ static void amdgpu_ttm_gart_bind(struct amdgpu_device *adev,
 		flags |= AMDGPU_PTE_TMZ;
 
 	if (abo->flags & AMDGPU_GEM_CREATE_CP_MQD_GFX9) {
-		uint64_t page_idx = 1;
-
-		amdgpu_gart_bind(adev, gtt->offset, page_idx,
-				 gtt->ttm.dma_address, flags);
-
-		/* The memory type of the first page defaults to UC. Now
-		 * modify the memory type to NC from the second page of
-		 * the BO onward.
-		 */
-		flags &= ~AMDGPU_PTE_MTYPE_VG10_MASK;
-		flags |= AMDGPU_PTE_MTYPE_VG10(AMDGPU_MTYPE_NC);
-
-		amdgpu_gart_bind(adev, gtt->offset + (page_idx << PAGE_SHIFT),
-				 ttm->num_pages - page_idx,
-				 &(gtt->ttm.dma_address[page_idx]), flags);
+		amdgpu_ttm_gart_bind_gfx9_mqd(adev, ttm, flags);
 	} else {
 		amdgpu_gart_bind(adev, gtt->offset, ttm->num_pages,
 				 gtt->ttm.dma_address, flags);
 	}
+	gtt->bound = true;
 }
 
 /*
@@ -857,7 +905,7 @@ static int amdgpu_ttm_backend_bind(struct ttm_device *bdev,
 	if (gtt->userptr) {
 		r = amdgpu_ttm_tt_pin_userptr(bdev, ttm);
 		if (r) {
-			DRM_ERROR("failed to pin userptr\n");
+			dev_err(adev->dev, "failed to pin userptr\n");
 			return r;
 		}
 	} else if (ttm->page_flags & TTM_TT_FLAG_EXTERNAL) {
@@ -922,16 +970,12 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
 		return 0;
 
 	addr = amdgpu_gmc_agp_addr(bo);
-	if (addr != AMDGPU_BO_INVALID_OFFSET) {
-		bo->resource->start = addr >> PAGE_SHIFT;
+	if (addr != AMDGPU_BO_INVALID_OFFSET)
 		return 0;
-	}
 
 	/* allocate GART space */
 	placement.num_placement = 1;
 	placement.placement = &placements;
-	placement.num_busy_placement = 1;
-	placement.busy_placement = &placements;
 	placements.fpfn = 0;
 	placements.lpfn = adev->gmc.gart_size >> PAGE_SHIFT;
 	placements.mem_type = TTM_PL_TT;
@@ -987,7 +1031,7 @@ static void amdgpu_ttm_backend_unbind(struct ttm_device *bdev,
 	/* if the pages have userptr pinning then clear that first */
 	if (gtt->userptr) {
 		amdgpu_ttm_tt_unpin_userptr(bdev, ttm);
-	} else if (ttm->sg && gtt->gobj->import_attach) {
+	} else if (ttm->sg && drm_gem_is_imported(gtt->gobj)) {
 		struct dma_buf_attachment *attach;
 
 		attach = gtt->gobj->import_attach;
@@ -1029,15 +1073,20 @@ static void amdgpu_ttm_backend_destroy(struct ttm_device *bdev,
 static struct ttm_tt *amdgpu_ttm_tt_create(struct ttm_buffer_object *bo,
 					   uint32_t page_flags)
 {
+	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);
 	struct amdgpu_bo *abo = ttm_to_amdgpu_bo(bo);
 	struct amdgpu_ttm_tt *gtt;
 	enum ttm_caching caching;
 
 	gtt = kzalloc(sizeof(struct amdgpu_ttm_tt), GFP_KERNEL);
-	if (gtt == NULL) {
+	if (!gtt)
 		return NULL;
-	}
+
 	gtt->gobj = &bo->base;
+	if (adev->gmc.mem_partitions && abo->xcp_id >= 0)
+		gtt->pool_id = KFD_XCP_MEM_ID(adev, abo->xcp_id);
+	else
+		gtt->pool_id = abo->xcp_id;
 
 	if (abo->flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC)
 		caching = ttm_write_combined;
@@ -1064,6 +1113,7 @@ static int amdgpu_ttm_tt_populate(struct ttm_device *bdev,
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bdev);
 	struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm);
+	struct ttm_pool *pool;
 	pgoff_t i;
 	int ret;
 
@@ -1078,7 +1128,11 @@ static int amdgpu_ttm_tt_populate(struct ttm_device *bdev,
 	if (ttm->page_flags & TTM_TT_FLAG_EXTERNAL)
 		return 0;
 
-	ret = ttm_pool_alloc(&adev->mman.bdev.pool, ttm, ctx);
+	if (adev->mman.ttm_pools && gtt->pool_id >= 0)
+		pool = &adev->mman.ttm_pools[gtt->pool_id];
+	else
+		pool = &adev->mman.bdev.pool;
+	ret = ttm_pool_alloc(pool, ttm, ctx);
 	if (ret)
 		return ret;
 
@@ -1099,6 +1153,7 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_device *bdev,
 {
 	struct amdgpu_ttm_tt *gtt = ttm_to_amdgpu_ttm_tt(ttm);
 	struct amdgpu_device *adev;
+	struct ttm_pool *pool;
 	pgoff_t i;
 
 	amdgpu_ttm_backend_unbind(bdev, ttm);
@@ -1117,7 +1172,13 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_device *bdev,
 		ttm->pages[i]->mapping = NULL;
 
 	adev = amdgpu_ttm_adev(bdev);
-	return ttm_pool_free(&adev->mman.bdev.pool, ttm);
+
+	if (adev->mman.ttm_pools && gtt->pool_id >= 0)
+		pool = &adev->mman.ttm_pools[gtt->pool_id];
+	else
+		pool = &adev->mman.bdev.pool;
+
+	return ttm_pool_free(pool, ttm);
 }
 
 /**
@@ -1263,10 +1324,12 @@ uint64_t amdgpu_ttm_tt_pde_flags(struct ttm_tt *ttm, struct ttm_resource *mem)
 		flags |= AMDGPU_PTE_VALID;
 
 	if (mem && (mem->mem_type == TTM_PL_TT ||
-		    mem->mem_type == AMDGPU_PL_PREEMPT)) {
+		    mem->mem_type == AMDGPU_PL_DOORBELL ||
+		    mem->mem_type == AMDGPU_PL_PREEMPT ||
+		    mem->mem_type == AMDGPU_PL_MMIO_REMAP)) {
 		flags |= AMDGPU_PTE_SYSTEM;
 
-		if (ttm->caching == ttm_cached)
+		if (ttm && ttm->caching == ttm_cached)
 			flags |= AMDGPU_PTE_SNOOPED;
 	}
 
@@ -1332,7 +1395,8 @@ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo,
 	 */
 	dma_resv_for_each_fence(&resv_cursor, bo->base.resv,
 				DMA_RESV_USAGE_BOOKKEEP, f) {
-		if (amdkfd_fence_check_mm(f, current->mm))
+		if (amdkfd_fence_check_mm(f, current->mm) &&
+		    !(place->flags & TTM_PL_FLAG_CONTIGUOUS))
 			return false;
 	}
 
@@ -1414,13 +1478,15 @@ static int amdgpu_ttm_access_memory_sdma(struct ttm_buffer_object *bo,
 		memcpy(adev->mman.sdma_access_ptr, buf, len);
 
 	num_dw = ALIGN(adev->mman.buffer_funcs->copy_num_dw, 8);
-	r = amdgpu_job_alloc_with_ib(adev, &adev->mman.entity,
+	r = amdgpu_job_alloc_with_ib(adev, &adev->mman.high_pr,
 				     AMDGPU_FENCE_OWNER_UNDEFINED,
 				     num_dw * 4, AMDGPU_IB_POOL_DELAYED,
-				     &job);
+				     &job,
+				     AMDGPU_KERNEL_JOB_ID_TTM_ACCESS_MEMORY_SDMA);
 	if (r)
 		goto out;
 
+	mutex_lock(&adev->mman.gtt_window_lock);
 	amdgpu_res_first(abo->tbo.resource, offset, len, &src_mm);
 	src_addr = amdgpu_ttm_domain_start(adev, bo->resource->mem_type) +
 		src_mm.start;
@@ -1429,12 +1495,13 @@ static int amdgpu_ttm_access_memory_sdma(struct ttm_buffer_object *bo,
 		swap(src_addr, dst_addr);
 
 	amdgpu_emit_copy_buffer(adev, &job->ibs[0], src_addr, dst_addr,
-				PAGE_SIZE, false);
+				PAGE_SIZE, 0);
 
 	amdgpu_ring_pad_ib(adev->mman.buffer_funcs_ring, &job->ibs[0]);
 	WARN_ON(job->ibs[0].length_dw > num_dw);
 
 	fence = amdgpu_job_submit(job);
+	mutex_unlock(&adev->mman.gtt_window_lock);
 
 	if (!dma_fence_wait_timeout(fence, false, adev->sdma_timeout))
 		r = -ETIMEDOUT;
@@ -1623,14 +1690,15 @@ static int amdgpu_ttm_training_reserve_vram_fini(struct amdgpu_device *adev)
 	return 0;
 }
 
-static void amdgpu_ttm_training_data_block_init(struct amdgpu_device *adev)
+static void amdgpu_ttm_training_data_block_init(struct amdgpu_device *adev,
+						uint32_t reserve_size)
 {
 	struct psp_memory_training_context *ctx = &adev->psp.mem_train_ctx;
 
 	memset(ctx, 0, sizeof(*ctx));
 
 	ctx->c2p_train_data_offset =
-		ALIGN((adev->gmc.mc_vram_size - adev->mman.discovery_tmr_size - SZ_1M), SZ_1M);
+		ALIGN((adev->gmc.mc_vram_size - reserve_size - SZ_1M), SZ_1M);
 	ctx->p2c_train_data_offset =
 		(adev->gmc.mc_vram_size - GDDR6_MEM_TRAINING_OFFSET);
 	ctx->train_data_size =
@@ -1648,11 +1716,12 @@ static void amdgpu_ttm_training_data_block_init(struct amdgpu_device *adev)
  */
 static int amdgpu_ttm_reserve_tmr(struct amdgpu_device *adev)
 {
-	int ret;
 	struct psp_memory_training_context *ctx = &adev->psp.mem_train_ctx;
 	bool mem_train_support = false;
+	uint32_t reserve_size = 0;
+	int ret;
 
-	if (!amdgpu_sriov_vf(adev)) {
+	if (adev->bios && !amdgpu_sriov_vf(adev)) {
 		if (amdgpu_atomfirmware_mem_training_supported(adev))
 			mem_train_support = true;
 		else
@@ -1666,41 +1735,135 @@ static int amdgpu_ttm_reserve_tmr(struct amdgpu_device *adev)
 	 * Otherwise, fallback to legacy approach to check and reserve tmr block for ip
 	 * discovery data and G6 memory training data respectively
 	 */
-	adev->mman.discovery_tmr_size =
-		amdgpu_atomfirmware_get_fw_reserved_fb_size(adev);
-	if (!adev->mman.discovery_tmr_size)
-		adev->mman.discovery_tmr_size = DISCOVERY_TMR_OFFSET;
+	if (adev->bios)
+		reserve_size =
+			amdgpu_atomfirmware_get_fw_reserved_fb_size(adev);
+
+	if (!adev->bios &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3) ||
+	     amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 4) ||
+	     amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 5, 0)))
+		reserve_size = max(reserve_size, (uint32_t)280 << 20);
+	else if (!reserve_size)
+		reserve_size = DISCOVERY_TMR_OFFSET;
 
 	if (mem_train_support) {
 		/* reserve vram for mem train according to TMR location */
-		amdgpu_ttm_training_data_block_init(adev);
+		amdgpu_ttm_training_data_block_init(adev, reserve_size);
 		ret = amdgpu_bo_create_kernel_at(adev,
 						 ctx->c2p_train_data_offset,
 						 ctx->train_data_size,
 						 &ctx->c2p_bo,
 						 NULL);
 		if (ret) {
-			DRM_ERROR("alloc c2p_bo failed(%d)!\n", ret);
+			dev_err(adev->dev, "alloc c2p_bo failed(%d)!\n", ret);
 			amdgpu_ttm_training_reserve_vram_fini(adev);
 			return ret;
 		}
 		ctx->init = PSP_MEM_TRAIN_RESERVE_SUCCESS;
 	}
 
-	ret = amdgpu_bo_create_kernel_at(adev,
-					 adev->gmc.real_vram_size - adev->mman.discovery_tmr_size,
-					 adev->mman.discovery_tmr_size,
-					 &adev->mman.discovery_memory,
-					 NULL);
+	ret = amdgpu_bo_create_kernel_at(
+		adev, adev->gmc.real_vram_size - reserve_size, reserve_size,
+		&adev->mman.fw_reserved_memory, NULL);
 	if (ret) {
-		DRM_ERROR("alloc tmr failed(%d)!\n", ret);
-		amdgpu_bo_free_kernel(&adev->mman.discovery_memory, NULL, NULL);
+		dev_err(adev->dev, "alloc tmr failed(%d)!\n", ret);
+		amdgpu_bo_free_kernel(&adev->mman.fw_reserved_memory, NULL,
+				      NULL);
 		return ret;
 	}
 
 	return 0;
 }
 
+static int amdgpu_ttm_pools_init(struct amdgpu_device *adev)
+{
+	int i;
+
+	if (!adev->gmc.is_app_apu || !adev->gmc.num_mem_partitions)
+		return 0;
+
+	adev->mman.ttm_pools = kcalloc(adev->gmc.num_mem_partitions,
+				       sizeof(*adev->mman.ttm_pools),
+				       GFP_KERNEL);
+	if (!adev->mman.ttm_pools)
+		return -ENOMEM;
+
+	for (i = 0; i < adev->gmc.num_mem_partitions; i++) {
+		ttm_pool_init(&adev->mman.ttm_pools[i], adev->dev,
+			      adev->gmc.mem_partitions[i].numa.node,
+			      TTM_ALLOCATION_POOL_BENEFICIAL_ORDER(get_order(SZ_2M)));
+	}
+	return 0;
+}
+
+static void amdgpu_ttm_pools_fini(struct amdgpu_device *adev)
+{
+	int i;
+
+	if (!adev->gmc.is_app_apu || !adev->mman.ttm_pools)
+		return;
+
+	for (i = 0; i < adev->gmc.num_mem_partitions; i++)
+		ttm_pool_fini(&adev->mman.ttm_pools[i]);
+
+	kfree(adev->mman.ttm_pools);
+	adev->mman.ttm_pools = NULL;
+}
+
+/**
+ * amdgpu_ttm_mmio_remap_bo_init - Allocate the singleton 4K MMIO_REMAP BO
+ * @adev: amdgpu device
+ *
+ * Allocates a one-page (4K) GEM BO in AMDGPU_GEM_DOMAIN_MMIO_REMAP when the
+ * hardware exposes a remap base (adev->rmmio_remap.bus_addr) and the host
+ * PAGE_SIZE is <= AMDGPU_GPU_PAGE_SIZE (4K). The BO is created as a regular
+ * GEM object (amdgpu_bo_create).
+ *
+ * Return:
+ *  * 0 on success or intentional skip (feature not present/unsupported)
+ *  * negative errno on allocation failure
+ */
+static int amdgpu_ttm_mmio_remap_bo_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_bo_param bp;
+	int r;
+
+	/* Skip if HW doesn't expose remap, or if PAGE_SIZE > AMDGPU_GPU_PAGE_SIZE (4K). */
+	if (!adev->rmmio_remap.bus_addr || PAGE_SIZE > AMDGPU_GPU_PAGE_SIZE)
+		return 0;
+
+	memset(&bp, 0, sizeof(bp));
+
+	/* Create exactly one GEM BO in the MMIO_REMAP domain. */
+	bp.type        = ttm_bo_type_device;          /* userspace-mappable GEM */
+	bp.size        = AMDGPU_GPU_PAGE_SIZE;        /* 4K */
+	bp.byte_align  = AMDGPU_GPU_PAGE_SIZE;
+	bp.domain      = AMDGPU_GEM_DOMAIN_MMIO_REMAP;
+	bp.flags       = 0;
+	bp.resv        = NULL;
+	bp.bo_ptr_size = sizeof(struct amdgpu_bo);
+
+	r = amdgpu_bo_create(adev, &bp, &adev->rmmio_remap.bo);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+/**
+ * amdgpu_ttm_mmio_remap_bo_fini - Free the singleton MMIO_REMAP BO
+ * @adev: amdgpu device
+ *
+ * Frees the kernel-owned MMIO_REMAP BO if it was allocated by
+ * amdgpu_ttm_mmio_remap_bo_init().
+ */
+static void amdgpu_ttm_mmio_remap_bo_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_unref(&adev->rmmio_remap.bo);
+	adev->rmmio_remap.bo = NULL;
+}
+
 /*
  * amdgpu_ttm_init - Init the memory management (ttm) as well as various
  * gtt/vram related fields.
@@ -1717,24 +1880,37 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
 
 	mutex_init(&adev->mman.gtt_window_lock);
 
+	dma_set_max_seg_size(adev->dev, UINT_MAX);
 	/* No others user of address space so set it to 0 */
 	r = ttm_device_init(&adev->mman.bdev, &amdgpu_bo_driver, adev->dev,
 			       adev_to_drm(adev)->anon_inode->i_mapping,
 			       adev_to_drm(adev)->vma_offset_manager,
-			       adev->need_swiotlb,
-			       dma_addressing_limited(adev->dev));
+			       (adev->need_swiotlb ?
+				TTM_ALLOCATION_POOL_USE_DMA_ALLOC : 0) |
+			       (dma_addressing_limited(adev->dev) ?
+				TTM_ALLOCATION_POOL_USE_DMA32 : 0) |
+			       TTM_ALLOCATION_POOL_BENEFICIAL_ORDER(get_order(SZ_2M)));
 	if (r) {
-		DRM_ERROR("failed initializing buffer object driver(%d).\n", r);
+		dev_err(adev->dev,
+			"failed initializing buffer object driver(%d).\n", r);
 		return r;
 	}
-	adev->mman.initialized = true;
 
-	/* Initialize VRAM pool with all of VRAM divided into pages */
-	r = amdgpu_vram_mgr_init(adev);
+	r = amdgpu_ttm_pools_init(adev);
 	if (r) {
-		DRM_ERROR("Failed initializing VRAM heap.\n");
+		dev_err(adev->dev, "failed to init ttm pools(%d).\n", r);
 		return r;
 	}
+	adev->mman.initialized = true;
+
+	if (!adev->gmc.is_app_apu) {
+		/* Initialize VRAM pool with all of VRAM divided into pages */
+		r = amdgpu_vram_mgr_init(adev);
+		if (r) {
+			dev_err(adev->dev, "Failed initializing VRAM heap.\n");
+			return r;
+		}
+	}
 
 	/* Change the size here instead of the init above so only lpfn is affected */
 	amdgpu_ttm_set_buffer_funcs_status(adev, false);
@@ -1744,6 +1920,9 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
 		adev->mman.aper_base_kaddr = ioremap_cache(adev->gmc.aper_base,
 				adev->gmc.visible_vram_size);
 
+	else if (adev->gmc.is_app_apu)
+		DRM_DEBUG_DRIVER(
+			"No need to ioremap when real vram size is 0\n");
 	else
 #endif
 		adev->mman.aper_base_kaddr = ioremap_wc(adev->gmc.aper_base,
@@ -1755,24 +1934,23 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
 	 *place on the VRAM, so reserve it early.
 	 */
 	r = amdgpu_ttm_fw_reserve_vram_init(adev);
-	if (r) {
+	if (r)
 		return r;
-	}
 
 	/*
-	 *The reserved vram for driver must be pinned to the specified
-	 *place on the VRAM, so reserve it early.
+	 * The reserved VRAM for the driver must be pinned to a specific
+	 * location in VRAM, so reserve it early.
 	 */
 	r = amdgpu_ttm_drv_reserve_vram_init(adev);
 	if (r)
 		return r;
 
 	/*
-	 * only NAVI10 and onwards ASIC support for IP discovery.
-	 * If IP discovery enabled, a block of memory should be
-	 * reserved for IP discovey.
+	 * only NAVI10 and later ASICs support IP discovery.
+	 * If IP discovery is enabled, a block of memory should be
+	 * reserved for it.
 	 */
-	if (adev->mman.discovery_bin) {
+	if (adev->discovery.reserve_tmr) {
 		r = amdgpu_ttm_reserve_tmr(adev);
 		if (r)
 			return r;
@@ -1781,84 +1959,120 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
 	/* allocate memory as required for VGA
 	 * This is used for VGA emulation and pre-OS scanout buffers to
 	 * avoid display artifacts while transitioning between pre-OS
-	 * and driver.  */
-	r = amdgpu_bo_create_kernel_at(adev, 0, adev->mman.stolen_vga_size,
-				       &adev->mman.stolen_vga_memory,
-				       NULL);
-	if (r)
-		return r;
-	r = amdgpu_bo_create_kernel_at(adev, adev->mman.stolen_vga_size,
-				       adev->mman.stolen_extended_size,
-				       &adev->mman.stolen_extended_memory,
-				       NULL);
-	if (r)
-		return r;
-	r = amdgpu_bo_create_kernel_at(adev, adev->mman.stolen_reserved_offset,
-				       adev->mman.stolen_reserved_size,
-				       &adev->mman.stolen_reserved_memory,
-				       NULL);
-	if (r)
-		return r;
+	 * and driver.
+	 */
+	if (!adev->gmc.is_app_apu) {
+		r = amdgpu_bo_create_kernel_at(adev, 0,
+					       adev->mman.stolen_vga_size,
+					       &adev->mman.stolen_vga_memory,
+					       NULL);
+		if (r)
+			return r;
 
-	DRM_INFO("amdgpu: %uM of VRAM memory ready\n",
-		 (unsigned) (adev->gmc.real_vram_size / (1024 * 1024)));
-
-	/* Compute GTT size, either based on 1/2 the size of RAM size
-	 * or whatever the user passed on module init */
-	if (amdgpu_gtt_size == -1) {
-		struct sysinfo si;
-
-		si_meminfo(&si);
-		/* Certain GL unit tests for large textures can cause problems
-		 * with the OOM killer since there is no way to link this memory
-		 * to a process.  This was originally mitigated (but not necessarily
-		 * eliminated) by limiting the GTT size.  The problem is this limit
-		 * is often too low for many modern games so just make the limit 1/2
-		 * of system memory which aligns with TTM. The OOM accounting needs
-		 * to be addressed, but we shouldn't prevent common 3D applications
-		 * from being usable just to potentially mitigate that corner case.
-		 */
-		gtt_size = max((AMDGPU_DEFAULT_GTT_SIZE_MB << 20),
-			       (u64)si.totalram * si.mem_unit / 2);
+		r = amdgpu_bo_create_kernel_at(adev, adev->mman.stolen_vga_size,
+					       adev->mman.stolen_extended_size,
+					       &adev->mman.stolen_extended_memory,
+					       NULL);
+
+		if (r)
+			return r;
+
+		r = amdgpu_bo_create_kernel_at(adev,
+					       adev->mman.stolen_reserved_offset,
+					       adev->mman.stolen_reserved_size,
+					       &adev->mman.stolen_reserved_memory,
+					       NULL);
+		if (r)
+			return r;
 	} else {
-		gtt_size = (uint64_t)amdgpu_gtt_size << 20;
+		DRM_DEBUG_DRIVER("Skipped stolen memory reservation\n");
+	}
+
+	dev_info(adev->dev, "amdgpu: %uM of VRAM memory ready\n",
+		 (unsigned int)(adev->gmc.real_vram_size / (1024 * 1024)));
+
+	/* Compute GTT size, either based on TTM limit
+	 * or whatever the user passed on module init.
+	 */
+	gtt_size = ttm_tt_pages_limit() << PAGE_SHIFT;
+	if (amdgpu_gtt_size != -1) {
+		uint64_t configured_size = (uint64_t)amdgpu_gtt_size << 20;
+
+		drm_warn(&adev->ddev,
+			"Configuring gttsize via module parameter is deprecated, please use ttm.pages_limit\n");
+		if (gtt_size != configured_size)
+			drm_warn(&adev->ddev,
+				"GTT size has been set as %llu but TTM size has been set as %llu, this is unusual\n",
+				configured_size, gtt_size);
+
+		gtt_size = configured_size;
 	}
 
 	/* Initialize GTT memory pool */
 	r = amdgpu_gtt_mgr_init(adev, gtt_size);
 	if (r) {
-		DRM_ERROR("Failed initializing GTT heap.\n");
+		dev_err(adev->dev, "Failed initializing GTT heap.\n");
 		return r;
 	}
-	DRM_INFO("amdgpu: %uM of GTT memory ready.\n",
-		 (unsigned)(gtt_size / (1024 * 1024)));
+	dev_info(adev->dev, "amdgpu: %uM of GTT memory ready.\n",
+		 (unsigned int)(gtt_size / (1024 * 1024)));
+
+	if (adev->flags & AMD_IS_APU) {
+		if (adev->gmc.real_vram_size < gtt_size)
+			adev->apu_prefer_gtt = true;
+	}
+
+	/* Initialize doorbell pool on PCI BAR */
+	r = amdgpu_ttm_init_on_chip(adev, AMDGPU_PL_DOORBELL, adev->doorbell.size / PAGE_SIZE);
+	if (r) {
+		dev_err(adev->dev, "Failed initializing doorbell heap.\n");
+		return r;
+	}
+
+	/* Create a boorbell page for kernel usages */
+	r = amdgpu_doorbell_create_kernel_doorbells(adev);
+	if (r) {
+		dev_err(adev->dev, "Failed to initialize kernel doorbells.\n");
+		return r;
+	}
+
+	/* Initialize MMIO-remap pool (single page 4K) */
+	r = amdgpu_ttm_init_on_chip(adev, AMDGPU_PL_MMIO_REMAP, 1);
+	if (r) {
+		dev_err(adev->dev, "Failed initializing MMIO-remap heap.\n");
+		return r;
+	}
+
+	/* Allocate the singleton MMIO_REMAP BO (4K) if supported */
+	r = amdgpu_ttm_mmio_remap_bo_init(adev);
+	if (r)
+		return r;
 
 	/* Initialize preemptible memory pool */
 	r = amdgpu_preempt_mgr_init(adev);
 	if (r) {
-		DRM_ERROR("Failed initializing PREEMPT heap.\n");
+		dev_err(adev->dev, "Failed initializing PREEMPT heap.\n");
 		return r;
 	}
 
 	/* Initialize various on-chip memory pools */
 	r = amdgpu_ttm_init_on_chip(adev, AMDGPU_PL_GDS, adev->gds.gds_size);
 	if (r) {
-		DRM_ERROR("Failed initializing GDS heap.\n");
+		dev_err(adev->dev, "Failed initializing GDS heap.\n");
 		return r;
 	}
 
 	r = amdgpu_ttm_init_on_chip(adev, AMDGPU_PL_GWS, adev->gds.gws_size);
 	if (r) {
-		DRM_ERROR("Failed initializing gws heap.\n");
+		dev_err(adev->dev, "Failed initializing gws heap.\n");
 		return r;
 	}
 
 	r = amdgpu_ttm_init_on_chip(adev, AMDGPU_PL_OA, adev->gds.oa_size);
 	if (r) {
-		DRM_ERROR("Failed initializing oa heap.\n");
+		dev_err(adev->dev, "Failed initializing oa heap.\n");
 		return r;
 	}
-
 	if (amdgpu_bo_create_kernel(adev, PAGE_SIZE, PAGE_SIZE,
 				AMDGPU_GEM_DOMAIN_GTT,
 				&adev->mman.sdma_access_bo, NULL,
@@ -1874,20 +2088,30 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
 void amdgpu_ttm_fini(struct amdgpu_device *adev)
 {
 	int idx;
+
 	if (!adev->mman.initialized)
 		return;
 
+	amdgpu_ttm_pools_fini(adev);
+
 	amdgpu_ttm_training_reserve_vram_fini(adev);
 	/* return the stolen vga memory back to VRAM */
-	amdgpu_bo_free_kernel(&adev->mman.stolen_vga_memory, NULL, NULL);
-	amdgpu_bo_free_kernel(&adev->mman.stolen_extended_memory, NULL, NULL);
-	/* return the IP Discovery TMR memory back to VRAM */
-	amdgpu_bo_free_kernel(&adev->mman.discovery_memory, NULL, NULL);
-	if (adev->mman.stolen_reserved_size)
-		amdgpu_bo_free_kernel(&adev->mman.stolen_reserved_memory,
-				      NULL, NULL);
+	if (!adev->gmc.is_app_apu) {
+		amdgpu_bo_free_kernel(&adev->mman.stolen_vga_memory, NULL, NULL);
+		amdgpu_bo_free_kernel(&adev->mman.stolen_extended_memory, NULL, NULL);
+		/* return the FW reserved memory back to VRAM */
+		amdgpu_bo_free_kernel(&adev->mman.fw_reserved_memory, NULL,
+				      NULL);
+		amdgpu_bo_free_kernel(&adev->mman.fw_reserved_memory_extend, NULL,
+				      NULL);
+		if (adev->mman.stolen_reserved_size)
+			amdgpu_bo_free_kernel(&adev->mman.stolen_reserved_memory,
+					      NULL, NULL);
+	}
 	amdgpu_bo_free_kernel(&adev->mman.sdma_access_bo, NULL,
 					&adev->mman.sdma_access_ptr);
+
+	amdgpu_ttm_mmio_remap_bo_fini(adev);
 	amdgpu_ttm_fw_reserve_vram_fini(adev);
 	amdgpu_ttm_drv_reserve_vram_fini(adev);
 
@@ -1900,15 +2124,20 @@ void amdgpu_ttm_fini(struct amdgpu_device *adev)
 		drm_dev_exit(idx);
 	}
 
-	amdgpu_vram_mgr_fini(adev);
+	if (!adev->gmc.is_app_apu)
+		amdgpu_vram_mgr_fini(adev);
 	amdgpu_gtt_mgr_fini(adev);
 	amdgpu_preempt_mgr_fini(adev);
+	amdgpu_doorbell_fini(adev);
+
 	ttm_range_man_fini(&adev->mman.bdev, AMDGPU_PL_GDS);
 	ttm_range_man_fini(&adev->mman.bdev, AMDGPU_PL_GWS);
 	ttm_range_man_fini(&adev->mman.bdev, AMDGPU_PL_OA);
+	ttm_range_man_fini(&adev->mman.bdev, AMDGPU_PL_DOORBELL);
+	ttm_range_man_fini(&adev->mman.bdev, AMDGPU_PL_MMIO_REMAP);
 	ttm_device_fini(&adev->mman.bdev);
 	adev->mman.initialized = false;
-	DRM_INFO("amdgpu: ttm finalized\n");
+	dev_info(adev->dev, "amdgpu: ttm finalized\n");
 }
 
 /**
@@ -1927,7 +2156,7 @@ void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable)
 	int r;
 
 	if (!adev->mman.initialized || amdgpu_in_reset(adev) ||
-	    adev->mman.buffer_funcs_enabled == enable)
+	    adev->mman.buffer_funcs_enabled == enable || adev->gmc.is_app_apu)
 		return;
 
 	if (enable) {
@@ -1936,18 +2165,32 @@ void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable)
 
 		ring = adev->mman.buffer_funcs_ring;
 		sched = &ring->sched;
-		r = drm_sched_entity_init(&adev->mman.entity,
+		r = drm_sched_entity_init(&adev->mman.high_pr,
 					  DRM_SCHED_PRIORITY_KERNEL, &sched,
 					  1, NULL);
 		if (r) {
-			DRM_ERROR("Failed setting up TTM BO move entity (%d)\n",
-				  r);
+			dev_err(adev->dev,
+				"Failed setting up TTM BO move entity (%d)\n",
+				r);
 			return;
 		}
+
+		r = drm_sched_entity_init(&adev->mman.low_pr,
+					  DRM_SCHED_PRIORITY_NORMAL, &sched,
+					  1, NULL);
+		if (r) {
+			dev_err(adev->dev,
+				"Failed setting up TTM BO move entity (%d)\n",
+				r);
+			goto error_free_entity;
+		}
 	} else {
-		drm_sched_entity_destroy(&adev->mman.entity);
-		dma_fence_put(man->move);
-		man->move = NULL;
+		drm_sched_entity_destroy(&adev->mman.high_pr);
+		drm_sched_entity_destroy(&adev->mman.low_pr);
+		/* Drop all the old fences since re-creating the scheduler entities
+		 * will allocate new contexts.
+		 */
+		ttm_resource_manager_cleanup(man);
 	}
 
 	/* this just adjusts TTM size idea, which sets lpfn to the correct value */
@@ -1957,6 +2200,11 @@ void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable)
 		size = adev->gmc.visible_vram_size;
 	man->size = size;
 	adev->mman.buffer_funcs_enabled = enable;
+
+	return;
+
+error_free_entity:
+	drm_sched_entity_destroy(&adev->mman.high_pr);
 }
 
 static int amdgpu_ttm_prepare_job(struct amdgpu_device *adev,
@@ -1964,16 +2212,18 @@ static int amdgpu_ttm_prepare_job(struct amdgpu_device *adev,
 				  unsigned int num_dw,
 				  struct dma_resv *resv,
 				  bool vm_needs_flush,
-				  struct amdgpu_job **job)
+				  struct amdgpu_job **job,
+				  bool delayed, u64 k_job_id)
 {
 	enum amdgpu_ib_pool_type pool = direct_submit ?
 		AMDGPU_IB_POOL_DIRECT :
 		AMDGPU_IB_POOL_DELAYED;
 	int r;
-
-	r = amdgpu_job_alloc_with_ib(adev, &adev->mman.entity,
+	struct drm_sched_entity *entity = delayed ? &adev->mman.low_pr :
+						    &adev->mman.high_pr;
+	r = amdgpu_job_alloc_with_ib(adev, entity,
 				     AMDGPU_FENCE_OWNER_UNDEFINED,
-				     num_dw * 4, pool, job);
+				     num_dw * 4, pool, job, k_job_id);
 	if (r)
 		return r;
 
@@ -1994,17 +2244,18 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
 		       uint64_t dst_offset, uint32_t byte_count,
 		       struct dma_resv *resv,
 		       struct dma_fence **fence, bool direct_submit,
-		       bool vm_needs_flush, bool tmz)
+		       bool vm_needs_flush, uint32_t copy_flags)
 {
 	struct amdgpu_device *adev = ring->adev;
-	unsigned num_loops, num_dw;
+	unsigned int num_loops, num_dw;
 	struct amdgpu_job *job;
 	uint32_t max_bytes;
-	unsigned i;
+	unsigned int i;
 	int r;
 
 	if (!direct_submit && !ring->sched.ready) {
-		DRM_ERROR("Trying to move memory with ring turned off.\n");
+		dev_err(adev->dev,
+			"Trying to move memory with ring turned off.\n");
 		return -EINVAL;
 	}
 
@@ -2012,7 +2263,8 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
 	num_loops = DIV_ROUND_UP(byte_count, max_bytes);
 	num_dw = ALIGN(num_loops * adev->mman.buffer_funcs->copy_num_dw, 8);
 	r = amdgpu_ttm_prepare_job(adev, direct_submit, num_dw,
-				   resv, vm_needs_flush, &job);
+				   resv, vm_needs_flush, &job, false,
+				   AMDGPU_KERNEL_JOB_ID_TTM_COPY_BUFFER);
 	if (r)
 		return r;
 
@@ -2020,8 +2272,7 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
 		uint32_t cur_size_in_bytes = min(byte_count, max_bytes);
 
 		amdgpu_emit_copy_buffer(adev, &job->ibs[0], src_offset,
-					dst_offset, cur_size_in_bytes, tmz);
-
+					dst_offset, cur_size_in_bytes, copy_flags);
 		src_offset += cur_size_in_bytes;
 		dst_offset += cur_size_in_bytes;
 		byte_count -= cur_size_in_bytes;
@@ -2040,7 +2291,7 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
 
 error_free:
 	amdgpu_job_free(job);
-	DRM_ERROR("Error scheduling IBs (%d)\n", r);
+	dev_err(adev->dev, "Error scheduling IBs (%d)\n", r);
 	return r;
 }
 
@@ -2048,7 +2299,8 @@ static int amdgpu_ttm_fill_mem(struct amdgpu_ring *ring, uint32_t src_data,
 			       uint64_t dst_addr, uint32_t byte_count,
 			       struct dma_resv *resv,
 			       struct dma_fence **fence,
-			       bool vm_needs_flush)
+			       bool vm_needs_flush, bool delayed,
+			       u64 k_job_id)
 {
 	struct amdgpu_device *adev = ring->adev;
 	unsigned int num_loops, num_dw;
@@ -2061,7 +2313,7 @@ static int amdgpu_ttm_fill_mem(struct amdgpu_ring *ring, uint32_t src_data,
 	num_loops = DIV_ROUND_UP_ULL(byte_count, max_bytes);
 	num_dw = ALIGN(num_loops * adev->mman.buffer_funcs->fill_num_dw, 8);
 	r = amdgpu_ttm_prepare_job(adev, false, num_dw, resv, vm_needs_flush,
-				   &job);
+				   &job, delayed, k_job_id);
 	if (r)
 		return r;
 
@@ -2081,10 +2333,78 @@ static int amdgpu_ttm_fill_mem(struct amdgpu_ring *ring, uint32_t src_data,
 	return 0;
 }
 
+/**
+ * amdgpu_ttm_clear_buffer - clear memory buffers
+ * @bo: amdgpu buffer object
+ * @resv: reservation object
+ * @fence: dma_fence associated with the operation
+ *
+ * Clear the memory buffer resource.
+ *
+ * Returns:
+ * 0 for success or a negative error code on failure.
+ */
+int amdgpu_ttm_clear_buffer(struct amdgpu_bo *bo,
+			    struct dma_resv *resv,
+			    struct dma_fence **fence)
+{
+	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
+	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
+	struct amdgpu_res_cursor cursor;
+	u64 addr;
+	int r = 0;
+
+	if (!adev->mman.buffer_funcs_enabled)
+		return -EINVAL;
+
+	if (!fence)
+		return -EINVAL;
+
+	*fence = dma_fence_get_stub();
+
+	amdgpu_res_first(bo->tbo.resource, 0, amdgpu_bo_size(bo), &cursor);
+
+	mutex_lock(&adev->mman.gtt_window_lock);
+	while (cursor.remaining) {
+		struct dma_fence *next = NULL;
+		u64 size;
+
+		if (amdgpu_res_cleared(&cursor)) {
+			amdgpu_res_next(&cursor, cursor.size);
+			continue;
+		}
+
+		/* Never clear more than 256MiB at once to avoid timeouts */
+		size = min(cursor.size, 256ULL << 20);
+
+		r = amdgpu_ttm_map_buffer(&bo->tbo, bo->tbo.resource, &cursor,
+					  1, ring, false, &size, &addr);
+		if (r)
+			goto err;
+
+		r = amdgpu_ttm_fill_mem(ring, 0, addr, size, resv,
+					&next, true, true,
+					AMDGPU_KERNEL_JOB_ID_TTM_CLEAR_BUFFER);
+		if (r)
+			goto err;
+
+		dma_fence_put(*fence);
+		*fence = next;
+
+		amdgpu_res_next(&cursor, size);
+	}
+err:
+	mutex_unlock(&adev->mman.gtt_window_lock);
+
+	return r;
+}
+
 int amdgpu_fill_buffer(struct amdgpu_bo *bo,
 			uint32_t src_data,
 			struct dma_resv *resv,
-			struct dma_fence **f)
+			struct dma_fence **f,
+			bool delayed,
+			u64 k_job_id)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
 	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
@@ -2093,7 +2413,8 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
 	int r;
 
 	if (!adev->mman.buffer_funcs_enabled) {
-		DRM_ERROR("Trying to clear memory with ring turned off.\n");
+		dev_err(adev->dev,
+			"Trying to clear memory with ring turned off.\n");
 		return -EINVAL;
 	}
 
@@ -2113,7 +2434,7 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
 			goto error;
 
 		r = amdgpu_ttm_fill_mem(ring, src_data, to, cur_size, resv,
-					&next, true);
+					&next, true, delayed, k_job_id);
 		if (r)
 			goto error;
 
@@ -2153,7 +2474,7 @@ int amdgpu_ttm_evict_resources(struct amdgpu_device *adev, int mem_type)
 		man = ttm_manager_type(&adev->mman.bdev, mem_type);
 		break;
 	default:
-		DRM_ERROR("Trying to evict invalid memory type\n");
+		dev_err(adev->dev, "Trying to evict invalid memory type\n");
 		return -EINVAL;
 	}
 
@@ -2164,7 +2485,7 @@ int amdgpu_ttm_evict_resources(struct amdgpu_device *adev, int mem_type)
 
 static int amdgpu_ttm_page_pool_show(struct seq_file *m, void *unused)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)m->private;
+	struct amdgpu_device *adev = m->private;
 
 	return ttm_pool_debugfs(&adev->mman.bdev.pool, m);
 }
@@ -2278,7 +2599,7 @@ static ssize_t amdgpu_iomem_read(struct file *f, char __user *buf,
 		struct page *p;
 		void *ptr;
 
-		bytes = bytes < size ? bytes : size;
+		bytes = min(bytes, size);
 
 		/* Translate the bus address to a physical address.  If
 		 * the domain is NULL it means there is no IOMMU active
@@ -2333,7 +2654,7 @@ static ssize_t amdgpu_iomem_write(struct file *f, const char __user *buf,
 		struct page *p;
 		void *ptr;
 
-		bytes = bytes < size ? bytes : size;
+		bytes = min(bytes, size);
 
 		addr = dom ? iommu_iova_to_phys(dom, addr) : addr;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
index e2cd5894afc9..577ee04ce0bf 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
@@ -26,18 +26,23 @@
 
 #include <linux/dma-direction.h>
 #include <drm/gpu_scheduler.h>
+#include <drm/ttm/ttm_placement.h>
 #include "amdgpu_vram_mgr.h"
-#include "amdgpu.h"
+#include "amdgpu_hmm.h"
 
 #define AMDGPU_PL_GDS		(TTM_PL_PRIV + 0)
 #define AMDGPU_PL_GWS		(TTM_PL_PRIV + 1)
 #define AMDGPU_PL_OA		(TTM_PL_PRIV + 2)
 #define AMDGPU_PL_PREEMPT	(TTM_PL_PRIV + 3)
+#define AMDGPU_PL_DOORBELL	(TTM_PL_PRIV + 4)
+#define AMDGPU_PL_MMIO_REMAP	(TTM_PL_PRIV + 5)
+#define __AMDGPU_PL_NUM	(TTM_PL_PRIV + 6)
 
 #define AMDGPU_GTT_MAX_TRANSFER_SIZE	512
 #define AMDGPU_GTT_NUM_TRANSFER_WINDOWS	2
 
-#define AMDGPU_POISON	0xd0bed0be
+extern const struct attribute_group amdgpu_vram_mgr_attr_group;
+extern const struct attribute_group amdgpu_gtt_mgr_attr_group;
 
 struct hmm_range;
 
@@ -49,6 +54,7 @@ struct amdgpu_gtt_mgr {
 
 struct amdgpu_mman {
 	struct ttm_device		bdev;
+	struct ttm_pool			*ttm_pools;
 	bool				initialized;
 	void __iomem			*aper_base_kaddr;
 
@@ -58,8 +64,10 @@ struct amdgpu_mman {
 	bool					buffer_funcs_enabled;
 
 	struct mutex				gtt_window_lock;
-	/* Scheduler entity for buffer moves */
-	struct drm_sched_entity			entity;
+	/* High priority scheduler entity for buffer moves */
+	struct drm_sched_entity			high_pr;
+	/* Low priority scheduler entity for VRAM clearing */
+	struct drm_sched_entity			low_pr;
 
 	struct amdgpu_vram_mgr vram_mgr;
 	struct amdgpu_gtt_mgr gtt_mgr;
@@ -75,10 +83,9 @@ struct amdgpu_mman {
 	uint64_t		stolen_reserved_offset;
 	uint64_t		stolen_reserved_size;
 
-	/* discovery */
-	uint8_t				*discovery_bin;
-	uint32_t			discovery_tmr_size;
-	struct amdgpu_bo		*discovery_memory;
+	/* fw reserved memory */
+	struct amdgpu_bo		*fw_reserved_memory;
+	struct amdgpu_bo		*fw_reserved_memory_extend;
 
 	/* firmware VRAM reservation */
 	u64		fw_vram_usage_start_offset;
@@ -103,6 +110,23 @@ struct amdgpu_copy_mem {
 	unsigned long			offset;
 };
 
+#define AMDGPU_COPY_FLAGS_TMZ		(1 << 0)
+#define AMDGPU_COPY_FLAGS_READ_DECOMPRESSED	(1 << 1)
+#define AMDGPU_COPY_FLAGS_WRITE_COMPRESSED	(1 << 2)
+#define AMDGPU_COPY_FLAGS_MAX_COMPRESSED_SHIFT		3
+#define AMDGPU_COPY_FLAGS_MAX_COMPRESSED_MASK		0x03
+#define AMDGPU_COPY_FLAGS_NUMBER_TYPE_SHIFT		5
+#define AMDGPU_COPY_FLAGS_NUMBER_TYPE_MASK		0x07
+#define AMDGPU_COPY_FLAGS_DATA_FORMAT_SHIFT		8
+#define AMDGPU_COPY_FLAGS_DATA_FORMAT_MASK		0x3f
+#define AMDGPU_COPY_FLAGS_WRITE_COMPRESS_DISABLE_SHIFT	14
+#define AMDGPU_COPY_FLAGS_WRITE_COMPRESS_DISABLE_MASK	0x1
+
+#define AMDGPU_COPY_FLAGS_SET(field, value) \
+	(((__u32)(value) & AMDGPU_COPY_FLAGS_##field##_MASK) << AMDGPU_COPY_FLAGS_##field##_SHIFT)
+#define AMDGPU_COPY_FLAGS_GET(value, field) \
+	(((__u32)(value) >> AMDGPU_COPY_FLAGS_##field##_SHIFT) & AMDGPU_COPY_FLAGS_##field##_MASK)
+
 int amdgpu_gtt_mgr_init(struct amdgpu_device *adev, uint64_t gtt_size);
 void amdgpu_gtt_mgr_fini(struct amdgpu_device *adev);
 int amdgpu_preempt_mgr_init(struct amdgpu_device *adev);
@@ -130,58 +154,46 @@ int amdgpu_vram_mgr_reserve_range(struct amdgpu_vram_mgr *mgr,
 				  uint64_t start, uint64_t size);
 int amdgpu_vram_mgr_query_page_status(struct amdgpu_vram_mgr *mgr,
 				      uint64_t start);
+void amdgpu_vram_mgr_clear_reset_blocks(struct amdgpu_device *adev);
+
+bool amdgpu_res_cpu_visible(struct amdgpu_device *adev,
+			    struct ttm_resource *res);
 
 int amdgpu_ttm_init(struct amdgpu_device *adev);
 void amdgpu_ttm_fini(struct amdgpu_device *adev);
 void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev,
 					bool enable);
-
 int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset,
 		       uint64_t dst_offset, uint32_t byte_count,
 		       struct dma_resv *resv,
 		       struct dma_fence **fence, bool direct_submit,
-		       bool vm_needs_flush, bool tmz);
-int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
-			       const struct amdgpu_copy_mem *src,
-			       const struct amdgpu_copy_mem *dst,
-			       uint64_t size, bool tmz,
-			       struct dma_resv *resv,
-			       struct dma_fence **f);
+		       bool vm_needs_flush, uint32_t copy_flags);
+int amdgpu_ttm_clear_buffer(struct amdgpu_bo *bo,
+			    struct dma_resv *resv,
+			    struct dma_fence **fence);
 int amdgpu_fill_buffer(struct amdgpu_bo *bo,
 			uint32_t src_data,
 			struct dma_resv *resv,
-			struct dma_fence **fence);
+			struct dma_fence **fence,
+			bool delayed,
+			u64 k_job_id);
 
 int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo);
 void amdgpu_ttm_recover_gart(struct ttm_buffer_object *tbo);
 uint64_t amdgpu_ttm_domain_start(struct amdgpu_device *adev, uint32_t type);
 
 #if IS_ENABLED(CONFIG_DRM_AMDGPU_USERPTR)
-int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo, struct page **pages,
-				 struct hmm_range **range);
-void amdgpu_ttm_tt_discard_user_pages(struct ttm_tt *ttm,
-				      struct hmm_range *range);
-bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm,
-				       struct hmm_range *range);
+int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo,
+				 struct amdgpu_hmm_range *range);
 #else
 static inline int amdgpu_ttm_tt_get_user_pages(struct amdgpu_bo *bo,
-					       struct page **pages,
-					       struct hmm_range **range)
+					       struct amdgpu_hmm_range *range)
 {
 	return -EPERM;
 }
-static inline void amdgpu_ttm_tt_discard_user_pages(struct ttm_tt *ttm,
-						    struct hmm_range *range)
-{
-}
-static inline bool amdgpu_ttm_tt_get_user_pages_done(struct ttm_tt *ttm,
-						     struct hmm_range *range)
-{
-	return false;
-}
 #endif
 
-void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages);
+void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct amdgpu_hmm_range *range);
 int amdgpu_ttm_tt_get_userptr(const struct ttm_buffer_object *tbo,
 			      uint64_t *user_addr);
 int amdgpu_ttm_tt_set_userptr(struct ttm_buffer_object *bo,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c
index f76b1cb8baf8..e96f24e9ad57 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c
@@ -28,6 +28,13 @@
 #include "amdgpu.h"
 #include "amdgpu_ucode.h"
 
+#define AMDGPU_UCODE_NAME_MAX		(128)
+
+static const struct kicker_device kicker_device_list[] = {
+	{0x744B, 0x00},
+	{0x7551, 0xC8}
+};
+
 static void amdgpu_ucode_print_common_hdr(const struct common_firmware_header *hdr)
 {
 	DRM_DEBUG("size_bytes: %u\n", le32_to_cpu(hdr->size_bytes));
@@ -323,6 +330,12 @@ void amdgpu_ucode_print_sdma_hdr(const struct common_firmware_header *hdr)
 		DRM_DEBUG("ctl_ucode_offset: %u\n", le32_to_cpu(sdma_hdr->ctl_ucode_offset));
 		DRM_DEBUG("ctl_jt_offset: %u\n", le32_to_cpu(sdma_hdr->ctl_jt_offset));
 		DRM_DEBUG("ctl_jt_size: %u\n", le32_to_cpu(sdma_hdr->ctl_jt_size));
+	} else if (version_major == 3) {
+		const struct sdma_firmware_header_v3_0 *sdma_hdr =
+			container_of(hdr, struct sdma_firmware_header_v3_0, header);
+
+		DRM_DEBUG("ucode_reversion: %u\n",
+			  le32_to_cpu(sdma_hdr->ucode_feature_version));
 	} else {
 		DRM_ERROR("Unknown SDMA ucode version: %u.%u\n",
 			  version_major, version_minor);
@@ -556,6 +569,8 @@ amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type)
 	default:
 		if (!load_type)
 			return AMDGPU_FW_LOAD_DIRECT;
+		else if (load_type == 3)
+			return AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO;
 		else
 			return AMDGPU_FW_LOAD_PSP;
 	}
@@ -642,6 +657,8 @@ const char *amdgpu_ucode_name(enum AMDGPU_UCODE_ID ucode_id)
 		return "SMC";
 	case AMDGPU_UCODE_ID_PPTABLE:
 		return "PPTABLE";
+	case AMDGPU_UCODE_ID_P2S_TABLE:
+		return "P2STABLE";
 	case AMDGPU_UCODE_ID_UVD:
 		return "UVD";
 	case AMDGPU_UCODE_ID_UVD1:
@@ -664,20 +681,70 @@ const char *amdgpu_ucode_name(enum AMDGPU_UCODE_ID ucode_id)
 		return "DMCUB";
 	case AMDGPU_UCODE_ID_CAP:
 		return "CAP";
+	case AMDGPU_UCODE_ID_VPE_CTX:
+		return "VPE_CTX";
+	case AMDGPU_UCODE_ID_VPE_CTL:
+		return "VPE_CTL";
+	case AMDGPU_UCODE_ID_VPE:
+		return "VPE";
+	case AMDGPU_UCODE_ID_UMSCH_MM_UCODE:
+		return "UMSCH_MM_UCODE";
+	case AMDGPU_UCODE_ID_UMSCH_MM_DATA:
+		return "UMSCH_MM_DATA";
+	case AMDGPU_UCODE_ID_UMSCH_MM_CMD_BUFFER:
+		return "UMSCH_MM_CMD_BUFFER";
+	case AMDGPU_UCODE_ID_JPEG_RAM:
+		return "JPEG";
+	case AMDGPU_UCODE_ID_SDMA_RS64:
+		return "RS64_SDMA";
+	case AMDGPU_UCODE_ID_CP_RS64_PFP:
+		return "RS64_PFP";
+	case AMDGPU_UCODE_ID_CP_RS64_ME:
+		return "RS64_ME";
+	case AMDGPU_UCODE_ID_CP_RS64_MEC:
+		return "RS64_MEC";
+	case AMDGPU_UCODE_ID_CP_RS64_PFP_P0_STACK:
+		return "RS64_PFP_P0_STACK";
+	case AMDGPU_UCODE_ID_CP_RS64_PFP_P1_STACK:
+		return "RS64_PFP_P1_STACK";
+	case AMDGPU_UCODE_ID_CP_RS64_ME_P0_STACK:
+		return "RS64_ME_P0_STACK";
+	case AMDGPU_UCODE_ID_CP_RS64_ME_P1_STACK:
+		return "RS64_ME_P1_STACK";
+	case AMDGPU_UCODE_ID_CP_RS64_MEC_P0_STACK:
+		return "RS64_MEC_P0_STACK";
+	case AMDGPU_UCODE_ID_CP_RS64_MEC_P1_STACK:
+		return "RS64_MEC_P1_STACK";
+	case AMDGPU_UCODE_ID_CP_RS64_MEC_P2_STACK:
+		return "RS64_MEC_P2_STACK";
+	case AMDGPU_UCODE_ID_CP_RS64_MEC_P3_STACK:
+		return "RS64_MEC_P3_STACK";
+	case AMDGPU_UCODE_ID_ISP:
+		return "ISP";
 	default:
 		return "UNKNOWN UCODE";
 	}
 }
 
+static inline int amdgpu_ucode_is_valid(uint32_t fw_version)
+{
+	if (!fw_version)
+		return -EINVAL;
+
+	return 0;
+}
+
 #define FW_VERSION_ATTR(name, mode, field)				\
 static ssize_t show_##name(struct device *dev,				\
-			  struct device_attribute *attr,		\
-			  char *buf)					\
+			   struct device_attribute *attr, char *buf)	\
 {									\
 	struct drm_device *ddev = dev_get_drvdata(dev);			\
 	struct amdgpu_device *adev = drm_to_adev(ddev);			\
 									\
-	return sysfs_emit(buf, "0x%08x\n", adev->field);	\
+	if (!buf)							\
+		return amdgpu_ucode_is_valid(adev->field);		\
+									\
+	return sysfs_emit(buf, "0x%08x\n", adev->field);		\
 }									\
 static DEVICE_ATTR(name, mode, show_##name, NULL)
 
@@ -703,6 +770,10 @@ FW_VERSION_ATTR(sdma_fw_version, 0444, sdma.instance[0].fw_version);
 FW_VERSION_ATTR(sdma2_fw_version, 0444, sdma.instance[1].fw_version);
 FW_VERSION_ATTR(vcn_fw_version, 0444, vcn.fw_version);
 FW_VERSION_ATTR(dmcu_fw_version, 0444, dm.dmcu_fw_version);
+FW_VERSION_ATTR(dmcub_fw_version, 0444, dm.dmcub_fw_version);
+FW_VERSION_ATTR(mes_fw_version, 0444, mes.sched_version & AMDGPU_MES_VERSION_MASK);
+FW_VERSION_ATTR(mes_kiq_fw_version, 0444, mes.kiq_version & AMDGPU_MES_VERSION_MASK);
+FW_VERSION_ATTR(pldm_fw_version, 0444, firmware.pldm_version);
 
 static struct attribute *fw_attrs[] = {
 	&dev_attr_vce_fw_version.attr, &dev_attr_uvd_fw_version.attr,
@@ -715,13 +786,30 @@ static struct attribute *fw_attrs[] = {
 	&dev_attr_ta_ras_fw_version.attr, &dev_attr_ta_xgmi_fw_version.attr,
 	&dev_attr_smc_fw_version.attr, &dev_attr_sdma_fw_version.attr,
 	&dev_attr_sdma2_fw_version.attr, &dev_attr_vcn_fw_version.attr,
-	&dev_attr_dmcu_fw_version.attr, &dev_attr_imu_fw_version.attr,
+	&dev_attr_dmcu_fw_version.attr, &dev_attr_dmcub_fw_version.attr,
+	&dev_attr_imu_fw_version.attr, &dev_attr_mes_fw_version.attr,
+	&dev_attr_mes_kiq_fw_version.attr, &dev_attr_pldm_fw_version.attr,
 	NULL
 };
 
+#define to_dev_attr(x) container_of(x, struct device_attribute, attr)
+
+static umode_t amdgpu_ucode_sys_visible(struct kobject *kobj,
+					struct attribute *attr, int idx)
+{
+	struct device_attribute *dev_attr = to_dev_attr(attr);
+	struct device *dev = kobj_to_dev(kobj);
+
+	if (dev_attr->show(dev, dev_attr, NULL) == -EINVAL)
+		return 0;
+
+	return attr->mode;
+}
+
 static const struct attribute_group fw_attr_group = {
 	.name = "fw_version",
-	.attrs = fw_attrs
+	.attrs = fw_attrs,
+	.is_visible = amdgpu_ucode_sys_visible
 };
 
 int amdgpu_ucode_sysfs_init(struct amdgpu_device *adev)
@@ -745,10 +833,13 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev,
 	const struct dmcub_firmware_header_v1_0 *dmcub_hdr = NULL;
 	const struct mes_firmware_header_v1_0 *mes_hdr = NULL;
 	const struct sdma_firmware_header_v2_0 *sdma_hdr = NULL;
+	const struct sdma_firmware_header_v3_0 *sdmav3_hdr = NULL;
 	const struct imu_firmware_header_v1_0 *imu_hdr = NULL;
+	const struct vpe_firmware_header_v1_0 *vpe_hdr = NULL;
+	const struct umsch_mm_firmware_header_v1_0 *umsch_mm_hdr = NULL;
 	u8 *ucode_addr;
 
-	if (NULL == ucode->fw)
+	if (!ucode->fw)
 		return 0;
 
 	ucode->mc_addr = mc_addr;
@@ -764,7 +855,10 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev,
 	dmcub_hdr = (const struct dmcub_firmware_header_v1_0 *)ucode->fw->data;
 	mes_hdr = (const struct mes_firmware_header_v1_0 *)ucode->fw->data;
 	sdma_hdr = (const struct sdma_firmware_header_v2_0 *)ucode->fw->data;
+	sdmav3_hdr = (const struct sdma_firmware_header_v3_0 *)ucode->fw->data;
 	imu_hdr = (const struct imu_firmware_header_v1_0 *)ucode->fw->data;
+	vpe_hdr = (const struct vpe_firmware_header_v1_0 *)ucode->fw->data;
+	umsch_mm_hdr = (const struct umsch_mm_firmware_header_v1_0 *)ucode->fw->data;
 
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 		switch (ucode->ucode_id) {
@@ -778,6 +872,11 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev,
 			ucode_addr = (u8 *)ucode->fw->data +
 				le32_to_cpu(sdma_hdr->ctl_ucode_offset);
 			break;
+		case AMDGPU_UCODE_ID_SDMA_RS64:
+			ucode->ucode_size = le32_to_cpu(sdmav3_hdr->ucode_size_bytes);
+			ucode_addr = (u8 *)ucode->fw->data +
+				le32_to_cpu(sdmav3_hdr->header.ucode_array_offset_bytes);
+			break;
 		case AMDGPU_UCODE_ID_CP_MEC1:
 		case AMDGPU_UCODE_ID_CP_MEC2:
 			ucode->ucode_size = le32_to_cpu(header->ucode_size_bytes) -
@@ -881,6 +980,10 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev,
 			ucode->ucode_size = ucode->fw->size;
 			ucode_addr = (u8 *)ucode->fw->data;
 			break;
+		case AMDGPU_UCODE_ID_P2S_TABLE:
+			ucode->ucode_size = ucode->fw->size;
+			ucode_addr = (u8 *)ucode->fw->data;
+			break;
 		case AMDGPU_UCODE_ID_IMU_I:
 			ucode->ucode_size = le32_to_cpu(imu_hdr->imu_iram_ucode_size_bytes);
 			ucode_addr = (u8 *)ucode->fw->data +
@@ -947,6 +1050,26 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev,
 			ucode_addr = (u8 *)ucode->fw->data +
 				le32_to_cpu(cpv2_hdr->data_offset_bytes);
 			break;
+		case AMDGPU_UCODE_ID_VPE_CTX:
+			ucode->ucode_size = le32_to_cpu(vpe_hdr->ctx_ucode_size_bytes);
+			ucode_addr = (u8 *)ucode->fw->data +
+				le32_to_cpu(vpe_hdr->header.ucode_array_offset_bytes);
+			break;
+		case AMDGPU_UCODE_ID_VPE_CTL:
+			ucode->ucode_size = le32_to_cpu(vpe_hdr->ctl_ucode_size_bytes);
+			ucode_addr = (u8 *)ucode->fw->data +
+				le32_to_cpu(vpe_hdr->ctl_ucode_offset);
+			break;
+		case AMDGPU_UCODE_ID_UMSCH_MM_UCODE:
+			ucode->ucode_size = le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_size_bytes);
+			ucode_addr = (u8 *)ucode->fw->data +
+				le32_to_cpu(umsch_mm_hdr->header.ucode_array_offset_bytes);
+			break;
+		case AMDGPU_UCODE_ID_UMSCH_MM_DATA:
+			ucode->ucode_size = le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_data_size_bytes);
+			ucode_addr = (u8 *)ucode->fw->data +
+				le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_data_offset_bytes);
+			break;
 		default:
 			ucode->ucode_size = le32_to_cpu(header->ucode_size_bytes);
 			ucode_addr = (u8 *)ucode->fw->data +
@@ -972,7 +1095,7 @@ static int amdgpu_ucode_patch_jt(struct amdgpu_firmware_info *ucode,
 	uint8_t *src_addr = NULL;
 	uint8_t *dst_addr = NULL;
 
-	if (NULL == ucode->fw)
+	if (!ucode->fw)
 		return 0;
 
 	comm_hdr = (const struct common_firmware_header *)ucode->fw->data;
@@ -990,9 +1113,11 @@ static int amdgpu_ucode_patch_jt(struct amdgpu_firmware_info *ucode,
 
 int amdgpu_ucode_create_bo(struct amdgpu_device *adev)
 {
-	if (adev->firmware.load_type != AMDGPU_FW_LOAD_DIRECT) {
+	if ((adev->firmware.load_type != AMDGPU_FW_LOAD_DIRECT) &&
+	    (adev->firmware.load_type != AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO)) {
 		amdgpu_bo_create_kernel(adev, adev->firmware.fw_size, PAGE_SIZE,
-			amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT,
+			(amdgpu_sriov_vf(adev) || adev->debug_use_vram_fw_buf) ?
+			AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT,
 			&adev->firmware.fw_buf,
 			&adev->firmware.fw_buf_mc,
 			&adev->firmware.fw_buf_ptr);
@@ -1035,6 +1160,9 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev)
 		adev->firmware.max_ucodes = AMDGPU_UCODE_ID_MAXIMUM;
 	}
 
+	if (amdgpu_virt_xgmi_migrate_enabled(adev) && adev->firmware.fw_buf)
+		adev->firmware.fw_buf_mc = amdgpu_bo_fb_aper_addr(adev->firmware.fw_buf);
+
 	for (i = 0; i < adev->firmware.max_ucodes; i++) {
 		ucode = &adev->firmware.ucode[i];
 		if (ucode->fw) {
@@ -1043,6 +1171,7 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev)
 			if (i == AMDGPU_UCODE_ID_CP_MEC1 &&
 			    adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
 				const struct gfx_firmware_header_v1_0 *cp_hdr;
+
 				cp_hdr = (const struct gfx_firmware_header_v1_0 *)ucode->fw->data;
 				amdgpu_ucode_patch_jt(ucode,  adev->firmware.fw_buf_mc + fw_offset,
 						    adev->firmware.fw_buf_ptr + fw_offset);
@@ -1057,7 +1186,7 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev)
 static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int block_type)
 {
 	if (block_type == MP0_HWIP) {
-		switch (adev->ip_versions[MP0_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 		case IP_VERSION(9, 0, 0):
 			switch (adev->asic_type) {
 			case CHIP_VEGA10:
@@ -1098,6 +1227,7 @@ static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int bl
 		case IP_VERSION(11, 0, 13):
 			return "beige_goby";
 		case IP_VERSION(11, 5, 0):
+		case IP_VERSION(11, 5, 2):
 			return "vangogh";
 		case IP_VERSION(12, 0, 1):
 			return "green_sardine";
@@ -1108,7 +1238,7 @@ static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int bl
 			return "yellow_carp";
 		}
 	} else if (block_type == MP1_HWIP) {
-		switch (adev->ip_versions[MP1_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, MP1_HWIP, 0)) {
 		case IP_VERSION(9, 0, 0):
 		case IP_VERSION(10, 0, 0):
 		case IP_VERSION(10, 0, 1):
@@ -1134,7 +1264,7 @@ static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int bl
 			return "aldebaran_smc";
 		}
 	} else if (block_type == SDMA0_HWIP) {
-		switch (adev->ip_versions[SDMA0_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) {
 		case IP_VERSION(4, 0, 0):
 			return "vega10_sdma";
 		case IP_VERSION(4, 0, 1):
@@ -1178,7 +1308,7 @@ static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int bl
 			return "vangogh_sdma";
 		}
 	} else if (block_type == UVD_HWIP) {
-		switch (adev->ip_versions[UVD_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, UVD_HWIP, 0)) {
 		case IP_VERSION(1, 0, 0):
 		case IP_VERSION(1, 0, 1):
 			if (adev->apu_flags & AMD_APU_IS_RAVEN2)
@@ -1203,7 +1333,8 @@ static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int bl
 		case IP_VERSION(3, 0, 0):
 		case IP_VERSION(3, 0, 64):
 		case IP_VERSION(3, 0, 192):
-			if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 0))
+			if (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			    IP_VERSION(10, 3, 0))
 				return "sienna_cichlid_vcn";
 			return "navy_flounder_vcn";
 		case IP_VERSION(3, 0, 2):
@@ -1216,7 +1347,7 @@ static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int bl
 			return "yellow_carp_vcn";
 		}
 	} else if (block_type == GC_HWIP) {
-		switch (adev->ip_versions[GC_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 		case IP_VERSION(9, 0, 1):
 			return "vega10";
 		case IP_VERSION(9, 2, 1):
@@ -1264,12 +1395,25 @@ static const char *amdgpu_ucode_legacy_naming(struct amdgpu_device *adev, int bl
 	return NULL;
 }
 
+bool amdgpu_is_kicker_fw(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(kicker_device_list); i++) {
+		if (adev->pdev->device == kicker_device_list[i].device &&
+		    adev->pdev->revision == kicker_device_list[i].revision)
+			return true;
+	}
+
+	return false;
+}
+
 void amdgpu_ucode_ip_version_decode(struct amdgpu_device *adev, int block_type, char *ucode_prefix, int len)
 {
 	int maj, min, rev;
 	char *ip_name;
 	const char *legacy;
-	uint32_t version = adev->ip_versions[block_type][0];
+	uint32_t version = amdgpu_ip_version(adev, block_type, 0);
 
 	legacy = amdgpu_ucode_legacy_naming(adev, block_type);
 	if (legacy) {
@@ -1293,6 +1437,12 @@ void amdgpu_ucode_ip_version_decode(struct amdgpu_device *adev, int block_type,
 	case UVD_HWIP:
 		ip_name = "vcn";
 		break;
+	case VPE_HWIP:
+		ip_name = "vpe";
+		break;
+	case ISP_HWIP:
+		ip_name = "isp";
+		break;
 	default:
 		BUG();
 	}
@@ -1309,24 +1459,49 @@ void amdgpu_ucode_ip_version_decode(struct amdgpu_device *adev, int block_type,
  *
  * @adev: amdgpu device
  * @fw: pointer to load firmware to
- * @fw_name: firmware to load
+ * @required: whether the firmware is required
+ * @fmt: firmware name format string
+ * @...: variable arguments
  *
  * This is a helper that will use request_firmware and amdgpu_ucode_validate
  * to load and run basic validation on firmware. If the load fails, remap
  * the error code to -ENODEV, so that early_init functions will fail to load.
  */
 int amdgpu_ucode_request(struct amdgpu_device *adev, const struct firmware **fw,
-			 const char *fw_name)
+			 enum amdgpu_ucode_required required, const char *fmt, ...)
 {
-	int err = request_firmware(fw, fw_name, adev->dev);
+	char fname[AMDGPU_UCODE_NAME_MAX];
+	va_list ap;
+	int r;
+
+	va_start(ap, fmt);
+	r = vsnprintf(fname, sizeof(fname), fmt, ap);
+	va_end(ap);
+	if (r == sizeof(fname)) {
+		dev_warn(adev->dev, "amdgpu firmware name buffer overflow\n");
+		return -EOVERFLOW;
+	}
 
-	if (err)
+	if (required == AMDGPU_UCODE_REQUIRED)
+		r = request_firmware(fw, fname, adev->dev);
+	else {
+		r = firmware_request_nowarn(fw, fname, adev->dev);
+		if (r)
+			drm_info(&adev->ddev, "Optional firmware \"%s\" was not found\n", fname);
+	}
+	if (r)
 		return -ENODEV;
-	err = amdgpu_ucode_validate(*fw);
-	if (err)
-		dev_dbg(adev->dev, "\"%s\" failed to validate\n", fw_name);
 
-	return err;
+	r = amdgpu_ucode_validate(*fw);
+	if (r)
+		/*
+		 * The amdgpu_ucode_request() should be paired with amdgpu_ucode_release()
+		 * regardless of success/failure, and the amdgpu_ucode_release() takes care of
+		 * firmware release and need to avoid redundant release FW operation here.
+		 */
+		dev_dbg(adev->dev, "\"%s\" failed to validate\n", fname);
+
+	return r;
 }
 
 /*
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h
index b03321e7d2d8..6349aad6da35 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h
@@ -25,6 +25,8 @@
 
 #include "amdgpu_socbb.h"
 
+#define RS64_FW_UC_START_ADDR_LO 0x3000
+
 struct common_firmware_header {
 	uint32_t size_bytes; /* size of the entire header+image(s) in bytes */
 	uint32_t header_size_bytes; /* size of just the header in bytes */
@@ -125,6 +127,8 @@ enum psp_fw_type {
 	PSP_FW_TYPE_PSP_INTF_DRV,
 	PSP_FW_TYPE_PSP_DBG_DRV,
 	PSP_FW_TYPE_PSP_RAS_DRV,
+	PSP_FW_TYPE_PSP_IPKEYMGR_DRV,
+	PSP_FW_TYPE_PSP_SPDM_DRV,
 	PSP_FW_TYPE_MAX_INDEX,
 };
 
@@ -135,6 +139,14 @@ struct psp_firmware_header_v2_0 {
 	struct psp_fw_bin_desc psp_fw_bin[];
 };
 
+/* version_major=2, version_minor=1 */
+struct psp_firmware_header_v2_1 {
+	struct common_firmware_header header;
+	uint32_t psp_fw_bin_count;
+	uint32_t psp_aux_fw_bin_index;
+	struct psp_fw_bin_desc psp_fw_bin[];
+};
+
 /* version_major=1, version_minor=0 */
 struct ta_firmware_header_v1_0 {
 	struct common_firmware_header header;
@@ -154,6 +166,7 @@ enum ta_fw_type {
 	TA_FW_TYPE_PSP_DTM,
 	TA_FW_TYPE_PSP_RAP,
 	TA_FW_TYPE_PSP_SECUREDISPLAY,
+	TA_FW_TYPE_PSP_XGMI_AUX,
 	TA_FW_TYPE_MAX_INDEX,
 };
 
@@ -315,6 +328,44 @@ struct sdma_firmware_header_v2_0 {
 	uint32_t ctl_jt_size; /* control thread size of jt */
 };
 
+/* version_major=1, version_minor=0 */
+struct vpe_firmware_header_v1_0 {
+	struct common_firmware_header header;
+	uint32_t ucode_feature_version;
+	uint32_t ctx_ucode_size_bytes; /* context thread ucode size */
+	uint32_t ctx_jt_offset; /* context thread jt location */
+	uint32_t ctx_jt_size; /* context thread size of jt */
+	uint32_t ctl_ucode_offset;
+	uint32_t ctl_ucode_size_bytes; /* control thread ucode size */
+	uint32_t ctl_jt_offset; /* control thread jt location */
+	uint32_t ctl_jt_size; /* control thread size of jt */
+};
+
+/* version_major=1, version_minor=0 */
+struct umsch_mm_firmware_header_v1_0 {
+	struct common_firmware_header header;
+	uint32_t umsch_mm_ucode_version;
+	uint32_t umsch_mm_ucode_size_bytes;
+	uint32_t umsch_mm_ucode_offset_bytes;
+	uint32_t umsch_mm_ucode_data_version;
+	uint32_t umsch_mm_ucode_data_size_bytes;
+	uint32_t umsch_mm_ucode_data_offset_bytes;
+	uint32_t umsch_mm_irq_start_addr_lo;
+	uint32_t umsch_mm_irq_start_addr_hi;
+	uint32_t umsch_mm_uc_start_addr_lo;
+	uint32_t umsch_mm_uc_start_addr_hi;
+	uint32_t umsch_mm_data_start_addr_lo;
+	uint32_t umsch_mm_data_start_addr_hi;
+};
+
+/* version_major=3, version_minor=0 */
+struct sdma_firmware_header_v3_0 {
+    struct common_firmware_header header;
+    uint32_t ucode_feature_version;
+    uint32_t ucode_offset_bytes;
+    uint32_t ucode_size_bytes;
+};
+
 /* gpu info payload */
 struct gpu_info_firmware_v1_0 {
 	uint32_t gc_num_se;
@@ -387,6 +438,7 @@ union amdgpu_firmware_header {
 	struct psp_firmware_header_v1_1 psp_v1_1;
 	struct psp_firmware_header_v1_3 psp_v1_3;
 	struct psp_firmware_header_v2_0 psp_v2_0;
+	struct psp_firmware_header_v2_0 psp_v2_1;
 	struct ta_firmware_header_v1_0 ta;
 	struct ta_firmware_header_v2_0 ta_v2_0;
 	struct gfx_firmware_header_v1_0 gfx;
@@ -400,6 +452,7 @@ union amdgpu_firmware_header {
 	struct sdma_firmware_header_v1_0 sdma;
 	struct sdma_firmware_header_v1_1 sdma_v1_1;
 	struct sdma_firmware_header_v2_0 sdma_v2_0;
+	struct sdma_firmware_header_v3_0 sdma_v3_0;
 	struct gpu_info_firmware_header_v1_0 gpu_info;
 	struct dmcu_firmware_header_v1_0 dmcu;
 	struct dmcub_firmware_header_v1_0 dmcub;
@@ -407,7 +460,7 @@ union amdgpu_firmware_header {
 	uint8_t raw[0x100];
 };
 
-#define UCODE_MAX_PSP_PACKAGING ((sizeof(union amdgpu_firmware_header) - sizeof(struct common_firmware_header) - 4) / sizeof(struct psp_fw_bin_desc))
+#define UCODE_MAX_PSP_PACKAGING (((sizeof(union amdgpu_firmware_header) - sizeof(struct common_firmware_header) - 4) / sizeof(struct psp_fw_bin_desc)) * 2)
 
 /*
  * fw loading support
@@ -424,6 +477,7 @@ enum AMDGPU_UCODE_ID {
 	AMDGPU_UCODE_ID_SDMA7,
 	AMDGPU_UCODE_ID_SDMA_UCODE_TH0,
 	AMDGPU_UCODE_ID_SDMA_UCODE_TH1,
+	AMDGPU_UCODE_ID_SDMA_RS64,
 	AMDGPU_UCODE_ID_CP_CE,
 	AMDGPU_UCODE_ID_CP_PFP,
 	AMDGPU_UCODE_ID_CP_ME,
@@ -474,6 +528,15 @@ enum AMDGPU_UCODE_ID {
 	AMDGPU_UCODE_ID_VCN0_RAM,
 	AMDGPU_UCODE_ID_VCN1_RAM,
 	AMDGPU_UCODE_ID_DMCUB,
+	AMDGPU_UCODE_ID_VPE_CTX,
+	AMDGPU_UCODE_ID_VPE_CTL,
+	AMDGPU_UCODE_ID_VPE,
+	AMDGPU_UCODE_ID_UMSCH_MM_UCODE,
+	AMDGPU_UCODE_ID_UMSCH_MM_DATA,
+	AMDGPU_UCODE_ID_UMSCH_MM_CMD_BUFFER,
+	AMDGPU_UCODE_ID_P2S_TABLE,
+	AMDGPU_UCODE_ID_JPEG_RAM,
+	AMDGPU_UCODE_ID_ISP,
 	AMDGPU_UCODE_ID_MAXIMUM,
 };
 
@@ -491,6 +554,11 @@ enum amdgpu_firmware_load_type {
 	AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO,
 };
 
+enum amdgpu_ucode_required {
+	AMDGPU_UCODE_OPTIONAL,
+	AMDGPU_UCODE_REQUIRED,
+};
+
 /* conform to smu_ucode_xfer_cz.h */
 #define AMDGPU_SDMA0_UCODE_LOADED	0x00000001
 #define AMDGPU_SDMA1_UCODE_LOADED	0x00000002
@@ -534,6 +602,12 @@ struct amdgpu_firmware {
 
 	void *fw_buf_ptr;
 	uint64_t fw_buf_mc;
+	uint32_t pldm_version;
+};
+
+struct kicker_device{
+	unsigned short device;
+	u8 revision;
 };
 
 void amdgpu_ucode_print_mc_hdr(const struct common_firmware_header *hdr);
@@ -544,8 +618,9 @@ void amdgpu_ucode_print_rlc_hdr(const struct common_firmware_header *hdr);
 void amdgpu_ucode_print_sdma_hdr(const struct common_firmware_header *hdr);
 void amdgpu_ucode_print_psp_hdr(const struct common_firmware_header *hdr);
 void amdgpu_ucode_print_gpu_info_hdr(const struct common_firmware_header *hdr);
+__printf(4, 5)
 int amdgpu_ucode_request(struct amdgpu_device *adev, const struct firmware **fw,
-			 const char *fw_name);
+			 enum amdgpu_ucode_required required, const char *fmt, ...);
 void amdgpu_ucode_release(const struct firmware **fw);
 bool amdgpu_ucode_hdr_version(union amdgpu_firmware_header *hdr,
 				uint16_t hdr_major, uint16_t hdr_minor);
@@ -562,5 +637,6 @@ amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type);
 const char *amdgpu_ucode_name(enum AMDGPU_UCODE_ID ucode_id);
 
 void amdgpu_ucode_ip_version_decode(struct amdgpu_device *adev, int block_type, char *ucode_prefix, int len);
+bool amdgpu_is_kicker_fw(struct amdgpu_device *adev);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c
index 9e2e97207e53..3f0b0e9af4f3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.c
@@ -21,14 +21,19 @@
  *
  */
 
+#include <linux/sort.h>
 #include "amdgpu.h"
 #include "umc_v6_7.h"
+#include "amdgpu_ras_mgr.h"
+#define MAX_UMC_POISON_POLLING_TIME_SYNC   20  //ms
+
+#define MAX_UMC_HASH_STRING_SIZE  256
 
 static int amdgpu_umc_convert_error_address(struct amdgpu_device *adev,
 				    struct ras_err_data *err_data, uint64_t err_addr,
 				    uint32_t ch_inst, uint32_t umc_inst)
 {
-	switch (adev->ip_versions[UMC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, UMC_HWIP, 0)) {
 	case IP_VERSION(6, 7, 0):
 		umc_v6_7_convert_error_address(adev,
 				err_data, err_addr, ch_inst, umc_inst);
@@ -45,8 +50,12 @@ static int amdgpu_umc_convert_error_address(struct amdgpu_device *adev,
 int amdgpu_umc_page_retirement_mca(struct amdgpu_device *adev,
 			uint64_t err_addr, uint32_t ch_inst, uint32_t umc_inst)
 {
-	struct ras_err_data err_data = {0, 0, 0, NULL};
-	int ret = AMDGPU_RAS_FAIL;
+	struct ras_err_data err_data;
+	int ret;
+
+	ret = amdgpu_ras_error_data_init(&err_data);
+	if (ret)
+		return ret;
 
 	err_data.err_addr =
 		kcalloc(adev->umc.max_ras_err_cnt_per_query,
@@ -54,145 +63,241 @@ int amdgpu_umc_page_retirement_mca(struct amdgpu_device *adev,
 	if (!err_data.err_addr) {
 		dev_warn(adev->dev,
 			"Failed to alloc memory for umc error record in MCA notifier!\n");
-		return AMDGPU_RAS_FAIL;
+		ret = AMDGPU_RAS_FAIL;
+		goto out_fini_err_data;
 	}
 
+	err_data.err_addr_len = adev->umc.max_ras_err_cnt_per_query;
+
 	/*
 	 * Translate UMC channel address to Physical address
 	 */
 	ret = amdgpu_umc_convert_error_address(adev, &err_data, err_addr,
 					ch_inst, umc_inst);
 	if (ret)
-		goto out;
+		goto out_free_err_addr;
 
 	if (amdgpu_bad_page_threshold != 0) {
 		amdgpu_ras_add_bad_pages(adev, err_data.err_addr,
-						err_data.err_addr_cnt);
+						err_data.err_addr_cnt, false);
 		amdgpu_ras_save_bad_pages(adev, NULL);
 	}
 
-out:
+out_free_err_addr:
 	kfree(err_data.err_addr);
+
+out_fini_err_data:
+	amdgpu_ras_error_data_fini(&err_data);
+
 	return ret;
 }
 
-static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev,
-		void *ras_error_status,
-		struct amdgpu_iv_entry *entry,
-		bool reset)
+void amdgpu_umc_handle_bad_pages(struct amdgpu_device *adev,
+			void *ras_error_status)
 {
 	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
 	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct amdgpu_ras_eeprom_control *control = &con->eeprom_control;
+	unsigned int error_query_mode;
 	int ret = 0;
+	unsigned long err_count;
 
-	kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
-	ret = amdgpu_dpm_get_ecc_info(adev, (void *)&(con->umc_ecc));
-	if (ret == -EOPNOTSUPP) {
-		if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops &&
-		    adev->umc.ras->ras_block.hw_ops->query_ras_error_count)
-		    adev->umc.ras->ras_block.hw_ops->query_ras_error_count(adev, ras_error_status);
-
-		if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops &&
-		    adev->umc.ras->ras_block.hw_ops->query_ras_error_address &&
-		    adev->umc.max_ras_err_cnt_per_query) {
-			err_data->err_addr =
-				kcalloc(adev->umc.max_ras_err_cnt_per_query,
-					sizeof(struct eeprom_table_record), GFP_KERNEL);
-
-			/* still call query_ras_error_address to clear error status
-			 * even NOMEM error is encountered
-			 */
-			if(!err_data->err_addr)
-				dev_warn(adev->dev, "Failed to alloc memory for "
-						"umc error address record!\n");
+	amdgpu_ras_get_error_query_mode(adev, &error_query_mode);
 
-			/* umc query_ras_error_address is also responsible for clearing
-			 * error status
-			 */
-			adev->umc.ras->ras_block.hw_ops->query_ras_error_address(adev, ras_error_status);
-		}
-	} else if (!ret) {
-		if (adev->umc.ras &&
-		    adev->umc.ras->ecc_info_query_ras_error_count)
-		    adev->umc.ras->ecc_info_query_ras_error_count(adev, ras_error_status);
-
-		if (adev->umc.ras &&
-		    adev->umc.ras->ecc_info_query_ras_error_address &&
-		    adev->umc.max_ras_err_cnt_per_query) {
-			err_data->err_addr =
-				kcalloc(adev->umc.max_ras_err_cnt_per_query,
-					sizeof(struct eeprom_table_record), GFP_KERNEL);
-
-			/* still call query_ras_error_address to clear error status
-			 * even NOMEM error is encountered
-			 */
-			if(!err_data->err_addr)
-				dev_warn(adev->dev, "Failed to alloc memory for "
-						"umc error address record!\n");
+	err_data->err_addr =
+		kcalloc(adev->umc.max_ras_err_cnt_per_query,
+			sizeof(struct eeprom_table_record), GFP_KERNEL);
 
-			/* umc query_ras_error_address is also responsible for clearing
-			 * error status
-			 */
-			adev->umc.ras->ecc_info_query_ras_error_address(adev, ras_error_status);
+	/* still call query_ras_error_address to clear error status
+	 * even NOMEM error is encountered
+	 */
+	if (!err_data->err_addr)
+		dev_warn(adev->dev,
+			"Failed to alloc memory for umc error address record!\n");
+	else
+		err_data->err_addr_len = adev->umc.max_ras_err_cnt_per_query;
+
+	mutex_lock(&con->page_retirement_lock);
+	if (!amdgpu_ras_smu_eeprom_supported(adev)) {
+		ret = amdgpu_dpm_get_ecc_info(adev, (void *)&(con->umc_ecc));
+		if (ret == -EOPNOTSUPP &&
+		    error_query_mode == AMDGPU_RAS_DIRECT_ERROR_QUERY) {
+			if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops &&
+			    adev->umc.ras->ras_block.hw_ops->query_ras_error_count)
+				adev->umc.ras->ras_block.hw_ops->query_ras_error_count(adev,
+								ras_error_status);
+
+			if (adev->umc.ras && adev->umc.ras->ras_block.hw_ops &&
+			    adev->umc.ras->ras_block.hw_ops->query_ras_error_address &&
+			    adev->umc.max_ras_err_cnt_per_query) {
+				err_data->err_addr =
+					kcalloc(adev->umc.max_ras_err_cnt_per_query,
+						sizeof(struct eeprom_table_record), GFP_KERNEL);
+
+				/* still call query_ras_error_address to clear error status
+				 * even NOMEM error is encountered
+				 */
+				if (!err_data->err_addr)
+					dev_warn(adev->dev,
+						"Failed to alloc memory for umc error address record!\n");
+				else
+					err_data->err_addr_len =
+						adev->umc.max_ras_err_cnt_per_query;
+
+				/* umc query_ras_error_address is also responsible for clearing
+				 * error status
+				 */
+				adev->umc.ras->ras_block.hw_ops->query_ras_error_address(adev,
+								ras_error_status);
+			}
+		} else if (error_query_mode == AMDGPU_RAS_FIRMWARE_ERROR_QUERY ||
+		    (!ret && error_query_mode == AMDGPU_RAS_DIRECT_ERROR_QUERY)) {
+			if (adev->umc.ras &&
+			    adev->umc.ras->ecc_info_query_ras_error_count)
+				adev->umc.ras->ecc_info_query_ras_error_count(adev,
+								ras_error_status);
+
+			if (adev->umc.ras &&
+			    adev->umc.ras->ecc_info_query_ras_error_address &&
+			    adev->umc.max_ras_err_cnt_per_query) {
+				err_data->err_addr =
+					kcalloc(adev->umc.max_ras_err_cnt_per_query,
+						sizeof(struct eeprom_table_record), GFP_KERNEL);
+
+				/* still call query_ras_error_address to clear error status
+				 * even NOMEM error is encountered
+				 */
+				if (!err_data->err_addr)
+					dev_warn(adev->dev,
+						"Failed to alloc memory for umc error address record!\n");
+				else
+					err_data->err_addr_len =
+						adev->umc.max_ras_err_cnt_per_query;
+
+				/* umc query_ras_error_address is also responsible for clearing
+				 * error status
+				 */
+				adev->umc.ras->ecc_info_query_ras_error_address(adev,
+								ras_error_status);
+			}
+		}
+	} else {
+		if (!amdgpu_ras_eeprom_update_record_num(control)) {
+			err_data->err_addr_cnt = err_data->de_count =
+				control->ras_num_recs -	control->ras_num_recs_old;
+			amdgpu_ras_eeprom_read_idx(control, err_data->err_addr,
+				control->ras_num_recs_old, err_data->de_count);
 		}
 	}
 
 	/* only uncorrectable error needs gpu reset */
-	if (err_data->ue_count) {
-		dev_info(adev->dev, "%ld uncorrectable hardware errors "
-				"detected in UMC block\n",
-				err_data->ue_count);
-
+	if (err_data->ue_count || err_data->de_count) {
+		err_count = err_data->ue_count + err_data->de_count;
 		if ((amdgpu_bad_page_threshold != 0) &&
 			err_data->err_addr_cnt) {
 			amdgpu_ras_add_bad_pages(adev, err_data->err_addr,
-						err_data->err_addr_cnt);
-			amdgpu_ras_save_bad_pages(adev, &(err_data->ue_count));
+				err_data->err_addr_cnt, amdgpu_ras_smu_eeprom_supported(adev));
+			amdgpu_ras_save_bad_pages(adev, &err_count);
 
-			amdgpu_dpm_send_hbm_bad_pages_num(adev, con->eeprom_control.ras_num_recs);
+			amdgpu_dpm_send_hbm_bad_pages_num(adev,
+					con->eeprom_control.ras_num_bad_pages);
 
 			if (con->update_channel_flag == true) {
 				amdgpu_dpm_send_hbm_bad_channel_flag(adev, con->eeprom_control.bad_channel_bitmap);
 				con->update_channel_flag = false;
 			}
 		}
-
-		if (reset)
-			amdgpu_ras_reset_gpu(adev);
 	}
 
 	kfree(err_data->err_addr);
+	err_data->err_addr = NULL;
+
+	mutex_unlock(&con->page_retirement_lock);
+}
+
+static int amdgpu_umc_do_page_retirement(struct amdgpu_device *adev,
+		void *ras_error_status,
+		struct amdgpu_iv_entry *entry,
+		uint32_t reset)
+{
+	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+
+	kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
+	amdgpu_umc_handle_bad_pages(adev, ras_error_status);
+
+	if ((err_data->ue_count || err_data->de_count) &&
+	    (reset || amdgpu_ras_is_rma(adev))) {
+		con->gpu_reset_flags |= reset;
+		amdgpu_ras_reset_gpu(adev);
+	}
+
 	return AMDGPU_RAS_SUCCESS;
 }
 
-int amdgpu_umc_poison_handler(struct amdgpu_device *adev, bool reset)
+int amdgpu_umc_pasid_poison_handler(struct amdgpu_device *adev,
+			enum amdgpu_ras_block block, uint16_t pasid,
+			pasid_notify pasid_fn, void *data, uint32_t reset)
 {
 	int ret = AMDGPU_RAS_SUCCESS;
 
+	if (adev->gmc.xgmi.connected_to_cpu ||
+		adev->gmc.is_app_apu) {
+		if (reset) {
+			/* MCA poison handler is only responsible for GPU reset,
+			 * let MCA notifier do page retirement.
+			 */
+			kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
+			amdgpu_ras_reset_gpu(adev);
+		}
+		return ret;
+	}
+
 	if (!amdgpu_sriov_vf(adev)) {
-		if (!adev->gmc.xgmi.connected_to_cpu) {
-			struct ras_err_data err_data = {0, 0, 0, NULL};
+		if (amdgpu_ip_version(adev, UMC_HWIP, 0) < IP_VERSION(12, 0, 0)) {
+			struct ras_err_data err_data;
 			struct ras_common_if head = {
 				.block = AMDGPU_RAS_BLOCK__UMC,
 			};
 			struct ras_manager *obj = amdgpu_ras_find_obj(adev, &head);
 
+			ret = amdgpu_ras_error_data_init(&err_data);
+			if (ret)
+				return ret;
+
 			ret = amdgpu_umc_do_page_retirement(adev, &err_data, NULL, reset);
 
 			if (ret == AMDGPU_RAS_SUCCESS && obj) {
 				obj->err_data.ue_count += err_data.ue_count;
 				obj->err_data.ce_count += err_data.ce_count;
+				obj->err_data.de_count += err_data.de_count;
+			}
+
+			amdgpu_ras_error_data_fini(&err_data);
+		} else if (amdgpu_uniras_enabled(adev)) {
+			struct ras_ih_info ih_info = {0};
+
+			ih_info.block = block;
+			ih_info.pasid = pasid;
+			ih_info.reset = reset;
+			ih_info.pasid_fn = pasid_fn;
+			ih_info.data = data;
+			amdgpu_ras_mgr_handle_consumer_interrupt(adev, &ih_info);
+		} else {
+			struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+			int ret;
+
+			ret = amdgpu_ras_put_poison_req(adev,
+				block, pasid, pasid_fn, data, reset);
+			if (!ret) {
+				atomic_inc(&con->page_retirement_req_cnt);
+				atomic_inc(&con->poison_consumption_count);
+				wake_up(&con->page_retirement_wq);
 			}
-		} else if (reset) {
-			/* MCA poison handler is only responsible for GPU reset,
-			 * let MCA notifier do page retirement.
-			 */
-			kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
-			amdgpu_ras_reset_gpu(adev);
 		}
 	} else {
 		if (adev->virt.ops && adev->virt.ops->ras_poison_handler)
-			adev->virt.ops->ras_poison_handler(adev);
+			adev->virt.ops->ras_poison_handler(adev, block);
 		else
 			dev_warn(adev->dev,
 				"No ras_poison_handler interface in SRIOV!\n");
@@ -201,11 +306,19 @@ int amdgpu_umc_poison_handler(struct amdgpu_device *adev, bool reset)
 	return ret;
 }
 
+int amdgpu_umc_poison_handler(struct amdgpu_device *adev,
+			enum amdgpu_ras_block block, uint32_t reset)
+{
+	return amdgpu_umc_pasid_poison_handler(adev,
+				block, 0, NULL, NULL, reset);
+}
+
 int amdgpu_umc_process_ras_data_cb(struct amdgpu_device *adev,
 		void *ras_error_status,
 		struct amdgpu_iv_entry *entry)
 {
-	return amdgpu_umc_do_page_retirement(adev, ras_error_status, entry, true);
+	return amdgpu_umc_do_page_retirement(adev, ras_error_status, entry,
+				AMDGPU_RAS_GPU_RESET_MODE1_RESET);
 }
 
 int amdgpu_umc_ras_sw_init(struct amdgpu_device *adev)
@@ -246,6 +359,9 @@ int amdgpu_umc_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *r
 	if (r)
 		return r;
 
+	if (amdgpu_sriov_vf(adev))
+		return r;
+
 	if (amdgpu_ras_is_supported(adev, ras_block->block)) {
 		r = amdgpu_irq_get(adev, &adev->gmc.ecc_irq, 0);
 		if (r)
@@ -282,14 +398,20 @@ int amdgpu_umc_process_ecc_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-void amdgpu_umc_fill_error_record(struct ras_err_data *err_data,
+int amdgpu_umc_fill_error_record(struct ras_err_data *err_data,
 		uint64_t err_addr,
 		uint64_t retired_page,
 		uint32_t channel_index,
 		uint32_t umc_inst)
 {
-	struct eeprom_table_record *err_rec =
-		&err_data->err_addr[err_data->err_addr_cnt];
+	struct eeprom_table_record *err_rec;
+
+	if (!err_data ||
+	    !err_data->err_addr ||
+	    (err_data->err_addr_cnt >= err_data->err_addr_len))
+		return -EINVAL;
+
+	err_rec = &err_data->err_addr[err_data->err_addr_cnt];
 
 	err_rec->address = err_addr;
 	/* page frame address is saved */
@@ -301,4 +423,205 @@ void amdgpu_umc_fill_error_record(struct ras_err_data *err_data,
 	err_rec->mcumc_id = umc_inst;
 
 	err_data->err_addr_cnt++;
+
+	return 0;
+}
+
+static int amdgpu_umc_loop_all_aid(struct amdgpu_device *adev, umc_func func,
+				   void *data)
+{
+	uint32_t umc_node_inst;
+	uint32_t node_inst;
+	uint32_t umc_inst;
+	uint32_t ch_inst;
+	int ret;
+
+	/*
+	 * This loop is done based on the following -
+	 * umc.active mask = mask of active umc instances across all nodes
+	 * umc.umc_inst_num = maximum number of umc instancess per node
+	 * umc.node_inst_num = maximum number of node instances
+	 * Channel instances are not assumed to be harvested.
+	 */
+	dev_dbg(adev->dev, "active umcs :%lx umc_inst per node: %d",
+		adev->umc.active_mask, adev->umc.umc_inst_num);
+	for_each_set_bit(umc_node_inst, &(adev->umc.active_mask),
+			 adev->umc.node_inst_num * adev->umc.umc_inst_num) {
+		node_inst = umc_node_inst / adev->umc.umc_inst_num;
+		umc_inst = umc_node_inst % adev->umc.umc_inst_num;
+		LOOP_UMC_CH_INST(ch_inst) {
+			dev_dbg(adev->dev,
+				"node_inst :%d umc_inst: %d ch_inst: %d",
+				node_inst, umc_inst, ch_inst);
+			ret = func(adev, node_inst, umc_inst, ch_inst, data);
+			if (ret) {
+				dev_err(adev->dev,
+					"Node %d umc %d ch %d func returns %d\n",
+					node_inst, umc_inst, ch_inst, ret);
+				return ret;
+			}
+		}
+	}
+
+	return 0;
+}
+
+int amdgpu_umc_loop_channels(struct amdgpu_device *adev,
+			umc_func func, void *data)
+{
+	uint32_t node_inst       = 0;
+	uint32_t umc_inst        = 0;
+	uint32_t ch_inst         = 0;
+	int ret = 0;
+
+	if (adev->aid_mask)
+		return amdgpu_umc_loop_all_aid(adev, func, data);
+
+	if (adev->umc.node_inst_num) {
+		LOOP_UMC_EACH_NODE_INST_AND_CH(node_inst, umc_inst, ch_inst) {
+			ret = func(adev, node_inst, umc_inst, ch_inst, data);
+			if (ret) {
+				dev_err(adev->dev, "Node %d umc %d ch %d func returns %d\n",
+					node_inst, umc_inst, ch_inst, ret);
+				return ret;
+			}
+		}
+	} else {
+		LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) {
+			ret = func(adev, 0, umc_inst, ch_inst, data);
+			if (ret) {
+				dev_err(adev->dev, "Umc %d ch %d func returns %d\n",
+					umc_inst, ch_inst, ret);
+				return ret;
+			}
+		}
+	}
+
+	return 0;
+}
+
+int amdgpu_umc_update_ecc_status(struct amdgpu_device *adev,
+				uint64_t status, uint64_t ipid, uint64_t addr)
+{
+	if (adev->umc.ras->update_ecc_status)
+		return adev->umc.ras->update_ecc_status(adev,
+					status, ipid, addr);
+	return 0;
+}
+
+int amdgpu_umc_logs_ecc_err(struct amdgpu_device *adev,
+		struct radix_tree_root *ecc_tree, struct ras_ecc_err *ecc_err)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+	struct ras_ecc_log_info *ecc_log;
+	int ret;
+
+	ecc_log = &con->umc_ecc_log;
+
+	mutex_lock(&ecc_log->lock);
+	ret = radix_tree_insert(ecc_tree, ecc_err->pa_pfn, ecc_err);
+	if (!ret)
+		radix_tree_tag_set(ecc_tree,
+			ecc_err->pa_pfn, UMC_ECC_NEW_DETECTED_TAG);
+	mutex_unlock(&ecc_log->lock);
+
+	return ret;
+}
+
+int amdgpu_umc_pages_in_a_row(struct amdgpu_device *adev,
+			struct ras_err_data *err_data, uint64_t pa_addr)
+{
+	struct ta_ras_query_address_output addr_out;
+
+	/* reinit err_data */
+	err_data->err_addr_cnt = 0;
+	err_data->err_addr_len = adev->umc.retire_unit;
+
+	addr_out.pa.pa = pa_addr;
+	if (adev->umc.ras && adev->umc.ras->convert_ras_err_addr)
+		return adev->umc.ras->convert_ras_err_addr(adev, err_data, NULL,
+				&addr_out, false);
+	else
+		return -EINVAL;
+}
+
+int amdgpu_umc_lookup_bad_pages_in_a_row(struct amdgpu_device *adev,
+			uint64_t pa_addr, uint64_t *pfns, int len)
+{
+	int i, ret;
+	struct ras_err_data err_data;
+
+	err_data.err_addr = kcalloc(adev->umc.retire_unit,
+				sizeof(struct eeprom_table_record), GFP_KERNEL);
+	if (!err_data.err_addr) {
+		dev_warn(adev->dev, "Failed to alloc memory in bad page lookup!\n");
+		return 0;
+	}
+
+	ret = amdgpu_umc_pages_in_a_row(adev, &err_data, pa_addr);
+	if (ret)
+		goto out;
+
+	for (i = 0; i < adev->umc.retire_unit; i++) {
+		if (i >= len)
+			goto out;
+
+		pfns[i] = err_data.err_addr[i].retired_page;
+	}
+	ret = i;
+	adev->umc.err_addr_cnt = err_data.err_addr_cnt;
+
+out:
+	kfree(err_data.err_addr);
+	return ret;
+}
+
+int amdgpu_umc_mca_to_addr(struct amdgpu_device *adev,
+			uint64_t err_addr, uint32_t ch, uint32_t umc,
+			uint32_t node, uint32_t socket,
+			struct ta_ras_query_address_output *addr_out, bool dump_addr)
+{
+	struct ta_ras_query_address_input addr_in;
+	int ret;
+
+	memset(&addr_in, 0, sizeof(addr_in));
+	addr_in.ma.err_addr = err_addr;
+	addr_in.ma.ch_inst = ch;
+	addr_in.ma.umc_inst = umc;
+	addr_in.ma.node_inst = node;
+	addr_in.ma.socket_id = socket;
+
+	if (adev->umc.ras && adev->umc.ras->convert_ras_err_addr) {
+		ret = adev->umc.ras->convert_ras_err_addr(adev, NULL, &addr_in,
+				addr_out, dump_addr);
+		if (ret)
+			return ret;
+	} else {
+		return 0;
+	}
+
+	return 0;
+}
+
+int amdgpu_umc_pa2mca(struct amdgpu_device *adev,
+		uint64_t pa, uint64_t *mca, enum amdgpu_memory_partition nps)
+{
+	struct ta_ras_query_address_input addr_in;
+	struct ta_ras_query_address_output addr_out;
+	int ret;
+
+	/* nps: the pa belongs to */
+	addr_in.pa.pa = pa | ((uint64_t)nps << 58);
+	addr_in.addr_type = TA_RAS_PA_TO_MCA;
+	ret = psp_ras_query_address(&adev->psp, &addr_in, &addr_out);
+	if (ret) {
+		dev_warn(adev->dev, "Failed to query RAS MCA address for 0x%llx",
+			pa);
+
+		return ret;
+	}
+
+	*mca = addr_out.ma.err_addr;
+
+	return 0;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h
index d7f1229ff11f..28dff750c47e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umc.h
@@ -21,7 +21,7 @@
 #ifndef __AMDGPU_UMC_H__
 #define __AMDGPU_UMC_H__
 #include "amdgpu_ras.h"
-
+#include "amdgpu_mca.h"
 /*
  * (addr / 256) * 4096, the higher 26 bits in ErrorAddr
  * is the index of 4KB block
@@ -32,6 +32,11 @@
  * is the index of 8KB block
  */
 #define ADDR_OF_8KB_BLOCK(addr)			(((addr) & ~0xffULL) << 5)
+/*
+ * (addr / 256) * 32768, the higher 26 bits in ErrorAddr
+ * is the index of 8KB block
+ */
+#define ADDR_OF_32KB_BLOCK(addr)			(((addr) & ~0xffULL) << 7)
 /* channel index is the index of 256B block */
 #define ADDR_OF_256B_BLOCK(channel_index)	((channel_index) << 8)
 /* offset in 256B block */
@@ -47,6 +52,47 @@
 #define LOOP_UMC_EACH_NODE_INST_AND_CH(node_inst, umc_inst, ch_inst) \
 		LOOP_UMC_NODE_INST((node_inst)) LOOP_UMC_INST_AND_CH((umc_inst), (ch_inst))
 
+/* Page retirement tag */
+#define UMC_ECC_NEW_DETECTED_TAG       0x1
+/*
+ * a flag to indicate v2 of channel index stored in eeprom
+ *
+ * v1 (legacy way): store channel index within a umc instance in eeprom
+ *    range in UMC v12: 0 ~ 7
+ * v2: store global channel index in eeprom
+ *    range in UMC v12: 0 ~ 127
+ *
+ * NOTE: it's better to store it in eeprom_table_record.mem_channel,
+ * but there is only 8 bits in mem_channel, and the channel number may
+ * increase in the future, we decide to save it in
+ * eeprom_table_record.retired_page. retired_page is useless in v2,
+ * we depend on eeprom_table_record.address instead of retired_page in v2.
+ * Only 48 bits are saved on eeprom, use bit 47 here.
+ */
+#define UMC_CHANNEL_IDX_V2	BIT_ULL(47)
+
+/*
+ * save nps value to eeprom_table_record.retired_page[47:40],
+ * the channel index flag above will be retired.
+ */
+#define UMC_NPS_SHIFT 40
+#define UMC_NPS_MASK 0xffULL
+
+/* three column bits and one row bit in MCA address flip
+ * in bad page retirement
+ */
+#define RETIRE_FLIP_BITS_NUM 4
+
+struct amdgpu_umc_flip_bits {
+	uint32_t flip_bits_in_pa[RETIRE_FLIP_BITS_NUM];
+	uint32_t flip_row_bit;
+	uint32_t r13_in_pa;
+	uint32_t bit_num;
+};
+
+typedef int (*umc_func)(struct amdgpu_device *adev, uint32_t node_inst,
+			uint32_t umc_inst, uint32_t ch_inst, void *data);
+
 struct amdgpu_umc_ras {
 	struct amdgpu_ras_block_object ras_block;
 	void (*err_cnt_init)(struct amdgpu_device *adev);
@@ -55,6 +101,20 @@ struct amdgpu_umc_ras {
 				      void *ras_error_status);
 	void (*ecc_info_query_ras_error_address)(struct amdgpu_device *adev,
 					void *ras_error_status);
+	bool (*check_ecc_err_status)(struct amdgpu_device *adev,
+			enum amdgpu_mca_error_type type, void *ras_error_status);
+	int (*update_ecc_status)(struct amdgpu_device *adev,
+			uint64_t status, uint64_t ipid, uint64_t addr);
+	int (*convert_ras_err_addr)(struct amdgpu_device *adev,
+			struct ras_err_data *err_data,
+			struct ta_ras_query_address_input *addr_in,
+			struct ta_ras_query_address_output *addr_out,
+			bool dump_addr);
+	uint32_t (*get_die_id_from_pa)(struct amdgpu_device *adev,
+			uint64_t mca_addr, uint64_t retired_page);
+	void (*get_retire_flip_bits)(struct amdgpu_device *adev);
+	void (*mca_ipid_parse)(struct amdgpu_device *adev, uint64_t ipid,
+			uint32_t *did, uint32_t *ch, uint32_t *umc_inst, uint32_t *sid);
 };
 
 struct amdgpu_umc_funcs {
@@ -85,15 +145,23 @@ struct amdgpu_umc {
 
 	/* active mask for umc node instance */
 	unsigned long active_mask;
+
+	struct amdgpu_umc_flip_bits flip_bits;
+
+	unsigned long err_addr_cnt;
 };
 
 int amdgpu_umc_ras_sw_init(struct amdgpu_device *adev);
 int amdgpu_umc_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block);
-int amdgpu_umc_poison_handler(struct amdgpu_device *adev, bool reset);
+int amdgpu_umc_poison_handler(struct amdgpu_device *adev,
+			enum amdgpu_ras_block block, uint32_t reset);
+int amdgpu_umc_pasid_poison_handler(struct amdgpu_device *adev,
+			enum amdgpu_ras_block block, uint16_t pasid,
+			pasid_notify pasid_fn, void *data, uint32_t reset);
 int amdgpu_umc_process_ecc_irq(struct amdgpu_device *adev,
 		struct amdgpu_irq_src *source,
 		struct amdgpu_iv_entry *entry);
-void amdgpu_umc_fill_error_record(struct ras_err_data *err_data,
+int amdgpu_umc_fill_error_record(struct ras_err_data *err_data,
 		uint64_t err_addr,
 		uint64_t retired_page,
 		uint32_t channel_index,
@@ -104,4 +172,25 @@ int amdgpu_umc_process_ras_data_cb(struct amdgpu_device *adev,
 		struct amdgpu_iv_entry *entry);
 int amdgpu_umc_page_retirement_mca(struct amdgpu_device *adev,
 			uint64_t err_addr, uint32_t ch_inst, uint32_t umc_inst);
+
+int amdgpu_umc_loop_channels(struct amdgpu_device *adev,
+			umc_func func, void *data);
+
+int amdgpu_umc_update_ecc_status(struct amdgpu_device *adev,
+				uint64_t status, uint64_t ipid, uint64_t addr);
+int amdgpu_umc_logs_ecc_err(struct amdgpu_device *adev,
+		struct radix_tree_root *ecc_tree, struct ras_ecc_err *ecc_err);
+
+void amdgpu_umc_handle_bad_pages(struct amdgpu_device *adev,
+			void *ras_error_status);
+int amdgpu_umc_pages_in_a_row(struct amdgpu_device *adev,
+			struct ras_err_data *err_data, uint64_t pa_addr);
+int amdgpu_umc_lookup_bad_pages_in_a_row(struct amdgpu_device *adev,
+			uint64_t pa_addr, uint64_t *pfns, int len);
+int amdgpu_umc_mca_to_addr(struct amdgpu_device *adev,
+			uint64_t err_addr, uint32_t ch, uint32_t umc,
+			uint32_t node, uint32_t socket,
+			struct ta_ras_query_address_output *addr_out, bool dump_addr);
+int amdgpu_umc_pa2mca(struct amdgpu_device *adev,
+		uint64_t pa, uint64_t *mca, enum amdgpu_memory_partition nps);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umr.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_umr.h
index 919d9d401750..5b27fc41ffbf 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_umr.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umr.h
@@ -35,17 +35,51 @@ struct amdgpu_debugfs_regs2_iocdata {
 	} srbm;
 };
 
+struct amdgpu_debugfs_regs2_iocdata_v2 {
+	__u32 use_srbm, use_grbm, pg_lock;
+	struct {
+		__u32 se, sh, instance;
+	} grbm;
+	struct {
+		__u32 me, pipe, queue, vmid;
+	} srbm;
+	u32 xcc_id;
+};
+
+struct amdgpu_debugfs_gprwave_iocdata {
+	u32 gpr_or_wave, se, sh, cu, wave, simd, xcc_id;
+	struct {
+		u32 thread, vpgr_or_sgpr;
+	} gpr;
+};
+
 /*
  * MMIO debugfs state data (per file* handle)
  */
 struct amdgpu_debugfs_regs2_data {
 	struct amdgpu_device *adev;
 	struct mutex lock;
-	struct amdgpu_debugfs_regs2_iocdata id;
+	struct amdgpu_debugfs_regs2_iocdata_v2 id;
+};
+
+struct amdgpu_debugfs_gprwave_data {
+	struct amdgpu_device *adev;
+	struct mutex lock;
+	struct amdgpu_debugfs_gprwave_iocdata id;
 };
 
 enum AMDGPU_DEBUGFS_REGS2_CMDS {
-	AMDGPU_DEBUGFS_REGS2_CMD_SET_STATE=0,
+	AMDGPU_DEBUGFS_REGS2_CMD_SET_STATE = 0,
+	AMDGPU_DEBUGFS_REGS2_CMD_SET_STATE_V2,
+};
+
+enum AMDGPU_DEBUGFS_GPRWAVE_CMDS {
+	AMDGPU_DEBUGFS_GPRWAVE_CMD_SET_STATE = 0,
 };
 
+//reg2 interface
 #define AMDGPU_DEBUGFS_REGS2_IOC_SET_STATE _IOWR(0x20, AMDGPU_DEBUGFS_REGS2_CMD_SET_STATE, struct amdgpu_debugfs_regs2_iocdata)
+#define AMDGPU_DEBUGFS_REGS2_IOC_SET_STATE_V2 _IOWR(0x20, AMDGPU_DEBUGFS_REGS2_CMD_SET_STATE_V2, struct amdgpu_debugfs_regs2_iocdata_v2)
+
+//gprwave interface
+#define AMDGPU_DEBUGFS_GPRWAVE_IOC_SET_STATE _IOWR(0x20, AMDGPU_DEBUGFS_GPRWAVE_CMD_SET_STATE, struct amdgpu_debugfs_gprwave_iocdata)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c
new file mode 100644
index 000000000000..cd707d70a0bf
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.c
@@ -0,0 +1,550 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/firmware.h>
+#include <linux/module.h>
+#include <linux/debugfs.h>
+#include <drm/drm_exec.h>
+#include <drm/drm_drv.h>
+
+#include "amdgpu.h"
+#include "amdgpu_umsch_mm.h"
+#include "umsch_mm_v4_0.h"
+
+MODULE_FIRMWARE("amdgpu/umsch_mm_4_0_0.bin");
+
+int amdgpu_umsch_mm_submit_pkt(struct amdgpu_umsch_mm *umsch, void *pkt, int ndws)
+{
+	struct amdgpu_ring *ring = &umsch->ring;
+
+	if (amdgpu_ring_alloc(ring, ndws))
+		return -ENOMEM;
+
+	amdgpu_ring_write_multiple(ring, pkt, ndws);
+	amdgpu_ring_commit(ring);
+
+	return 0;
+}
+
+int amdgpu_umsch_mm_query_fence(struct amdgpu_umsch_mm *umsch)
+{
+	struct amdgpu_ring *ring = &umsch->ring;
+	struct amdgpu_device *adev = ring->adev;
+	int r;
+
+	r = amdgpu_fence_wait_polling(ring, ring->fence_drv.sync_seq, adev->usec_timeout);
+	if (r < 1) {
+		dev_err(adev->dev, "ring umsch timeout, emitted fence %u\n",
+			ring->fence_drv.sync_seq);
+		return -ETIMEDOUT;
+	}
+
+	return 0;
+}
+
+static void umsch_mm_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_umsch_mm *umsch = (struct amdgpu_umsch_mm *)ring;
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->use_doorbell)
+		WDOORBELL32(ring->doorbell_index, ring->wptr << 2);
+	else
+		WREG32(umsch->rb_wptr, ring->wptr << 2);
+}
+
+static u64 umsch_mm_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_umsch_mm *umsch = (struct amdgpu_umsch_mm *)ring;
+	struct amdgpu_device *adev = ring->adev;
+
+	return RREG32(umsch->rb_rptr);
+}
+
+static u64 umsch_mm_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_umsch_mm *umsch = (struct amdgpu_umsch_mm *)ring;
+	struct amdgpu_device *adev = ring->adev;
+
+	return RREG32(umsch->rb_wptr);
+}
+
+static const struct amdgpu_ring_funcs umsch_v4_0_ring_funcs = {
+	.type = AMDGPU_RING_TYPE_UMSCH_MM,
+	.align_mask = 0,
+	.nop = 0,
+	.support_64bit_ptrs = false,
+	.get_rptr = umsch_mm_ring_get_rptr,
+	.get_wptr = umsch_mm_ring_get_wptr,
+	.set_wptr = umsch_mm_ring_set_wptr,
+	.insert_nop = amdgpu_ring_insert_nop,
+};
+
+int amdgpu_umsch_mm_ring_init(struct amdgpu_umsch_mm *umsch)
+{
+	struct amdgpu_device *adev = container_of(umsch, struct amdgpu_device, umsch_mm);
+	struct amdgpu_ring *ring = &umsch->ring;
+
+	ring->vm_hub = AMDGPU_MMHUB0(0);
+	ring->use_doorbell = true;
+	ring->no_scheduler = true;
+	ring->doorbell_index = (AMDGPU_NAVI10_DOORBELL64_VCN0_1 << 1) + 6;
+
+	snprintf(ring->name, sizeof(ring->name), "umsch");
+
+	return amdgpu_ring_init(adev, ring, 1024, NULL, 0, AMDGPU_RING_PRIO_DEFAULT, NULL);
+}
+
+int amdgpu_umsch_mm_init_microcode(struct amdgpu_umsch_mm *umsch)
+{
+	const struct umsch_mm_firmware_header_v1_0 *umsch_mm_hdr;
+	struct amdgpu_device *adev = umsch->ring.adev;
+	const char *fw_name = NULL;
+	int r;
+
+	switch (amdgpu_ip_version(adev, VCN_HWIP, 0)) {
+	case IP_VERSION(4, 0, 5):
+	case IP_VERSION(4, 0, 6):
+		fw_name = "4_0_0";
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	r = amdgpu_ucode_request(adev, &adev->umsch_mm.fw, AMDGPU_UCODE_REQUIRED,
+				 "amdgpu/umsch_mm_%s.bin", fw_name);
+	if (r) {
+		release_firmware(adev->umsch_mm.fw);
+		adev->umsch_mm.fw = NULL;
+		return r;
+	}
+
+	umsch_mm_hdr = (const struct umsch_mm_firmware_header_v1_0 *)adev->umsch_mm.fw->data;
+
+	adev->umsch_mm.ucode_size = le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_size_bytes);
+	adev->umsch_mm.data_size = le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_data_size_bytes);
+
+	adev->umsch_mm.irq_start_addr =
+		le32_to_cpu(umsch_mm_hdr->umsch_mm_irq_start_addr_lo) |
+		((uint64_t)(le32_to_cpu(umsch_mm_hdr->umsch_mm_irq_start_addr_hi)) << 32);
+	adev->umsch_mm.uc_start_addr =
+		le32_to_cpu(umsch_mm_hdr->umsch_mm_uc_start_addr_lo) |
+		((uint64_t)(le32_to_cpu(umsch_mm_hdr->umsch_mm_uc_start_addr_hi)) << 32);
+	adev->umsch_mm.data_start_addr =
+		le32_to_cpu(umsch_mm_hdr->umsch_mm_data_start_addr_lo) |
+		((uint64_t)(le32_to_cpu(umsch_mm_hdr->umsch_mm_data_start_addr_hi)) << 32);
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		struct amdgpu_firmware_info *info;
+
+		info = &adev->firmware.ucode[AMDGPU_UCODE_ID_UMSCH_MM_UCODE];
+		info->ucode_id = AMDGPU_UCODE_ID_UMSCH_MM_UCODE;
+		info->fw = adev->umsch_mm.fw;
+		adev->firmware.fw_size +=
+			ALIGN(le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_size_bytes), PAGE_SIZE);
+
+		info = &adev->firmware.ucode[AMDGPU_UCODE_ID_UMSCH_MM_DATA];
+		info->ucode_id = AMDGPU_UCODE_ID_UMSCH_MM_DATA;
+		info->fw = adev->umsch_mm.fw;
+		adev->firmware.fw_size +=
+			ALIGN(le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_data_size_bytes), PAGE_SIZE);
+	}
+
+	return 0;
+}
+
+int amdgpu_umsch_mm_allocate_ucode_buffer(struct amdgpu_umsch_mm *umsch)
+{
+	const struct umsch_mm_firmware_header_v1_0 *umsch_mm_hdr;
+	struct amdgpu_device *adev = umsch->ring.adev;
+	const __le32 *fw_data;
+	uint32_t fw_size;
+	int r;
+
+	umsch_mm_hdr = (const struct umsch_mm_firmware_header_v1_0 *)
+		       adev->umsch_mm.fw->data;
+
+	fw_data = (const __le32 *)(adev->umsch_mm.fw->data +
+		  le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_offset_bytes));
+	fw_size = le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_size_bytes);
+
+	r = amdgpu_bo_create_reserved(adev, fw_size,
+				      4 * 1024, AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->umsch_mm.ucode_fw_obj,
+				      &adev->umsch_mm.ucode_fw_gpu_addr,
+				      (void **)&adev->umsch_mm.ucode_fw_ptr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create umsch_mm fw ucode bo\n", r);
+		return r;
+	}
+
+	memcpy(adev->umsch_mm.ucode_fw_ptr, fw_data, fw_size);
+
+	amdgpu_bo_kunmap(adev->umsch_mm.ucode_fw_obj);
+	amdgpu_bo_unreserve(adev->umsch_mm.ucode_fw_obj);
+	return 0;
+}
+
+int amdgpu_umsch_mm_allocate_ucode_data_buffer(struct amdgpu_umsch_mm *umsch)
+{
+	const struct umsch_mm_firmware_header_v1_0 *umsch_mm_hdr;
+	struct amdgpu_device *adev = umsch->ring.adev;
+	const __le32 *fw_data;
+	uint32_t fw_size;
+	int r;
+
+	umsch_mm_hdr = (const struct umsch_mm_firmware_header_v1_0 *)
+		       adev->umsch_mm.fw->data;
+
+	fw_data = (const __le32 *)(adev->umsch_mm.fw->data +
+		  le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_data_offset_bytes));
+	fw_size = le32_to_cpu(umsch_mm_hdr->umsch_mm_ucode_data_size_bytes);
+
+	r = amdgpu_bo_create_reserved(adev, fw_size,
+				      64 * 1024, AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->umsch_mm.data_fw_obj,
+				      &adev->umsch_mm.data_fw_gpu_addr,
+				      (void **)&adev->umsch_mm.data_fw_ptr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create umsch_mm fw data bo\n", r);
+		return r;
+	}
+
+	memcpy(adev->umsch_mm.data_fw_ptr, fw_data, fw_size);
+
+	amdgpu_bo_kunmap(adev->umsch_mm.data_fw_obj);
+	amdgpu_bo_unreserve(adev->umsch_mm.data_fw_obj);
+	return 0;
+}
+
+int amdgpu_umsch_mm_psp_execute_cmd_buf(struct amdgpu_umsch_mm *umsch)
+{
+	struct amdgpu_device *adev = umsch->ring.adev;
+	struct amdgpu_firmware_info ucode = {
+		.ucode_id = AMDGPU_UCODE_ID_UMSCH_MM_CMD_BUFFER,
+		.mc_addr = adev->umsch_mm.cmd_buf_gpu_addr,
+		.ucode_size = ((uintptr_t)adev->umsch_mm.cmd_buf_curr_ptr -
+			      (uintptr_t)adev->umsch_mm.cmd_buf_ptr),
+	};
+
+	return psp_execute_ip_fw_load(&adev->psp, &ucode);
+}
+
+static void umsch_mm_agdb_index_init(struct amdgpu_device *adev)
+{
+	uint32_t umsch_mm_agdb_start;
+	int i;
+
+	umsch_mm_agdb_start = adev->doorbell_index.max_assignment + 1;
+	umsch_mm_agdb_start = roundup(umsch_mm_agdb_start, 1024);
+	umsch_mm_agdb_start += (AMDGPU_NAVI10_DOORBELL64_VCN0_1 << 1);
+
+	for (i = 0; i < CONTEXT_PRIORITY_NUM_LEVELS; i++)
+		adev->umsch_mm.agdb_index[i] = umsch_mm_agdb_start + i;
+}
+
+static int umsch_mm_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	adev->umsch_mm.vmid_mask_mm_vpe = 0xf00;
+	adev->umsch_mm.engine_mask = (1 << UMSCH_SWIP_ENGINE_TYPE_VPE);
+	adev->umsch_mm.vpe_hqd_mask = 0xfe;
+
+	r = amdgpu_device_wb_get(adev, &adev->umsch_mm.wb_index);
+	if (r) {
+		dev_err(adev->dev, "failed to alloc wb for umsch: %d\n", r);
+		return r;
+	}
+
+	adev->umsch_mm.sch_ctx_gpu_addr = adev->wb.gpu_addr +
+					  (adev->umsch_mm.wb_index * 4);
+
+	r = amdgpu_bo_create_kernel(adev, PAGE_SIZE, PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_GTT,
+				    &adev->umsch_mm.cmd_buf_obj,
+				    &adev->umsch_mm.cmd_buf_gpu_addr,
+				    (void **)&adev->umsch_mm.cmd_buf_ptr);
+	if (r) {
+		dev_err(adev->dev, "failed to allocate cmdbuf bo %d\n", r);
+		amdgpu_device_wb_free(adev, adev->umsch_mm.wb_index);
+		return r;
+	}
+
+	r = amdgpu_bo_create_kernel(adev, AMDGPU_UMSCHFW_LOG_SIZE, PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_VRAM |
+				    AMDGPU_GEM_DOMAIN_GTT,
+				    &adev->umsch_mm.dbglog_bo,
+				    &adev->umsch_mm.log_gpu_addr,
+				    &adev->umsch_mm.log_cpu_addr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to allocate umsch debug bo\n", r);
+		return r;
+	}
+
+	mutex_init(&adev->umsch_mm.mutex_hidden);
+
+	umsch_mm_agdb_index_init(adev);
+
+	return 0;
+}
+
+
+static int umsch_mm_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	switch (amdgpu_ip_version(adev, VCN_HWIP, 0)) {
+	case IP_VERSION(4, 0, 5):
+	case IP_VERSION(4, 0, 6):
+		umsch_mm_v4_0_set_funcs(&adev->umsch_mm);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	adev->umsch_mm.ring.funcs = &umsch_v4_0_ring_funcs;
+	umsch_mm_set_regs(&adev->umsch_mm);
+
+	return 0;
+}
+
+static int umsch_mm_late_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (amdgpu_in_reset(adev) || adev->in_s0ix || adev->in_suspend)
+		return 0;
+
+	return 0;
+}
+
+static int umsch_mm_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = umsch_mm_init(adev);
+	if (r)
+		return r;
+
+	amdgpu_umsch_fwlog_init(&adev->umsch_mm);
+	r = umsch_mm_ring_init(&adev->umsch_mm);
+	if (r)
+		return r;
+
+	r = umsch_mm_init_microcode(&adev->umsch_mm);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static int umsch_mm_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	release_firmware(adev->umsch_mm.fw);
+	adev->umsch_mm.fw = NULL;
+
+	amdgpu_ring_fini(&adev->umsch_mm.ring);
+
+	mutex_destroy(&adev->umsch_mm.mutex_hidden);
+
+	amdgpu_bo_free_kernel(&adev->umsch_mm.cmd_buf_obj,
+			      &adev->umsch_mm.cmd_buf_gpu_addr,
+			      (void **)&adev->umsch_mm.cmd_buf_ptr);
+
+	amdgpu_bo_free_kernel(&adev->umsch_mm.dbglog_bo,
+				    &adev->umsch_mm.log_gpu_addr,
+				    (void **)&adev->umsch_mm.log_cpu_addr);
+
+	amdgpu_device_wb_free(adev, adev->umsch_mm.wb_index);
+
+	return 0;
+}
+
+static int umsch_mm_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = umsch_mm_load_microcode(&adev->umsch_mm);
+	if (r)
+		return r;
+
+	umsch_mm_ring_start(&adev->umsch_mm);
+
+	r = umsch_mm_set_hw_resources(&adev->umsch_mm);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static int umsch_mm_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	umsch_mm_ring_stop(&adev->umsch_mm);
+
+	amdgpu_bo_free_kernel(&adev->umsch_mm.data_fw_obj,
+			      &adev->umsch_mm.data_fw_gpu_addr,
+			      (void **)&adev->umsch_mm.data_fw_ptr);
+
+	amdgpu_bo_free_kernel(&adev->umsch_mm.ucode_fw_obj,
+			      &adev->umsch_mm.ucode_fw_gpu_addr,
+			      (void **)&adev->umsch_mm.ucode_fw_ptr);
+	return 0;
+}
+
+static int umsch_mm_suspend(struct amdgpu_ip_block *ip_block)
+{
+	return umsch_mm_hw_fini(ip_block);
+}
+
+static int umsch_mm_resume(struct amdgpu_ip_block *ip_block)
+{
+	return umsch_mm_hw_init(ip_block);
+}
+
+void amdgpu_umsch_fwlog_init(struct amdgpu_umsch_mm *umsch_mm)
+{
+#if defined(CONFIG_DEBUG_FS)
+	void *fw_log_cpu_addr = umsch_mm->log_cpu_addr;
+	volatile struct amdgpu_umsch_fwlog *log_buf = fw_log_cpu_addr;
+
+	log_buf->header_size = sizeof(struct amdgpu_umsch_fwlog);
+	log_buf->buffer_size = AMDGPU_UMSCHFW_LOG_SIZE;
+	log_buf->rptr = log_buf->header_size;
+	log_buf->wptr = log_buf->header_size;
+	log_buf->wrapped = 0;
+#endif
+}
+
+/*
+ * debugfs for mapping umsch firmware log buffer.
+ */
+#if defined(CONFIG_DEBUG_FS)
+static ssize_t amdgpu_debugfs_umsch_fwlog_read(struct file *f, char __user *buf,
+					     size_t size, loff_t *pos)
+{
+	struct amdgpu_umsch_mm *umsch_mm;
+	void *log_buf;
+	volatile struct amdgpu_umsch_fwlog *plog;
+	unsigned int read_pos, write_pos, available, i, read_bytes = 0;
+	unsigned int read_num[2] = {0};
+
+	umsch_mm = file_inode(f)->i_private;
+	if (!umsch_mm)
+		return -ENODEV;
+
+	if (!umsch_mm->log_cpu_addr)
+		return -EFAULT;
+
+	log_buf = umsch_mm->log_cpu_addr;
+
+	plog = (volatile struct amdgpu_umsch_fwlog *)log_buf;
+	read_pos = plog->rptr;
+	write_pos = plog->wptr;
+
+	if (read_pos > AMDGPU_UMSCHFW_LOG_SIZE || write_pos > AMDGPU_UMSCHFW_LOG_SIZE)
+		return -EFAULT;
+
+	if (!size || (read_pos == write_pos))
+		return 0;
+
+	if (write_pos > read_pos) {
+		available = write_pos - read_pos;
+		read_num[0] = min_t(size_t, size, available);
+	} else {
+		read_num[0] = AMDGPU_UMSCHFW_LOG_SIZE - read_pos;
+		available = read_num[0] + write_pos - plog->header_size;
+		if (size > available)
+			read_num[1] = write_pos - plog->header_size;
+		else if (size > read_num[0])
+			read_num[1] = size - read_num[0];
+		else
+			read_num[0] = size;
+	}
+
+	for (i = 0; i < 2; i++) {
+		if (read_num[i]) {
+			if (read_pos == AMDGPU_UMSCHFW_LOG_SIZE)
+				read_pos = plog->header_size;
+			if (read_num[i] == copy_to_user((buf + read_bytes),
+							(log_buf + read_pos), read_num[i]))
+				return -EFAULT;
+
+			read_bytes += read_num[i];
+			read_pos += read_num[i];
+		}
+	}
+
+	plog->rptr = read_pos;
+	*pos += read_bytes;
+	return read_bytes;
+}
+
+static const struct file_operations amdgpu_debugfs_umschfwlog_fops = {
+	.owner = THIS_MODULE,
+	.read = amdgpu_debugfs_umsch_fwlog_read,
+	.llseek = default_llseek
+};
+#endif
+
+void amdgpu_debugfs_umsch_fwlog_init(struct amdgpu_device *adev,
+			struct amdgpu_umsch_mm *umsch_mm)
+{
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+	char name[32];
+
+	sprintf(name, "amdgpu_umsch_fwlog");
+	debugfs_create_file_size(name, S_IFREG | 0444, root, umsch_mm,
+				 &amdgpu_debugfs_umschfwlog_fops,
+				 AMDGPU_UMSCHFW_LOG_SIZE);
+#endif
+}
+
+static const struct amd_ip_funcs umsch_mm_v4_0_ip_funcs = {
+	.name = "umsch_mm_v4_0",
+	.early_init = umsch_mm_early_init,
+	.late_init = umsch_mm_late_init,
+	.sw_init = umsch_mm_sw_init,
+	.sw_fini = umsch_mm_sw_fini,
+	.hw_init = umsch_mm_hw_init,
+	.hw_fini = umsch_mm_hw_fini,
+	.suspend = umsch_mm_suspend,
+	.resume = umsch_mm_resume,
+};
+
+const struct amdgpu_ip_block_version umsch_mm_v4_0_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_UMSCH_MM,
+	.major = 4,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &umsch_mm_v4_0_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h
new file mode 100644
index 000000000000..2c771a753778
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_umsch_mm.h
@@ -0,0 +1,246 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __AMDGPU_UMSCH_MM_H__
+#define __AMDGPU_UMSCH_MM_H__
+
+enum UMSCH_SWIP_ENGINE_TYPE {
+	UMSCH_SWIP_ENGINE_TYPE_VCN0 = 0,
+	UMSCH_SWIP_ENGINE_TYPE_VCN1 = 1,
+	UMSCH_SWIP_ENGINE_TYPE_VCN = 2,
+	UMSCH_SWIP_ENGINE_TYPE_VPE = 3,
+	UMSCH_SWIP_ENGINE_TYPE_MAX
+};
+
+enum UMSCH_CONTEXT_PRIORITY_LEVEL {
+	CONTEXT_PRIORITY_LEVEL_IDLE = 0,
+	CONTEXT_PRIORITY_LEVEL_NORMAL = 1,
+	CONTEXT_PRIORITY_LEVEL_FOCUS = 2,
+	CONTEXT_PRIORITY_LEVEL_REALTIME = 3,
+	CONTEXT_PRIORITY_NUM_LEVELS
+};
+
+struct umsch_mm_set_resource_input {
+	uint32_t vmid_mask_mm_vcn;
+	uint32_t vmid_mask_mm_vpe;
+	uint32_t collaboration_mask_vpe;
+	uint32_t logging_vmid;
+	uint32_t engine_mask;
+	union {
+		struct {
+			uint32_t disable_reset : 1;
+			uint32_t disable_umsch_mm_log : 1;
+			uint32_t use_rs64mem_for_proc_ctx_csa : 1;
+			uint32_t reserved : 29;
+		};
+		uint32_t uint32_all;
+	};
+};
+
+struct amdgpu_umsch_fwlog {
+	uint32_t rptr;
+	uint32_t wptr;
+	uint32_t buffer_size;
+	uint32_t header_size;
+	uint32_t wrapped;
+};
+
+struct umsch_mm_add_queue_input {
+	uint32_t process_id;
+	uint64_t page_table_base_addr;
+	uint64_t process_va_start;
+	uint64_t process_va_end;
+	uint64_t process_quantum;
+	uint64_t process_csa_addr;
+	uint64_t context_quantum;
+	uint64_t context_csa_addr;
+	uint32_t inprocess_context_priority;
+	enum UMSCH_CONTEXT_PRIORITY_LEVEL context_global_priority_level;
+	uint32_t doorbell_offset_0;
+	uint32_t doorbell_offset_1;
+	enum UMSCH_SWIP_ENGINE_TYPE engine_type;
+	uint32_t affinity;
+	uint64_t mqd_addr;
+	uint64_t h_context;
+	uint64_t h_queue;
+	uint32_t vm_context_cntl;
+
+	uint32_t process_csa_array_index;
+	uint32_t context_csa_array_index;
+
+	struct {
+		uint32_t is_context_suspended : 1;
+		uint32_t collaboration_mode : 1;
+		uint32_t reserved : 30;
+	};
+};
+
+struct umsch_mm_remove_queue_input {
+	uint32_t doorbell_offset_0;
+	uint32_t doorbell_offset_1;
+	uint64_t context_csa_addr;
+	uint32_t context_csa_array_index;
+};
+
+struct MQD_INFO {
+	uint32_t rb_base_hi;
+	uint32_t rb_base_lo;
+	uint32_t rb_size;
+	uint32_t wptr_val;
+	uint32_t rptr_val;
+	uint32_t unmapped;
+	uint32_t vmid;
+};
+
+struct amdgpu_umsch_mm;
+
+struct umsch_mm_funcs {
+	int (*set_hw_resources)(struct amdgpu_umsch_mm *umsch);
+	int (*add_queue)(struct amdgpu_umsch_mm *umsch,
+			 struct umsch_mm_add_queue_input *input);
+	int (*remove_queue)(struct amdgpu_umsch_mm *umsch,
+			    struct umsch_mm_remove_queue_input *input);
+	int (*set_regs)(struct amdgpu_umsch_mm *umsch);
+	int (*init_microcode)(struct amdgpu_umsch_mm *umsch);
+	int (*load_microcode)(struct amdgpu_umsch_mm *umsch);
+	int (*ring_init)(struct amdgpu_umsch_mm *umsch);
+	int (*ring_start)(struct amdgpu_umsch_mm *umsch);
+	int (*ring_stop)(struct amdgpu_umsch_mm *umsch);
+	int (*ring_fini)(struct amdgpu_umsch_mm *umsch);
+};
+
+struct amdgpu_umsch_mm {
+	struct amdgpu_ring		ring;
+
+	uint32_t			rb_wptr;
+	uint32_t			rb_rptr;
+
+	const struct umsch_mm_funcs	*funcs;
+
+	const struct firmware		*fw;
+	uint32_t			fw_version;
+	uint32_t			feature_version;
+
+	struct amdgpu_bo		*ucode_fw_obj;
+	uint64_t			ucode_fw_gpu_addr;
+	uint32_t			*ucode_fw_ptr;
+	uint64_t			irq_start_addr;
+	uint64_t			uc_start_addr;
+	uint32_t			ucode_size;
+
+	struct amdgpu_bo		*data_fw_obj;
+	uint64_t			data_fw_gpu_addr;
+	uint32_t			*data_fw_ptr;
+	uint64_t			data_start_addr;
+	uint32_t			data_size;
+
+	struct amdgpu_bo		*cmd_buf_obj;
+	uint64_t			cmd_buf_gpu_addr;
+	uint32_t			*cmd_buf_ptr;
+	uint32_t			*cmd_buf_curr_ptr;
+
+	uint32_t			wb_index;
+	uint64_t			sch_ctx_gpu_addr;
+	uint32_t			*sch_ctx_cpu_addr;
+
+	uint32_t			vmid_mask_mm_vcn;
+	uint32_t			vmid_mask_mm_vpe;
+	uint32_t			engine_mask;
+	uint32_t			vcn0_hqd_mask;
+	uint32_t			vcn1_hqd_mask;
+	uint32_t			vcn_hqd_mask[2];
+	uint32_t			vpe_hqd_mask;
+	uint32_t			agdb_index[CONTEXT_PRIORITY_NUM_LEVELS];
+
+	struct mutex			mutex_hidden;
+	struct amdgpu_bo		*dbglog_bo;
+	void				*log_cpu_addr;
+	uint64_t			log_gpu_addr;
+	uint32_t			mem_size;
+	uint32_t			log_offset;
+};
+
+int amdgpu_umsch_mm_submit_pkt(struct amdgpu_umsch_mm *umsch, void *pkt, int ndws);
+int amdgpu_umsch_mm_query_fence(struct amdgpu_umsch_mm *umsch);
+
+int amdgpu_umsch_mm_init_microcode(struct amdgpu_umsch_mm *umsch);
+int amdgpu_umsch_mm_allocate_ucode_buffer(struct amdgpu_umsch_mm *umsch);
+int amdgpu_umsch_mm_allocate_ucode_data_buffer(struct amdgpu_umsch_mm *umsch);
+
+int amdgpu_umsch_mm_psp_execute_cmd_buf(struct amdgpu_umsch_mm *umsch);
+
+int amdgpu_umsch_mm_ring_init(struct amdgpu_umsch_mm *umsch);
+
+void amdgpu_debugfs_umsch_fwlog_init(struct amdgpu_device *adev,
+			struct amdgpu_umsch_mm *umsch);
+
+void amdgpu_umsch_fwlog_init(struct amdgpu_umsch_mm *umsch_mm);
+
+#define WREG32_SOC15_UMSCH(reg, value)								\
+	do {											\
+		uint32_t reg_offset = adev->reg_offset[VCN_HWIP][0][reg##_BASE_IDX] + reg;	\
+		if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {				\
+			*adev->umsch_mm.cmd_buf_curr_ptr++ = (reg_offset << 2);			\
+			*adev->umsch_mm.cmd_buf_curr_ptr++ = value;				\
+		} else	{									\
+			WREG32(reg_offset, value);						\
+		}										\
+	} while (0)
+
+#define umsch_mm_set_hw_resources(umsch) \
+	((umsch)->funcs->set_hw_resources ? (umsch)->funcs->set_hw_resources((umsch)) : 0)
+#define umsch_mm_add_queue(umsch, input) \
+	((umsch)->funcs->add_queue ? (umsch)->funcs->add_queue((umsch), (input)) : 0)
+#define umsch_mm_remove_queue(umsch, input) \
+	((umsch)->funcs->remove_queue ? (umsch)->funcs->remove_queue((umsch), (input)) : 0)
+
+#define umsch_mm_set_regs(umsch) \
+	((umsch)->funcs->set_regs ? (umsch)->funcs->set_regs((umsch)) : 0)
+#define umsch_mm_init_microcode(umsch) \
+	((umsch)->funcs->init_microcode ? (umsch)->funcs->init_microcode((umsch)) : 0)
+#define umsch_mm_load_microcode(umsch) \
+	((umsch)->funcs->load_microcode ? (umsch)->funcs->load_microcode((umsch)) : 0)
+
+#define umsch_mm_ring_init(umsch) \
+	((umsch)->funcs->ring_init ? (umsch)->funcs->ring_init((umsch)) : 0)
+#define umsch_mm_ring_start(umsch) \
+	((umsch)->funcs->ring_start ? (umsch)->funcs->ring_start((umsch)) : 0)
+#define umsch_mm_ring_stop(umsch) \
+	((umsch)->funcs->ring_stop ? (umsch)->funcs->ring_stop((umsch)) : 0)
+#define umsch_mm_ring_fini(umsch) \
+	((umsch)->funcs->ring_fini ? (umsch)->funcs->ring_fini((umsch)) : 0)
+
+static inline void amdgpu_umsch_mm_lock(struct amdgpu_umsch_mm *umsch)
+{
+	mutex_lock(&umsch->mutex_hidden);
+}
+
+static inline void amdgpu_umsch_mm_unlock(struct amdgpu_umsch_mm *umsch)
+{
+	mutex_unlock(&umsch->mutex_hidden);
+}
+
+extern const struct amdgpu_ip_block_version umsch_mm_v4_0_ip_block;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
new file mode 100644
index 000000000000..9a969175900e
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c
@@ -0,0 +1,1482 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <drm/drm_auth.h>
+#include <drm/drm_exec.h>
+#include <linux/pm_runtime.h>
+#include <drm/drm_drv.h>
+
+#include "amdgpu.h"
+#include "amdgpu_reset.h"
+#include "amdgpu_vm.h"
+#include "amdgpu_userq.h"
+#include "amdgpu_hmm.h"
+#include "amdgpu_userq_fence.h"
+
+u32 amdgpu_userq_get_supported_ip_mask(struct amdgpu_device *adev)
+{
+	int i;
+	u32 userq_ip_mask = 0;
+
+	for (i = 0; i < AMDGPU_HW_IP_NUM; i++) {
+		if (adev->userq_funcs[i])
+			userq_ip_mask |= (1 << i);
+	}
+
+	return userq_ip_mask;
+}
+
+static bool amdgpu_userq_is_reset_type_supported(struct amdgpu_device *adev,
+				enum amdgpu_ring_type ring_type, int reset_type)
+{
+
+	if (ring_type < 0 || ring_type >= AMDGPU_RING_TYPE_MAX)
+		return false;
+
+	switch (ring_type) {
+	case AMDGPU_RING_TYPE_GFX:
+		if (adev->gfx.gfx_supported_reset & reset_type)
+			return true;
+		break;
+	case AMDGPU_RING_TYPE_COMPUTE:
+		if (adev->gfx.compute_supported_reset & reset_type)
+			return true;
+		break;
+	case AMDGPU_RING_TYPE_SDMA:
+		if (adev->sdma.supported_reset & reset_type)
+			return true;
+		break;
+	case AMDGPU_RING_TYPE_VCN_DEC:
+	case AMDGPU_RING_TYPE_VCN_ENC:
+		if (adev->vcn.supported_reset & reset_type)
+			return true;
+		break;
+	case AMDGPU_RING_TYPE_VCN_JPEG:
+		if (adev->jpeg.supported_reset & reset_type)
+			return true;
+		break;
+	default:
+		break;
+	}
+	return false;
+}
+
+static void amdgpu_userq_gpu_reset(struct amdgpu_device *adev)
+{
+	if (amdgpu_device_should_recover_gpu(adev)) {
+		amdgpu_reset_domain_schedule(adev->reset_domain,
+					     &adev->userq_reset_work);
+		/* Wait for the reset job to complete */
+		flush_work(&adev->userq_reset_work);
+	}
+}
+
+static int
+amdgpu_userq_detect_and_reset_queues(struct amdgpu_userq_mgr *uq_mgr)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	const int queue_types[] = {
+		AMDGPU_RING_TYPE_COMPUTE,
+		AMDGPU_RING_TYPE_GFX,
+		AMDGPU_RING_TYPE_SDMA
+	};
+	const int num_queue_types = ARRAY_SIZE(queue_types);
+	bool gpu_reset = false;
+	int r = 0;
+	int i;
+
+	/* Warning if current process mutex is not held */
+	WARN_ON(!mutex_is_locked(&uq_mgr->userq_mutex));
+
+	if (unlikely(adev->debug_disable_gpu_ring_reset)) {
+		dev_err(adev->dev, "userq reset disabled by debug mask\n");
+		return 0;
+	}
+
+	/*
+	 * If GPU recovery feature is disabled system-wide,
+	 * skip all reset detection logic
+	 */
+	if (!amdgpu_gpu_recovery)
+		return 0;
+
+	/*
+	 * Iterate through all queue types to detect and reset problematic queues
+	 * Process each queue type in the defined order
+	 */
+	for (i = 0; i < num_queue_types; i++) {
+		int ring_type = queue_types[i];
+		const struct amdgpu_userq_funcs *funcs = adev->userq_funcs[ring_type];
+
+		if (!amdgpu_userq_is_reset_type_supported(adev, ring_type, AMDGPU_RESET_TYPE_PER_QUEUE))
+				continue;
+
+		if (atomic_read(&uq_mgr->userq_count[ring_type]) > 0 &&
+		    funcs && funcs->detect_and_reset) {
+			r = funcs->detect_and_reset(adev, ring_type);
+			if (r) {
+				gpu_reset = true;
+				break;
+			}
+		}
+	}
+
+	if (gpu_reset)
+		amdgpu_userq_gpu_reset(adev);
+
+	return r;
+}
+
+static int amdgpu_userq_buffer_va_list_add(struct amdgpu_usermode_queue *queue,
+					   struct amdgpu_bo_va_mapping *va_map, u64 addr)
+{
+	struct amdgpu_userq_va_cursor *va_cursor;
+	struct userq_va_list;
+
+	va_cursor = kzalloc(sizeof(*va_cursor), GFP_KERNEL);
+	if (!va_cursor)
+		return -ENOMEM;
+
+	INIT_LIST_HEAD(&va_cursor->list);
+	va_cursor->gpu_addr = addr;
+	atomic_set(&va_map->bo_va->userq_va_mapped, 1);
+	list_add(&va_cursor->list, &queue->userq_va_list);
+
+	return 0;
+}
+
+int amdgpu_userq_input_va_validate(struct amdgpu_usermode_queue *queue,
+				   u64 addr, u64 expected_size)
+{
+	struct amdgpu_bo_va_mapping *va_map;
+	struct amdgpu_vm *vm = queue->vm;
+	u64 user_addr;
+	u64 size;
+	int r = 0;
+
+	user_addr = (addr & AMDGPU_GMC_HOLE_MASK) >> AMDGPU_GPU_PAGE_SHIFT;
+	size = expected_size >> AMDGPU_GPU_PAGE_SHIFT;
+
+	r = amdgpu_bo_reserve(vm->root.bo, false);
+	if (r)
+		return r;
+
+	va_map = amdgpu_vm_bo_lookup_mapping(vm, user_addr);
+	if (!va_map) {
+		r = -EINVAL;
+		goto out_err;
+	}
+	/* Only validate the userq whether resident in the VM mapping range */
+	if (user_addr >= va_map->start  &&
+	    va_map->last - user_addr + 1 >= size) {
+		amdgpu_userq_buffer_va_list_add(queue, va_map, user_addr);
+		amdgpu_bo_unreserve(vm->root.bo);
+		return 0;
+	}
+
+	r = -EINVAL;
+out_err:
+	amdgpu_bo_unreserve(vm->root.bo);
+	return r;
+}
+
+static bool amdgpu_userq_buffer_va_mapped(struct amdgpu_vm *vm, u64 addr)
+{
+	struct amdgpu_bo_va_mapping *mapping;
+	bool r;
+
+	if (amdgpu_bo_reserve(vm->root.bo, false))
+		return false;
+
+	mapping = amdgpu_vm_bo_lookup_mapping(vm, addr);
+	if (!IS_ERR_OR_NULL(mapping) && atomic_read(&mapping->bo_va->userq_va_mapped))
+		r = true;
+	else
+		r = false;
+	amdgpu_bo_unreserve(vm->root.bo);
+
+	return r;
+}
+
+static bool amdgpu_userq_buffer_vas_mapped(struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_userq_va_cursor *va_cursor, *tmp;
+	int r = 0;
+
+	list_for_each_entry_safe(va_cursor, tmp, &queue->userq_va_list, list) {
+		r += amdgpu_userq_buffer_va_mapped(queue->vm, va_cursor->gpu_addr);
+		dev_dbg(queue->userq_mgr->adev->dev,
+			"validate the userq mapping:%p va:%llx r:%d\n",
+			queue, va_cursor->gpu_addr, r);
+	}
+
+	if (r != 0)
+		return true;
+
+	return false;
+}
+
+static void amdgpu_userq_buffer_va_list_del(struct amdgpu_bo_va_mapping *mapping,
+					    struct amdgpu_userq_va_cursor *va_cursor)
+{
+	atomic_set(&mapping->bo_va->userq_va_mapped, 0);
+	list_del(&va_cursor->list);
+	kfree(va_cursor);
+}
+
+static int amdgpu_userq_buffer_vas_list_cleanup(struct amdgpu_device *adev,
+						struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_userq_va_cursor *va_cursor, *tmp;
+	struct amdgpu_bo_va_mapping *mapping;
+	int r;
+
+	r = amdgpu_bo_reserve(queue->vm->root.bo, false);
+	if (r)
+		return r;
+
+	list_for_each_entry_safe(va_cursor, tmp, &queue->userq_va_list, list) {
+		mapping = amdgpu_vm_bo_lookup_mapping(queue->vm, va_cursor->gpu_addr);
+		if (!mapping) {
+			r = -EINVAL;
+			goto err;
+		}
+		dev_dbg(adev->dev, "delete the userq:%p va:%llx\n",
+			queue, va_cursor->gpu_addr);
+		amdgpu_userq_buffer_va_list_del(mapping, va_cursor);
+	}
+err:
+	amdgpu_bo_unreserve(queue->vm->root.bo);
+	return r;
+}
+
+static int
+amdgpu_userq_preempt_helper(struct amdgpu_userq_mgr *uq_mgr,
+			  struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	const struct amdgpu_userq_funcs *userq_funcs =
+		adev->userq_funcs[queue->queue_type];
+	bool found_hung_queue = false;
+	int r = 0;
+
+	if (queue->state == AMDGPU_USERQ_STATE_MAPPED) {
+		r = userq_funcs->preempt(uq_mgr, queue);
+		if (r) {
+			queue->state = AMDGPU_USERQ_STATE_HUNG;
+			found_hung_queue = true;
+		} else {
+			queue->state = AMDGPU_USERQ_STATE_PREEMPTED;
+		}
+	}
+
+	if (found_hung_queue)
+		amdgpu_userq_detect_and_reset_queues(uq_mgr);
+
+	return r;
+}
+
+static int
+amdgpu_userq_restore_helper(struct amdgpu_userq_mgr *uq_mgr,
+			struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	const struct amdgpu_userq_funcs *userq_funcs =
+		adev->userq_funcs[queue->queue_type];
+	int r = 0;
+
+	if (queue->state == AMDGPU_USERQ_STATE_PREEMPTED) {
+		r = userq_funcs->restore(uq_mgr, queue);
+		if (r) {
+			queue->state = AMDGPU_USERQ_STATE_HUNG;
+		} else {
+			queue->state = AMDGPU_USERQ_STATE_MAPPED;
+		}
+	}
+
+	return r;
+}
+
+static int
+amdgpu_userq_unmap_helper(struct amdgpu_userq_mgr *uq_mgr,
+			  struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	const struct amdgpu_userq_funcs *userq_funcs =
+		adev->userq_funcs[queue->queue_type];
+	bool found_hung_queue = false;
+	int r = 0;
+
+	if ((queue->state == AMDGPU_USERQ_STATE_MAPPED) ||
+		(queue->state == AMDGPU_USERQ_STATE_PREEMPTED)) {
+		r = userq_funcs->unmap(uq_mgr, queue);
+		if (r) {
+			queue->state = AMDGPU_USERQ_STATE_HUNG;
+			found_hung_queue = true;
+		} else {
+			queue->state = AMDGPU_USERQ_STATE_UNMAPPED;
+		}
+	}
+
+	if (found_hung_queue)
+		amdgpu_userq_detect_and_reset_queues(uq_mgr);
+
+	return r;
+}
+
+static int
+amdgpu_userq_map_helper(struct amdgpu_userq_mgr *uq_mgr,
+			struct amdgpu_usermode_queue *queue)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	const struct amdgpu_userq_funcs *userq_funcs =
+		adev->userq_funcs[queue->queue_type];
+	int r = 0;
+
+	if (queue->state == AMDGPU_USERQ_STATE_UNMAPPED) {
+		r = userq_funcs->map(uq_mgr, queue);
+		if (r) {
+			queue->state = AMDGPU_USERQ_STATE_HUNG;
+			amdgpu_userq_detect_and_reset_queues(uq_mgr);
+		} else {
+			queue->state = AMDGPU_USERQ_STATE_MAPPED;
+		}
+	}
+
+	return r;
+}
+
+static int
+amdgpu_userq_wait_for_last_fence(struct amdgpu_userq_mgr *uq_mgr,
+				 struct amdgpu_usermode_queue *queue)
+{
+	struct dma_fence *f = queue->last_fence;
+	int ret = 0;
+
+	if (f && !dma_fence_is_signaled(f)) {
+		ret = dma_fence_wait_timeout(f, true, MAX_SCHEDULE_TIMEOUT);
+		if (ret <= 0) {
+			drm_file_err(uq_mgr->file, "Timed out waiting for fence=%llu:%llu\n",
+				     f->context, f->seqno);
+			queue->state = AMDGPU_USERQ_STATE_HUNG;
+			return -ETIME;
+		}
+	}
+
+	return ret;
+}
+
+static void
+amdgpu_userq_cleanup(struct amdgpu_userq_mgr *uq_mgr,
+		     struct amdgpu_usermode_queue *queue,
+		     int queue_id)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	const struct amdgpu_userq_funcs *uq_funcs = adev->userq_funcs[queue->queue_type];
+
+	/* Wait for mode-1 reset to complete */
+	down_read(&adev->reset_domain->sem);
+
+	/* Drop the userq reference. */
+	amdgpu_userq_buffer_vas_list_cleanup(adev, queue);
+	uq_funcs->mqd_destroy(uq_mgr, queue);
+	amdgpu_userq_fence_driver_free(queue);
+	/* Use interrupt-safe locking since IRQ handlers may access these XArrays */
+	xa_erase_irq(&uq_mgr->userq_mgr_xa, (unsigned long)queue_id);
+	xa_erase_irq(&adev->userq_doorbell_xa, queue->doorbell_index);
+	queue->userq_mgr = NULL;
+	list_del(&queue->userq_va_list);
+	kfree(queue);
+
+	up_read(&adev->reset_domain->sem);
+}
+
+static struct amdgpu_usermode_queue *
+amdgpu_userq_find(struct amdgpu_userq_mgr *uq_mgr, int qid)
+{
+	return xa_load(&uq_mgr->userq_mgr_xa, qid);
+}
+
+void
+amdgpu_userq_ensure_ev_fence(struct amdgpu_userq_mgr *uq_mgr,
+			     struct amdgpu_eviction_fence_mgr *evf_mgr)
+{
+	struct amdgpu_eviction_fence *ev_fence;
+
+retry:
+	/* Flush any pending resume work to create ev_fence */
+	flush_delayed_work(&uq_mgr->resume_work);
+
+	mutex_lock(&uq_mgr->userq_mutex);
+	spin_lock(&evf_mgr->ev_fence_lock);
+	ev_fence = evf_mgr->ev_fence;
+	spin_unlock(&evf_mgr->ev_fence_lock);
+	if (!ev_fence || dma_fence_is_signaled(&ev_fence->base)) {
+		mutex_unlock(&uq_mgr->userq_mutex);
+		/*
+		 * Looks like there was no pending resume work,
+		 * add one now to create a valid eviction fence
+		 */
+		schedule_delayed_work(&uq_mgr->resume_work, 0);
+		goto retry;
+	}
+}
+
+int amdgpu_userq_create_object(struct amdgpu_userq_mgr *uq_mgr,
+			       struct amdgpu_userq_obj *userq_obj,
+			       int size)
+{
+	struct amdgpu_device *adev = uq_mgr->adev;
+	struct amdgpu_bo_param bp;
+	int r;
+
+	memset(&bp, 0, sizeof(bp));
+	bp.byte_align = PAGE_SIZE;
+	bp.domain = AMDGPU_GEM_DOMAIN_GTT;
+	bp.flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS |
+		   AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
+	bp.type = ttm_bo_type_kernel;
+	bp.size = size;
+	bp.resv = NULL;
+	bp.bo_ptr_size = sizeof(struct amdgpu_bo);
+
+	r = amdgpu_bo_create(adev, &bp, &userq_obj->obj);
+	if (r) {
+		drm_file_err(uq_mgr->file, "Failed to allocate BO for userqueue (%d)", r);
+		return r;
+	}
+
+	r = amdgpu_bo_reserve(userq_obj->obj, true);
+	if (r) {
+		drm_file_err(uq_mgr->file, "Failed to reserve BO to map (%d)", r);
+		goto free_obj;
+	}
+
+	r = amdgpu_ttm_alloc_gart(&(userq_obj->obj)->tbo);
+	if (r) {
+		drm_file_err(uq_mgr->file, "Failed to alloc GART for userqueue object (%d)", r);
+		goto unresv;
+	}
+
+	r = amdgpu_bo_kmap(userq_obj->obj, &userq_obj->cpu_ptr);
+	if (r) {
+		drm_file_err(uq_mgr->file, "Failed to map BO for userqueue (%d)", r);
+		goto unresv;
+	}
+
+	userq_obj->gpu_addr = amdgpu_bo_gpu_offset(userq_obj->obj);
+	amdgpu_bo_unreserve(userq_obj->obj);
+	memset(userq_obj->cpu_ptr, 0, size);
+	return 0;
+
+unresv:
+	amdgpu_bo_unreserve(userq_obj->obj);
+
+free_obj:
+	amdgpu_bo_unref(&userq_obj->obj);
+	return r;
+}
+
+void amdgpu_userq_destroy_object(struct amdgpu_userq_mgr *uq_mgr,
+				 struct amdgpu_userq_obj *userq_obj)
+{
+	amdgpu_bo_kunmap(userq_obj->obj);
+	amdgpu_bo_unref(&userq_obj->obj);
+}
+
+uint64_t
+amdgpu_userq_get_doorbell_index(struct amdgpu_userq_mgr *uq_mgr,
+				struct amdgpu_db_info *db_info,
+				struct drm_file *filp)
+{
+	uint64_t index;
+	struct drm_gem_object *gobj;
+	struct amdgpu_userq_obj *db_obj = db_info->db_obj;
+	int r, db_size;
+
+	gobj = drm_gem_object_lookup(filp, db_info->doorbell_handle);
+	if (gobj == NULL) {
+		drm_file_err(uq_mgr->file, "Can't find GEM object for doorbell\n");
+		return -EINVAL;
+	}
+
+	db_obj->obj = amdgpu_bo_ref(gem_to_amdgpu_bo(gobj));
+	drm_gem_object_put(gobj);
+
+	r = amdgpu_bo_reserve(db_obj->obj, true);
+	if (r) {
+		drm_file_err(uq_mgr->file, "[Usermode queues] Failed to pin doorbell object\n");
+		goto unref_bo;
+	}
+
+	/* Pin the BO before generating the index, unpin in queue destroy */
+	r = amdgpu_bo_pin(db_obj->obj, AMDGPU_GEM_DOMAIN_DOORBELL);
+	if (r) {
+		drm_file_err(uq_mgr->file, "[Usermode queues] Failed to pin doorbell object\n");
+		goto unresv_bo;
+	}
+
+	switch (db_info->queue_type) {
+	case AMDGPU_HW_IP_GFX:
+	case AMDGPU_HW_IP_COMPUTE:
+	case AMDGPU_HW_IP_DMA:
+		db_size = sizeof(u64);
+		break;
+	default:
+		drm_file_err(uq_mgr->file, "[Usermode queues] IP %d not support\n",
+			     db_info->queue_type);
+		r = -EINVAL;
+		goto unpin_bo;
+	}
+
+	index = amdgpu_doorbell_index_on_bar(uq_mgr->adev, db_obj->obj,
+					     db_info->doorbell_offset, db_size);
+	drm_dbg_driver(adev_to_drm(uq_mgr->adev),
+		       "[Usermode queues] doorbell index=%lld\n", index);
+	amdgpu_bo_unreserve(db_obj->obj);
+	return index;
+
+unpin_bo:
+	amdgpu_bo_unpin(db_obj->obj);
+unresv_bo:
+	amdgpu_bo_unreserve(db_obj->obj);
+unref_bo:
+	amdgpu_bo_unref(&db_obj->obj);
+	return r;
+}
+
+static int
+amdgpu_userq_destroy(struct drm_file *filp, int queue_id)
+{
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
+	struct amdgpu_userq_mgr *uq_mgr = &fpriv->userq_mgr;
+	struct amdgpu_device *adev = uq_mgr->adev;
+	struct amdgpu_usermode_queue *queue;
+	int r = 0;
+
+	cancel_delayed_work_sync(&uq_mgr->resume_work);
+	mutex_lock(&uq_mgr->userq_mutex);
+
+	queue = amdgpu_userq_find(uq_mgr, queue_id);
+	if (!queue) {
+		drm_dbg_driver(adev_to_drm(uq_mgr->adev), "Invalid queue id to destroy\n");
+		mutex_unlock(&uq_mgr->userq_mutex);
+		return -EINVAL;
+	}
+	amdgpu_userq_wait_for_last_fence(uq_mgr, queue);
+	r = amdgpu_bo_reserve(queue->db_obj.obj, true);
+	if (!r) {
+		amdgpu_bo_unpin(queue->db_obj.obj);
+		amdgpu_bo_unreserve(queue->db_obj.obj);
+	}
+	amdgpu_bo_unref(&queue->db_obj.obj);
+	atomic_dec(&uq_mgr->userq_count[queue->queue_type]);
+#if defined(CONFIG_DEBUG_FS)
+	debugfs_remove_recursive(queue->debugfs_queue);
+#endif
+	amdgpu_userq_detect_and_reset_queues(uq_mgr);
+	r = amdgpu_userq_unmap_helper(uq_mgr, queue);
+	/*TODO: It requires a reset for userq hw unmap error*/
+	if (unlikely(r != AMDGPU_USERQ_STATE_UNMAPPED)) {
+		drm_warn(adev_to_drm(uq_mgr->adev), "trying to destroy a HW mapping userq\n");
+		queue->state = AMDGPU_USERQ_STATE_HUNG;
+	}
+	amdgpu_userq_cleanup(uq_mgr, queue, queue_id);
+	mutex_unlock(&uq_mgr->userq_mutex);
+
+	pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
+
+	return r;
+}
+
+static int amdgpu_userq_priority_permit(struct drm_file *filp,
+					int priority)
+{
+	if (priority < AMDGPU_USERQ_CREATE_FLAGS_QUEUE_PRIORITY_HIGH)
+		return 0;
+
+	if (capable(CAP_SYS_NICE))
+		return 0;
+
+	if (drm_is_current_master(filp))
+		return 0;
+
+	return -EACCES;
+}
+
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_mqd_info_read(struct seq_file *m, void *unused)
+{
+	struct amdgpu_usermode_queue *queue = m->private;
+	struct amdgpu_bo *bo;
+	int r;
+
+	if (!queue || !queue->mqd.obj)
+		return -EINVAL;
+
+	bo = amdgpu_bo_ref(queue->mqd.obj);
+	r = amdgpu_bo_reserve(bo, true);
+	if (r) {
+		amdgpu_bo_unref(&bo);
+		return -EINVAL;
+	}
+
+	seq_printf(m, "queue_type: %d\n", queue->queue_type);
+	seq_printf(m, "mqd_gpu_address: 0x%llx\n", amdgpu_bo_gpu_offset(queue->mqd.obj));
+
+	amdgpu_bo_unreserve(bo);
+	amdgpu_bo_unref(&bo);
+
+	return 0;
+}
+
+static int amdgpu_mqd_info_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, amdgpu_mqd_info_read, inode->i_private);
+}
+
+static const struct file_operations amdgpu_mqd_info_fops = {
+	.owner = THIS_MODULE,
+	.open = amdgpu_mqd_info_open,
+	.read = seq_read,
+	.llseek = seq_lseek,
+	.release = single_release,
+};
+#endif
+
+static int
+amdgpu_userq_create(struct drm_file *filp, union drm_amdgpu_userq *args)
+{
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
+	struct amdgpu_userq_mgr *uq_mgr = &fpriv->userq_mgr;
+	struct amdgpu_device *adev = uq_mgr->adev;
+	const struct amdgpu_userq_funcs *uq_funcs;
+	struct amdgpu_usermode_queue *queue;
+	struct amdgpu_db_info db_info;
+	char *queue_name;
+	bool skip_map_queue;
+	u32 qid;
+	uint64_t index;
+	int r = 0;
+	int priority =
+		(args->in.flags & AMDGPU_USERQ_CREATE_FLAGS_QUEUE_PRIORITY_MASK) >>
+		AMDGPU_USERQ_CREATE_FLAGS_QUEUE_PRIORITY_SHIFT;
+
+	r = amdgpu_userq_priority_permit(filp, priority);
+	if (r)
+		return r;
+
+	r = pm_runtime_get_sync(adev_to_drm(adev)->dev);
+	if (r < 0) {
+		drm_file_err(uq_mgr->file, "pm_runtime_get_sync() failed for userqueue create\n");
+		pm_runtime_put_autosuspend(adev_to_drm(adev)->dev);
+		return r;
+	}
+
+	/*
+	 * There could be a situation that we are creating a new queue while
+	 * the other queues under this UQ_mgr are suspended. So if there is any
+	 * resume work pending, wait for it to get done.
+	 *
+	 * This will also make sure we have a valid eviction fence ready to be used.
+	 */
+	amdgpu_userq_ensure_ev_fence(&fpriv->userq_mgr, &fpriv->evf_mgr);
+
+	uq_funcs = adev->userq_funcs[args->in.ip_type];
+	if (!uq_funcs) {
+		drm_file_err(uq_mgr->file, "Usermode queue is not supported for this IP (%u)\n",
+			     args->in.ip_type);
+		r = -EINVAL;
+		goto unlock;
+	}
+
+	queue = kzalloc(sizeof(struct amdgpu_usermode_queue), GFP_KERNEL);
+	if (!queue) {
+		drm_file_err(uq_mgr->file, "Failed to allocate memory for queue\n");
+		r = -ENOMEM;
+		goto unlock;
+	}
+
+	INIT_LIST_HEAD(&queue->userq_va_list);
+	queue->doorbell_handle = args->in.doorbell_handle;
+	queue->queue_type = args->in.ip_type;
+	queue->vm = &fpriv->vm;
+	queue->priority = priority;
+
+	db_info.queue_type = queue->queue_type;
+	db_info.doorbell_handle = queue->doorbell_handle;
+	db_info.db_obj = &queue->db_obj;
+	db_info.doorbell_offset = args->in.doorbell_offset;
+
+	/* Validate the userq virtual address.*/
+	if (amdgpu_userq_input_va_validate(queue, args->in.queue_va, args->in.queue_size) ||
+	    amdgpu_userq_input_va_validate(queue, args->in.rptr_va, AMDGPU_GPU_PAGE_SIZE) ||
+	    amdgpu_userq_input_va_validate(queue, args->in.wptr_va, AMDGPU_GPU_PAGE_SIZE)) {
+		r = -EINVAL;
+		kfree(queue);
+		goto unlock;
+	}
+
+	/* Convert relative doorbell offset into absolute doorbell index */
+	index = amdgpu_userq_get_doorbell_index(uq_mgr, &db_info, filp);
+	if (index == (uint64_t)-EINVAL) {
+		drm_file_err(uq_mgr->file, "Failed to get doorbell for queue\n");
+		kfree(queue);
+		r = -EINVAL;
+		goto unlock;
+	}
+
+	queue->doorbell_index = index;
+	xa_init_flags(&queue->fence_drv_xa, XA_FLAGS_ALLOC);
+	r = amdgpu_userq_fence_driver_alloc(adev, queue);
+	if (r) {
+		drm_file_err(uq_mgr->file, "Failed to alloc fence driver\n");
+		goto unlock;
+	}
+
+	r = uq_funcs->mqd_create(uq_mgr, &args->in, queue);
+	if (r) {
+		drm_file_err(uq_mgr->file, "Failed to create Queue\n");
+		amdgpu_userq_fence_driver_free(queue);
+		kfree(queue);
+		goto unlock;
+	}
+
+	/* Wait for mode-1 reset to complete */
+	down_read(&adev->reset_domain->sem);
+	r = xa_err(xa_store_irq(&adev->userq_doorbell_xa, index, queue, GFP_KERNEL));
+	if (r) {
+		kfree(queue);
+		up_read(&adev->reset_domain->sem);
+		goto unlock;
+	}
+
+	r = xa_alloc(&uq_mgr->userq_mgr_xa, &qid, queue, XA_LIMIT(1, AMDGPU_MAX_USERQ_COUNT), GFP_KERNEL);
+	if (r) {
+		drm_file_err(uq_mgr->file, "Failed to allocate a queue id\n");
+		amdgpu_userq_fence_driver_free(queue);
+		uq_funcs->mqd_destroy(uq_mgr, queue);
+		kfree(queue);
+		r = -ENOMEM;
+		up_read(&adev->reset_domain->sem);
+		goto unlock;
+	}
+	up_read(&adev->reset_domain->sem);
+	queue->userq_mgr = uq_mgr;
+
+	/* don't map the queue if scheduling is halted */
+	if (adev->userq_halt_for_enforce_isolation &&
+	    ((queue->queue_type == AMDGPU_HW_IP_GFX) ||
+	     (queue->queue_type == AMDGPU_HW_IP_COMPUTE)))
+		skip_map_queue = true;
+	else
+		skip_map_queue = false;
+	if (!skip_map_queue) {
+		r = amdgpu_userq_map_helper(uq_mgr, queue);
+		if (r) {
+			drm_file_err(uq_mgr->file, "Failed to map Queue\n");
+			xa_erase(&uq_mgr->userq_mgr_xa, qid);
+			amdgpu_userq_fence_driver_free(queue);
+			uq_funcs->mqd_destroy(uq_mgr, queue);
+			kfree(queue);
+			goto unlock;
+		}
+	}
+
+	queue_name = kasprintf(GFP_KERNEL, "queue-%d", qid);
+	if (!queue_name) {
+		r = -ENOMEM;
+		goto unlock;
+	}
+
+#if defined(CONFIG_DEBUG_FS)
+	/* Queue dentry per client to hold MQD information   */
+	queue->debugfs_queue = debugfs_create_dir(queue_name, filp->debugfs_client);
+	debugfs_create_file("mqd_info", 0444, queue->debugfs_queue, queue, &amdgpu_mqd_info_fops);
+#endif
+	kfree(queue_name);
+
+	args->out.queue_id = qid;
+	atomic_inc(&uq_mgr->userq_count[queue->queue_type]);
+
+unlock:
+	mutex_unlock(&uq_mgr->userq_mutex);
+
+	return r;
+}
+
+static int amdgpu_userq_input_args_validate(struct drm_device *dev,
+					union drm_amdgpu_userq *args,
+					struct drm_file *filp)
+{
+	struct amdgpu_device *adev = drm_to_adev(dev);
+
+	switch (args->in.op) {
+	case AMDGPU_USERQ_OP_CREATE:
+		if (args->in.flags & ~(AMDGPU_USERQ_CREATE_FLAGS_QUEUE_PRIORITY_MASK |
+				       AMDGPU_USERQ_CREATE_FLAGS_QUEUE_SECURE))
+			return -EINVAL;
+		/* Usermode queues are only supported for GFX IP as of now */
+		if (args->in.ip_type != AMDGPU_HW_IP_GFX &&
+		    args->in.ip_type != AMDGPU_HW_IP_DMA &&
+		    args->in.ip_type != AMDGPU_HW_IP_COMPUTE) {
+			drm_file_err(filp, "Usermode queue doesn't support IP type %u\n",
+				     args->in.ip_type);
+			return -EINVAL;
+		}
+
+		if ((args->in.flags & AMDGPU_USERQ_CREATE_FLAGS_QUEUE_SECURE) &&
+		    (args->in.ip_type != AMDGPU_HW_IP_GFX) &&
+		    (args->in.ip_type != AMDGPU_HW_IP_COMPUTE) &&
+		    !amdgpu_is_tmz(adev)) {
+			drm_file_err(filp, "Secure only supported on GFX/Compute queues\n");
+			return -EINVAL;
+		}
+
+		if (args->in.queue_va == AMDGPU_BO_INVALID_OFFSET ||
+		    args->in.queue_va == 0 ||
+		    args->in.queue_size == 0) {
+			drm_file_err(filp, "invalidate userq queue va or size\n");
+			return -EINVAL;
+		}
+		if (!args->in.wptr_va || !args->in.rptr_va) {
+			drm_file_err(filp, "invalidate userq queue rptr or wptr\n");
+			return -EINVAL;
+		}
+		break;
+	case AMDGPU_USERQ_OP_FREE:
+		if (args->in.ip_type ||
+		    args->in.doorbell_handle ||
+		    args->in.doorbell_offset ||
+		    args->in.flags ||
+		    args->in.queue_va ||
+		    args->in.queue_size ||
+		    args->in.rptr_va ||
+		    args->in.wptr_va ||
+		    args->in.mqd ||
+		    args->in.mqd_size)
+			return -EINVAL;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+int amdgpu_userq_ioctl(struct drm_device *dev, void *data,
+		       struct drm_file *filp)
+{
+	union drm_amdgpu_userq *args = data;
+	int r;
+
+	if (amdgpu_userq_input_args_validate(dev, args, filp) < 0)
+		return -EINVAL;
+
+	switch (args->in.op) {
+	case AMDGPU_USERQ_OP_CREATE:
+		r = amdgpu_userq_create(filp, args);
+		if (r)
+			drm_file_err(filp, "Failed to create usermode queue\n");
+		break;
+
+	case AMDGPU_USERQ_OP_FREE:
+		r = amdgpu_userq_destroy(filp, args->in.queue_id);
+		if (r)
+			drm_file_err(filp, "Failed to destroy usermode queue\n");
+		break;
+
+	default:
+		drm_dbg_driver(dev, "Invalid user queue op specified: %d\n", args->in.op);
+		return -EINVAL;
+	}
+
+	return r;
+}
+
+static int
+amdgpu_userq_restore_all(struct amdgpu_userq_mgr *uq_mgr)
+{
+	struct amdgpu_usermode_queue *queue;
+	unsigned long queue_id;
+	int ret = 0, r;
+
+	/* Resume all the queues for this process */
+	xa_for_each(&uq_mgr->userq_mgr_xa, queue_id, queue) {
+
+		if (!amdgpu_userq_buffer_vas_mapped(queue)) {
+			drm_file_err(uq_mgr->file,
+				     "trying restore queue without va mapping\n");
+			queue->state = AMDGPU_USERQ_STATE_INVALID_VA;
+			continue;
+		}
+
+		r = amdgpu_userq_restore_helper(uq_mgr, queue);
+		if (r)
+			ret = r;
+	}
+
+	if (ret)
+		drm_file_err(uq_mgr->file, "Failed to map all the queues\n");
+	return ret;
+}
+
+static int amdgpu_userq_validate_vm(void *param, struct amdgpu_bo *bo)
+{
+	struct ttm_operation_ctx ctx = { false, false };
+
+	amdgpu_bo_placement_from_domain(bo, bo->allowed_domains);
+	return ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+}
+
+/* Handle all BOs on the invalidated list, validate them and update the PTs */
+static int
+amdgpu_userq_bo_validate(struct amdgpu_device *adev, struct drm_exec *exec,
+			 struct amdgpu_vm *vm)
+{
+	struct ttm_operation_ctx ctx = { false, false };
+	struct amdgpu_bo_va *bo_va;
+	struct amdgpu_bo *bo;
+	int ret;
+
+	spin_lock(&vm->status_lock);
+	while (!list_empty(&vm->invalidated)) {
+		bo_va = list_first_entry(&vm->invalidated,
+					 struct amdgpu_bo_va,
+					 base.vm_status);
+		spin_unlock(&vm->status_lock);
+
+		bo = bo_va->base.bo;
+		ret = drm_exec_prepare_obj(exec, &bo->tbo.base, 2);
+		if (unlikely(ret))
+			return ret;
+
+		amdgpu_bo_placement_from_domain(bo, bo->allowed_domains);
+		ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+		if (ret)
+			return ret;
+
+		/* This moves the bo_va to the done list */
+		ret = amdgpu_vm_bo_update(adev, bo_va, false);
+		if (ret)
+			return ret;
+
+		spin_lock(&vm->status_lock);
+	}
+	spin_unlock(&vm->status_lock);
+
+	return 0;
+}
+
+/* Make sure the whole VM is ready to be used */
+static int
+amdgpu_userq_vm_validate(struct amdgpu_userq_mgr *uq_mgr)
+{
+	struct amdgpu_fpriv *fpriv = uq_mgr_to_fpriv(uq_mgr);
+	bool invalidated = false, new_addition = false;
+	struct ttm_operation_ctx ctx = { true, false };
+	struct amdgpu_device *adev = uq_mgr->adev;
+	struct amdgpu_hmm_range *range;
+	struct amdgpu_vm *vm = &fpriv->vm;
+	unsigned long key, tmp_key;
+	struct amdgpu_bo_va *bo_va;
+	struct amdgpu_bo *bo;
+	struct drm_exec exec;
+	struct xarray xa;
+	int ret;
+
+	xa_init(&xa);
+
+retry_lock:
+	drm_exec_init(&exec, DRM_EXEC_IGNORE_DUPLICATES, 0);
+	drm_exec_until_all_locked(&exec) {
+		ret = amdgpu_vm_lock_pd(vm, &exec, 1);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(ret))
+			goto unlock_all;
+
+		ret = amdgpu_vm_lock_done_list(vm, &exec, 1);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(ret))
+			goto unlock_all;
+
+		/* This validates PDs, PTs and per VM BOs */
+		ret = amdgpu_vm_validate(adev, vm, NULL,
+					 amdgpu_userq_validate_vm,
+					 NULL);
+		if (unlikely(ret))
+			goto unlock_all;
+
+		/* This locks and validates the remaining evicted BOs */
+		ret = amdgpu_userq_bo_validate(adev, &exec, vm);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(ret))
+			goto unlock_all;
+	}
+
+	if (invalidated) {
+		xa_for_each(&xa, tmp_key, range) {
+			bo = range->bo;
+			amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU);
+			ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+			if (ret)
+				goto unlock_all;
+
+			amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, range);
+
+			amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT);
+			ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
+			if (ret)
+				goto unlock_all;
+		}
+		invalidated = false;
+	}
+
+	ret = amdgpu_vm_handle_moved(adev, vm, NULL);
+	if (ret)
+		goto unlock_all;
+
+	key = 0;
+	/* Validate User Ptr BOs */
+	list_for_each_entry(bo_va, &vm->done, base.vm_status) {
+		bo = bo_va->base.bo;
+		if (!bo)
+			continue;
+
+		if (!amdgpu_ttm_tt_is_userptr(bo->tbo.ttm))
+			continue;
+
+		range = xa_load(&xa, key);
+		if (range && range->bo != bo) {
+			xa_erase(&xa, key);
+			amdgpu_hmm_range_free(range);
+			range = NULL;
+		}
+
+		if (!range) {
+			range = amdgpu_hmm_range_alloc(bo);
+			if (!range) {
+				ret = -ENOMEM;
+				goto unlock_all;
+			}
+
+			xa_store(&xa, key, range, GFP_KERNEL);
+			new_addition = true;
+		}
+		key++;
+	}
+
+	if (new_addition) {
+		drm_exec_fini(&exec);
+		xa_for_each(&xa, tmp_key, range) {
+			if (!range)
+				continue;
+			bo = range->bo;
+			ret = amdgpu_ttm_tt_get_user_pages(bo, range);
+			if (ret)
+				goto unlock_all;
+		}
+
+		invalidated = true;
+		new_addition = false;
+		goto retry_lock;
+	}
+
+	ret = amdgpu_vm_update_pdes(adev, vm, false);
+	if (ret)
+		goto unlock_all;
+
+	/*
+	 * We need to wait for all VM updates to finish before restarting the
+	 * queues. Using the done list like that is now ok since everything is
+	 * locked in place.
+	 */
+	list_for_each_entry(bo_va, &vm->done, base.vm_status)
+		dma_fence_wait(bo_va->last_pt_update, false);
+	dma_fence_wait(vm->last_update, false);
+
+	ret = amdgpu_eviction_fence_replace_fence(&fpriv->evf_mgr, &exec);
+	if (ret)
+		drm_file_err(uq_mgr->file, "Failed to replace eviction fence\n");
+
+unlock_all:
+	drm_exec_fini(&exec);
+	xa_for_each(&xa, tmp_key, range) {
+		if (!range)
+			continue;
+		bo = range->bo;
+		amdgpu_hmm_range_free(range);
+	}
+	xa_destroy(&xa);
+	return ret;
+}
+
+static void amdgpu_userq_restore_worker(struct work_struct *work)
+{
+	struct amdgpu_userq_mgr *uq_mgr = work_to_uq_mgr(work, resume_work.work);
+	struct amdgpu_fpriv *fpriv = uq_mgr_to_fpriv(uq_mgr);
+	int ret;
+
+	flush_delayed_work(&fpriv->evf_mgr.suspend_work);
+
+	mutex_lock(&uq_mgr->userq_mutex);
+
+	ret = amdgpu_userq_vm_validate(uq_mgr);
+	if (ret) {
+		drm_file_err(uq_mgr->file, "Failed to validate BOs to restore\n");
+		goto unlock;
+	}
+
+	ret = amdgpu_userq_restore_all(uq_mgr);
+	if (ret) {
+		drm_file_err(uq_mgr->file, "Failed to restore all queues\n");
+		goto unlock;
+	}
+
+unlock:
+	mutex_unlock(&uq_mgr->userq_mutex);
+}
+
+static int
+amdgpu_userq_evict_all(struct amdgpu_userq_mgr *uq_mgr)
+{
+	struct amdgpu_usermode_queue *queue;
+	unsigned long queue_id;
+	int ret = 0, r;
+
+	amdgpu_userq_detect_and_reset_queues(uq_mgr);
+	/* Try to unmap all the queues in this process ctx */
+	xa_for_each(&uq_mgr->userq_mgr_xa, queue_id, queue) {
+		r = amdgpu_userq_preempt_helper(uq_mgr, queue);
+		if (r)
+			ret = r;
+	}
+
+	if (ret)
+		drm_file_err(uq_mgr->file, "Couldn't unmap all the queues\n");
+	return ret;
+}
+
+void amdgpu_userq_reset_work(struct work_struct *work)
+{
+	struct amdgpu_device *adev = container_of(work, struct amdgpu_device,
+						  userq_reset_work);
+	struct amdgpu_reset_context reset_context;
+
+	memset(&reset_context, 0, sizeof(reset_context));
+
+	reset_context.method = AMD_RESET_METHOD_NONE;
+	reset_context.reset_req_dev = adev;
+	reset_context.src = AMDGPU_RESET_SRC_USERQ;
+	set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
+	/*set_bit(AMDGPU_SKIP_COREDUMP, &reset_context.flags);*/
+
+	amdgpu_device_gpu_recover(adev, NULL, &reset_context);
+}
+
+static int
+amdgpu_userq_wait_for_signal(struct amdgpu_userq_mgr *uq_mgr)
+{
+	struct amdgpu_usermode_queue *queue;
+	unsigned long queue_id;
+	int ret;
+
+	xa_for_each(&uq_mgr->userq_mgr_xa, queue_id, queue) {
+		struct dma_fence *f = queue->last_fence;
+
+		if (!f || dma_fence_is_signaled(f))
+			continue;
+		ret = dma_fence_wait_timeout(f, true, msecs_to_jiffies(100));
+		if (ret <= 0) {
+			drm_file_err(uq_mgr->file, "Timed out waiting for fence=%llu:%llu\n",
+				     f->context, f->seqno);
+			return -ETIMEDOUT;
+		}
+	}
+
+	return 0;
+}
+
+void
+amdgpu_userq_evict(struct amdgpu_userq_mgr *uq_mgr,
+		   struct amdgpu_eviction_fence *ev_fence)
+{
+	struct amdgpu_fpriv *fpriv = uq_mgr_to_fpriv(uq_mgr);
+	struct amdgpu_eviction_fence_mgr *evf_mgr = &fpriv->evf_mgr;
+	struct amdgpu_device *adev = uq_mgr->adev;
+	int ret;
+
+	/* Wait for any pending userqueue fence work to finish */
+	ret = amdgpu_userq_wait_for_signal(uq_mgr);
+	if (ret)
+		dev_err(adev->dev, "Not evicting userqueue, timeout waiting for work\n");
+
+	ret = amdgpu_userq_evict_all(uq_mgr);
+	if (ret)
+		dev_err(adev->dev, "Failed to evict userqueue\n");
+
+	/* Signal current eviction fence */
+	amdgpu_eviction_fence_signal(evf_mgr, ev_fence);
+
+	if (evf_mgr->fd_closing) {
+		cancel_delayed_work_sync(&uq_mgr->resume_work);
+		return;
+	}
+
+	/* Schedule a resume work */
+	schedule_delayed_work(&uq_mgr->resume_work, 0);
+}
+
+int amdgpu_userq_mgr_init(struct amdgpu_userq_mgr *userq_mgr, struct drm_file *file_priv,
+			  struct amdgpu_device *adev)
+{
+	mutex_init(&userq_mgr->userq_mutex);
+	xa_init_flags(&userq_mgr->userq_mgr_xa, XA_FLAGS_ALLOC);
+	userq_mgr->adev = adev;
+	userq_mgr->file = file_priv;
+
+	INIT_DELAYED_WORK(&userq_mgr->resume_work, amdgpu_userq_restore_worker);
+	return 0;
+}
+
+void amdgpu_userq_mgr_fini(struct amdgpu_userq_mgr *userq_mgr)
+{
+	struct amdgpu_usermode_queue *queue;
+	unsigned long queue_id;
+
+	cancel_delayed_work_sync(&userq_mgr->resume_work);
+
+	mutex_lock(&userq_mgr->userq_mutex);
+	amdgpu_userq_detect_and_reset_queues(userq_mgr);
+	xa_for_each(&userq_mgr->userq_mgr_xa, queue_id, queue) {
+		amdgpu_userq_wait_for_last_fence(userq_mgr, queue);
+		amdgpu_userq_unmap_helper(userq_mgr, queue);
+		amdgpu_userq_cleanup(userq_mgr, queue, queue_id);
+	}
+
+	xa_destroy(&userq_mgr->userq_mgr_xa);
+	mutex_unlock(&userq_mgr->userq_mutex);
+	mutex_destroy(&userq_mgr->userq_mutex);
+}
+
+int amdgpu_userq_suspend(struct amdgpu_device *adev)
+{
+	u32 ip_mask = amdgpu_userq_get_supported_ip_mask(adev);
+	struct amdgpu_usermode_queue *queue;
+	struct amdgpu_userq_mgr *uqm;
+	unsigned long queue_id;
+	int r;
+
+	if (!ip_mask)
+		return 0;
+
+	xa_for_each(&adev->userq_doorbell_xa, queue_id, queue) {
+		uqm = queue->userq_mgr;
+		cancel_delayed_work_sync(&uqm->resume_work);
+		guard(mutex)(&uqm->userq_mutex);
+		amdgpu_userq_detect_and_reset_queues(uqm);
+		if (adev->in_s0ix)
+			r = amdgpu_userq_preempt_helper(uqm, queue);
+		else
+			r = amdgpu_userq_unmap_helper(uqm, queue);
+		if (r)
+			return r;
+	}
+	return 0;
+}
+
+int amdgpu_userq_resume(struct amdgpu_device *adev)
+{
+	u32 ip_mask = amdgpu_userq_get_supported_ip_mask(adev);
+	struct amdgpu_usermode_queue *queue;
+	struct amdgpu_userq_mgr *uqm;
+	unsigned long queue_id;
+	int r;
+
+	if (!ip_mask)
+		return 0;
+
+	xa_for_each(&adev->userq_doorbell_xa, queue_id, queue) {
+		uqm = queue->userq_mgr;
+		guard(mutex)(&uqm->userq_mutex);
+		if (adev->in_s0ix)
+			r = amdgpu_userq_restore_helper(uqm, queue);
+		else
+			r = amdgpu_userq_map_helper(uqm, queue);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+int amdgpu_userq_stop_sched_for_enforce_isolation(struct amdgpu_device *adev,
+						  u32 idx)
+{
+	u32 ip_mask = amdgpu_userq_get_supported_ip_mask(adev);
+	struct amdgpu_usermode_queue *queue;
+	struct amdgpu_userq_mgr *uqm;
+	unsigned long queue_id;
+	int ret = 0, r;
+
+	/* only need to stop gfx/compute */
+	if (!(ip_mask & ((1 << AMDGPU_HW_IP_GFX) | (1 << AMDGPU_HW_IP_COMPUTE))))
+		return 0;
+
+	if (adev->userq_halt_for_enforce_isolation)
+		dev_warn(adev->dev, "userq scheduling already stopped!\n");
+	adev->userq_halt_for_enforce_isolation = true;
+	xa_for_each(&adev->userq_doorbell_xa, queue_id, queue) {
+		uqm = queue->userq_mgr;
+		cancel_delayed_work_sync(&uqm->resume_work);
+		mutex_lock(&uqm->userq_mutex);
+		if (((queue->queue_type == AMDGPU_HW_IP_GFX) ||
+		     (queue->queue_type == AMDGPU_HW_IP_COMPUTE)) &&
+		    (queue->xcp_id == idx)) {
+			amdgpu_userq_detect_and_reset_queues(uqm);
+			r = amdgpu_userq_preempt_helper(uqm, queue);
+			if (r)
+				ret = r;
+		}
+		mutex_unlock(&uqm->userq_mutex);
+	}
+
+	return ret;
+}
+
+int amdgpu_userq_start_sched_for_enforce_isolation(struct amdgpu_device *adev,
+						   u32 idx)
+{
+	u32 ip_mask = amdgpu_userq_get_supported_ip_mask(adev);
+	struct amdgpu_usermode_queue *queue;
+	struct amdgpu_userq_mgr *uqm;
+	unsigned long queue_id;
+	int ret = 0, r;
+
+	/* only need to stop gfx/compute */
+	if (!(ip_mask & ((1 << AMDGPU_HW_IP_GFX) | (1 << AMDGPU_HW_IP_COMPUTE))))
+		return 0;
+
+	if (!adev->userq_halt_for_enforce_isolation)
+		dev_warn(adev->dev, "userq scheduling already started!\n");
+	adev->userq_halt_for_enforce_isolation = false;
+	xa_for_each(&adev->userq_doorbell_xa, queue_id, queue) {
+		uqm = queue->userq_mgr;
+		mutex_lock(&uqm->userq_mutex);
+			if (((queue->queue_type == AMDGPU_HW_IP_GFX) ||
+			     (queue->queue_type == AMDGPU_HW_IP_COMPUTE)) &&
+			    (queue->xcp_id == idx)) {
+				r = amdgpu_userq_restore_helper(uqm, queue);
+				if (r)
+					ret = r;
+			}
+		mutex_unlock(&uqm->userq_mutex);
+	}
+
+	return ret;
+}
+
+int amdgpu_userq_gem_va_unmap_validate(struct amdgpu_device *adev,
+				       struct amdgpu_bo_va_mapping *mapping,
+				       uint64_t saddr)
+{
+	u32 ip_mask = amdgpu_userq_get_supported_ip_mask(adev);
+	struct amdgpu_bo_va *bo_va = mapping->bo_va;
+	struct dma_resv *resv = bo_va->base.bo->tbo.base.resv;
+	int ret = 0;
+
+	if (!ip_mask)
+		return 0;
+
+	dev_warn_once(adev->dev, "now unmapping a vital queue va:%llx\n", saddr);
+	/**
+	 * The userq VA mapping reservation should include the eviction fence,
+	 * if the eviction fence can't signal successfully during unmapping,
+	 * then driver will warn to flag this improper unmap of the userq VA.
+	 * Note: The eviction fence may be attached to different BOs, and this
+	 * unmap is only for one kind of userq VAs, so at this point suppose
+	 * the eviction fence is always unsignaled.
+	 */
+	if (!dma_resv_test_signaled(resv, DMA_RESV_USAGE_BOOKKEEP)) {
+		ret = dma_resv_wait_timeout(resv, DMA_RESV_USAGE_BOOKKEEP, true,
+					    MAX_SCHEDULE_TIMEOUT);
+		if (ret <= 0)
+			return -EBUSY;
+	}
+
+	return 0;
+}
+
+void amdgpu_userq_pre_reset(struct amdgpu_device *adev)
+{
+	const struct amdgpu_userq_funcs *userq_funcs;
+	struct amdgpu_usermode_queue *queue;
+	struct amdgpu_userq_mgr *uqm;
+	unsigned long queue_id;
+
+	xa_for_each(&adev->userq_doorbell_xa, queue_id, queue) {
+		uqm = queue->userq_mgr;
+		cancel_delayed_work_sync(&uqm->resume_work);
+		if (queue->state == AMDGPU_USERQ_STATE_MAPPED) {
+			amdgpu_userq_wait_for_last_fence(uqm, queue);
+			userq_funcs = adev->userq_funcs[queue->queue_type];
+			userq_funcs->unmap(uqm, queue);
+			/* just mark all queues as hung at this point.
+			 * if unmap succeeds, we could map again
+			 * in amdgpu_userq_post_reset() if vram is not lost
+			 */
+			queue->state = AMDGPU_USERQ_STATE_HUNG;
+			amdgpu_userq_fence_driver_force_completion(queue);
+		}
+	}
+}
+
+int amdgpu_userq_post_reset(struct amdgpu_device *adev, bool vram_lost)
+{
+	/* if any queue state is AMDGPU_USERQ_STATE_UNMAPPED
+	 * at this point, we should be able to map it again
+	 * and continue if vram is not lost.
+	 */
+	struct amdgpu_userq_mgr *uqm;
+	struct amdgpu_usermode_queue *queue;
+	const struct amdgpu_userq_funcs *userq_funcs;
+	unsigned long queue_id;
+	int r = 0;
+
+	xa_for_each(&adev->userq_doorbell_xa, queue_id, queue) {
+		uqm = queue->userq_mgr;
+		if (queue->state == AMDGPU_USERQ_STATE_HUNG && !vram_lost) {
+			userq_funcs = adev->userq_funcs[queue->queue_type];
+			/* Re-map queue */
+			r = userq_funcs->map(uqm, queue);
+			if (r) {
+				dev_err(adev->dev, "Failed to remap queue %ld\n", queue_id);
+				continue;
+			}
+			queue->state = AMDGPU_USERQ_STATE_MAPPED;
+		}
+	}
+
+	return r;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h
new file mode 100644
index 000000000000..c37444427a14
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h
@@ -0,0 +1,161 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef AMDGPU_USERQ_H_
+#define AMDGPU_USERQ_H_
+#include "amdgpu_eviction_fence.h"
+
+#define AMDGPU_MAX_USERQ_COUNT 512
+
+#define to_ev_fence(f) container_of(f, struct amdgpu_eviction_fence, base)
+#define uq_mgr_to_fpriv(u) container_of(u, struct amdgpu_fpriv, userq_mgr)
+#define work_to_uq_mgr(w, name) container_of(w, struct amdgpu_userq_mgr, name)
+
+enum amdgpu_userq_state {
+	AMDGPU_USERQ_STATE_UNMAPPED = 0,
+	AMDGPU_USERQ_STATE_MAPPED,
+	AMDGPU_USERQ_STATE_PREEMPTED,
+	AMDGPU_USERQ_STATE_HUNG,
+	AMDGPU_USERQ_STATE_INVALID_VA,
+};
+
+struct amdgpu_mqd_prop;
+
+struct amdgpu_userq_obj {
+	void		 *cpu_ptr;
+	uint64_t	 gpu_addr;
+	struct amdgpu_bo *obj;
+};
+
+struct amdgpu_userq_va_cursor {
+	u64			gpu_addr;
+	struct list_head	list;
+};
+
+struct amdgpu_usermode_queue {
+	int			queue_type;
+	enum amdgpu_userq_state state;
+	uint64_t		doorbell_handle;
+	uint64_t		doorbell_index;
+	uint64_t		flags;
+	struct amdgpu_mqd_prop	*userq_prop;
+	struct amdgpu_userq_mgr *userq_mgr;
+	struct amdgpu_vm	*vm;
+	struct amdgpu_userq_obj mqd;
+	struct amdgpu_userq_obj	db_obj;
+	struct amdgpu_userq_obj fw_obj;
+	struct amdgpu_userq_obj wptr_obj;
+	struct xarray		fence_drv_xa;
+	struct amdgpu_userq_fence_driver *fence_drv;
+	struct dma_fence	*last_fence;
+	u32			xcp_id;
+	int			priority;
+	struct dentry		*debugfs_queue;
+
+	struct list_head	userq_va_list;
+};
+
+struct amdgpu_userq_funcs {
+	int (*mqd_create)(struct amdgpu_userq_mgr *uq_mgr,
+			  struct drm_amdgpu_userq_in *args,
+			  struct amdgpu_usermode_queue *queue);
+	void (*mqd_destroy)(struct amdgpu_userq_mgr *uq_mgr,
+			    struct amdgpu_usermode_queue *uq);
+	int (*unmap)(struct amdgpu_userq_mgr *uq_mgr,
+		     struct amdgpu_usermode_queue *queue);
+	int (*map)(struct amdgpu_userq_mgr *uq_mgr,
+		   struct amdgpu_usermode_queue *queue);
+	int (*preempt)(struct amdgpu_userq_mgr *uq_mgr,
+		   struct amdgpu_usermode_queue *queue);
+	int (*restore)(struct amdgpu_userq_mgr *uq_mgr,
+		   struct amdgpu_usermode_queue *queue);
+	int (*detect_and_reset)(struct amdgpu_device *adev,
+		  int queue_type);
+};
+
+/* Usermode queues for gfx */
+struct amdgpu_userq_mgr {
+	/**
+	 * @userq_mgr_xa: Per-process user queue map (queue ID → queue)
+	 * Key: queue_id (unique ID within the process's userq manager)
+	 * Value: struct amdgpu_usermode_queue
+	 */
+	struct xarray			userq_mgr_xa;
+	struct mutex			userq_mutex;
+	struct amdgpu_device		*adev;
+	struct delayed_work		resume_work;
+	struct drm_file			*file;
+	atomic_t                        userq_count[AMDGPU_RING_TYPE_MAX];
+};
+
+struct amdgpu_db_info {
+	uint64_t doorbell_handle;
+	uint32_t queue_type;
+	uint32_t doorbell_offset;
+	struct amdgpu_userq_obj	*db_obj;
+};
+
+int amdgpu_userq_ioctl(struct drm_device *dev, void *data, struct drm_file *filp);
+
+int amdgpu_userq_mgr_init(struct amdgpu_userq_mgr *userq_mgr, struct drm_file *file_priv,
+			  struct amdgpu_device *adev);
+
+void amdgpu_userq_mgr_fini(struct amdgpu_userq_mgr *userq_mgr);
+
+int amdgpu_userq_create_object(struct amdgpu_userq_mgr *uq_mgr,
+			       struct amdgpu_userq_obj *userq_obj,
+			       int size);
+
+void amdgpu_userq_destroy_object(struct amdgpu_userq_mgr *uq_mgr,
+				 struct amdgpu_userq_obj *userq_obj);
+
+void amdgpu_userq_evict(struct amdgpu_userq_mgr *uq_mgr,
+			struct amdgpu_eviction_fence *ev_fence);
+
+void amdgpu_userq_ensure_ev_fence(struct amdgpu_userq_mgr *userq_mgr,
+				  struct amdgpu_eviction_fence_mgr *evf_mgr);
+
+uint64_t amdgpu_userq_get_doorbell_index(struct amdgpu_userq_mgr *uq_mgr,
+					 struct amdgpu_db_info *db_info,
+					     struct drm_file *filp);
+
+u32 amdgpu_userq_get_supported_ip_mask(struct amdgpu_device *adev);
+
+int amdgpu_userq_suspend(struct amdgpu_device *adev);
+int amdgpu_userq_resume(struct amdgpu_device *adev);
+
+int amdgpu_userq_stop_sched_for_enforce_isolation(struct amdgpu_device *adev,
+						  u32 idx);
+int amdgpu_userq_start_sched_for_enforce_isolation(struct amdgpu_device *adev,
+						   u32 idx);
+void amdgpu_userq_reset_work(struct work_struct *work);
+void amdgpu_userq_pre_reset(struct amdgpu_device *adev);
+int amdgpu_userq_post_reset(struct amdgpu_device *adev, bool vram_lost);
+
+int amdgpu_userq_input_va_validate(struct amdgpu_usermode_queue *queue,
+				   u64 addr, u64 expected_size);
+int amdgpu_userq_gem_va_unmap_validate(struct amdgpu_device *adev,
+				       struct amdgpu_bo_va_mapping *mapping,
+				       uint64_t saddr);
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c
new file mode 100644
index 000000000000..eba9fb359047
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.c
@@ -0,0 +1,1011 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/kref.h>
+#include <linux/slab.h>
+#include <linux/dma-fence-unwrap.h>
+
+#include <drm/drm_exec.h>
+#include <drm/drm_syncobj.h>
+
+#include "amdgpu.h"
+#include "amdgpu_userq_fence.h"
+
+static const struct dma_fence_ops amdgpu_userq_fence_ops;
+static struct kmem_cache *amdgpu_userq_fence_slab;
+
+int amdgpu_userq_fence_slab_init(void)
+{
+	amdgpu_userq_fence_slab = kmem_cache_create("amdgpu_userq_fence",
+						    sizeof(struct amdgpu_userq_fence),
+						    0,
+						    SLAB_HWCACHE_ALIGN,
+						    NULL);
+	if (!amdgpu_userq_fence_slab)
+		return -ENOMEM;
+
+	return 0;
+}
+
+void amdgpu_userq_fence_slab_fini(void)
+{
+	rcu_barrier();
+	kmem_cache_destroy(amdgpu_userq_fence_slab);
+}
+
+static inline struct amdgpu_userq_fence *to_amdgpu_userq_fence(struct dma_fence *f)
+{
+	if (!f || f->ops != &amdgpu_userq_fence_ops)
+		return NULL;
+
+	return container_of(f, struct amdgpu_userq_fence, base);
+}
+
+static u64 amdgpu_userq_fence_read(struct amdgpu_userq_fence_driver *fence_drv)
+{
+	return le64_to_cpu(*fence_drv->cpu_addr);
+}
+
+static void
+amdgpu_userq_fence_write(struct amdgpu_userq_fence_driver *fence_drv,
+			 u64 seq)
+{
+	if (fence_drv->cpu_addr)
+		*fence_drv->cpu_addr = cpu_to_le64(seq);
+}
+
+int amdgpu_userq_fence_driver_alloc(struct amdgpu_device *adev,
+				    struct amdgpu_usermode_queue *userq)
+{
+	struct amdgpu_userq_fence_driver *fence_drv;
+	unsigned long flags;
+	int r;
+
+	fence_drv = kzalloc(sizeof(*fence_drv), GFP_KERNEL);
+	if (!fence_drv)
+		return -ENOMEM;
+
+	/* Acquire seq64 memory */
+	r = amdgpu_seq64_alloc(adev, &fence_drv->va, &fence_drv->gpu_addr,
+			       &fence_drv->cpu_addr);
+	if (r)
+		goto free_fence_drv;
+
+	memset(fence_drv->cpu_addr, 0, sizeof(u64));
+
+	kref_init(&fence_drv->refcount);
+	INIT_LIST_HEAD(&fence_drv->fences);
+	spin_lock_init(&fence_drv->fence_list_lock);
+
+	fence_drv->adev = adev;
+	fence_drv->context = dma_fence_context_alloc(1);
+	get_task_comm(fence_drv->timeline_name, current);
+
+	xa_lock_irqsave(&adev->userq_xa, flags);
+	r = xa_err(__xa_store(&adev->userq_xa, userq->doorbell_index,
+			      fence_drv, GFP_KERNEL));
+	xa_unlock_irqrestore(&adev->userq_xa, flags);
+	if (r)
+		goto free_seq64;
+
+	userq->fence_drv = fence_drv;
+
+	return 0;
+
+free_seq64:
+	amdgpu_seq64_free(adev, fence_drv->va);
+free_fence_drv:
+	kfree(fence_drv);
+
+	return r;
+}
+
+static void amdgpu_userq_walk_and_drop_fence_drv(struct xarray *xa)
+{
+	struct amdgpu_userq_fence_driver *fence_drv;
+	unsigned long index;
+
+	if (xa_empty(xa))
+		return;
+
+	xa_lock(xa);
+	xa_for_each(xa, index, fence_drv) {
+		__xa_erase(xa, index);
+		amdgpu_userq_fence_driver_put(fence_drv);
+	}
+
+	xa_unlock(xa);
+}
+
+void
+amdgpu_userq_fence_driver_free(struct amdgpu_usermode_queue *userq)
+{
+	amdgpu_userq_walk_and_drop_fence_drv(&userq->fence_drv_xa);
+	xa_destroy(&userq->fence_drv_xa);
+	/* Drop the fence_drv reference held by user queue */
+	amdgpu_userq_fence_driver_put(userq->fence_drv);
+}
+
+void amdgpu_userq_fence_driver_process(struct amdgpu_userq_fence_driver *fence_drv)
+{
+	struct amdgpu_userq_fence *userq_fence, *tmp;
+	struct dma_fence *fence;
+	unsigned long flags;
+	u64 rptr;
+	int i;
+
+	if (!fence_drv)
+		return;
+
+	spin_lock_irqsave(&fence_drv->fence_list_lock, flags);
+	rptr = amdgpu_userq_fence_read(fence_drv);
+
+	list_for_each_entry_safe(userq_fence, tmp, &fence_drv->fences, link) {
+		fence = &userq_fence->base;
+
+		if (rptr < fence->seqno)
+			break;
+
+		dma_fence_signal(fence);
+
+		for (i = 0; i < userq_fence->fence_drv_array_count; i++)
+			amdgpu_userq_fence_driver_put(userq_fence->fence_drv_array[i]);
+
+		list_del(&userq_fence->link);
+		dma_fence_put(fence);
+	}
+	spin_unlock_irqrestore(&fence_drv->fence_list_lock, flags);
+}
+
+void amdgpu_userq_fence_driver_destroy(struct kref *ref)
+{
+	struct amdgpu_userq_fence_driver *fence_drv = container_of(ref,
+					 struct amdgpu_userq_fence_driver,
+					 refcount);
+	struct amdgpu_userq_fence_driver *xa_fence_drv;
+	struct amdgpu_device *adev = fence_drv->adev;
+	struct amdgpu_userq_fence *fence, *tmp;
+	struct xarray *xa = &adev->userq_xa;
+	unsigned long index, flags;
+	struct dma_fence *f;
+
+	spin_lock_irqsave(&fence_drv->fence_list_lock, flags);
+	list_for_each_entry_safe(fence, tmp, &fence_drv->fences, link) {
+		f = &fence->base;
+
+		if (!dma_fence_is_signaled(f)) {
+			dma_fence_set_error(f, -ECANCELED);
+			dma_fence_signal(f);
+		}
+
+		list_del(&fence->link);
+		dma_fence_put(f);
+	}
+	spin_unlock_irqrestore(&fence_drv->fence_list_lock, flags);
+
+	xa_lock_irqsave(xa, flags);
+	xa_for_each(xa, index, xa_fence_drv)
+		if (xa_fence_drv == fence_drv)
+			__xa_erase(xa, index);
+	xa_unlock_irqrestore(xa, flags);
+
+	/* Free seq64 memory */
+	amdgpu_seq64_free(adev, fence_drv->va);
+	kfree(fence_drv);
+}
+
+void amdgpu_userq_fence_driver_get(struct amdgpu_userq_fence_driver *fence_drv)
+{
+	kref_get(&fence_drv->refcount);
+}
+
+void amdgpu_userq_fence_driver_put(struct amdgpu_userq_fence_driver *fence_drv)
+{
+	kref_put(&fence_drv->refcount, amdgpu_userq_fence_driver_destroy);
+}
+
+static int amdgpu_userq_fence_alloc(struct amdgpu_userq_fence **userq_fence)
+{
+	*userq_fence = kmem_cache_alloc(amdgpu_userq_fence_slab, GFP_ATOMIC);
+	return *userq_fence ? 0 : -ENOMEM;
+}
+
+static int amdgpu_userq_fence_create(struct amdgpu_usermode_queue *userq,
+				     struct amdgpu_userq_fence *userq_fence,
+				     u64 seq, struct dma_fence **f)
+{
+	struct amdgpu_userq_fence_driver *fence_drv;
+	struct dma_fence *fence;
+	unsigned long flags;
+
+	fence_drv = userq->fence_drv;
+	if (!fence_drv)
+		return -EINVAL;
+
+	spin_lock_init(&userq_fence->lock);
+	INIT_LIST_HEAD(&userq_fence->link);
+	fence = &userq_fence->base;
+	userq_fence->fence_drv = fence_drv;
+
+	dma_fence_init64(fence, &amdgpu_userq_fence_ops, &userq_fence->lock,
+			 fence_drv->context, seq);
+
+	amdgpu_userq_fence_driver_get(fence_drv);
+	dma_fence_get(fence);
+
+	if (!xa_empty(&userq->fence_drv_xa)) {
+		struct amdgpu_userq_fence_driver *stored_fence_drv;
+		unsigned long index, count = 0;
+		int i = 0;
+
+		xa_lock(&userq->fence_drv_xa);
+		xa_for_each(&userq->fence_drv_xa, index, stored_fence_drv)
+			count++;
+
+		userq_fence->fence_drv_array =
+			kvmalloc_array(count,
+				       sizeof(struct amdgpu_userq_fence_driver *),
+				       GFP_ATOMIC);
+
+		if (userq_fence->fence_drv_array) {
+			xa_for_each(&userq->fence_drv_xa, index, stored_fence_drv) {
+				userq_fence->fence_drv_array[i] = stored_fence_drv;
+				__xa_erase(&userq->fence_drv_xa, index);
+				i++;
+			}
+		}
+
+		userq_fence->fence_drv_array_count = i;
+		xa_unlock(&userq->fence_drv_xa);
+	} else {
+		userq_fence->fence_drv_array = NULL;
+		userq_fence->fence_drv_array_count = 0;
+	}
+
+	/* Check if hardware has already processed the job */
+	spin_lock_irqsave(&fence_drv->fence_list_lock, flags);
+	if (!dma_fence_is_signaled(fence))
+		list_add_tail(&userq_fence->link, &fence_drv->fences);
+	else
+		dma_fence_put(fence);
+
+	spin_unlock_irqrestore(&fence_drv->fence_list_lock, flags);
+
+	*f = fence;
+
+	return 0;
+}
+
+static const char *amdgpu_userq_fence_get_driver_name(struct dma_fence *f)
+{
+	return "amdgpu_userq_fence";
+}
+
+static const char *amdgpu_userq_fence_get_timeline_name(struct dma_fence *f)
+{
+	struct amdgpu_userq_fence *fence = to_amdgpu_userq_fence(f);
+
+	return fence->fence_drv->timeline_name;
+}
+
+static bool amdgpu_userq_fence_signaled(struct dma_fence *f)
+{
+	struct amdgpu_userq_fence *fence = to_amdgpu_userq_fence(f);
+	struct amdgpu_userq_fence_driver *fence_drv = fence->fence_drv;
+	u64 rptr, wptr;
+
+	rptr = amdgpu_userq_fence_read(fence_drv);
+	wptr = fence->base.seqno;
+
+	if (rptr >= wptr)
+		return true;
+
+	return false;
+}
+
+static void amdgpu_userq_fence_free(struct rcu_head *rcu)
+{
+	struct dma_fence *fence = container_of(rcu, struct dma_fence, rcu);
+	struct amdgpu_userq_fence *userq_fence = to_amdgpu_userq_fence(fence);
+	struct amdgpu_userq_fence_driver *fence_drv = userq_fence->fence_drv;
+
+	/* Release the fence driver reference */
+	amdgpu_userq_fence_driver_put(fence_drv);
+
+	kvfree(userq_fence->fence_drv_array);
+	kmem_cache_free(amdgpu_userq_fence_slab, userq_fence);
+}
+
+static void amdgpu_userq_fence_release(struct dma_fence *f)
+{
+	call_rcu(&f->rcu, amdgpu_userq_fence_free);
+}
+
+static const struct dma_fence_ops amdgpu_userq_fence_ops = {
+	.get_driver_name = amdgpu_userq_fence_get_driver_name,
+	.get_timeline_name = amdgpu_userq_fence_get_timeline_name,
+	.signaled = amdgpu_userq_fence_signaled,
+	.release = amdgpu_userq_fence_release,
+};
+
+/**
+ * amdgpu_userq_fence_read_wptr - Read the userq wptr value
+ *
+ * @queue: user mode queue structure pointer
+ * @wptr: write pointer value
+ *
+ * Read the wptr value from userq's MQD. The userq signal IOCTL
+ * creates a dma_fence for the shared buffers that expects the
+ * RPTR value written to seq64 memory >= WPTR.
+ *
+ * Returns wptr value on success, error on failure.
+ */
+static int amdgpu_userq_fence_read_wptr(struct amdgpu_usermode_queue *queue,
+					u64 *wptr)
+{
+	struct amdgpu_bo_va_mapping *mapping;
+	struct amdgpu_bo *bo;
+	u64 addr, *ptr;
+	int r;
+
+	r = amdgpu_bo_reserve(queue->vm->root.bo, false);
+	if (r)
+		return r;
+
+	addr = queue->userq_prop->wptr_gpu_addr;
+	addr &= AMDGPU_GMC_HOLE_MASK;
+
+	mapping = amdgpu_vm_bo_lookup_mapping(queue->vm, addr >> PAGE_SHIFT);
+	if (!mapping) {
+		amdgpu_bo_unreserve(queue->vm->root.bo);
+		DRM_ERROR("Failed to lookup amdgpu_bo_va_mapping\n");
+		return -EINVAL;
+	}
+
+	bo = amdgpu_bo_ref(mapping->bo_va->base.bo);
+	amdgpu_bo_unreserve(queue->vm->root.bo);
+	r = amdgpu_bo_reserve(bo, true);
+	if (r) {
+		amdgpu_bo_unref(&bo);
+		DRM_ERROR("Failed to reserve userqueue wptr bo");
+		return r;
+	}
+
+	r = amdgpu_bo_kmap(bo, (void **)&ptr);
+	if (r) {
+		DRM_ERROR("Failed mapping the userqueue wptr bo");
+		goto map_error;
+	}
+
+	*wptr = le64_to_cpu(*ptr);
+
+	amdgpu_bo_kunmap(bo);
+	amdgpu_bo_unreserve(bo);
+	amdgpu_bo_unref(&bo);
+
+	return 0;
+
+map_error:
+	amdgpu_bo_unreserve(bo);
+	amdgpu_bo_unref(&bo);
+
+	return r;
+}
+
+static void amdgpu_userq_fence_cleanup(struct dma_fence *fence)
+{
+	dma_fence_put(fence);
+}
+
+static void
+amdgpu_userq_fence_driver_set_error(struct amdgpu_userq_fence *fence,
+				    int error)
+{
+	struct amdgpu_userq_fence_driver *fence_drv = fence->fence_drv;
+	unsigned long flags;
+	struct dma_fence *f;
+
+	spin_lock_irqsave(&fence_drv->fence_list_lock, flags);
+
+	f = rcu_dereference_protected(&fence->base,
+				      lockdep_is_held(&fence_drv->fence_list_lock));
+	if (f && !dma_fence_is_signaled_locked(f))
+		dma_fence_set_error(f, error);
+	spin_unlock_irqrestore(&fence_drv->fence_list_lock, flags);
+}
+
+void
+amdgpu_userq_fence_driver_force_completion(struct amdgpu_usermode_queue *userq)
+{
+	struct dma_fence *f = userq->last_fence;
+
+	if (f) {
+		struct amdgpu_userq_fence *fence = to_amdgpu_userq_fence(f);
+		struct amdgpu_userq_fence_driver *fence_drv = fence->fence_drv;
+		u64 wptr = fence->base.seqno;
+
+		amdgpu_userq_fence_driver_set_error(fence, -ECANCELED);
+		amdgpu_userq_fence_write(fence_drv, wptr);
+		amdgpu_userq_fence_driver_process(fence_drv);
+
+	}
+}
+
+int amdgpu_userq_signal_ioctl(struct drm_device *dev, void *data,
+			      struct drm_file *filp)
+{
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
+	struct amdgpu_userq_mgr *userq_mgr = &fpriv->userq_mgr;
+	struct drm_amdgpu_userq_signal *args = data;
+	struct drm_gem_object **gobj_write = NULL;
+	struct drm_gem_object **gobj_read = NULL;
+	struct amdgpu_usermode_queue *queue;
+	struct amdgpu_userq_fence *userq_fence;
+	struct drm_syncobj **syncobj = NULL;
+	u32 *bo_handles_write, num_write_bo_handles;
+	u32 *syncobj_handles, num_syncobj_handles;
+	u32 *bo_handles_read, num_read_bo_handles;
+	int r, i, entry, rentry, wentry;
+	struct dma_fence *fence;
+	struct drm_exec exec;
+	u64 wptr;
+
+	num_syncobj_handles = args->num_syncobj_handles;
+	syncobj_handles = memdup_user(u64_to_user_ptr(args->syncobj_handles),
+				      size_mul(sizeof(u32), num_syncobj_handles));
+	if (IS_ERR(syncobj_handles))
+		return PTR_ERR(syncobj_handles);
+
+	/* Array of pointers to the looked up syncobjs */
+	syncobj = kmalloc_array(num_syncobj_handles, sizeof(*syncobj), GFP_KERNEL);
+	if (!syncobj) {
+		r = -ENOMEM;
+		goto free_syncobj_handles;
+	}
+
+	for (entry = 0; entry < num_syncobj_handles; entry++) {
+		syncobj[entry] = drm_syncobj_find(filp, syncobj_handles[entry]);
+		if (!syncobj[entry]) {
+			r = -ENOENT;
+			goto free_syncobj;
+		}
+	}
+
+	num_read_bo_handles = args->num_bo_read_handles;
+	bo_handles_read = memdup_user(u64_to_user_ptr(args->bo_read_handles),
+				      sizeof(u32) * num_read_bo_handles);
+	if (IS_ERR(bo_handles_read)) {
+		r = PTR_ERR(bo_handles_read);
+		goto free_syncobj;
+	}
+
+	/* Array of pointers to the GEM read objects */
+	gobj_read = kmalloc_array(num_read_bo_handles, sizeof(*gobj_read), GFP_KERNEL);
+	if (!gobj_read) {
+		r = -ENOMEM;
+		goto free_bo_handles_read;
+	}
+
+	for (rentry = 0; rentry < num_read_bo_handles; rentry++) {
+		gobj_read[rentry] = drm_gem_object_lookup(filp, bo_handles_read[rentry]);
+		if (!gobj_read[rentry]) {
+			r = -ENOENT;
+			goto put_gobj_read;
+		}
+	}
+
+	num_write_bo_handles = args->num_bo_write_handles;
+	bo_handles_write = memdup_user(u64_to_user_ptr(args->bo_write_handles),
+				       sizeof(u32) * num_write_bo_handles);
+	if (IS_ERR(bo_handles_write)) {
+		r = PTR_ERR(bo_handles_write);
+		goto put_gobj_read;
+	}
+
+	/* Array of pointers to the GEM write objects */
+	gobj_write = kmalloc_array(num_write_bo_handles, sizeof(*gobj_write), GFP_KERNEL);
+	if (!gobj_write) {
+		r = -ENOMEM;
+		goto free_bo_handles_write;
+	}
+
+	for (wentry = 0; wentry < num_write_bo_handles; wentry++) {
+		gobj_write[wentry] = drm_gem_object_lookup(filp, bo_handles_write[wentry]);
+		if (!gobj_write[wentry]) {
+			r = -ENOENT;
+			goto put_gobj_write;
+		}
+	}
+
+	/* Retrieve the user queue */
+	queue = xa_load(&userq_mgr->userq_mgr_xa, args->queue_id);
+	if (!queue) {
+		r = -ENOENT;
+		goto put_gobj_write;
+	}
+
+	r = amdgpu_userq_fence_read_wptr(queue, &wptr);
+	if (r)
+		goto put_gobj_write;
+
+	r = amdgpu_userq_fence_alloc(&userq_fence);
+	if (r)
+		goto put_gobj_write;
+
+	/* We are here means UQ is active, make sure the eviction fence is valid */
+	amdgpu_userq_ensure_ev_fence(&fpriv->userq_mgr, &fpriv->evf_mgr);
+
+	/* Create a new fence */
+	r = amdgpu_userq_fence_create(queue, userq_fence, wptr, &fence);
+	if (r) {
+		mutex_unlock(&userq_mgr->userq_mutex);
+		kmem_cache_free(amdgpu_userq_fence_slab, userq_fence);
+		goto put_gobj_write;
+	}
+
+	dma_fence_put(queue->last_fence);
+	queue->last_fence = dma_fence_get(fence);
+	mutex_unlock(&userq_mgr->userq_mutex);
+
+	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT,
+		      (num_read_bo_handles + num_write_bo_handles));
+
+	/* Lock all BOs with retry handling */
+	drm_exec_until_all_locked(&exec) {
+		r = drm_exec_prepare_array(&exec, gobj_read, num_read_bo_handles, 1);
+		drm_exec_retry_on_contention(&exec);
+		if (r) {
+			amdgpu_userq_fence_cleanup(fence);
+			goto exec_fini;
+		}
+
+		r = drm_exec_prepare_array(&exec, gobj_write, num_write_bo_handles, 1);
+		drm_exec_retry_on_contention(&exec);
+		if (r) {
+			amdgpu_userq_fence_cleanup(fence);
+			goto exec_fini;
+		}
+	}
+
+	for (i = 0; i < num_read_bo_handles; i++) {
+		if (!gobj_read || !gobj_read[i]->resv)
+			continue;
+
+		dma_resv_add_fence(gobj_read[i]->resv, fence,
+				   DMA_RESV_USAGE_READ);
+	}
+
+	for (i = 0; i < num_write_bo_handles; i++) {
+		if (!gobj_write || !gobj_write[i]->resv)
+			continue;
+
+		dma_resv_add_fence(gobj_write[i]->resv, fence,
+				   DMA_RESV_USAGE_WRITE);
+	}
+
+	/* Add the created fence to syncobj/BO's */
+	for (i = 0; i < num_syncobj_handles; i++)
+		drm_syncobj_replace_fence(syncobj[i], fence);
+
+	/* drop the reference acquired in fence creation function */
+	dma_fence_put(fence);
+
+exec_fini:
+	drm_exec_fini(&exec);
+put_gobj_write:
+	while (wentry-- > 0)
+		drm_gem_object_put(gobj_write[wentry]);
+	kfree(gobj_write);
+free_bo_handles_write:
+	kfree(bo_handles_write);
+put_gobj_read:
+	while (rentry-- > 0)
+		drm_gem_object_put(gobj_read[rentry]);
+	kfree(gobj_read);
+free_bo_handles_read:
+	kfree(bo_handles_read);
+free_syncobj:
+	while (entry-- > 0)
+		if (syncobj[entry])
+			drm_syncobj_put(syncobj[entry]);
+	kfree(syncobj);
+free_syncobj_handles:
+	kfree(syncobj_handles);
+
+	return r;
+}
+
+int amdgpu_userq_wait_ioctl(struct drm_device *dev, void *data,
+			    struct drm_file *filp)
+{
+	u32 *syncobj_handles, *timeline_points, *timeline_handles, *bo_handles_read, *bo_handles_write;
+	u32 num_syncobj, num_read_bo_handles, num_write_bo_handles;
+	struct drm_amdgpu_userq_fence_info *fence_info = NULL;
+	struct drm_amdgpu_userq_wait *wait_info = data;
+	struct amdgpu_fpriv *fpriv = filp->driver_priv;
+	struct amdgpu_userq_mgr *userq_mgr = &fpriv->userq_mgr;
+	struct amdgpu_usermode_queue *waitq;
+	struct drm_gem_object **gobj_write;
+	struct drm_gem_object **gobj_read;
+	struct dma_fence **fences = NULL;
+	u16 num_points, num_fences = 0;
+	int r, i, rentry, wentry, cnt;
+	struct drm_exec exec;
+
+	num_read_bo_handles = wait_info->num_bo_read_handles;
+	bo_handles_read = memdup_user(u64_to_user_ptr(wait_info->bo_read_handles),
+				      size_mul(sizeof(u32), num_read_bo_handles));
+	if (IS_ERR(bo_handles_read))
+		return PTR_ERR(bo_handles_read);
+
+	num_write_bo_handles = wait_info->num_bo_write_handles;
+	bo_handles_write = memdup_user(u64_to_user_ptr(wait_info->bo_write_handles),
+				       size_mul(sizeof(u32), num_write_bo_handles));
+	if (IS_ERR(bo_handles_write)) {
+		r = PTR_ERR(bo_handles_write);
+		goto free_bo_handles_read;
+	}
+
+	num_syncobj = wait_info->num_syncobj_handles;
+	syncobj_handles = memdup_user(u64_to_user_ptr(wait_info->syncobj_handles),
+				      size_mul(sizeof(u32), num_syncobj));
+	if (IS_ERR(syncobj_handles)) {
+		r = PTR_ERR(syncobj_handles);
+		goto free_bo_handles_write;
+	}
+
+	num_points = wait_info->num_syncobj_timeline_handles;
+	timeline_handles = memdup_user(u64_to_user_ptr(wait_info->syncobj_timeline_handles),
+				       sizeof(u32) * num_points);
+	if (IS_ERR(timeline_handles)) {
+		r = PTR_ERR(timeline_handles);
+		goto free_syncobj_handles;
+	}
+
+	timeline_points = memdup_user(u64_to_user_ptr(wait_info->syncobj_timeline_points),
+				      sizeof(u32) * num_points);
+	if (IS_ERR(timeline_points)) {
+		r = PTR_ERR(timeline_points);
+		goto free_timeline_handles;
+	}
+
+	gobj_read = kmalloc_array(num_read_bo_handles, sizeof(*gobj_read), GFP_KERNEL);
+	if (!gobj_read) {
+		r = -ENOMEM;
+		goto free_timeline_points;
+	}
+
+	for (rentry = 0; rentry < num_read_bo_handles; rentry++) {
+		gobj_read[rentry] = drm_gem_object_lookup(filp, bo_handles_read[rentry]);
+		if (!gobj_read[rentry]) {
+			r = -ENOENT;
+			goto put_gobj_read;
+		}
+	}
+
+	gobj_write = kmalloc_array(num_write_bo_handles, sizeof(*gobj_write), GFP_KERNEL);
+	if (!gobj_write) {
+		r = -ENOMEM;
+		goto put_gobj_read;
+	}
+
+	for (wentry = 0; wentry < num_write_bo_handles; wentry++) {
+		gobj_write[wentry] = drm_gem_object_lookup(filp, bo_handles_write[wentry]);
+		if (!gobj_write[wentry]) {
+			r = -ENOENT;
+			goto put_gobj_write;
+		}
+	}
+
+	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT,
+		      (num_read_bo_handles + num_write_bo_handles));
+
+	/* Lock all BOs with retry handling */
+	drm_exec_until_all_locked(&exec) {
+		r = drm_exec_prepare_array(&exec, gobj_read, num_read_bo_handles, 1);
+		drm_exec_retry_on_contention(&exec);
+		if (r) {
+			drm_exec_fini(&exec);
+			goto put_gobj_write;
+		}
+
+		r = drm_exec_prepare_array(&exec, gobj_write, num_write_bo_handles, 1);
+		drm_exec_retry_on_contention(&exec);
+		if (r) {
+			drm_exec_fini(&exec);
+			goto put_gobj_write;
+		}
+	}
+
+	if (!wait_info->num_fences) {
+		if (num_points) {
+			struct dma_fence_unwrap iter;
+			struct dma_fence *fence;
+			struct dma_fence *f;
+
+			for (i = 0; i < num_points; i++) {
+				r = drm_syncobj_find_fence(filp, timeline_handles[i],
+							   timeline_points[i],
+							   DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT,
+							   &fence);
+				if (r)
+					goto exec_fini;
+
+				dma_fence_unwrap_for_each(f, &iter, fence)
+					num_fences++;
+
+				dma_fence_put(fence);
+			}
+		}
+
+		/* Count syncobj's fence */
+		for (i = 0; i < num_syncobj; i++) {
+			struct dma_fence *fence;
+
+			r = drm_syncobj_find_fence(filp, syncobj_handles[i],
+						   0,
+						   DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT,
+						   &fence);
+			if (r)
+				goto exec_fini;
+
+			num_fences++;
+			dma_fence_put(fence);
+		}
+
+		/* Count GEM objects fence */
+		for (i = 0; i < num_read_bo_handles; i++) {
+			struct dma_resv_iter resv_cursor;
+			struct dma_fence *fence;
+
+			dma_resv_for_each_fence(&resv_cursor, gobj_read[i]->resv,
+						DMA_RESV_USAGE_READ, fence)
+				num_fences++;
+		}
+
+		for (i = 0; i < num_write_bo_handles; i++) {
+			struct dma_resv_iter resv_cursor;
+			struct dma_fence *fence;
+
+			dma_resv_for_each_fence(&resv_cursor, gobj_write[i]->resv,
+						DMA_RESV_USAGE_WRITE, fence)
+				num_fences++;
+		}
+
+		/*
+		 * Passing num_fences = 0 means that userspace doesn't want to
+		 * retrieve userq_fence_info. If num_fences = 0 we skip filling
+		 * userq_fence_info and return the actual number of fences on
+		 * args->num_fences.
+		 */
+		wait_info->num_fences = num_fences;
+	} else {
+		/* Array of fence info */
+		fence_info = kmalloc_array(wait_info->num_fences, sizeof(*fence_info), GFP_KERNEL);
+		if (!fence_info) {
+			r = -ENOMEM;
+			goto exec_fini;
+		}
+
+		/* Array of fences */
+		fences = kmalloc_array(wait_info->num_fences, sizeof(*fences), GFP_KERNEL);
+		if (!fences) {
+			r = -ENOMEM;
+			goto free_fence_info;
+		}
+
+		/* Retrieve GEM read objects fence */
+		for (i = 0; i < num_read_bo_handles; i++) {
+			struct dma_resv_iter resv_cursor;
+			struct dma_fence *fence;
+
+			dma_resv_for_each_fence(&resv_cursor, gobj_read[i]->resv,
+						DMA_RESV_USAGE_READ, fence) {
+				if (WARN_ON_ONCE(num_fences >= wait_info->num_fences)) {
+					r = -EINVAL;
+					goto free_fences;
+				}
+
+				fences[num_fences++] = fence;
+				dma_fence_get(fence);
+			}
+		}
+
+		/* Retrieve GEM write objects fence */
+		for (i = 0; i < num_write_bo_handles; i++) {
+			struct dma_resv_iter resv_cursor;
+			struct dma_fence *fence;
+
+			dma_resv_for_each_fence(&resv_cursor, gobj_write[i]->resv,
+						DMA_RESV_USAGE_WRITE, fence) {
+				if (WARN_ON_ONCE(num_fences >= wait_info->num_fences)) {
+					r = -EINVAL;
+					goto free_fences;
+				}
+
+				fences[num_fences++] = fence;
+				dma_fence_get(fence);
+			}
+		}
+
+		if (num_points) {
+			struct dma_fence_unwrap iter;
+			struct dma_fence *fence;
+			struct dma_fence *f;
+
+			for (i = 0; i < num_points; i++) {
+				r = drm_syncobj_find_fence(filp, timeline_handles[i],
+							   timeline_points[i],
+							   DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT,
+							   &fence);
+				if (r)
+					goto free_fences;
+
+				dma_fence_unwrap_for_each(f, &iter, fence) {
+					if (WARN_ON_ONCE(num_fences >= wait_info->num_fences)) {
+						r = -EINVAL;
+						goto free_fences;
+					}
+
+					dma_fence_get(f);
+					fences[num_fences++] = f;
+				}
+
+				dma_fence_put(fence);
+			}
+		}
+
+		/* Retrieve syncobj's fence */
+		for (i = 0; i < num_syncobj; i++) {
+			struct dma_fence *fence;
+
+			r = drm_syncobj_find_fence(filp, syncobj_handles[i],
+						   0,
+						   DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT,
+						   &fence);
+			if (r)
+				goto free_fences;
+
+			if (WARN_ON_ONCE(num_fences >= wait_info->num_fences)) {
+				r = -EINVAL;
+				goto free_fences;
+			}
+
+			fences[num_fences++] = fence;
+		}
+
+		/*
+		 * Keep only the latest fences to reduce the number of values
+		 * given back to userspace.
+		 */
+		num_fences = dma_fence_dedup_array(fences, num_fences);
+
+		waitq = xa_load(&userq_mgr->userq_mgr_xa, wait_info->waitq_id);
+		if (!waitq) {
+			r = -EINVAL;
+			goto free_fences;
+		}
+
+		for (i = 0, cnt = 0; i < num_fences; i++) {
+			struct amdgpu_userq_fence_driver *fence_drv;
+			struct amdgpu_userq_fence *userq_fence;
+			u32 index;
+
+			userq_fence = to_amdgpu_userq_fence(fences[i]);
+			if (!userq_fence) {
+				/*
+				 * Just waiting on other driver fences should
+				 * be good for now
+				 */
+				r = dma_fence_wait(fences[i], true);
+				if (r) {
+					dma_fence_put(fences[i]);
+					goto free_fences;
+				}
+
+				dma_fence_put(fences[i]);
+				continue;
+			}
+
+			fence_drv = userq_fence->fence_drv;
+			/*
+			 * We need to make sure the user queue release their reference
+			 * to the fence drivers at some point before queue destruction.
+			 * Otherwise, we would gather those references until we don't
+			 * have any more space left and crash.
+			 */
+			r = xa_alloc(&waitq->fence_drv_xa, &index, fence_drv,
+				     xa_limit_32b, GFP_KERNEL);
+			if (r)
+				goto free_fences;
+
+			amdgpu_userq_fence_driver_get(fence_drv);
+
+			/* Store drm syncobj's gpu va address and value */
+			fence_info[cnt].va = fence_drv->va;
+			fence_info[cnt].value = fences[i]->seqno;
+
+			dma_fence_put(fences[i]);
+			/* Increment the actual userq fence count */
+			cnt++;
+		}
+
+		wait_info->num_fences = cnt;
+		/* Copy userq fence info to user space */
+		if (copy_to_user(u64_to_user_ptr(wait_info->out_fences),
+				 fence_info, wait_info->num_fences * sizeof(*fence_info))) {
+			r = -EFAULT;
+			goto free_fences;
+		}
+
+		kfree(fences);
+		kfree(fence_info);
+	}
+
+	drm_exec_fini(&exec);
+	for (i = 0; i < num_read_bo_handles; i++)
+		drm_gem_object_put(gobj_read[i]);
+	kfree(gobj_read);
+
+	for (i = 0; i < num_write_bo_handles; i++)
+		drm_gem_object_put(gobj_write[i]);
+	kfree(gobj_write);
+
+	kfree(timeline_points);
+	kfree(timeline_handles);
+	kfree(syncobj_handles);
+	kfree(bo_handles_write);
+	kfree(bo_handles_read);
+
+	return 0;
+
+free_fences:
+	while (num_fences-- > 0)
+		dma_fence_put(fences[num_fences]);
+	kfree(fences);
+free_fence_info:
+	kfree(fence_info);
+exec_fini:
+	drm_exec_fini(&exec);
+put_gobj_write:
+	while (wentry-- > 0)
+		drm_gem_object_put(gobj_write[wentry]);
+	kfree(gobj_write);
+put_gobj_read:
+	while (rentry-- > 0)
+		drm_gem_object_put(gobj_read[rentry]);
+	kfree(gobj_read);
+free_timeline_points:
+	kfree(timeline_points);
+free_timeline_handles:
+	kfree(timeline_handles);
+free_syncobj_handles:
+	kfree(syncobj_handles);
+free_bo_handles_write:
+	kfree(bo_handles_write);
+free_bo_handles_read:
+	kfree(bo_handles_read);
+
+	return r;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.h
new file mode 100644
index 000000000000..d76add2afc77
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_userq_fence.h
@@ -0,0 +1,77 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __AMDGPU_USERQ_FENCE_H__
+#define __AMDGPU_USERQ_FENCE_H__
+
+#include <linux/types.h>
+
+#include "amdgpu_userq.h"
+
+struct amdgpu_userq_fence {
+	struct dma_fence base;
+	/*
+	 * This lock is necessary to synchronize the
+	 * userqueue dma fence operations.
+	 */
+	spinlock_t lock;
+	struct list_head link;
+	unsigned long fence_drv_array_count;
+	struct amdgpu_userq_fence_driver *fence_drv;
+	struct amdgpu_userq_fence_driver **fence_drv_array;
+};
+
+struct amdgpu_userq_fence_driver {
+	struct kref refcount;
+	u64 va;
+	u64 gpu_addr;
+	u64 *cpu_addr;
+	u64 context;
+	/*
+	 * This lock is necesaary to synchronize the access
+	 * to the fences list by the fence driver.
+	 */
+	spinlock_t fence_list_lock;
+	struct list_head fences;
+	struct amdgpu_device *adev;
+	char timeline_name[TASK_COMM_LEN];
+};
+
+int amdgpu_userq_fence_slab_init(void);
+void amdgpu_userq_fence_slab_fini(void);
+
+void amdgpu_userq_fence_driver_get(struct amdgpu_userq_fence_driver *fence_drv);
+void amdgpu_userq_fence_driver_put(struct amdgpu_userq_fence_driver *fence_drv);
+int amdgpu_userq_fence_driver_alloc(struct amdgpu_device *adev,
+				    struct amdgpu_usermode_queue *userq);
+void amdgpu_userq_fence_driver_free(struct amdgpu_usermode_queue *userq);
+void amdgpu_userq_fence_driver_process(struct amdgpu_userq_fence_driver *fence_drv);
+void amdgpu_userq_fence_driver_force_completion(struct amdgpu_usermode_queue *userq);
+void amdgpu_userq_fence_driver_destroy(struct kref *ref);
+int amdgpu_userq_signal_ioctl(struct drm_device *dev, void *data,
+			      struct drm_file *filp);
+int amdgpu_userq_wait_ioctl(struct drm_device *dev, void *data,
+			    struct drm_file *filp);
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_utils.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_utils.h
new file mode 100644
index 000000000000..1e40ca3b1584
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_utils.h
@@ -0,0 +1,91 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef AMDGPU_UTILS_H_
+#define AMDGPU_UTILS_H_
+
+/* ---------- Generic 2‑bit capability attribute encoding ----------
+ * 00 INVALID, 01 RO, 10 WO, 11 RW
+ */
+enum amdgpu_cap_attr {
+	AMDGPU_CAP_ATTR_INVALID = 0,
+	AMDGPU_CAP_ATTR_RO      = 1 << 0,
+	AMDGPU_CAP_ATTR_WO      = 1 << 1,
+	AMDGPU_CAP_ATTR_RW      = (AMDGPU_CAP_ATTR_RO | AMDGPU_CAP_ATTR_WO),
+};
+
+#define AMDGPU_CAP_ATTR_BITS 2
+#define AMDGPU_CAP_ATTR_MAX  ((1U << AMDGPU_CAP_ATTR_BITS) - 1)
+
+/* Internal helper to build helpers for a given enum NAME */
+#define DECLARE_ATTR_CAP_CLASS_HELPERS(NAME)							\
+enum { NAME##_BITMAP_BITS = NAME##_COUNT * AMDGPU_CAP_ATTR_BITS };				\
+struct NAME##_caps {										\
+	DECLARE_BITMAP(bmap, NAME##_BITMAP_BITS);						\
+};												\
+static inline unsigned int NAME##_ATTR_START(enum NAME##_cap_id cap)				\
+{ return (unsigned int)cap * AMDGPU_CAP_ATTR_BITS; }						\
+static inline void NAME##_attr_init(struct NAME##_caps *c)					\
+{ if (c) bitmap_zero(c->bmap, NAME##_BITMAP_BITS); }						\
+static inline int NAME##_attr_set(struct NAME##_caps *c,					\
+				  enum NAME##_cap_id cap, enum amdgpu_cap_attr attr)		\
+{												\
+	if (!c)											\
+		return -EINVAL;									\
+	if (cap >= NAME##_COUNT)								\
+		return -EINVAL;									\
+	if ((unsigned int)attr > AMDGPU_CAP_ATTR_MAX)						\
+		return -EINVAL;									\
+	bitmap_write(c->bmap, (unsigned long)attr,						\
+			NAME##_ATTR_START(cap), AMDGPU_CAP_ATTR_BITS);				\
+	return 0;										\
+}												\
+static inline int NAME##_attr_get(const struct NAME##_caps *c,					\
+				  enum NAME##_cap_id cap, enum amdgpu_cap_attr *out)		\
+{												\
+	unsigned long v;									\
+	if (!c || !out)										\
+		return -EINVAL;									\
+	if (cap >= NAME##_COUNT)								\
+		return -EINVAL;									\
+	v = bitmap_read(c->bmap, NAME##_ATTR_START(cap), AMDGPU_CAP_ATTR_BITS);			\
+	*out = (enum amdgpu_cap_attr)v;								\
+	return 0;										\
+}												\
+static inline bool NAME##_cap_is_ro(const struct NAME##_caps *c, enum NAME##_cap_id id)		\
+{ enum amdgpu_cap_attr a; return !NAME##_attr_get(c, id, &a) && a == AMDGPU_CAP_ATTR_RO; }	\
+static inline bool NAME##_cap_is_wo(const struct NAME##_caps *c, enum NAME##_cap_id id)		\
+{ enum amdgpu_cap_attr a; return !NAME##_attr_get(c, id, &a) && a == AMDGPU_CAP_ATTR_WO; }	\
+static inline bool NAME##_cap_is_rw(const struct NAME##_caps *c, enum NAME##_cap_id id)		\
+{ enum amdgpu_cap_attr a; return !NAME##_attr_get(c, id, &a) && a == AMDGPU_CAP_ATTR_RW; }
+
+/* Element expander for enum creation */
+#define _CAP_ENUM_ELEM(x) x,
+
+/* Public macro: declare enum + helpers from an X‑macro list */
+#define DECLARE_ATTR_CAP_CLASS(NAME, LIST_MACRO)						\
+	enum NAME##_cap_id { LIST_MACRO(_CAP_ENUM_ELEM) NAME##_COUNT };				\
+	DECLARE_ATTR_CAP_CLASS_HELPERS(NAME)
+
+#endif /* AMDGPU_UTILS_H_ */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
index 6887109abb13..5c38f0d30c87 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
@@ -96,16 +96,16 @@
  */
 struct amdgpu_uvd_cs_ctx {
 	struct amdgpu_cs_parser *parser;
-	unsigned reg, count;
-	unsigned data0, data1;
-	unsigned idx;
+	unsigned int reg, count;
+	unsigned int data0, data1;
+	unsigned int idx;
 	struct amdgpu_ib *ib;
 
 	/* does the IB has a msg command */
 	bool has_msg_cmd;
 
 	/* minimum buffer sizes */
-	unsigned *buf_sizes;
+	unsigned int *buf_sizes;
 };
 
 #ifdef CONFIG_DRM_AMDGPU_SI
@@ -186,7 +186,7 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev)
 	unsigned long bo_size;
 	const char *fw_name;
 	const struct common_firmware_header *hdr;
-	unsigned family_id;
+	unsigned int family_id;
 	int i, j, r;
 
 	INIT_DELAYED_WORK(&adev->uvd.idle_work, amdgpu_uvd_idle_work_handler);
@@ -260,7 +260,7 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev)
 		return -EINVAL;
 	}
 
-	r = amdgpu_ucode_request(adev, &adev->uvd.fw, fw_name);
+	r = amdgpu_ucode_request(adev, &adev->uvd.fw, AMDGPU_UCODE_REQUIRED, "%s", fw_name);
 	if (r) {
 		dev_err(adev->dev, "amdgpu_uvd: Can't validate firmware \"%s\"\n",
 			fw_name);
@@ -275,7 +275,7 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev)
 	family_id = le32_to_cpu(hdr->ucode_version) & 0xff;
 
 	if (adev->asic_type < CHIP_VEGA20) {
-		unsigned version_major, version_minor;
+		unsigned int version_major, version_minor;
 
 		version_major = (le32_to_cpu(hdr->ucode_version) >> 24) & 0xff;
 		version_minor = (le32_to_cpu(hdr->ucode_version) >> 8) & 0xff;
@@ -398,32 +398,32 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev)
  * amdgpu_uvd_entity_init - init entity
  *
  * @adev: amdgpu_device pointer
+ * @ring: amdgpu_ring pointer to check
  *
+ * Initialize the entity used for handle management in the kernel driver.
  */
-int amdgpu_uvd_entity_init(struct amdgpu_device *adev)
+int amdgpu_uvd_entity_init(struct amdgpu_device *adev, struct amdgpu_ring *ring)
 {
-	struct amdgpu_ring *ring;
-	struct drm_gpu_scheduler *sched;
-	int r;
+	if (ring == &adev->uvd.inst[0].ring) {
+		struct drm_gpu_scheduler *sched = &ring->sched;
+		int r;
 
-	ring = &adev->uvd.inst[0].ring;
-	sched = &ring->sched;
-	r = drm_sched_entity_init(&adev->uvd.entity, DRM_SCHED_PRIORITY_NORMAL,
-				  &sched, 1, NULL);
-	if (r) {
-		DRM_ERROR("Failed setting up UVD kernel entity.\n");
-		return r;
+		r = drm_sched_entity_init(&adev->uvd.entity, DRM_SCHED_PRIORITY_NORMAL,
+					  &sched, 1, NULL);
+		if (r) {
+			DRM_ERROR("Failed setting up UVD kernel entity.\n");
+			return r;
+		}
 	}
 
 	return 0;
 }
 
-int amdgpu_uvd_suspend(struct amdgpu_device *adev)
+int amdgpu_uvd_prepare_suspend(struct amdgpu_device *adev)
 {
-	unsigned size;
+	unsigned int size;
 	void *ptr;
 	int i, j, idx;
-	bool in_ras_intr = amdgpu_ras_intr_triggered();
 
 	cancel_delayed_work_sync(&adev->uvd.idle_work);
 
@@ -452,7 +452,7 @@ int amdgpu_uvd_suspend(struct amdgpu_device *adev)
 
 		if (drm_dev_enter(adev_to_drm(adev), &idx)) {
 			/* re-write 0 since err_event_athub will corrupt VCPU buffer */
-			if (in_ras_intr)
+			if (amdgpu_ras_intr_triggered())
 				memset(adev->uvd.inst[j].saved_bo, 0, size);
 			else
 				memcpy_fromio(adev->uvd.inst[j].saved_bo, ptr, size);
@@ -461,7 +461,12 @@ int amdgpu_uvd_suspend(struct amdgpu_device *adev)
 		}
 	}
 
-	if (in_ras_intr)
+	return 0;
+}
+
+int amdgpu_uvd_suspend(struct amdgpu_device *adev)
+{
+	if (amdgpu_ras_intr_triggered())
 		DRM_WARN("UVD VCPU state may lost due to RAS ERREVENT_ATHUB_INTERRUPT\n");
 
 	return 0;
@@ -469,7 +474,7 @@ int amdgpu_uvd_suspend(struct amdgpu_device *adev)
 
 int amdgpu_uvd_resume(struct amdgpu_device *adev)
 {
-	unsigned size;
+	unsigned int size;
 	void *ptr;
 	int i, idx;
 
@@ -491,7 +496,7 @@ int amdgpu_uvd_resume(struct amdgpu_device *adev)
 			adev->uvd.inst[i].saved_bo = NULL;
 		} else {
 			const struct common_firmware_header *hdr;
-			unsigned offset;
+			unsigned int offset;
 
 			hdr = (const struct common_firmware_header *)adev->uvd.fw->data;
 			if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
@@ -542,9 +547,12 @@ void amdgpu_uvd_free_handles(struct amdgpu_device *adev, struct drm_file *filp)
 static void amdgpu_uvd_force_into_uvd_segment(struct amdgpu_bo *abo)
 {
 	int i;
+
 	for (i = 0; i < abo->placement.num_placement; ++i) {
 		abo->placements[i].fpfn = 0 >> PAGE_SHIFT;
 		abo->placements[i].lpfn = (256 * 1024 * 1024) >> PAGE_SHIFT;
+		if (abo->placements[i].mem_type == TTM_PL_VRAM)
+			abo->placements[i].flags |= TTM_PL_FLAG_CONTIGUOUS;
 	}
 }
 
@@ -579,7 +587,7 @@ static int amdgpu_uvd_cs_pass1(struct amdgpu_uvd_cs_ctx *ctx)
 
 	r = amdgpu_cs_find_mapping(ctx->parser, addr, &bo, &mapping);
 	if (r) {
-		DRM_ERROR("Can't find BO for addr 0x%08Lx\n", addr);
+		DRM_ERROR("Can't find BO for addr 0x%08llx\n", addr);
 		return r;
 	}
 
@@ -589,6 +597,7 @@ static int amdgpu_uvd_cs_pass1(struct amdgpu_uvd_cs_ctx *ctx)
 		if (cmd == 0x0 || cmd == 0x3) {
 			/* yes, force it into VRAM */
 			uint32_t domain = AMDGPU_GEM_DOMAIN_VRAM;
+
 			amdgpu_bo_placement_from_domain(bo, domain);
 		}
 		amdgpu_uvd_force_into_uvd_segment(bo);
@@ -609,21 +618,21 @@ static int amdgpu_uvd_cs_pass1(struct amdgpu_uvd_cs_ctx *ctx)
  * Peek into the decode message and calculate the necessary buffer sizes.
  */
 static int amdgpu_uvd_cs_msg_decode(struct amdgpu_device *adev, uint32_t *msg,
-	unsigned buf_sizes[])
+	unsigned int buf_sizes[])
 {
-	unsigned stream_type = msg[4];
-	unsigned width = msg[6];
-	unsigned height = msg[7];
-	unsigned dpb_size = msg[9];
-	unsigned pitch = msg[28];
-	unsigned level = msg[57];
+	unsigned int stream_type = msg[4];
+	unsigned int width = msg[6];
+	unsigned int height = msg[7];
+	unsigned int dpb_size = msg[9];
+	unsigned int pitch = msg[28];
+	unsigned int level = msg[57];
 
-	unsigned width_in_mb = width / 16;
-	unsigned height_in_mb = ALIGN(height / 16, 2);
-	unsigned fs_in_mb = width_in_mb * height_in_mb;
+	unsigned int width_in_mb = width / 16;
+	unsigned int height_in_mb = ALIGN(height / 16, 2);
+	unsigned int fs_in_mb = width_in_mb * height_in_mb;
 
-	unsigned image_size, tmp, min_dpb_size, num_dpb_buffer;
-	unsigned min_ctx_size = ~0;
+	unsigned int image_size, tmp, min_dpb_size, num_dpb_buffer;
+	unsigned int min_ctx_size = ~0;
 
 	image_size = width * height;
 	image_size += image_size / 2;
@@ -631,7 +640,7 @@ static int amdgpu_uvd_cs_msg_decode(struct amdgpu_device *adev, uint32_t *msg,
 
 	switch (stream_type) {
 	case 0: /* H264 */
-		switch(level) {
+		switch (level) {
 		case 30:
 			num_dpb_buffer = 8100 / fs_in_mb;
 			break;
@@ -709,7 +718,7 @@ static int amdgpu_uvd_cs_msg_decode(struct amdgpu_device *adev, uint32_t *msg,
 		break;
 
 	case 7: /* H264 Perf */
-		switch(level) {
+		switch (level) {
 		case 30:
 			num_dpb_buffer = 8100 / fs_in_mb;
 			break;
@@ -742,7 +751,7 @@ static int amdgpu_uvd_cs_msg_decode(struct amdgpu_device *adev, uint32_t *msg,
 		/* reference picture buffer */
 		min_dpb_size = image_size * num_dpb_buffer;
 
-		if (!adev->uvd.use_ctx_buf){
+		if (!adev->uvd.use_ctx_buf) {
 			/* macroblock context buffer */
 			min_dpb_size +=
 				width_in_mb * height_in_mb * num_dpb_buffer * 192;
@@ -805,7 +814,7 @@ static int amdgpu_uvd_cs_msg_decode(struct amdgpu_device *adev, uint32_t *msg,
  * Make sure that we don't open up to many sessions.
  */
 static int amdgpu_uvd_cs_msg(struct amdgpu_uvd_cs_ctx *ctx,
-			     struct amdgpu_bo *bo, unsigned offset)
+			     struct amdgpu_bo *bo, unsigned int offset)
 {
 	struct amdgpu_device *adev = ctx->parser->adev;
 	int32_t *msg, msg_type, handle;
@@ -911,7 +920,7 @@ static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx)
 
 	r = amdgpu_cs_find_mapping(ctx->parser, addr, &bo, &mapping);
 	if (r) {
-		DRM_ERROR("Can't find BO for addr 0x%08Lx\n", addr);
+		DRM_ERROR("Can't find BO for addr 0x%08llx\n", addr);
 		return r;
 	}
 
@@ -930,7 +939,7 @@ static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx)
 	if (cmd < 0x4) {
 		if ((end - start) < ctx->buf_sizes[cmd]) {
 			DRM_ERROR("buffer (%d) to small (%d / %d)!\n", cmd,
-				  (unsigned)(end - start),
+				  (unsigned int)(end - start),
 				  ctx->buf_sizes[cmd]);
 			return -EINVAL;
 		}
@@ -938,7 +947,7 @@ static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx)
 	} else if (cmd == 0x206) {
 		if ((end - start) < ctx->buf_sizes[4]) {
 			DRM_ERROR("buffer (%d) to small (%d / %d)!\n", cmd,
-					  (unsigned)(end - start),
+					  (unsigned int)(end - start),
 					  ctx->buf_sizes[4]);
 			return -EINVAL;
 		}
@@ -949,14 +958,14 @@ static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx)
 
 	if (!ctx->parser->adev->uvd.address_64_bit) {
 		if ((start >> 28) != ((end - 1) >> 28)) {
-			DRM_ERROR("reloc %LX-%LX crossing 256MB boundary!\n",
+			DRM_ERROR("reloc %llx-%llx crossing 256MB boundary!\n",
 				  start, end);
 			return -EINVAL;
 		}
 
 		if ((cmd == 0 || cmd == 0x3) &&
 		    (start >> 28) != (ctx->parser->adev->uvd.inst->gpu_addr >> 28)) {
-			DRM_ERROR("msg/fb buffer %LX-%LX out of 256MB segment!\n",
+			DRM_ERROR("msg/fb buffer %llx-%llx out of 256MB segment!\n",
 				  start, end);
 			return -EINVAL;
 		}
@@ -990,7 +999,7 @@ static int amdgpu_uvd_cs_reg(struct amdgpu_uvd_cs_ctx *ctx,
 
 	ctx->idx++;
 	for (i = 0; i <= ctx->count; ++i) {
-		unsigned reg = ctx->reg + i;
+		unsigned int reg = ctx->reg + i;
 
 		if (ctx->idx >= ctx->ib->length_dw) {
 			DRM_ERROR("Register command after end of CS!\n");
@@ -1036,7 +1045,8 @@ static int amdgpu_uvd_cs_packets(struct amdgpu_uvd_cs_ctx *ctx,
 
 	for (ctx->idx = 0 ; ctx->idx < ctx->ib->length_dw; ) {
 		uint32_t cmd = amdgpu_ib_get_value(ctx->ib, ctx->idx);
-		unsigned type = CP_PACKET_GET_TYPE(cmd);
+		unsigned int type = CP_PACKET_GET_TYPE(cmd);
+
 		switch (type) {
 		case PACKET_TYPE0:
 			ctx->reg = CP_PACKET0_GET_REG(cmd);
@@ -1070,7 +1080,7 @@ int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser,
 			     struct amdgpu_ib *ib)
 {
 	struct amdgpu_uvd_cs_ctx ctx = {};
-	unsigned buf_sizes[] = {
+	unsigned int buf_sizes[] = {
 		[0x00000000]	=	2048,
 		[0x00000001]	=	0xFFFFFFFF,
 		[0x00000002]	=	0xFFFFFFFF,
@@ -1080,7 +1090,6 @@ int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser,
 	int r;
 
 	job->vm = NULL;
-	ib->gpu_addr = amdgpu_sa_bo_gpu_addr(ib->sa_bo);
 
 	if (ib->length_dw % 16) {
 		DRM_ERROR("UVD IB length (%d) not 16 dwords aligned!\n",
@@ -1127,7 +1136,8 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo,
 	r = amdgpu_job_alloc_with_ib(ring->adev, &adev->uvd.entity,
 				     AMDGPU_FENCE_OWNER_UNDEFINED,
 				     64, direct ? AMDGPU_IB_POOL_DIRECT :
-				     AMDGPU_IB_POOL_DELAYED, &job);
+				     AMDGPU_IB_POOL_DELAYED, &job,
+				     AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		return r;
 
@@ -1185,8 +1195,9 @@ err_free:
 }
 
 /* multiple fence commands without any stream commands in between can
-   crash the vcpu so just try to emmit a dummy create/destroy msg to
-   avoid this */
+ * crash the vcpu so just try to emmit a dummy create/destroy msg to
+ * avoid this
+ */
 int amdgpu_uvd_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
 			      struct dma_fence **fence)
 {
@@ -1252,15 +1263,14 @@ static void amdgpu_uvd_idle_work_handler(struct work_struct *work)
 {
 	struct amdgpu_device *adev =
 		container_of(work, struct amdgpu_device, uvd.idle_work.work);
-	unsigned fences = 0, i, j;
+	unsigned int fences = 0, i, j;
 
 	for (i = 0; i < adev->uvd.num_uvd_inst; ++i) {
 		if (adev->uvd.harvest_config & (1 << i))
 			continue;
 		fences += amdgpu_fence_count_emitted(&adev->uvd.inst[i].ring);
-		for (j = 0; j < adev->uvd.num_enc_rings; ++j) {
+		for (j = 0; j < adev->uvd.num_enc_rings; ++j)
 			fences += amdgpu_fence_count_emitted(&adev->uvd.inst[i].ring_enc[j]);
-		}
 	}
 
 	if (fences == 0) {
@@ -1356,7 +1366,7 @@ error:
  */
 uint32_t amdgpu_uvd_used_handles(struct amdgpu_device *adev)
 {
-	unsigned i;
+	unsigned int i;
 	uint32_t used_handles = 0;
 
 	for (i = 0; i < adev->uvd.max_handles; ++i) {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h
index 9f89bb7cd60b..9dfad2f48ef4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h
@@ -73,7 +73,8 @@ struct amdgpu_uvd {
 
 int amdgpu_uvd_sw_init(struct amdgpu_device *adev);
 int amdgpu_uvd_sw_fini(struct amdgpu_device *adev);
-int amdgpu_uvd_entity_init(struct amdgpu_device *adev);
+int amdgpu_uvd_entity_init(struct amdgpu_device *adev, struct amdgpu_ring *ring);
+int amdgpu_uvd_prepare_suspend(struct amdgpu_device *adev);
 int amdgpu_uvd_suspend(struct amdgpu_device *adev);
 int amdgpu_uvd_resume(struct amdgpu_device *adev);
 int amdgpu_uvd_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
index e2b7324a70cb..a7d8f1ce6ac2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
@@ -41,6 +41,9 @@
 #define VCE_IDLE_TIMEOUT	msecs_to_jiffies(1000)
 
 /* Firmware Names */
+#ifdef CONFIG_DRM_AMDGPU_SI
+#define FIRMWARE_VCE_V1_0	"amdgpu/vce_1_0_0.bin"
+#endif
 #ifdef CONFIG_DRM_AMDGPU_CIK
 #define FIRMWARE_BONAIRE	"amdgpu/bonaire_vce.bin"
 #define FIRMWARE_KABINI	"amdgpu/kabini_vce.bin"
@@ -61,6 +64,9 @@
 #define FIRMWARE_VEGA12		"amdgpu/vega12_vce.bin"
 #define FIRMWARE_VEGA20		"amdgpu/vega20_vce.bin"
 
+#ifdef CONFIG_DRM_AMDGPU_SI
+MODULE_FIRMWARE(FIRMWARE_VCE_V1_0);
+#endif
 #ifdef CONFIG_DRM_AMDGPU_CIK
 MODULE_FIRMWARE(FIRMWARE_BONAIRE);
 MODULE_FIRMWARE(FIRMWARE_KABINI);
@@ -88,82 +94,93 @@ static int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle,
 				      bool direct, struct dma_fence **fence);
 
 /**
- * amdgpu_vce_sw_init - allocate memory, load vce firmware
+ * amdgpu_vce_firmware_name() - determine the firmware file name for VCE
  *
  * @adev: amdgpu_device pointer
- * @size: size for the new BO
  *
- * First step to get VCE online, allocate memory and load the firmware
+ * Each chip that has VCE IP may need a different firmware.
+ * This function returns the name of the VCE firmware file
+ * appropriate for the current chip.
  */
-int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size)
+static const char *amdgpu_vce_firmware_name(struct amdgpu_device *adev)
 {
-	const char *fw_name;
-	const struct common_firmware_header *hdr;
-	unsigned ucode_version, version_major, version_minor, binary_id;
-	int i, r;
-
 	switch (adev->asic_type) {
+#ifdef CONFIG_DRM_AMDGPU_SI
+	case CHIP_PITCAIRN:
+	case CHIP_TAHITI:
+	case CHIP_VERDE:
+		return FIRMWARE_VCE_V1_0;
+#endif
 #ifdef CONFIG_DRM_AMDGPU_CIK
 	case CHIP_BONAIRE:
-		fw_name = FIRMWARE_BONAIRE;
-		break;
+		return FIRMWARE_BONAIRE;
 	case CHIP_KAVERI:
-		fw_name = FIRMWARE_KAVERI;
-		break;
+		return FIRMWARE_KAVERI;
 	case CHIP_KABINI:
-		fw_name = FIRMWARE_KABINI;
-		break;
+		return FIRMWARE_KABINI;
 	case CHIP_HAWAII:
-		fw_name = FIRMWARE_HAWAII;
-		break;
+		return FIRMWARE_HAWAII;
 	case CHIP_MULLINS:
-		fw_name = FIRMWARE_MULLINS;
-		break;
+		return FIRMWARE_MULLINS;
 #endif
 	case CHIP_TONGA:
-		fw_name = FIRMWARE_TONGA;
-		break;
+		return  FIRMWARE_TONGA;
 	case CHIP_CARRIZO:
-		fw_name = FIRMWARE_CARRIZO;
-		break;
+		return  FIRMWARE_CARRIZO;
 	case CHIP_FIJI:
-		fw_name = FIRMWARE_FIJI;
-		break;
+		return  FIRMWARE_FIJI;
 	case CHIP_STONEY:
-		fw_name = FIRMWARE_STONEY;
-		break;
+		return  FIRMWARE_STONEY;
 	case CHIP_POLARIS10:
-		fw_name = FIRMWARE_POLARIS10;
-		break;
+		return  FIRMWARE_POLARIS10;
 	case CHIP_POLARIS11:
-		fw_name = FIRMWARE_POLARIS11;
-		break;
+		return  FIRMWARE_POLARIS11;
 	case CHIP_POLARIS12:
-		fw_name = FIRMWARE_POLARIS12;
-		break;
+		return  FIRMWARE_POLARIS12;
 	case CHIP_VEGAM:
-		fw_name = FIRMWARE_VEGAM;
-		break;
+		return  FIRMWARE_VEGAM;
 	case CHIP_VEGA10:
-		fw_name = FIRMWARE_VEGA10;
-		break;
+		return  FIRMWARE_VEGA10;
 	case CHIP_VEGA12:
-		fw_name = FIRMWARE_VEGA12;
-		break;
+		return  FIRMWARE_VEGA12;
 	case CHIP_VEGA20:
-		fw_name = FIRMWARE_VEGA20;
-		break;
+		return  FIRMWARE_VEGA20;
 
 	default:
-		return -EINVAL;
+		return NULL;
 	}
+}
 
-	r = amdgpu_ucode_request(adev, &adev->vce.fw, fw_name);
+/**
+ * amdgpu_vce_early_init() - try to load VCE firmware
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Tries to load the VCE firmware.
+ *
+ * When not found, returns ENOENT so that the driver can
+ * still load and initialize the rest of the IP blocks.
+ * The GPU can function just fine without VCE, they will just
+ * not support video encoding.
+ */
+int amdgpu_vce_early_init(struct amdgpu_device *adev)
+{
+	const char *fw_name = amdgpu_vce_firmware_name(adev);
+	const struct common_firmware_header *hdr;
+	unsigned int ucode_version, version_major, version_minor, binary_id;
+	int r;
+
+	if (!fw_name)
+		return -ENOENT;
+
+	r = amdgpu_ucode_request(adev, &adev->vce.fw, AMDGPU_UCODE_REQUIRED, "%s", fw_name);
 	if (r) {
-		dev_err(adev->dev, "amdgpu_vce: Can't validate firmware \"%s\"\n",
-			fw_name);
+		dev_err(adev->dev,
+			"amdgpu_vce: Firmware \"%s\" not found or failed to validate (%d)\n",
+			fw_name, r);
+
 		amdgpu_ucode_release(&adev->vce.fw);
-		return r;
+		return -ENOENT;
 	}
 
 	hdr = (const struct common_firmware_header *)adev->vce.fw->data;
@@ -172,11 +189,35 @@ int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size)
 	version_major = (ucode_version >> 20) & 0xfff;
 	version_minor = (ucode_version >> 8) & 0xfff;
 	binary_id = ucode_version & 0xff;
-	DRM_INFO("Found VCE firmware Version: %d.%d Binary ID: %d\n",
+	dev_info(adev->dev, "Found VCE firmware Version: %d.%d Binary ID: %d\n",
 		version_major, version_minor, binary_id);
 	adev->vce.fw_version = ((version_major << 24) | (version_minor << 16) |
 				(binary_id << 8));
 
+	return 0;
+}
+
+/**
+ * amdgpu_vce_sw_init() - allocate memory for VCE BO
+ *
+ * @adev: amdgpu_device pointer
+ * @size: size for the new BO
+ *
+ * First step to get VCE online: allocate memory for VCE BO.
+ * The VCE firmware binary is copied into the VCE BO later,
+ * in amdgpu_vce_resume. The VCE executes its code from the
+ * VCE BO and also uses the space in this BO for its stack and data.
+ *
+ * Ideally this BO should be placed in VRAM for optimal performance,
+ * although technically it also runs from system RAM (albeit slowly).
+ */
+int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size)
+{
+	int i, r;
+
+	if (!adev->vce.fw)
+		return -ENOENT;
+
 	r = amdgpu_bo_create_kernel(adev, size, PAGE_SIZE,
 				    AMDGPU_GEM_DOMAIN_VRAM |
 				    AMDGPU_GEM_DOMAIN_GTT,
@@ -207,22 +248,22 @@ int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size)
  */
 int amdgpu_vce_sw_fini(struct amdgpu_device *adev)
 {
-	unsigned i;
+	unsigned int i;
 
 	if (adev->vce.vcpu_bo == NULL)
 		return 0;
 
 	drm_sched_entity_destroy(&adev->vce.entity);
 
-	amdgpu_bo_free_kernel(&adev->vce.vcpu_bo, &adev->vce.gpu_addr,
-		(void **)&adev->vce.cpu_addr);
-
 	for (i = 0; i < adev->vce.num_rings; i++)
 		amdgpu_ring_fini(&adev->vce.ring[i]);
 
 	amdgpu_ucode_release(&adev->vce.fw);
 	mutex_destroy(&adev->vce.idle_mutex);
 
+	amdgpu_bo_free_kernel(&adev->vce.vcpu_bo, &adev->vce.gpu_addr,
+		(void **)&adev->vce.cpu_addr);
+
 	return 0;
 }
 
@@ -230,21 +271,22 @@ int amdgpu_vce_sw_fini(struct amdgpu_device *adev)
  * amdgpu_vce_entity_init - init entity
  *
  * @adev: amdgpu_device pointer
+ * @ring: amdgpu_ring pointer to check
  *
+ * Initialize the entity used for handle management in the kernel driver.
  */
-int amdgpu_vce_entity_init(struct amdgpu_device *adev)
+int amdgpu_vce_entity_init(struct amdgpu_device *adev, struct amdgpu_ring *ring)
 {
-	struct amdgpu_ring *ring;
-	struct drm_gpu_scheduler *sched;
-	int r;
-
-	ring = &adev->vce.ring[0];
-	sched = &ring->sched;
-	r = drm_sched_entity_init(&adev->vce.entity, DRM_SCHED_PRIORITY_NORMAL,
-				  &sched, 1, NULL);
-	if (r != 0) {
-		DRM_ERROR("Failed setting up VCE run queue.\n");
-		return r;
+	if (ring == &adev->vce.ring[0]) {
+		struct drm_gpu_scheduler *sched = &ring->sched;
+		int r;
+
+		r = drm_sched_entity_init(&adev->vce.entity, DRM_SCHED_PRIORITY_NORMAL,
+					  &sched, 1, NULL);
+		if (r != 0) {
+			DRM_ERROR("Failed setting up VCE run queue.\n");
+			return r;
+		}
 	}
 
 	return 0;
@@ -284,40 +326,23 @@ int amdgpu_vce_suspend(struct amdgpu_device *adev)
  */
 int amdgpu_vce_resume(struct amdgpu_device *adev)
 {
-	void *cpu_addr;
 	const struct common_firmware_header *hdr;
-	unsigned offset;
-	int r, idx;
+	unsigned int offset;
+	int idx;
 
 	if (adev->vce.vcpu_bo == NULL)
 		return -EINVAL;
 
-	r = amdgpu_bo_reserve(adev->vce.vcpu_bo, false);
-	if (r) {
-		dev_err(adev->dev, "(%d) failed to reserve VCE bo\n", r);
-		return r;
-	}
-
-	r = amdgpu_bo_kmap(adev->vce.vcpu_bo, &cpu_addr);
-	if (r) {
-		amdgpu_bo_unreserve(adev->vce.vcpu_bo);
-		dev_err(adev->dev, "(%d) VCE map failed\n", r);
-		return r;
-	}
-
 	hdr = (const struct common_firmware_header *)adev->vce.fw->data;
 	offset = le32_to_cpu(hdr->ucode_array_offset_bytes);
 
 	if (drm_dev_enter(adev_to_drm(adev), &idx)) {
-		memcpy_toio(cpu_addr, adev->vce.fw->data + offset,
+		memset_io(adev->vce.cpu_addr, 0, amdgpu_bo_size(adev->vce.vcpu_bo));
+		memcpy_toio(adev->vce.cpu_addr, adev->vce.fw->data + offset,
 			    adev->vce.fw->size - offset);
 		drm_dev_exit(idx);
 	}
 
-	amdgpu_bo_kunmap(adev->vce.vcpu_bo);
-
-	amdgpu_bo_unreserve(adev->vce.vcpu_bo);
-
 	return 0;
 }
 
@@ -332,7 +357,7 @@ static void amdgpu_vce_idle_work_handler(struct work_struct *work)
 {
 	struct amdgpu_device *adev =
 		container_of(work, struct amdgpu_device, vce.idle_work.work);
-	unsigned i, count = 0;
+	unsigned int i, count = 0;
 
 	for (i = 0; i < adev->vce.num_rings; i++)
 		count += amdgpu_fence_count_emitted(&adev->vce.ring[i]);
@@ -409,6 +434,7 @@ void amdgpu_vce_free_handles(struct amdgpu_device *adev, struct drm_file *filp)
 {
 	struct amdgpu_ring *ring = &adev->vce.ring[0];
 	int i, r;
+
 	for (i = 0; i < AMDGPU_MAX_VCE_HANDLES; ++i) {
 		uint32_t handle = atomic_read(&adev->vce.handles[i]);
 
@@ -425,6 +451,24 @@ void amdgpu_vce_free_handles(struct amdgpu_device *adev, struct drm_file *filp)
 }
 
 /**
+ * amdgpu_vce_required_gart_pages() - gets number of GART pages required by VCE
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Returns how many GART pages we need before GTT for the VCE IP block.
+ * For VCE1, see vce_v1_0_ensure_vcpu_bo_32bit_addr for details.
+ * For VCE2+, this is not needed so return zero.
+ */
+u32 amdgpu_vce_required_gart_pages(struct amdgpu_device *adev)
+{
+	/* VCE IP block not added yet, so can't use amdgpu_ip_version */
+	if (adev->family == AMDGPU_FAMILY_SI)
+		return 512;
+
+	return 0;
+}
+
+/**
  * amdgpu_vce_get_create_msg - generate a VCE create msg
  *
  * @ring: ring we should submit the msg to
@@ -436,7 +480,7 @@ void amdgpu_vce_free_handles(struct amdgpu_device *adev, struct drm_file *filp)
 static int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
 				     struct dma_fence **fence)
 {
-	const unsigned ib_size_dw = 1024;
+	const unsigned int ib_size_dw = 1024;
 	struct amdgpu_job *job;
 	struct amdgpu_ib *ib;
 	struct amdgpu_ib ib_msg;
@@ -447,7 +491,7 @@ static int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
 	r = amdgpu_job_alloc_with_ib(ring->adev, &ring->adev->vce.entity,
 				     AMDGPU_FENCE_OWNER_UNDEFINED,
 				     ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT,
-				     &job);
+				     &job, AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		return r;
 
@@ -501,7 +545,7 @@ static int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
 		ib->ptr[i] = 0x0;
 
 	r = amdgpu_job_submit_direct(job, ring, &f);
-	amdgpu_ib_free(ring->adev, &ib_msg, f);
+	amdgpu_ib_free(&ib_msg, f);
 	if (r)
 		goto err;
 
@@ -528,7 +572,7 @@ err:
 static int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle,
 				      bool direct, struct dma_fence **fence)
 {
-	const unsigned ib_size_dw = 1024;
+	const unsigned int ib_size_dw = 1024;
 	struct amdgpu_job *job;
 	struct amdgpu_ib *ib;
 	struct dma_fence *f = NULL;
@@ -538,7 +582,8 @@ static int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle,
 				     AMDGPU_FENCE_OWNER_UNDEFINED,
 				     ib_size_dw * 4,
 				     direct ? AMDGPU_IB_POOL_DIRECT :
-				     AMDGPU_IB_POOL_DELAYED, &job);
+				     AMDGPU_IB_POOL_DELAYED, &job,
+				     AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		return r;
 
@@ -596,12 +641,12 @@ err:
  */
 static int amdgpu_vce_validate_bo(struct amdgpu_cs_parser *p,
 				  struct amdgpu_ib *ib, int lo, int hi,
-				  unsigned size, int32_t index)
+				  unsigned int size, int32_t index)
 {
 	int64_t offset = ((uint64_t)size) * ((int64_t)index);
 	struct ttm_operation_ctx ctx = { false, false };
 	struct amdgpu_bo_va_mapping *mapping;
-	unsigned i, fpfn, lpfn;
+	unsigned int i, fpfn, lpfn;
 	struct amdgpu_bo *bo;
 	uint64_t addr;
 	int r;
@@ -619,7 +664,7 @@ static int amdgpu_vce_validate_bo(struct amdgpu_cs_parser *p,
 
 	r = amdgpu_cs_find_mapping(p, addr, &bo, &mapping);
 	if (r) {
-		DRM_ERROR("Can't find BO for addr 0x%010Lx %d %d %d %d\n",
+		DRM_ERROR("Can't find BO for addr 0x%010llx %d %d %d %d\n",
 			  addr, lo, hi, size, index);
 		return r;
 	}
@@ -646,7 +691,7 @@ static int amdgpu_vce_validate_bo(struct amdgpu_cs_parser *p,
  * Patch relocation inside command stream with real buffer address
  */
 static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, struct amdgpu_ib *ib,
-			       int lo, int hi, unsigned size, uint32_t index)
+			       int lo, int hi, unsigned int size, uint32_t index)
 {
 	struct amdgpu_bo_va_mapping *mapping;
 	struct amdgpu_bo *bo;
@@ -662,14 +707,14 @@ static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, struct amdgpu_ib *ib,
 
 	r = amdgpu_cs_find_mapping(p, addr, &bo, &mapping);
 	if (r) {
-		DRM_ERROR("Can't find BO for addr 0x%010Lx %d %d %d %d\n",
+		DRM_ERROR("Can't find BO for addr 0x%010llx %d %d %d %d\n",
 			  addr, lo, hi, size, index);
 		return r;
 	}
 
 	if ((addr + (uint64_t)size) >
 	    (mapping->last + 1) * AMDGPU_GPU_PAGE_SIZE) {
-		DRM_ERROR("BO too small for addr 0x%010Lx %d %d\n",
+		DRM_ERROR("BO too small for addr 0x%010llx %d %d\n",
 			  addr, lo, hi);
 		return -EINVAL;
 	}
@@ -692,12 +737,12 @@ static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, struct amdgpu_ib *ib,
  * @allocated: allocated a new handle?
  *
  * Validates the handle and return the found session index or -EINVAL
- * we we don't have another free session index.
+ * we don't have another free session index.
  */
 static int amdgpu_vce_validate_handle(struct amdgpu_cs_parser *p,
 				      uint32_t handle, uint32_t *allocated)
 {
-	unsigned i;
+	unsigned int i;
 
 	/* validate the handle */
 	for (i = 0; i < AMDGPU_MAX_VCE_HANDLES; ++i) {
@@ -735,18 +780,18 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p,
 			     struct amdgpu_job *job,
 			     struct amdgpu_ib *ib)
 {
-	unsigned fb_idx = 0, bs_idx = 0;
+	unsigned int fb_idx = 0, bs_idx = 0;
 	int session_idx = -1;
 	uint32_t destroyed = 0;
 	uint32_t created = 0;
 	uint32_t allocated = 0;
 	uint32_t tmp, handle = 0;
-	uint32_t *size = &tmp;
-	unsigned idx;
+	uint32_t dummy = 0xffffffff;
+	uint32_t *size = &dummy;
+	unsigned int idx;
 	int i, r = 0;
 
 	job->vm = NULL;
-	ib->gpu_addr = amdgpu_sa_bo_gpu_addr(ib->sa_bo);
 
 	for (idx = 0; idx < ib->length_dw;) {
 		uint32_t len = amdgpu_ib_get_value(ib, idx);
@@ -1041,7 +1086,6 @@ out:
 	if (!r) {
 		/* No error, free all destroyed handle slots */
 		tmp = destroyed;
-		amdgpu_ib_free(p->adev, ib, NULL);
 	} else {
 		/* Error during parsing, free all allocated handle slots */
 		tmp = allocated;
@@ -1084,7 +1128,7 @@ void amdgpu_vce_ring_emit_ib(struct amdgpu_ring *ring,
  *
  */
 void amdgpu_vce_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
-				unsigned flags)
+				unsigned int flags)
 {
 	WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
 
@@ -1106,7 +1150,7 @@ int amdgpu_vce_ring_test_ring(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 	uint32_t rptr;
-	unsigned i;
+	unsigned int i;
 	int r, timeout = adev->usec_timeout;
 
 	/* skip ring test for sriov*/
@@ -1171,7 +1215,7 @@ error:
 
 enum amdgpu_ring_priority_level amdgpu_vce_get_ring_prio(int ring)
 {
-	switch(ring) {
+	switch (ring) {
 	case 0:
 		return AMDGPU_RING_PRIO_0;
 	case 1:
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h
index ea680fc9a6c3..1c3464ce5037 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h
@@ -51,14 +51,17 @@ struct amdgpu_vce {
 	struct drm_sched_entity	entity;
 	uint32_t                srbm_soft_reset;
 	unsigned		num_rings;
+	uint32_t		keyselect;
 };
 
+int amdgpu_vce_early_init(struct amdgpu_device *adev);
 int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size);
 int amdgpu_vce_sw_fini(struct amdgpu_device *adev);
-int amdgpu_vce_entity_init(struct amdgpu_device *adev);
+int amdgpu_vce_entity_init(struct amdgpu_device *adev, struct amdgpu_ring *ring);
 int amdgpu_vce_suspend(struct amdgpu_device *adev);
 int amdgpu_vce_resume(struct amdgpu_device *adev);
 void amdgpu_vce_free_handles(struct amdgpu_device *adev, struct drm_file *filp);
+u32 amdgpu_vce_required_gart_pages(struct amdgpu_device *adev);
 int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, struct amdgpu_job *job,
 			     struct amdgpu_ib *ib);
 int amdgpu_vce_ring_parse_cs_vm(struct amdgpu_cs_parser *p,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
index e63fcc58e8e0..5e0786ea911b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2016 Advanced Micro Devices, Inc.
+ * Copyright 2016-2024 Advanced Micro Devices, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -56,7 +56,13 @@
 #define FIRMWARE_VCN_3_1_2		"amdgpu/vcn_3_1_2.bin"
 #define FIRMWARE_VCN4_0_0		"amdgpu/vcn_4_0_0.bin"
 #define FIRMWARE_VCN4_0_2		"amdgpu/vcn_4_0_2.bin"
+#define FIRMWARE_VCN4_0_3		"amdgpu/vcn_4_0_3.bin"
 #define FIRMWARE_VCN4_0_4		"amdgpu/vcn_4_0_4.bin"
+#define FIRMWARE_VCN4_0_5		"amdgpu/vcn_4_0_5.bin"
+#define FIRMWARE_VCN4_0_6		"amdgpu/vcn_4_0_6.bin"
+#define FIRMWARE_VCN4_0_6_1		"amdgpu/vcn_4_0_6_1.bin"
+#define FIRMWARE_VCN5_0_0		"amdgpu/vcn_5_0_0.bin"
+#define FIRMWARE_VCN5_0_1		"amdgpu/vcn_5_0_1.bin"
 
 MODULE_FIRMWARE(FIRMWARE_RAVEN);
 MODULE_FIRMWARE(FIRMWARE_PICASSO);
@@ -77,43 +83,65 @@ MODULE_FIRMWARE(FIRMWARE_YELLOW_CARP);
 MODULE_FIRMWARE(FIRMWARE_VCN_3_1_2);
 MODULE_FIRMWARE(FIRMWARE_VCN4_0_0);
 MODULE_FIRMWARE(FIRMWARE_VCN4_0_2);
+MODULE_FIRMWARE(FIRMWARE_VCN4_0_3);
 MODULE_FIRMWARE(FIRMWARE_VCN4_0_4);
+MODULE_FIRMWARE(FIRMWARE_VCN4_0_5);
+MODULE_FIRMWARE(FIRMWARE_VCN4_0_6);
+MODULE_FIRMWARE(FIRMWARE_VCN4_0_6_1);
+MODULE_FIRMWARE(FIRMWARE_VCN5_0_0);
+MODULE_FIRMWARE(FIRMWARE_VCN5_0_1);
 
 static void amdgpu_vcn_idle_work_handler(struct work_struct *work);
+static void amdgpu_vcn_reg_dump_fini(struct amdgpu_device *adev);
 
-int amdgpu_vcn_early_init(struct amdgpu_device *adev)
+int amdgpu_vcn_early_init(struct amdgpu_device *adev, int i)
 {
-	char ucode_prefix[30];
-	char fw_name[40];
+	char ucode_prefix[25];
 	int r;
 
+	adev->vcn.inst[i].adev = adev;
+	adev->vcn.inst[i].inst = i;
 	amdgpu_ucode_ip_version_decode(adev, UVD_HWIP, ucode_prefix, sizeof(ucode_prefix));
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s.bin", ucode_prefix);
-	r = amdgpu_ucode_request(adev, &adev->vcn.fw, fw_name);
-	if (r)
-		amdgpu_ucode_release(&adev->vcn.fw);
+
+	if (i != 0 && adev->vcn.per_inst_fw) {
+		r = amdgpu_ucode_request(adev, &adev->vcn.inst[i].fw,
+					 AMDGPU_UCODE_REQUIRED,
+					 "amdgpu/%s_%d.bin", ucode_prefix, i);
+		if (r)
+			amdgpu_ucode_release(&adev->vcn.inst[i].fw);
+	} else {
+		if (!adev->vcn.inst[0].fw) {
+			r = amdgpu_ucode_request(adev, &adev->vcn.inst[0].fw,
+						 AMDGPU_UCODE_REQUIRED,
+						 "amdgpu/%s.bin", ucode_prefix);
+			if (r)
+				amdgpu_ucode_release(&adev->vcn.inst[0].fw);
+		} else {
+			r = 0;
+		}
+		adev->vcn.inst[i].fw = adev->vcn.inst[0].fw;
+	}
 
 	return r;
 }
 
-int amdgpu_vcn_sw_init(struct amdgpu_device *adev)
+int amdgpu_vcn_sw_init(struct amdgpu_device *adev, int i)
 {
 	unsigned long bo_size;
 	const struct common_firmware_header *hdr;
 	unsigned char fw_check;
 	unsigned int fw_shared_size, log_offset;
-	int i, r;
-
-	INIT_DELAYED_WORK(&adev->vcn.idle_work, amdgpu_vcn_idle_work_handler);
-	mutex_init(&adev->vcn.vcn_pg_lock);
-	mutex_init(&adev->vcn.vcn1_jpeg1_workaround);
-	atomic_set(&adev->vcn.total_submission_cnt, 0);
-	for (i = 0; i < adev->vcn.num_vcn_inst; i++)
-		atomic_set(&adev->vcn.inst[i].dpg_enc_submission_cnt, 0);
+	int r;
 
+	mutex_init(&adev->vcn.inst[i].vcn1_jpeg1_workaround);
+	mutex_init(&adev->vcn.inst[i].vcn_pg_lock);
+	mutex_init(&adev->vcn.inst[i].engine_reset_mutex);
+	atomic_set(&adev->vcn.inst[i].total_submission_cnt, 0);
+	INIT_DELAYED_WORK(&adev->vcn.inst[i].idle_work, amdgpu_vcn_idle_work_handler);
+	atomic_set(&adev->vcn.inst[i].dpg_enc_submission_cnt, 0);
 	if ((adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) &&
 	    (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG))
-		adev->vcn.indirect_sram = true;
+		adev->vcn.inst[i].indirect_sram = true;
 
 	/*
 	 * Some Steam Deck's BIOS versions are incompatible with the
@@ -122,18 +150,23 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev)
 	 * Hence, check for these versions here - notice this is
 	 * restricted to Vangogh (Deck's APU).
 	 */
-	if (adev->ip_versions[UVD_HWIP][0] == IP_VERSION(3, 0, 2)) {
+	if (amdgpu_ip_version(adev, UVD_HWIP, 0) == IP_VERSION(3, 0, 2)) {
 		const char *bios_ver = dmi_get_system_info(DMI_BIOS_VERSION);
 
 		if (bios_ver && (!strncmp("F7A0113", bios_ver, 7) ||
-		     !strncmp("F7A0114", bios_ver, 7))) {
-			adev->vcn.indirect_sram = false;
+				 !strncmp("F7A0114", bios_ver, 7))) {
+			adev->vcn.inst[i].indirect_sram = false;
 			dev_info(adev->dev,
-				"Steam Deck quirk: indirect SRAM disabled on BIOS %s\n", bios_ver);
+				 "Steam Deck quirk: indirect SRAM disabled on BIOS %s\n", bios_ver);
 		}
 	}
 
-	hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
+	/* from vcn4 and above, only unified queue is used */
+	adev->vcn.inst[i].using_unified_queue =
+		amdgpu_ip_version(adev, UVD_HWIP, 0) >= IP_VERSION(4, 0, 0);
+
+	hdr = (const struct common_firmware_header *)adev->vcn.inst[i].fw->data;
+	adev->vcn.inst[i].fw_version = le32_to_cpu(hdr->ucode_version);
 	adev->vcn.fw_version = le32_to_cpu(hdr->ucode_version);
 
 	/* Bit 20-23, it is encode major and non-zero for new naming convention.
@@ -151,23 +184,27 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev)
 		enc_major = fw_check;
 		dec_ver = (le32_to_cpu(hdr->ucode_version) >> 24) & 0xf;
 		vep = (le32_to_cpu(hdr->ucode_version) >> 28) & 0xf;
-		DRM_INFO("Found VCN firmware Version ENC: %u.%u DEC: %u VEP: %u Revision: %u\n",
-			enc_major, enc_minor, dec_ver, vep, fw_rev);
+		dev_info(adev->dev,
+			 "[VCN instance %d] Found VCN firmware Version ENC: %u.%u DEC: %u VEP: %u Revision: %u\n",
+			 i, enc_major, enc_minor, dec_ver, vep, fw_rev);
 	} else {
 		unsigned int version_major, version_minor, family_id;
 
 		family_id = le32_to_cpu(hdr->ucode_version) & 0xff;
 		version_major = (le32_to_cpu(hdr->ucode_version) >> 24) & 0xff;
 		version_minor = (le32_to_cpu(hdr->ucode_version) >> 8) & 0xff;
-		DRM_INFO("Found VCN firmware Version: %u.%u Family ID: %u\n",
-			version_major, version_minor, family_id);
+		dev_info(adev->dev, "[VCN instance %d] Found VCN firmware Version: %u.%u Family ID: %u\n",
+			 i, version_major, version_minor, family_id);
 	}
 
 	bo_size = AMDGPU_VCN_STACK_SIZE + AMDGPU_VCN_CONTEXT_SIZE;
 	if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP)
 		bo_size += AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8);
 
-	if (adev->ip_versions[UVD_HWIP][0] >= IP_VERSION(4, 0, 0)){
+	if (amdgpu_ip_version(adev, UVD_HWIP, 0) >= IP_VERSION(5, 0, 0)) {
+		fw_shared_size = AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn5_fw_shared));
+		log_offset = offsetof(struct amdgpu_vcn5_fw_shared, fw_log);
+	} else if (amdgpu_ip_version(adev, UVD_HWIP, 0) >= IP_VERSION(4, 0, 0)) {
 		fw_shared_size = AMDGPU_GPU_PAGE_ALIGN(sizeof(struct amdgpu_vcn4_fw_shared));
 		log_offset = offsetof(struct amdgpu_vcn4_fw_shared, fw_log);
 	} else {
@@ -180,285 +217,345 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev)
 	if (amdgpu_vcnfw_log)
 		bo_size += AMDGPU_VCNFW_LOG_SIZE;
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
-		if (adev->vcn.harvest_config & (1 << i))
-			continue;
+	r = amdgpu_bo_create_kernel(adev, bo_size, PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_VRAM |
+				    AMDGPU_GEM_DOMAIN_GTT,
+				    &adev->vcn.inst[i].vcpu_bo,
+				    &adev->vcn.inst[i].gpu_addr,
+				    &adev->vcn.inst[i].cpu_addr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to allocate vcn bo\n", r);
+		return r;
+	}
+
+	adev->vcn.inst[i].fw_shared.cpu_addr = adev->vcn.inst[i].cpu_addr +
+		bo_size - fw_shared_size;
+	adev->vcn.inst[i].fw_shared.gpu_addr = adev->vcn.inst[i].gpu_addr +
+		bo_size - fw_shared_size;
+
+	adev->vcn.inst[i].fw_shared.mem_size = fw_shared_size;
 
-		r = amdgpu_bo_create_kernel(adev, bo_size, PAGE_SIZE,
+	if (amdgpu_vcnfw_log) {
+		adev->vcn.inst[i].fw_shared.cpu_addr -= AMDGPU_VCNFW_LOG_SIZE;
+		adev->vcn.inst[i].fw_shared.gpu_addr -= AMDGPU_VCNFW_LOG_SIZE;
+		adev->vcn.inst[i].fw_shared.log_offset = log_offset;
+	}
+
+	if (adev->vcn.inst[i].indirect_sram) {
+		r = amdgpu_bo_create_kernel(adev, 64 * 2 * 4, PAGE_SIZE,
 					    AMDGPU_GEM_DOMAIN_VRAM |
 					    AMDGPU_GEM_DOMAIN_GTT,
-					    &adev->vcn.inst[i].vcpu_bo,
-					    &adev->vcn.inst[i].gpu_addr,
-					    &adev->vcn.inst[i].cpu_addr);
+					    &adev->vcn.inst[i].dpg_sram_bo,
+					    &adev->vcn.inst[i].dpg_sram_gpu_addr,
+					    &adev->vcn.inst[i].dpg_sram_cpu_addr);
 		if (r) {
-			dev_err(adev->dev, "(%d) failed to allocate vcn bo\n", r);
+			dev_err(adev->dev, "VCN %d (%d) failed to allocate DPG bo\n", i, r);
 			return r;
 		}
-
-		adev->vcn.inst[i].fw_shared.cpu_addr = adev->vcn.inst[i].cpu_addr +
-				bo_size - fw_shared_size;
-		adev->vcn.inst[i].fw_shared.gpu_addr = adev->vcn.inst[i].gpu_addr +
-				bo_size - fw_shared_size;
-
-		adev->vcn.inst[i].fw_shared.mem_size = fw_shared_size;
-
-		if (amdgpu_vcnfw_log) {
-			adev->vcn.inst[i].fw_shared.cpu_addr -= AMDGPU_VCNFW_LOG_SIZE;
-			adev->vcn.inst[i].fw_shared.gpu_addr -= AMDGPU_VCNFW_LOG_SIZE;
-			adev->vcn.inst[i].fw_shared.log_offset = log_offset;
-		}
-
-		if (adev->vcn.indirect_sram) {
-			r = amdgpu_bo_create_kernel(adev, 64 * 2 * 4, PAGE_SIZE,
-					AMDGPU_GEM_DOMAIN_VRAM |
-					AMDGPU_GEM_DOMAIN_GTT,
-					&adev->vcn.inst[i].dpg_sram_bo,
-					&adev->vcn.inst[i].dpg_sram_gpu_addr,
-					&adev->vcn.inst[i].dpg_sram_cpu_addr);
-			if (r) {
-				dev_err(adev->dev, "VCN %d (%d) failed to allocate DPG bo\n", i, r);
-				return r;
-			}
-		}
 	}
 
 	return 0;
 }
 
-int amdgpu_vcn_sw_fini(struct amdgpu_device *adev)
+void amdgpu_vcn_sw_fini(struct amdgpu_device *adev, int i)
 {
-	int i, j;
-
-	for (j = 0; j < adev->vcn.num_vcn_inst; ++j) {
-		if (adev->vcn.harvest_config & (1 << j))
-			continue;
+	int j;
 
-		if (adev->vcn.indirect_sram) {
-			amdgpu_bo_free_kernel(&adev->vcn.inst[j].dpg_sram_bo,
-						  &adev->vcn.inst[j].dpg_sram_gpu_addr,
-						  (void **)&adev->vcn.inst[j].dpg_sram_cpu_addr);
-		}
-		kvfree(adev->vcn.inst[j].saved_bo);
+	if (adev->vcn.harvest_config & (1 << i))
+		return;
 
-		amdgpu_bo_free_kernel(&adev->vcn.inst[j].vcpu_bo,
-					  &adev->vcn.inst[j].gpu_addr,
-					  (void **)&adev->vcn.inst[j].cpu_addr);
+	amdgpu_bo_free_kernel(
+		&adev->vcn.inst[i].dpg_sram_bo,
+		&adev->vcn.inst[i].dpg_sram_gpu_addr,
+		(void **)&adev->vcn.inst[i].dpg_sram_cpu_addr);
 
-		amdgpu_ring_fini(&adev->vcn.inst[j].ring_dec);
+	kvfree(adev->vcn.inst[i].saved_bo);
 
-		for (i = 0; i < adev->vcn.num_enc_rings; ++i)
-			amdgpu_ring_fini(&adev->vcn.inst[j].ring_enc[i]);
-	}
+	amdgpu_bo_free_kernel(&adev->vcn.inst[i].vcpu_bo,
+			      &adev->vcn.inst[i].gpu_addr,
+			      (void **)&adev->vcn.inst[i].cpu_addr);
 
-	amdgpu_ucode_release(&adev->vcn.fw);
-	mutex_destroy(&adev->vcn.vcn1_jpeg1_workaround);
-	mutex_destroy(&adev->vcn.vcn_pg_lock);
+	amdgpu_ring_fini(&adev->vcn.inst[i].ring_dec);
 
-	return 0;
-}
+	for (j = 0; j < adev->vcn.inst[i].num_enc_rings; ++j)
+		amdgpu_ring_fini(&adev->vcn.inst[i].ring_enc[j]);
 
-/* from vcn4 and above, only unified queue is used */
-static bool amdgpu_vcn_using_unified_queue(struct amdgpu_ring *ring)
-{
-	struct amdgpu_device *adev = ring->adev;
-	bool ret = false;
+	if (adev->vcn.per_inst_fw) {
+		amdgpu_ucode_release(&adev->vcn.inst[i].fw);
+	} else {
+		amdgpu_ucode_release(&adev->vcn.inst[0].fw);
+		adev->vcn.inst[i].fw = NULL;
+	}
 
-	if (adev->ip_versions[UVD_HWIP][0] >= IP_VERSION(4, 0, 0))
-		ret = true;
+	if (adev->vcn.reg_list)
+		amdgpu_vcn_reg_dump_fini(adev);
 
-	return ret;
+	mutex_destroy(&adev->vcn.inst[i].vcn_pg_lock);
+	mutex_destroy(&adev->vcn.inst[i].vcn1_jpeg1_workaround);
 }
 
 bool amdgpu_vcn_is_disabled_vcn(struct amdgpu_device *adev, enum vcn_ring_type type, uint32_t vcn_instance)
 {
 	bool ret = false;
-	int vcn_config = adev->vcn.vcn_config[vcn_instance];
+	int vcn_config = adev->vcn.inst[vcn_instance].vcn_config;
 
-	if ((type == VCN_ENCODE_RING) && (vcn_config & VCN_BLOCK_ENCODE_DISABLE_MASK)) {
+	if ((type == VCN_ENCODE_RING) && (vcn_config & VCN_BLOCK_ENCODE_DISABLE_MASK))
 		ret = true;
-	} else if ((type == VCN_DECODE_RING) && (vcn_config & VCN_BLOCK_DECODE_DISABLE_MASK)) {
+	else if ((type == VCN_DECODE_RING) && (vcn_config & VCN_BLOCK_DECODE_DISABLE_MASK))
 		ret = true;
-	} else if ((type == VCN_UNIFIED_RING) && (vcn_config & VCN_BLOCK_QUEUE_DISABLE_MASK)) {
+	else if ((type == VCN_UNIFIED_RING) && (vcn_config & VCN_BLOCK_QUEUE_DISABLE_MASK))
 		ret = true;
-	}
 
 	return ret;
 }
 
-int amdgpu_vcn_suspend(struct amdgpu_device *adev)
+static int amdgpu_vcn_save_vcpu_bo_inst(struct amdgpu_device *adev, int i)
 {
-	unsigned size;
+	unsigned int size;
 	void *ptr;
-	int i, idx;
+	int idx;
 
-	cancel_delayed_work_sync(&adev->vcn.idle_work);
+	if (adev->vcn.harvest_config & (1 << i))
+		return 0;
+	if (adev->vcn.inst[i].vcpu_bo == NULL)
+		return 0;
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
-		if (adev->vcn.harvest_config & (1 << i))
-			continue;
-		if (adev->vcn.inst[i].vcpu_bo == NULL)
-			return 0;
+	size = amdgpu_bo_size(adev->vcn.inst[i].vcpu_bo);
+	ptr = adev->vcn.inst[i].cpu_addr;
 
-		size = amdgpu_bo_size(adev->vcn.inst[i].vcpu_bo);
-		ptr = adev->vcn.inst[i].cpu_addr;
+	adev->vcn.inst[i].saved_bo = kvmalloc(size, GFP_KERNEL);
+	if (!adev->vcn.inst[i].saved_bo)
+		return -ENOMEM;
 
-		adev->vcn.inst[i].saved_bo = kvmalloc(size, GFP_KERNEL);
-		if (!adev->vcn.inst[i].saved_bo)
-			return -ENOMEM;
+	if (drm_dev_enter(adev_to_drm(adev), &idx)) {
+		memcpy_fromio(adev->vcn.inst[i].saved_bo, ptr, size);
+		drm_dev_exit(idx);
+	}
 
-		if (drm_dev_enter(adev_to_drm(adev), &idx)) {
-			memcpy_fromio(adev->vcn.inst[i].saved_bo, ptr, size);
-			drm_dev_exit(idx);
-		}
+	return 0;
+}
+
+int amdgpu_vcn_save_vcpu_bo(struct amdgpu_device *adev)
+{
+	int ret, i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		ret = amdgpu_vcn_save_vcpu_bo_inst(adev, i);
+		if (ret)
+			return ret;
 	}
+
 	return 0;
 }
 
-int amdgpu_vcn_resume(struct amdgpu_device *adev)
+int amdgpu_vcn_suspend(struct amdgpu_device *adev, int i)
 {
-	unsigned size;
+	bool in_ras_intr = amdgpu_ras_intr_triggered();
+
+	if (adev->vcn.harvest_config & (1 << i))
+		return 0;
+
+	/* err_event_athub and dpc recovery will corrupt VCPU buffer, so we need to
+	 * restore fw data and clear buffer in amdgpu_vcn_resume() */
+	if (in_ras_intr || adev->pcie_reset_ctx.in_link_reset)
+		return 0;
+
+	return amdgpu_vcn_save_vcpu_bo_inst(adev, i);
+}
+
+int amdgpu_vcn_resume(struct amdgpu_device *adev, int i)
+{
+	unsigned int size;
 	void *ptr;
-	int i, idx;
+	int idx;
 
-	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
-		if (adev->vcn.harvest_config & (1 << i))
-			continue;
-		if (adev->vcn.inst[i].vcpu_bo == NULL)
-			return -EINVAL;
+	if (adev->vcn.harvest_config & (1 << i))
+		return 0;
+	if (adev->vcn.inst[i].vcpu_bo == NULL)
+		return -EINVAL;
 
-		size = amdgpu_bo_size(adev->vcn.inst[i].vcpu_bo);
-		ptr = adev->vcn.inst[i].cpu_addr;
+	size = amdgpu_bo_size(adev->vcn.inst[i].vcpu_bo);
+	ptr = adev->vcn.inst[i].cpu_addr;
 
-		if (adev->vcn.inst[i].saved_bo != NULL) {
+	if (adev->vcn.inst[i].saved_bo != NULL) {
+		if (drm_dev_enter(adev_to_drm(adev), &idx)) {
+			memcpy_toio(ptr, adev->vcn.inst[i].saved_bo, size);
+			drm_dev_exit(idx);
+		}
+		kvfree(adev->vcn.inst[i].saved_bo);
+		adev->vcn.inst[i].saved_bo = NULL;
+	} else {
+		const struct common_firmware_header *hdr;
+		unsigned int offset;
+
+		hdr = (const struct common_firmware_header *)adev->vcn.inst[i].fw->data;
+		if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
+			offset = le32_to_cpu(hdr->ucode_array_offset_bytes);
 			if (drm_dev_enter(adev_to_drm(adev), &idx)) {
-				memcpy_toio(ptr, adev->vcn.inst[i].saved_bo, size);
+				memcpy_toio(adev->vcn.inst[i].cpu_addr,
+					    adev->vcn.inst[i].fw->data + offset,
+					    le32_to_cpu(hdr->ucode_size_bytes));
 				drm_dev_exit(idx);
 			}
-			kvfree(adev->vcn.inst[i].saved_bo);
-			adev->vcn.inst[i].saved_bo = NULL;
-		} else {
-			const struct common_firmware_header *hdr;
-			unsigned offset;
-
-			hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
-			if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
-				offset = le32_to_cpu(hdr->ucode_array_offset_bytes);
-				if (drm_dev_enter(adev_to_drm(adev), &idx)) {
-					memcpy_toio(adev->vcn.inst[i].cpu_addr, adev->vcn.fw->data + offset,
-						    le32_to_cpu(hdr->ucode_size_bytes));
-					drm_dev_exit(idx);
-				}
-				size -= le32_to_cpu(hdr->ucode_size_bytes);
-				ptr += le32_to_cpu(hdr->ucode_size_bytes);
-			}
-			memset_io(ptr, 0, size);
+			size -= le32_to_cpu(hdr->ucode_size_bytes);
+			ptr += le32_to_cpu(hdr->ucode_size_bytes);
 		}
+		memset_io(ptr, 0, size);
 	}
+
 	return 0;
 }
 
-static void amdgpu_vcn_idle_work_handler(struct work_struct *work)
+void amdgpu_vcn_get_profile(struct amdgpu_device *adev)
 {
-	struct amdgpu_device *adev =
-		container_of(work, struct amdgpu_device, vcn.idle_work.work);
-	unsigned int fences = 0, fence[AMDGPU_MAX_VCN_INSTANCES] = {0};
-	unsigned int i, j;
-	int r = 0;
+	int r;
 
-	for (j = 0; j < adev->vcn.num_vcn_inst; ++j) {
-		if (adev->vcn.harvest_config & (1 << j))
-			continue;
+	mutex_lock(&adev->vcn.workload_profile_mutex);
+
+	if (adev->vcn.workload_profile_active) {
+		mutex_unlock(&adev->vcn.workload_profile_mutex);
+		return;
+	}
+	r = amdgpu_dpm_switch_power_profile(adev, PP_SMC_POWER_PROFILE_VIDEO,
+					    true);
+	if (r)
+		dev_warn(adev->dev,
+			 "(%d) failed to enable video power profile mode\n", r);
+	else
+		adev->vcn.workload_profile_active = true;
+	mutex_unlock(&adev->vcn.workload_profile_mutex);
+}
 
-		for (i = 0; i < adev->vcn.num_enc_rings; ++i) {
-			fence[j] += amdgpu_fence_count_emitted(&adev->vcn.inst[j].ring_enc[i]);
+void amdgpu_vcn_put_profile(struct amdgpu_device *adev)
+{
+	bool pg = true;
+	int r, i;
+
+	mutex_lock(&adev->vcn.workload_profile_mutex);
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		if (adev->vcn.inst[i].cur_state != AMD_PG_STATE_GATE) {
+			pg = false;
+			break;
 		}
+	}
 
-		if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)	{
-			struct dpg_pause_state new_state;
+	if (pg) {
+		r = amdgpu_dpm_switch_power_profile(
+			adev, PP_SMC_POWER_PROFILE_VIDEO, false);
+		if (r)
+			dev_warn(
+				adev->dev,
+				"(%d) failed to disable video power profile mode\n",
+				r);
+		else
+			adev->vcn.workload_profile_active = false;
+	}
 
-			if (fence[j] ||
-				unlikely(atomic_read(&adev->vcn.inst[j].dpg_enc_submission_cnt)))
-				new_state.fw_based = VCN_DPG_STATE__PAUSE;
-			else
-				new_state.fw_based = VCN_DPG_STATE__UNPAUSE;
+	mutex_unlock(&adev->vcn.workload_profile_mutex);
+}
 
-			adev->vcn.pause_dpg_mode(adev, j, &new_state);
-		}
+static void amdgpu_vcn_idle_work_handler(struct work_struct *work)
+{
+	struct amdgpu_vcn_inst *vcn_inst =
+		container_of(work, struct amdgpu_vcn_inst, idle_work.work);
+	struct amdgpu_device *adev = vcn_inst->adev;
+	unsigned int fences = 0, fence[AMDGPU_MAX_VCN_INSTANCES] = {0};
+	unsigned int i = vcn_inst->inst, j;
 
-		fence[j] += amdgpu_fence_count_emitted(&adev->vcn.inst[j].ring_dec);
-		fences += fence[j];
+	if (adev->vcn.harvest_config & (1 << i))
+		return;
+
+	for (j = 0; j < adev->vcn.inst[i].num_enc_rings; ++j)
+		fence[i] += amdgpu_fence_count_emitted(&vcn_inst->ring_enc[j]);
+
+	/* Only set DPG pause for VCN3 or below, VCN4 and above will be handled by FW */
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG &&
+	    !adev->vcn.inst[i].using_unified_queue) {
+		struct dpg_pause_state new_state;
+
+		if (fence[i] ||
+		    unlikely(atomic_read(&vcn_inst->dpg_enc_submission_cnt)))
+			new_state.fw_based = VCN_DPG_STATE__PAUSE;
+		else
+			new_state.fw_based = VCN_DPG_STATE__UNPAUSE;
+
+		adev->vcn.inst[i].pause_dpg_mode(vcn_inst, &new_state);
 	}
 
-	if (!fences && !atomic_read(&adev->vcn.total_submission_cnt)) {
-		amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN,
-		       AMD_PG_STATE_GATE);
-		r = amdgpu_dpm_switch_power_profile(adev, PP_SMC_POWER_PROFILE_VIDEO,
-				false);
-		if (r)
-			dev_warn(adev->dev, "(%d) failed to disable video power profile mode\n", r);
+	fence[i] += amdgpu_fence_count_emitted(&vcn_inst->ring_dec);
+	fences += fence[i];
+
+	if (!fences && !atomic_read(&vcn_inst->total_submission_cnt)) {
+		mutex_lock(&vcn_inst->vcn_pg_lock);
+		vcn_inst->set_pg_state(vcn_inst, AMD_PG_STATE_GATE);
+		mutex_unlock(&vcn_inst->vcn_pg_lock);
+		amdgpu_vcn_put_profile(adev);
+
 	} else {
-		schedule_delayed_work(&adev->vcn.idle_work, VCN_IDLE_TIMEOUT);
+		schedule_delayed_work(&vcn_inst->idle_work, VCN_IDLE_TIMEOUT);
 	}
 }
 
 void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	int r = 0;
+	struct amdgpu_vcn_inst *vcn_inst = &adev->vcn.inst[ring->me];
 
-	atomic_inc(&adev->vcn.total_submission_cnt);
+	atomic_inc(&vcn_inst->total_submission_cnt);
 
-	if (!cancel_delayed_work_sync(&adev->vcn.idle_work)) {
-		r = amdgpu_dpm_switch_power_profile(adev, PP_SMC_POWER_PROFILE_VIDEO,
-				true);
-		if (r)
-			dev_warn(adev->dev, "(%d) failed to switch to video power profile mode\n", r);
-	}
+	cancel_delayed_work_sync(&vcn_inst->idle_work);
 
-	mutex_lock(&adev->vcn.vcn_pg_lock);
-	amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN,
-	       AMD_PG_STATE_UNGATE);
+	mutex_lock(&vcn_inst->vcn_pg_lock);
+	vcn_inst->set_pg_state(vcn_inst, AMD_PG_STATE_UNGATE);
 
-	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)	{
+	/* Only set DPG pause for VCN3 or below, VCN4 and above will be handled by FW */
+	if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG &&
+	    !vcn_inst->using_unified_queue) {
 		struct dpg_pause_state new_state;
 
 		if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC) {
-			atomic_inc(&adev->vcn.inst[ring->me].dpg_enc_submission_cnt);
+			atomic_inc(&vcn_inst->dpg_enc_submission_cnt);
 			new_state.fw_based = VCN_DPG_STATE__PAUSE;
 		} else {
 			unsigned int fences = 0;
 			unsigned int i;
 
-			for (i = 0; i < adev->vcn.num_enc_rings; ++i)
-				fences += amdgpu_fence_count_emitted(&adev->vcn.inst[ring->me].ring_enc[i]);
+			for (i = 0; i < vcn_inst->num_enc_rings; ++i)
+				fences += amdgpu_fence_count_emitted(&vcn_inst->ring_enc[i]);
 
-			if (fences || atomic_read(&adev->vcn.inst[ring->me].dpg_enc_submission_cnt))
+			if (fences || atomic_read(&vcn_inst->dpg_enc_submission_cnt))
 				new_state.fw_based = VCN_DPG_STATE__PAUSE;
 			else
 				new_state.fw_based = VCN_DPG_STATE__UNPAUSE;
 		}
 
-		adev->vcn.pause_dpg_mode(adev, ring->me, &new_state);
+		vcn_inst->pause_dpg_mode(vcn_inst, &new_state);
 	}
-	mutex_unlock(&adev->vcn.vcn_pg_lock);
+	mutex_unlock(&vcn_inst->vcn_pg_lock);
+	amdgpu_vcn_get_profile(adev);
 }
 
 void amdgpu_vcn_ring_end_use(struct amdgpu_ring *ring)
 {
+	struct amdgpu_device *adev = ring->adev;
+
+	/* Only set DPG pause for VCN3 or below, VCN4 and above will be handled by FW */
 	if (ring->adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG &&
-		ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC)
+	    ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC &&
+	    !adev->vcn.inst[ring->me].using_unified_queue)
 		atomic_dec(&ring->adev->vcn.inst[ring->me].dpg_enc_submission_cnt);
 
-	atomic_dec(&ring->adev->vcn.total_submission_cnt);
+	atomic_dec(&ring->adev->vcn.inst[ring->me].total_submission_cnt);
 
-	schedule_delayed_work(&ring->adev->vcn.idle_work, VCN_IDLE_TIMEOUT);
+	schedule_delayed_work(&ring->adev->vcn.inst[ring->me].idle_work,
+			      VCN_IDLE_TIMEOUT);
 }
 
 int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 	uint32_t tmp = 0;
-	unsigned i;
+	unsigned int i;
 	int r;
 
 	/* VCN in SRIOV does not support direct register read/write */
@@ -469,7 +566,7 @@ int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring)
 	r = amdgpu_ring_alloc(ring, 3);
 	if (r)
 		return r;
-	amdgpu_ring_write(ring, PACKET0(adev->vcn.internal.scratch9, 0));
+	amdgpu_ring_write(ring, PACKET0(adev->vcn.inst[ring->me].internal.scratch9, 0));
 	amdgpu_ring_write(ring, 0xDEADBEEF);
 	amdgpu_ring_commit(ring);
 	for (i = 0; i < adev->usec_timeout; i++) {
@@ -529,19 +626,19 @@ static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring,
 
 	r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL,
 				     64, AMDGPU_IB_POOL_DIRECT,
-				     &job);
+				     &job, AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		goto err;
 
 	ib = &job->ibs[0];
-	ib->ptr[0] = PACKET0(adev->vcn.internal.data0, 0);
+	ib->ptr[0] = PACKET0(adev->vcn.inst[ring->me].internal.data0, 0);
 	ib->ptr[1] = addr;
-	ib->ptr[2] = PACKET0(adev->vcn.internal.data1, 0);
+	ib->ptr[2] = PACKET0(adev->vcn.inst[ring->me].internal.data1, 0);
 	ib->ptr[3] = addr >> 32;
-	ib->ptr[4] = PACKET0(adev->vcn.internal.cmd, 0);
+	ib->ptr[4] = PACKET0(adev->vcn.inst[ring->me].internal.cmd, 0);
 	ib->ptr[5] = 0;
 	for (i = 6; i < 16; i += 2) {
-		ib->ptr[i] = PACKET0(adev->vcn.internal.nop, 0);
+		ib->ptr[i] = PACKET0(adev->vcn.inst[ring->me].internal.nop, 0);
 		ib->ptr[i+1] = 0;
 	}
 	ib->length_dw = 16;
@@ -550,7 +647,7 @@ static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring,
 	if (r)
 		goto err_free;
 
-	amdgpu_ib_free(adev, ib_msg, f);
+	amdgpu_ib_free(ib_msg, f);
 
 	if (fence)
 		*fence = dma_fence_get(f);
@@ -561,7 +658,7 @@ static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring,
 err_free:
 	amdgpu_job_free(job);
 err:
-	amdgpu_ib_free(adev, ib_msg, f);
+	amdgpu_ib_free(ib_msg, f);
 	return r;
 }
 
@@ -700,17 +797,16 @@ static int amdgpu_vcn_dec_sw_send_msg(struct amdgpu_ring *ring,
 	struct amdgpu_job *job;
 	struct amdgpu_ib *ib;
 	uint64_t addr = AMDGPU_GPU_PAGE_ALIGN(ib_msg->gpu_addr);
-	bool sq = amdgpu_vcn_using_unified_queue(ring);
 	uint32_t *ib_checksum;
 	uint32_t ib_pack_in_dw;
 	int i, r;
 
-	if (sq)
+	if (adev->vcn.inst[ring->me].using_unified_queue)
 		ib_size_dw += 8;
 
 	r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL,
 				     ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT,
-				     &job);
+				     &job, AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		goto err;
 
@@ -718,7 +814,7 @@ static int amdgpu_vcn_dec_sw_send_msg(struct amdgpu_ring *ring,
 	ib->length_dw = 0;
 
 	/* single queue headers */
-	if (sq) {
+	if (adev->vcn.inst[ring->me].using_unified_queue) {
 		ib_pack_in_dw = sizeof(struct amdgpu_vcn_decode_buffer) / sizeof(uint32_t)
 						+ 4 + 2; /* engine info + decoding ib in dw */
 		ib_checksum = amdgpu_vcn_unified_ring_ib_header(ib, ib_pack_in_dw, false);
@@ -737,14 +833,14 @@ static int amdgpu_vcn_dec_sw_send_msg(struct amdgpu_ring *ring,
 	for (i = ib->length_dw; i < ib_size_dw; ++i)
 		ib->ptr[i] = 0x0;
 
-	if (sq)
+	if (adev->vcn.inst[ring->me].using_unified_queue)
 		amdgpu_vcn_unified_ring_ib_checksum(&ib_checksum, ib_pack_in_dw);
 
 	r = amdgpu_job_submit_direct(job, ring, &f);
 	if (r)
 		goto err_free;
 
-	amdgpu_ib_free(adev, ib_msg, f);
+	amdgpu_ib_free(ib_msg, f);
 
 	if (fence)
 		*fence = dma_fence_get(f);
@@ -755,7 +851,7 @@ static int amdgpu_vcn_dec_sw_send_msg(struct amdgpu_ring *ring,
 err_free:
 	amdgpu_job_free(job);
 err:
-	amdgpu_ib_free(adev, ib_msg, f);
+	amdgpu_ib_free(ib_msg, f);
 	return r;
 }
 
@@ -795,7 +891,7 @@ int amdgpu_vcn_enc_ring_test_ring(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 	uint32_t rptr;
-	unsigned i;
+	unsigned int i;
 	int r;
 
 	if (amdgpu_sriov_vf(adev))
@@ -827,20 +923,20 @@ static int amdgpu_vcn_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t hand
 					 struct dma_fence **fence)
 {
 	unsigned int ib_size_dw = 16;
+	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_job *job;
 	struct amdgpu_ib *ib;
 	struct dma_fence *f = NULL;
 	uint32_t *ib_checksum = NULL;
 	uint64_t addr;
-	bool sq = amdgpu_vcn_using_unified_queue(ring);
 	int i, r;
 
-	if (sq)
+	if (adev->vcn.inst[ring->me].using_unified_queue)
 		ib_size_dw += 8;
 
 	r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL,
 				     ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT,
-				     &job);
+				     &job, AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		return r;
 
@@ -849,7 +945,7 @@ static int amdgpu_vcn_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t hand
 
 	ib->length_dw = 0;
 
-	if (sq)
+	if (adev->vcn.inst[ring->me].using_unified_queue)
 		ib_checksum = amdgpu_vcn_unified_ring_ib_header(ib, 0x11, true);
 
 	ib->ptr[ib->length_dw++] = 0x00000018;
@@ -857,7 +953,7 @@ static int amdgpu_vcn_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t hand
 	ib->ptr[ib->length_dw++] = handle;
 	ib->ptr[ib->length_dw++] = upper_32_bits(addr);
 	ib->ptr[ib->length_dw++] = addr;
-	ib->ptr[ib->length_dw++] = 0x0000000b;
+	ib->ptr[ib->length_dw++] = 0x00000000;
 
 	ib->ptr[ib->length_dw++] = 0x00000014;
 	ib->ptr[ib->length_dw++] = 0x00000002; /* task info */
@@ -871,7 +967,7 @@ static int amdgpu_vcn_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t hand
 	for (i = ib->length_dw; i < ib_size_dw; ++i)
 		ib->ptr[i] = 0x0;
 
-	if (sq)
+	if (adev->vcn.inst[ring->me].using_unified_queue)
 		amdgpu_vcn_unified_ring_ib_checksum(&ib_checksum, 0x11);
 
 	r = amdgpu_job_submit_direct(job, ring, &f);
@@ -894,20 +990,20 @@ static int amdgpu_vcn_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han
 					  struct dma_fence **fence)
 {
 	unsigned int ib_size_dw = 16;
+	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_job *job;
 	struct amdgpu_ib *ib;
 	struct dma_fence *f = NULL;
 	uint32_t *ib_checksum = NULL;
 	uint64_t addr;
-	bool sq = amdgpu_vcn_using_unified_queue(ring);
 	int i, r;
 
-	if (sq)
+	if (adev->vcn.inst[ring->me].using_unified_queue)
 		ib_size_dw += 8;
 
 	r = amdgpu_job_alloc_with_ib(ring->adev, NULL, NULL,
 				     ib_size_dw * 4, AMDGPU_IB_POOL_DIRECT,
-				     &job);
+				     &job, AMDGPU_KERNEL_JOB_ID_VCN_RING_TEST);
 	if (r)
 		return r;
 
@@ -916,7 +1012,7 @@ static int amdgpu_vcn_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han
 
 	ib->length_dw = 0;
 
-	if (sq)
+	if (adev->vcn.inst[ring->me].using_unified_queue)
 		ib_checksum = amdgpu_vcn_unified_ring_ib_header(ib, 0x11, true);
 
 	ib->ptr[ib->length_dw++] = 0x00000018;
@@ -924,7 +1020,7 @@ static int amdgpu_vcn_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han
 	ib->ptr[ib->length_dw++] = handle;
 	ib->ptr[ib->length_dw++] = upper_32_bits(addr);
 	ib->ptr[ib->length_dw++] = addr;
-	ib->ptr[ib->length_dw++] = 0x0000000b;
+	ib->ptr[ib->length_dw++] = 0x00000000;
 
 	ib->ptr[ib->length_dw++] = 0x00000014;
 	ib->ptr[ib->length_dw++] = 0x00000002;
@@ -938,7 +1034,7 @@ static int amdgpu_vcn_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han
 	for (i = ib->length_dw; i < ib_size_dw; ++i)
 		ib->ptr[i] = 0x0;
 
-	if (sq)
+	if (adev->vcn.inst[ring->me].using_unified_queue)
 		amdgpu_vcn_unified_ring_ib_checksum(&ib_checksum, 0x11);
 
 	r = amdgpu_job_submit_direct(job, ring, &f);
@@ -985,7 +1081,7 @@ int amdgpu_vcn_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = 0;
 
 error:
-	amdgpu_ib_free(adev, &ib, fence);
+	amdgpu_ib_free(&ib, fence);
 	dma_fence_put(fence);
 
 	return r;
@@ -993,11 +1089,15 @@ error:
 
 int amdgpu_vcn_unified_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 {
+	struct amdgpu_device *adev = ring->adev;
 	long r;
 
-	r = amdgpu_vcn_enc_ring_test_ib(ring, timeout);
-	if (r)
-		goto error;
+	if ((amdgpu_ip_version(adev, UVD_HWIP, 0) != IP_VERSION(4, 0, 3)) &&
+	    (amdgpu_ip_version(adev, UVD_HWIP, 0) != IP_VERSION(5, 0, 1))) {
+		r = amdgpu_vcn_enc_ring_test_ib(ring, timeout);
+		if (r)
+			goto error;
+	}
 
 	r =  amdgpu_vcn_dec_sw_ring_test_ib(ring, timeout);
 
@@ -1007,7 +1107,7 @@ error:
 
 enum amdgpu_ring_priority_level amdgpu_vcn_get_enc_ring_prio(int ring)
 {
-	switch(ring) {
+	switch (ring) {
 	case 0:
 		return AMDGPU_RING_PRIO_0;
 	case 1:
@@ -1019,30 +1119,32 @@ enum amdgpu_ring_priority_level amdgpu_vcn_get_enc_ring_prio(int ring)
 	}
 }
 
-void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev)
+void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev, int i)
 {
-	int i;
 	unsigned int idx;
 
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
 		const struct common_firmware_header *hdr;
-		hdr = (const struct common_firmware_header *)adev->vcn.fw->data;
 
-		for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
-			if (adev->vcn.harvest_config & (1 << i))
-				continue;
-			/* currently only support 2 FW instances */
-			if (i >= 2) {
-				dev_info(adev->dev, "More then 2 VCN FW instances!\n");
-				break;
-			}
-			idx = AMDGPU_UCODE_ID_VCN + i;
-			adev->firmware.ucode[idx].ucode_id = idx;
-			adev->firmware.ucode[idx].fw = adev->vcn.fw;
-			adev->firmware.fw_size +=
-				ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE);
+		if (adev->vcn.harvest_config & (1 << i))
+			return;
+
+		if ((amdgpu_ip_version(adev, UVD_HWIP, 0) == IP_VERSION(4, 0, 3) ||
+		     amdgpu_ip_version(adev, UVD_HWIP, 0) == IP_VERSION(5, 0, 1))
+		    && (i > 0))
+			return;
+
+		hdr = (const struct common_firmware_header *)adev->vcn.inst[i].fw->data;
+		/* currently only support 2 FW instances */
+		if (i >= 2) {
+			dev_info(adev->dev, "More then 2 VCN FW instances!\n");
+			return;
 		}
-		dev_info(adev->dev, "Will use PSP to load VCN firmware\n");
+		idx = AMDGPU_UCODE_ID_VCN + i;
+		adev->firmware.ucode[idx].ucode_id = idx;
+		adev->firmware.ucode[idx].fw = adev->vcn.inst[i].fw;
+		adev->firmware.fw_size +=
+			ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE);
 	}
 }
 
@@ -1051,11 +1153,11 @@ void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev)
  */
 #if defined(CONFIG_DEBUG_FS)
 static ssize_t amdgpu_debugfs_vcn_fwlog_read(struct file *f, char __user *buf,
-                                             size_t size, loff_t *pos)
+					     size_t size, loff_t *pos)
 {
 	struct amdgpu_vcn_inst *vcn;
 	void *log_buf;
-	volatile struct amdgpu_vcn_fwlog *plog;
+	struct amdgpu_vcn_fwlog *plog;
 	unsigned int read_pos, write_pos, available, i, read_bytes = 0;
 	unsigned int read_num[2] = {0};
 
@@ -1068,7 +1170,7 @@ static ssize_t amdgpu_debugfs_vcn_fwlog_read(struct file *f, char __user *buf,
 
 	log_buf = vcn->fw_shared.cpu_addr + vcn->fw_shared.mem_size;
 
-	plog = (volatile struct amdgpu_vcn_fwlog *)log_buf;
+	plog = (struct amdgpu_vcn_fwlog *)log_buf;
 	read_pos = plog->rptr;
 	write_pos = plog->wptr;
 
@@ -1080,7 +1182,7 @@ static ssize_t amdgpu_debugfs_vcn_fwlog_read(struct file *f, char __user *buf,
 
 	if (write_pos > read_pos) {
 		available = write_pos - read_pos;
-		read_num[0] = min(size, (size_t)available);
+		read_num[0] = min_t(size_t, size, available);
 	} else {
 		read_num[0] = AMDGPU_VCNFW_LOG_SIZE - read_pos;
 		available = read_num[0] + write_pos - plog->header_size;
@@ -1097,7 +1199,7 @@ static ssize_t amdgpu_debugfs_vcn_fwlog_read(struct file *f, char __user *buf,
 			if (read_pos == AMDGPU_VCNFW_LOG_SIZE)
 				read_pos = plog->header_size;
 			if (read_num[i] == copy_to_user((buf + read_bytes),
-			                                (log_buf + read_pos), read_num[i]))
+							(log_buf + read_pos), read_num[i]))
 				return -EFAULT;
 
 			read_bytes += read_num[i];
@@ -1118,7 +1220,7 @@ static const struct file_operations amdgpu_debugfs_vcnfwlog_fops = {
 #endif
 
 void amdgpu_debugfs_vcn_fwlog_init(struct amdgpu_device *adev, uint8_t i,
-                                   struct amdgpu_vcn_inst *vcn)
+				   struct amdgpu_vcn_inst *vcn)
 {
 #if defined(CONFIG_DEBUG_FS)
 	struct drm_minor *minor = adev_to_drm(adev)->primary;
@@ -1126,7 +1228,7 @@ void amdgpu_debugfs_vcn_fwlog_init(struct amdgpu_device *adev, uint8_t i,
 	char name[32];
 
 	sprintf(name, "amdgpu_vcn_%d_fwlog", i);
-	debugfs_create_file_size(name, S_IFREG | S_IRUGO, root, vcn,
+	debugfs_create_file_size(name, S_IFREG | 0444, root, vcn,
 				 &amdgpu_debugfs_vcnfwlog_fops,
 				 AMDGPU_VCNFW_LOG_SIZE);
 #endif
@@ -1135,12 +1237,12 @@ void amdgpu_debugfs_vcn_fwlog_init(struct amdgpu_device *adev, uint8_t i,
 void amdgpu_vcn_fwlog_init(struct amdgpu_vcn_inst *vcn)
 {
 #if defined(CONFIG_DEBUG_FS)
-	volatile uint32_t *flag = vcn->fw_shared.cpu_addr;
+	uint32_t *flag = vcn->fw_shared.cpu_addr;
 	void *fw_log_cpu_addr = vcn->fw_shared.cpu_addr + vcn->fw_shared.mem_size;
 	uint64_t fw_log_gpu_addr = vcn->fw_shared.gpu_addr + vcn->fw_shared.mem_size;
-	volatile struct amdgpu_vcn_fwlog *log_buf = fw_log_cpu_addr;
-	volatile struct amdgpu_fw_shared_fw_logging *fw_log = vcn->fw_shared.cpu_addr
-                                                         + vcn->fw_shared.log_offset;
+	struct amdgpu_vcn_fwlog *log_buf = fw_log_cpu_addr;
+	struct amdgpu_fw_shared_fw_logging *fw_log = vcn->fw_shared.cpu_addr
+							 + vcn->fw_shared.log_offset;
 	*flag |= cpu_to_le32(AMDGPU_VCN_FW_LOGGING_FLAG);
 	fw_log->is_enabled = 1;
 	fw_log->addr_lo = cpu_to_le32(fw_log_gpu_addr & 0xFFFFFFFF);
@@ -1172,7 +1274,7 @@ int amdgpu_vcn_process_poison_irq(struct amdgpu_device *adev,
 		amdgpu_ras_interrupt_dispatch(adev, &ih_data);
 	} else {
 		if (adev->virt.ops && adev->virt.ops->ras_poison_handler)
-			adev->virt.ops->ras_poison_handler(adev);
+			adev->virt.ops->ras_poison_handler(adev, ras_if->block);
 		else
 			dev_warn(adev->dev,
 				"No ras_poison_handler interface in SRIOV for VCN!\n");
@@ -1181,6 +1283,32 @@ int amdgpu_vcn_process_poison_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
+int amdgpu_vcn_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
+{
+	int r, i;
+
+	r = amdgpu_ras_block_late_init(adev, ras_block);
+	if (r)
+		return r;
+
+	if (amdgpu_ras_is_supported(adev, ras_block->block)) {
+		for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+			if (adev->vcn.harvest_config & (1 << i) ||
+			    !adev->vcn.inst[i].ras_poison_irq.funcs)
+				continue;
+
+			r = amdgpu_irq_get(adev, &adev->vcn.inst[i].ras_poison_irq, 0);
+			if (r)
+				goto late_fini;
+		}
+	}
+	return 0;
+
+late_fini:
+	amdgpu_ras_block_late_fini(adev, ras_block);
+	return r;
+}
+
 int amdgpu_vcn_ras_sw_init(struct amdgpu_device *adev)
 {
 	int err;
@@ -1202,7 +1330,308 @@ int amdgpu_vcn_ras_sw_init(struct amdgpu_device *adev)
 	adev->vcn.ras_if = &ras->ras_block.ras_comm;
 
 	if (!ras->ras_block.ras_late_init)
-		ras->ras_block.ras_late_init = amdgpu_ras_block_late_init;
+		ras->ras_block.ras_late_init = amdgpu_vcn_ras_late_init;
+
+	return 0;
+}
+
+int amdgpu_vcn_psp_update_sram(struct amdgpu_device *adev, int inst_idx,
+			       enum AMDGPU_UCODE_ID ucode_id)
+{
+	struct amdgpu_firmware_info ucode = {
+		.ucode_id = (ucode_id ? ucode_id :
+			    (inst_idx ? AMDGPU_UCODE_ID_VCN1_RAM :
+					AMDGPU_UCODE_ID_VCN0_RAM)),
+		.mc_addr = adev->vcn.inst[inst_idx].dpg_sram_gpu_addr,
+		.ucode_size = ((uintptr_t)adev->vcn.inst[inst_idx].dpg_sram_curr_addr -
+			      (uintptr_t)adev->vcn.inst[inst_idx].dpg_sram_cpu_addr),
+	};
+
+	return psp_execute_ip_fw_load(&adev->psp, &ucode);
+}
+
+static ssize_t amdgpu_get_vcn_reset_mask(struct device *dev,
+						struct device_attribute *attr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (!adev)
+		return -ENODEV;
+
+	return amdgpu_show_reset_mask(buf, adev->vcn.supported_reset);
+}
+
+static DEVICE_ATTR(vcn_reset_mask, 0444,
+		   amdgpu_get_vcn_reset_mask, NULL);
+
+int amdgpu_vcn_sysfs_reset_mask_init(struct amdgpu_device *adev)
+{
+	int r = 0;
+
+	if (adev->vcn.num_vcn_inst) {
+		r = device_create_file(adev->dev, &dev_attr_vcn_reset_mask);
+		if (r)
+			return r;
+	}
+
+	return r;
+}
+
+void amdgpu_vcn_sysfs_reset_mask_fini(struct amdgpu_device *adev)
+{
+	if (adev->dev->kobj.sd) {
+		if (adev->vcn.num_vcn_inst)
+			device_remove_file(adev->dev, &dev_attr_vcn_reset_mask);
+	}
+}
+
+/*
+ * debugfs to enable/disable vcn job submission to specific core or
+ * instance. It is created only if the queue type is unified.
+ */
+#if defined(CONFIG_DEBUG_FS)
+static int amdgpu_debugfs_vcn_sched_mask_set(void *data, u64 val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i;
+	u64 mask;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+
+	mask = (1ULL << adev->vcn.num_vcn_inst) - 1;
+	if ((val & mask) == 0)
+		return -EINVAL;
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		ring = &adev->vcn.inst[i].ring_enc[0];
+		if (val & (1ULL << i))
+			ring->sched.ready = true;
+		else
+			ring->sched.ready = false;
+	}
+	/* publish sched.ready flag update effective immediately across smp */
+	smp_rmb();
+	return 0;
+}
+
+static int amdgpu_debugfs_vcn_sched_mask_get(void *data, u64 *val)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)data;
+	u32 i;
+	u64 mask = 0;
+	struct amdgpu_ring *ring;
+
+	if (!adev)
+		return -ENODEV;
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		ring = &adev->vcn.inst[i].ring_enc[0];
+		if (ring->sched.ready)
+			mask |= 1ULL << i;
+		}
+	*val = mask;
+	return 0;
+}
+
+DEFINE_DEBUGFS_ATTRIBUTE(amdgpu_debugfs_vcn_sched_mask_fops,
+			 amdgpu_debugfs_vcn_sched_mask_get,
+			 amdgpu_debugfs_vcn_sched_mask_set, "%llx\n");
+#endif
+
+void amdgpu_debugfs_vcn_sched_mask_init(struct amdgpu_device *adev)
+{
+#if defined(CONFIG_DEBUG_FS)
+	struct drm_minor *minor = adev_to_drm(adev)->primary;
+	struct dentry *root = minor->debugfs_root;
+	char name[32];
+
+	if (adev->vcn.num_vcn_inst <= 1 || !adev->vcn.inst[0].using_unified_queue)
+		return;
+	sprintf(name, "amdgpu_vcn_sched_mask");
+	debugfs_create_file(name, 0600, root, adev,
+			    &amdgpu_debugfs_vcn_sched_mask_fops);
+#endif
+}
+
+/**
+ * vcn_set_powergating_state - set VCN block powergating state
+ *
+ * @ip_block: amdgpu_ip_block pointer
+ * @state: power gating state
+ *
+ * Set VCN block powergating state
+ */
+int vcn_set_powergating_state(struct amdgpu_ip_block *ip_block,
+			      enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int ret = 0, i;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
+		struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[i];
+
+		ret |= vinst->set_pg_state(vinst, state);
+	}
+
+	return ret;
+}
+
+/**
+ * amdgpu_vcn_reset_engine - Reset a specific VCN engine
+ * @adev: Pointer to the AMDGPU device
+ * @instance_id: VCN engine instance to reset
+ *
+ * Returns: 0 on success, or a negative error code on failure.
+ */
+static int amdgpu_vcn_reset_engine(struct amdgpu_device *adev,
+				   uint32_t instance_id)
+{
+	struct amdgpu_vcn_inst *vinst = &adev->vcn.inst[instance_id];
+	int r, i;
+
+	mutex_lock(&vinst->engine_reset_mutex);
+	/* Stop the scheduler's work queue for the dec and enc rings if they are running.
+	 * This ensures that no new tasks are submitted to the queues while
+	 * the reset is in progress.
+	 */
+	drm_sched_wqueue_stop(&vinst->ring_dec.sched);
+	for (i = 0; i < vinst->num_enc_rings; i++)
+		drm_sched_wqueue_stop(&vinst->ring_enc[i].sched);
+
+	/* Perform the VCN reset for the specified instance */
+	r = vinst->reset(vinst);
+	if (r)
+		goto unlock;
+	r = amdgpu_ring_test_ring(&vinst->ring_dec);
+	if (r)
+		goto unlock;
+	for (i = 0; i < vinst->num_enc_rings; i++) {
+		r = amdgpu_ring_test_ring(&vinst->ring_enc[i]);
+		if (r)
+			goto unlock;
+	}
+	amdgpu_fence_driver_force_completion(&vinst->ring_dec);
+	for (i = 0; i < vinst->num_enc_rings; i++)
+		amdgpu_fence_driver_force_completion(&vinst->ring_enc[i]);
+
+	/* Restart the scheduler's work queue for the dec and enc rings
+	 * if they were stopped by this function. This allows new tasks
+	 * to be submitted to the queues after the reset is complete.
+	 */
+	drm_sched_wqueue_start(&vinst->ring_dec.sched);
+	for (i = 0; i < vinst->num_enc_rings; i++)
+		drm_sched_wqueue_start(&vinst->ring_enc[i].sched);
+
+unlock:
+	mutex_unlock(&vinst->engine_reset_mutex);
+
+	return r;
+}
+
+/**
+ * amdgpu_vcn_ring_reset - Reset a VCN ring
+ * @ring: ring to reset
+ * @vmid: vmid of guilty job
+ * @timedout_fence: fence of timed out job
+ *
+ * This helper is for VCN blocks without unified queues because
+ * resetting the engine resets all queues in that case.  With
+ * unified queues we have one queue per engine.
+ * Returns: 0 on success, or a negative error code on failure.
+ */
+int amdgpu_vcn_ring_reset(struct amdgpu_ring *ring,
+			  unsigned int vmid,
+			  struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (adev->vcn.inst[ring->me].using_unified_queue)
+		return -EINVAL;
+
+	return amdgpu_vcn_reset_engine(adev, ring->me);
+}
+
+int amdgpu_vcn_reg_dump_init(struct amdgpu_device *adev,
+			     const struct amdgpu_hwip_reg_entry *reg, u32 count)
+{
+	adev->vcn.ip_dump = kcalloc(adev->vcn.num_vcn_inst * count,
+				     sizeof(uint32_t), GFP_KERNEL);
+	if (!adev->vcn.ip_dump)
+		return -ENOMEM;
+	adev->vcn.reg_list = reg;
+	adev->vcn.reg_count = count;
 
 	return 0;
 }
+
+static void amdgpu_vcn_reg_dump_fini(struct amdgpu_device *adev)
+{
+	kfree(adev->vcn.ip_dump);
+	adev->vcn.ip_dump = NULL;
+	adev->vcn.reg_list = NULL;
+	adev->vcn.reg_count = 0;
+}
+
+void amdgpu_vcn_dump_ip_state(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	bool is_powered;
+	u32 inst_off;
+
+	if (!adev->vcn.ip_dump)
+		return;
+
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		if (adev->vcn.harvest_config & (1 << i))
+			continue;
+
+		inst_off = i * adev->vcn.reg_count;
+		/* mmUVD_POWER_STATUS is always readable and is the first in reg_list */
+		adev->vcn.ip_dump[inst_off] =
+			RREG32(SOC15_REG_ENTRY_OFFSET_INST(adev->vcn.reg_list[0], i));
+		is_powered = (adev->vcn.ip_dump[inst_off] &
+			      UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF) !=
+			      UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF;
+
+		if (is_powered)
+			for (j = 1; j < adev->vcn.reg_count; j++)
+				adev->vcn.ip_dump[inst_off + j] =
+				RREG32(SOC15_REG_ENTRY_OFFSET_INST(adev->vcn.reg_list[j], i));
+	}
+}
+
+void amdgpu_vcn_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, j;
+	bool is_powered;
+	u32 inst_off;
+
+	if (!adev->vcn.ip_dump)
+		return;
+
+	drm_printf(p, "num_instances:%d\n", adev->vcn.num_vcn_inst);
+	for (i = 0; i < adev->vcn.num_vcn_inst; i++) {
+		if (adev->vcn.harvest_config & (1 << i)) {
+			drm_printf(p, "\nHarvested Instance:VCN%d Skipping dump\n", i);
+			continue;
+		}
+
+		inst_off = i * adev->vcn.reg_count;
+		is_powered = (adev->vcn.ip_dump[inst_off] &
+			      UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF) !=
+			      UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF;
+
+		if (is_powered) {
+			drm_printf(p, "\nActive Instance:VCN%d\n", i);
+			for (j = 0; j < adev->vcn.reg_count; j++)
+				drm_printf(p, "%-50s \t 0x%08x\n", adev->vcn.reg_list[j].reg_name,
+					   adev->vcn.ip_dump[inst_off + j]);
+		} else {
+			drm_printf(p, "\nInactive Instance:VCN%d\n", i);
+		}
+	}
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
index c730949ece7d..82624b44e661 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h
@@ -1,5 +1,5 @@
 /*
- * Copyright 2016 Advanced Micro Devices, Inc.
+ * Copyright 2016-2024 Advanced Micro Devices, Inc. All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -32,8 +32,8 @@
 #define AMDGPU_VCN_FIRMWARE_OFFSET	256
 #define AMDGPU_VCN_MAX_ENC_RINGS	3
 
-#define AMDGPU_MAX_VCN_INSTANCES	2
-#define AMDGPU_MAX_VCN_ENC_RINGS  AMDGPU_VCN_MAX_ENC_RINGS * AMDGPU_MAX_VCN_INSTANCES
+#define AMDGPU_MAX_VCN_INSTANCES	4
+#define AMDGPU_MAX_VCN_ENC_RINGS  (AMDGPU_VCN_MAX_ENC_RINGS * AMDGPU_MAX_VCN_INSTANCES)
 
 #define AMDGPU_VCN_HARVEST_VCN0 (1 << 0)
 #define AMDGPU_VCN_HARVEST_VCN1 (1 << 1)
@@ -66,7 +66,6 @@
 #define VCN_ENC_CMD_REG_WAIT		0x0000000c
 
 #define VCN_AON_SOC_ADDRESS_2_0 	0x1f800
-#define VCN1_AON_SOC_ADDRESS_3_0 	0x48000
 #define VCN_VID_IP_ADDRESS_2_0		0x0
 #define VCN_AON_IP_ADDRESS_2_0		0x30000
 
@@ -101,7 +100,8 @@
 
 #define SOC15_DPG_MODE_OFFSET(ip, inst_idx, reg) 						\
 	({											\
-		uint32_t internal_reg_offset, addr;						\
+		/* To avoid a -Wunused-but-set-variable warning. */				\
+		uint32_t internal_reg_offset __maybe_unused, addr;				\
 		bool video_range, video1_range, aon_range, aon1_range;				\
 												\
 		addr = (adev->reg_offset[ip##_HWIP][inst_idx][reg##_BASE_IDX] + reg);		\
@@ -141,21 +141,80 @@
 		RREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_DATA);				\
 	})
 
-#define WREG32_SOC15_DPG_MODE(inst_idx, offset, value, mask_en, indirect)			\
-	do {											\
-		if (!indirect) {								\
-			WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_DATA, value);			\
-			WREG32_SOC15(VCN, inst_idx, mmUVD_DPG_LMA_CTL, 				\
-				(0x1 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT |			\
-				 mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT |			\
-				 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT));		\
-		} else {									\
-			*adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = offset;		\
-			*adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ = value;			\
-		}										\
+#define WREG32_SOC15_DPG_MODE(inst_idx, offset, value, mask_en, indirect)             \
+	do {                                                                          \
+		if (!indirect) {                                                      \
+			WREG32_SOC15(VCN, GET_INST(VCN, inst_idx),                    \
+				     mmUVD_DPG_LMA_DATA, value);                      \
+			WREG32_SOC15(                                                 \
+				VCN, GET_INST(VCN, inst_idx),                         \
+				mmUVD_DPG_LMA_CTL,                                    \
+				(0x1 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT |          \
+				 mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT |         \
+				 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT)); \
+		} else {                                                              \
+			*adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ =              \
+				offset;                                               \
+			*adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ =              \
+				value;                                                \
+		}                                                                     \
+	} while (0)
+
+#define SOC24_DPG_MODE_OFFSET(ip, inst_idx, reg)						\
+	({											\
+		/* To avoid a -Wunused-but-set-variable warning. */				\
+		uint32_t internal_reg_offset __maybe_unused, addr;				\
+		bool video_range, video1_range, aon_range, aon1_range;				\
+												\
+		addr = (adev->reg_offset[ip##_HWIP][inst_idx][reg##_BASE_IDX] + reg);		\
+		addr <<= 2;									\
+		video_range = ((((0xFFFFF & addr) >= (VCN_VID_SOC_ADDRESS)) &&			\
+				((0xFFFFF & addr) < ((VCN_VID_SOC_ADDRESS + 0x2600)))));	\
+		video1_range = ((((0xFFFFF & addr) >= (VCN1_VID_SOC_ADDRESS)) &&		\
+				((0xFFFFF & addr) < ((VCN1_VID_SOC_ADDRESS + 0x2600)))));	\
+		aon_range   = ((((0xFFFFF & addr) >= (VCN_AON_SOC_ADDRESS)) &&			\
+				((0xFFFFF & addr) < ((VCN_AON_SOC_ADDRESS + 0x600)))));		\
+		aon1_range   = ((((0xFFFFF & addr) >= (VCN1_AON_SOC_ADDRESS)) &&		\
+				((0xFFFFF & addr) < ((VCN1_AON_SOC_ADDRESS + 0x600)))));	\
+		if (video_range)								\
+			internal_reg_offset = ((0xFFFFF & addr) - (VCN_VID_SOC_ADDRESS) +	\
+				(VCN_VID_IP_ADDRESS));						\
+		else if (aon_range)								\
+			internal_reg_offset = ((0xFFFFF & addr) - (VCN_AON_SOC_ADDRESS) +	\
+				(VCN_AON_IP_ADDRESS));						\
+		else if (video1_range)								\
+			internal_reg_offset = ((0xFFFFF & addr) - (VCN1_VID_SOC_ADDRESS) +	\
+				(VCN_VID_IP_ADDRESS));						\
+		else if (aon1_range)								\
+			internal_reg_offset = ((0xFFFFF & addr) - (VCN1_AON_SOC_ADDRESS) +	\
+				(VCN_AON_IP_ADDRESS));						\
+		else										\
+			internal_reg_offset = (0xFFFFF & addr);					\
+												\
+		internal_reg_offset >>= 2;							\
+	})
+
+#define WREG32_SOC24_DPG_MODE(inst_idx, offset, value, mask_en, indirect)		\
+	do {										\
+		if (!indirect) {							\
+			WREG32_SOC15(VCN, GET_INST(VCN, inst_idx),			\
+				     regUVD_DPG_LMA_DATA, value);			\
+			WREG32_SOC15(							\
+				VCN, GET_INST(VCN, inst_idx),				\
+				regUVD_DPG_LMA_CTL,					\
+				(0x1 << UVD_DPG_LMA_CTL__READ_WRITE__SHIFT |		\
+				 mask_en << UVD_DPG_LMA_CTL__MASK_EN__SHIFT |		\
+				 offset << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT));	\
+		} else {								\
+			*adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ =		\
+				offset;							\
+			*adev->vcn.inst[inst_idx].dpg_sram_curr_addr++ =		\
+				value;							\
+		}									\
 	} while (0)
 
 #define AMDGPU_FW_SHARED_FLAG_0_UNIFIED_QUEUE (1 << 2)
+#define AMDGPU_FW_SHARED_FLAG_0_DRM_KEY_INJECT (1 << 4)
 #define AMDGPU_VCN_FW_SHARED_FLAG_0_RB	(1 << 6)
 #define AMDGPU_VCN_MULTI_QUEUE_FLAG	(1 << 8)
 #define AMDGPU_VCN_SW_RING_FLAG		(1 << 9)
@@ -163,6 +222,9 @@
 #define AMDGPU_VCN_SMU_VERSION_INFO_FLAG (1 << 11)
 #define AMDGPU_VCN_SMU_DPM_INTERFACE_FLAG (1 << 11)
 #define AMDGPU_VCN_VF_RB_SETUP_FLAG (1 << 14)
+#define AMDGPU_VCN_VF_RB_DECOUPLE_FLAG (1 << 15)
+
+#define MAX_NUM_VCN_RB_SETUP 4
 
 #define AMDGPU_VCN_IB_FLAG_DECODE_BUFFER	0x00000001
 #define AMDGPU_VCN_CMD_FLAG_MSG_BUFFER		0x00000001
@@ -175,6 +237,16 @@
 #define AMDGPU_VCN_SMU_DPM_INTERFACE_DGPU (0)
 #define AMDGPU_VCN_SMU_DPM_INTERFACE_APU (1)
 
+#define AMDGPU_DRM_KEY_INJECT_WORKAROUND_VCNFW_ASD_HANDSHAKING 2
+
+struct amdgpu_hwip_reg_entry;
+
+enum amdgpu_vcn_caps {
+	AMDGPU_VCN_RRMT_ENABLED,
+};
+
+#define AMDGPU_VCN_CAPS(caps) BIT(AMDGPU_VCN_##caps)
+
 enum fw_queue_mode {
 	FW_QUEUE_RING_RESET = 1,
 	FW_QUEUE_DPG_HOLD_OFF = 2,
@@ -226,6 +298,8 @@ struct amdgpu_vcn_fw_shared {
 };
 
 struct amdgpu_vcn_inst {
+	struct amdgpu_device	*adev;
+	int			inst;
 	struct amdgpu_bo	*vcpu_bo;
 	void			*cpu_addr;
 	uint64_t		gpu_addr;
@@ -234,6 +308,7 @@ struct amdgpu_vcn_inst {
 	struct amdgpu_ring	ring_enc[AMDGPU_VCN_MAX_ENC_RINGS];
 	atomic_t		sched_score;
 	struct amdgpu_irq_src	irq;
+	struct amdgpu_irq_src	ras_poison_irq;
 	struct amdgpu_vcn_reg	external;
 	struct amdgpu_bo	*dpg_sram_bo;
 	struct dpg_pause_state	pause_state;
@@ -242,6 +317,26 @@ struct amdgpu_vcn_inst {
 	uint32_t		*dpg_sram_curr_addr;
 	atomic_t		dpg_enc_submission_cnt;
 	struct amdgpu_vcn_fw_shared fw_shared;
+	uint8_t			aid_id;
+	const struct firmware	*fw; /* VCN firmware */
+	uint8_t			vcn_config;
+	uint32_t		vcn_codec_disable_mask;
+	atomic_t		total_submission_cnt;
+	struct mutex		vcn_pg_lock;
+	enum amd_powergating_state cur_state;
+	struct delayed_work	idle_work;
+	unsigned		fw_version;
+	unsigned		num_enc_rings;
+	bool			indirect_sram;
+	struct amdgpu_vcn_reg	 internal;
+	struct mutex		vcn1_jpeg1_workaround;
+	int (*pause_dpg_mode)(struct amdgpu_vcn_inst *vinst,
+			      struct dpg_pause_state *new_state);
+	int (*set_pg_state)(struct amdgpu_vcn_inst *vinst,
+			    enum amd_powergating_state state);
+	int (*reset)(struct amdgpu_vcn_inst *vinst);
+	bool using_unified_queue;
+	struct mutex		engine_reset_mutex;
 };
 
 struct amdgpu_vcn_ras {
@@ -249,28 +344,30 @@ struct amdgpu_vcn_ras {
 };
 
 struct amdgpu_vcn {
-	unsigned		fw_version;
-	struct delayed_work	idle_work;
-	const struct firmware	*fw;	/* VCN firmware */
-	unsigned		num_enc_rings;
-	enum amd_powergating_state cur_state;
-	bool			indirect_sram;
-
 	uint8_t	num_vcn_inst;
 	struct amdgpu_vcn_inst	 inst[AMDGPU_MAX_VCN_INSTANCES];
-	uint8_t			 vcn_config[AMDGPU_MAX_VCN_INSTANCES];
-	uint32_t		 vcn_codec_disable_mask[AMDGPU_MAX_VCN_INSTANCES];
-	struct amdgpu_vcn_reg	 internal;
-	struct mutex		 vcn_pg_lock;
-	struct mutex		vcn1_jpeg1_workaround;
-	atomic_t		 total_submission_cnt;
 
 	unsigned	harvest_config;
-	int (*pause_dpg_mode)(struct amdgpu_device *adev,
-		int inst_idx, struct dpg_pause_state *new_state);
 
 	struct ras_common_if    *ras_if;
 	struct amdgpu_vcn_ras   *ras;
+
+	uint16_t inst_mask;
+	uint8_t	num_inst_per_aid;
+
+	/* IP reg dump */
+	uint32_t		*ip_dump;
+
+	uint32_t		supported_reset;
+	uint32_t		caps;
+
+	bool			per_inst_fw;
+	unsigned		fw_version;
+
+	bool			workload_profile_active;
+	struct mutex            workload_profile_mutex;
+	u32 reg_count;
+	const struct amdgpu_hwip_reg_entry *reg_list;
 };
 
 struct amdgpu_fw_shared_rb_ptrs_struct {
@@ -322,15 +419,40 @@ struct amdgpu_fw_shared {
 	struct amdgpu_fw_shared_smu_interface_info smu_interface_info;
 };
 
+struct amdgpu_vcn_rb_setup_info {
+	uint32_t  rb_addr_lo;
+	uint32_t  rb_addr_hi;
+	uint32_t  rb_size;
+};
+
 struct amdgpu_fw_shared_rb_setup {
 	uint32_t is_rb_enabled_flags;
-	uint32_t rb_addr_lo;
-	uint32_t rb_addr_hi;
-	uint32_t  rb_size;
-	uint32_t  rb4_addr_lo;
-	uint32_t  rb4_addr_hi;
-	uint32_t  rb4_size;
-	uint32_t  reserved[6];
+
+	union {
+		struct {
+			uint32_t rb_addr_lo;
+			uint32_t rb_addr_hi;
+			uint32_t  rb_size;
+			uint32_t  rb4_addr_lo;
+			uint32_t  rb4_addr_hi;
+			uint32_t  rb4_size;
+			uint32_t  reserved[6];
+		};
+
+		struct {
+			struct amdgpu_vcn_rb_setup_info rb_info[MAX_NUM_VCN_RB_SETUP];
+		};
+	};
+};
+
+struct amdgpu_fw_shared_drm_key_wa {
+	uint8_t  method;
+	uint8_t  reserved[3];
+};
+
+struct amdgpu_fw_shared_queue_decouple {
+	uint8_t  is_enabled;
+	uint8_t  reserved[7];
 };
 
 struct amdgpu_vcn4_fw_shared {
@@ -342,6 +464,9 @@ struct amdgpu_vcn4_fw_shared {
 	uint8_t pad2[20];
 	struct amdgpu_fw_shared_rb_setup rb_setup;
 	struct amdgpu_fw_shared_smu_interface_info smu_dpm_interface;
+	struct amdgpu_fw_shared_drm_key_wa drm_key_wa;
+	uint8_t pad3[9];
+	struct amdgpu_fw_shared_queue_decouple decouple;
 };
 
 struct amdgpu_vcn_fwlog {
@@ -359,6 +484,28 @@ struct amdgpu_vcn_decode_buffer {
 	uint32_t pad[30];
 };
 
+struct amdgpu_vcn_rb_metadata {
+	uint32_t size;
+	uint32_t present_flag_0;
+
+	uint8_t version;
+	uint8_t ring_id;
+	uint8_t pad[26];
+};
+
+struct amdgpu_vcn5_fw_shared {
+	uint32_t present_flag_0;
+	uint8_t pad[12];
+	struct amdgpu_fw_shared_unified_queue_struct sq;
+	uint8_t pad1[8];
+	struct amdgpu_fw_shared_fw_logging fw_log;
+	uint8_t pad2[20];
+	struct amdgpu_fw_shared_rb_setup rb_setup;
+	struct amdgpu_fw_shared_smu_interface_info smu_dpm_interface;
+	struct amdgpu_fw_shared_drm_key_wa drm_key_wa;
+	uint8_t pad3[404];
+};
+
 #define VCN_BLOCK_ENCODE_DISABLE_MASK 0x80
 #define VCN_BLOCK_DECODE_DISABLE_MASK 0x40
 #define VCN_BLOCK_QUEUE_DISABLE_MASK 0xC0
@@ -369,11 +516,11 @@ enum vcn_ring_type {
 	VCN_UNIFIED_RING,
 };
 
-int amdgpu_vcn_early_init(struct amdgpu_device *adev);
-int amdgpu_vcn_sw_init(struct amdgpu_device *adev);
-int amdgpu_vcn_sw_fini(struct amdgpu_device *adev);
-int amdgpu_vcn_suspend(struct amdgpu_device *adev);
-int amdgpu_vcn_resume(struct amdgpu_device *adev);
+int amdgpu_vcn_early_init(struct amdgpu_device *adev, int i);
+int amdgpu_vcn_sw_init(struct amdgpu_device *adev, int i);
+void amdgpu_vcn_sw_fini(struct amdgpu_device *adev, int i);
+int amdgpu_vcn_suspend(struct amdgpu_device *adev, int i);
+int amdgpu_vcn_resume(struct amdgpu_device *adev, int i);
 void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring);
 void amdgpu_vcn_ring_end_use(struct amdgpu_ring *ring);
 
@@ -391,7 +538,7 @@ int amdgpu_vcn_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout);
 
 enum amdgpu_ring_priority_level amdgpu_vcn_get_enc_ring_prio(int ring);
 
-void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev);
+void amdgpu_vcn_setup_ucode(struct amdgpu_device *adev, int i);
 
 void amdgpu_vcn_fwlog_init(struct amdgpu_vcn_inst *vcn);
 void amdgpu_debugfs_vcn_fwlog_init(struct amdgpu_device *adev,
@@ -400,6 +547,27 @@ void amdgpu_debugfs_vcn_fwlog_init(struct amdgpu_device *adev,
 int amdgpu_vcn_process_poison_irq(struct amdgpu_device *adev,
 			struct amdgpu_irq_src *source,
 			struct amdgpu_iv_entry *entry);
+int amdgpu_vcn_ras_late_init(struct amdgpu_device *adev,
+			struct ras_common_if *ras_block);
 int amdgpu_vcn_ras_sw_init(struct amdgpu_device *adev);
 
+int amdgpu_vcn_psp_update_sram(struct amdgpu_device *adev, int inst_idx,
+			       enum AMDGPU_UCODE_ID ucode_id);
+int amdgpu_vcn_save_vcpu_bo(struct amdgpu_device *adev);
+int amdgpu_vcn_sysfs_reset_mask_init(struct amdgpu_device *adev);
+void amdgpu_vcn_sysfs_reset_mask_fini(struct amdgpu_device *adev);
+void amdgpu_debugfs_vcn_sched_mask_init(struct amdgpu_device *adev);
+
+int vcn_set_powergating_state(struct amdgpu_ip_block *ip_block,
+			      enum amd_powergating_state state);
+int amdgpu_vcn_ring_reset(struct amdgpu_ring *ring,
+			  unsigned int vmid,
+			  struct amdgpu_fence *guilty_fence);
+int amdgpu_vcn_reg_dump_init(struct amdgpu_device *adev,
+			     const struct amdgpu_hwip_reg_entry *reg, u32 count);
+void amdgpu_vcn_dump_ip_state(struct amdgpu_ip_block *ip_block);
+void amdgpu_vcn_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p);
+void amdgpu_vcn_get_profile(struct amdgpu_device *adev);
+void amdgpu_vcn_put_profile(struct amdgpu_device *adev);
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
index f2e2cbaa7fde..47a6ce4fdc74 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
@@ -32,6 +32,8 @@
 
 #include "amdgpu.h"
 #include "amdgpu_ras.h"
+#include "amdgpu_reset.h"
+#include "amdgpu_dpm.h"
 #include "vi.h"
 #include "soc15.h"
 #include "nv.h"
@@ -42,6 +44,18 @@
 		vf2pf_info->ucode_info[ucode].version = ver; \
 	} while (0)
 
+#define mmRCC_CONFIG_MEMSIZE    0xde3
+
+const char *amdgpu_virt_dynamic_crit_table_name[] = {
+	"IP DISCOVERY",
+	"VBIOS IMG",
+	"RAS TELEMETRY",
+	"DATA EXCHANGE",
+	"BAD PAGE INFO",
+	"INIT HEADER",
+	"LAST",
+};
+
 bool amdgpu_virt_mmio_blocked(struct amdgpu_device *adev)
 {
 	/* By now all MMIO pages except mailbox are blocked */
@@ -56,7 +70,8 @@ void amdgpu_virt_init_setting(struct amdgpu_device *adev)
 
 	/* enable virtual display */
 	if (adev->asic_type != CHIP_ALDEBARAN &&
-	    adev->asic_type != CHIP_ARCTURUS) {
+	    adev->asic_type != CHIP_ARCTURUS &&
+	    ((adev->pdev->class >> 8) != PCI_CLASS_ACCELERATOR_PROCESSING)) {
 		if (adev->mode_info.num_crtc == 0)
 			adev->mode_info.num_crtc = 1;
 		adev->enable_virtual_display = true;
@@ -65,61 +80,9 @@ void amdgpu_virt_init_setting(struct amdgpu_device *adev)
 	adev->cg_flags = 0;
 	adev->pg_flags = 0;
 
-	/* enable mcbp for sriov asic_type before soc21 */
-	amdgpu_mcbp = (adev->asic_type < CHIP_IP_DISCOVERY) ? 1 : 0;
-
-}
-
-void amdgpu_virt_kiq_reg_write_reg_wait(struct amdgpu_device *adev,
-					uint32_t reg0, uint32_t reg1,
-					uint32_t ref, uint32_t mask)
-{
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-	struct amdgpu_ring *ring = &kiq->ring;
-	signed long r, cnt = 0;
-	unsigned long flags;
-	uint32_t seq;
-
-	if (adev->mes.ring.sched.ready) {
-		amdgpu_mes_reg_write_reg_wait(adev, reg0, reg1,
-					      ref, mask);
-		return;
-	}
-
-	spin_lock_irqsave(&kiq->ring_lock, flags);
-	amdgpu_ring_alloc(ring, 32);
-	amdgpu_ring_emit_reg_write_reg_wait(ring, reg0, reg1,
-					    ref, mask);
-	r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
-	if (r)
-		goto failed_undo;
-
-	amdgpu_ring_commit(ring);
-	spin_unlock_irqrestore(&kiq->ring_lock, flags);
-
-	r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
-
-	/* don't wait anymore for IRQ context */
-	if (r < 1 && in_interrupt())
-		goto failed_kiq;
-
-	might_sleep();
-	while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
-
-		msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
-		r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
-	}
-
-	if (cnt > MAX_KIQ_REG_TRY)
-		goto failed_kiq;
-
-	return;
-
-failed_undo:
-	amdgpu_ring_undo(ring);
-	spin_unlock_irqrestore(&kiq->ring_lock, flags);
-failed_kiq:
-	dev_err(adev->dev, "failed to write reg %x wait reg %x\n", reg0, reg1);
+	/* Reduce kcq number to 2 to reduce latency */
+	if (amdgpu_num_kcq == -1)
+		amdgpu_num_kcq = 2;
 }
 
 /**
@@ -136,8 +99,10 @@ int amdgpu_virt_request_full_gpu(struct amdgpu_device *adev, bool init)
 
 	if (virt->ops && virt->ops->req_full_gpu) {
 		r = virt->ops->req_full_gpu(adev, init);
-		if (r)
+		if (r) {
+			adev->no_hw_access = true;
 			return r;
+		}
 
 		adev->virt.caps &= ~AMDGPU_SRIOV_CAPS_RUNTIME;
 	}
@@ -197,9 +162,24 @@ void amdgpu_virt_request_init_data(struct amdgpu_device *adev)
 		virt->ops->req_init_data(adev);
 
 	if (adev->virt.req_init_data_ver > 0)
-		DRM_INFO("host supports REQ_INIT_DATA handshake\n");
+		dev_info(adev->dev, "host supports REQ_INIT_DATA handshake of critical_region_version %d\n",
+				 adev->virt.req_init_data_ver);
 	else
-		DRM_WARN("host doesn't support REQ_INIT_DATA handshake\n");
+		dev_warn(adev->dev, "host doesn't support REQ_INIT_DATA handshake\n");
+}
+
+/**
+ * amdgpu_virt_ready_to_reset() - send ready to reset to host
+ * @adev:	amdgpu device.
+ * Send ready to reset message to GPU hypervisor to signal we have stopped GPU
+ * activity and is ready for host FLR
+ */
+void amdgpu_virt_ready_to_reset(struct amdgpu_device *adev)
+{
+	struct amdgpu_virt *virt = &adev->virt;
+
+	if (virt->ops && virt->ops->reset_gpu)
+		virt->ops->ready_to_reset(adev);
 }
 
 /**
@@ -238,12 +218,12 @@ int amdgpu_virt_alloc_mm_table(struct amdgpu_device *adev)
 				    &adev->virt.mm_table.gpu_addr,
 				    (void *)&adev->virt.mm_table.cpu_addr);
 	if (r) {
-		DRM_ERROR("failed to alloc mm table and error = %d.\n", r);
+		dev_err(adev->dev, "failed to alloc mm table and error = %d.\n", r);
 		return r;
 	}
 
 	memset((void *)adev->virt.mm_table.cpu_addr, 0, PAGE_SIZE);
-	DRM_INFO("MM table gpu addr = 0x%llx, cpu addr = %p.\n",
+	dev_info(adev->dev, "MM table gpu addr = 0x%llx, cpu addr = %p.\n",
 		 adev->virt.mm_table.gpu_addr,
 		 adev->virt.mm_table.cpu_addr);
 	return 0;
@@ -265,6 +245,22 @@ void amdgpu_virt_free_mm_table(struct amdgpu_device *adev)
 	adev->virt.mm_table.gpu_addr = 0;
 }
 
+/**
+ * amdgpu_virt_rcvd_ras_interrupt() - receive ras interrupt
+ * @adev:	amdgpu device.
+ * Check whether host sent RAS error message
+ * Return: true if found, otherwise false
+ */
+bool amdgpu_virt_rcvd_ras_interrupt(struct amdgpu_device *adev)
+{
+	struct amdgpu_virt *virt = &adev->virt;
+
+	if (!virt->ops || !virt->ops->rcvd_ras_intr)
+		return false;
+
+	return virt->ops->rcvd_ras_intr(adev);
+}
+
 
 unsigned int amd_sriov_msg_checksum(void *obj,
 				unsigned long obj_size,
@@ -301,11 +297,11 @@ static int amdgpu_virt_init_ras_err_handler_data(struct amdgpu_device *adev)
 	if (!*data)
 		goto data_failure;
 
-	bps = kmalloc_array(align_space, sizeof((*data)->bps), GFP_KERNEL);
+	bps = kmalloc_array(align_space, sizeof(*(*data)->bps), GFP_KERNEL);
 	if (!bps)
 		goto bps_failure;
 
-	bps_bo = kmalloc_array(align_space, sizeof((*data)->bps_bo), GFP_KERNEL);
+	bps_bo = kmalloc_array(align_space, sizeof(*(*data)->bps_bo), GFP_KERNEL);
 	if (!bps_bo)
 		goto bps_bo_failure;
 
@@ -338,8 +334,10 @@ static void amdgpu_virt_ras_release_bp(struct amdgpu_device *adev)
 
 	for (i = data->last_reserved - 1; i >= 0; i--) {
 		bo = data->bps_bo[i];
-		amdgpu_bo_free_kernel(&bo, NULL, NULL);
-		data->bps_bo[i] = bo;
+		if (bo) {
+			amdgpu_bo_free_kernel(&bo, NULL, NULL);
+			data->bps_bo[i] = bo;
+		}
 		data->last_reserved = i;
 	}
 }
@@ -379,6 +377,8 @@ static void amdgpu_virt_ras_reserve_bps(struct amdgpu_device *adev)
 {
 	struct amdgpu_virt *virt = &adev->virt;
 	struct amdgpu_virt_ras_err_handler_data *data = virt->virt_eh_data;
+	struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr;
+	struct ttm_resource_manager *man = &mgr->manager;
 	struct amdgpu_bo *bo = NULL;
 	uint64_t bp;
 	int i;
@@ -394,12 +394,20 @@ static void amdgpu_virt_ras_reserve_bps(struct amdgpu_device *adev)
 		 * 2) a ras bad page has been reserved (duplicate error injection
 		 *    for one page);
 		 */
-		if (amdgpu_bo_create_kernel_at(adev, bp << AMDGPU_GPU_PAGE_SHIFT,
-					       AMDGPU_GPU_PAGE_SIZE,
-					       &bo, NULL))
-			DRM_DEBUG("RAS WARN: reserve vram for retired page %llx fail\n", bp);
-
-		data->bps_bo[i] = bo;
+		if  (ttm_resource_manager_used(man)) {
+			amdgpu_vram_mgr_reserve_range(&adev->mman.vram_mgr,
+				bp << AMDGPU_GPU_PAGE_SHIFT,
+				AMDGPU_GPU_PAGE_SIZE);
+			data->bps_bo[i] = NULL;
+		} else {
+			if (amdgpu_bo_create_kernel_at(adev, bp << AMDGPU_GPU_PAGE_SHIFT,
+							AMDGPU_GPU_PAGE_SIZE,
+							&bo, NULL))
+				dev_dbg(adev->dev,
+						"RAS WARN: reserve vram for retired page %llx fail\n",
+						bp);
+			data->bps_bo[i] = bo;
+		}
 		data->last_reserved = i + 1;
 		bo = NULL;
 	}
@@ -435,6 +443,8 @@ static void amdgpu_virt_add_bad_page(struct amdgpu_device *adev,
 	else
 		vram_usage_va = adev->mman.drv_vram_usage_va;
 
+	memset(&bp, 0, sizeof(bp));
+
 	if (bp_block_size) {
 		bp_cnt = bp_block_size / sizeof(uint64_t);
 		for (bp_idx = 0; bp_idx < bp_cnt; bp_idx++) {
@@ -465,7 +475,7 @@ static int amdgpu_virt_read_pf2vf_data(struct amdgpu_device *adev)
 		return -EINVAL;
 
 	if (pf2vf_info->size > 1024) {
-		DRM_ERROR("invalid pf2vf message size\n");
+		dev_err(adev->dev, "invalid pf2vf message size: 0x%x\n", pf2vf_info->size);
 		return -EINVAL;
 	}
 
@@ -476,7 +486,9 @@ static int amdgpu_virt_read_pf2vf_data(struct amdgpu_device *adev)
 			adev->virt.fw_reserve.p_pf2vf, pf2vf_info->size,
 			adev->virt.fw_reserve.checksum_key, checksum);
 		if (checksum != checkval) {
-			DRM_ERROR("invalid pf2vf message\n");
+			dev_err(adev->dev,
+				"invalid pf2vf message: header checksum=0x%x calculated checksum=0x%x\n",
+				checksum, checkval);
 			return -EINVAL;
 		}
 
@@ -490,7 +502,9 @@ static int amdgpu_virt_read_pf2vf_data(struct amdgpu_device *adev)
 			adev->virt.fw_reserve.p_pf2vf, pf2vf_info->size,
 			0, checksum);
 		if (checksum != checkval) {
-			DRM_ERROR("invalid pf2vf message\n");
+			dev_err(adev->dev,
+				"invalid pf2vf message: header checksum=0x%x calculated checksum=0x%x\n",
+				checksum, checkval);
 			return -EINVAL;
 		}
 
@@ -519,14 +533,17 @@ static int amdgpu_virt_read_pf2vf_data(struct amdgpu_device *adev)
 			tmp = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->mm_bw_management[i].encode_max_frame_pixels;
 			adev->virt.encode_max_frame_pixels = max(tmp, adev->virt.encode_max_frame_pixels);
 		}
-		if((adev->virt.decode_max_dimension_pixels > 0) || (adev->virt.encode_max_dimension_pixels > 0))
+		if ((adev->virt.decode_max_dimension_pixels > 0) || (adev->virt.encode_max_dimension_pixels > 0))
 			adev->virt.is_mm_bw_enabled = true;
 
 		adev->unique_id =
 			((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->uuid;
+		adev->virt.ras_en_caps.all = ((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->ras_en_caps.all;
+		adev->virt.ras_telemetry_en_caps.all =
+			((struct amd_sriov_msg_pf2vf_info *)pf2vf_info)->ras_telemetry_en_caps.all;
 		break;
 	default:
-		DRM_ERROR("invalid pf2vf version\n");
+		dev_err(adev->dev, "invalid pf2vf version: 0x%x\n", pf2vf_info->version);
 		return -EINVAL;
 	}
 
@@ -557,7 +574,6 @@ static void amdgpu_virt_populate_vf2pf_ucode_info(struct amdgpu_device *adev)
 	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_RLC_SRLS, adev->gfx.rlc_srls_fw_version);
 	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_MEC,      adev->gfx.mec_fw_version);
 	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_MEC2,     adev->gfx.mec2_fw_version);
-	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_IMU,      adev->gfx.imu_fw_version);
 	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_SOS,      adev->psp.sos.fw_version);
 	POPULATE_UCODE_INFO(vf2pf_info, AMD_SRIOV_UCODE_ID_ASD,
 			    adev->psp.asd_context.bin_desc.fw_version);
@@ -597,8 +613,8 @@ static int amdgpu_virt_write_vf2pf_data(struct amdgpu_device *adev)
 	vf2pf_info->driver_cert = 0;
 	vf2pf_info->os_info.all = 0;
 
-	vf2pf_info->fb_usage =
-		ttm_resource_manager_usage(&adev->mman.vram_mgr.manager) >> 20;
+	vf2pf_info->fb_usage = ttm_resource_manager_used(&adev->mman.vram_mgr.manager) ?
+		 ttm_resource_manager_usage(&adev->mman.vram_mgr.manager) >> 20 : 0;
 	vf2pf_info->fb_vis_usage =
 		amdgpu_vram_mgr_vis_usage(&adev->mman.vram_mgr) >> 20;
 	vf2pf_info->fb_size = adev->gmc.real_vram_size >> 20;
@@ -613,9 +629,15 @@ static int amdgpu_virt_write_vf2pf_data(struct amdgpu_device *adev)
 	vf2pf_info->decode_usage = 0;
 
 	vf2pf_info->dummy_page_addr = (uint64_t)adev->dummy_page_addr;
+	if (amdgpu_sriov_is_mes_info_enable(adev)) {
+		vf2pf_info->mes_info_addr =
+			(uint64_t)(adev->mes.resource_1_gpu_addr[0] + AMDGPU_GPU_PAGE_SIZE);
+		vf2pf_info->mes_info_size =
+			adev->mes.resource_1[0]->tbo.base.size - AMDGPU_GPU_PAGE_SIZE;
+	}
 	vf2pf_info->checksum =
 		amd_sriov_msg_checksum(
-		vf2pf_info, vf2pf_info->header.size, 0, 0);
+		vf2pf_info, sizeof(*vf2pf_info), 0, 0);
 
 	return 0;
 }
@@ -626,18 +648,59 @@ static void amdgpu_virt_update_vf2pf_work_item(struct work_struct *work)
 	int ret;
 
 	ret = amdgpu_virt_read_pf2vf_data(adev);
-	if (ret)
+	if (ret) {
+		adev->virt.vf2pf_update_retry_cnt++;
+
+		if ((amdgpu_virt_rcvd_ras_interrupt(adev) ||
+			adev->virt.vf2pf_update_retry_cnt >= AMDGPU_VF2PF_UPDATE_MAX_RETRY_LIMIT) &&
+			amdgpu_sriov_runtime(adev)) {
+
+			amdgpu_ras_set_fed(adev, true);
+			if (amdgpu_reset_domain_schedule(adev->reset_domain,
+							&adev->kfd.reset_work))
+				return;
+			else
+				dev_err(adev->dev, "Failed to queue work! at %s", __func__);
+		}
+
 		goto out;
+	}
+
+	adev->virt.vf2pf_update_retry_cnt = 0;
 	amdgpu_virt_write_vf2pf_data(adev);
 
 out:
 	schedule_delayed_work(&(adev->virt.vf2pf_work), adev->virt.vf2pf_update_interval_ms);
 }
 
+static int amdgpu_virt_read_exchange_data_from_mem(struct amdgpu_device *adev, uint32_t *pfvf_data)
+{
+	uint32_t dataexchange_offset =
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID].offset;
+	uint32_t dataexchange_size =
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID].size_kb << 10;
+	uint64_t pos = 0;
+
+	dev_info(adev->dev,
+			"Got data exchange info from dynamic crit_region_table at offset 0x%x with size of 0x%x bytes.\n",
+			dataexchange_offset, dataexchange_size);
+
+	if (!IS_ALIGNED(dataexchange_offset, 4) || !IS_ALIGNED(dataexchange_size, 4)) {
+		dev_err(adev->dev, "Data exchange data not aligned to 4 bytes\n");
+		return -EINVAL;
+	}
+
+	pos = (uint64_t)dataexchange_offset;
+	amdgpu_device_vram_access(adev, pos, pfvf_data,
+			dataexchange_size, false);
+
+	return 0;
+}
+
 void amdgpu_virt_fini_data_exchange(struct amdgpu_device *adev)
 {
 	if (adev->virt.vf2pf_update_interval_ms != 0) {
-		DRM_INFO("clean up the vf2pf work item\n");
+		dev_info(adev->dev, "clean up the vf2pf work item\n");
 		cancel_delayed_work_sync(&adev->virt.vf2pf_work);
 		adev->virt.vf2pf_update_interval_ms = 0;
 	}
@@ -645,12 +708,15 @@ void amdgpu_virt_fini_data_exchange(struct amdgpu_device *adev)
 
 void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev)
 {
+	uint32_t *pfvf_data = NULL;
+
 	adev->virt.fw_reserve.p_pf2vf = NULL;
 	adev->virt.fw_reserve.p_vf2pf = NULL;
 	adev->virt.vf2pf_update_interval_ms = 0;
+	adev->virt.vf2pf_update_retry_cnt = 0;
 
 	if (adev->mman.fw_vram_usage_va && adev->mman.drv_vram_usage_va) {
-		DRM_WARN("Currently fw_vram and drv_vram should not have values at the same time!");
+		dev_warn(adev->dev, "Currently fw_vram and drv_vram should not have values at the same time!");
 	} else if (adev->mman.fw_vram_usage_va || adev->mman.drv_vram_usage_va) {
 		/* go through this logic in ip_init and reset to init workqueue*/
 		amdgpu_virt_exchange_data(adev);
@@ -659,11 +725,34 @@ void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev)
 		schedule_delayed_work(&(adev->virt.vf2pf_work), msecs_to_jiffies(adev->virt.vf2pf_update_interval_ms));
 	} else if (adev->bios != NULL) {
 		/* got through this logic in early init stage to get necessary flags, e.g. rlcg_acc related*/
-		adev->virt.fw_reserve.p_pf2vf =
-			(struct amd_sriov_msg_pf2vf_info_header *)
-			(adev->bios + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB << 10));
+		if (adev->virt.req_init_data_ver == GPU_CRIT_REGION_V2) {
+			pfvf_data =
+				kzalloc(adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID].size_kb << 10,
+					GFP_KERNEL);
+			if (!pfvf_data) {
+				dev_err(adev->dev, "Failed to allocate memory for pfvf_data\n");
+				return;
+			}
 
-		amdgpu_virt_read_pf2vf_data(adev);
+			if (amdgpu_virt_read_exchange_data_from_mem(adev, pfvf_data))
+				goto free_pfvf_data;
+
+			adev->virt.fw_reserve.p_pf2vf =
+				(struct amd_sriov_msg_pf2vf_info_header *)pfvf_data;
+
+			amdgpu_virt_read_pf2vf_data(adev);
+
+free_pfvf_data:
+			kfree(pfvf_data);
+			pfvf_data = NULL;
+			adev->virt.fw_reserve.p_pf2vf = NULL;
+		} else {
+			adev->virt.fw_reserve.p_pf2vf =
+				(struct amd_sriov_msg_pf2vf_info_header *)
+				(adev->bios + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB_V1 << 10));
+
+			amdgpu_virt_read_pf2vf_data(adev);
+		}
 	}
 }
 
@@ -676,19 +765,38 @@ void amdgpu_virt_exchange_data(struct amdgpu_device *adev)
 
 	if (adev->mman.fw_vram_usage_va || adev->mman.drv_vram_usage_va) {
 		if (adev->mman.fw_vram_usage_va) {
-			adev->virt.fw_reserve.p_pf2vf =
-				(struct amd_sriov_msg_pf2vf_info_header *)
-				(adev->mman.fw_vram_usage_va + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB << 10));
-			adev->virt.fw_reserve.p_vf2pf =
-				(struct amd_sriov_msg_vf2pf_info_header *)
-				(adev->mman.fw_vram_usage_va + (AMD_SRIOV_MSG_VF2PF_OFFSET_KB << 10));
+			if (adev->virt.req_init_data_ver == GPU_CRIT_REGION_V2) {
+				adev->virt.fw_reserve.p_pf2vf =
+					(struct amd_sriov_msg_pf2vf_info_header *)
+					(adev->mman.fw_vram_usage_va +
+					adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID].offset);
+				adev->virt.fw_reserve.p_vf2pf =
+					(struct amd_sriov_msg_vf2pf_info_header *)
+					(adev->mman.fw_vram_usage_va +
+					adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID].offset +
+					(AMD_SRIOV_MSG_SIZE_KB << 10));
+				adev->virt.fw_reserve.ras_telemetry =
+					(adev->mman.fw_vram_usage_va +
+					adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_RAS_TELEMETRY_TABLE_ID].offset);
+			} else {
+				adev->virt.fw_reserve.p_pf2vf =
+					(struct amd_sriov_msg_pf2vf_info_header *)
+					(adev->mman.fw_vram_usage_va + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB_V1 << 10));
+				adev->virt.fw_reserve.p_vf2pf =
+					(struct amd_sriov_msg_vf2pf_info_header *)
+					(adev->mman.fw_vram_usage_va + (AMD_SRIOV_MSG_VF2PF_OFFSET_KB_V1 << 10));
+				adev->virt.fw_reserve.ras_telemetry =
+					(adev->mman.fw_vram_usage_va + (AMD_SRIOV_MSG_RAS_TELEMETRY_OFFSET_KB_V1 << 10));
+			}
 		} else if (adev->mman.drv_vram_usage_va) {
 			adev->virt.fw_reserve.p_pf2vf =
 				(struct amd_sriov_msg_pf2vf_info_header *)
-				(adev->mman.drv_vram_usage_va + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB << 10));
+				(adev->mman.drv_vram_usage_va + (AMD_SRIOV_MSG_PF2VF_OFFSET_KB_V1 << 10));
 			adev->virt.fw_reserve.p_vf2pf =
 				(struct amd_sriov_msg_vf2pf_info_header *)
-				(adev->mman.drv_vram_usage_va + (AMD_SRIOV_MSG_VF2PF_OFFSET_KB << 10));
+				(adev->mman.drv_vram_usage_va + (AMD_SRIOV_MSG_VF2PF_OFFSET_KB_V1 << 10));
+			adev->virt.fw_reserve.ras_telemetry =
+				(adev->mman.drv_vram_usage_va + (AMD_SRIOV_MSG_RAS_TELEMETRY_OFFSET_KB_V1 << 10));
 		}
 
 		amdgpu_virt_read_pf2vf_data(adev);
@@ -711,7 +819,7 @@ void amdgpu_virt_exchange_data(struct amdgpu_device *adev)
 	}
 }
 
-void amdgpu_detect_virtualization(struct amdgpu_device *adev)
+static u32 amdgpu_virt_init_detect_asic(struct amdgpu_device *adev)
 {
 	uint32_t reg;
 
@@ -747,14 +855,17 @@ void amdgpu_detect_virtualization(struct amdgpu_device *adev)
 			adev->virt.caps |= AMDGPU_PASSTHROUGH_MODE;
 	}
 
-	if (amdgpu_sriov_vf(adev) && adev->asic_type == CHIP_SIENNA_CICHLID)
-		/* VF MMIO access (except mailbox range) from CPU
-		 * will be blocked during sriov runtime
-		 */
-		adev->virt.caps |= AMDGPU_VF_MMIO_ACCESS_PROTECT;
+	return reg;
+}
+
+static bool amdgpu_virt_init_req_data(struct amdgpu_device *adev, u32 reg)
+{
+	bool is_sriov = false;
 
 	/* we have the ability to check now */
 	if (amdgpu_sriov_vf(adev)) {
+		is_sriov = true;
+
 		switch (adev->asic_type) {
 		case CHIP_TONGA:
 		case CHIP_FIJI:
@@ -783,10 +894,252 @@ void amdgpu_detect_virtualization(struct amdgpu_device *adev)
 			amdgpu_virt_request_init_data(adev);
 			break;
 		default: /* other chip doesn't support SRIOV */
-			DRM_ERROR("Unknown asic type: %d!\n", adev->asic_type);
+			is_sriov = false;
+			dev_err(adev->dev, "Unknown asic type: %d!\n", adev->asic_type);
 			break;
 		}
 	}
+
+	return is_sriov;
+}
+
+static void amdgpu_virt_init_ras(struct amdgpu_device *adev)
+{
+	ratelimit_state_init(&adev->virt.ras.ras_error_cnt_rs, 5 * HZ, 1);
+	ratelimit_state_init(&adev->virt.ras.ras_cper_dump_rs, 5 * HZ, 1);
+	ratelimit_state_init(&adev->virt.ras.ras_chk_criti_rs, 5 * HZ, 1);
+
+	ratelimit_set_flags(&adev->virt.ras.ras_error_cnt_rs,
+			    RATELIMIT_MSG_ON_RELEASE);
+	ratelimit_set_flags(&adev->virt.ras.ras_cper_dump_rs,
+			    RATELIMIT_MSG_ON_RELEASE);
+	ratelimit_set_flags(&adev->virt.ras.ras_chk_criti_rs,
+			    RATELIMIT_MSG_ON_RELEASE);
+
+	mutex_init(&adev->virt.ras.ras_telemetry_mutex);
+	mutex_init(&adev->virt.access_req_mutex);
+
+	adev->virt.ras.cper_rptr = 0;
+}
+
+static uint8_t amdgpu_virt_crit_region_calc_checksum(uint8_t *buf_start, uint8_t *buf_end)
+{
+	uint32_t sum = 0;
+
+	if (buf_start >= buf_end)
+		return 0;
+
+	for (; buf_start < buf_end; buf_start++)
+		sum += buf_start[0];
+
+	return 0xffffffff - sum;
+}
+
+int amdgpu_virt_init_critical_region(struct amdgpu_device *adev)
+{
+	struct amd_sriov_msg_init_data_header *init_data_hdr = NULL;
+	u64 init_hdr_offset = adev->virt.init_data_header.offset;
+	u64 init_hdr_size = (u64)adev->virt.init_data_header.size_kb << 10; /* KB → bytes */
+	u64 vram_size;
+	u64 end;
+	int r = 0;
+	uint8_t checksum = 0;
+
+	/* Skip below init if critical region version != v2 */
+	if (adev->virt.req_init_data_ver != GPU_CRIT_REGION_V2)
+		return 0;
+
+	if (init_hdr_offset < 0) {
+		dev_err(adev->dev, "Invalid init header offset\n");
+		return -EINVAL;
+	}
+
+	vram_size = RREG32(mmRCC_CONFIG_MEMSIZE);
+	if (!vram_size || vram_size == U32_MAX)
+		return -EINVAL;
+	vram_size <<= 20;
+
+	if (check_add_overflow(init_hdr_offset, init_hdr_size, &end) || end > vram_size) {
+		dev_err(adev->dev, "init_data_header exceeds VRAM size, exiting\n");
+		return -EINVAL;
+	}
+
+	/* Allocate for init_data_hdr */
+	init_data_hdr = kzalloc(sizeof(struct amd_sriov_msg_init_data_header), GFP_KERNEL);
+	if (!init_data_hdr)
+		return -ENOMEM;
+
+	amdgpu_device_vram_access(adev, (uint64_t)init_hdr_offset, (uint32_t *)init_data_hdr,
+					sizeof(struct amd_sriov_msg_init_data_header), false);
+
+	/* Table validation */
+	if (strncmp(init_data_hdr->signature,
+				AMDGPU_SRIOV_CRIT_DATA_SIGNATURE,
+				AMDGPU_SRIOV_CRIT_DATA_SIG_LEN) != 0) {
+		dev_err(adev->dev, "Invalid init data signature: %.4s\n",
+			init_data_hdr->signature);
+		r = -EINVAL;
+		goto out;
+	}
+
+	checksum = amdgpu_virt_crit_region_calc_checksum(
+			(uint8_t *)&init_data_hdr->initdata_offset,
+			(uint8_t *)init_data_hdr +
+			sizeof(struct amd_sriov_msg_init_data_header));
+	if (checksum != init_data_hdr->checksum) {
+		dev_err(adev->dev, "Found unmatching checksum from calculation 0x%x and init_data 0x%x\n",
+				checksum, init_data_hdr->checksum);
+		r = -EINVAL;
+		goto out;
+	}
+
+	memset(&adev->virt.crit_regn, 0, sizeof(adev->virt.crit_regn));
+	memset(adev->virt.crit_regn_tbl, 0, sizeof(adev->virt.crit_regn_tbl));
+
+	adev->virt.crit_regn.offset = init_data_hdr->initdata_offset;
+	adev->virt.crit_regn.size_kb = init_data_hdr->initdata_size_in_kb;
+
+	/* Validation and initialization for each table entry */
+	if (IS_SRIOV_CRIT_REGN_ENTRY_VALID(init_data_hdr, AMD_SRIOV_MSG_IPD_TABLE_ID)) {
+		if (!init_data_hdr->ip_discovery_size_in_kb ||
+				init_data_hdr->ip_discovery_size_in_kb > DISCOVERY_TMR_SIZE) {
+			dev_err(adev->dev, "Invalid %s size: 0x%x\n",
+				amdgpu_virt_dynamic_crit_table_name[AMD_SRIOV_MSG_IPD_TABLE_ID],
+				init_data_hdr->ip_discovery_size_in_kb);
+			r = -EINVAL;
+			goto out;
+		}
+
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_IPD_TABLE_ID].offset =
+			init_data_hdr->ip_discovery_offset;
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_IPD_TABLE_ID].size_kb =
+			init_data_hdr->ip_discovery_size_in_kb;
+	}
+
+	if (IS_SRIOV_CRIT_REGN_ENTRY_VALID(init_data_hdr, AMD_SRIOV_MSG_VBIOS_IMG_TABLE_ID)) {
+		if (!init_data_hdr->vbios_img_size_in_kb) {
+			dev_err(adev->dev, "Invalid %s size: 0x%x\n",
+				amdgpu_virt_dynamic_crit_table_name[AMD_SRIOV_MSG_VBIOS_IMG_TABLE_ID],
+				init_data_hdr->vbios_img_size_in_kb);
+			r = -EINVAL;
+			goto out;
+		}
+
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_VBIOS_IMG_TABLE_ID].offset =
+			init_data_hdr->vbios_img_offset;
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_VBIOS_IMG_TABLE_ID].size_kb =
+			init_data_hdr->vbios_img_size_in_kb;
+	}
+
+	if (IS_SRIOV_CRIT_REGN_ENTRY_VALID(init_data_hdr, AMD_SRIOV_MSG_RAS_TELEMETRY_TABLE_ID)) {
+		if (!init_data_hdr->ras_tele_info_size_in_kb) {
+			dev_err(adev->dev, "Invalid %s size: 0x%x\n",
+				amdgpu_virt_dynamic_crit_table_name[AMD_SRIOV_MSG_RAS_TELEMETRY_TABLE_ID],
+				init_data_hdr->ras_tele_info_size_in_kb);
+			r = -EINVAL;
+			goto out;
+		}
+
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_RAS_TELEMETRY_TABLE_ID].offset =
+			init_data_hdr->ras_tele_info_offset;
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_RAS_TELEMETRY_TABLE_ID].size_kb =
+			init_data_hdr->ras_tele_info_size_in_kb;
+	}
+
+	if (IS_SRIOV_CRIT_REGN_ENTRY_VALID(init_data_hdr, AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID)) {
+		if (!init_data_hdr->dataexchange_size_in_kb) {
+			dev_err(adev->dev, "Invalid %s size: 0x%x\n",
+				amdgpu_virt_dynamic_crit_table_name[AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID],
+				init_data_hdr->dataexchange_size_in_kb);
+			r = -EINVAL;
+			goto out;
+		}
+
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID].offset =
+			init_data_hdr->dataexchange_offset;
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID].size_kb =
+			init_data_hdr->dataexchange_size_in_kb;
+	}
+
+	if (IS_SRIOV_CRIT_REGN_ENTRY_VALID(init_data_hdr, AMD_SRIOV_MSG_BAD_PAGE_INFO_TABLE_ID)) {
+		if (!init_data_hdr->bad_page_size_in_kb) {
+			dev_err(adev->dev, "Invalid %s size: 0x%x\n",
+				amdgpu_virt_dynamic_crit_table_name[AMD_SRIOV_MSG_BAD_PAGE_INFO_TABLE_ID],
+				init_data_hdr->bad_page_size_in_kb);
+			r = -EINVAL;
+			goto out;
+		}
+
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_BAD_PAGE_INFO_TABLE_ID].offset =
+			init_data_hdr->bad_page_info_offset;
+		adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_BAD_PAGE_INFO_TABLE_ID].size_kb =
+			init_data_hdr->bad_page_size_in_kb;
+	}
+
+	/* Validation for critical region info */
+	if (adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_IPD_TABLE_ID].size_kb > DISCOVERY_TMR_SIZE) {
+		dev_err(adev->dev, "Invalid IP discovery size: 0x%x\n",
+				adev->virt.crit_regn_tbl[AMD_SRIOV_MSG_IPD_TABLE_ID].size_kb);
+		r = -EINVAL;
+		goto out;
+	}
+
+	/* reserved memory starts from crit region base offset with the size of 5MB */
+	adev->mman.fw_vram_usage_start_offset = adev->virt.crit_regn.offset;
+	adev->mman.fw_vram_usage_size = adev->virt.crit_regn.size_kb << 10;
+	dev_info(adev->dev,
+		"critical region v%d requested to reserve memory start at %08llx with %llu KB.\n",
+			init_data_hdr->version,
+			adev->mman.fw_vram_usage_start_offset,
+			adev->mman.fw_vram_usage_size >> 10);
+
+	adev->virt.is_dynamic_crit_regn_enabled = true;
+
+out:
+	kfree(init_data_hdr);
+	init_data_hdr = NULL;
+
+	return r;
+}
+
+int amdgpu_virt_get_dynamic_data_info(struct amdgpu_device *adev,
+	int data_id, uint8_t *binary, u32 *size)
+{
+	uint32_t data_offset = 0;
+	uint32_t data_size = 0;
+	enum amd_sriov_msg_table_id_enum data_table_id = data_id;
+
+	if (data_table_id >= AMD_SRIOV_MSG_MAX_TABLE_ID)
+		return -EINVAL;
+
+	data_offset = adev->virt.crit_regn_tbl[data_table_id].offset;
+	data_size = adev->virt.crit_regn_tbl[data_table_id].size_kb << 10;
+
+	/* Validate on input params */
+	if (!binary || !size || *size < (uint64_t)data_size)
+		return -EINVAL;
+
+	/* Proceed to copy the dynamic content */
+	amdgpu_device_vram_access(adev,
+			(uint64_t)data_offset, (uint32_t *)binary, data_size, false);
+	*size = (uint64_t)data_size;
+
+	dev_dbg(adev->dev,
+		"Got %s info from dynamic crit_region_table at offset 0x%x with size of 0x%x bytes.\n",
+		amdgpu_virt_dynamic_crit_table_name[data_id], data_offset, data_size);
+
+	return 0;
+}
+
+void amdgpu_virt_init(struct amdgpu_device *adev)
+{
+	bool is_sriov = false;
+	uint32_t reg = amdgpu_virt_init_detect_asic(adev);
+
+	is_sriov = amdgpu_virt_init_req_data(adev, reg);
+
+	if (is_sriov)
+		amdgpu_virt_init_ras(adev);
 }
 
 static bool amdgpu_virt_access_debugfs_is_mmio(struct amdgpu_device *adev)
@@ -835,9 +1188,28 @@ enum amdgpu_sriov_vf_mode amdgpu_virt_get_sriov_vf_mode(struct amdgpu_device *ad
 	return mode;
 }
 
+void amdgpu_virt_pre_reset(struct amdgpu_device *adev)
+{
+	/* stop the data exchange thread */
+	amdgpu_virt_fini_data_exchange(adev);
+	amdgpu_dpm_set_mp1_state(adev, PP_MP1_STATE_FLR);
+}
+
+void amdgpu_virt_post_reset(struct amdgpu_device *adev)
+{
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 0, 3)) {
+		/* force set to GFXOFF state after reset,
+		 * to avoid some invalid operation before GC enable
+		 */
+		adev->gfx.is_poweron = false;
+	}
+
+	adev->mes.ring[0].sched.ready = false;
+}
+
 bool amdgpu_virt_fw_load_skip_check(struct amdgpu_device *adev, uint32_t ucode_id)
 {
-	switch (adev->ip_versions[MP0_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MP0_HWIP, 0)) {
 	case IP_VERSION(13, 0, 0):
 		/* no vf autoload, white list */
 		if (ucode_id == AMDGPU_UCODE_ID_VCN1 ||
@@ -845,6 +1217,17 @@ bool amdgpu_virt_fw_load_skip_check(struct amdgpu_device *adev, uint32_t ucode_i
 			return false;
 		else
 			return true;
+	case IP_VERSION(11, 0, 9):
+	case IP_VERSION(11, 0, 7):
+		/* black list for CHIP_NAVI12 and CHIP_SIENNA_CICHLID */
+		if (ucode_id == AMDGPU_UCODE_ID_RLC_G
+		    || ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL
+		    || ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM
+		    || ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM
+		    || ucode_id == AMDGPU_UCODE_ID_SMC)
+			return true;
+		else
+			return false;
 	case IP_VERSION(13, 0, 10):
 		/* white list */
 		if (ucode_id == AMDGPU_UCODE_ID_CAP
@@ -921,7 +1304,7 @@ void amdgpu_virt_update_sriov_video_codec(struct amdgpu_device *adev,
 	}
 }
 
-static bool amdgpu_virt_get_rlcg_reg_access_flag(struct amdgpu_device *adev,
+bool amdgpu_virt_get_rlcg_reg_access_flag(struct amdgpu_device *adev,
 						 u32 acc_flags, u32 hwip,
 						 bool write, u32 *rlcg_flag)
 {
@@ -954,7 +1337,7 @@ static bool amdgpu_virt_get_rlcg_reg_access_flag(struct amdgpu_device *adev,
 	return ret;
 }
 
-static u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v, u32 flag)
+u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v, u32 flag, u32 xcc_id)
 {
 	struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl;
 	uint32_t timeout = 50000;
@@ -965,6 +1348,7 @@ static u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v
 	void *scratch_reg2;
 	void *scratch_reg3;
 	void *spare_int;
+	unsigned long flags;
 
 	if (!adev->gfx.rlc.rlcg_reg_access_supported) {
 		dev_err(adev->dev,
@@ -972,11 +1356,22 @@ static u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v
 		return 0;
 	}
 
-	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl;
+	if (adev->gfx.xcc_mask && (((1 << xcc_id) & adev->gfx.xcc_mask) == 0)) {
+		dev_err(adev->dev, "invalid xcc\n");
+		return 0;
+	}
+
+	if (amdgpu_device_skip_hw_access(adev))
+		return 0;
+
+	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl[xcc_id];
 	scratch_reg0 = (void __iomem *)adev->rmmio + 4 * reg_access_ctrl->scratch_reg0;
 	scratch_reg1 = (void __iomem *)adev->rmmio + 4 * reg_access_ctrl->scratch_reg1;
 	scratch_reg2 = (void __iomem *)adev->rmmio + 4 * reg_access_ctrl->scratch_reg2;
 	scratch_reg3 = (void __iomem *)adev->rmmio + 4 * reg_access_ctrl->scratch_reg3;
+
+	spin_lock_irqsave(&adev->virt.rlcg_reg_lock, flags);
+
 	if (reg_access_ctrl->spare_int)
 		spare_int = (void __iomem *)adev->rmmio + 4 * reg_access_ctrl->spare_int;
 
@@ -995,7 +1390,7 @@ static u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v
 		 * SCRATCH_REG0 	= read/write value
 		 * SCRATCH_REG1[30:28]	= command
 		 * SCRATCH_REG1[19:0]	= address in dword
-		 * SCRATCH_REG1[26:24]	= Error reporting
+		 * SCRATCH_REG1[27:24]	= Error reporting
 		 */
 		writel(v, scratch_reg0);
 		writel((offset | flag), scratch_reg1);
@@ -1009,7 +1404,8 @@ static u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v
 			udelay(10);
 		}
 
-		if (i >= timeout) {
+		tmp = readl(scratch_reg1);
+		if (i >= timeout || (tmp & AMDGPU_RLCG_SCRATCH1_ERROR_MASK) != 0) {
 			if (amdgpu_sriov_rlcg_error_report_enabled(adev)) {
 				if (tmp & AMDGPU_RLCG_VFGATE_DISABLED) {
 					dev_err(adev->dev,
@@ -1032,18 +1428,24 @@ static u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v
 	}
 
 	ret = readl(scratch_reg0);
+
+	spin_unlock_irqrestore(&adev->virt.rlcg_reg_lock, flags);
+
 	return ret;
 }
 
 void amdgpu_sriov_wreg(struct amdgpu_device *adev,
 		       u32 offset, u32 value,
-		       u32 acc_flags, u32 hwip)
+		       u32 acc_flags, u32 hwip, u32 xcc_id)
 {
 	u32 rlcg_flag;
 
+	if (amdgpu_device_skip_hw_access(adev))
+		return;
+
 	if (!amdgpu_sriov_runtime(adev) &&
 		amdgpu_virt_get_rlcg_reg_access_flag(adev, acc_flags, hwip, true, &rlcg_flag)) {
-		amdgpu_virt_rlcg_reg_rw(adev, offset, value, rlcg_flag);
+		amdgpu_virt_rlcg_reg_rw(adev, offset, value, rlcg_flag, xcc_id);
 		return;
 	}
 
@@ -1054,16 +1456,392 @@ void amdgpu_sriov_wreg(struct amdgpu_device *adev,
 }
 
 u32 amdgpu_sriov_rreg(struct amdgpu_device *adev,
-		      u32 offset, u32 acc_flags, u32 hwip)
+		      u32 offset, u32 acc_flags, u32 hwip, u32 xcc_id)
 {
 	u32 rlcg_flag;
 
+	if (amdgpu_device_skip_hw_access(adev))
+		return 0;
+
 	if (!amdgpu_sriov_runtime(adev) &&
 		amdgpu_virt_get_rlcg_reg_access_flag(adev, acc_flags, hwip, false, &rlcg_flag))
-		return amdgpu_virt_rlcg_reg_rw(adev, offset, 0, rlcg_flag);
+		return amdgpu_virt_rlcg_reg_rw(adev, offset, 0, rlcg_flag, xcc_id);
 
 	if (acc_flags & AMDGPU_REGS_NO_KIQ)
 		return RREG32_NO_KIQ(offset);
 	else
 		return RREG32(offset);
 }
+
+bool amdgpu_sriov_xnack_support(struct amdgpu_device *adev)
+{
+	bool xnack_mode = true;
+
+	if (amdgpu_sriov_vf(adev) &&
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2))
+		xnack_mode = false;
+
+	return xnack_mode;
+}
+
+bool amdgpu_virt_get_ras_capability(struct amdgpu_device *adev)
+{
+	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+
+	if (!amdgpu_sriov_ras_caps_en(adev))
+		return false;
+
+	if (adev->virt.ras_en_caps.bits.block_umc)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__UMC);
+	if (adev->virt.ras_en_caps.bits.block_sdma)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__SDMA);
+	if (adev->virt.ras_en_caps.bits.block_gfx)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__GFX);
+	if (adev->virt.ras_en_caps.bits.block_mmhub)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__MMHUB);
+	if (adev->virt.ras_en_caps.bits.block_athub)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__ATHUB);
+	if (adev->virt.ras_en_caps.bits.block_pcie_bif)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__PCIE_BIF);
+	if (adev->virt.ras_en_caps.bits.block_hdp)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__HDP);
+	if (adev->virt.ras_en_caps.bits.block_xgmi_wafl)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__XGMI_WAFL);
+	if (adev->virt.ras_en_caps.bits.block_df)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__DF);
+	if (adev->virt.ras_en_caps.bits.block_smn)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__SMN);
+	if (adev->virt.ras_en_caps.bits.block_sem)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__SEM);
+	if (adev->virt.ras_en_caps.bits.block_mp0)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__MP0);
+	if (adev->virt.ras_en_caps.bits.block_mp1)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__MP1);
+	if (adev->virt.ras_en_caps.bits.block_fuse)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__FUSE);
+	if (adev->virt.ras_en_caps.bits.block_mca)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__MCA);
+	if (adev->virt.ras_en_caps.bits.block_vcn)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__VCN);
+	if (adev->virt.ras_en_caps.bits.block_jpeg)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__JPEG);
+	if (adev->virt.ras_en_caps.bits.block_ih)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__IH);
+	if (adev->virt.ras_en_caps.bits.block_mpio)
+		adev->ras_hw_enabled |= BIT(AMDGPU_RAS_BLOCK__MPIO);
+
+	if (adev->virt.ras_en_caps.bits.poison_propogation_mode)
+		con->poison_supported = true; /* Poison is handled by host */
+
+	return true;
+}
+
+static inline enum amd_sriov_ras_telemetry_gpu_block
+amdgpu_ras_block_to_sriov(struct amdgpu_device *adev, enum amdgpu_ras_block block) {
+	switch (block) {
+	case AMDGPU_RAS_BLOCK__UMC:
+		return RAS_TELEMETRY_GPU_BLOCK_UMC;
+	case AMDGPU_RAS_BLOCK__SDMA:
+		return RAS_TELEMETRY_GPU_BLOCK_SDMA;
+	case AMDGPU_RAS_BLOCK__GFX:
+		return RAS_TELEMETRY_GPU_BLOCK_GFX;
+	case AMDGPU_RAS_BLOCK__MMHUB:
+		return RAS_TELEMETRY_GPU_BLOCK_MMHUB;
+	case AMDGPU_RAS_BLOCK__ATHUB:
+		return RAS_TELEMETRY_GPU_BLOCK_ATHUB;
+	case AMDGPU_RAS_BLOCK__PCIE_BIF:
+		return RAS_TELEMETRY_GPU_BLOCK_PCIE_BIF;
+	case AMDGPU_RAS_BLOCK__HDP:
+		return RAS_TELEMETRY_GPU_BLOCK_HDP;
+	case AMDGPU_RAS_BLOCK__XGMI_WAFL:
+		return RAS_TELEMETRY_GPU_BLOCK_XGMI_WAFL;
+	case AMDGPU_RAS_BLOCK__DF:
+		return RAS_TELEMETRY_GPU_BLOCK_DF;
+	case AMDGPU_RAS_BLOCK__SMN:
+		return RAS_TELEMETRY_GPU_BLOCK_SMN;
+	case AMDGPU_RAS_BLOCK__SEM:
+		return RAS_TELEMETRY_GPU_BLOCK_SEM;
+	case AMDGPU_RAS_BLOCK__MP0:
+		return RAS_TELEMETRY_GPU_BLOCK_MP0;
+	case AMDGPU_RAS_BLOCK__MP1:
+		return RAS_TELEMETRY_GPU_BLOCK_MP1;
+	case AMDGPU_RAS_BLOCK__FUSE:
+		return RAS_TELEMETRY_GPU_BLOCK_FUSE;
+	case AMDGPU_RAS_BLOCK__MCA:
+		return RAS_TELEMETRY_GPU_BLOCK_MCA;
+	case AMDGPU_RAS_BLOCK__VCN:
+		return RAS_TELEMETRY_GPU_BLOCK_VCN;
+	case AMDGPU_RAS_BLOCK__JPEG:
+		return RAS_TELEMETRY_GPU_BLOCK_JPEG;
+	case AMDGPU_RAS_BLOCK__IH:
+		return RAS_TELEMETRY_GPU_BLOCK_IH;
+	case AMDGPU_RAS_BLOCK__MPIO:
+		return RAS_TELEMETRY_GPU_BLOCK_MPIO;
+	default:
+		dev_warn(adev->dev, "Unsupported SRIOV RAS telemetry block 0x%x\n",
+			      block);
+		return RAS_TELEMETRY_GPU_BLOCK_COUNT;
+	}
+}
+
+static int amdgpu_virt_cache_host_error_counts(struct amdgpu_device *adev,
+					       struct amdsriov_ras_telemetry *host_telemetry)
+{
+	struct amd_sriov_ras_telemetry_error_count *tmp = NULL;
+	uint32_t checksum, used_size;
+
+	checksum = host_telemetry->header.checksum;
+	used_size = host_telemetry->header.used_size;
+
+	if (used_size > (AMD_SRIOV_MSG_RAS_TELEMETRY_SIZE_KB_V1 << 10))
+		return 0;
+
+	tmp = kmemdup(&host_telemetry->body.error_count, used_size, GFP_KERNEL);
+	if (!tmp)
+		return -ENOMEM;
+
+	if (checksum != amd_sriov_msg_checksum(tmp, used_size, 0, 0))
+		goto out;
+
+	memcpy(&adev->virt.count_cache, tmp,
+	       min(used_size, sizeof(adev->virt.count_cache)));
+out:
+	kfree(tmp);
+
+	return 0;
+}
+
+static int amdgpu_virt_req_ras_err_count_internal(struct amdgpu_device *adev, bool force_update)
+{
+	struct amdgpu_virt *virt = &adev->virt;
+
+	if (!virt->ops || !virt->ops->req_ras_err_count)
+		return -EOPNOTSUPP;
+
+	/* Host allows 15 ras telemetry requests per 60 seconds. Afterwhich, the Host
+	 * will ignore incoming guest messages. Ratelimit the guest messages to
+	 * prevent guest self DOS.
+	 */
+	if (__ratelimit(&virt->ras.ras_error_cnt_rs) || force_update) {
+		mutex_lock(&virt->ras.ras_telemetry_mutex);
+		if (!virt->ops->req_ras_err_count(adev))
+			amdgpu_virt_cache_host_error_counts(adev,
+				virt->fw_reserve.ras_telemetry);
+		mutex_unlock(&virt->ras.ras_telemetry_mutex);
+	}
+
+	return 0;
+}
+
+/* Bypass ACA interface and query ECC counts directly from host */
+int amdgpu_virt_req_ras_err_count(struct amdgpu_device *adev, enum amdgpu_ras_block block,
+				  struct ras_err_data *err_data)
+{
+	enum amd_sriov_ras_telemetry_gpu_block sriov_block;
+
+	sriov_block = amdgpu_ras_block_to_sriov(adev, block);
+
+	if (sriov_block >= RAS_TELEMETRY_GPU_BLOCK_COUNT ||
+	    !amdgpu_sriov_ras_telemetry_block_en(adev, sriov_block))
+		return -EOPNOTSUPP;
+
+	/* Host Access may be lost during reset, just return last cached data. */
+	if (down_read_trylock(&adev->reset_domain->sem)) {
+		amdgpu_virt_req_ras_err_count_internal(adev, false);
+		up_read(&adev->reset_domain->sem);
+	}
+
+	err_data->ue_count = adev->virt.count_cache.block[sriov_block].ue_count;
+	err_data->ce_count = adev->virt.count_cache.block[sriov_block].ce_count;
+	err_data->de_count = adev->virt.count_cache.block[sriov_block].de_count;
+
+	return 0;
+}
+
+static int
+amdgpu_virt_write_cpers_to_ring(struct amdgpu_device *adev,
+				struct amdsriov_ras_telemetry *host_telemetry,
+				u32 *more)
+{
+	struct amd_sriov_ras_cper_dump *cper_dump = NULL;
+	struct cper_hdr *entry = NULL;
+	struct amdgpu_ring *ring = &adev->cper.ring_buf;
+	uint32_t checksum, used_size, i;
+	int ret = 0;
+
+	checksum = host_telemetry->header.checksum;
+	used_size = host_telemetry->header.used_size;
+
+	if (used_size > (AMD_SRIOV_MSG_RAS_TELEMETRY_SIZE_KB_V1 << 10))
+		return -EINVAL;
+
+	cper_dump = kmemdup(&host_telemetry->body.cper_dump, used_size, GFP_KERNEL);
+	if (!cper_dump)
+		return -ENOMEM;
+
+	if (checksum != amd_sriov_msg_checksum(cper_dump, used_size, 0, 0)) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	*more = cper_dump->more;
+
+	if (cper_dump->wptr < adev->virt.ras.cper_rptr) {
+		dev_warn(
+			adev->dev,
+			"guest specified rptr that was too high! guest rptr: 0x%llx, host rptr: 0x%llx\n",
+			adev->virt.ras.cper_rptr, cper_dump->wptr);
+
+		adev->virt.ras.cper_rptr = cper_dump->wptr;
+		goto out;
+	}
+
+	entry = (struct cper_hdr *)&cper_dump->buf[0];
+
+	for (i = 0; i < cper_dump->count; i++) {
+		amdgpu_cper_ring_write(ring, entry, entry->record_length);
+		entry = (struct cper_hdr *)((char *)entry +
+					    entry->record_length);
+	}
+
+	if (cper_dump->overflow_count)
+		dev_warn(adev->dev,
+			 "host reported CPER overflow of 0x%llx entries!\n",
+			 cper_dump->overflow_count);
+
+	adev->virt.ras.cper_rptr = cper_dump->wptr;
+out:
+	kfree(cper_dump);
+
+	return ret;
+}
+
+static int amdgpu_virt_req_ras_cper_dump_internal(struct amdgpu_device *adev)
+{
+	struct amdgpu_virt *virt = &adev->virt;
+	int ret = 0;
+	uint32_t more = 0;
+
+	if (!virt->ops || !virt->ops->req_ras_cper_dump)
+		return -EOPNOTSUPP;
+
+	do {
+		if (!virt->ops->req_ras_cper_dump(adev, virt->ras.cper_rptr))
+			ret = amdgpu_virt_write_cpers_to_ring(
+				adev, virt->fw_reserve.ras_telemetry, &more);
+		else
+			ret = 0;
+	} while (more && !ret);
+
+	return ret;
+}
+
+int amdgpu_virt_req_ras_cper_dump(struct amdgpu_device *adev, bool force_update)
+{
+	struct amdgpu_virt *virt = &adev->virt;
+	int ret = 0;
+
+	if (!amdgpu_sriov_ras_cper_en(adev))
+		return -EOPNOTSUPP;
+
+	if ((__ratelimit(&virt->ras.ras_cper_dump_rs) || force_update) &&
+	    down_read_trylock(&adev->reset_domain->sem)) {
+		mutex_lock(&virt->ras.ras_telemetry_mutex);
+		ret = amdgpu_virt_req_ras_cper_dump_internal(adev);
+		mutex_unlock(&virt->ras.ras_telemetry_mutex);
+		up_read(&adev->reset_domain->sem);
+	}
+
+	return ret;
+}
+
+int amdgpu_virt_ras_telemetry_post_reset(struct amdgpu_device *adev)
+{
+	unsigned long ue_count, ce_count;
+
+	if (amdgpu_sriov_ras_telemetry_en(adev)) {
+		amdgpu_virt_req_ras_err_count_internal(adev, true);
+		amdgpu_ras_query_error_count(adev, &ce_count, &ue_count, NULL);
+	}
+
+	return 0;
+}
+
+bool amdgpu_virt_ras_telemetry_block_en(struct amdgpu_device *adev,
+					enum amdgpu_ras_block block)
+{
+	enum amd_sriov_ras_telemetry_gpu_block sriov_block;
+
+	sriov_block = amdgpu_ras_block_to_sriov(adev, block);
+
+	if (sriov_block >= RAS_TELEMETRY_GPU_BLOCK_COUNT ||
+	    !amdgpu_sriov_ras_telemetry_block_en(adev, sriov_block))
+		return false;
+
+	return true;
+}
+
+/*
+ * amdgpu_virt_request_bad_pages() - request bad pages
+ * @adev: amdgpu device.
+ * Send command to GPU hypervisor to write new bad pages into the shared PF2VF region
+ */
+void amdgpu_virt_request_bad_pages(struct amdgpu_device *adev)
+{
+	struct amdgpu_virt *virt = &adev->virt;
+
+	if (virt->ops && virt->ops->req_bad_pages)
+		virt->ops->req_bad_pages(adev);
+}
+
+static int amdgpu_virt_cache_chk_criti_hit(struct amdgpu_device *adev,
+					   struct amdsriov_ras_telemetry *host_telemetry,
+					   bool *hit)
+{
+	struct amd_sriov_ras_chk_criti *tmp = NULL;
+	uint32_t checksum, used_size;
+
+	checksum = host_telemetry->header.checksum;
+	used_size = host_telemetry->header.used_size;
+
+	if (used_size > (AMD_SRIOV_MSG_RAS_TELEMETRY_SIZE_KB_V1 << 10))
+		return 0;
+
+	tmp = kmemdup(&host_telemetry->body.chk_criti, used_size, GFP_KERNEL);
+	if (!tmp)
+		return -ENOMEM;
+
+	if (checksum != amd_sriov_msg_checksum(tmp, used_size, 0, 0))
+		goto out;
+
+	if (hit)
+		*hit = tmp->hit ? true : false;
+
+out:
+	kfree(tmp);
+
+	return 0;
+}
+
+int amdgpu_virt_check_vf_critical_region(struct amdgpu_device *adev, u64 addr, bool *hit)
+{
+	struct amdgpu_virt *virt = &adev->virt;
+	int r = -EPERM;
+
+	if (!virt->ops || !virt->ops->req_ras_chk_criti)
+		return -EOPNOTSUPP;
+
+	/* Host allows 15 ras telemetry requests per 60 seconds. Afterwhich, the Host
+	 * will ignore incoming guest messages. Ratelimit the guest messages to
+	 * prevent guest self DOS.
+	 */
+	if (__ratelimit(&virt->ras.ras_chk_criti_rs)) {
+		mutex_lock(&virt->ras.ras_telemetry_mutex);
+		if (!virt->ops->req_ras_chk_criti(adev, addr))
+			r = amdgpu_virt_cache_chk_criti_hit(
+				adev, virt->fw_reserve.ras_telemetry, hit);
+		mutex_unlock(&virt->ras.ras_telemetry_mutex);
+	}
+
+	return r;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
index 4f7bab52282a..01d5bca2dee1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.h
@@ -45,12 +45,21 @@
 #define AMDGPU_RLCG_REG_NOT_IN_RANGE		0x1000000
 
 #define AMDGPU_RLCG_SCRATCH1_ADDRESS_MASK	0xFFFFF
+#define AMDGPU_RLCG_SCRATCH1_ERROR_MASK	0xF000000
 
 /* all asic after AI use this offset */
 #define mmRCC_IOV_FUNC_IDENTIFIER 0xDE5
 /* tonga/fiji use this offset */
 #define mmBIF_IOV_FUNC_IDENTIFIER 0x1503
 
+#define AMDGPU_VF2PF_UPDATE_MAX_RETRY_LIMIT 2
+
+/* Signature used to validate the SR-IOV dynamic critical region init data header ("INDA") */
+#define AMDGPU_SRIOV_CRIT_DATA_SIGNATURE "INDA"
+#define AMDGPU_SRIOV_CRIT_DATA_SIG_LEN   4
+
+#define IS_SRIOV_CRIT_REGN_ENTRY_VALID(hdr, id) ((hdr)->valid_tables & (1 << (id)))
+
 enum amdgpu_sriov_vf_mode {
 	SRIOV_VF_MODE_BARE_METAL = 0,
 	SRIOV_VF_MODE_ONE_VF,
@@ -85,10 +94,17 @@ struct amdgpu_virt_ops {
 	int (*rel_full_gpu)(struct amdgpu_device *adev, bool init);
 	int (*req_init_data)(struct amdgpu_device *adev);
 	int (*reset_gpu)(struct amdgpu_device *adev);
+	void (*ready_to_reset)(struct amdgpu_device *adev);
 	int (*wait_reset)(struct amdgpu_device *adev);
 	void (*trans_msg)(struct amdgpu_device *adev, enum idh_request req,
 			  u32 data1, u32 data2, u32 data3);
-	void (*ras_poison_handler)(struct amdgpu_device *adev);
+	void (*ras_poison_handler)(struct amdgpu_device *adev,
+					enum amdgpu_ras_block block);
+	bool (*rcvd_ras_intr)(struct amdgpu_device *adev);
+	int (*req_ras_err_count)(struct amdgpu_device *adev);
+	int (*req_ras_cper_dump)(struct amdgpu_device *adev, u64 vf_rptr);
+	int (*req_bad_pages)(struct amdgpu_device *adev);
+	int (*req_ras_chk_criti)(struct amdgpu_device *adev, u64 addr);
 };
 
 /*
@@ -97,6 +113,7 @@ struct amdgpu_virt_ops {
 struct amdgpu_virt_fw_reserve {
 	struct amd_sriov_msg_pf2vf_info_header *p_pf2vf;
 	struct amd_sriov_msg_vf2pf_info_header *p_vf2pf;
+	void *ras_telemetry;
 	unsigned int checksum_key;
 };
 
@@ -126,15 +143,27 @@ enum AMDGIM_FEATURE_FLAG {
 	AMDGIM_FEATURE_INDIRECT_REG_ACCESS = (1 << 5),
 	/* AV1 Support MODE*/
 	AMDGIM_FEATURE_AV1_SUPPORT = (1 << 6),
+	/* VCN RB decouple */
+	AMDGIM_FEATURE_VCN_RB_DECOUPLE = (1 << 7),
+	/* MES info */
+	AMDGIM_FEATURE_MES_INFO_ENABLE = (1 << 8),
+	AMDGIM_FEATURE_RAS_CAPS = (1 << 9),
+	AMDGIM_FEATURE_RAS_TELEMETRY = (1 << 10),
+	AMDGIM_FEATURE_RAS_CPER = (1 << 11),
+	AMDGIM_FEATURE_XGMI_TA_EXT_PEER_LINK = (1 << 12),
 };
 
 enum AMDGIM_REG_ACCESS_FLAG {
 	/* Use PSP to program IH_RB_CNTL */
-	AMDGIM_FEATURE_IH_REG_PSP_EN     = (1 << 0),
+	AMDGIM_FEATURE_IH_REG_PSP_EN      = (1 << 0),
 	/* Use RLC to program MMHUB regs */
-	AMDGIM_FEATURE_MMHUB_REG_RLC_EN  = (1 << 1),
+	AMDGIM_FEATURE_MMHUB_REG_RLC_EN   = (1 << 1),
 	/* Use RLC to program GC regs */
-	AMDGIM_FEATURE_GC_REG_RLC_EN     = (1 << 2),
+	AMDGIM_FEATURE_GC_REG_RLC_EN      = (1 << 2),
+	/* Use PSP to program L1_TLB_CNTL */
+	AMDGIM_FEATURE_L1_TLB_CNTL_PSP_EN = (1 << 3),
+	/* Use RLCG to program SQ_CONFIG1 */
+	AMDGIM_FEATURE_REG_ACCESS_SQ_CONFIG = (1 << 4),
 };
 
 struct amdgim_pf2vf_info_v1 {
@@ -228,6 +257,23 @@ struct amdgpu_virt_ras_err_handler_data {
 	int last_reserved;
 };
 
+struct amdgpu_virt_ras {
+	struct ratelimit_state ras_error_cnt_rs;
+	struct ratelimit_state ras_cper_dump_rs;
+	struct ratelimit_state ras_chk_criti_rs;
+	struct mutex ras_telemetry_mutex;
+	uint64_t cper_rptr;
+};
+
+#define AMDGPU_VIRT_CAPS_LIST(X) X(AMDGPU_VIRT_CAP_POWER_LIMIT)
+
+DECLARE_ATTR_CAP_CLASS(amdgpu_virt, AMDGPU_VIRT_CAPS_LIST);
+
+struct amdgpu_virt_region {
+	uint32_t offset;
+	uint32_t size_kb;
+};
+
 /* GPU virtualization */
 struct amdgpu_virt {
 	uint32_t			caps;
@@ -237,11 +283,16 @@ struct amdgpu_virt {
 	uint32_t			reg_val_offs;
 	struct amdgpu_irq_src		ack_irq;
 	struct amdgpu_irq_src		rcv_irq;
+
 	struct work_struct		flr_work;
+	struct work_struct		req_bad_pages_work;
+	struct work_struct		handle_bad_pages_work;
+
 	struct amdgpu_mm_table		mm_table;
 	const struct amdgpu_virt_ops	*ops;
 	struct amdgpu_vf_error_buffer	vf_errors;
 	struct amdgpu_virt_fw_reserve	fw_reserve;
+	struct amdgpu_virt_caps virt_caps;
 	uint32_t gim_feature;
 	uint32_t reg_access_mode;
 	int req_init_data_ver;
@@ -250,9 +301,16 @@ struct amdgpu_virt {
 	bool ras_init_done;
 	uint32_t reg_access;
 
+	/* dynamic(v2) critical regions */
+	struct amdgpu_virt_region init_data_header;
+	struct amdgpu_virt_region crit_regn;
+	struct amdgpu_virt_region crit_regn_tbl[AMD_SRIOV_MSG_MAX_TABLE_ID];
+	bool is_dynamic_crit_regn_enabled;
+
 	/* vf2pf message */
 	struct delayed_work vf2pf_work;
 	uint32_t vf2pf_update_interval_ms;
+	int vf2pf_update_retry_cnt;
 
 	/* multimedia bandwidth config */
 	bool     is_mm_bw_enabled;
@@ -263,6 +321,19 @@ struct amdgpu_virt {
 
 	/* the ucode id to signal the autoload */
 	uint32_t autoload_ucode_id;
+
+	/* Spinlock to protect access to the RLCG register interface */
+	spinlock_t rlcg_reg_lock;
+
+	struct mutex access_req_mutex;
+
+	union amd_sriov_ras_caps ras_en_caps;
+	union amd_sriov_ras_caps ras_telemetry_en_caps;
+	struct amdgpu_virt_ras ras;
+	struct amd_sriov_ras_telemetry_error_count count_cache;
+
+	/* hibernate and resume with different VF feature for xgmi enabled system */
+	bool is_xgmi_node_migrate_enabled;
 };
 
 struct amdgpu_video_codec_info;
@@ -298,15 +369,38 @@ struct amdgpu_video_codec_info;
 (amdgpu_sriov_vf((adev)) && \
 	((adev)->virt.reg_access & (AMDGIM_FEATURE_GC_REG_RLC_EN)))
 
+#define amdgpu_sriov_reg_indirect_l1_tlb_cntl(adev) \
+(amdgpu_sriov_vf((adev)) && \
+	((adev)->virt.reg_access & (AMDGIM_FEATURE_L1_TLB_CNTL_PSP_EN)))
+
 #define amdgpu_sriov_rlcg_error_report_enabled(adev) \
         (amdgpu_sriov_reg_indirect_mmhub(adev) || amdgpu_sriov_reg_indirect_gc(adev))
 
+#define amdgpu_sriov_reg_access_sq_config(adev) \
+(amdgpu_sriov_vf((adev)) && \
+	((adev)->virt.reg_access & (AMDGIM_FEATURE_REG_ACCESS_SQ_CONFIG)))
+
 #define amdgpu_passthrough(adev) \
 ((adev)->virt.caps & AMDGPU_PASSTHROUGH_MODE)
 
 #define amdgpu_sriov_vf_mmio_access_protection(adev) \
 ((adev)->virt.caps & AMDGPU_VF_MMIO_ACCESS_PROTECT)
 
+#define amdgpu_sriov_ras_caps_en(adev) \
+((adev)->virt.gim_feature & AMDGIM_FEATURE_RAS_CAPS)
+
+#define amdgpu_sriov_ras_telemetry_en(adev) \
+(((adev)->virt.gim_feature & AMDGIM_FEATURE_RAS_TELEMETRY) && (adev)->virt.fw_reserve.ras_telemetry)
+
+#define amdgpu_sriov_ras_telemetry_block_en(adev, sriov_blk) \
+(amdgpu_sriov_ras_telemetry_en((adev)) && (adev)->virt.ras_telemetry_en_caps.all & BIT(sriov_blk))
+
+#define amdgpu_sriov_ras_cper_en(adev) \
+((adev)->virt.gim_feature & AMDGIM_FEATURE_RAS_CPER)
+
+#define amdgpu_sriov_xgmi_ta_ext_peer_link_en(adev) \
+((adev)->virt.gim_feature & AMDGIM_FEATURE_XGMI_TA_EXT_PEER_LINK)
+
 static inline bool is_virtual_machine(void)
 {
 #if defined(CONFIG_X86)
@@ -320,29 +414,42 @@ static inline bool is_virtual_machine(void)
 
 #define amdgpu_sriov_is_pp_one_vf(adev) \
 	((adev)->virt.gim_feature & AMDGIM_FEATURE_PP_ONE_VF)
+#define amdgpu_sriov_multi_vf_mode(adev) \
+	(amdgpu_sriov_vf(adev) && !amdgpu_sriov_is_pp_one_vf(adev))
 #define amdgpu_sriov_is_debug(adev) \
 	((!amdgpu_in_reset(adev)) && adev->virt.tdr_debug)
 #define amdgpu_sriov_is_normal(adev) \
 	((!amdgpu_in_reset(adev)) && (!adev->virt.tdr_debug))
 #define amdgpu_sriov_is_av1_support(adev) \
 	((adev)->virt.gim_feature & AMDGIM_FEATURE_AV1_SUPPORT)
+#define amdgpu_sriov_is_vcn_rb_decouple(adev) \
+	((adev)->virt.gim_feature & AMDGIM_FEATURE_VCN_RB_DECOUPLE)
+#define amdgpu_sriov_is_mes_info_enable(adev) \
+	((adev)->virt.gim_feature & AMDGIM_FEATURE_MES_INFO_ENABLE)
+
+#define amdgpu_virt_xgmi_migrate_enabled(adev) \
+	((adev)->virt.is_xgmi_node_migrate_enabled && (adev)->gmc.xgmi.node_segment_size != 0)
+
 bool amdgpu_virt_mmio_blocked(struct amdgpu_device *adev);
 void amdgpu_virt_init_setting(struct amdgpu_device *adev);
-void amdgpu_virt_kiq_reg_write_reg_wait(struct amdgpu_device *adev,
-					uint32_t reg0, uint32_t rreg1,
-					uint32_t ref, uint32_t mask);
 int amdgpu_virt_request_full_gpu(struct amdgpu_device *adev, bool init);
 int amdgpu_virt_release_full_gpu(struct amdgpu_device *adev, bool init);
 int amdgpu_virt_reset_gpu(struct amdgpu_device *adev);
 void amdgpu_virt_request_init_data(struct amdgpu_device *adev);
+void amdgpu_virt_ready_to_reset(struct amdgpu_device *adev);
 int amdgpu_virt_wait_reset(struct amdgpu_device *adev);
 int amdgpu_virt_alloc_mm_table(struct amdgpu_device *adev);
 void amdgpu_virt_free_mm_table(struct amdgpu_device *adev);
+bool amdgpu_virt_rcvd_ras_interrupt(struct amdgpu_device *adev);
 void amdgpu_virt_release_ras_err_handler_data(struct amdgpu_device *adev);
 void amdgpu_virt_init_data_exchange(struct amdgpu_device *adev);
 void amdgpu_virt_exchange_data(struct amdgpu_device *adev);
 void amdgpu_virt_fini_data_exchange(struct amdgpu_device *adev);
-void amdgpu_detect_virtualization(struct amdgpu_device *adev);
+void amdgpu_virt_init(struct amdgpu_device *adev);
+
+int amdgpu_virt_init_critical_region(struct amdgpu_device *adev);
+int amdgpu_virt_get_dynamic_data_info(struct amdgpu_device *adev,
+	int data_id, uint8_t *binary, u32 *size);
 
 bool amdgpu_virt_can_access_debugfs(struct amdgpu_device *adev);
 int amdgpu_virt_enable_access_debugfs(struct amdgpu_device *adev);
@@ -355,9 +462,25 @@ void amdgpu_virt_update_sriov_video_codec(struct amdgpu_device *adev,
 			struct amdgpu_video_codec_info *decode, uint32_t decode_array_size);
 void amdgpu_sriov_wreg(struct amdgpu_device *adev,
 		       u32 offset, u32 value,
-		       u32 acc_flags, u32 hwip);
+		       u32 acc_flags, u32 hwip, u32 xcc_id);
 u32 amdgpu_sriov_rreg(struct amdgpu_device *adev,
-		      u32 offset, u32 acc_flags, u32 hwip);
+		      u32 offset, u32 acc_flags, u32 hwip, u32 xcc_id);
 bool amdgpu_virt_fw_load_skip_check(struct amdgpu_device *adev,
 			uint32_t ucode_id);
+void amdgpu_virt_pre_reset(struct amdgpu_device *adev);
+void amdgpu_virt_post_reset(struct amdgpu_device *adev);
+bool amdgpu_sriov_xnack_support(struct amdgpu_device *adev);
+bool amdgpu_virt_get_rlcg_reg_access_flag(struct amdgpu_device *adev,
+					  u32 acc_flags, u32 hwip,
+					  bool write, u32 *rlcg_flag);
+u32 amdgpu_virt_rlcg_reg_rw(struct amdgpu_device *adev, u32 offset, u32 v, u32 flag, u32 xcc_id);
+bool amdgpu_virt_get_ras_capability(struct amdgpu_device *adev);
+int amdgpu_virt_req_ras_err_count(struct amdgpu_device *adev, enum amdgpu_ras_block block,
+				  struct ras_err_data *err_data);
+int amdgpu_virt_req_ras_cper_dump(struct amdgpu_device *adev, bool force_update);
+int amdgpu_virt_ras_telemetry_post_reset(struct amdgpu_device *adev);
+bool amdgpu_virt_ras_telemetry_block_en(struct amdgpu_device *adev,
+					enum amdgpu_ras_block block);
+void amdgpu_virt_request_bad_pages(struct amdgpu_device *adev);
+int amdgpu_virt_check_vf_critical_region(struct amdgpu_device *adev, u64 addr, bool *hit);
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c
index 53ff91fc6cf6..79bad9cbe2ab 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c
@@ -1,7 +1,9 @@
 // SPDX-License-Identifier: GPL-2.0+
 
 #include <drm/drm_atomic_helper.h>
+#include <drm/drm_edid.h>
 #include <drm/drm_simple_kms_helper.h>
+#include <drm/drm_gem_framebuffer_helper.h>
 #include <drm/drm_vblank.h>
 
 #include "amdgpu.h"
@@ -12,7 +14,6 @@
 #include "dce_v8_0.h"
 #endif
 #include "dce_v10_0.h"
-#include "dce_v11_0.h"
 #include "ivsrcid/ivsrcid_vislands30.h"
 #include "amdgpu_vkms.h"
 #include "amdgpu_display.h"
@@ -55,17 +56,16 @@ static enum hrtimer_restart amdgpu_vkms_vblank_simulate(struct hrtimer *timer)
 		DRM_WARN("%s: vblank timer overrun\n", __func__);
 
 	ret = drm_crtc_handle_vblank(crtc);
+	/* Don't queue timer again when vblank is disabled. */
 	if (!ret)
-		DRM_ERROR("amdgpu_vkms failure on handling vblank");
+		return HRTIMER_NORESTART;
 
 	return HRTIMER_RESTART;
 }
 
 static int amdgpu_vkms_enable_vblank(struct drm_crtc *crtc)
 {
-	struct drm_device *dev = crtc->dev;
-	unsigned int pipe = drm_crtc_index(crtc);
-	struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
+	struct drm_vblank_crtc *vblank = drm_crtc_vblank_crtc(crtc);
 	struct amdgpu_vkms_output *out = drm_crtc_to_amdgpu_vkms_output(crtc);
 	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
 
@@ -81,7 +81,7 @@ static void amdgpu_vkms_disable_vblank(struct drm_crtc *crtc)
 {
 	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
 
-	hrtimer_cancel(&amdgpu_crtc->vblank_timer);
+	hrtimer_try_to_cancel(&amdgpu_crtc->vblank_timer);
 }
 
 static bool amdgpu_vkms_get_vblank_timestamp(struct drm_crtc *crtc,
@@ -89,10 +89,8 @@ static bool amdgpu_vkms_get_vblank_timestamp(struct drm_crtc *crtc,
 					     ktime_t *vblank_time,
 					     bool in_vblank_irq)
 {
-	struct drm_device *dev = crtc->dev;
-	unsigned int pipe = crtc->index;
 	struct amdgpu_vkms_output *output = drm_crtc_to_amdgpu_vkms_output(crtc);
-	struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
+	struct drm_vblank_crtc *vblank = drm_crtc_vblank_crtc(crtc);
 	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
 
 	if (!READ_ONCE(vblank->enabled)) {
@@ -189,8 +187,8 @@ static int amdgpu_vkms_crtc_init(struct drm_device *dev, struct drm_crtc *crtc,
 	amdgpu_crtc->connector = NULL;
 	amdgpu_crtc->vsync_timer_enabled = AMDGPU_IRQ_STATE_DISABLE;
 
-	hrtimer_init(&amdgpu_crtc->vblank_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
-	amdgpu_crtc->vblank_timer.function = &amdgpu_vkms_vblank_simulate;
+	hrtimer_setup(&amdgpu_crtc->vblank_timer, &amdgpu_vkms_vblank_simulate, CLOCK_MONOTONIC,
+		      HRTIMER_MODE_REL);
 
 	return ret;
 }
@@ -238,6 +236,8 @@ static int amdgpu_vkms_conn_get_modes(struct drm_connector *connector)
 
 	for (i = 0; i < ARRAY_SIZE(common_modes); i++) {
 		mode = drm_cvt_mode(dev, common_modes[i].w, common_modes[i].h, 60, false, false, false);
+		if (!mode)
+			continue;
 		drm_mode_probed_add(connector, mode);
 	}
 
@@ -310,7 +310,13 @@ static int amdgpu_vkms_prepare_fb(struct drm_plane *plane,
 		return 0;
 	}
 	afb = to_amdgpu_framebuffer(new_state->fb);
-	obj = new_state->fb->obj[0];
+
+	obj = drm_gem_fb_get_obj(new_state->fb, 0);
+	if (!obj) {
+		DRM_ERROR("Failed to get obj from framebuffer\n");
+		return -EINVAL;
+	}
+
 	rbo = gem_to_amdgpu_bo(obj);
 	adev = amdgpu_ttm_adev(rbo->tbo.bdev);
 
@@ -331,6 +337,7 @@ static int amdgpu_vkms_prepare_fb(struct drm_plane *plane,
 	else
 		domain = AMDGPU_GEM_DOMAIN_VRAM;
 
+	rbo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 	r = amdgpu_bo_pin(rbo, domain);
 	if (unlikely(r != 0)) {
 		if (r != -ERESTARTSYS)
@@ -364,12 +371,19 @@ static void amdgpu_vkms_cleanup_fb(struct drm_plane *plane,
 				   struct drm_plane_state *old_state)
 {
 	struct amdgpu_bo *rbo;
+	struct drm_gem_object *obj;
 	int r;
 
 	if (!old_state->fb)
 		return;
 
-	rbo = gem_to_amdgpu_bo(old_state->fb->obj[0]);
+	obj = drm_gem_fb_get_obj(old_state->fb, 0);
+	if (!obj) {
+		DRM_ERROR("Failed to get obj from framebuffer\n");
+		return;
+	}
+
+	rbo = gem_to_amdgpu_bo(obj);
 	r = amdgpu_bo_reserve(rbo, false);
 	if (unlikely(r)) {
 		DRM_ERROR("failed to reserve rbo before unpin\n");
@@ -478,10 +492,10 @@ const struct drm_mode_config_funcs amdgpu_vkms_mode_funcs = {
 	.atomic_commit = drm_atomic_helper_commit,
 };
 
-static int amdgpu_vkms_sw_init(void *handle)
+static int amdgpu_vkms_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->amdgpu_vkms_output = kcalloc(adev->mode_info.num_crtc,
 		sizeof(struct amdgpu_vkms_output), GFP_KERNEL);
@@ -500,8 +514,6 @@ static int amdgpu_vkms_sw_init(void *handle)
 
 	adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true;
 
-	adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true;
-
 	r = amdgpu_display_modeset_create_props(adev);
 	if (r)
 		return r;
@@ -523,9 +535,9 @@ static int amdgpu_vkms_sw_init(void *handle)
 	return 0;
 }
 
-static int amdgpu_vkms_sw_fini(void *handle)
+static int amdgpu_vkms_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i = 0;
 
 	for (i = 0; i < adev->mode_info.num_crtc; i++)
@@ -537,14 +549,14 @@ static int amdgpu_vkms_sw_fini(void *handle)
 
 	adev->mode_info.mode_config_initialized = false;
 
-	kfree(adev->mode_info.bios_hardcoded_edid);
+	drm_edid_free(adev->mode_info.bios_hardcoded_edid);
 	kfree(adev->amdgpu_vkms_output);
 	return 0;
 }
 
-static int amdgpu_vkms_hw_init(void *handle)
+static int amdgpu_vkms_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	switch (adev->asic_type) {
 #ifdef CONFIG_DRM_AMDGPU_SI
@@ -568,13 +580,6 @@ static int amdgpu_vkms_hw_init(void *handle)
 	case CHIP_TONGA:
 		dce_v10_0_disable_dce(adev);
 		break;
-	case CHIP_CARRIZO:
-	case CHIP_STONEY:
-	case CHIP_POLARIS10:
-	case CHIP_POLARIS11:
-	case CHIP_VEGAM:
-		dce_v11_0_disable_dce(adev);
-		break;
 	case CHIP_TOPAZ:
 #ifdef CONFIG_DRM_AMDGPU_SI
 	case CHIP_HAINAN:
@@ -587,55 +592,45 @@ static int amdgpu_vkms_hw_init(void *handle)
 	return 0;
 }
 
-static int amdgpu_vkms_hw_fini(void *handle)
+static int amdgpu_vkms_hw_fini(struct amdgpu_ip_block *ip_block)
 {
 	return 0;
 }
 
-static int amdgpu_vkms_suspend(void *handle)
+static int amdgpu_vkms_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = drm_mode_config_helper_suspend(adev_to_drm(adev));
 	if (r)
 		return r;
-	return amdgpu_vkms_hw_fini(handle);
+
+	return 0;
 }
 
-static int amdgpu_vkms_resume(void *handle)
+static int amdgpu_vkms_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	int r;
 
-	r = amdgpu_vkms_hw_init(handle);
+	r = amdgpu_vkms_hw_init(ip_block);
 	if (r)
 		return r;
-	return drm_mode_config_helper_resume(adev_to_drm(adev));
+	return drm_mode_config_helper_resume(adev_to_drm(ip_block->adev));
 }
 
-static bool amdgpu_vkms_is_idle(void *handle)
+static bool amdgpu_vkms_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	return true;
 }
 
-static int amdgpu_vkms_wait_for_idle(void *handle)
-{
-	return 0;
-}
-
-static int amdgpu_vkms_soft_reset(void *handle)
-{
-	return 0;
-}
-
-static int amdgpu_vkms_set_clockgating_state(void *handle,
+static int amdgpu_vkms_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int amdgpu_vkms_set_powergating_state(void *handle,
+static int amdgpu_vkms_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -643,8 +638,6 @@ static int amdgpu_vkms_set_powergating_state(void *handle,
 
 static const struct amd_ip_funcs amdgpu_vkms_ip_funcs = {
 	.name = "amdgpu_vkms",
-	.early_init = NULL,
-	.late_init = NULL,
 	.sw_init = amdgpu_vkms_sw_init,
 	.sw_fini = amdgpu_vkms_sw_fini,
 	.hw_init = amdgpu_vkms_hw_init,
@@ -652,14 +645,11 @@ static const struct amd_ip_funcs amdgpu_vkms_ip_funcs = {
 	.suspend = amdgpu_vkms_suspend,
 	.resume = amdgpu_vkms_resume,
 	.is_idle = amdgpu_vkms_is_idle,
-	.wait_for_idle = amdgpu_vkms_wait_for_idle,
-	.soft_reset = amdgpu_vkms_soft_reset,
 	.set_clockgating_state = amdgpu_vkms_set_clockgating_state,
 	.set_powergating_state = amdgpu_vkms_set_powergating_state,
 };
 
-const struct amdgpu_ip_block_version amdgpu_vkms_ip_block =
-{
+const struct amdgpu_ip_block_version amdgpu_vkms_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 1,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index 286e326bb4bd..a67285118c37 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -34,7 +34,9 @@
 #include <drm/amdgpu_drm.h>
 #include <drm/drm_drv.h>
 #include <drm/ttm/ttm_tt.h>
+#include <drm/drm_exec.h>
 #include "amdgpu.h"
+#include "amdgpu_vm.h"
 #include "amdgpu_trace.h"
 #include "amdgpu_amdkfd.h"
 #include "amdgpu_gmc.h"
@@ -111,9 +113,9 @@ struct amdgpu_prt_cb {
 };
 
 /**
- * struct amdgpu_vm_tlb_seq_cb - Helper to increment the TLB flush sequence
+ * struct amdgpu_vm_tlb_seq_struct - Helper to increment the TLB flush sequence
  */
-struct amdgpu_vm_tlb_seq_cb {
+struct amdgpu_vm_tlb_seq_struct {
 	/**
 	 * @vm: pointer to the amdgpu_vm structure to set the fence sequence on
 	 */
@@ -126,43 +128,14 @@ struct amdgpu_vm_tlb_seq_cb {
 };
 
 /**
- * amdgpu_vm_set_pasid - manage pasid and vm ptr mapping
- *
- * @adev: amdgpu_device pointer
- * @vm: amdgpu_vm pointer
- * @pasid: the pasid the VM is using on this GPU
- *
- * Set the pasid this VM is using on this GPU, can also be used to remove the
- * pasid by passing in zero.
+ * amdgpu_vm_assert_locked - check if VM is correctly locked
+ * @vm: the VM which schould be tested
  *
+ * Asserts that the VM root PD is locked.
  */
-int amdgpu_vm_set_pasid(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			u32 pasid)
+static void amdgpu_vm_assert_locked(struct amdgpu_vm *vm)
 {
-	int r;
-
-	if (vm->pasid == pasid)
-		return 0;
-
-	if (vm->pasid) {
-		r = xa_err(xa_erase_irq(&adev->vm_manager.pasids, vm->pasid));
-		if (r < 0)
-			return r;
-
-		vm->pasid = 0;
-	}
-
-	if (pasid) {
-		r = xa_err(xa_store_irq(&adev->vm_manager.pasids, pasid, vm,
-					GFP_KERNEL));
-		if (r < 0)
-			return r;
-
-		vm->pasid = pasid;
-	}
-
-
-	return 0;
+	dma_resv_assert_held(vm->root.bo->tbo.base.resv);
 }
 
 /**
@@ -179,6 +152,7 @@ static void amdgpu_vm_bo_evicted(struct amdgpu_vm_bo_base *vm_bo)
 	struct amdgpu_bo *bo = vm_bo->bo;
 
 	vm_bo->moved = true;
+	amdgpu_vm_assert_locked(vm);
 	spin_lock(&vm_bo->vm->status_lock);
 	if (bo->tbo.type == ttm_bo_type_kernel)
 		list_move(&vm_bo->vm_status, &vm->evicted);
@@ -196,6 +170,7 @@ static void amdgpu_vm_bo_evicted(struct amdgpu_vm_bo_base *vm_bo)
  */
 static void amdgpu_vm_bo_moved(struct amdgpu_vm_bo_base *vm_bo)
 {
+	amdgpu_vm_assert_locked(vm_bo->vm);
 	spin_lock(&vm_bo->vm->status_lock);
 	list_move(&vm_bo->vm_status, &vm_bo->vm->moved);
 	spin_unlock(&vm_bo->vm->status_lock);
@@ -211,6 +186,7 @@ static void amdgpu_vm_bo_moved(struct amdgpu_vm_bo_base *vm_bo)
  */
 static void amdgpu_vm_bo_idle(struct amdgpu_vm_bo_base *vm_bo)
 {
+	amdgpu_vm_assert_locked(vm_bo->vm);
 	spin_lock(&vm_bo->vm->status_lock);
 	list_move(&vm_bo->vm_status, &vm_bo->vm->idle);
 	spin_unlock(&vm_bo->vm->status_lock);
@@ -233,6 +209,22 @@ static void amdgpu_vm_bo_invalidated(struct amdgpu_vm_bo_base *vm_bo)
 }
 
 /**
+ * amdgpu_vm_bo_evicted_user - vm_bo is evicted
+ *
+ * @vm_bo: vm_bo which is evicted
+ *
+ * State for BOs used by user mode queues which are not at the location they
+ * should be.
+ */
+static void amdgpu_vm_bo_evicted_user(struct amdgpu_vm_bo_base *vm_bo)
+{
+	vm_bo->moved = true;
+	spin_lock(&vm_bo->vm->status_lock);
+	list_move(&vm_bo->vm_status, &vm_bo->vm->evicted_user);
+	spin_unlock(&vm_bo->vm->status_lock);
+}
+
+/**
  * amdgpu_vm_bo_relocated - vm_bo is reloacted
  *
  * @vm_bo: vm_bo which is relocated
@@ -242,6 +234,7 @@ static void amdgpu_vm_bo_invalidated(struct amdgpu_vm_bo_base *vm_bo)
  */
 static void amdgpu_vm_bo_relocated(struct amdgpu_vm_bo_base *vm_bo)
 {
+	amdgpu_vm_assert_locked(vm_bo->vm);
 	if (vm_bo->bo->parent) {
 		spin_lock(&vm_bo->vm->status_lock);
 		list_move(&vm_bo->vm_status, &vm_bo->vm->relocated);
@@ -261,12 +254,149 @@ static void amdgpu_vm_bo_relocated(struct amdgpu_vm_bo_base *vm_bo)
  */
 static void amdgpu_vm_bo_done(struct amdgpu_vm_bo_base *vm_bo)
 {
+	amdgpu_vm_assert_locked(vm_bo->vm);
 	spin_lock(&vm_bo->vm->status_lock);
 	list_move(&vm_bo->vm_status, &vm_bo->vm->done);
 	spin_unlock(&vm_bo->vm->status_lock);
 }
 
 /**
+ * amdgpu_vm_bo_reset_state_machine - reset the vm_bo state machine
+ * @vm: the VM which state machine to reset
+ *
+ * Move all vm_bo object in the VM into a state where they will be updated
+ * again during validation.
+ */
+static void amdgpu_vm_bo_reset_state_machine(struct amdgpu_vm *vm)
+{
+	struct amdgpu_vm_bo_base *vm_bo, *tmp;
+
+	amdgpu_vm_assert_locked(vm);
+
+	spin_lock(&vm->status_lock);
+	list_splice_init(&vm->done, &vm->invalidated);
+	list_for_each_entry(vm_bo, &vm->invalidated, vm_status)
+		vm_bo->moved = true;
+
+	list_for_each_entry_safe(vm_bo, tmp, &vm->idle, vm_status) {
+		struct amdgpu_bo *bo = vm_bo->bo;
+
+		vm_bo->moved = true;
+		if (!bo || bo->tbo.type != ttm_bo_type_kernel)
+			list_move(&vm_bo->vm_status, &vm_bo->vm->moved);
+		else if (bo->parent)
+			list_move(&vm_bo->vm_status, &vm_bo->vm->relocated);
+	}
+	spin_unlock(&vm->status_lock);
+}
+
+/**
+ * amdgpu_vm_update_shared - helper to update shared memory stat
+ * @base: base structure for tracking BO usage in a VM
+ *
+ * Takes the vm status_lock and updates the shared memory stat. If the basic
+ * stat changed (e.g. buffer was moved) amdgpu_vm_update_stats need to be called
+ * as well.
+ */
+static void amdgpu_vm_update_shared(struct amdgpu_vm_bo_base *base)
+{
+	struct amdgpu_vm *vm = base->vm;
+	struct amdgpu_bo *bo = base->bo;
+	uint64_t size = amdgpu_bo_size(bo);
+	uint32_t bo_memtype = amdgpu_bo_mem_stats_placement(bo);
+	bool shared;
+
+	dma_resv_assert_held(bo->tbo.base.resv);
+	spin_lock(&vm->status_lock);
+	shared = drm_gem_object_is_shared_for_memory_stats(&bo->tbo.base);
+	if (base->shared != shared) {
+		base->shared = shared;
+		if (shared) {
+			vm->stats[bo_memtype].drm.shared += size;
+			vm->stats[bo_memtype].drm.private -= size;
+		} else {
+			vm->stats[bo_memtype].drm.shared -= size;
+			vm->stats[bo_memtype].drm.private += size;
+		}
+	}
+	spin_unlock(&vm->status_lock);
+}
+
+/**
+ * amdgpu_vm_bo_update_shared - callback when bo gets shared/unshared
+ * @bo: amdgpu buffer object
+ *
+ * Update the per VM stats for all the vm if needed from private to shared or
+ * vice versa.
+ */
+void amdgpu_vm_bo_update_shared(struct amdgpu_bo *bo)
+{
+	struct amdgpu_vm_bo_base *base;
+
+	for (base = bo->vm_bo; base; base = base->next)
+		amdgpu_vm_update_shared(base);
+}
+
+/**
+ * amdgpu_vm_update_stats_locked - helper to update normal memory stat
+ * @base: base structure for tracking BO usage in a VM
+ * @res:  the ttm_resource to use for the purpose of accounting, may or may not
+ *        be bo->tbo.resource
+ * @sign: if we should add (+1) or subtract (-1) from the stat
+ *
+ * Caller need to have the vm status_lock held. Useful for when multiple update
+ * need to happen at the same time.
+ */
+static void amdgpu_vm_update_stats_locked(struct amdgpu_vm_bo_base *base,
+			    struct ttm_resource *res, int sign)
+{
+	struct amdgpu_vm *vm = base->vm;
+	struct amdgpu_bo *bo = base->bo;
+	int64_t size = sign * amdgpu_bo_size(bo);
+	uint32_t bo_memtype = amdgpu_bo_mem_stats_placement(bo);
+
+	/* For drm-total- and drm-shared-, BO are accounted by their preferred
+	 * placement, see also amdgpu_bo_mem_stats_placement.
+	 */
+	if (base->shared)
+		vm->stats[bo_memtype].drm.shared += size;
+	else
+		vm->stats[bo_memtype].drm.private += size;
+
+	if (res && res->mem_type < __AMDGPU_PL_NUM) {
+		uint32_t res_memtype = res->mem_type;
+
+		vm->stats[res_memtype].drm.resident += size;
+		/* BO only count as purgeable if it is resident,
+		 * since otherwise there's nothing to purge.
+		 */
+		if (bo->flags & AMDGPU_GEM_CREATE_DISCARDABLE)
+			vm->stats[res_memtype].drm.purgeable += size;
+		if (!(bo->preferred_domains & amdgpu_mem_type_to_domain(res_memtype)))
+			vm->stats[bo_memtype].evicted += size;
+	}
+}
+
+/**
+ * amdgpu_vm_update_stats - helper to update normal memory stat
+ * @base: base structure for tracking BO usage in a VM
+ * @res:  the ttm_resource to use for the purpose of accounting, may or may not
+ *        be bo->tbo.resource
+ * @sign: if we should add (+1) or subtract (-1) from the stat
+ *
+ * Updates the basic memory stat when bo is added/deleted/moved.
+ */
+void amdgpu_vm_update_stats(struct amdgpu_vm_bo_base *base,
+			    struct ttm_resource *res, int sign)
+{
+	struct amdgpu_vm *vm = base->vm;
+
+	spin_lock(&vm->status_lock);
+	amdgpu_vm_update_stats_locked(base, res, sign);
+	spin_unlock(&vm->status_lock);
+}
+
+/**
  * amdgpu_vm_bo_base_init - Adds bo to the list of bos associated with the vm
  *
  * @base: base structure for tracking BO usage in a VM
@@ -289,7 +419,12 @@ void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base,
 	base->next = bo->vm_bo;
 	bo->vm_bo = base;
 
-	if (bo->tbo.base.resv != vm->root.bo->tbo.base.resv)
+	spin_lock(&vm->status_lock);
+	base->shared = drm_gem_object_is_shared_for_memory_stats(&bo->tbo.base);
+	amdgpu_vm_update_stats_locked(base, bo->tbo.resource, +1);
+	spin_unlock(&vm->status_lock);
+
+	if (!amdgpu_vm_is_bo_always_valid(vm, bo))
 		return;
 
 	dma_resv_assert_held(vm->root.bo->tbo.base.resv);
@@ -313,25 +448,60 @@ void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base,
 }
 
 /**
- * amdgpu_vm_get_pd_bo - add the VM PD to a validation list
+ * amdgpu_vm_lock_pd - lock PD in drm_exec
  *
  * @vm: vm providing the BOs
- * @validated: head of validation list
- * @entry: entry to add
+ * @exec: drm execution context
+ * @num_fences: number of extra fences to reserve
  *
- * Add the page directory to the list of BOs to
- * validate for command submission.
+ * Lock the VM root PD in the DRM execution context.
  */
-void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
-			 struct list_head *validated,
-			 struct amdgpu_bo_list_entry *entry)
+int amdgpu_vm_lock_pd(struct amdgpu_vm *vm, struct drm_exec *exec,
+		      unsigned int num_fences)
 {
-	entry->priority = 0;
-	entry->tv.bo = &vm->root.bo->tbo;
-	/* Two for VM updates, one for TTM and one for the CS job */
-	entry->tv.num_shared = 4;
-	entry->user_pages = NULL;
-	list_add(&entry->tv.head, validated);
+	/* We need at least two fences for the VM PD/PT updates */
+	return drm_exec_prepare_obj(exec, &vm->root.bo->tbo.base,
+				    2 + num_fences);
+}
+
+/**
+ * amdgpu_vm_lock_done_list - lock all BOs on the done list
+ * @vm: vm providing the BOs
+ * @exec: drm execution context
+ * @num_fences: number of extra fences to reserve
+ *
+ * Lock the BOs on the done list in the DRM execution context.
+ */
+int amdgpu_vm_lock_done_list(struct amdgpu_vm *vm, struct drm_exec *exec,
+			     unsigned int num_fences)
+{
+	struct list_head *prev = &vm->done;
+	struct amdgpu_bo_va *bo_va;
+	struct amdgpu_bo *bo;
+	int ret;
+
+	/* We can only trust prev->next while holding the lock */
+	spin_lock(&vm->status_lock);
+	while (!list_is_head(prev->next, &vm->done)) {
+		bo_va = list_entry(prev->next, typeof(*bo_va), base.vm_status);
+
+		bo = bo_va->base.bo;
+		if (bo) {
+			amdgpu_bo_ref(bo);
+			spin_unlock(&vm->status_lock);
+
+			ret = drm_exec_prepare_obj(exec, &bo->tbo.base, 1);
+			amdgpu_bo_unref(&bo);
+			if (unlikely(ret))
+				return ret;
+
+			spin_lock(&vm->status_lock);
+		}
+		prev = prev->next;
+	}
+	spin_unlock(&vm->status_lock);
+
+	return 0;
 }
 
 /**
@@ -351,28 +521,93 @@ void amdgpu_vm_move_to_lru_tail(struct amdgpu_device *adev,
 	spin_unlock(&adev->mman.bdev.lru_lock);
 }
 
+/* Create scheduler entities for page table updates */
+static int amdgpu_vm_init_entities(struct amdgpu_device *adev,
+				   struct amdgpu_vm *vm)
+{
+	int r;
+
+	r = drm_sched_entity_init(&vm->immediate, DRM_SCHED_PRIORITY_NORMAL,
+				  adev->vm_manager.vm_pte_scheds,
+				  adev->vm_manager.vm_pte_num_scheds, NULL);
+	if (r)
+		goto error;
+
+	return drm_sched_entity_init(&vm->delayed, DRM_SCHED_PRIORITY_NORMAL,
+				     adev->vm_manager.vm_pte_scheds,
+				     adev->vm_manager.vm_pte_num_scheds, NULL);
+
+error:
+	drm_sched_entity_destroy(&vm->immediate);
+	return r;
+}
+
+/* Destroy the entities for page table updates again */
+static void amdgpu_vm_fini_entities(struct amdgpu_vm *vm)
+{
+	drm_sched_entity_destroy(&vm->immediate);
+	drm_sched_entity_destroy(&vm->delayed);
+}
+
 /**
- * amdgpu_vm_validate_pt_bos - validate the page table BOs
+ * amdgpu_vm_generation - return the page table re-generation counter
+ * @adev: the amdgpu_device
+ * @vm: optional VM to check, might be NULL
+ *
+ * Returns a page table re-generation token to allow checking if submissions
+ * are still valid to use this VM. The VM parameter might be NULL in which case
+ * just the VRAM lost counter will be used.
+ */
+uint64_t amdgpu_vm_generation(struct amdgpu_device *adev, struct amdgpu_vm *vm)
+{
+	uint64_t result = (u64)atomic_read(&adev->vram_lost_counter) << 32;
+
+	if (!vm)
+		return result;
+
+	result += lower_32_bits(vm->generation);
+	/* Add one if the page tables will be re-generated on next CS */
+	if (drm_sched_entity_error(&vm->delayed))
+		++result;
+
+	return result;
+}
+
+/**
+ * amdgpu_vm_validate - validate evicted BOs tracked in the VM
  *
  * @adev: amdgpu device pointer
  * @vm: vm providing the BOs
+ * @ticket: optional reservation ticket used to reserve the VM
  * @validate: callback to do the validation
  * @param: parameter for the validation callback
  *
- * Validate the page table BOs on command submission if neccessary.
+ * Validate the page table BOs and per-VM BOs on command submission if
+ * necessary. If a ticket is given, also try to validate evicted user queue
+ * BOs. They must already be reserved with the given ticket.
  *
  * Returns:
  * Validation result.
  */
-int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			      int (*validate)(void *p, struct amdgpu_bo *bo),
-			      void *param)
+int amdgpu_vm_validate(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+		       struct ww_acquire_ctx *ticket,
+		       int (*validate)(void *p, struct amdgpu_bo *bo),
+		       void *param)
 {
+	uint64_t new_vm_generation = amdgpu_vm_generation(adev, vm);
 	struct amdgpu_vm_bo_base *bo_base;
-	struct amdgpu_bo *shadow;
 	struct amdgpu_bo *bo;
 	int r;
 
+	if (vm->generation != new_vm_generation) {
+		vm->generation = new_vm_generation;
+		amdgpu_vm_bo_reset_state_machine(vm);
+		amdgpu_vm_fini_entities(vm);
+		r = amdgpu_vm_init_entities(adev, vm);
+		if (r)
+			return r;
+	}
+
 	spin_lock(&vm->status_lock);
 	while (!list_empty(&vm->evicted)) {
 		bo_base = list_first_entry(&vm->evicted,
@@ -381,16 +616,10 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		spin_unlock(&vm->status_lock);
 
 		bo = bo_base->bo;
-		shadow = amdgpu_bo_shadowed(bo);
 
 		r = validate(param, bo);
 		if (r)
 			return r;
-		if (shadow) {
-			r = validate(param, shadow);
-			if (r)
-				return r;
-		}
 
 		if (bo->tbo.type != ttm_bo_type_kernel) {
 			amdgpu_vm_bo_moved(bo_base);
@@ -400,6 +629,23 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		}
 		spin_lock(&vm->status_lock);
 	}
+	while (ticket && !list_empty(&vm->evicted_user)) {
+		bo_base = list_first_entry(&vm->evicted_user,
+					   struct amdgpu_vm_bo_base,
+					   vm_status);
+		spin_unlock(&vm->status_lock);
+
+		bo = bo_base->bo;
+		dma_resv_assert_held(bo->tbo.base.resv);
+
+		r = validate(param, bo);
+		if (r)
+			return r;
+
+		amdgpu_vm_bo_invalidated(bo_base);
+
+		spin_lock(&vm->status_lock);
+	}
 	spin_unlock(&vm->status_lock);
 
 	amdgpu_vm_eviction_lock(vm);
@@ -417,22 +663,31 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
  * Check if all VM PDs/PTs are ready for updates
  *
  * Returns:
- * True if VM is not evicting.
+ * True if VM is not evicting and all VM entities are not stopped
  */
 bool amdgpu_vm_ready(struct amdgpu_vm *vm)
 {
-	bool empty;
 	bool ret;
 
+	amdgpu_vm_assert_locked(vm);
+
 	amdgpu_vm_eviction_lock(vm);
 	ret = !vm->evicting;
 	amdgpu_vm_eviction_unlock(vm);
 
 	spin_lock(&vm->status_lock);
-	empty = list_empty(&vm->evicted);
+	ret &= list_empty(&vm->evicted);
 	spin_unlock(&vm->status_lock);
 
-	return ret && empty;
+	spin_lock(&vm->immediate.lock);
+	ret &= !vm->immediate.stopped;
+	spin_unlock(&vm->immediate.lock);
+
+	spin_lock(&vm->delayed.lock);
+	ret &= !vm->delayed.stopped;
+	spin_unlock(&vm->delayed.lock);
+
+	return ret;
 }
 
 /**
@@ -483,7 +738,7 @@ bool amdgpu_vm_need_pipeline_sync(struct amdgpu_ring *ring,
 				  struct amdgpu_job *job)
 {
 	struct amdgpu_device *adev = ring->adev;
-	unsigned vmhub = ring->funcs->vmhub;
+	unsigned vmhub = ring->vm_hub;
 	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
 
 	if (job->vmid == 0)
@@ -517,16 +772,18 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 		    bool need_pipe_sync)
 {
 	struct amdgpu_device *adev = ring->adev;
-	unsigned vmhub = ring->funcs->vmhub;
+	struct amdgpu_isolation *isolation = &adev->isolation[ring->xcp_id];
+	unsigned vmhub = ring->vm_hub;
 	struct amdgpu_vmid_mgr *id_mgr = &adev->vm_manager.id_mgr[vmhub];
 	struct amdgpu_vmid *id = &id_mgr->ids[job->vmid];
 	bool spm_update_needed = job->spm_update_needed;
 	bool gds_switch_needed = ring->funcs->emit_gds_switch &&
 		job->gds_switch_needed;
 	bool vm_flush_needed = job->vm_needs_flush;
-	struct dma_fence *fence = NULL;
+	bool cleaner_shader_needed = false;
 	bool pasid_mapping_needed = false;
-	unsigned patch_offset = 0;
+	struct dma_fence *fence = NULL;
+	unsigned int patch;
 	int r;
 
 	if (amdgpu_vmid_had_gpu_reset(adev, id)) {
@@ -548,16 +805,26 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 	pasid_mapping_needed &= adev->gmc.gmc_funcs->emit_pasid_mapping &&
 		ring->funcs->emit_wreg;
 
-	if (!vm_flush_needed && !gds_switch_needed && !need_pipe_sync)
+	cleaner_shader_needed = job->run_cleaner_shader &&
+		adev->gfx.enable_cleaner_shader &&
+		ring->funcs->emit_cleaner_shader && job->base.s_fence &&
+		&job->base.s_fence->scheduled == isolation->spearhead;
+
+	if (!vm_flush_needed && !gds_switch_needed && !need_pipe_sync &&
+	    !cleaner_shader_needed)
 		return 0;
 
 	amdgpu_ring_ib_begin(ring);
 	if (ring->funcs->init_cond_exec)
-		patch_offset = amdgpu_ring_init_cond_exec(ring);
+		patch = amdgpu_ring_init_cond_exec(ring,
+						   ring->cond_exe_gpu_addr);
 
 	if (need_pipe_sync)
 		amdgpu_ring_emit_pipeline_sync(ring);
 
+	if (cleaner_shader_needed)
+		ring->funcs->emit_cleaner_shader(ring);
+
 	if (vm_flush_needed) {
 		trace_amdgpu_vm_flush(ring, job->vmid, job->vm_pd_addr);
 		amdgpu_ring_emit_vm_flush(ring, job->vmid, job->vm_pd_addr);
@@ -567,9 +834,9 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 		amdgpu_gmc_emit_pasid_mapping(ring, job->vmid, job->pasid);
 
 	if (spm_update_needed && adev->gfx.rlc.funcs->update_spm_vmid)
-		adev->gfx.rlc.funcs->update_spm_vmid(adev, job->vmid);
+		adev->gfx.rlc.funcs->update_spm_vmid(adev, ring, job->vmid);
 
-	if (!ring->is_mes_queue && ring->funcs->emit_gds_switch &&
+	if (ring->funcs->emit_gds_switch &&
 	    gds_switch_needed) {
 		amdgpu_ring_emit_gds_switch(ring, job->vmid, job->gds_base,
 					    job->gds_size, job->gws_base,
@@ -577,10 +844,13 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 					    job->oa_size);
 	}
 
-	if (vm_flush_needed || pasid_mapping_needed) {
-		r = amdgpu_fence_emit(ring, &fence, NULL, 0);
+	if (vm_flush_needed || pasid_mapping_needed || cleaner_shader_needed) {
+		r = amdgpu_fence_emit(ring, job->hw_vm_fence, 0);
 		if (r)
 			return r;
+		fence = &job->hw_vm_fence->base;
+		/* get a ref for the job */
+		dma_fence_get(fence);
 	}
 
 	if (vm_flush_needed) {
@@ -599,16 +869,28 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job,
 		id->pasid_mapping = dma_fence_get(fence);
 		mutex_unlock(&id_mgr->lock);
 	}
+
+	/*
+	 * Make sure that all other submissions wait for the cleaner shader to
+	 * finish before we push them to the HW.
+	 */
+	if (cleaner_shader_needed) {
+		trace_amdgpu_cleaner_shader(ring, fence);
+		mutex_lock(&adev->enforce_isolation_mutex);
+		dma_fence_put(isolation->spearhead);
+		isolation->spearhead = dma_fence_get(fence);
+		mutex_unlock(&adev->enforce_isolation_mutex);
+	}
 	dma_fence_put(fence);
 
-	if (ring->funcs->patch_cond_exec)
-		amdgpu_ring_patch_cond_exec(ring, patch_offset);
+	amdgpu_ring_patch_cond_exec(ring, patch);
 
 	/* the double SWITCH_BUFFER here *cannot* be skipped by COND_EXEC */
 	if (ring->funcs->emit_switch_buffer) {
 		amdgpu_ring_emit_switch_buffer(ring);
 		amdgpu_ring_emit_switch_buffer(ring);
 	}
+
 	amdgpu_ring_ib_end(ring);
 	return 0;
 }
@@ -690,6 +972,8 @@ int amdgpu_vm_update_pdes(struct amdgpu_device *adev,
 	LIST_HEAD(relocated);
 	int r, idx;
 
+	amdgpu_vm_assert_locked(vm);
+
 	spin_lock(&vm->status_lock);
 	list_splice_init(&vm->relocated, &relocated);
 	spin_unlock(&vm->status_lock);
@@ -705,7 +989,8 @@ int amdgpu_vm_update_pdes(struct amdgpu_device *adev,
 	params.vm = vm;
 	params.immediate = immediate;
 
-	r = vm->update_funcs->prepare(&params, NULL, AMDGPU_SYNC_EXPLICIT);
+	r = vm->update_funcs->prepare(&params, NULL,
+				      AMDGPU_KERNEL_JOB_ID_VM_UPDATE_PDES);
 	if (r)
 		goto error;
 
@@ -746,7 +1031,7 @@ error:
 static void amdgpu_vm_tlb_seq_cb(struct dma_fence *fence,
 				 struct dma_fence_cb *cb)
 {
-	struct amdgpu_vm_tlb_seq_cb *tlb_cb;
+	struct amdgpu_vm_tlb_seq_struct *tlb_cb;
 
 	tlb_cb = container_of(cb, typeof(*tlb_cb), cb);
 	atomic64_inc(&tlb_cb->vm->tlb_seq);
@@ -754,6 +1039,46 @@ static void amdgpu_vm_tlb_seq_cb(struct dma_fence *fence,
 }
 
 /**
+ * amdgpu_vm_tlb_flush - prepare TLB flush
+ *
+ * @params: parameters for update
+ * @fence: input fence to sync TLB flush with
+ * @tlb_cb: the callback structure
+ *
+ * Increments the tlb sequence to make sure that future CS execute a VM flush.
+ */
+static void
+amdgpu_vm_tlb_flush(struct amdgpu_vm_update_params *params,
+		    struct dma_fence **fence,
+		    struct amdgpu_vm_tlb_seq_struct *tlb_cb)
+{
+	struct amdgpu_vm *vm = params->vm;
+
+	tlb_cb->vm = vm;
+	if (!fence || !*fence) {
+		amdgpu_vm_tlb_seq_cb(NULL, &tlb_cb->cb);
+		return;
+	}
+
+	if (!dma_fence_add_callback(*fence, &tlb_cb->cb,
+				    amdgpu_vm_tlb_seq_cb)) {
+		dma_fence_put(vm->last_tlb_flush);
+		vm->last_tlb_flush = dma_fence_get(*fence);
+	} else {
+		amdgpu_vm_tlb_seq_cb(NULL, &tlb_cb->cb);
+	}
+
+	/* Prepare a TLB flush fence to be attached to PTs */
+	if (!params->unlocked) {
+		amdgpu_vm_tlb_fence_create(params->adev, vm, fence);
+
+		/* Makes sure no PD/PT is freed before the flush */
+		dma_resv_add_fence(vm->root.bo->tbo.base.resv, *fence,
+				   DMA_RESV_USAGE_BOOKKEEP);
+	}
+}
+
+/**
  * amdgpu_vm_update_range - update a range in the vm page table
  *
  * @adev: amdgpu_device pointer to use for commands
@@ -761,7 +1086,8 @@ static void amdgpu_vm_tlb_seq_cb(struct dma_fence *fence,
  * @immediate: immediate submission in a page fault
  * @unlocked: unlocked invalidation during MM callback
  * @flush_tlb: trigger tlb invalidation after update completed
- * @resv: fences we need to sync to
+ * @allow_override: change MTYPE for local NUMA nodes
+ * @sync: fences we need to sync to
  * @start: start of mapped range
  * @last: last mapped entry
  * @flags: flags for the entries
@@ -778,15 +1104,15 @@ static void amdgpu_vm_tlb_seq_cb(struct dma_fence *fence,
  */
 int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			   bool immediate, bool unlocked, bool flush_tlb,
-			   struct dma_resv *resv, uint64_t start, uint64_t last,
-			   uint64_t flags, uint64_t offset, uint64_t vram_base,
+			   bool allow_override, struct amdgpu_sync *sync,
+			   uint64_t start, uint64_t last, uint64_t flags,
+			   uint64_t offset, uint64_t vram_base,
 			   struct ttm_resource *res, dma_addr_t *pages_addr,
 			   struct dma_fence **fence)
 {
+	struct amdgpu_vm_tlb_seq_struct *tlb_cb;
 	struct amdgpu_vm_update_params params;
-	struct amdgpu_vm_tlb_seq_cb *tlb_cb;
 	struct amdgpu_res_cursor cursor;
-	enum amdgpu_sync_mode sync_mode;
 	int r, idx;
 
 	if (!drm_dev_enter(adev_to_drm(adev), &idx))
@@ -794,20 +1120,20 @@ int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 
 	tlb_cb = kmalloc(sizeof(*tlb_cb), GFP_KERNEL);
 	if (!tlb_cb) {
-		r = -ENOMEM;
-		goto error_unlock;
+		drm_dev_exit(idx);
+		return -ENOMEM;
 	}
 
 	/* Vega20+XGMI where PTEs get inadvertently cached in L2 texture cache,
 	 * heavy-weight flush TLB unconditionally.
 	 */
 	flush_tlb |= adev->gmc.xgmi.num_physical_nodes &&
-		     adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 0);
+		     amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 0);
 
 	/*
 	 * On GFX8 and older any 8 PTE block with a valid bit set enters the TLB
 	 */
-	flush_tlb |= adev->ip_versions[GC_HWIP][0] < IP_VERSION(9, 0, 0);
+	flush_tlb |= amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(9, 0, 0);
 
 	memset(&params, 0, sizeof(params));
 	params.adev = adev;
@@ -815,14 +1141,9 @@ int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	params.immediate = immediate;
 	params.pages_addr = pages_addr;
 	params.unlocked = unlocked;
-
-	/* Implicitly sync to command submissions in the same VM before
-	 * unmapping. Sync to moving fences before mapping.
-	 */
-	if (!(flags & AMDGPU_PTE_VALID))
-		sync_mode = AMDGPU_SYNC_EQ_OWNER;
-	else
-		sync_mode = AMDGPU_SYNC_EXPLICIT;
+	params.needs_flush = flush_tlb;
+	params.allow_override = allow_override;
+	INIT_LIST_HEAD(&params.tlb_flush_waitlist);
 
 	amdgpu_vm_eviction_lock(vm);
 	if (vm->evicting) {
@@ -838,7 +1159,8 @@ int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		dma_fence_put(tmp);
 	}
 
-	r = vm->update_funcs->prepare(&params, resv, sync_mode);
+	r = vm->update_funcs->prepare(&params, sync,
+				      AMDGPU_KERNEL_JOB_ID_VM_UPDATE_RANGE);
 	if (r)
 		goto error_free;
 
@@ -881,7 +1203,7 @@ int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 				params.pages_addr = NULL;
 			}
 
-		} else if (flags & (AMDGPU_PTE_VALID | AMDGPU_PTE_PRT)) {
+		} else if (flags & (AMDGPU_PTE_VALID | AMDGPU_PTE_PRT_FLAG(adev))) {
 			addr = vram_base + cursor.start;
 		} else {
 			addr = 0;
@@ -897,65 +1219,28 @@ int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	}
 
 	r = vm->update_funcs->commit(&params, fence);
+	if (r)
+		goto error_free;
 
-	if (flush_tlb || params.table_freed) {
-		tlb_cb->vm = vm;
-		if (fence && *fence &&
-		    !dma_fence_add_callback(*fence, &tlb_cb->cb,
-					   amdgpu_vm_tlb_seq_cb)) {
-			dma_fence_put(vm->last_tlb_flush);
-			vm->last_tlb_flush = dma_fence_get(*fence);
-		} else {
-			amdgpu_vm_tlb_seq_cb(NULL, &tlb_cb->cb);
-		}
+	if (params.needs_flush) {
+		amdgpu_vm_tlb_flush(&params, fence, tlb_cb);
 		tlb_cb = NULL;
 	}
 
+	amdgpu_vm_pt_free_list(adev, &params);
+
 error_free:
 	kfree(tlb_cb);
-
-error_unlock:
 	amdgpu_vm_eviction_unlock(vm);
 	drm_dev_exit(idx);
 	return r;
 }
 
 void amdgpu_vm_get_memory(struct amdgpu_vm *vm,
-			  struct amdgpu_mem_stats *stats)
+			  struct amdgpu_mem_stats stats[__AMDGPU_PL_NUM])
 {
-	struct amdgpu_bo_va *bo_va, *tmp;
-
 	spin_lock(&vm->status_lock);
-	list_for_each_entry_safe(bo_va, tmp, &vm->idle, base.vm_status) {
-		if (!bo_va->base.bo)
-			continue;
-		amdgpu_bo_get_memory(bo_va->base.bo, stats);
-	}
-	list_for_each_entry_safe(bo_va, tmp, &vm->evicted, base.vm_status) {
-		if (!bo_va->base.bo)
-			continue;
-		amdgpu_bo_get_memory(bo_va->base.bo, stats);
-	}
-	list_for_each_entry_safe(bo_va, tmp, &vm->relocated, base.vm_status) {
-		if (!bo_va->base.bo)
-			continue;
-		amdgpu_bo_get_memory(bo_va->base.bo, stats);
-	}
-	list_for_each_entry_safe(bo_va, tmp, &vm->moved, base.vm_status) {
-		if (!bo_va->base.bo)
-			continue;
-		amdgpu_bo_get_memory(bo_va->base.bo, stats);
-	}
-	list_for_each_entry_safe(bo_va, tmp, &vm->invalidated, base.vm_status) {
-		if (!bo_va->base.bo)
-			continue;
-		amdgpu_bo_get_memory(bo_va->base.bo, stats);
-	}
-	list_for_each_entry_safe(bo_va, tmp, &vm->done, base.vm_status) {
-		if (!bo_va->base.bo)
-			continue;
-		amdgpu_bo_get_memory(bo_va->base.bo, stats);
-	}
+	memcpy(stats, vm->stats, sizeof(*stats) * __AMDGPU_PL_NUM);
 	spin_unlock(&vm->status_lock);
 }
 
@@ -977,34 +1262,59 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va,
 	struct amdgpu_bo *bo = bo_va->base.bo;
 	struct amdgpu_vm *vm = bo_va->base.vm;
 	struct amdgpu_bo_va_mapping *mapping;
+	struct dma_fence **last_update;
 	dma_addr_t *pages_addr = NULL;
 	struct ttm_resource *mem;
-	struct dma_fence **last_update;
+	struct amdgpu_sync sync;
 	bool flush_tlb = clear;
-	struct dma_resv *resv;
 	uint64_t vram_base;
 	uint64_t flags;
+	bool uncached;
 	int r;
 
-	if (clear || !bo) {
+	amdgpu_sync_create(&sync);
+	if (clear) {
+		mem = NULL;
+
+		/* Implicitly sync to command submissions in the same VM before
+		 * unmapping.
+		 */
+		r = amdgpu_sync_resv(adev, &sync, vm->root.bo->tbo.base.resv,
+				     AMDGPU_SYNC_EQ_OWNER, vm);
+		if (r)
+			goto error_free;
+		if (bo) {
+			r = amdgpu_sync_kfd(&sync, bo->tbo.base.resv);
+			if (r)
+				goto error_free;
+		}
+	} else if (!bo) {
 		mem = NULL;
-		resv = vm->root.bo->tbo.base.resv;
+
+		/* PRT map operations don't need to sync to anything. */
+
 	} else {
 		struct drm_gem_object *obj = &bo->tbo.base;
 
-		resv = bo->tbo.base.resv;
-		if (obj->import_attach && bo_va->is_xgmi) {
+		if (drm_gem_is_imported(obj) && bo_va->is_xgmi) {
 			struct dma_buf *dma_buf = obj->import_attach->dmabuf;
 			struct drm_gem_object *gobj = dma_buf->priv;
 			struct amdgpu_bo *abo = gem_to_amdgpu_bo(gobj);
 
-			if (abo->tbo.resource->mem_type == TTM_PL_VRAM)
+			if (abo->tbo.resource &&
+			    abo->tbo.resource->mem_type == TTM_PL_VRAM)
 				bo = gem_to_amdgpu_bo(gobj);
 		}
 		mem = bo->tbo.resource;
-		if (mem->mem_type == TTM_PL_TT ||
-		    mem->mem_type == AMDGPU_PL_PREEMPT)
+		if (mem && (mem->mem_type == TTM_PL_TT ||
+			    mem->mem_type == AMDGPU_PL_PREEMPT))
 			pages_addr = bo->tbo.ttm->dma_address;
+
+		/* Implicitly sync to moving fences before mapping anything */
+		r = amdgpu_sync_resv(adev, &sync, bo->tbo.base.resv,
+				     AMDGPU_SYNC_EXPLICIT, vm);
+		if (r)
+			goto error_free;
 	}
 
 	if (bo) {
@@ -1017,13 +1327,14 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va,
 
 		bo_adev = amdgpu_ttm_adev(bo->tbo.bdev);
 		vram_base = bo_adev->vm_manager.vram_base_offset;
+		uncached = (bo->flags & AMDGPU_GEM_CREATE_UNCACHED) != 0;
 	} else {
 		flags = 0x0;
 		vram_base = 0;
+		uncached = false;
 	}
 
-	if (clear || (bo && bo->tbo.base.resv ==
-		      vm->root.bo->tbo.base.resv))
+	if (clear || amdgpu_vm_is_bo_always_valid(vm, bo))
 		last_update = &vm->last_update;
 	else
 		last_update = &bo_va->last_pt_update;
@@ -1042,34 +1353,34 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va,
 		/* normally,bo_va->flags only contians READABLE and WIRTEABLE bit go here
 		 * but in case of something, we filter the flags in first place
 		 */
-		if (!(mapping->flags & AMDGPU_PTE_READABLE))
+		if (!(mapping->flags & AMDGPU_VM_PAGE_READABLE))
 			update_flags &= ~AMDGPU_PTE_READABLE;
-		if (!(mapping->flags & AMDGPU_PTE_WRITEABLE))
+		if (!(mapping->flags & AMDGPU_VM_PAGE_WRITEABLE))
 			update_flags &= ~AMDGPU_PTE_WRITEABLE;
 
 		/* Apply ASIC specific mapping flags */
-		amdgpu_gmc_get_vm_pte(adev, mapping, &update_flags);
+		amdgpu_gmc_get_vm_pte(adev, vm, bo, mapping->flags,
+				      &update_flags);
 
 		trace_amdgpu_vm_bo_update(mapping);
 
 		r = amdgpu_vm_update_range(adev, vm, false, false, flush_tlb,
-					   resv, mapping->start, mapping->last,
-					   update_flags, mapping->offset,
-					   vram_base, mem, pages_addr,
-					   last_update);
+					   !uncached, &sync, mapping->start,
+					   mapping->last, update_flags,
+					   mapping->offset, vram_base, mem,
+					   pages_addr, last_update);
 		if (r)
-			return r;
+			goto error_free;
 	}
 
 	/* If the BO is not in its preferred location add it back to
 	 * the evicted list so that it gets validated again on the
 	 * next command submission.
 	 */
-	if (bo && bo->tbo.base.resv == vm->root.bo->tbo.base.resv) {
-		uint32_t mem_type = bo->tbo.resource->mem_type;
-
-		if (!(bo->preferred_domains &
-		      amdgpu_mem_type_to_domain(mem_type)))
+	if (amdgpu_vm_is_bo_always_valid(vm, bo)) {
+		if (bo->tbo.resource &&
+		    !(bo->preferred_domains &
+		      amdgpu_mem_type_to_domain(bo->tbo.resource->mem_type)))
 			amdgpu_vm_bo_evicted(&bo_va->base);
 		else
 			amdgpu_vm_bo_idle(&bo_va->base);
@@ -1086,7 +1397,9 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va,
 			trace_amdgpu_vm_bo_mapping(mapping);
 	}
 
-	return 0;
+error_free:
+	amdgpu_sync_free(&sync);
+	return r;
 }
 
 /**
@@ -1188,7 +1501,7 @@ static void amdgpu_vm_free_mapping(struct amdgpu_device *adev,
 				   struct amdgpu_bo_va_mapping *mapping,
 				   struct dma_fence *fence)
 {
-	if (mapping->flags & AMDGPU_PTE_PRT)
+	if (mapping->flags & AMDGPU_VM_PAGE_PRT)
 		amdgpu_vm_add_prt_cb(adev, fence);
 	kfree(mapping);
 }
@@ -1233,29 +1546,34 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
 			  struct amdgpu_vm *vm,
 			  struct dma_fence **fence)
 {
-	struct dma_resv *resv = vm->root.bo->tbo.base.resv;
 	struct amdgpu_bo_va_mapping *mapping;
-	uint64_t init_pte_value = 0;
 	struct dma_fence *f = NULL;
+	struct amdgpu_sync sync;
 	int r;
 
+
+	/*
+	 * Implicitly sync to command submissions in the same VM before
+	 * unmapping.
+	 */
+	amdgpu_sync_create(&sync);
+	r = amdgpu_sync_resv(adev, &sync, vm->root.bo->tbo.base.resv,
+			     AMDGPU_SYNC_EQ_OWNER, vm);
+	if (r)
+		goto error_free;
+
 	while (!list_empty(&vm->freed)) {
 		mapping = list_first_entry(&vm->freed,
 			struct amdgpu_bo_va_mapping, list);
 		list_del(&mapping->list);
 
-		if (vm->pte_support_ats &&
-		    mapping->start < AMDGPU_GMC_HOLE_START)
-			init_pte_value = AMDGPU_PTE_DEFAULT_ATC;
-
-		r = amdgpu_vm_update_range(adev, vm, false, false, true, resv,
-					   mapping->start, mapping->last,
-					   init_pte_value, 0, 0, NULL, NULL,
-					   &f);
+		r = amdgpu_vm_update_range(adev, vm, false, false, true, false,
+					   &sync, mapping->start, mapping->last,
+					   0, 0, 0, NULL, NULL, &f);
 		amdgpu_vm_free_mapping(adev, vm, mapping, f);
 		if (r) {
 			dma_fence_put(f);
-			return r;
+			goto error_free;
 		}
 	}
 
@@ -1266,7 +1584,9 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
 		dma_fence_put(f);
 	}
 
-	return 0;
+error_free:
+	amdgpu_sync_free(&sync);
+	return r;
 
 }
 
@@ -1275,6 +1595,7 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
  *
  * @adev: amdgpu_device pointer
  * @vm: requested vm
+ * @ticket: optional reservation ticket used to reserve the VM
  *
  * Make sure all BOs which are moved are updated in the PTs.
  *
@@ -1284,11 +1605,12 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
  * PTs have to be reserved!
  */
 int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
-			   struct amdgpu_vm *vm)
+			   struct amdgpu_vm *vm,
+			   struct ww_acquire_ctx *ticket)
 {
 	struct amdgpu_bo_va *bo_va;
 	struct dma_resv *resv;
-	bool clear;
+	bool clear, unlock;
 	int r;
 
 	spin_lock(&vm->status_lock);
@@ -1311,18 +1633,35 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
 		spin_unlock(&vm->status_lock);
 
 		/* Try to reserve the BO to avoid clearing its ptes */
-		if (!amdgpu_vm_debug && dma_resv_trylock(resv))
+		if (!adev->debug_vm && dma_resv_trylock(resv)) {
+			clear = false;
+			unlock = true;
+		/* The caller is already holding the reservation lock */
+		} else if (ticket && dma_resv_locking_ctx(resv) == ticket) {
 			clear = false;
+			unlock = false;
 		/* Somebody else is using the BO right now */
-		else
+		} else {
 			clear = true;
+			unlock = false;
+		}
 
 		r = amdgpu_vm_bo_update(adev, bo_va, clear);
+
+		if (unlock)
+			dma_resv_unlock(resv);
 		if (r)
 			return r;
 
-		if (!clear)
-			dma_resv_unlock(resv);
+		/* Remember evicted DMABuf imports in compute VMs for later
+		 * validation
+		 */
+		if (vm->is_compute_context &&
+		    drm_gem_is_imported(&bo_va->base.bo->tbo.base) &&
+		    (!bo_va->base.bo->tbo.resource ||
+		     bo_va->base.bo->tbo.resource->mem_type == TTM_PL_SYSTEM))
+			amdgpu_vm_bo_evicted_user(&bo_va->base);
+
 		spin_lock(&vm->status_lock);
 	}
 	spin_unlock(&vm->status_lock);
@@ -1331,6 +1670,51 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
 }
 
 /**
+ * amdgpu_vm_flush_compute_tlb - Flush TLB on compute VM
+ *
+ * @adev: amdgpu_device pointer
+ * @vm: requested vm
+ * @flush_type: flush type
+ * @xcc_mask: mask of XCCs that belong to the compute partition in need of a TLB flush.
+ *
+ * Flush TLB if needed for a compute VM.
+ *
+ * Returns:
+ * 0 for success.
+ */
+int amdgpu_vm_flush_compute_tlb(struct amdgpu_device *adev,
+				struct amdgpu_vm *vm,
+				uint32_t flush_type,
+				uint32_t xcc_mask)
+{
+	uint64_t tlb_seq = amdgpu_vm_tlb_seq(vm);
+	bool all_hub = false;
+	int xcc = 0, r = 0;
+
+	WARN_ON_ONCE(!vm->is_compute_context);
+
+	/*
+	 * It can be that we race and lose here, but that is extremely unlikely
+	 * and the worst thing which could happen is that we flush the changes
+	 * into the TLB once more which is harmless.
+	 */
+	if (atomic64_xchg(&vm->kfd_last_flushed_seq, tlb_seq) == tlb_seq)
+		return 0;
+
+	if (adev->family == AMDGPU_FAMILY_AI ||
+	    adev->family == AMDGPU_FAMILY_RV)
+		all_hub = true;
+
+	for_each_inst(xcc, xcc_mask) {
+		r = amdgpu_gmc_flush_gpu_tlb_pasid(adev, vm->pasid, flush_type,
+						   all_hub, xcc);
+		if (r)
+			break;
+	}
+	return r;
+}
+
+/**
  * amdgpu_vm_bo_add - add a bo to a specific vm
  *
  * @adev: amdgpu_device pointer
@@ -1358,6 +1742,7 @@ struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev,
 	amdgpu_vm_bo_base_init(&bo_va->base, vm, bo);
 
 	bo_va->ref_count = 1;
+	bo_va->last_pt_update = dma_fence_get_stub();
 	INIT_LIST_HEAD(&bo_va->valids);
 	INIT_LIST_HEAD(&bo_va->invalids);
 
@@ -1395,16 +1780,46 @@ static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev,
 	list_add(&mapping->list, &bo_va->invalids);
 	amdgpu_vm_it_insert(mapping, &vm->va);
 
-	if (mapping->flags & AMDGPU_PTE_PRT)
+	if (mapping->flags & AMDGPU_VM_PAGE_PRT)
 		amdgpu_vm_prt_get(adev);
 
-	if (bo && bo->tbo.base.resv == vm->root.bo->tbo.base.resv &&
-	    !bo_va->base.moved) {
+	if (amdgpu_vm_is_bo_always_valid(vm, bo) && !bo_va->base.moved)
 		amdgpu_vm_bo_moved(&bo_va->base);
-	}
+
 	trace_amdgpu_vm_bo_map(bo_va, mapping);
 }
 
+/* Validate operation parameters to prevent potential abuse */
+static int amdgpu_vm_verify_parameters(struct amdgpu_device *adev,
+					  struct amdgpu_bo *bo,
+					  uint64_t saddr,
+					  uint64_t offset,
+					  uint64_t size)
+{
+	uint64_t tmp, lpfn;
+
+	if (saddr & AMDGPU_GPU_PAGE_MASK
+	    || offset & AMDGPU_GPU_PAGE_MASK
+	    || size & AMDGPU_GPU_PAGE_MASK)
+		return -EINVAL;
+
+	if (check_add_overflow(saddr, size, &tmp)
+	    || check_add_overflow(offset, size, &tmp)
+	    || size == 0 /* which also leads to end < begin */)
+		return -EINVAL;
+
+	/* make sure object fit at this offset */
+	if (bo && offset + size > amdgpu_bo_size(bo))
+		return -EINVAL;
+
+	/* Ensure last pfn not exceed max_pfn */
+	lpfn = (saddr + size - 1) >> AMDGPU_GPU_PAGE_SHIFT;
+	if (lpfn >= adev->vm_manager.max_pfn)
+		return -EINVAL;
+
+	return 0;
+}
+
 /**
  * amdgpu_vm_bo_map - map bo inside a vm
  *
@@ -1425,27 +1840,20 @@ static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev,
 int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 		     struct amdgpu_bo_va *bo_va,
 		     uint64_t saddr, uint64_t offset,
-		     uint64_t size, uint64_t flags)
+		     uint64_t size, uint32_t flags)
 {
 	struct amdgpu_bo_va_mapping *mapping, *tmp;
 	struct amdgpu_bo *bo = bo_va->base.bo;
 	struct amdgpu_vm *vm = bo_va->base.vm;
 	uint64_t eaddr;
+	int r;
 
-	/* validate the parameters */
-	if (saddr & ~PAGE_MASK || offset & ~PAGE_MASK ||
-	    size == 0 || size & ~PAGE_MASK)
-		return -EINVAL;
-
-	/* make sure object fit at this offset */
-	eaddr = saddr + size - 1;
-	if (saddr >= eaddr ||
-	    (bo && offset + size > amdgpu_bo_size(bo)) ||
-	    (eaddr >= adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT))
-		return -EINVAL;
+	r = amdgpu_vm_verify_parameters(adev, bo, saddr, offset, size);
+	if (r)
+		return r;
 
 	saddr /= AMDGPU_GPU_PAGE_SIZE;
-	eaddr /= AMDGPU_GPU_PAGE_SIZE;
+	eaddr = saddr + (size - 1) / AMDGPU_GPU_PAGE_SIZE;
 
 	tmp = amdgpu_vm_it_iter_first(&vm->va, saddr, eaddr);
 	if (tmp) {
@@ -1491,24 +1899,16 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev,
 			     struct amdgpu_bo_va *bo_va,
 			     uint64_t saddr, uint64_t offset,
-			     uint64_t size, uint64_t flags)
+			     uint64_t size, uint32_t flags)
 {
 	struct amdgpu_bo_va_mapping *mapping;
 	struct amdgpu_bo *bo = bo_va->base.bo;
 	uint64_t eaddr;
 	int r;
 
-	/* validate the parameters */
-	if (saddr & ~PAGE_MASK || offset & ~PAGE_MASK ||
-	    size == 0 || size & ~PAGE_MASK)
-		return -EINVAL;
-
-	/* make sure object fit at this offset */
-	eaddr = saddr + size - 1;
-	if (saddr >= eaddr ||
-	    (bo && offset + size > amdgpu_bo_size(bo)) ||
-	    (eaddr >= adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT))
-		return -EINVAL;
+	r = amdgpu_vm_verify_parameters(adev, bo, saddr, offset, size);
+	if (r)
+		return r;
 
 	/* Allocate all the needed memory */
 	mapping = kmalloc(sizeof(*mapping), GFP_KERNEL);
@@ -1522,7 +1922,7 @@ int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev,
 	}
 
 	saddr /= AMDGPU_GPU_PAGE_SIZE;
-	eaddr /= AMDGPU_GPU_PAGE_SIZE;
+	eaddr = saddr + (size - 1) / AMDGPU_GPU_PAGE_SIZE;
 
 	mapping->start = saddr;
 	mapping->last = eaddr;
@@ -1555,6 +1955,7 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
 	struct amdgpu_bo_va_mapping *mapping;
 	struct amdgpu_vm *vm = bo_va->base.vm;
 	bool valid = true;
+	int r;
 
 	saddr /= AMDGPU_GPU_PAGE_SIZE;
 
@@ -1575,6 +1976,17 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
 			return -ENOENT;
 	}
 
+	/* It's unlikely to happen that the mapping userq hasn't been idled
+	 * during user requests GEM unmap IOCTL except for forcing the unmap
+	 * from user space.
+	 */
+	if (unlikely(atomic_read(&bo_va->userq_va_mapped) > 0)) {
+		r = amdgpu_userq_gem_va_unmap_validate(adev, mapping, saddr);
+		if (unlikely(r == -EBUSY))
+			dev_warn_once(adev->dev,
+				      "Attempt to unmap an active userq buffer\n");
+	}
+
 	list_del(&mapping->list);
 	amdgpu_vm_it_remove(mapping, &vm->va);
 	mapping->bo_va = NULL;
@@ -1609,10 +2021,14 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,
 	struct amdgpu_bo_va_mapping *before, *after, *tmp, *next;
 	LIST_HEAD(removed);
 	uint64_t eaddr;
+	int r;
+
+	r = amdgpu_vm_verify_parameters(adev, NULL, saddr, 0, size);
+	if (r)
+		return r;
 
-	eaddr = saddr + size - 1;
 	saddr /= AMDGPU_GPU_PAGE_SIZE;
-	eaddr /= AMDGPU_GPU_PAGE_SIZE;
+	eaddr = saddr + (size - 1) / AMDGPU_GPU_PAGE_SIZE;
 
 	/* Allocate all the needed memory */
 	before = kzalloc(sizeof(*before), GFP_KERNEL);
@@ -1674,18 +2090,30 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,
 
 	/* Insert partial mapping before the range */
 	if (!list_empty(&before->list)) {
+		struct amdgpu_bo *bo = before->bo_va->base.bo;
+
 		amdgpu_vm_it_insert(before, &vm->va);
-		if (before->flags & AMDGPU_PTE_PRT)
+		if (before->flags & AMDGPU_VM_PAGE_PRT)
 			amdgpu_vm_prt_get(adev);
+
+		if (amdgpu_vm_is_bo_always_valid(vm, bo) &&
+		    !before->bo_va->base.moved)
+			amdgpu_vm_bo_moved(&before->bo_va->base);
 	} else {
 		kfree(before);
 	}
 
 	/* Insert partial mapping after the range */
 	if (!list_empty(&after->list)) {
+		struct amdgpu_bo *bo = after->bo_va->base.bo;
+
 		amdgpu_vm_it_insert(after, &vm->va);
-		if (after->flags & AMDGPU_PTE_PRT)
+		if (after->flags & AMDGPU_VM_PAGE_PRT)
 			amdgpu_vm_prt_get(adev);
+
+		if (amdgpu_vm_is_bo_always_valid(vm, bo) &&
+		    !after->bo_va->base.moved)
+			amdgpu_vm_bo_moved(&after->bo_va->base);
 	} else {
 		kfree(after);
 	}
@@ -1763,7 +2191,7 @@ void amdgpu_vm_bo_del(struct amdgpu_device *adev,
 
 	if (bo) {
 		dma_resv_assert_held(bo->tbo.base.resv);
-		if (bo->tbo.base.resv == vm->root.bo->tbo.base.resv)
+		if (amdgpu_vm_is_bo_always_valid(vm, bo))
 			ttm_bo_set_bulk_move(&bo->tbo, NULL);
 
 		for (base = &bo_va->base.bo->vm_bo; *base;
@@ -1771,6 +2199,7 @@ void amdgpu_vm_bo_del(struct amdgpu_device *adev,
 			if (*base != &bo_va->base)
 				continue;
 
+			amdgpu_vm_update_stats(*base, bo->tbo.resource, -1);
 			*base = bo_va->base.next;
 			break;
 		}
@@ -1839,25 +2268,19 @@ bool amdgpu_vm_evictable(struct amdgpu_bo *bo)
 /**
  * amdgpu_vm_bo_invalidate - mark the bo as invalid
  *
- * @adev: amdgpu_device pointer
  * @bo: amdgpu buffer object
  * @evicted: is the BO evicted
  *
  * Mark @bo as invalid.
  */
-void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
-			     struct amdgpu_bo *bo, bool evicted)
+void amdgpu_vm_bo_invalidate(struct amdgpu_bo *bo, bool evicted)
 {
 	struct amdgpu_vm_bo_base *bo_base;
 
-	/* shadow bo doesn't have bo base, its validation needs its parent */
-	if (bo->parent && (amdgpu_bo_shadowed(bo->parent) == bo))
-		bo = bo->parent;
-
 	for (bo_base = bo->vm_bo; bo_base; bo_base = bo_base->next) {
 		struct amdgpu_vm *vm = bo_base->vm;
 
-		if (evicted && bo->tbo.base.resv == vm->root.bo->tbo.base.resv) {
+		if (evicted && amdgpu_vm_is_bo_always_valid(vm, bo)) {
 			amdgpu_vm_bo_evicted(bo_base);
 			continue;
 		}
@@ -1868,7 +2291,7 @@ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
 
 		if (bo->tbo.type == ttm_bo_type_kernel)
 			amdgpu_vm_bo_relocated(bo_base);
-		else if (bo->tbo.base.resv == vm->root.bo->tbo.base.resv)
+		else if (amdgpu_vm_is_bo_always_valid(vm, bo))
 			amdgpu_vm_bo_moved(bo_base);
 		else
 			amdgpu_vm_bo_invalidated(bo_base);
@@ -1876,6 +2299,32 @@ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
 }
 
 /**
+ * amdgpu_vm_bo_move - handle BO move
+ *
+ * @bo: amdgpu buffer object
+ * @new_mem: the new placement of the BO move
+ * @evicted: is the BO evicted
+ *
+ * Update the memory stats for the new placement and mark @bo as invalid.
+ */
+void amdgpu_vm_bo_move(struct amdgpu_bo *bo, struct ttm_resource *new_mem,
+		       bool evicted)
+{
+	struct amdgpu_vm_bo_base *bo_base;
+
+	for (bo_base = bo->vm_bo; bo_base; bo_base = bo_base->next) {
+		struct amdgpu_vm *vm = bo_base->vm;
+
+		spin_lock(&vm->status_lock);
+		amdgpu_vm_update_stats_locked(bo_base, bo->tbo.resource, -1);
+		amdgpu_vm_update_stats_locked(bo_base, new_mem, +1);
+		spin_unlock(&vm->status_lock);
+	}
+
+	amdgpu_vm_bo_invalidate(bo, evicted);
+}
+
+/**
  * amdgpu_vm_get_block_size - calculate VM page table size as power of two
  *
  * @vm_size: VM size
@@ -1945,7 +2394,7 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t min_vm_size,
 		phys_ram_gb = ((uint64_t)si.totalram * si.mem_unit +
 			       (1 << 30) - 1) >> 30;
 		vm_size = roundup_pow_of_two(
-			min(max(phys_ram_gb * 3, min_vm_size), max_size));
+			clamp(phys_ram_gb * 3, min_vm_size, max_size));
 	}
 
 	adev->vm_manager.max_pfn = (uint64_t)vm_size << 18;
@@ -1954,7 +2403,7 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t min_vm_size,
 	if (amdgpu_vm_block_size != -1)
 		tmp >>= amdgpu_vm_block_size - 9;
 	tmp = DIV_ROUND_UP(fls64(tmp) - 1, 9) - 1;
-	adev->vm_manager.num_level = min(max_level, (unsigned)tmp);
+	adev->vm_manager.num_level = min_t(unsigned int, max_level, tmp);
 	switch (adev->vm_manager.num_level) {
 	case 3:
 		adev->vm_manager.root_level = AMDGPU_VM_PDB2;
@@ -1984,10 +2433,11 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t min_vm_size,
 	else
 		adev->vm_manager.fragment_size = amdgpu_vm_fragment_size;
 
-	DRM_INFO("vm size is %u GB, %u levels, block size is %u-bit, fragment size is %u-bit\n",
-		 vm_size, adev->vm_manager.num_level + 1,
-		 adev->vm_manager.block_size,
-		 adev->vm_manager.fragment_size);
+	dev_info(
+		adev->dev,
+		"vm size is %u GB, %u levels, block size is %u-bit, fragment size is %u-bit\n",
+		vm_size, adev->vm_manager.num_level + 1,
+		adev->vm_manager.block_size, adev->vm_manager.fragment_size);
 }
 
 /**
@@ -1998,13 +2448,114 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t min_vm_size,
  */
 long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout)
 {
-	timeout = dma_resv_wait_timeout(vm->root.bo->tbo.base.resv,
-					DMA_RESV_USAGE_BOOKKEEP,
-					true, timeout);
+	timeout = drm_sched_entity_flush(&vm->immediate, timeout);
 	if (timeout <= 0)
 		return timeout;
 
-	return dma_fence_wait_timeout(vm->last_unlocked, true, timeout);
+	return drm_sched_entity_flush(&vm->delayed, timeout);
+}
+
+static void amdgpu_vm_destroy_task_info(struct kref *kref)
+{
+	struct amdgpu_task_info *ti = container_of(kref, struct amdgpu_task_info, refcount);
+
+	kfree(ti);
+}
+
+static inline struct amdgpu_vm *
+amdgpu_vm_get_vm_from_pasid(struct amdgpu_device *adev, u32 pasid)
+{
+	struct amdgpu_vm *vm;
+	unsigned long flags;
+
+	xa_lock_irqsave(&adev->vm_manager.pasids, flags);
+	vm = xa_load(&adev->vm_manager.pasids, pasid);
+	xa_unlock_irqrestore(&adev->vm_manager.pasids, flags);
+
+	return vm;
+}
+
+/**
+ * amdgpu_vm_put_task_info - reference down the vm task_info ptr
+ *
+ * @task_info: task_info struct under discussion.
+ *
+ * frees the vm task_info ptr at the last put
+ */
+void amdgpu_vm_put_task_info(struct amdgpu_task_info *task_info)
+{
+	if (task_info)
+		kref_put(&task_info->refcount, amdgpu_vm_destroy_task_info);
+}
+
+/**
+ * amdgpu_vm_get_task_info_vm - Extracts task info for a vm.
+ *
+ * @vm: VM to get info from
+ *
+ * Returns the reference counted task_info structure, which must be
+ * referenced down with amdgpu_vm_put_task_info.
+ */
+struct amdgpu_task_info *
+amdgpu_vm_get_task_info_vm(struct amdgpu_vm *vm)
+{
+	struct amdgpu_task_info *ti = NULL;
+
+	if (vm) {
+		ti = vm->task_info;
+		kref_get(&vm->task_info->refcount);
+	}
+
+	return ti;
+}
+
+/**
+ * amdgpu_vm_get_task_info_pasid - Extracts task info for a PASID.
+ *
+ * @adev: drm device pointer
+ * @pasid: PASID identifier for VM
+ *
+ * Returns the reference counted task_info structure, which must be
+ * referenced down with amdgpu_vm_put_task_info.
+ */
+struct amdgpu_task_info *
+amdgpu_vm_get_task_info_pasid(struct amdgpu_device *adev, u32 pasid)
+{
+	return amdgpu_vm_get_task_info_vm(
+			amdgpu_vm_get_vm_from_pasid(adev, pasid));
+}
+
+static int amdgpu_vm_create_task_info(struct amdgpu_vm *vm)
+{
+	vm->task_info = kzalloc(sizeof(struct amdgpu_task_info), GFP_KERNEL);
+	if (!vm->task_info)
+		return -ENOMEM;
+
+	kref_init(&vm->task_info->refcount);
+	return 0;
+}
+
+/**
+ * amdgpu_vm_set_task_info - Sets VMs task info.
+ *
+ * @vm: vm for which to set the info
+ */
+void amdgpu_vm_set_task_info(struct amdgpu_vm *vm)
+{
+	if (!vm->task_info)
+		return;
+
+	if (vm->task_info->task.pid == current->pid)
+		return;
+
+	vm->task_info->task.pid = current->pid;
+	get_task_comm(vm->task_info->task.comm, current);
+
+	if (current->group_leader->mm != current->mm)
+		return;
+
+	vm->task_info->tgid = current->group_leader->pid;
+	get_task_comm(vm->task_info->process_name, current->group_leader);
 }
 
 /**
@@ -2012,13 +2563,16 @@ long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout)
  *
  * @adev: amdgpu_device pointer
  * @vm: requested vm
+ * @xcp_id: GPU partition selection id
+ * @pasid: the pasid the VM is using on this GPU
  *
  * Init @vm fields.
  *
  * Returns:
  * 0 for success, error for failure.
  */
-int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm)
+int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+		   int32_t xcp_id, uint32_t pasid)
 {
 	struct amdgpu_bo *root_bo;
 	struct amdgpu_bo_vm *root;
@@ -2028,6 +2582,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++)
 		vm->reserved_vmid[i] = NULL;
 	INIT_LIST_HEAD(&vm->evicted);
+	INIT_LIST_HEAD(&vm->evicted_user);
 	INIT_LIST_HEAD(&vm->relocated);
 	INIT_LIST_HEAD(&vm->moved);
 	INIT_LIST_HEAD(&vm->idle);
@@ -2035,30 +2590,21 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 	spin_lock_init(&vm->status_lock);
 	INIT_LIST_HEAD(&vm->freed);
 	INIT_LIST_HEAD(&vm->done);
-	INIT_LIST_HEAD(&vm->pt_freed);
-	INIT_WORK(&vm->pt_free_work, amdgpu_vm_pt_free_work);
+	INIT_KFIFO(vm->faults);
 
-	/* create scheduler entities for page table updates */
-	r = drm_sched_entity_init(&vm->immediate, DRM_SCHED_PRIORITY_NORMAL,
-				  adev->vm_manager.vm_pte_scheds,
-				  adev->vm_manager.vm_pte_num_scheds, NULL);
+	r = amdgpu_vm_init_entities(adev, vm);
 	if (r)
 		return r;
 
-	r = drm_sched_entity_init(&vm->delayed, DRM_SCHED_PRIORITY_NORMAL,
-				  adev->vm_manager.vm_pte_scheds,
-				  adev->vm_manager.vm_pte_num_scheds, NULL);
-	if (r)
-		goto error_free_immediate;
+	ttm_lru_bulk_move_init(&vm->lru_bulk_move);
 
-	vm->pte_support_ats = false;
 	vm->is_compute_context = false;
 
 	vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode &
 				    AMDGPU_VM_USE_CPU_FOR_GFX);
 
-	DRM_DEBUG_DRIVER("VM update mode is %s\n",
-			 vm->use_cpu_for_update ? "CPU" : "SDMA");
+	dev_dbg(adev->dev, "VM update mode is %s\n",
+		vm->use_cpu_for_update ? "CPU" : "SDMA");
 	WARN_ONCE((vm->use_cpu_for_update &&
 		   !amdgpu_gmc_vram_full_visible(&adev->gmc)),
 		  "CPU update of VM recommended only for large BAR system\n");
@@ -2067,53 +2613,70 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 		vm->update_funcs = &amdgpu_vm_cpu_funcs;
 	else
 		vm->update_funcs = &amdgpu_vm_sdma_funcs;
-	vm->last_update = NULL;
+
+	vm->last_update = dma_fence_get_stub();
 	vm->last_unlocked = dma_fence_get_stub();
 	vm->last_tlb_flush = dma_fence_get_stub();
+	vm->generation = amdgpu_vm_generation(adev, NULL);
 
 	mutex_init(&vm->eviction_lock);
 	vm->evicting = false;
+	vm->tlb_fence_context = dma_fence_context_alloc(1);
 
 	r = amdgpu_vm_pt_create(adev, vm, adev->vm_manager.root_level,
-				false, &root);
+				false, &root, xcp_id);
 	if (r)
 		goto error_free_delayed;
-	root_bo = &root->bo;
+
+	root_bo = amdgpu_bo_ref(&root->bo);
 	r = amdgpu_bo_reserve(root_bo, true);
-	if (r)
-		goto error_free_root;
+	if (r) {
+		amdgpu_bo_unref(&root_bo);
+		goto error_free_delayed;
+	}
 
+	amdgpu_vm_bo_base_init(&vm->root, vm, root_bo);
 	r = dma_resv_reserve_fences(root_bo->tbo.base.resv, 1);
 	if (r)
-		goto error_unreserve;
-
-	amdgpu_vm_bo_base_init(&vm->root, vm, root_bo);
+		goto error_free_root;
 
 	r = amdgpu_vm_pt_clear(adev, vm, root, false);
 	if (r)
-		goto error_unreserve;
+		goto error_free_root;
 
-	amdgpu_bo_unreserve(vm->root.bo);
+	r = amdgpu_vm_create_task_info(vm);
+	if (r)
+		dev_dbg(adev->dev, "Failed to create task info for VM\n");
 
-	INIT_KFIFO(vm->faults);
+	/* Store new PASID in XArray (if non-zero) */
+	if (pasid != 0) {
+		r = xa_err(xa_store_irq(&adev->vm_manager.pasids, pasid, vm, GFP_KERNEL));
+		if (r < 0)
+			goto error_free_root;
 
-	return 0;
+		vm->pasid = pasid;
+	}
 
-error_unreserve:
 	amdgpu_bo_unreserve(vm->root.bo);
+	amdgpu_bo_unref(&root_bo);
+
+	return 0;
 
 error_free_root:
-	amdgpu_bo_unref(&root->shadow);
+	/* If PASID was partially set, erase it from XArray before failing */
+	if (vm->pasid != 0) {
+		xa_erase_irq(&adev->vm_manager.pasids, vm->pasid);
+		vm->pasid = 0;
+	}
+	amdgpu_vm_pt_free_root(adev, vm);
+	amdgpu_bo_unreserve(vm->root.bo);
 	amdgpu_bo_unref(&root_bo);
-	vm->root.bo = NULL;
 
 error_free_delayed:
 	dma_fence_put(vm->last_tlb_flush);
 	dma_fence_put(vm->last_unlocked);
-	drm_sched_entity_destroy(&vm->delayed);
-
-error_free_immediate:
-	drm_sched_entity_destroy(&vm->immediate);
+	ttm_lru_bulk_move_fini(&adev->mman.bdev, &vm->lru_bulk_move);
+	amdgpu_vm_fini_entities(vm);
 
 	return r;
 }
@@ -2139,35 +2702,17 @@ error_free_immediate:
  */
 int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 {
-	bool pte_support_ats = (adev->asic_type == CHIP_RAVEN);
 	int r;
 
 	r = amdgpu_bo_reserve(vm->root.bo, true);
 	if (r)
 		return r;
 
-	/* Sanity checks */
-	if (!amdgpu_vm_pt_is_root_clean(adev, vm)) {
-		r = -EINVAL;
-		goto unreserve_bo;
-	}
-
-	/* Check if PD needs to be reinitialized and do it before
-	 * changing any other state, in case it fails.
-	 */
-	if (pte_support_ats != vm->pte_support_ats) {
-		vm->pte_support_ats = pte_support_ats;
-		r = amdgpu_vm_pt_clear(adev, vm, to_amdgpu_bo_vm(vm->root.bo),
-				       false);
-		if (r)
-			goto unreserve_bo;
-	}
-
 	/* Update VM state */
 	vm->use_cpu_for_update = !!(adev->vm_manager.vm_update_mode &
 				    AMDGPU_VM_USE_CPU_FOR_COMPUTE);
-	DRM_DEBUG_DRIVER("VM update mode is %s\n",
-			 vm->use_cpu_for_update ? "CPU" : "SDMA");
+	dev_dbg(adev->dev, "VM update mode is %s\n",
+		vm->use_cpu_for_update ? "CPU" : "SDMA");
 	WARN_ONCE((vm->use_cpu_for_update &&
 		   !amdgpu_gmc_vram_full_visible(&adev->gmc)),
 		  "CPU update of VM recommended only for large BAR system\n");
@@ -2180,43 +2725,31 @@ int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 			goto unreserve_bo;
 
 		vm->update_funcs = &amdgpu_vm_cpu_funcs;
+		r = amdgpu_vm_pt_map_tables(adev, vm);
+		if (r)
+			goto unreserve_bo;
+
 	} else {
 		vm->update_funcs = &amdgpu_vm_sdma_funcs;
 	}
-	/*
-	 * Make sure root PD gets mapped. As vm_update_mode could be changed
-	 * when turning a GFX VM into a compute VM.
-	 */
-	r = vm->update_funcs->map_table(to_amdgpu_bo_vm(vm->root.bo));
-	if (r)
-		goto unreserve_bo;
 
 	dma_fence_put(vm->last_update);
-	vm->last_update = NULL;
+	vm->last_update = dma_fence_get_stub();
 	vm->is_compute_context = true;
 
-	/* Free the shadow bo for compute VM */
-	amdgpu_bo_unref(&to_amdgpu_bo_vm(vm->root.bo)->shadow);
-
-	goto unreserve_bo;
-
 unreserve_bo:
 	amdgpu_bo_unreserve(vm->root.bo);
 	return r;
 }
 
-/**
- * amdgpu_vm_release_compute - release a compute vm
- * @adev: amdgpu_device pointer
- * @vm: a vm turned into compute vm by calling amdgpu_vm_make_compute
- *
- * This is a correspondant of amdgpu_vm_make_compute. It decouples compute
- * pasid from vm. Compute should stop use of vm after this call.
- */
-void amdgpu_vm_release_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm)
+static int amdgpu_vm_stats_is_zero(struct amdgpu_vm *vm)
 {
-	amdgpu_vm_set_pasid(adev, vm, 0);
-	vm->is_compute_context = false;
+	for (int i = 0; i < __AMDGPU_PL_NUM; ++i) {
+		if (!(drm_memory_stats_is_zero(&vm->stats[i].drm) &&
+		      vm->stats[i].evicted == 0))
+			return false;
+	}
+	return true;
 }
 
 /**
@@ -2238,11 +2771,13 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 
 	amdgpu_amdkfd_gpuvm_destroy_cb(adev, vm);
 
-	flush_work(&vm->pt_free_work);
-
 	root = amdgpu_bo_ref(vm->root.bo);
 	amdgpu_bo_reserve(root, true);
-	amdgpu_vm_set_pasid(adev, vm, 0);
+	/* Remove PASID mapping before destroying VM */
+	if (vm->pasid != 0) {
+		xa_erase_irq(&adev->vm_manager.pasids, vm->pasid);
+		vm->pasid = 0;
+	}
 	dma_fence_wait(vm->last_unlocked, false);
 	dma_fence_put(vm->last_unlocked);
 	dma_fence_wait(vm->last_tlb_flush, false);
@@ -2252,7 +2787,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 	dma_fence_put(vm->last_tlb_flush);
 
 	list_for_each_entry_safe(mapping, tmp, &vm->freed, list) {
-		if (mapping->flags & AMDGPU_PTE_PRT && prt_fini_needed) {
+		if (mapping->flags & AMDGPU_VM_PAGE_PRT && prt_fini_needed) {
 			amdgpu_vm_prt_fini(adev, vm);
 			prt_fini_needed = false;
 		}
@@ -2266,8 +2801,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 	amdgpu_bo_unref(&root);
 	WARN_ON(vm->root.bo);
 
-	drm_sched_entity_destroy(&vm->immediate);
-	drm_sched_entity_destroy(&vm->delayed);
+	amdgpu_vm_fini_entities(vm);
 
 	if (!RB_EMPTY_ROOT(&vm->va.rb_root)) {
 		dev_err(adev->dev, "still active bo inside vm\n");
@@ -2282,8 +2816,22 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 	}
 
 	dma_fence_put(vm->last_update);
-	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++)
+
+	for (i = 0; i < AMDGPU_MAX_VMHUBS; i++) {
 		amdgpu_vmid_free_reserved(adev, vm, i);
+	}
+
+	ttm_lru_bulk_move_fini(&adev->mman.bdev, &vm->lru_bulk_move);
+
+	if (!amdgpu_vm_stats_is_zero(vm)) {
+		struct amdgpu_task_info *ti = vm->task_info;
+
+		dev_warn(adev->dev,
+			 "VM memory stats for proc %s(%d) task %s(%d) is non-zero when fini\n",
+			 ti->process_name, ti->task.pid, ti->task.comm, ti->tgid);
+	}
+
+	amdgpu_vm_put_task_info(vm->task_info);
 }
 
 /**
@@ -2295,8 +2843,6 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
  */
 void amdgpu_vm_manager_init(struct amdgpu_device *adev)
 {
-	unsigned i;
-
 	/* Concurrent flushes are only possible starting with Vega10 and
 	 * are broken on Navi10 and Navi14.
 	 */
@@ -2305,11 +2851,6 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev)
 					      adev->asic_type == CHIP_NAVI14);
 	amdgpu_vmid_mgr_init(adev);
 
-	adev->vm_manager.fence_context =
-		dma_fence_context_alloc(AMDGPU_MAX_RINGS);
-	for (i = 0; i < AMDGPU_MAX_RINGS; ++i)
-		adev->vm_manager.seqno[i] = 0;
-
 	spin_lock_init(&adev->vm_manager.prt_lock);
 	atomic_set(&adev->vm_manager.num_prt_users, 0);
 
@@ -2366,18 +2907,18 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	union drm_amdgpu_vm *args = data;
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct amdgpu_fpriv *fpriv = filp->driver_priv;
-	int r;
+	struct amdgpu_vm *vm = &fpriv->vm;
+
+	/* No valid flags defined yet */
+	if (args->in.flags)
+		return -EINVAL;
 
 	switch (args->in.op) {
 	case AMDGPU_VM_OP_RESERVE_VMID:
 		/* We only have requirement to reserve vmid from gfxhub */
-		r = amdgpu_vmid_alloc_reserved(adev, &fpriv->vm,
-					       AMDGPU_GFXHUB_0);
-		if (r)
-			return r;
-		break;
+		return amdgpu_vmid_alloc_reserved(adev, vm, AMDGPU_GFXHUB(0));
 	case AMDGPU_VM_OP_UNRESERVE_VMID:
-		amdgpu_vmid_free_reserved(adev, &fpriv->vm, AMDGPU_GFXHUB_0);
+		amdgpu_vmid_free_reserved(adev, vm, AMDGPU_GFXHUB(0));
 		break;
 	default:
 		return -EINVAL;
@@ -2387,51 +2928,13 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 }
 
 /**
- * amdgpu_vm_get_task_info - Extracts task info for a PASID.
- *
- * @adev: drm device pointer
- * @pasid: PASID identifier for VM
- * @task_info: task_info to fill.
- */
-void amdgpu_vm_get_task_info(struct amdgpu_device *adev, u32 pasid,
-			 struct amdgpu_task_info *task_info)
-{
-	struct amdgpu_vm *vm;
-	unsigned long flags;
-
-	xa_lock_irqsave(&adev->vm_manager.pasids, flags);
-
-	vm = xa_load(&adev->vm_manager.pasids, pasid);
-	if (vm)
-		*task_info = vm->task_info;
-
-	xa_unlock_irqrestore(&adev->vm_manager.pasids, flags);
-}
-
-/**
- * amdgpu_vm_set_task_info - Sets VMs task info.
- *
- * @vm: vm for which to set the info
- */
-void amdgpu_vm_set_task_info(struct amdgpu_vm *vm)
-{
-	if (vm->task_info.pid)
-		return;
-
-	vm->task_info.pid = current->pid;
-	get_task_comm(vm->task_info.task_name, current);
-
-	if (current->group_leader->mm != current->mm)
-		return;
-
-	vm->task_info.tgid = current->group_leader->pid;
-	get_task_comm(vm->task_info.process_name, current->group_leader);
-}
-
-/**
  * amdgpu_vm_handle_fault - graceful handling of VM faults.
  * @adev: amdgpu device pointer
  * @pasid: PASID of the VM
+ * @ts: Timestamp of the fault
+ * @vmid: VMID, only used for GFX 9.4.3.
+ * @node_id: Node_id received in IH cookie. Only applicable for
+ *           GFX 9.4.3.
  * @addr: Address of the fault
  * @write_fault: true is write fault, false is read fault
  *
@@ -2439,7 +2942,8 @@ void amdgpu_vm_set_task_info(struct amdgpu_vm *vm)
  * shouldn't be reported any more.
  */
 bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid,
-			    uint64_t addr, bool write_fault)
+			    u32 vmid, u32 node_id, uint64_t addr, uint64_t ts,
+			    bool write_fault)
 {
 	bool is_compute_context = false;
 	struct amdgpu_bo *root;
@@ -2463,8 +2967,8 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid,
 
 	addr /= AMDGPU_GPU_PAGE_SIZE;
 
-	if (is_compute_context &&
-	    !svm_range_restore_pages(adev, pasid, addr, write_fault)) {
+	if (is_compute_context && !svm_range_restore_pages(adev, pasid, vmid,
+	    node_id, addr, ts, write_fault)) {
 		amdgpu_bo_unref(&root);
 		return true;
 	}
@@ -2489,7 +2993,7 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid,
 		/* Intentionally setting invalid PTE flag
 		 * combination to force a no-retry-fault
 		 */
-		flags = AMDGPU_PTE_SNOOPED | AMDGPU_PTE_PRT;
+		flags = AMDGPU_VM_NORETRY_FLAGS;
 		value = 0;
 	} else if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_NEVER) {
 		/* Redirect the access to the dummy page */
@@ -2508,8 +3012,8 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid,
 		goto error_unlock;
 	}
 
-	r = amdgpu_vm_update_range(adev, vm, true, false, false, NULL, addr,
-				   addr, flags, value, 0, NULL, NULL, NULL);
+	r = amdgpu_vm_update_range(adev, vm, true, false, false, false,
+				   NULL, addr, addr, flags, value, 0, NULL, NULL, NULL);
 	if (r)
 		goto error_unlock;
 
@@ -2518,7 +3022,7 @@ bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid,
 error_unlock:
 	amdgpu_bo_unreserve(root);
 	if (r < 0)
-		DRM_ERROR("Can't handle page fault (%d)\n", r);
+		dev_err(adev->dev, "Can't handle page fault (%d)\n", r);
 
 error_unref:
 	amdgpu_bo_unref(&root);
@@ -2552,6 +3056,8 @@ void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m)
 	unsigned int total_done_objs = 0;
 	unsigned int id = 0;
 
+	amdgpu_vm_assert_locked(vm);
+
 	spin_lock(&vm->status_lock);
 	seq_puts(m, "\tIdle BOs:\n");
 	list_for_each_entry_safe(bo_va, tmp, &vm->idle, base.vm_status) {
@@ -2621,3 +3127,83 @@ void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m)
 		   total_done_objs);
 }
 #endif
+
+/**
+ * amdgpu_vm_update_fault_cache - update cached fault into.
+ * @adev: amdgpu device pointer
+ * @pasid: PASID of the VM
+ * @addr: Address of the fault
+ * @status: GPUVM fault status register
+ * @vmhub: which vmhub got the fault
+ *
+ * Cache the fault info for later use by userspace in debugging.
+ */
+void amdgpu_vm_update_fault_cache(struct amdgpu_device *adev,
+				  unsigned int pasid,
+				  uint64_t addr,
+				  uint32_t status,
+				  unsigned int vmhub)
+{
+	struct amdgpu_vm *vm;
+	unsigned long flags;
+
+	xa_lock_irqsave(&adev->vm_manager.pasids, flags);
+
+	vm = xa_load(&adev->vm_manager.pasids, pasid);
+	/* Don't update the fault cache if status is 0.  In the multiple
+	 * fault case, subsequent faults will return a 0 status which is
+	 * useless for userspace and replaces the useful fault status, so
+	 * only update if status is non-0.
+	 */
+	if (vm && status) {
+		vm->fault_info.addr = addr;
+		vm->fault_info.status = status;
+		/*
+		 * Update the fault information globally for later usage
+		 * when vm could be stale or freed.
+		 */
+		adev->vm_manager.fault_info.addr = addr;
+		adev->vm_manager.fault_info.vmhub = vmhub;
+		adev->vm_manager.fault_info.status = status;
+
+		if (AMDGPU_IS_GFXHUB(vmhub)) {
+			vm->fault_info.vmhub = AMDGPU_VMHUB_TYPE_GFX;
+			vm->fault_info.vmhub |=
+				(vmhub - AMDGPU_GFXHUB_START) << AMDGPU_VMHUB_IDX_SHIFT;
+		} else if (AMDGPU_IS_MMHUB0(vmhub)) {
+			vm->fault_info.vmhub = AMDGPU_VMHUB_TYPE_MM0;
+			vm->fault_info.vmhub |=
+				(vmhub - AMDGPU_MMHUB0_START) << AMDGPU_VMHUB_IDX_SHIFT;
+		} else if (AMDGPU_IS_MMHUB1(vmhub)) {
+			vm->fault_info.vmhub = AMDGPU_VMHUB_TYPE_MM1;
+			vm->fault_info.vmhub |=
+				(vmhub - AMDGPU_MMHUB1_START) << AMDGPU_VMHUB_IDX_SHIFT;
+		} else {
+			WARN_ONCE(1, "Invalid vmhub %u\n", vmhub);
+		}
+	}
+	xa_unlock_irqrestore(&adev->vm_manager.pasids, flags);
+}
+
+/**
+ * amdgpu_vm_is_bo_always_valid - check if the BO is VM always valid
+ *
+ * @vm: VM to test against.
+ * @bo: BO to be tested.
+ *
+ * Returns true if the BO shares the dma_resv object with the root PD and is
+ * always guaranteed to be valid inside the VM.
+ */
+bool amdgpu_vm_is_bo_always_valid(struct amdgpu_vm *vm, struct amdgpu_bo *bo)
+{
+	return bo && bo->tbo.base.resv == vm->root.bo->tbo.base.resv;
+}
+
+void amdgpu_vm_print_task_info(struct amdgpu_device *adev,
+			       struct amdgpu_task_info *task_info)
+{
+	dev_err(adev->dev,
+		" Process %s pid %d thread %s pid %d\n",
+		task_info->process_name, task_info->tgid,
+		task_info->task.comm, task_info->task.pid);
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
index 6f085f0b4ef3..15d757c016cb 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
@@ -35,12 +35,14 @@
 #include "amdgpu_sync.h"
 #include "amdgpu_ring.h"
 #include "amdgpu_ids.h"
+#include "amdgpu_ttm.h"
+
+struct drm_exec;
 
 struct amdgpu_bo_va;
 struct amdgpu_job;
 struct amdgpu_bo_list_entry;
 struct amdgpu_bo_vm;
-struct amdgpu_mem_stats;
 
 /*
  * GPUVM handling
@@ -84,10 +86,19 @@ struct amdgpu_mem_stats;
 /* PDE Block Fragment Size for VEGA10 */
 #define AMDGPU_PDE_BFS(a)	((uint64_t)a << 59)
 
+/* Flag combination to set no-retry with TF disabled */
+#define AMDGPU_VM_NORETRY_FLAGS	(AMDGPU_PTE_EXECUTABLE | AMDGPU_PDE_PTE | \
+				AMDGPU_PTE_TF)
 
+/* Flag combination to set no-retry with TF enabled */
+#define AMDGPU_VM_NORETRY_FLAGS_TF (AMDGPU_PTE_VALID | AMDGPU_PTE_SYSTEM | \
+				   AMDGPU_PTE_PRT)
 /* For GFX9 */
-#define AMDGPU_PTE_MTYPE_VG10(a)	((uint64_t)(a) << 57)
-#define AMDGPU_PTE_MTYPE_VG10_MASK	AMDGPU_PTE_MTYPE_VG10(3ULL)
+#define AMDGPU_PTE_MTYPE_VG10_SHIFT(mtype)	((uint64_t)(mtype) << 57)
+#define AMDGPU_PTE_MTYPE_VG10_MASK	AMDGPU_PTE_MTYPE_VG10_SHIFT(3ULL)
+#define AMDGPU_PTE_MTYPE_VG10(flags, mtype)			\
+	(((uint64_t)(flags) & (~AMDGPU_PTE_MTYPE_VG10_MASK)) |	\
+	  AMDGPU_PTE_MTYPE_VG10_SHIFT(mtype))
 
 #define AMDGPU_MTYPE_NC 0
 #define AMDGPU_MTYPE_CC 2
@@ -100,25 +111,74 @@ struct amdgpu_mem_stats;
                                 | AMDGPU_PTE_MTYPE_VG10(AMDGPU_MTYPE_CC))
 
 /* gfx10 */
-#define AMDGPU_PTE_MTYPE_NV10(a)       ((uint64_t)(a) << 48)
-#define AMDGPU_PTE_MTYPE_NV10_MASK     AMDGPU_PTE_MTYPE_NV10(7ULL)
+#define AMDGPU_PTE_MTYPE_NV10_SHIFT(mtype)	((uint64_t)(mtype) << 48)
+#define AMDGPU_PTE_MTYPE_NV10_MASK     AMDGPU_PTE_MTYPE_NV10_SHIFT(7ULL)
+#define AMDGPU_PTE_MTYPE_NV10(flags, mtype)			\
+	(((uint64_t)(flags) & (~AMDGPU_PTE_MTYPE_NV10_MASK)) |	\
+	  AMDGPU_PTE_MTYPE_NV10_SHIFT(mtype))
+
+/* gfx12 */
+#define AMDGPU_PTE_PRT_GFX12		(1ULL << 56)
+#define AMDGPU_PTE_PRT_FLAG(adev)	\
+	((amdgpu_ip_version((adev), GC_HWIP, 0) >= IP_VERSION(12, 0, 0)) ? AMDGPU_PTE_PRT_GFX12 : AMDGPU_PTE_PRT)
+
+#define AMDGPU_PTE_MTYPE_GFX12_SHIFT(mtype)	((uint64_t)(mtype) << 54)
+#define AMDGPU_PTE_MTYPE_GFX12_MASK	AMDGPU_PTE_MTYPE_GFX12_SHIFT(3ULL)
+#define AMDGPU_PTE_MTYPE_GFX12(flags, mtype)				\
+	(((uint64_t)(flags) & (~AMDGPU_PTE_MTYPE_GFX12_MASK)) |	\
+	  AMDGPU_PTE_MTYPE_GFX12_SHIFT(mtype))
+
+#define AMDGPU_PTE_DCC			(1ULL << 58)
+#define AMDGPU_PTE_IS_PTE		(1ULL << 63)
+
+/* PDE Block Fragment Size for gfx v12 */
+#define AMDGPU_PDE_BFS_GFX12(a)		((uint64_t)((a) & 0x1fULL) << 58)
+#define AMDGPU_PDE_BFS_FLAG(adev, a)	\
+	((amdgpu_ip_version((adev), GC_HWIP, 0) >= IP_VERSION(12, 0, 0)) ? AMDGPU_PDE_BFS_GFX12(a) : AMDGPU_PDE_BFS(a))
+/* PDE is handled as PTE for gfx v12 */
+#define AMDGPU_PDE_PTE_GFX12		(1ULL << 63)
+#define AMDGPU_PDE_PTE_FLAG(adev)	\
+	((amdgpu_ip_version((adev), GC_HWIP, 0) >= IP_VERSION(12, 0, 0)) ? AMDGPU_PDE_PTE_GFX12 : AMDGPU_PDE_PTE)
 
 /* How to program VM fault handling */
 #define AMDGPU_VM_FAULT_STOP_NEVER	0
 #define AMDGPU_VM_FAULT_STOP_FIRST	1
 #define AMDGPU_VM_FAULT_STOP_ALWAYS	2
 
-/* Reserve 4MB VRAM for page tables */
+/* How much VRAM be reserved for page tables */
 #define AMDGPU_VM_RESERVED_VRAM		(8ULL << 20)
 
-/* max number of VMHUB */
-#define AMDGPU_MAX_VMHUBS			3
-#define AMDGPU_GFXHUB_0				0
-#define AMDGPU_MMHUB_0				1
-#define AMDGPU_MMHUB_1				2
-
-/* Reserve 2MB at top/bottom of address space for kernel use */
-#define AMDGPU_VA_RESERVED_SIZE			(2ULL << 20)
+/*
+ * max number of VMHUB
+ * layout: max 8 GFXHUB + 4 MMHUB0 + 1 MMHUB1
+ */
+#define AMDGPU_MAX_VMHUBS			13
+#define AMDGPU_GFXHUB_START			0
+#define AMDGPU_MMHUB0_START			8
+#define AMDGPU_MMHUB1_START			12
+#define AMDGPU_GFXHUB(x)			(AMDGPU_GFXHUB_START + (x))
+#define AMDGPU_MMHUB0(x)			(AMDGPU_MMHUB0_START + (x))
+#define AMDGPU_MMHUB1(x)			(AMDGPU_MMHUB1_START + (x))
+
+#define AMDGPU_IS_GFXHUB(x) ((x) >= AMDGPU_GFXHUB_START && (x) < AMDGPU_MMHUB0_START)
+#define AMDGPU_IS_MMHUB0(x) ((x) >= AMDGPU_MMHUB0_START && (x) < AMDGPU_MMHUB1_START)
+#define AMDGPU_IS_MMHUB1(x) ((x) >= AMDGPU_MMHUB1_START && (x) < AMDGPU_MAX_VMHUBS)
+
+/* Reserve space at top/bottom of address space for kernel use */
+#define AMDGPU_VA_RESERVED_CSA_SIZE		(2ULL << 20)
+#define AMDGPU_VA_RESERVED_CSA_START(adev)	(((adev)->vm_manager.max_pfn \
+						  << AMDGPU_GPU_PAGE_SHIFT)  \
+						 - AMDGPU_VA_RESERVED_CSA_SIZE)
+#define AMDGPU_VA_RESERVED_SEQ64_SIZE		(2ULL << 20)
+#define AMDGPU_VA_RESERVED_SEQ64_START(adev)	(AMDGPU_VA_RESERVED_CSA_START(adev) \
+						 - AMDGPU_VA_RESERVED_SEQ64_SIZE)
+#define AMDGPU_VA_RESERVED_TRAP_SIZE		(2ULL << 12)
+#define AMDGPU_VA_RESERVED_TRAP_START(adev)	(AMDGPU_VA_RESERVED_SEQ64_START(adev) \
+						 - AMDGPU_VA_RESERVED_TRAP_SIZE)
+#define AMDGPU_VA_RESERVED_BOTTOM		(1ULL << 16)
+#define AMDGPU_VA_RESERVED_TOP			(AMDGPU_VA_RESERVED_TRAP_SIZE + \
+						 AMDGPU_VA_RESERVED_SEQ64_SIZE + \
+						 AMDGPU_VA_RESERVED_CSA_SIZE)
 
 /* See vm_update_mode */
 #define AMDGPU_VM_USE_CPU_FOR_GFX (1 << 0)
@@ -143,9 +203,13 @@ struct amdgpu_vm_bo_base {
 	/* protected by bo being reserved */
 	struct amdgpu_vm_bo_base	*next;
 
-	/* protected by spinlock */
+	/* protected by vm status_lock */
 	struct list_head		vm_status;
 
+	/* if the bo is counted as shared in mem stats
+	 * protected by vm status_lock */
+	bool				shared;
+
 	/* protected by the BO being reserved */
 	bool				moved;
 };
@@ -172,10 +236,10 @@ struct amdgpu_vm_pte_funcs {
 };
 
 struct amdgpu_task_info {
-	char	process_name[TASK_COMM_LEN];
-	char	task_name[TASK_COMM_LEN];
-	pid_t	pid;
-	pid_t	tgid;
+	struct drm_wedge_task_info task;
+	char		process_name[TASK_COMM_LEN];
+	pid_t		tgid;
+	struct kref	refcount;
 };
 
 /**
@@ -225,15 +289,26 @@ struct amdgpu_vm_update_params {
 	unsigned int num_dw_left;
 
 	/**
-	 * @table_freed: return true if page table is freed when updating
+	 * @needs_flush: true whenever we need to invalidate the TLB
+	 */
+	bool needs_flush;
+
+	/**
+	 * @allow_override: true for memory that is not uncached: allows MTYPE
+	 * to be overridden for NUMA local memory.
+	 */
+	bool allow_override;
+
+	/**
+	 * @tlb_flush_waitlist: temporary storage for BOs until tlb_flush
 	 */
-	bool table_freed;
+	struct list_head tlb_flush_waitlist;
 };
 
 struct amdgpu_vm_update_funcs {
 	int (*map_table)(struct amdgpu_bo_vm *bo);
-	int (*prepare)(struct amdgpu_vm_update_params *p, struct dma_resv *resv,
-		       enum amdgpu_sync_mode sync_mode);
+	int (*prepare)(struct amdgpu_vm_update_params *p,
+		       struct amdgpu_sync *sync, u64 k_job_id);
 	int (*update)(struct amdgpu_vm_update_params *p,
 		      struct amdgpu_bo_vm *bo, uint64_t pe, uint64_t addr,
 		      unsigned count, uint32_t incr, uint64_t flags);
@@ -241,6 +316,22 @@ struct amdgpu_vm_update_funcs {
 		      struct dma_fence **fence);
 };
 
+struct amdgpu_vm_fault_info {
+	/* fault address */
+	uint64_t	addr;
+	/* fault status register */
+	uint32_t	status;
+	/* which vmhub? gfxhub, mmhub, etc. */
+	unsigned int	vmhub;
+};
+
+struct amdgpu_mem_stats {
+	struct drm_memory_stats drm;
+
+	/* buffers that requested this placement but are currently evicted */
+	uint64_t evicted;
+};
+
 struct amdgpu_vm {
 	/* tree of virtual addresses mapped */
 	struct rb_root_cached	va;
@@ -255,7 +346,17 @@ struct amdgpu_vm {
 	/* Lock to protect vm_bo add/del/move on all lists of vm */
 	spinlock_t		status_lock;
 
-	/* BOs who needs a validation */
+	/* Memory statistics for this vm, protected by status_lock */
+	struct amdgpu_mem_stats stats[__AMDGPU_PL_NUM];
+
+	/*
+	 * The following lists contain amdgpu_vm_bo_base objects for either
+	 * PDs, PTs or per VM BOs. The state transits are:
+	 *
+	 * evicted -> relocated (PDs, PTs) or moved (per VM BOs) -> idle
+	 */
+
+	/* Per-VM and PT BOs who needs a validation */
 	struct list_head	evicted;
 
 	/* PT BOs which relocated and their parent need an update */
@@ -267,18 +368,28 @@ struct amdgpu_vm {
 	/* All BOs of this VM not currently in the state machine */
 	struct list_head	idle;
 
+	/*
+	 * The following lists contain amdgpu_vm_bo_base objects for BOs which
+	 * have their own dma_resv object and not depend on the root PD. Their
+	 * state transits are:
+	 *
+	 * evicted_user or invalidated -> done
+	 */
+
+	/* BOs for user mode queues that need a validation */
+	struct list_head	evicted_user;
+
 	/* regular invalidated BOs, but not yet updated in the PT */
 	struct list_head	invalidated;
 
-	/* BO mappings freed, but not yet updated in the PT */
-	struct list_head	freed;
-
 	/* BOs which are invalidated, has been updated in the PTs */
 	struct list_head        done;
 
-	/* PT BOs scheduled to free and fill with zero if vm_resv is not hold */
-	struct list_head	pt_freed;
-	struct work_struct	pt_free_work;
+	/*
+	 * This list contains amdgpu_bo_va_mapping objects which have been freed
+	 * but not updated in the PTs
+	 */
+	struct list_head	freed;
 
 	/* contains the page directory */
 	struct amdgpu_vm_bo_base     root;
@@ -291,12 +402,17 @@ struct amdgpu_vm {
 	/* Last finished delayed update */
 	atomic64_t		tlb_seq;
 	struct dma_fence	*last_tlb_flush;
+	atomic64_t		kfd_last_flushed_seq;
+	uint64_t		tlb_fence_context;
+
+	/* How many times we had to re-generate the page tables */
+	uint64_t		generation;
 
 	/* Last unlocked submission to the scheduler entities */
 	struct dma_fence	*last_unlocked;
 
 	unsigned int		pasid;
-	bool			reserved_vmid[AMDGPU_MAX_VMHUBS];
+	struct amdgpu_vmid	*reserved_vmid[AMDGPU_MAX_VMHUBS];
 
 	/* Flag to indicate if VM tables are updated by CPU or GPU (SDMA) */
 	bool					use_cpu_for_update;
@@ -304,9 +420,6 @@ struct amdgpu_vm {
 	/* Functions to use for VM table updates */
 	const struct amdgpu_vm_update_funcs	*update_funcs;
 
-	/* Flag to indicate ATS support from PTE for GFX9 */
-	bool			pte_support_ats;
-
 	/* Up to 128 pending retry page faults */
 	DECLARE_KFIFO(faults, u64, 128);
 
@@ -320,12 +433,18 @@ struct amdgpu_vm {
 	uint64_t		pd_phys_addr;
 
 	/* Some basic info about the task */
-	struct amdgpu_task_info task_info;
+	struct amdgpu_task_info *task_info;
 
 	/* Store positions of group of BOs */
 	struct ttm_lru_bulk_move lru_bulk_move;
 	/* Flag to indicate if VM is used for compute */
 	bool			is_compute_context;
+
+	/* Memory partition number, -1 means any partition */
+	int8_t			mem_id;
+
+	/* cached fault info */
+	struct amdgpu_vm_fault_info fault_info;
 };
 
 struct amdgpu_vm_manager {
@@ -334,10 +453,6 @@ struct amdgpu_vm_manager {
 	unsigned int				first_kfd_vmid;
 	bool					concurrent_flush;
 
-	/* Handling of VM fences */
-	u64					fence_context;
-	unsigned				seqno[AMDGPU_MAX_RINGS];
-
 	uint64_t				max_pfn;
 	uint32_t				num_level;
 	uint32_t				block_size;
@@ -365,6 +480,8 @@ struct amdgpu_vm_manager {
 	 * look up VM of a page fault
 	 */
 	struct xarray				pasids;
+	/* Global registration of recent page fault information */
+	struct amdgpu_vm_fault_info	fault_info;
 };
 
 struct amdgpu_bo_va_mapping;
@@ -379,21 +496,20 @@ extern const struct amdgpu_vm_update_funcs amdgpu_vm_sdma_funcs;
 void amdgpu_vm_manager_init(struct amdgpu_device *adev);
 void amdgpu_vm_manager_fini(struct amdgpu_device *adev);
 
-int amdgpu_vm_set_pasid(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			u32 pasid);
-
 long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout);
-int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm);
+int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, int32_t xcp_id, uint32_t pasid);
 int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm);
-void amdgpu_vm_release_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm);
 void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm);
-void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
-			 struct list_head *validated,
-			 struct amdgpu_bo_list_entry *entry);
+int amdgpu_vm_lock_pd(struct amdgpu_vm *vm, struct drm_exec *exec,
+		      unsigned int num_fences);
+int amdgpu_vm_lock_done_list(struct amdgpu_vm *vm, struct drm_exec *exec,
+			     unsigned int num_fences);
 bool amdgpu_vm_ready(struct amdgpu_vm *vm);
-int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			      int (*callback)(void *p, struct amdgpu_bo *bo),
-			      void *param);
+uint64_t amdgpu_vm_generation(struct amdgpu_device *adev, struct amdgpu_vm *vm);
+int amdgpu_vm_validate(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+		       struct ww_acquire_ctx *ticket,
+		       int (*callback)(void *p, struct amdgpu_bo *bo),
+		       void *param);
 int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_pipe_sync);
 int amdgpu_vm_update_pdes(struct amdgpu_device *adev,
 			  struct amdgpu_vm *vm, bool immediate);
@@ -401,21 +517,31 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
 			  struct amdgpu_vm *vm,
 			  struct dma_fence **fence);
 int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
-			   struct amdgpu_vm *vm);
+			   struct amdgpu_vm *vm,
+			   struct ww_acquire_ctx *ticket);
+int amdgpu_vm_flush_compute_tlb(struct amdgpu_device *adev,
+				struct amdgpu_vm *vm,
+				uint32_t flush_type,
+				uint32_t xcc_mask);
 void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base,
 			    struct amdgpu_vm *vm, struct amdgpu_bo *bo);
 int amdgpu_vm_update_range(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			   bool immediate, bool unlocked, bool flush_tlb,
-			   struct dma_resv *resv, uint64_t start, uint64_t last,
-			   uint64_t flags, uint64_t offset, uint64_t vram_base,
+			   bool allow_override, struct amdgpu_sync *sync,
+			   uint64_t start, uint64_t last, uint64_t flags,
+			   uint64_t offset, uint64_t vram_base,
 			   struct ttm_resource *res, dma_addr_t *pages_addr,
 			   struct dma_fence **fence);
 int amdgpu_vm_bo_update(struct amdgpu_device *adev,
 			struct amdgpu_bo_va *bo_va,
 			bool clear);
 bool amdgpu_vm_evictable(struct amdgpu_bo *bo);
-void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev,
-			     struct amdgpu_bo *bo, bool evicted);
+void amdgpu_vm_bo_invalidate(struct amdgpu_bo *bo, bool evicted);
+void amdgpu_vm_update_stats(struct amdgpu_vm_bo_base *base,
+			    struct ttm_resource *new_res, int sign);
+void amdgpu_vm_bo_update_shared(struct amdgpu_bo *bo);
+void amdgpu_vm_bo_move(struct amdgpu_bo *bo, struct ttm_resource *new_mem,
+		       bool evicted);
 uint64_t amdgpu_vm_map_gart(const dma_addr_t *pages_addr, uint64_t addr);
 struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm,
 				       struct amdgpu_bo *bo);
@@ -425,11 +551,11 @@ struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev,
 int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 		     struct amdgpu_bo_va *bo_va,
 		     uint64_t addr, uint64_t offset,
-		     uint64_t size, uint64_t flags);
+		     uint64_t size, uint32_t flags);
 int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev,
 			     struct amdgpu_bo_va *bo_va,
 			     uint64_t addr, uint64_t offset,
-			     uint64_t size, uint64_t flags);
+			     uint64_t size, uint32_t flags);
 int amdgpu_vm_bo_unmap(struct amdgpu_device *adev,
 		       struct amdgpu_bo_va *bo_va,
 		       uint64_t addr);
@@ -449,25 +575,31 @@ bool amdgpu_vm_need_pipeline_sync(struct amdgpu_ring *ring,
 				  struct amdgpu_job *job);
 void amdgpu_vm_check_compute_bug(struct amdgpu_device *adev);
 
-void amdgpu_vm_get_task_info(struct amdgpu_device *adev, u32 pasid,
-			     struct amdgpu_task_info *task_info);
+struct amdgpu_task_info *
+amdgpu_vm_get_task_info_pasid(struct amdgpu_device *adev, u32 pasid);
+
+struct amdgpu_task_info *
+amdgpu_vm_get_task_info_vm(struct amdgpu_vm *vm);
+
+void amdgpu_vm_put_task_info(struct amdgpu_task_info *task_info);
+
 bool amdgpu_vm_handle_fault(struct amdgpu_device *adev, u32 pasid,
-			    uint64_t addr, bool write_fault);
+			    u32 vmid, u32 node_id, uint64_t addr, uint64_t ts,
+			    bool write_fault);
 
 void amdgpu_vm_set_task_info(struct amdgpu_vm *vm);
 
 void amdgpu_vm_move_to_lru_tail(struct amdgpu_device *adev,
 				struct amdgpu_vm *vm);
 void amdgpu_vm_get_memory(struct amdgpu_vm *vm,
-			  struct amdgpu_mem_stats *stats);
+			  struct amdgpu_mem_stats stats[__AMDGPU_PL_NUM]);
 
 int amdgpu_vm_pt_clear(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		       struct amdgpu_bo_vm *vmbo, bool immediate);
 int amdgpu_vm_pt_create(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			int level, bool immediate, struct amdgpu_bo_vm **vmbo);
+			int level, bool immediate, struct amdgpu_bo_vm **vmbo,
+			int32_t xcp_id);
 void amdgpu_vm_pt_free_root(struct amdgpu_device *adev, struct amdgpu_vm *vm);
-bool amdgpu_vm_pt_is_root_clean(struct amdgpu_device *adev,
-				struct amdgpu_vm *vm);
 
 int amdgpu_vm_pde_update(struct amdgpu_vm_update_params *params,
 			 struct amdgpu_vm_bo_base *entry);
@@ -475,11 +607,17 @@ int amdgpu_vm_ptes_update(struct amdgpu_vm_update_params *params,
 			  uint64_t start, uint64_t end,
 			  uint64_t dst, uint64_t flags);
 void amdgpu_vm_pt_free_work(struct work_struct *work);
+void amdgpu_vm_pt_free_list(struct amdgpu_device *adev,
+			    struct amdgpu_vm_update_params *params);
 
 #if defined(CONFIG_DEBUG_FS)
 void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m);
 #endif
 
+int amdgpu_vm_pt_map_tables(struct amdgpu_device *adev, struct amdgpu_vm *vm);
+
+bool amdgpu_vm_is_bo_always_valid(struct amdgpu_vm *vm, struct amdgpu_bo *bo);
+
 /**
  * amdgpu_vm_tlb_seq - return tlb flush sequence number
  * @vm: the amdgpu_vm structure to query
@@ -533,4 +671,21 @@ static inline void amdgpu_vm_eviction_unlock(struct amdgpu_vm *vm)
 	mutex_unlock(&vm->eviction_lock);
 }
 
+void amdgpu_vm_update_fault_cache(struct amdgpu_device *adev,
+				  unsigned int pasid,
+				  uint64_t addr,
+				  uint32_t status,
+				  unsigned int vmhub);
+void amdgpu_vm_tlb_fence_create(struct amdgpu_device *adev,
+				 struct amdgpu_vm *vm,
+				 struct dma_fence **fence);
+
+void amdgpu_vm_print_task_info(struct amdgpu_device *adev,
+			       struct amdgpu_task_info *task_info);
+
+#define amdgpu_vm_bo_va_for_each_valid_mapping(bo_va, mapping) \
+		list_for_each_entry(mapping, &(bo_va)->valids, list)
+#define amdgpu_vm_bo_va_for_each_invalid_mapping(bo_va, mapping) \
+		list_for_each_entry(mapping, &(bo_va)->invalids, list)
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_cpu.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_cpu.c
index 31913ae86de6..22e2e5b47341 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_cpu.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_cpu.c
@@ -31,6 +31,7 @@
  */
 static int amdgpu_vm_cpu_map_table(struct amdgpu_bo_vm *table)
 {
+	table->bo.flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED;
 	return amdgpu_bo_kmap(&table->bo, NULL);
 }
 
@@ -38,20 +39,20 @@ static int amdgpu_vm_cpu_map_table(struct amdgpu_bo_vm *table)
  * amdgpu_vm_cpu_prepare - prepare page table update with the CPU
  *
  * @p: see amdgpu_vm_update_params definition
- * @resv: reservation object with embedded fence
- * @sync_mode: synchronization mode
+ * @sync: sync obj with fences to wait on
+ * @k_job_id: the id for tracing/debug purposes
  *
  * Returns:
  * Negativ errno, 0 for success.
  */
 static int amdgpu_vm_cpu_prepare(struct amdgpu_vm_update_params *p,
-				 struct dma_resv *resv,
-				 enum amdgpu_sync_mode sync_mode)
+				 struct amdgpu_sync *sync,
+				 u64 k_job_id)
 {
-	if (!resv)
+	if (!sync)
 		return 0;
 
-	return amdgpu_bo_sync_wait_resv(p->adev, resv, sync_mode, p->vm, true);
+	return amdgpu_sync_wait(sync, true);
 }
 
 /**
@@ -107,7 +108,9 @@ static int amdgpu_vm_cpu_update(struct amdgpu_vm_update_params *p,
 static int amdgpu_vm_cpu_commit(struct amdgpu_vm_update_params *p,
 				struct dma_fence **fence)
 {
-	/* Flush HDP */
+	if (p->needs_flush)
+		atomic64_inc(&p->vm->tlb_seq);
+
 	mb();
 	amdgpu_device_flush_hdp(p->adev, NULL);
 	return 0;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c
index df63dc3bca18..f794fb1cc06e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_pt.c
@@ -26,6 +26,7 @@
 #include "amdgpu.h"
 #include "amdgpu_trace.h"
 #include "amdgpu_vm.h"
+#include "amdgpu_job.h"
 
 /*
  * amdgpu_vm_pt_cursor - state for for_each_amdgpu_vm_pt
@@ -90,22 +91,6 @@ static unsigned int amdgpu_vm_pt_num_entries(struct amdgpu_device *adev,
 }
 
 /**
- * amdgpu_vm_pt_num_ats_entries - return the number of ATS entries in the root PD
- *
- * @adev: amdgpu_device pointer
- *
- * Returns:
- * The number of entries in the root page directory which needs the ATS setting.
- */
-static unsigned int amdgpu_vm_pt_num_ats_entries(struct amdgpu_device *adev)
-{
-	unsigned int shift;
-
-	shift = amdgpu_vm_pt_level_shift(adev, adev->vm_manager.root_level);
-	return AMDGPU_GMC_HOLE_START >> (shift + AMDGPU_GPU_PAGE_SHIFT);
-}
-
-/**
  * amdgpu_vm_pt_entries_mask - the mask to get the entry number of a PD/PT
  *
  * @adev: amdgpu_device pointer
@@ -379,7 +364,7 @@ int amdgpu_vm_pt_clear(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	struct ttm_operation_ctx ctx = { true, false };
 	struct amdgpu_vm_update_params params;
 	struct amdgpu_bo *ancestor = &vmbo->bo;
-	unsigned int entries, ats_entries;
+	unsigned int entries;
 	struct amdgpu_bo *bo = &vmbo->bo;
 	uint64_t addr;
 	int r, idx;
@@ -394,40 +379,11 @@ int amdgpu_vm_pt_clear(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	}
 
 	entries = amdgpu_bo_size(bo) / 8;
-	if (!vm->pte_support_ats) {
-		ats_entries = 0;
-
-	} else if (!bo->parent) {
-		ats_entries = amdgpu_vm_pt_num_ats_entries(adev);
-		ats_entries = min(ats_entries, entries);
-		entries -= ats_entries;
-
-	} else {
-		struct amdgpu_vm_bo_base *pt;
-
-		pt = ancestor->vm_bo;
-		ats_entries = amdgpu_vm_pt_num_ats_entries(adev);
-		if ((pt - to_amdgpu_bo_vm(vm->root.bo)->entries) >=
-		    ats_entries) {
-			ats_entries = 0;
-		} else {
-			ats_entries = entries;
-			entries = 0;
-		}
-	}
 
 	r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 	if (r)
 		return r;
 
-	if (vmbo->shadow) {
-		struct amdgpu_bo *shadow = vmbo->shadow;
-
-		r = ttm_bo_validate(&shadow->tbo, &shadow->placement, &ctx);
-		if (r)
-			return r;
-	}
-
 	if (!drm_dev_enter(adev_to_drm(adev), &idx))
 		return -ENODEV;
 
@@ -440,49 +396,30 @@ int amdgpu_vm_pt_clear(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	params.vm = vm;
 	params.immediate = immediate;
 
-	r = vm->update_funcs->prepare(&params, NULL, AMDGPU_SYNC_EXPLICIT);
+	r = vm->update_funcs->prepare(&params, NULL,
+				      AMDGPU_KERNEL_JOB_ID_VM_PT_CLEAR);
 	if (r)
 		goto exit;
 
 	addr = 0;
-	if (ats_entries) {
-		uint64_t value = 0, flags;
 
-		flags = AMDGPU_PTE_DEFAULT_ATC;
+	uint64_t value = 0, flags = 0;
+	if (adev->asic_type >= CHIP_VEGA10) {
 		if (level != AMDGPU_VM_PTB) {
 			/* Handle leaf PDEs as PTEs */
-			flags |= AMDGPU_PDE_PTE;
-			amdgpu_gmc_get_vm_pde(adev, level, &value, &flags);
+			flags |= AMDGPU_PDE_PTE_FLAG(adev);
+			amdgpu_gmc_get_vm_pde(adev, level,
+					      &value, &flags);
+		} else {
+			/* Workaround for fault priority problem on GMC9 */
+			flags = AMDGPU_PTE_EXECUTABLE;
 		}
-
-		r = vm->update_funcs->update(&params, vmbo, addr, 0,
-					     ats_entries, value, flags);
-		if (r)
-			goto exit;
-
-		addr += ats_entries * 8;
 	}
 
-	if (entries) {
-		uint64_t value = 0, flags = 0;
-
-		if (adev->asic_type >= CHIP_VEGA10) {
-			if (level != AMDGPU_VM_PTB) {
-				/* Handle leaf PDEs as PTEs */
-				flags |= AMDGPU_PDE_PTE;
-				amdgpu_gmc_get_vm_pde(adev, level,
-						      &value, &flags);
-			} else {
-				/* Workaround for fault priority problem on GMC9 */
-				flags = AMDGPU_PTE_EXECUTABLE;
-			}
-		}
-
-		r = vm->update_funcs->update(&params, vmbo, addr, 0, entries,
-					     value, flags);
-		if (r)
-			goto exit;
-	}
+	r = vm->update_funcs->update(&params, vmbo, addr, 0, entries,
+				     value, flags);
+	if (r)
+		goto exit;
 
 	r = vm->update_funcs->commit(&params, NULL);
 exit:
@@ -498,21 +435,25 @@ exit:
  * @level: the page table level
  * @immediate: use a immediate update
  * @vmbo: pointer to the buffer object pointer
+ * @xcp_id: GPU partition id
  */
 int amdgpu_vm_pt_create(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-			int level, bool immediate, struct amdgpu_bo_vm **vmbo)
+			int level, bool immediate, struct amdgpu_bo_vm **vmbo,
+			int32_t xcp_id)
 {
 	struct amdgpu_bo_param bp;
-	struct amdgpu_bo *bo;
-	struct dma_resv *resv;
 	unsigned int num_entries;
-	int r;
 
 	memset(&bp, 0, sizeof(bp));
 
 	bp.size = amdgpu_vm_pt_size(adev, level);
 	bp.byte_align = AMDGPU_GPU_PAGE_SIZE;
-	bp.domain = AMDGPU_GEM_DOMAIN_VRAM;
+
+	if (!adev->gmc.is_app_apu)
+		bp.domain = AMDGPU_GEM_DOMAIN_VRAM;
+	else
+		bp.domain = AMDGPU_GEM_DOMAIN_GTT;
+
 	bp.domain = amdgpu_bo_get_preferred_domain(adev, bp.domain);
 	bp.flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS |
 		AMDGPU_GEM_CREATE_CPU_GTT_USWC;
@@ -529,45 +470,12 @@ int amdgpu_vm_pt_create(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 
 	bp.type = ttm_bo_type_kernel;
 	bp.no_wait_gpu = immediate;
+	bp.xcp_id_plus1 = xcp_id + 1;
+
 	if (vm->root.bo)
 		bp.resv = vm->root.bo->tbo.base.resv;
 
-	r = amdgpu_bo_create_vm(adev, &bp, vmbo);
-	if (r)
-		return r;
-
-	bo = &(*vmbo)->bo;
-	if (vm->is_compute_context || (adev->flags & AMD_IS_APU)) {
-		(*vmbo)->shadow = NULL;
-		return 0;
-	}
-
-	if (!bp.resv)
-		WARN_ON(dma_resv_lock(bo->tbo.base.resv,
-				      NULL));
-	resv = bp.resv;
-	memset(&bp, 0, sizeof(bp));
-	bp.size = amdgpu_vm_pt_size(adev, level);
-	bp.domain = AMDGPU_GEM_DOMAIN_GTT;
-	bp.flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC;
-	bp.type = ttm_bo_type_kernel;
-	bp.resv = bo->tbo.base.resv;
-	bp.bo_ptr_size = sizeof(struct amdgpu_bo);
-
-	r = amdgpu_bo_create(adev, &bp, &(*vmbo)->shadow);
-
-	if (!resv)
-		dma_resv_unlock(bo->tbo.base.resv);
-
-	if (r) {
-		amdgpu_bo_unref(&bo);
-		return r;
-	}
-
-	(*vmbo)->shadow->parent = amdgpu_bo_ref(bo);
-	amdgpu_bo_add_to_shadow_list(*vmbo);
-
-	return 0;
+	return amdgpu_bo_create_vm(adev, &bp, vmbo);
 }
 
 /**
@@ -598,7 +506,8 @@ static int amdgpu_vm_pt_alloc(struct amdgpu_device *adev,
 		return 0;
 
 	amdgpu_vm_eviction_unlock(vm);
-	r = amdgpu_vm_pt_create(adev, vm, cursor->level, immediate, &pt);
+	r = amdgpu_vm_pt_create(adev, vm, cursor->level, immediate, &pt,
+				vm->root.bo->xcp_id);
 	amdgpu_vm_eviction_lock(vm);
 	if (r)
 		return r;
@@ -616,7 +525,6 @@ static int amdgpu_vm_pt_alloc(struct amdgpu_device *adev,
 	return 0;
 
 error_free_pt:
-	amdgpu_bo_unref(&pt->shadow);
 	amdgpu_bo_unref(&pt_bo);
 	return r;
 }
@@ -628,17 +536,12 @@ error_free_pt:
  */
 static void amdgpu_vm_pt_free(struct amdgpu_vm_bo_base *entry)
 {
-	struct amdgpu_bo *shadow;
-
 	if (!entry->bo)
 		return;
-	shadow = amdgpu_bo_shadowed(entry->bo);
-	if (shadow) {
-		ttm_bo_set_bulk_move(&shadow->tbo, NULL);
-		amdgpu_bo_unref(&shadow);
-	}
-	ttm_bo_set_bulk_move(&entry->bo->tbo, NULL);
+
+	amdgpu_vm_update_stats(entry, entry->bo->tbo.resource, -1);
 	entry->bo->vm_bo = NULL;
+	ttm_bo_set_bulk_move(&entry->bo->tbo, NULL);
 
 	spin_lock(&entry->vm->status_lock);
 	list_del(&entry->vm_status);
@@ -646,62 +549,55 @@ static void amdgpu_vm_pt_free(struct amdgpu_vm_bo_base *entry)
 	amdgpu_bo_unref(&entry->bo);
 }
 
-void amdgpu_vm_pt_free_work(struct work_struct *work)
+/**
+ * amdgpu_vm_pt_free_list - free PD/PT levels
+ *
+ * @adev: amdgpu device structure
+ * @params: see amdgpu_vm_update_params definition
+ *
+ * Free the page directory objects saved in the flush list
+ */
+void amdgpu_vm_pt_free_list(struct amdgpu_device *adev,
+			    struct amdgpu_vm_update_params *params)
 {
 	struct amdgpu_vm_bo_base *entry, *next;
-	struct amdgpu_vm *vm;
-	LIST_HEAD(pt_freed);
-
-	vm = container_of(work, struct amdgpu_vm, pt_free_work);
+	bool unlocked = params->unlocked;
 
-	spin_lock(&vm->status_lock);
-	list_splice_init(&vm->pt_freed, &pt_freed);
-	spin_unlock(&vm->status_lock);
+	if (list_empty(&params->tlb_flush_waitlist))
+		return;
 
-	/* flush_work in amdgpu_vm_fini ensure vm->root.bo is valid. */
-	amdgpu_bo_reserve(vm->root.bo, true);
+	/*
+	 * unlocked unmap clear page table leaves, warning to free the page entry.
+	 */
+	WARN_ON(unlocked);
 
-	list_for_each_entry_safe(entry, next, &pt_freed, vm_status)
+	list_for_each_entry_safe(entry, next, &params->tlb_flush_waitlist, vm_status)
 		amdgpu_vm_pt_free(entry);
-
-	amdgpu_bo_unreserve(vm->root.bo);
 }
 
 /**
- * amdgpu_vm_pt_free_dfs - free PD/PT levels
+ * amdgpu_vm_pt_add_list - add PD/PT level to the flush list
  *
- * @adev: amdgpu device structure
- * @vm: amdgpu vm structure
- * @start: optional cursor where to start freeing PDs/PTs
- * @unlocked: vm resv unlock status
+ * @params: parameters for the update
+ * @cursor: first PT entry to start DF search from, non NULL
  *
- * Free the page directory or page table level and all sub levels.
+ * This list will be freed after TLB flush.
  */
-static void amdgpu_vm_pt_free_dfs(struct amdgpu_device *adev,
-				  struct amdgpu_vm *vm,
-				  struct amdgpu_vm_pt_cursor *start,
-				  bool unlocked)
+static void amdgpu_vm_pt_add_list(struct amdgpu_vm_update_params *params,
+				  struct amdgpu_vm_pt_cursor *cursor)
 {
-	struct amdgpu_vm_pt_cursor cursor;
+	struct amdgpu_vm_pt_cursor seek;
 	struct amdgpu_vm_bo_base *entry;
 
-	if (unlocked) {
-		spin_lock(&vm->status_lock);
-		for_each_amdgpu_vm_pt_dfs_safe(adev, vm, start, cursor, entry)
-			list_move(&entry->vm_status, &vm->pt_freed);
-
-		if (start)
-			list_move(&start->entry->vm_status, &vm->pt_freed);
-		spin_unlock(&vm->status_lock);
-		schedule_work(&vm->pt_free_work);
-		return;
+	spin_lock(&params->vm->status_lock);
+	for_each_amdgpu_vm_pt_dfs_safe(params->adev, params->vm, cursor, seek, entry) {
+		if (entry && entry->bo)
+			list_move(&entry->vm_status, &params->tlb_flush_waitlist);
 	}
 
-	for_each_amdgpu_vm_pt_dfs_safe(adev, vm, start, cursor, entry)
-		amdgpu_vm_pt_free(entry);
-
-	if (start)
-		amdgpu_vm_pt_free(start->entry);
+	/* enter start node now */
+	list_move(&cursor->entry->vm_status, &params->tlb_flush_waitlist);
+	spin_unlock(&params->vm->status_lock);
 }
 
 /**
@@ -713,34 +609,13 @@ static void amdgpu_vm_pt_free_dfs(struct amdgpu_device *adev,
  */
 void amdgpu_vm_pt_free_root(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 {
-	amdgpu_vm_pt_free_dfs(adev, vm, NULL, false);
-}
-
-/**
- * amdgpu_vm_pt_is_root_clean - check if a root PD is clean
- *
- * @adev: amdgpu_device pointer
- * @vm: the VM to check
- *
- * Check all entries of the root PD, if any subsequent PDs are allocated,
- * it means there are page table creating and filling, and is no a clean
- * VM
- *
- * Returns:
- *	0 if this VM is clean
- */
-bool amdgpu_vm_pt_is_root_clean(struct amdgpu_device *adev,
-				struct amdgpu_vm *vm)
-{
-	enum amdgpu_vm_level root = adev->vm_manager.root_level;
-	unsigned int entries = amdgpu_vm_pt_num_entries(adev, root);
-	unsigned int i = 0;
+	struct amdgpu_vm_pt_cursor cursor;
+	struct amdgpu_vm_bo_base *entry;
 
-	for (i = 0; i < entries; i++) {
-		if (to_amdgpu_bo_vm(vm->root.bo)->entries[i].bo)
-			return false;
+	for_each_amdgpu_vm_pt_dfs_safe(adev, vm, NULL, cursor, entry) {
+		if (entry)
+			amdgpu_vm_pt_free(entry);
 	}
-	return true;
 }
 
 /**
@@ -755,11 +630,15 @@ int amdgpu_vm_pde_update(struct amdgpu_vm_update_params *params,
 			 struct amdgpu_vm_bo_base *entry)
 {
 	struct amdgpu_vm_bo_base *parent = amdgpu_vm_pt_parent(entry);
-	struct amdgpu_bo *bo = parent->bo, *pbo;
+	struct amdgpu_bo *bo, *pbo;
 	struct amdgpu_vm *vm = params->vm;
 	uint64_t pde, pt, flags;
 	unsigned int level;
 
+	if (WARN_ON(!parent))
+		return -EINVAL;
+
+	bo = parent->bo;
 	for (level = 0, pbo = bo->parent; pbo; ++level)
 		pbo = pbo->parent;
 
@@ -770,6 +649,27 @@ int amdgpu_vm_pde_update(struct amdgpu_vm_update_params *params,
 					1, 0, flags);
 }
 
+/**
+ * amdgpu_vm_pte_update_noretry_flags - Update PTE no-retry flags
+ *
+ * @adev: amdgpu_device pointer
+ * @flags: pointer to PTE flags
+ *
+ * Update PTE no-retry flags when TF is enabled.
+ */
+static void amdgpu_vm_pte_update_noretry_flags(struct amdgpu_device *adev,
+						uint64_t *flags)
+{
+	/*
+	 * Update no-retry flags with the corresponding TF
+	 * no-retry combination.
+	 */
+	if ((*flags & AMDGPU_VM_NORETRY_FLAGS) == AMDGPU_VM_NORETRY_FLAGS) {
+		*flags &= ~AMDGPU_VM_NORETRY_FLAGS;
+		*flags |= adev->gmc.noretry_flags;
+	}
+}
+
 /*
  * amdgpu_vm_pte_update_flags - figure out flags for PTE updates
  *
@@ -781,20 +681,40 @@ static void amdgpu_vm_pte_update_flags(struct amdgpu_vm_update_params *params,
 				       uint64_t pe, uint64_t addr,
 				       unsigned int count, uint32_t incr,
 				       uint64_t flags)
-
 {
+	struct amdgpu_device *adev = params->adev;
+
 	if (level != AMDGPU_VM_PTB) {
-		flags |= AMDGPU_PDE_PTE;
-		amdgpu_gmc_get_vm_pde(params->adev, level, &addr, &flags);
+		flags |= AMDGPU_PDE_PTE_FLAG(params->adev);
+		amdgpu_gmc_get_vm_pde(adev, level, &addr, &flags);
 
-	} else if (params->adev->asic_type >= CHIP_VEGA10 &&
+	} else if (adev->asic_type >= CHIP_VEGA10 &&
 		   !(flags & AMDGPU_PTE_VALID) &&
-		   !(flags & AMDGPU_PTE_PRT)) {
+		   !(flags & AMDGPU_PTE_PRT_FLAG(params->adev))) {
 
 		/* Workaround for fault priority problem on GMC9 */
 		flags |= AMDGPU_PTE_EXECUTABLE;
 	}
 
+	/*
+	 * Update no-retry flags to use the no-retry flag combination
+	 * with TF enabled. The AMDGPU_VM_NORETRY_FLAGS flag combination
+	 * does not work when TF is enabled. So, replace them with
+	 * AMDGPU_VM_NORETRY_FLAGS_TF flag combination which works for
+	 * all cases.
+	 */
+	if (level == AMDGPU_VM_PTB)
+		amdgpu_vm_pte_update_noretry_flags(adev, &flags);
+
+	/* APUs mapping system memory may need different MTYPEs on different
+	 * NUMA nodes. Only do this for contiguous ranges that can be assumed
+	 * to be on the same NUMA node.
+	 */
+	if ((flags & AMDGPU_PTE_SYSTEM) && (adev->flags & AMD_IS_APU) &&
+	    adev->gmc.gmc_funcs->override_vm_pte_flags &&
+	    num_possible_nodes() > 1 && !params->pages_addr && params->allow_override)
+		amdgpu_gmc_override_vm_pte_flags(adev, params->vm, addr, &flags);
+
 	params->vm->update_funcs->update(params, pt, pe, addr, count, incr,
 					 flags);
 }
@@ -957,7 +877,15 @@ int amdgpu_vm_ptes_update(struct amdgpu_vm_update_params *params,
 		incr = (uint64_t)AMDGPU_GPU_PAGE_SIZE << shift;
 		mask = amdgpu_vm_pt_entries_mask(adev, cursor.level);
 		pe_start = ((cursor.pfn >> shift) & mask) * 8;
-		entry_end = ((uint64_t)mask + 1) << shift;
+
+		if (cursor.level < AMDGPU_VM_PTB && params->unlocked)
+			/*
+			 * MMU notifier callback unlocked unmap huge page, leave is PDE entry,
+			 * only clear one entry. Next entry search again for PDE or PTE leave.
+			 */
+			entry_end = 1ULL << shift;
+		else
+			entry_end = ((uint64_t)mask + 1) << shift;
 		entry_end += cursor.pfn & ~(entry_end - 1);
 		entry_end = min(entry_end, end);
 
@@ -975,7 +903,7 @@ int amdgpu_vm_ptes_update(struct amdgpu_vm_update_params *params,
 			trace_amdgpu_vm_update_ptes(params, frag_start, upd_end,
 						    min(nptes, 32u), dst, incr,
 						    upd_flags,
-						    vm->task_info.tgid,
+						    vm->task_info ? vm->task_info->tgid : 0,
 						    vm->immediate.fence_context);
 			amdgpu_vm_pte_update_flags(params, to_amdgpu_bo_vm(pt),
 						   cursor.level, pe_start, dst,
@@ -1004,10 +932,8 @@ int amdgpu_vm_ptes_update(struct amdgpu_vm_update_params *params,
 			while (cursor.pfn < frag_start) {
 				/* Make sure previous mapping is freed */
 				if (cursor.entry->bo) {
-					params->table_freed = true;
-					amdgpu_vm_pt_free_dfs(adev, params->vm,
-							      &cursor,
-							      params->unlocked);
+					params->needs_flush = true;
+					amdgpu_vm_pt_add_list(params, &cursor);
 				}
 				amdgpu_vm_pt_next(adev, &cursor);
 			}
@@ -1020,3 +946,31 @@ int amdgpu_vm_ptes_update(struct amdgpu_vm_update_params *params,
 
 	return 0;
 }
+
+/**
+ * amdgpu_vm_pt_map_tables - have bo of root PD cpu accessible
+ * @adev: amdgpu device structure
+ * @vm: amdgpu vm structure
+ *
+ * make root page directory and everything below it cpu accessible.
+ */
+int amdgpu_vm_pt_map_tables(struct amdgpu_device *adev, struct amdgpu_vm *vm)
+{
+	struct amdgpu_vm_pt_cursor cursor;
+	struct amdgpu_vm_bo_base *entry;
+
+	for_each_amdgpu_vm_pt_dfs_safe(adev, vm, NULL, cursor, entry) {
+
+		struct amdgpu_bo_vm *bo;
+		int r;
+
+		if (entry->bo) {
+			bo = to_amdgpu_bo_vm(entry->bo);
+			r = vm->update_funcs->map_table(bo);
+			if (r)
+				return r;
+		}
+	}
+
+	return 0;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
index 349416e176a1..36805dcfa159 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
@@ -35,21 +35,12 @@
  */
 static int amdgpu_vm_sdma_map_table(struct amdgpu_bo_vm *table)
 {
-	int r;
-
-	r = amdgpu_ttm_alloc_gart(&table->bo.tbo);
-	if (r)
-		return r;
-
-	if (table->shadow)
-		r = amdgpu_ttm_alloc_gart(&table->shadow->tbo);
-
-	return r;
+	return amdgpu_ttm_alloc_gart(&table->bo.tbo);
 }
 
 /* Allocate a new job for @count PTE updates */
 static int amdgpu_vm_sdma_alloc_job(struct amdgpu_vm_update_params *p,
-				    unsigned int count)
+				    unsigned int count, u64 k_job_id)
 {
 	enum amdgpu_ib_pool_type pool = p->immediate ? AMDGPU_IB_POOL_IMMEDIATE
 		: AMDGPU_IB_POOL_DELAYED;
@@ -65,7 +56,7 @@ static int amdgpu_vm_sdma_alloc_job(struct amdgpu_vm_update_params *p,
 	ndw = min(ndw, AMDGPU_VM_SDMA_MAX_NUM_DW);
 
 	r = amdgpu_job_alloc_with_ib(p->adev, entity, AMDGPU_FENCE_OWNER_VM,
-				     ndw * 4, pool, &p->job);
+				     ndw * 4, pool, &p->job, k_job_id);
 	if (r)
 		return r;
 
@@ -77,31 +68,29 @@ static int amdgpu_vm_sdma_alloc_job(struct amdgpu_vm_update_params *p,
  * amdgpu_vm_sdma_prepare - prepare SDMA command submission
  *
  * @p: see amdgpu_vm_update_params definition
- * @resv: reservation object with embedded fence
- * @sync_mode: synchronization mode
+ * @sync: amdgpu_sync object with fences to wait for
+ * @k_job_id: identifier of the job, for tracing purpose
  *
  * Returns:
  * Negativ errno, 0 for success.
  */
 static int amdgpu_vm_sdma_prepare(struct amdgpu_vm_update_params *p,
-				  struct dma_resv *resv,
-				  enum amdgpu_sync_mode sync_mode)
+				  struct amdgpu_sync *sync, u64 k_job_id)
 {
-	struct amdgpu_sync sync;
 	int r;
 
-	r = amdgpu_vm_sdma_alloc_job(p, 0);
+	r = amdgpu_vm_sdma_alloc_job(p, 0, k_job_id);
 	if (r)
 		return r;
 
-	if (!resv)
+	if (!sync)
 		return 0;
 
-	amdgpu_sync_create(&sync);
-	r = amdgpu_sync_resv(p->adev, &sync, resv, sync_mode, p->vm);
-	if (!r)
-		r = amdgpu_sync_push_to_job(&sync, p->job);
-	amdgpu_sync_free(&sync);
+	r = amdgpu_sync_push_to_job(sync, p->job);
+	if (r) {
+		p->num_dw_left = 0;
+		amdgpu_job_free(p->job);
+	}
 	return r;
 }
 
@@ -126,6 +115,10 @@ static int amdgpu_vm_sdma_commit(struct amdgpu_vm_update_params *p,
 
 	WARN_ON(ib->length_dw == 0);
 	amdgpu_ring_pad_ib(ring, ib);
+
+	if (p->needs_flush)
+		atomic64_inc(&p->vm->tlb_seq);
+
 	WARN_ON(ib->length_dw > p->num_dw_left);
 	f = amdgpu_job_submit(p->job);
 
@@ -257,24 +250,21 @@ static int amdgpu_vm_sdma_update(struct amdgpu_vm_update_params *p,
 			if (r)
 				return r;
 
-			r = amdgpu_vm_sdma_alloc_job(p, count);
+			r = amdgpu_vm_sdma_alloc_job(p, count,
+						     AMDGPU_KERNEL_JOB_ID_VM_UPDATE);
 			if (r)
 				return r;
 		}
 
 		if (!p->pages_addr) {
 			/* set page commands needed */
-			if (vmbo->shadow)
-				amdgpu_vm_sdma_set_ptes(p, vmbo->shadow, pe, addr,
-							count, incr, flags);
 			amdgpu_vm_sdma_set_ptes(p, bo, pe, addr, count,
 						incr, flags);
 			return 0;
 		}
 
 		/* copy commands needed */
-		ndw -= p->adev->vm_manager.vm_pte_funcs->copy_pte_num_dw *
-			(vmbo->shadow ? 2 : 1);
+		ndw -= p->adev->vm_manager.vm_pte_funcs->copy_pte_num_dw;
 
 		/* for padding */
 		ndw -= 7;
@@ -289,8 +279,6 @@ static int amdgpu_vm_sdma_update(struct amdgpu_vm_update_params *p,
 			pte[i] |= flags;
 		}
 
-		if (vmbo->shadow)
-			amdgpu_vm_sdma_copy_ptes(p, vmbo->shadow, pe, nptes);
 		amdgpu_vm_sdma_copy_ptes(p, bo, pe, nptes);
 
 		pe += nptes * 8;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_tlb_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_tlb_fence.c
new file mode 100644
index 000000000000..5d26797356a3
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_tlb_fence.c
@@ -0,0 +1,111 @@
+// SPDX-License-Identifier: GPL-2.0 OR MIT
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include <linux/dma-fence.h>
+#include <linux/workqueue.h>
+
+#include "amdgpu.h"
+#include "amdgpu_vm.h"
+#include "amdgpu_gmc.h"
+
+struct amdgpu_tlb_fence {
+	struct dma_fence	base;
+	struct amdgpu_device	*adev;
+	struct dma_fence	*dependency;
+	struct work_struct	work;
+	spinlock_t		lock;
+	uint16_t		pasid;
+
+};
+
+static const char *amdgpu_tlb_fence_get_driver_name(struct dma_fence *fence)
+{
+	return "amdgpu tlb fence";
+}
+
+static const char *amdgpu_tlb_fence_get_timeline_name(struct dma_fence *f)
+{
+	return "amdgpu tlb timeline";
+}
+
+static void amdgpu_tlb_fence_work(struct work_struct *work)
+{
+	struct amdgpu_tlb_fence *f = container_of(work, typeof(*f), work);
+	int r;
+
+	if (f->dependency) {
+		dma_fence_wait(f->dependency, false);
+		dma_fence_put(f->dependency);
+		f->dependency = NULL;
+	}
+
+	r = amdgpu_gmc_flush_gpu_tlb_pasid(f->adev, f->pasid, 2, true, 0);
+	if (r) {
+		dev_err(f->adev->dev, "TLB flush failed for PASID %d.\n",
+			f->pasid);
+		dma_fence_set_error(&f->base, r);
+	}
+
+	dma_fence_signal(&f->base);
+	dma_fence_put(&f->base);
+}
+
+static const struct dma_fence_ops amdgpu_tlb_fence_ops = {
+	.get_driver_name = amdgpu_tlb_fence_get_driver_name,
+	.get_timeline_name = amdgpu_tlb_fence_get_timeline_name
+};
+
+void amdgpu_vm_tlb_fence_create(struct amdgpu_device *adev, struct amdgpu_vm *vm,
+				struct dma_fence **fence)
+{
+	struct amdgpu_tlb_fence *f;
+
+	f = kmalloc(sizeof(*f), GFP_KERNEL);
+	if (!f) {
+		/*
+		 * We can't fail since the PDEs and PTEs are already updated, so
+		 * just block for the dependency and execute the TLB flush
+		 */
+		if (*fence)
+			dma_fence_wait(*fence, false);
+
+		amdgpu_gmc_flush_gpu_tlb_pasid(adev, vm->pasid, 2, true, 0);
+		*fence = dma_fence_get_stub();
+		return;
+	}
+
+	f->adev = adev;
+	f->dependency = *fence;
+	f->pasid = vm->pasid;
+	INIT_WORK(&f->work, amdgpu_tlb_fence_work);
+	spin_lock_init(&f->lock);
+
+	dma_fence_init64(&f->base, &amdgpu_tlb_fence_ops, &f->lock,
+			 vm->tlb_fence_context, atomic64_read(&vm->tlb_seq));
+
+	/* TODO: We probably need a separate wq here */
+	dma_fence_get(&f->base);
+	schedule_work(&f->work);
+
+	*fence = &f->base;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c
new file mode 100644
index 000000000000..aa78c2ee9e21
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.c
@@ -0,0 +1,1018 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include <linux/firmware.h>
+#include <drm/drm_drv.h>
+
+#include "amdgpu.h"
+#include "amdgpu_ucode.h"
+#include "amdgpu_vpe.h"
+#include "amdgpu_smu.h"
+#include "soc15_common.h"
+#include "vpe_v6_1.h"
+
+#define AMDGPU_CSA_VPE_SIZE 	64
+/* VPE CSA resides in the 4th page of CSA */
+#define AMDGPU_CSA_VPE_OFFSET 	(4096 * 3)
+
+/* 1 second timeout */
+#define VPE_IDLE_TIMEOUT	msecs_to_jiffies(1000)
+
+#define VPE_MAX_DPM_LEVEL			4
+#define FIXED1_8_BITS_PER_FRACTIONAL_PART	8
+#define GET_PRATIO_INTEGER_PART(x)		((x) >> FIXED1_8_BITS_PER_FRACTIONAL_PART)
+
+static void vpe_set_ring_funcs(struct amdgpu_device *adev);
+
+static inline uint16_t div16_u16_rem(uint16_t dividend, uint16_t divisor, uint16_t *remainder)
+{
+	*remainder = dividend % divisor;
+	return dividend / divisor;
+}
+
+static inline uint16_t complete_integer_division_u16(
+	uint16_t dividend,
+	uint16_t divisor,
+	uint16_t *remainder)
+{
+	return div16_u16_rem(dividend, divisor, (uint16_t *)remainder);
+}
+
+static uint16_t vpe_u1_8_from_fraction(uint16_t numerator, uint16_t denominator)
+{
+	u16 arg1_value = numerator;
+	u16 arg2_value = denominator;
+
+	uint16_t remainder;
+
+	/* determine integer part */
+	uint16_t res_value = complete_integer_division_u16(
+		arg1_value, arg2_value, &remainder);
+
+	if (res_value > 127 /* CHAR_MAX */)
+		return 0;
+
+	/* determine fractional part */
+	{
+		unsigned int i = FIXED1_8_BITS_PER_FRACTIONAL_PART;
+
+		do {
+			remainder <<= 1;
+
+			res_value <<= 1;
+
+			if (remainder >= arg2_value) {
+				res_value |= 1;
+				remainder -= arg2_value;
+			}
+		} while (--i != 0);
+	}
+
+	/* round up LSB */
+	{
+		uint16_t summand = (remainder << 1) >= arg2_value;
+
+		if ((res_value + summand) > 32767 /* SHRT_MAX */)
+			return 0;
+
+		res_value += summand;
+	}
+
+	return res_value;
+}
+
+static uint16_t vpe_internal_get_pratio(uint16_t from_frequency, uint16_t to_frequency)
+{
+	uint16_t pratio = vpe_u1_8_from_fraction(from_frequency, to_frequency);
+
+	if (GET_PRATIO_INTEGER_PART(pratio) > 1)
+		pratio = 0;
+
+	return pratio;
+}
+
+/*
+ * VPE has 4 DPM levels from level 0 (lowerest) to 3 (highest),
+ * VPE FW will dynamically decide which level should be used according to current loading.
+ *
+ * Get VPE and SOC clocks from PM, and select the appropriate four clock values,
+ * calculate the ratios of adjusting from one clock to another.
+ * The VPE FW can then request the appropriate frequency from the PMFW.
+ */
+int amdgpu_vpe_configure_dpm(struct amdgpu_vpe *vpe)
+{
+	struct amdgpu_device *adev = vpe->ring.adev;
+	uint32_t dpm_ctl;
+
+	if (adev->pm.dpm_enabled) {
+		struct dpm_clocks clock_table = { 0 };
+		struct dpm_clock *VPEClks;
+		struct dpm_clock *SOCClks;
+		uint32_t idx;
+		uint32_t vpeclk_enalbled_num = 0;
+		uint32_t pratio_vmax_vnorm = 0, pratio_vnorm_vmid = 0, pratio_vmid_vmin = 0;
+		uint16_t pratio_vmin_freq = 0, pratio_vmid_freq = 0, pratio_vnorm_freq = 0, pratio_vmax_freq = 0;
+
+		dpm_ctl = RREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_enable));
+		dpm_ctl |= 1; /* DPM enablement */
+		WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_enable), dpm_ctl);
+
+		/* Get VPECLK and SOCCLK */
+		if (amdgpu_dpm_get_dpm_clock_table(adev, &clock_table)) {
+			dev_dbg(adev->dev, "%s: get clock failed!\n", __func__);
+			goto disable_dpm;
+		}
+
+		SOCClks = clock_table.SocClocks;
+		VPEClks = clock_table.VPEClocks;
+
+		/* Comfirm enabled vpe clk num
+		 * Enabled VPE clocks are ordered from low to high in VPEClks
+		 * The highest valid clock index+1 is the number of VPEClks
+		 */
+		for (idx = PP_SMU_NUM_VPECLK_DPM_LEVELS; idx && !vpeclk_enalbled_num; idx--)
+			if (VPEClks[idx-1].Freq)
+				vpeclk_enalbled_num = idx;
+
+		/* vpe dpm only cares 4 levels. */
+		for (idx = 0; idx < VPE_MAX_DPM_LEVEL; idx++) {
+			uint32_t soc_dpm_level;
+			uint32_t min_freq;
+
+			if (idx == 0)
+				soc_dpm_level = 0;
+			else
+				soc_dpm_level = (idx * 2) + 1;
+
+			/* clamp the max level */
+			if (soc_dpm_level > vpeclk_enalbled_num - 1)
+				soc_dpm_level = vpeclk_enalbled_num - 1;
+
+			min_freq = (SOCClks[soc_dpm_level].Freq < VPEClks[soc_dpm_level].Freq) ?
+				   SOCClks[soc_dpm_level].Freq : VPEClks[soc_dpm_level].Freq;
+
+			switch (idx) {
+			case 0:
+				pratio_vmin_freq = min_freq;
+				break;
+			case 1:
+				pratio_vmid_freq = min_freq;
+				break;
+			case 2:
+				pratio_vnorm_freq = min_freq;
+				break;
+			case 3:
+				pratio_vmax_freq = min_freq;
+				break;
+			default:
+				break;
+			}
+		}
+
+		if (pratio_vmin_freq && pratio_vmid_freq && pratio_vnorm_freq && pratio_vmax_freq) {
+			uint32_t pratio_ctl;
+
+			pratio_vmax_vnorm = (uint32_t)vpe_internal_get_pratio(pratio_vmax_freq, pratio_vnorm_freq);
+			pratio_vnorm_vmid = (uint32_t)vpe_internal_get_pratio(pratio_vnorm_freq, pratio_vmid_freq);
+			pratio_vmid_vmin = (uint32_t)vpe_internal_get_pratio(pratio_vmid_freq, pratio_vmin_freq);
+
+			pratio_ctl = pratio_vmax_vnorm | (pratio_vnorm_vmid << 9) | (pratio_vmid_vmin << 18);
+			WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_pratio), pratio_ctl);		/* PRatio */
+			WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_request_interval), 24000);	/* 1ms, unit=1/24MHz */
+			WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_decision_threshold), 1200000);	/* 50ms */
+			WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_busy_clamp_threshold), 1200000);/* 50ms */
+			WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_idle_clamp_threshold), 1200000);/* 50ms */
+			dev_dbg(adev->dev, "%s: configure vpe dpm pratio done!\n", __func__);
+		} else {
+			dev_dbg(adev->dev, "%s: invalid pratio parameters!\n", __func__);
+			goto disable_dpm;
+		}
+	}
+	return 0;
+
+disable_dpm:
+	dpm_ctl = RREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_enable));
+	dpm_ctl &= 0xfffffffe; /* Disable DPM */
+	WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_enable), dpm_ctl);
+	dev_dbg(adev->dev, "%s: disable vpe dpm\n", __func__);
+	return -EINVAL;
+}
+
+int amdgpu_vpe_psp_update_sram(struct amdgpu_device *adev)
+{
+	struct amdgpu_firmware_info ucode = {
+		.ucode_id = AMDGPU_UCODE_ID_VPE,
+		.mc_addr = adev->vpe.cmdbuf_gpu_addr,
+		.ucode_size = 8,
+	};
+
+	return psp_execute_ip_fw_load(&adev->psp, &ucode);
+}
+
+int amdgpu_vpe_init_microcode(struct amdgpu_vpe *vpe)
+{
+	struct amdgpu_device *adev = vpe->ring.adev;
+	const struct vpe_firmware_header_v1_0 *vpe_hdr;
+	char fw_prefix[32];
+	int ret;
+
+	amdgpu_ucode_ip_version_decode(adev, VPE_HWIP, fw_prefix, sizeof(fw_prefix));
+	ret = amdgpu_ucode_request(adev, &adev->vpe.fw, AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s.bin", fw_prefix);
+	if (ret)
+		goto out;
+
+	vpe_hdr = (const struct vpe_firmware_header_v1_0 *)adev->vpe.fw->data;
+	adev->vpe.fw_version = le32_to_cpu(vpe_hdr->header.ucode_version);
+	adev->vpe.feature_version = le32_to_cpu(vpe_hdr->ucode_feature_version);
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		struct amdgpu_firmware_info *info;
+
+		info = &adev->firmware.ucode[AMDGPU_UCODE_ID_VPE_CTX];
+		info->ucode_id = AMDGPU_UCODE_ID_VPE_CTX;
+		info->fw = adev->vpe.fw;
+		adev->firmware.fw_size +=
+			ALIGN(le32_to_cpu(vpe_hdr->ctx_ucode_size_bytes), PAGE_SIZE);
+
+		info = &adev->firmware.ucode[AMDGPU_UCODE_ID_VPE_CTL];
+		info->ucode_id = AMDGPU_UCODE_ID_VPE_CTL;
+		info->fw = adev->vpe.fw;
+		adev->firmware.fw_size +=
+			ALIGN(le32_to_cpu(vpe_hdr->ctl_ucode_size_bytes), PAGE_SIZE);
+	}
+
+	return 0;
+out:
+	dev_err(adev->dev, "fail to initialize vpe microcode\n");
+	release_firmware(adev->vpe.fw);
+	adev->vpe.fw = NULL;
+	return ret;
+}
+
+int amdgpu_vpe_ring_init(struct amdgpu_vpe *vpe)
+{
+	struct amdgpu_device *adev = container_of(vpe, struct amdgpu_device, vpe);
+	struct amdgpu_ring *ring = &vpe->ring;
+	int ret;
+
+	ring->ring_obj = NULL;
+	ring->use_doorbell = true;
+	ring->vm_hub = AMDGPU_MMHUB0(0);
+	ring->doorbell_index = (adev->doorbell_index.vpe_ring << 1);
+	snprintf(ring->name, 4, "vpe");
+
+	ret = amdgpu_ring_init(adev, ring, 1024, &vpe->trap_irq, 0,
+			     AMDGPU_RING_PRIO_DEFAULT, NULL);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+int amdgpu_vpe_ring_fini(struct amdgpu_vpe *vpe)
+{
+	amdgpu_ring_fini(&vpe->ring);
+
+	return 0;
+}
+
+static int vpe_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+
+	switch (amdgpu_ip_version(adev, VPE_HWIP, 0)) {
+	case IP_VERSION(6, 1, 0):
+	case IP_VERSION(6, 1, 3):
+		vpe_v6_1_set_funcs(vpe);
+		break;
+	case IP_VERSION(6, 1, 1):
+		vpe_v6_1_set_funcs(vpe);
+		vpe->collaborate_mode = true;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	vpe_set_ring_funcs(adev);
+	vpe_set_regs(vpe);
+
+	dev_info(adev->dev, "VPE: collaborate mode %s", vpe->collaborate_mode ? "true" : "false");
+
+	return 0;
+}
+
+static bool vpe_need_dpm0_at_power_down(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, VPE_HWIP, 0)) {
+	case IP_VERSION(6, 1, 1):
+		return adev->pm.fw_version < 0x0a640500;
+	default:
+		return false;
+	}
+}
+
+static int vpe_get_dpm_level(struct amdgpu_device *adev)
+{
+	struct amdgpu_vpe *vpe = &adev->vpe;
+
+	if (!adev->pm.dpm_enabled)
+		return 0;
+
+	return RREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.dpm_request_lv));
+}
+
+static void vpe_idle_work_handler(struct work_struct *work)
+{
+	struct amdgpu_device *adev =
+		container_of(work, struct amdgpu_device, vpe.idle_work.work);
+	unsigned int fences = 0;
+
+	fences += amdgpu_fence_count_emitted(&adev->vpe.ring);
+	if (fences)
+		goto reschedule;
+
+	if (vpe_need_dpm0_at_power_down(adev) && vpe_get_dpm_level(adev) != 0)
+		goto reschedule;
+
+	amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE, AMD_PG_STATE_GATE);
+	return;
+
+reschedule:
+	schedule_delayed_work(&adev->vpe.idle_work, VPE_IDLE_TIMEOUT);
+}
+
+static int vpe_common_init(struct amdgpu_vpe *vpe)
+{
+	struct amdgpu_device *adev = container_of(vpe, struct amdgpu_device, vpe);
+	int r;
+
+	r = amdgpu_bo_create_kernel(adev, PAGE_SIZE, PAGE_SIZE,
+				    AMDGPU_GEM_DOMAIN_GTT,
+				    &adev->vpe.cmdbuf_obj,
+				    &adev->vpe.cmdbuf_gpu_addr,
+				    (void **)&adev->vpe.cmdbuf_cpu_addr);
+	if (r) {
+		dev_err(adev->dev, "VPE: failed to allocate cmdbuf bo %d\n", r);
+		return r;
+	}
+
+	vpe->context_started = false;
+	INIT_DELAYED_WORK(&adev->vpe.idle_work, vpe_idle_work_handler);
+
+	return 0;
+}
+
+static int vpe_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+	int ret;
+
+	ret = vpe_common_init(vpe);
+	if (ret)
+		goto out;
+
+	ret = vpe_irq_init(vpe);
+	if (ret)
+		goto out;
+
+	ret = vpe_ring_init(vpe);
+	if (ret)
+		goto out;
+
+	ret = vpe_init_microcode(vpe);
+	if (ret)
+		goto out;
+
+	adev->vpe.supported_reset =
+		 amdgpu_get_soft_full_reset_mask(&adev->vpe.ring);
+	if (!amdgpu_sriov_vf(adev))
+		adev->vpe.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+	ret = amdgpu_vpe_sysfs_reset_mask_init(adev);
+	if (ret)
+		goto out;
+out:
+	return ret;
+}
+
+static int vpe_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+
+	release_firmware(vpe->fw);
+	vpe->fw = NULL;
+
+	amdgpu_vpe_sysfs_reset_mask_fini(adev);
+	vpe_ring_fini(vpe);
+
+	amdgpu_bo_free_kernel(&adev->vpe.cmdbuf_obj,
+			      &adev->vpe.cmdbuf_gpu_addr,
+			      (void **)&adev->vpe.cmdbuf_cpu_addr);
+
+	return 0;
+}
+
+static int vpe_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+	int ret;
+
+	/* Power on VPE */
+	ret = amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE,
+						     AMD_PG_STATE_UNGATE);
+	if (ret)
+		return ret;
+
+	ret = vpe_load_microcode(vpe);
+	if (ret)
+		return ret;
+
+	ret = vpe_ring_start(vpe);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static int vpe_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+
+	cancel_delayed_work_sync(&adev->vpe.idle_work);
+
+	vpe_ring_stop(vpe);
+
+	/* Power off VPE */
+	amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE, AMD_PG_STATE_GATE);
+
+	return 0;
+}
+
+static int vpe_suspend(struct amdgpu_ip_block *ip_block)
+{
+	return vpe_hw_fini(ip_block);
+}
+
+static int vpe_resume(struct amdgpu_ip_block *ip_block)
+{
+	return vpe_hw_init(ip_block);
+}
+
+static void vpe_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
+{
+	int i;
+
+	for (i = 0; i < count; i++)
+		if (i == 0)
+			amdgpu_ring_write(ring, ring->funcs->nop |
+				VPE_CMD_NOP_HEADER_COUNT(count - 1));
+		else
+			amdgpu_ring_write(ring, ring->funcs->nop);
+}
+
+static uint64_t vpe_get_csa_mc_addr(struct amdgpu_ring *ring, uint32_t vmid)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t index = 0;
+	uint64_t csa_mc_addr;
+
+	if (amdgpu_sriov_vf(adev) || vmid == 0 || !adev->gfx.mcbp)
+		return 0;
+
+	csa_mc_addr = amdgpu_csa_vaddr(adev) + AMDGPU_CSA_VPE_OFFSET +
+		      index * AMDGPU_CSA_VPE_SIZE;
+
+	return csa_mc_addr;
+}
+
+static void vpe_ring_emit_pred_exec(struct amdgpu_ring *ring,
+				    uint32_t device_select,
+				    uint32_t exec_count)
+{
+	if (!ring->adev->vpe.collaborate_mode)
+		return;
+
+	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_PRED_EXE, 0) |
+				(device_select << 16));
+	amdgpu_ring_write(ring, exec_count & 0x1fff);
+}
+
+static void vpe_ring_emit_ib(struct amdgpu_ring *ring,
+			     struct amdgpu_job *job,
+			     struct amdgpu_ib *ib,
+			     uint32_t flags)
+{
+	uint32_t vmid = AMDGPU_JOB_GET_VMID(job);
+	uint64_t csa_mc_addr = vpe_get_csa_mc_addr(ring, vmid);
+
+	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_INDIRECT, 0) |
+				VPE_CMD_INDIRECT_HEADER_VMID(vmid & 0xf));
+
+	/* base must be 32 byte aligned */
+	amdgpu_ring_write(ring, ib->gpu_addr & 0xffffffe0);
+	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
+	amdgpu_ring_write(ring, ib->length_dw);
+	amdgpu_ring_write(ring, lower_32_bits(csa_mc_addr));
+	amdgpu_ring_write(ring, upper_32_bits(csa_mc_addr));
+}
+
+static void vpe_ring_emit_fence(struct amdgpu_ring *ring, uint64_t addr,
+				uint64_t seq, unsigned int flags)
+{
+	int i = 0;
+
+	do {
+		/* write the fence */
+		amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_FENCE, 0));
+		/* zero in first two bits */
+		WARN_ON_ONCE(addr & 0x3);
+		amdgpu_ring_write(ring, lower_32_bits(addr));
+		amdgpu_ring_write(ring, upper_32_bits(addr));
+		amdgpu_ring_write(ring, i == 0 ? lower_32_bits(seq) : upper_32_bits(seq));
+		addr += 4;
+	} while ((flags & AMDGPU_FENCE_FLAG_64BIT) && (i++ < 1));
+
+	if (flags & AMDGPU_FENCE_FLAG_INT) {
+		/* generate an interrupt */
+		amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_TRAP, 0));
+		amdgpu_ring_write(ring, 0);
+	}
+
+}
+
+static void vpe_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
+{
+	uint32_t seq = ring->fence_drv.sync_seq;
+	uint64_t addr = ring->fence_drv.gpu_addr;
+
+	vpe_ring_emit_pred_exec(ring, 0, 6);
+
+	/* wait for idle */
+	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_POLL_REGMEM,
+				VPE_POLL_REGMEM_SUBOP_REGMEM) |
+				VPE_CMD_POLL_REGMEM_HEADER_FUNC(3) | /* equal */
+				VPE_CMD_POLL_REGMEM_HEADER_MEM(1));
+	amdgpu_ring_write(ring, addr & 0xfffffffc);
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	amdgpu_ring_write(ring, seq); /* reference */
+	amdgpu_ring_write(ring, 0xffffffff); /* mask */
+	amdgpu_ring_write(ring, VPE_CMD_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) |
+				VPE_CMD_POLL_REGMEM_DW5_INTERVAL(4));
+}
+
+static void vpe_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, uint32_t val)
+{
+	vpe_ring_emit_pred_exec(ring, 0, 3);
+
+	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_REG_WRITE, 0));
+	amdgpu_ring_write(ring,	reg << 2);
+	amdgpu_ring_write(ring, val);
+}
+
+static void vpe_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
+				   uint32_t val, uint32_t mask)
+{
+	vpe_ring_emit_pred_exec(ring, 0, 6);
+
+	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_POLL_REGMEM,
+				VPE_POLL_REGMEM_SUBOP_REGMEM) |
+				VPE_CMD_POLL_REGMEM_HEADER_FUNC(3) | /* equal */
+				VPE_CMD_POLL_REGMEM_HEADER_MEM(0));
+	amdgpu_ring_write(ring, reg << 2);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, val); /* reference */
+	amdgpu_ring_write(ring, mask); /* mask */
+	amdgpu_ring_write(ring, VPE_CMD_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) |
+				VPE_CMD_POLL_REGMEM_DW5_INTERVAL(10));
+}
+
+static void vpe_ring_emit_vm_flush(struct amdgpu_ring *ring, unsigned int vmid,
+				   uint64_t pd_addr)
+{
+	amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
+}
+
+static unsigned int vpe_ring_init_cond_exec(struct amdgpu_ring *ring,
+					    uint64_t addr)
+{
+	unsigned int ret;
+
+	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_COND_EXE, 0));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	amdgpu_ring_write(ring, 1);
+	ret = ring->wptr & ring->buf_mask;
+	amdgpu_ring_write(ring, 0);
+
+	return ret;
+}
+
+static int vpe_ring_preempt_ib(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+	uint32_t preempt_reg = vpe->regs.queue0_preempt;
+	int i, r = 0;
+
+	/* assert preemption condition */
+	amdgpu_ring_set_preempt_cond_exec(ring, false);
+
+	/* emit the trailing fence */
+	ring->trail_seq += 1;
+	amdgpu_ring_alloc(ring, 10);
+	vpe_ring_emit_fence(ring, ring->trail_fence_gpu_addr, ring->trail_seq, 0);
+	amdgpu_ring_commit(ring);
+
+	/* assert IB preemption */
+	WREG32(vpe_get_reg_offset(vpe, ring->me, preempt_reg), 1);
+
+	/* poll the trailing fence */
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (ring->trail_seq ==
+		    le32_to_cpu(*(ring->trail_fence_cpu_addr)))
+			break;
+		udelay(1);
+	}
+
+	if (i >= adev->usec_timeout) {
+		r = -EINVAL;
+		dev_err(adev->dev, "ring %d failed to be preempted\n", ring->idx);
+	}
+
+	/* deassert IB preemption */
+	WREG32(vpe_get_reg_offset(vpe, ring->me, preempt_reg), 0);
+
+	/* deassert the preemption condition */
+	amdgpu_ring_set_preempt_cond_exec(ring, true);
+
+	return r;
+}
+
+static int vpe_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+				     enum amd_clockgating_state state)
+{
+	return 0;
+}
+
+static int vpe_set_powergating_state(struct amdgpu_ip_block *ip_block,
+				     enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+
+	if (!adev->pm.dpm_enabled)
+		dev_err(adev->dev, "Without PM, cannot support powergating\n");
+
+	dev_dbg(adev->dev, "%s: %s!\n", __func__, (state == AMD_PG_STATE_GATE) ? "GATE":"UNGATE");
+
+	if (state == AMD_PG_STATE_GATE) {
+		amdgpu_dpm_enable_vpe(adev, false);
+		vpe->context_started = false;
+	} else {
+		amdgpu_dpm_enable_vpe(adev, true);
+	}
+
+	return 0;
+}
+
+static uint64_t vpe_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+	uint64_t rptr;
+
+	if (ring->use_doorbell) {
+		rptr = atomic64_read((atomic64_t *)ring->rptr_cpu_addr);
+		dev_dbg(adev->dev, "rptr/doorbell before shift == 0x%016llx\n", rptr);
+	} else {
+		rptr = RREG32(vpe_get_reg_offset(vpe, ring->me, vpe->regs.queue0_rb_rptr_hi));
+		rptr = rptr << 32;
+		rptr |= RREG32(vpe_get_reg_offset(vpe, ring->me, vpe->regs.queue0_rb_rptr_lo));
+		dev_dbg(adev->dev, "rptr before shift [%i] == 0x%016llx\n", ring->me, rptr);
+	}
+
+	return (rptr >> 2);
+}
+
+static uint64_t vpe_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+	uint64_t wptr;
+
+	if (ring->use_doorbell) {
+		wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr);
+		dev_dbg(adev->dev, "wptr/doorbell before shift == 0x%016llx\n", wptr);
+	} else {
+		wptr = RREG32(vpe_get_reg_offset(vpe, ring->me, vpe->regs.queue0_rb_wptr_hi));
+		wptr = wptr << 32;
+		wptr |= RREG32(vpe_get_reg_offset(vpe, ring->me, vpe->regs.queue0_rb_wptr_lo));
+		dev_dbg(adev->dev, "wptr before shift [%i] == 0x%016llx\n", ring->me, wptr);
+	}
+
+	return (wptr >> 2);
+}
+
+static void vpe_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+
+	if (ring->use_doorbell) {
+		dev_dbg(adev->dev, "Using doorbell, \
+			wptr_offs == 0x%08x, \
+			lower_32_bits(ring->wptr) << 2 == 0x%08x, \
+			upper_32_bits(ring->wptr) << 2 == 0x%08x\n",
+			ring->wptr_offs,
+			lower_32_bits(ring->wptr << 2),
+			upper_32_bits(ring->wptr << 2));
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, ring->wptr << 2);
+		WDOORBELL64(ring->doorbell_index, ring->wptr << 2);
+		if (vpe->collaborate_mode)
+			WDOORBELL64(ring->doorbell_index + 4, ring->wptr << 2);
+	} else {
+		int i;
+
+		for (i = 0; i < vpe->num_instances; i++) {
+			dev_dbg(adev->dev, "Not using doorbell, \
+				regVPEC_QUEUE0_RB_WPTR == 0x%08x, \
+				regVPEC_QUEUE0_RB_WPTR_HI == 0x%08x\n",
+				lower_32_bits(ring->wptr << 2),
+				upper_32_bits(ring->wptr << 2));
+			WREG32(vpe_get_reg_offset(vpe, i, vpe->regs.queue0_rb_wptr_lo),
+			       lower_32_bits(ring->wptr << 2));
+			WREG32(vpe_get_reg_offset(vpe, i, vpe->regs.queue0_rb_wptr_hi),
+			       upper_32_bits(ring->wptr << 2));
+		}
+	}
+}
+
+static int vpe_ring_test_ring(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	const uint32_t test_pattern = 0xdeadbeef;
+	uint32_t index, i;
+	uint64_t wb_addr;
+	int ret;
+
+	ret = amdgpu_device_wb_get(adev, &index);
+	if (ret) {
+		dev_err(adev->dev, "(%d) failed to allocate wb slot\n", ret);
+		return ret;
+	}
+
+	adev->wb.wb[index] = 0;
+	wb_addr = adev->wb.gpu_addr + (index * 4);
+
+	ret = amdgpu_ring_alloc(ring, 4);
+	if (ret) {
+		dev_err(adev->dev, "amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, ret);
+		goto out;
+	}
+
+	amdgpu_ring_write(ring, VPE_CMD_HEADER(VPE_CMD_OPCODE_FENCE, 0));
+	amdgpu_ring_write(ring, lower_32_bits(wb_addr));
+	amdgpu_ring_write(ring, upper_32_bits(wb_addr));
+	amdgpu_ring_write(ring, test_pattern);
+	amdgpu_ring_commit(ring);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (le32_to_cpu(adev->wb.wb[index]) == test_pattern)
+			goto out;
+		udelay(1);
+	}
+
+	ret = -ETIMEDOUT;
+out:
+	amdgpu_device_wb_free(adev, index);
+
+	return ret;
+}
+
+static int vpe_ring_test_ib(struct amdgpu_ring *ring, long timeout)
+{
+	struct amdgpu_device *adev = ring->adev;
+	const uint32_t test_pattern = 0xdeadbeef;
+	struct amdgpu_ib ib = {};
+	struct dma_fence *f = NULL;
+	uint32_t index;
+	uint64_t wb_addr;
+	int ret;
+
+	ret = amdgpu_device_wb_get(adev, &index);
+	if (ret) {
+		dev_err(adev->dev, "(%d) failed to allocate wb slot\n", ret);
+		return ret;
+	}
+
+	adev->wb.wb[index] = 0;
+	wb_addr = adev->wb.gpu_addr + (index * 4);
+
+	ret = amdgpu_ib_get(adev, NULL, 256, AMDGPU_IB_POOL_DIRECT, &ib);
+	if (ret)
+		goto err0;
+
+	ib.ptr[0] = VPE_CMD_HEADER(VPE_CMD_OPCODE_FENCE, 0);
+	ib.ptr[1] = lower_32_bits(wb_addr);
+	ib.ptr[2] = upper_32_bits(wb_addr);
+	ib.ptr[3] = test_pattern;
+	ib.ptr[4] = VPE_CMD_HEADER(VPE_CMD_OPCODE_NOP, 0);
+	ib.ptr[5] = VPE_CMD_HEADER(VPE_CMD_OPCODE_NOP, 0);
+	ib.ptr[6] = VPE_CMD_HEADER(VPE_CMD_OPCODE_NOP, 0);
+	ib.ptr[7] = VPE_CMD_HEADER(VPE_CMD_OPCODE_NOP, 0);
+	ib.length_dw = 8;
+
+	ret = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f);
+	if (ret)
+		goto err1;
+
+	ret = dma_fence_wait_timeout(f, false, timeout);
+	if (ret <= 0) {
+		ret = ret ? : -ETIMEDOUT;
+		goto err1;
+	}
+
+	ret = (le32_to_cpu(adev->wb.wb[index]) == test_pattern) ? 0 : -EINVAL;
+
+err1:
+	amdgpu_ib_free(&ib, NULL);
+	dma_fence_put(f);
+err0:
+	amdgpu_device_wb_free(adev, index);
+
+	return ret;
+}
+
+static void vpe_ring_begin_use(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_vpe *vpe = &adev->vpe;
+
+	cancel_delayed_work_sync(&adev->vpe.idle_work);
+
+	/* Power on VPE and notify VPE of new context  */
+	if (!vpe->context_started) {
+		uint32_t context_notify;
+
+		/* Power on VPE */
+		amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE, AMD_PG_STATE_UNGATE);
+
+		/* Indicates that a job from a new context has been submitted. */
+		context_notify = RREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.context_indicator));
+		if ((context_notify & 0x1) == 0)
+			context_notify |= 0x1;
+		else
+			context_notify &= ~(0x1);
+		WREG32(vpe_get_reg_offset(vpe, 0, vpe->regs.context_indicator), context_notify);
+		vpe->context_started = true;
+	}
+}
+
+static void vpe_ring_end_use(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	schedule_delayed_work(&adev->vpe.idle_work, VPE_IDLE_TIMEOUT);
+}
+
+static int vpe_ring_reset(struct amdgpu_ring *ring,
+			  unsigned int vmid,
+			  struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int r;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	r = amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE,
+						   AMD_PG_STATE_GATE);
+	if (r)
+		return r;
+	r = amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VPE,
+						   AMD_PG_STATE_UNGATE);
+	if (r)
+		return r;
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static ssize_t amdgpu_get_vpe_reset_mask(struct device *dev,
+						struct device_attribute *attr,
+						char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (!adev)
+		return -ENODEV;
+
+	return amdgpu_show_reset_mask(buf, adev->vpe.supported_reset);
+}
+
+static DEVICE_ATTR(vpe_reset_mask, 0444,
+		   amdgpu_get_vpe_reset_mask, NULL);
+
+int amdgpu_vpe_sysfs_reset_mask_init(struct amdgpu_device *adev)
+{
+	int r = 0;
+
+	if (adev->vpe.num_instances) {
+		r = device_create_file(adev->dev, &dev_attr_vpe_reset_mask);
+		if (r)
+			return r;
+	}
+
+	return r;
+}
+
+void amdgpu_vpe_sysfs_reset_mask_fini(struct amdgpu_device *adev)
+{
+	if (adev->dev->kobj.sd) {
+		if (adev->vpe.num_instances)
+			device_remove_file(adev->dev, &dev_attr_vpe_reset_mask);
+	}
+}
+
+static const struct amdgpu_ring_funcs vpe_ring_funcs = {
+	.type = AMDGPU_RING_TYPE_VPE,
+	.align_mask = 0xf,
+	.nop = VPE_CMD_HEADER(VPE_CMD_OPCODE_NOP, 0),
+	.support_64bit_ptrs = true,
+	.get_rptr = vpe_ring_get_rptr,
+	.get_wptr = vpe_ring_get_wptr,
+	.set_wptr = vpe_ring_set_wptr,
+	.emit_frame_size =
+		5 + /* vpe_ring_init_cond_exec */
+		6 + /* vpe_ring_emit_pipeline_sync */
+		10 + 10 + 10 + /* vpe_ring_emit_fence */
+		/* vpe_ring_emit_vm_flush */
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 3 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 6,
+	.emit_ib_size = 7 + 6,
+	.emit_ib = vpe_ring_emit_ib,
+	.emit_pipeline_sync = vpe_ring_emit_pipeline_sync,
+	.emit_fence = vpe_ring_emit_fence,
+	.emit_vm_flush = vpe_ring_emit_vm_flush,
+	.emit_wreg = vpe_ring_emit_wreg,
+	.emit_reg_wait = vpe_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper,
+	.insert_nop = vpe_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.test_ring = vpe_ring_test_ring,
+	.test_ib = vpe_ring_test_ib,
+	.init_cond_exec = vpe_ring_init_cond_exec,
+	.preempt_ib = vpe_ring_preempt_ib,
+	.begin_use = vpe_ring_begin_use,
+	.end_use = vpe_ring_end_use,
+	.reset = vpe_ring_reset,
+};
+
+static void vpe_set_ring_funcs(struct amdgpu_device *adev)
+{
+	adev->vpe.ring.funcs = &vpe_ring_funcs;
+}
+
+const struct amd_ip_funcs vpe_ip_funcs = {
+	.name = "vpe_v6_1",
+	.early_init = vpe_early_init,
+	.sw_init = vpe_sw_init,
+	.sw_fini = vpe_sw_fini,
+	.hw_init = vpe_hw_init,
+	.hw_fini = vpe_hw_fini,
+	.suspend = vpe_suspend,
+	.resume = vpe_resume,
+	.set_clockgating_state = vpe_set_clockgating_state,
+	.set_powergating_state = vpe_set_powergating_state,
+};
+
+const struct amdgpu_ip_block_version vpe_v6_1_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_VPE,
+	.major = 6,
+	.minor = 1,
+	.rev = 0,
+	.funcs = &vpe_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h
new file mode 100644
index 000000000000..695da740a97e
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vpe.h
@@ -0,0 +1,111 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef __AMDGPU_VPE_H__
+#define __AMDGPU_VPE_H__
+
+#include "amdgpu_ring.h"
+#include "amdgpu_irq.h"
+#include "vpe_6_1_fw_if.h"
+
+#define AMDGPU_MAX_VPE_INSTANCES 2
+
+struct amdgpu_vpe;
+
+struct vpe_funcs {
+	uint32_t (*get_reg_offset)(struct amdgpu_vpe *vpe, uint32_t inst, uint32_t offset);
+	int (*set_regs)(struct amdgpu_vpe *vpe);
+	int (*irq_init)(struct amdgpu_vpe *vpe);
+	int (*init_microcode)(struct amdgpu_vpe *vpe);
+	int (*load_microcode)(struct amdgpu_vpe *vpe);
+	int (*ring_init)(struct amdgpu_vpe *vpe);
+	int (*ring_start)(struct amdgpu_vpe *vpe);
+	int (*ring_stop)(struct amdgpu_vpe *vpe);
+	int (*ring_fini)(struct amdgpu_vpe *vpe);
+};
+
+struct vpe_regs {
+	uint32_t queue0_rb_rptr_lo;
+	uint32_t queue0_rb_rptr_hi;
+	uint32_t queue0_rb_wptr_lo;
+	uint32_t queue0_rb_wptr_hi;
+	uint32_t queue0_preempt;
+
+	uint32_t dpm_enable;
+	uint32_t dpm_pratio;
+	uint32_t dpm_request_interval;
+	uint32_t dpm_decision_threshold;
+	uint32_t dpm_busy_clamp_threshold;
+	uint32_t dpm_idle_clamp_threshold;
+	uint32_t dpm_request_lv;
+	uint32_t context_indicator;
+};
+
+struct amdgpu_vpe {
+	struct amdgpu_ring		ring;
+	struct amdgpu_irq_src		trap_irq;
+
+	const struct vpe_funcs		*funcs;
+	struct vpe_regs			regs;
+
+	const struct firmware		*fw;
+	uint32_t			fw_version;
+	uint32_t			feature_version;
+
+	struct amdgpu_bo		*cmdbuf_obj;
+	uint64_t			cmdbuf_gpu_addr;
+	uint32_t			*cmdbuf_cpu_addr;
+	struct delayed_work		idle_work;
+	bool				context_started;
+
+	uint32_t			num_instances;
+	bool				collaborate_mode;
+	uint32_t			supported_reset;
+};
+
+int amdgpu_vpe_psp_update_sram(struct amdgpu_device *adev);
+int amdgpu_vpe_init_microcode(struct amdgpu_vpe *vpe);
+int amdgpu_vpe_ring_init(struct amdgpu_vpe *vpe);
+int amdgpu_vpe_ring_fini(struct amdgpu_vpe *vpe);
+int amdgpu_vpe_configure_dpm(struct amdgpu_vpe *vpe);
+void amdgpu_vpe_sysfs_reset_mask_fini(struct amdgpu_device *adev);
+int amdgpu_vpe_sysfs_reset_mask_init(struct amdgpu_device *adev);
+
+#define vpe_ring_init(vpe) ((vpe)->funcs->ring_init ? (vpe)->funcs->ring_init((vpe)) : 0)
+#define vpe_ring_start(vpe) ((vpe)->funcs->ring_start ? (vpe)->funcs->ring_start((vpe)) : 0)
+#define vpe_ring_stop(vpe) ((vpe)->funcs->ring_stop ? (vpe)->funcs->ring_stop((vpe)) : 0)
+#define vpe_ring_fini(vpe) ((vpe)->funcs->ring_fini ? (vpe)->funcs->ring_fini((vpe)) : 0)
+
+#define vpe_get_reg_offset(vpe, inst, offset) \
+		((vpe)->funcs->get_reg_offset ? (vpe)->funcs->get_reg_offset((vpe), (inst), (offset)) : 0)
+#define vpe_set_regs(vpe) \
+		((vpe)->funcs->set_regs ? (vpe)->funcs->set_regs((vpe)) : 0)
+#define vpe_irq_init(vpe) \
+		((vpe)->funcs->irq_init ? (vpe)->funcs->irq_init((vpe)) : 0)
+#define vpe_init_microcode(vpe) \
+		((vpe)->funcs->init_microcode ? (vpe)->funcs->init_microcode((vpe)) : 0)
+#define vpe_load_microcode(vpe) \
+		((vpe)->funcs->load_microcode ? (vpe)->funcs->load_microcode((vpe)) : 0)
+
+extern const struct amdgpu_ip_block_version vpe_v6_1_ip_block;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
index 43d6a9d6a538..9d934c07fa6b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c
@@ -24,13 +24,15 @@
 
 #include <linux/dma-mapping.h>
 #include <drm/ttm/ttm_range_manager.h>
+#include <drm/drm_drv.h>
 
 #include "amdgpu.h"
 #include "amdgpu_vm.h"
 #include "amdgpu_res_cursor.h"
-#include "amdgpu_atomfirmware.h"
 #include "atom.h"
 
+#define AMDGPU_MAX_SG_SEGMENT_SIZE	(2UL << 30)
+
 struct amdgpu_vram_reservation {
 	u64 start;
 	u64 size;
@@ -77,7 +79,16 @@ static inline bool amdgpu_is_vram_mgr_blocks_contiguous(struct list_head *head)
 	return true;
 }
 
+static inline u64 amdgpu_vram_mgr_blocks_size(struct list_head *head)
+{
+	struct drm_buddy_block *block;
+	u64 size = 0;
+
+	list_for_each_entry(block, head, link)
+		size += amdgpu_vram_mgr_block_size(block);
 
+	return size;
+}
 
 /**
  * DOC: mem_info_vram_total
@@ -212,8 +223,26 @@ static struct attribute *amdgpu_vram_mgr_attributes[] = {
 	NULL
 };
 
+static umode_t amdgpu_vram_attrs_is_visible(struct kobject *kobj,
+					    struct attribute *attr, int i)
+{
+	struct device *dev = kobj_to_dev(kobj);
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	if (attr == &dev_attr_mem_info_vram_vendor.attr &&
+	    !adev->gmc.vram_vendor)
+		return 0;
+
+	if (!ttm_resource_manager_used(&adev->mman.vram_mgr.manager))
+		return 0;
+
+	return attr->mode;
+}
+
 const struct attribute_group amdgpu_vram_mgr_attr_group = {
-	.attrs = amdgpu_vram_mgr_attributes
+	.attrs = amdgpu_vram_mgr_attributes,
+	.is_visible = amdgpu_vram_attrs_is_visible
 };
 
 /**
@@ -370,6 +399,35 @@ out:
 	return ret;
 }
 
+int amdgpu_vram_mgr_query_address_block_info(struct amdgpu_vram_mgr *mgr,
+			uint64_t address, struct amdgpu_vram_block_info *info)
+{
+	struct amdgpu_vram_mgr_resource *vres;
+	struct drm_buddy_block *block;
+	u64 start, size;
+	int ret = -ENOENT;
+
+	mutex_lock(&mgr->lock);
+	list_for_each_entry(vres, &mgr->allocated_vres_list, vres_node) {
+		list_for_each_entry(block, &vres->blocks, link) {
+			start = amdgpu_vram_mgr_block_start(block);
+			size = amdgpu_vram_mgr_block_size(block);
+			if ((start <= address) && (address < (start + size))) {
+				info->start = start;
+				info->size = size;
+				memcpy(&info->task, &vres->task, sizeof(vres->task));
+				ret = 0;
+				goto out;
+			}
+		}
+	}
+
+out:
+	mutex_unlock(&mgr->lock);
+
+	return ret;
+}
+
 /**
  * amdgpu_vram_mgr_new - allocate new ranges
  *
@@ -385,18 +443,20 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 			       const struct ttm_place *place,
 			       struct ttm_resource **res)
 {
-	u64 vis_usage = 0, max_bytes, cur_size, min_block_size;
 	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
 	struct amdgpu_device *adev = to_amdgpu_device(mgr);
+	struct amdgpu_bo *bo = ttm_to_amdgpu_bo(tbo);
+	u64 vis_usage = 0, max_bytes, min_block_size;
 	struct amdgpu_vram_mgr_resource *vres;
 	u64 size, remaining_size, lpfn, fpfn;
+	unsigned int adjust_dcc_size = 0;
 	struct drm_buddy *mm = &mgr->mm;
 	struct drm_buddy_block *block;
 	unsigned long pages_per_block;
 	int r;
 
 	lpfn = (u64)place->lpfn << PAGE_SHIFT;
-	if (!lpfn)
+	if (!lpfn || lpfn > man->size)
 		lpfn = man->size;
 
 	fpfn = (u64)place->fpfn << PAGE_SHIFT;
@@ -405,7 +465,7 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 	if (tbo->type != ttm_bo_type_kernel)
 		max_bytes -= AMDGPU_VM_RESERVED_VRAM;
 
-	if (place->flags & TTM_PL_FLAG_CONTIGUOUS) {
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS) {
 		pages_per_block = ~0ul;
 	} else {
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
@@ -414,7 +474,7 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 		/* default to 2MB */
 		pages_per_block = 2UL << (20UL - PAGE_SHIFT);
 #endif
-		pages_per_block = max_t(uint32_t, pages_per_block,
+		pages_per_block = max_t(u32, pages_per_block,
 					tbo->page_alignment);
 	}
 
@@ -435,11 +495,29 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 	if (place->flags & TTM_PL_FLAG_TOPDOWN)
 		vres->flags |= DRM_BUDDY_TOPDOWN_ALLOCATION;
 
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)
+		vres->flags |= DRM_BUDDY_CONTIGUOUS_ALLOCATION;
+
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CLEARED)
+		vres->flags |= DRM_BUDDY_CLEAR_ALLOCATION;
+
 	if (fpfn || lpfn != mgr->mm.size)
 		/* Allocate blocks in desired range */
 		vres->flags |= DRM_BUDDY_RANGE_ALLOCATION;
 
+	if (bo->flags & AMDGPU_GEM_CREATE_GFX12_DCC &&
+	    adev->gmc.gmc_funcs->get_dcc_alignment)
+		adjust_dcc_size = amdgpu_gmc_get_dcc_alignment(adev);
+
 	remaining_size = (u64)vres->base.size;
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS && adjust_dcc_size) {
+		unsigned int dcc_size;
+
+		dcc_size = roundup_pow_of_two(vres->base.size + adjust_dcc_size);
+		remaining_size = (u64)dcc_size;
+
+		vres->flags |= DRM_BUDDY_TRIM_DISABLE;
+	}
 
 	mutex_lock(&mgr->lock);
 	while (remaining_size) {
@@ -448,33 +526,15 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 		else
 			min_block_size = mgr->default_page_size;
 
-		BUG_ON(min_block_size < mm->chunk_size);
-
-		/* Limit maximum size to 2GiB due to SG table limitations */
-		size = min(remaining_size, 2ULL << 30);
+		size = remaining_size;
 
-		if ((size >= (u64)pages_per_block << PAGE_SHIFT) &&
-				!(size & (((u64)pages_per_block << PAGE_SHIFT) - 1)))
+		if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS && adjust_dcc_size)
+			min_block_size = size;
+		else if ((size >= (u64)pages_per_block << PAGE_SHIFT) &&
+			 !(size & (((u64)pages_per_block << PAGE_SHIFT) - 1)))
 			min_block_size = (u64)pages_per_block << PAGE_SHIFT;
 
-		cur_size = size;
-
-		if (fpfn + size != (u64)place->lpfn << PAGE_SHIFT) {
-			/*
-			 * Except for actual range allocation, modify the size and
-			 * min_block_size conforming to continuous flag enablement
-			 */
-			if (place->flags & TTM_PL_FLAG_CONTIGUOUS) {
-				size = roundup_pow_of_two(size);
-				min_block_size = size;
-			/*
-			 * Modify the size value if size is not
-			 * aligned with min_block_size
-			 */
-			} else if (!IS_ALIGNED(size, min_block_size)) {
-				size = round_up(size, min_block_size);
-			}
-		}
+		BUG_ON(min_block_size < mm->chunk_size);
 
 		r = drm_buddy_alloc_blocks(mm, fpfn,
 					   lpfn,
@@ -482,6 +542,16 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 					   min_block_size,
 					   &vres->blocks,
 					   vres->flags);
+
+		if (unlikely(r == -ENOSPC) && pages_per_block == ~0ul &&
+		    !(place->flags & TTM_PL_FLAG_CONTIGUOUS)) {
+			vres->flags &= ~DRM_BUDDY_CONTIGUOUS_ALLOCATION;
+			pages_per_block = max_t(u32, 2UL << (20UL - PAGE_SHIFT),
+						tbo->page_alignment);
+
+			continue;
+		}
+
 		if (unlikely(r))
 			goto error_free_blocks;
 
@@ -490,43 +560,31 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 		else
 			remaining_size -= size;
 	}
-	mutex_unlock(&mgr->lock);
-
-	if (cur_size != size) {
-		struct drm_buddy_block *block;
-		struct list_head *trim_list;
-		u64 original_size;
-		LIST_HEAD(temp);
-
-		trim_list = &vres->blocks;
-		original_size = (u64)vres->base.size;
-
-		/*
-		 * If size value is rounded up to min_block_size, trim the last
-		 * block to the required size
-		 */
-		if (!list_is_singular(&vres->blocks)) {
-			block = list_last_entry(&vres->blocks, typeof(*block), link);
-			list_move_tail(&block->link, &temp);
-			trim_list = &temp;
-			/*
-			 * Compute the original_size value by subtracting the
-			 * last block size with (aligned size - original size)
-			 */
-			original_size = amdgpu_vram_mgr_block_size(block) - (size - cur_size);
-		}
-
-		mutex_lock(&mgr->lock);
-		drm_buddy_block_trim(mm,
-				     original_size,
-				     trim_list);
-		mutex_unlock(&mgr->lock);
 
-		if (!list_empty(&temp))
-			list_splice_tail(trim_list, &vres->blocks);
+	vres->task.pid = task_pid_nr(current);
+	get_task_comm(vres->task.comm, current);
+	list_add_tail(&vres->vres_node, &mgr->allocated_vres_list);
+
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS && adjust_dcc_size) {
+		struct drm_buddy_block *dcc_block;
+		unsigned long dcc_start;
+		u64 trim_start;
+
+		dcc_block = amdgpu_vram_mgr_first_block(&vres->blocks);
+		/* Adjust the start address for DCC buffers only */
+		dcc_start =
+			roundup((unsigned long)amdgpu_vram_mgr_block_start(dcc_block),
+				adjust_dcc_size);
+		trim_start = (u64)dcc_start;
+		drm_buddy_block_trim(mm, &trim_start,
+				     (u64)vres->base.size,
+				     &vres->blocks);
 	}
+	mutex_unlock(&mgr->lock);
 
 	vres->base.start = 0;
+	size = max_t(u64, amdgpu_vram_mgr_blocks_size(&vres->blocks),
+		     vres->base.size);
 	list_for_each_entry(block, &vres->blocks, link) {
 		unsigned long start;
 
@@ -534,8 +592,8 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 			amdgpu_vram_mgr_block_size(block);
 		start >>= PAGE_SHIFT;
 
-		if (start > PFN_UP(vres->base.size))
-			start -= PFN_UP(vres->base.size);
+		if (start > PFN_UP(size))
+			start -= PFN_UP(size);
 		else
 			start = 0;
 		vres->base.start = max(vres->base.start, start);
@@ -556,7 +614,7 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man,
 	return 0;
 
 error_free_blocks:
-	drm_buddy_free_list(mm, &vres->blocks);
+	drm_buddy_free_list(mm, &vres->blocks, 0);
 	mutex_unlock(&mgr->lock);
 error_fini:
 	ttm_resource_fini(man, &vres->base);
@@ -584,12 +642,15 @@ static void amdgpu_vram_mgr_del(struct ttm_resource_manager *man,
 	uint64_t vis_usage = 0;
 
 	mutex_lock(&mgr->lock);
+
+	list_del(&vres->vres_node);
+	memset(&vres->task, 0, sizeof(vres->task));
+
 	list_for_each_entry(block, &vres->blocks, link)
 		vis_usage += amdgpu_vram_mgr_vis_size(adev, block);
 
+	drm_buddy_free_list(mm, &vres->blocks, vres->flags);
 	amdgpu_vram_mgr_do_reserve(man);
-
-	drm_buddy_free_list(mm, &vres->blocks);
 	mutex_unlock(&mgr->lock);
 
 	atomic64_sub(vis_usage, &mgr->vis_usage);
@@ -631,7 +692,7 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev,
 	amdgpu_res_first(res, offset, length, &cursor);
 	while (cursor.remaining) {
 		num_entries++;
-		amdgpu_res_next(&cursor, cursor.size);
+		amdgpu_res_next(&cursor, min(cursor.size, AMDGPU_MAX_SG_SEGMENT_SIZE));
 	}
 
 	r = sg_alloc_table(*sgt, num_entries, GFP_KERNEL);
@@ -651,7 +712,7 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev,
 	amdgpu_res_first(res, offset, length, &cursor);
 	for_each_sgtable_sg((*sgt), sg, i) {
 		phys_addr_t phys = cursor.start + adev->gmc.aper_base;
-		size_t size = cursor.size;
+		unsigned long size = min(cursor.size, AMDGPU_MAX_SG_SEGMENT_SIZE);
 		dma_addr_t addr;
 
 		addr = dma_map_resource(dev, phys, size, dir,
@@ -664,7 +725,7 @@ int amdgpu_vram_mgr_alloc_sgt(struct amdgpu_device *adev,
 		sg_dma_address(sg) = addr;
 		sg_dma_len(sg) = size;
 
-		amdgpu_res_next(&cursor, cursor.size);
+		amdgpu_res_next(&cursor, size);
 	}
 
 	return 0;
@@ -722,6 +783,23 @@ uint64_t amdgpu_vram_mgr_vis_usage(struct amdgpu_vram_mgr *mgr)
 }
 
 /**
+ * amdgpu_vram_mgr_clear_reset_blocks - reset clear blocks
+ *
+ * @adev: amdgpu device pointer
+ *
+ * Reset the cleared drm buddy blocks.
+ */
+void amdgpu_vram_mgr_clear_reset_blocks(struct amdgpu_device *adev)
+{
+	struct amdgpu_vram_mgr *mgr = &adev->mman.vram_mgr;
+	struct drm_buddy *mm = &mgr->mm;
+
+	mutex_lock(&mgr->lock);
+	drm_buddy_reset_clear(mm, false);
+	mutex_unlock(&mgr->lock);
+}
+
+/**
  * amdgpu_vram_mgr_intersects - test each drm buddy block for intersection
  *
  * @man: TTM memory type manager
@@ -800,7 +878,7 @@ static void amdgpu_vram_mgr_debug(struct ttm_resource_manager *man,
 {
 	struct amdgpu_vram_mgr *mgr = to_vram_mgr(man);
 	struct drm_buddy *mm = &mgr->mm;
-	struct drm_buddy_block *block;
+	struct amdgpu_vram_reservation *rsv;
 
 	drm_printf(printer, "  vis usage:%llu\n",
 		   amdgpu_vram_mgr_vis_usage(mgr));
@@ -812,8 +890,9 @@ static void amdgpu_vram_mgr_debug(struct ttm_resource_manager *man,
 	drm_buddy_print(mm, printer);
 
 	drm_printf(printer, "reserved:\n");
-	list_for_each_entry(block, &mgr->reserved_pages, link)
-		drm_buddy_block_print(mm, block, printer);
+	list_for_each_entry(rsv, &mgr->reserved_pages, blocks)
+		drm_printf(printer, "%#018llx-%#018llx: %llu\n",
+			rsv->start, rsv->start + rsv->size, rsv->size);
 	mutex_unlock(&mgr->lock);
 }
 
@@ -838,20 +917,23 @@ int amdgpu_vram_mgr_init(struct amdgpu_device *adev)
 	struct ttm_resource_manager *man = &mgr->manager;
 	int err;
 
+	man->cg = drmm_cgroup_register_region(adev_to_drm(adev), "vram", adev->gmc.real_vram_size);
+	if (IS_ERR(man->cg))
+		return PTR_ERR(man->cg);
 	ttm_resource_manager_init(man, &adev->mman.bdev,
 				  adev->gmc.real_vram_size);
 
-	man->func = &amdgpu_vram_mgr_func;
-
-	err = drm_buddy_init(&mgr->mm, man->size, PAGE_SIZE);
-	if (err)
-		return err;
-
 	mutex_init(&mgr->lock);
 	INIT_LIST_HEAD(&mgr->reservations_pending);
 	INIT_LIST_HEAD(&mgr->reserved_pages);
+	INIT_LIST_HEAD(&mgr->allocated_vres_list);
 	mgr->default_page_size = PAGE_SIZE;
 
+	man->func = &amdgpu_vram_mgr_func;
+	err = drm_buddy_init(&mgr->mm, man->size, PAGE_SIZE);
+	if (err)
+		return err;
+
 	ttm_set_driver_manager(&adev->mman.bdev, TTM_PL_VRAM, &mgr->manager);
 	ttm_resource_manager_set_used(man, true);
 	return 0;
@@ -883,10 +965,11 @@ void amdgpu_vram_mgr_fini(struct amdgpu_device *adev)
 		kfree(rsv);
 
 	list_for_each_entry_safe(rsv, temp, &mgr->reserved_pages, blocks) {
-		drm_buddy_free_list(&mgr->mm, &rsv->allocated);
+		drm_buddy_free_list(&mgr->mm, &rsv->allocated, 0);
 		kfree(rsv);
 	}
-	drm_buddy_fini(&mgr->mm);
+	if (!adev->gmc.is_app_apu)
+		drm_buddy_fini(&mgr->mm);
 	mutex_unlock(&mgr->lock);
 
 	ttm_resource_manager_cleanup(man);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.h
index 0e04e42cf809..5f5fd9a911c2 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.h
@@ -35,12 +35,26 @@ struct amdgpu_vram_mgr {
 	struct list_head reserved_pages;
 	atomic64_t vis_usage;
 	u64 default_page_size;
+	struct list_head allocated_vres_list;
+};
+
+struct amdgpu_vres_task {
+	pid_t pid;
+	char  comm[TASK_COMM_LEN];
+};
+
+struct amdgpu_vram_block_info {
+	u64 start;
+	u64 size;
+	struct amdgpu_vres_task task;
 };
 
 struct amdgpu_vram_mgr_resource {
 	struct ttm_resource base;
 	struct list_head blocks;
 	unsigned long flags;
+	struct list_head vres_node;
+	struct amdgpu_vres_task task;
 };
 
 static inline u64 amdgpu_vram_mgr_block_start(struct drm_buddy_block *block)
@@ -53,10 +67,26 @@ static inline u64 amdgpu_vram_mgr_block_size(struct drm_buddy_block *block)
 	return (u64)PAGE_SIZE << drm_buddy_block_order(block);
 }
 
+static inline bool amdgpu_vram_mgr_is_cleared(struct drm_buddy_block *block)
+{
+	return drm_buddy_block_is_clear(block);
+}
+
 static inline struct amdgpu_vram_mgr_resource *
 to_amdgpu_vram_mgr_resource(struct ttm_resource *res)
 {
 	return container_of(res, struct amdgpu_vram_mgr_resource, base);
 }
 
+static inline void amdgpu_vram_mgr_set_cleared(struct ttm_resource *res)
+{
+	struct amdgpu_vram_mgr_resource *ares = to_amdgpu_vram_mgr_resource(res);
+
+	WARN_ON(ares->flags & DRM_BUDDY_CLEARED);
+	ares->flags |= DRM_BUDDY_CLEARED;
+}
+
+int amdgpu_vram_mgr_query_address_block_info(struct amdgpu_vram_mgr *mgr,
+		uint64_t address, struct amdgpu_vram_block_info *info);
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.c
new file mode 100644
index 000000000000..1083db8cea2e
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.c
@@ -0,0 +1,1107 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "amdgpu_xcp.h"
+#include "amdgpu_drv.h"
+
+#include <drm/drm_drv.h>
+#include "../amdxcp/amdgpu_xcp_drv.h"
+
+static void amdgpu_xcp_sysfs_entries_init(struct amdgpu_xcp_mgr *xcp_mgr);
+static void amdgpu_xcp_sysfs_entries_update(struct amdgpu_xcp_mgr *xcp_mgr);
+
+static int __amdgpu_xcp_run(struct amdgpu_xcp_mgr *xcp_mgr,
+			    struct amdgpu_xcp_ip *xcp_ip, int xcp_state)
+{
+	int (*run_func)(void *handle, uint32_t inst_mask);
+	int ret = 0;
+
+	if (!xcp_ip || !xcp_ip->valid || !xcp_ip->ip_funcs)
+		return 0;
+
+	run_func = NULL;
+
+	switch (xcp_state) {
+	case AMDGPU_XCP_PREPARE_SUSPEND:
+		run_func = xcp_ip->ip_funcs->prepare_suspend;
+		break;
+	case AMDGPU_XCP_SUSPEND:
+		run_func = xcp_ip->ip_funcs->suspend;
+		break;
+	case AMDGPU_XCP_PREPARE_RESUME:
+		run_func = xcp_ip->ip_funcs->prepare_resume;
+		break;
+	case AMDGPU_XCP_RESUME:
+		run_func = xcp_ip->ip_funcs->resume;
+		break;
+	}
+
+	if (run_func)
+		ret = run_func(xcp_mgr->adev, xcp_ip->inst_mask);
+
+	return ret;
+}
+
+static int amdgpu_xcp_run_transition(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id,
+				     int state)
+{
+	struct amdgpu_xcp_ip *xcp_ip;
+	struct amdgpu_xcp *xcp;
+	int i, ret;
+
+	if (xcp_id >= MAX_XCP || !xcp_mgr->xcp[xcp_id].valid)
+		return -EINVAL;
+
+	xcp = &xcp_mgr->xcp[xcp_id];
+	for (i = 0; i < AMDGPU_XCP_MAX_BLOCKS; ++i) {
+		xcp_ip = &xcp->ip[i];
+		ret = __amdgpu_xcp_run(xcp_mgr, xcp_ip, state);
+		if (ret)
+			break;
+	}
+
+	return ret;
+}
+
+int amdgpu_xcp_prepare_suspend(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id)
+{
+	return amdgpu_xcp_run_transition(xcp_mgr, xcp_id,
+					 AMDGPU_XCP_PREPARE_SUSPEND);
+}
+
+int amdgpu_xcp_suspend(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id)
+{
+	return amdgpu_xcp_run_transition(xcp_mgr, xcp_id, AMDGPU_XCP_SUSPEND);
+}
+
+int amdgpu_xcp_prepare_resume(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id)
+{
+	return amdgpu_xcp_run_transition(xcp_mgr, xcp_id,
+					 AMDGPU_XCP_PREPARE_RESUME);
+}
+
+int amdgpu_xcp_resume(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id)
+{
+	return amdgpu_xcp_run_transition(xcp_mgr, xcp_id, AMDGPU_XCP_RESUME);
+}
+
+static void __amdgpu_xcp_add_block(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id,
+				   struct amdgpu_xcp_ip *ip)
+{
+	struct amdgpu_xcp *xcp;
+
+	if (!ip)
+		return;
+
+	xcp = &xcp_mgr->xcp[xcp_id];
+	xcp->ip[ip->ip_id] = *ip;
+	xcp->ip[ip->ip_id].valid = true;
+
+	xcp->valid = true;
+}
+
+static void __amdgpu_xcp_set_unique_id(struct amdgpu_xcp_mgr *xcp_mgr,
+				       int xcp_id)
+{
+	struct amdgpu_xcp *xcp = &xcp_mgr->xcp[xcp_id];
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	uint32_t inst_mask;
+	uint64_t uid;
+	int i;
+
+	if (!amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_GFX, &inst_mask) &&
+	    inst_mask) {
+		i = GET_INST(GC, (ffs(inst_mask) - 1));
+		uid = amdgpu_device_get_uid(xcp_mgr->adev->uid_info,
+					    AMDGPU_UID_TYPE_XCD, i);
+		if (uid)
+			xcp->unique_id = uid;
+	}
+}
+
+int amdgpu_xcp_init(struct amdgpu_xcp_mgr *xcp_mgr, int num_xcps, int mode)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	struct amdgpu_xcp_ip ip;
+	uint8_t mem_id;
+	int i, j, ret;
+
+	if (!num_xcps || num_xcps > MAX_XCP)
+		return -EINVAL;
+
+	xcp_mgr->mode = mode;
+
+	for (i = 0; i < MAX_XCP; ++i)
+		xcp_mgr->xcp[i].valid = false;
+
+	/* This is needed for figuring out memory id of xcp */
+	xcp_mgr->num_xcp_per_mem_partition = num_xcps / xcp_mgr->adev->gmc.num_mem_partitions;
+
+	for (i = 0; i < num_xcps; ++i) {
+		for (j = AMDGPU_XCP_GFXHUB; j < AMDGPU_XCP_MAX_BLOCKS; ++j) {
+			ret = xcp_mgr->funcs->get_ip_details(xcp_mgr, i, j,
+							     &ip);
+			if (ret)
+				continue;
+
+			__amdgpu_xcp_add_block(xcp_mgr, i, &ip);
+		}
+
+		xcp_mgr->xcp[i].id = i;
+
+		if (xcp_mgr->funcs->get_xcp_mem_id) {
+			ret = xcp_mgr->funcs->get_xcp_mem_id(
+				xcp_mgr, &xcp_mgr->xcp[i], &mem_id);
+			if (ret)
+				continue;
+			else
+				xcp_mgr->xcp[i].mem_id = mem_id;
+		}
+		__amdgpu_xcp_set_unique_id(xcp_mgr, i);
+	}
+
+	xcp_mgr->num_xcps = num_xcps;
+	amdgpu_xcp_update_partition_sched_list(adev);
+
+	return 0;
+}
+
+static int __amdgpu_xcp_switch_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr,
+					      int mode)
+{
+	int ret, curr_mode, num_xcps = 0;
+
+	if (!xcp_mgr->funcs || !xcp_mgr->funcs->switch_partition_mode)
+		return 0;
+
+	mutex_lock(&xcp_mgr->xcp_lock);
+
+	curr_mode = xcp_mgr->mode;
+	/* State set to transient mode */
+	xcp_mgr->mode = AMDGPU_XCP_MODE_TRANS;
+
+	ret = xcp_mgr->funcs->switch_partition_mode(xcp_mgr, mode, &num_xcps);
+
+	if (ret) {
+		/* Failed, get whatever mode it's at now */
+		if (xcp_mgr->funcs->query_partition_mode)
+			xcp_mgr->mode = amdgpu_xcp_query_partition_mode(
+				xcp_mgr, AMDGPU_XCP_FL_LOCKED);
+		else
+			xcp_mgr->mode = curr_mode;
+
+		goto out;
+	}
+	amdgpu_xcp_sysfs_entries_update(xcp_mgr);
+out:
+	mutex_unlock(&xcp_mgr->xcp_lock);
+
+	return ret;
+}
+
+int amdgpu_xcp_switch_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr, int mode)
+{
+	if (!xcp_mgr || mode == AMDGPU_XCP_MODE_NONE)
+		return -EINVAL;
+
+	if (xcp_mgr->mode == mode)
+		return 0;
+
+	return __amdgpu_xcp_switch_partition_mode(xcp_mgr, mode);
+}
+
+int amdgpu_xcp_restore_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	if (!xcp_mgr || xcp_mgr->mode == AMDGPU_XCP_MODE_NONE)
+		return 0;
+
+	return __amdgpu_xcp_switch_partition_mode(xcp_mgr, xcp_mgr->mode);
+}
+
+static bool __amdgpu_xcp_is_cached_mode_valid(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	if (!xcp_mgr->funcs || !xcp_mgr->funcs->query_partition_mode)
+		return true;
+
+	if (!amdgpu_sriov_vf(xcp_mgr->adev) &&
+	    xcp_mgr->mode == AMDGPU_XCP_MODE_NONE)
+		return true;
+
+	if (xcp_mgr->mode != AMDGPU_XCP_MODE_NONE &&
+	    xcp_mgr->mode != AMDGPU_XCP_MODE_TRANS)
+		return true;
+
+	return false;
+}
+
+int amdgpu_xcp_query_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr, u32 flags)
+{
+	int mode;
+
+	if (__amdgpu_xcp_is_cached_mode_valid(xcp_mgr))
+		return xcp_mgr->mode;
+
+	if (!(flags & AMDGPU_XCP_FL_LOCKED))
+		mutex_lock(&xcp_mgr->xcp_lock);
+	mode = xcp_mgr->funcs->query_partition_mode(xcp_mgr);
+
+	/* First time query for VF, set the mode here */
+	if (amdgpu_sriov_vf(xcp_mgr->adev) &&
+	    xcp_mgr->mode == AMDGPU_XCP_MODE_NONE)
+		xcp_mgr->mode = mode;
+
+	if (xcp_mgr->mode != AMDGPU_XCP_MODE_TRANS && mode != xcp_mgr->mode)
+		dev_WARN(
+			xcp_mgr->adev->dev,
+			"Cached partition mode %d not matching with device mode %d",
+			xcp_mgr->mode, mode);
+
+	if (!(flags & AMDGPU_XCP_FL_LOCKED))
+		mutex_unlock(&xcp_mgr->xcp_lock);
+
+	return mode;
+}
+
+static int amdgpu_xcp_dev_alloc(struct amdgpu_device *adev)
+{
+	struct drm_device *p_ddev;
+	struct drm_device *ddev;
+	int i, ret;
+
+	ddev = adev_to_drm(adev);
+
+	/* xcp #0 shares drm device setting with adev */
+	adev->xcp_mgr->xcp->ddev = ddev;
+
+	for (i = 1; i < MAX_XCP; i++) {
+		ret = amdgpu_xcp_drm_dev_alloc(&p_ddev);
+		if (ret == -ENOSPC) {
+			dev_warn(adev->dev,
+			"Skip xcp node #%d when out of drm node resource.", i);
+			ret = 0;
+			goto out;
+		} else if (ret) {
+			goto out;
+		}
+
+		/* Redirect all IOCTLs to the primary device */
+		adev->xcp_mgr->xcp[i].rdev = p_ddev->render->dev;
+		adev->xcp_mgr->xcp[i].pdev = p_ddev->primary->dev;
+		adev->xcp_mgr->xcp[i].driver = (struct drm_driver *)p_ddev->driver;
+		adev->xcp_mgr->xcp[i].vma_offset_manager = p_ddev->vma_offset_manager;
+		p_ddev->render->dev = ddev;
+		p_ddev->primary->dev = ddev;
+		p_ddev->vma_offset_manager = ddev->vma_offset_manager;
+		p_ddev->driver = &amdgpu_partition_driver;
+		adev->xcp_mgr->xcp[i].ddev = p_ddev;
+
+		dev_set_drvdata(p_ddev->dev, &adev->xcp_mgr->xcp[i]);
+	}
+	ret = 0;
+out:
+	amdgpu_xcp_sysfs_entries_init(adev->xcp_mgr);
+
+	return ret;
+}
+
+int amdgpu_xcp_mgr_init(struct amdgpu_device *adev, int init_mode,
+			int init_num_xcps,
+			struct amdgpu_xcp_mgr_funcs *xcp_funcs)
+{
+	struct amdgpu_xcp_mgr *xcp_mgr;
+	int i;
+
+	if (!xcp_funcs || !xcp_funcs->get_ip_details)
+		return -EINVAL;
+
+	xcp_mgr = kzalloc(sizeof(*xcp_mgr), GFP_KERNEL);
+
+	if (!xcp_mgr)
+		return -ENOMEM;
+
+	xcp_mgr->adev = adev;
+	xcp_mgr->funcs = xcp_funcs;
+	xcp_mgr->mode = init_mode;
+	mutex_init(&xcp_mgr->xcp_lock);
+
+	if (init_mode != AMDGPU_XCP_MODE_NONE)
+		amdgpu_xcp_init(xcp_mgr, init_num_xcps, init_mode);
+
+	adev->xcp_mgr = xcp_mgr;
+	for (i = 0; i < MAX_XCP; ++i)
+		xcp_mgr->xcp[i].xcp_mgr = xcp_mgr;
+
+	return amdgpu_xcp_dev_alloc(adev);
+}
+
+int amdgpu_xcp_get_partition(struct amdgpu_xcp_mgr *xcp_mgr,
+			     enum AMDGPU_XCP_IP_BLOCK ip, int instance)
+{
+	struct amdgpu_xcp *xcp;
+	int i, id_mask = 0;
+
+	if (ip >= AMDGPU_XCP_MAX_BLOCKS)
+		return -EINVAL;
+
+	for (i = 0; i < xcp_mgr->num_xcps; ++i) {
+		xcp = &xcp_mgr->xcp[i];
+		if ((xcp->valid) && (xcp->ip[ip].valid) &&
+		    (xcp->ip[ip].inst_mask & BIT(instance)))
+			id_mask |= BIT(i);
+	}
+
+	if (!id_mask)
+		id_mask = -ENXIO;
+
+	return id_mask;
+}
+
+int amdgpu_xcp_get_inst_details(struct amdgpu_xcp *xcp,
+				enum AMDGPU_XCP_IP_BLOCK ip,
+				uint32_t *inst_mask)
+{
+	if (!xcp->valid || !inst_mask || !(xcp->ip[ip].valid))
+		return -EINVAL;
+
+	*inst_mask = xcp->ip[ip].inst_mask;
+
+	return 0;
+}
+
+int amdgpu_xcp_dev_register(struct amdgpu_device *adev,
+			const struct pci_device_id *ent)
+{
+	int i, ret;
+
+	if (!adev->xcp_mgr)
+		return 0;
+
+	for (i = 1; i < MAX_XCP; i++) {
+		if (!adev->xcp_mgr->xcp[i].ddev)
+			break;
+
+		ret = drm_dev_register(adev->xcp_mgr->xcp[i].ddev, ent->driver_data);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
+void amdgpu_xcp_dev_unplug(struct amdgpu_device *adev)
+{
+	struct drm_device *p_ddev;
+	int i;
+
+	if (!adev->xcp_mgr)
+		return;
+
+	for (i = 1; i < MAX_XCP; i++) {
+		if (!adev->xcp_mgr->xcp[i].ddev)
+			break;
+
+		p_ddev = adev->xcp_mgr->xcp[i].ddev;
+		drm_dev_unplug(p_ddev);
+		p_ddev->render->dev = adev->xcp_mgr->xcp[i].rdev;
+		p_ddev->primary->dev = adev->xcp_mgr->xcp[i].pdev;
+		p_ddev->driver =  adev->xcp_mgr->xcp[i].driver;
+		p_ddev->vma_offset_manager = adev->xcp_mgr->xcp[i].vma_offset_manager;
+		amdgpu_xcp_drm_dev_free(p_ddev);
+	}
+}
+
+int amdgpu_xcp_open_device(struct amdgpu_device *adev,
+			   struct amdgpu_fpriv *fpriv,
+			   struct drm_file *file_priv)
+{
+	int i;
+
+	if (!adev->xcp_mgr)
+		return 0;
+
+	fpriv->xcp_id = AMDGPU_XCP_NO_PARTITION;
+	for (i = 0; i < MAX_XCP; ++i) {
+		if (!adev->xcp_mgr->xcp[i].ddev)
+			break;
+
+		if (file_priv->minor == adev->xcp_mgr->xcp[i].ddev->render) {
+			if (adev->xcp_mgr->xcp[i].valid == FALSE) {
+				dev_err(adev->dev, "renderD%d partition %d not valid!",
+						file_priv->minor->index, i);
+				return -ENOENT;
+			}
+			dev_dbg(adev->dev, "renderD%d partition %d opened!",
+					file_priv->minor->index, i);
+			fpriv->xcp_id = i;
+			break;
+		}
+	}
+
+	fpriv->vm.mem_id = fpriv->xcp_id == AMDGPU_XCP_NO_PARTITION ? -1 :
+				adev->xcp_mgr->xcp[fpriv->xcp_id].mem_id;
+	return 0;
+}
+
+void amdgpu_xcp_release_sched(struct amdgpu_device *adev,
+				  struct amdgpu_ctx_entity *entity)
+{
+	struct drm_gpu_scheduler *sched;
+	struct amdgpu_ring *ring;
+
+	if (!adev->xcp_mgr)
+		return;
+
+	sched = entity->entity.rq->sched;
+	if (drm_sched_wqueue_ready(sched)) {
+		ring = to_amdgpu_ring(entity->entity.rq->sched);
+		atomic_dec(&adev->xcp_mgr->xcp[ring->xcp_id].ref_cnt);
+	}
+}
+
+int amdgpu_xcp_select_scheds(struct amdgpu_device *adev,
+			     u32 hw_ip, u32 hw_prio,
+			     struct amdgpu_fpriv *fpriv,
+			     unsigned int *num_scheds,
+			     struct drm_gpu_scheduler ***scheds)
+{
+	u32 sel_xcp_id;
+	int i;
+	struct amdgpu_xcp_mgr *xcp_mgr = adev->xcp_mgr;
+
+	if (fpriv->xcp_id == AMDGPU_XCP_NO_PARTITION) {
+		u32 least_ref_cnt = ~0;
+
+		fpriv->xcp_id = 0;
+		for (i = 0; i < xcp_mgr->num_xcps; i++) {
+			u32 total_ref_cnt;
+
+			total_ref_cnt = atomic_read(&xcp_mgr->xcp[i].ref_cnt);
+			if (total_ref_cnt < least_ref_cnt) {
+				fpriv->xcp_id = i;
+				least_ref_cnt = total_ref_cnt;
+			}
+		}
+	}
+	sel_xcp_id = fpriv->xcp_id;
+
+	if (xcp_mgr->xcp[sel_xcp_id].gpu_sched[hw_ip][hw_prio].num_scheds) {
+		*num_scheds =
+			xcp_mgr->xcp[fpriv->xcp_id].gpu_sched[hw_ip][hw_prio].num_scheds;
+		*scheds =
+			xcp_mgr->xcp[fpriv->xcp_id].gpu_sched[hw_ip][hw_prio].sched;
+		atomic_inc(&adev->xcp_mgr->xcp[sel_xcp_id].ref_cnt);
+		dev_dbg(adev->dev, "Selected partition #%d", sel_xcp_id);
+	} else {
+		dev_err(adev->dev, "Failed to schedule partition #%d.", sel_xcp_id);
+		return -ENOENT;
+	}
+
+	return 0;
+}
+
+static void amdgpu_set_xcp_id(struct amdgpu_device *adev,
+			      uint32_t inst_idx,
+			      struct amdgpu_ring *ring)
+{
+	int xcp_id;
+	enum AMDGPU_XCP_IP_BLOCK ip_blk;
+	uint32_t inst_mask;
+
+	ring->xcp_id = AMDGPU_XCP_NO_PARTITION;
+	if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE)
+		adev->gfx.enforce_isolation[0].xcp_id = ring->xcp_id;
+	if ((adev->xcp_mgr->mode == AMDGPU_XCP_MODE_NONE) ||
+	    (ring->funcs->type == AMDGPU_RING_TYPE_CPER))
+		return;
+
+	inst_mask = 1 << inst_idx;
+
+	switch (ring->funcs->type) {
+	case AMDGPU_HW_IP_GFX:
+	case AMDGPU_RING_TYPE_COMPUTE:
+	case AMDGPU_RING_TYPE_KIQ:
+		ip_blk = AMDGPU_XCP_GFX;
+		break;
+	case AMDGPU_RING_TYPE_SDMA:
+		ip_blk = AMDGPU_XCP_SDMA;
+		break;
+	case AMDGPU_RING_TYPE_VCN_ENC:
+	case AMDGPU_RING_TYPE_VCN_JPEG:
+		ip_blk = AMDGPU_XCP_VCN;
+		break;
+	default:
+		dev_err(adev->dev, "Not support ring type %d!", ring->funcs->type);
+		return;
+	}
+
+	for (xcp_id = 0; xcp_id < adev->xcp_mgr->num_xcps; xcp_id++) {
+		if (adev->xcp_mgr->xcp[xcp_id].ip[ip_blk].inst_mask & inst_mask) {
+			ring->xcp_id = xcp_id;
+			dev_dbg(adev->dev, "ring:%s xcp_id :%u", ring->name,
+				ring->xcp_id);
+			if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE)
+				adev->gfx.enforce_isolation[xcp_id].xcp_id = xcp_id;
+			break;
+		}
+	}
+}
+
+static void amdgpu_xcp_gpu_sched_update(struct amdgpu_device *adev,
+					struct amdgpu_ring *ring,
+					unsigned int sel_xcp_id)
+{
+	unsigned int *num_gpu_sched;
+
+	num_gpu_sched = &adev->xcp_mgr->xcp[sel_xcp_id]
+			.gpu_sched[ring->funcs->type][ring->hw_prio].num_scheds;
+	adev->xcp_mgr->xcp[sel_xcp_id].gpu_sched[ring->funcs->type][ring->hw_prio]
+			.sched[(*num_gpu_sched)++] = &ring->sched;
+	dev_dbg(adev->dev, "%s :[%d] gpu_sched[%d][%d] = %d",
+		ring->name, sel_xcp_id, ring->funcs->type,
+		ring->hw_prio, *num_gpu_sched);
+}
+
+static int amdgpu_xcp_sched_list_update(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	int i;
+
+	for (i = 0; i < MAX_XCP; i++) {
+		atomic_set(&adev->xcp_mgr->xcp[i].ref_cnt, 0);
+		memset(adev->xcp_mgr->xcp[i].gpu_sched, 0, sizeof(adev->xcp_mgr->xcp->gpu_sched));
+	}
+
+	if (adev->xcp_mgr->mode == AMDGPU_XCP_MODE_NONE)
+		return 0;
+
+	for (i = 0; i < AMDGPU_MAX_RINGS; i++) {
+		ring = adev->rings[i];
+		if (!ring || !ring->sched.ready || ring->no_scheduler)
+			continue;
+
+		amdgpu_xcp_gpu_sched_update(adev, ring, ring->xcp_id);
+
+		/* VCN may be shared by two partitions under CPX MODE in certain
+		 * configs.
+		 */
+		if ((ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC ||
+		     ring->funcs->type == AMDGPU_RING_TYPE_VCN_JPEG) &&
+		    (adev->xcp_mgr->num_xcps > adev->vcn.num_vcn_inst))
+			amdgpu_xcp_gpu_sched_update(adev, ring, ring->xcp_id + 1);
+	}
+
+	return 0;
+}
+
+int amdgpu_xcp_update_partition_sched_list(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->num_rings; i++) {
+		struct amdgpu_ring *ring = adev->rings[i];
+
+		if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE ||
+			ring->funcs->type == AMDGPU_RING_TYPE_KIQ)
+			amdgpu_set_xcp_id(adev, ring->xcc_id, ring);
+		else
+			amdgpu_set_xcp_id(adev, ring->me, ring);
+	}
+
+	return amdgpu_xcp_sched_list_update(adev);
+}
+
+void amdgpu_xcp_update_supported_modes(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+
+	xcp_mgr->supp_xcp_modes = 0;
+
+	switch (NUM_XCC(adev->gfx.xcc_mask)) {
+	case 8:
+		xcp_mgr->supp_xcp_modes = BIT(AMDGPU_SPX_PARTITION_MODE) |
+					  BIT(AMDGPU_DPX_PARTITION_MODE) |
+					  BIT(AMDGPU_QPX_PARTITION_MODE) |
+					  BIT(AMDGPU_CPX_PARTITION_MODE);
+		break;
+	case 6:
+		xcp_mgr->supp_xcp_modes = BIT(AMDGPU_SPX_PARTITION_MODE) |
+					  BIT(AMDGPU_TPX_PARTITION_MODE) |
+					  BIT(AMDGPU_CPX_PARTITION_MODE);
+		break;
+	case 4:
+		xcp_mgr->supp_xcp_modes = BIT(AMDGPU_SPX_PARTITION_MODE) |
+					  BIT(AMDGPU_DPX_PARTITION_MODE) |
+					  BIT(AMDGPU_CPX_PARTITION_MODE);
+		break;
+	case 2:
+		xcp_mgr->supp_xcp_modes = BIT(AMDGPU_SPX_PARTITION_MODE) |
+					  BIT(AMDGPU_CPX_PARTITION_MODE);
+		break;
+	case 1:
+		xcp_mgr->supp_xcp_modes = BIT(AMDGPU_SPX_PARTITION_MODE) |
+					  BIT(AMDGPU_CPX_PARTITION_MODE);
+		break;
+
+	default:
+		break;
+	}
+}
+
+int amdgpu_xcp_pre_partition_switch(struct amdgpu_xcp_mgr *xcp_mgr, u32 flags)
+{
+	/* TODO:
+	 * Stop user queues and threads, and make sure GPU is empty of work.
+	 */
+
+	if (flags & AMDGPU_XCP_OPS_KFD)
+		amdgpu_amdkfd_device_fini_sw(xcp_mgr->adev);
+
+	return 0;
+}
+
+int amdgpu_xcp_post_partition_switch(struct amdgpu_xcp_mgr *xcp_mgr, u32 flags)
+{
+	int ret = 0;
+
+	if (flags & AMDGPU_XCP_OPS_KFD) {
+		amdgpu_amdkfd_device_probe(xcp_mgr->adev);
+		amdgpu_amdkfd_device_init(xcp_mgr->adev);
+		/* If KFD init failed, return failure */
+		if (!xcp_mgr->adev->kfd.init_complete)
+			ret = -EIO;
+	}
+
+	return ret;
+}
+
+/*====================== xcp sysfs - configuration ======================*/
+#define XCP_CFG_SYSFS_RES_ATTR_SHOW(_name)                         \
+	static ssize_t amdgpu_xcp_res_sysfs_##_name##_show(        \
+		struct amdgpu_xcp_res_details *xcp_res, char *buf) \
+	{                                                          \
+		return sysfs_emit(buf, "%d\n", xcp_res->_name);    \
+	}
+
+struct amdgpu_xcp_res_sysfs_attribute {
+	struct attribute attr;
+	ssize_t (*show)(struct amdgpu_xcp_res_details *xcp_res, char *buf);
+};
+
+#define XCP_CFG_SYSFS_RES_ATTR(_name)                                        \
+	struct amdgpu_xcp_res_sysfs_attribute xcp_res_sysfs_attr_##_name = { \
+		.attr = { .name = __stringify(_name), .mode = 0400 },        \
+		.show = amdgpu_xcp_res_sysfs_##_name##_show,                 \
+	}
+
+XCP_CFG_SYSFS_RES_ATTR_SHOW(num_inst)
+XCP_CFG_SYSFS_RES_ATTR(num_inst);
+XCP_CFG_SYSFS_RES_ATTR_SHOW(num_shared)
+XCP_CFG_SYSFS_RES_ATTR(num_shared);
+
+#define XCP_CFG_SYSFS_RES_ATTR_PTR(_name) xcp_res_sysfs_attr_##_name.attr
+
+static struct attribute *xcp_cfg_res_sysfs_attrs[] = {
+	&XCP_CFG_SYSFS_RES_ATTR_PTR(num_inst),
+	&XCP_CFG_SYSFS_RES_ATTR_PTR(num_shared), NULL
+};
+
+static const char *xcp_desc[] = {
+	[AMDGPU_SPX_PARTITION_MODE] = "SPX",
+	[AMDGPU_DPX_PARTITION_MODE] = "DPX",
+	[AMDGPU_TPX_PARTITION_MODE] = "TPX",
+	[AMDGPU_QPX_PARTITION_MODE] = "QPX",
+	[AMDGPU_CPX_PARTITION_MODE] = "CPX",
+};
+
+static const char *nps_desc[] = {
+	[UNKNOWN_MEMORY_PARTITION_MODE] = "UNKNOWN",
+	[AMDGPU_NPS1_PARTITION_MODE] = "NPS1",
+	[AMDGPU_NPS2_PARTITION_MODE] = "NPS2",
+	[AMDGPU_NPS3_PARTITION_MODE] = "NPS3",
+	[AMDGPU_NPS4_PARTITION_MODE] = "NPS4",
+	[AMDGPU_NPS6_PARTITION_MODE] = "NPS6",
+	[AMDGPU_NPS8_PARTITION_MODE] = "NPS8",
+};
+
+ATTRIBUTE_GROUPS(xcp_cfg_res_sysfs);
+
+#define to_xcp_attr(x) \
+	container_of(x, struct amdgpu_xcp_res_sysfs_attribute, attr)
+#define to_xcp_res(x) container_of(x, struct amdgpu_xcp_res_details, kobj)
+
+static ssize_t xcp_cfg_res_sysfs_attr_show(struct kobject *kobj,
+					   struct attribute *attr, char *buf)
+{
+	struct amdgpu_xcp_res_sysfs_attribute *attribute;
+	struct amdgpu_xcp_res_details *xcp_res;
+
+	attribute = to_xcp_attr(attr);
+	xcp_res = to_xcp_res(kobj);
+
+	if (!attribute->show)
+		return -EIO;
+
+	return attribute->show(xcp_res, buf);
+}
+
+static const struct sysfs_ops xcp_cfg_res_sysfs_ops = {
+	.show = xcp_cfg_res_sysfs_attr_show,
+};
+
+static const struct kobj_type xcp_cfg_res_sysfs_ktype = {
+	.sysfs_ops = &xcp_cfg_res_sysfs_ops,
+	.default_groups = xcp_cfg_res_sysfs_groups,
+};
+
+const char *xcp_res_names[] = {
+	[AMDGPU_XCP_RES_XCC] = "xcc",
+	[AMDGPU_XCP_RES_DMA] = "dma",
+	[AMDGPU_XCP_RES_DEC] = "dec",
+	[AMDGPU_XCP_RES_JPEG] = "jpeg",
+};
+
+static int amdgpu_xcp_get_res_info(struct amdgpu_xcp_mgr *xcp_mgr,
+				   int mode,
+				   struct amdgpu_xcp_cfg *xcp_cfg)
+{
+	if (xcp_mgr->funcs && xcp_mgr->funcs->get_xcp_res_info)
+		return xcp_mgr->funcs->get_xcp_res_info(xcp_mgr, mode, xcp_cfg);
+
+	return -EOPNOTSUPP;
+}
+
+#define to_xcp_cfg(x) container_of(x, struct amdgpu_xcp_cfg, kobj)
+static ssize_t supported_xcp_configs_show(struct kobject *kobj,
+					  struct kobj_attribute *attr, char *buf)
+{
+	struct amdgpu_xcp_cfg *xcp_cfg = to_xcp_cfg(kobj);
+	struct amdgpu_xcp_mgr *xcp_mgr = xcp_cfg->xcp_mgr;
+	int size = 0, mode;
+	char *sep = "";
+
+	if (!xcp_mgr || !xcp_mgr->supp_xcp_modes)
+		return sysfs_emit(buf, "Not supported\n");
+
+	for_each_inst(mode, xcp_mgr->supp_xcp_modes) {
+		size += sysfs_emit_at(buf, size, "%s%s", sep, xcp_desc[mode]);
+		sep = ", ";
+	}
+
+	size += sysfs_emit_at(buf, size, "\n");
+
+	return size;
+}
+
+static ssize_t supported_nps_configs_show(struct kobject *kobj,
+					  struct kobj_attribute *attr, char *buf)
+{
+	struct amdgpu_xcp_cfg *xcp_cfg = to_xcp_cfg(kobj);
+	int size = 0, mode;
+	char *sep = "";
+
+	if (!xcp_cfg || !xcp_cfg->compatible_nps_modes)
+		return sysfs_emit(buf, "Not supported\n");
+
+	for_each_inst(mode, xcp_cfg->compatible_nps_modes) {
+		size += sysfs_emit_at(buf, size, "%s%s", sep, nps_desc[mode]);
+		sep = ", ";
+	}
+
+	size += sysfs_emit_at(buf, size, "\n");
+
+	return size;
+}
+
+static ssize_t xcp_config_show(struct kobject *kobj,
+			       struct kobj_attribute *attr, char *buf)
+{
+	struct amdgpu_xcp_cfg *xcp_cfg = to_xcp_cfg(kobj);
+
+	return sysfs_emit(buf, "%s\n",
+			  amdgpu_gfx_compute_mode_desc(xcp_cfg->mode));
+}
+
+static ssize_t xcp_config_store(struct kobject *kobj,
+				struct kobj_attribute *attr,
+				const char *buf, size_t size)
+{
+	struct amdgpu_xcp_cfg *xcp_cfg = to_xcp_cfg(kobj);
+	int mode, r;
+
+	if (!strncasecmp("SPX", buf, strlen("SPX")))
+		mode = AMDGPU_SPX_PARTITION_MODE;
+	else if (!strncasecmp("DPX", buf, strlen("DPX")))
+		mode = AMDGPU_DPX_PARTITION_MODE;
+	else if (!strncasecmp("TPX", buf, strlen("TPX")))
+		mode = AMDGPU_TPX_PARTITION_MODE;
+	else if (!strncasecmp("QPX", buf, strlen("QPX")))
+		mode = AMDGPU_QPX_PARTITION_MODE;
+	else if (!strncasecmp("CPX", buf, strlen("CPX")))
+		mode = AMDGPU_CPX_PARTITION_MODE;
+	else
+		return -EINVAL;
+
+	r = amdgpu_xcp_get_res_info(xcp_cfg->xcp_mgr, mode, xcp_cfg);
+
+	if (r)
+		return r;
+
+	xcp_cfg->mode = mode;
+	return size;
+}
+
+static struct kobj_attribute xcp_cfg_sysfs_mode =
+	__ATTR_RW_MODE(xcp_config, 0644);
+
+static void xcp_cfg_sysfs_release(struct kobject *kobj)
+{
+	struct amdgpu_xcp_cfg *xcp_cfg = to_xcp_cfg(kobj);
+
+	kfree(xcp_cfg);
+}
+
+static const struct kobj_type xcp_cfg_sysfs_ktype = {
+	.release = xcp_cfg_sysfs_release,
+	.sysfs_ops = &kobj_sysfs_ops,
+};
+
+static struct kobj_attribute supp_part_sysfs_mode =
+	__ATTR_RO(supported_xcp_configs);
+
+static struct kobj_attribute supp_nps_sysfs_mode =
+	__ATTR_RO(supported_nps_configs);
+
+static const struct attribute *xcp_attrs[] = {
+	&supp_part_sysfs_mode.attr,
+	&xcp_cfg_sysfs_mode.attr,
+	NULL,
+};
+
+static void amdgpu_xcp_cfg_sysfs_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_xcp_res_details *xcp_res;
+	struct amdgpu_xcp_cfg *xcp_cfg;
+	int i, r, j, rid, mode;
+
+	if (!adev->xcp_mgr)
+		return;
+
+	xcp_cfg = kzalloc(sizeof(*xcp_cfg), GFP_KERNEL);
+	if (!xcp_cfg)
+		return;
+	xcp_cfg->xcp_mgr = adev->xcp_mgr;
+
+	r = kobject_init_and_add(&xcp_cfg->kobj, &xcp_cfg_sysfs_ktype,
+				 &adev->dev->kobj, "compute_partition_config");
+	if (r)
+		goto err1;
+
+	r = sysfs_create_files(&xcp_cfg->kobj, xcp_attrs);
+	if (r)
+		goto err1;
+
+	if (adev->gmc.supported_nps_modes != 0) {
+		r = sysfs_create_file(&xcp_cfg->kobj, &supp_nps_sysfs_mode.attr);
+		if (r) {
+			sysfs_remove_files(&xcp_cfg->kobj, xcp_attrs);
+			goto err1;
+		}
+	}
+
+	mode = (xcp_cfg->xcp_mgr->mode ==
+		AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE) ?
+		       AMDGPU_SPX_PARTITION_MODE :
+		       xcp_cfg->xcp_mgr->mode;
+	r = amdgpu_xcp_get_res_info(xcp_cfg->xcp_mgr, mode, xcp_cfg);
+	if (r) {
+		sysfs_remove_file(&xcp_cfg->kobj, &supp_nps_sysfs_mode.attr);
+		sysfs_remove_files(&xcp_cfg->kobj, xcp_attrs);
+		goto err1;
+	}
+
+	xcp_cfg->mode = mode;
+	for (i = 0; i < xcp_cfg->num_res; i++) {
+		xcp_res = &xcp_cfg->xcp_res[i];
+		rid = xcp_res->id;
+		r = kobject_init_and_add(&xcp_res->kobj,
+					 &xcp_cfg_res_sysfs_ktype,
+					 &xcp_cfg->kobj, "%s",
+					 xcp_res_names[rid]);
+		if (r)
+			goto err;
+	}
+
+	adev->xcp_mgr->xcp_cfg = xcp_cfg;
+	return;
+err:
+	for (j = 0; j < i; j++) {
+		xcp_res = &xcp_cfg->xcp_res[i];
+		kobject_put(&xcp_res->kobj);
+	}
+
+	sysfs_remove_file(&xcp_cfg->kobj, &supp_nps_sysfs_mode.attr);
+	sysfs_remove_files(&xcp_cfg->kobj, xcp_attrs);
+err1:
+	kobject_put(&xcp_cfg->kobj);
+}
+
+static void amdgpu_xcp_cfg_sysfs_fini(struct amdgpu_device *adev)
+{
+	struct amdgpu_xcp_res_details *xcp_res;
+	struct amdgpu_xcp_cfg *xcp_cfg;
+	int i;
+
+	if (!adev->xcp_mgr || !adev->xcp_mgr->xcp_cfg)
+		return;
+
+	xcp_cfg = adev->xcp_mgr->xcp_cfg;
+	for (i = 0; i < xcp_cfg->num_res; i++) {
+		xcp_res = &xcp_cfg->xcp_res[i];
+		kobject_put(&xcp_res->kobj);
+	}
+
+	sysfs_remove_file(&xcp_cfg->kobj, &supp_nps_sysfs_mode.attr);
+	sysfs_remove_files(&xcp_cfg->kobj, xcp_attrs);
+	kobject_put(&xcp_cfg->kobj);
+}
+
+/*====================== xcp sysfs - data entries ======================*/
+
+#define to_xcp(x) container_of(x, struct amdgpu_xcp, kobj)
+
+static ssize_t xcp_metrics_show(struct kobject *kobj,
+				struct kobj_attribute *attr, char *buf)
+{
+	struct amdgpu_xcp *xcp = to_xcp(kobj);
+	struct amdgpu_xcp_mgr *xcp_mgr;
+	ssize_t size;
+
+	xcp_mgr = xcp->xcp_mgr;
+	size = amdgpu_dpm_get_xcp_metrics(xcp_mgr->adev, xcp->id, NULL);
+	if (size <= 0)
+		return size;
+
+	if (size > PAGE_SIZE)
+		return -ENOSPC;
+
+	return amdgpu_dpm_get_xcp_metrics(xcp_mgr->adev, xcp->id, buf);
+}
+
+static umode_t amdgpu_xcp_attrs_is_visible(struct kobject *kobj,
+					   struct attribute *attr, int n)
+{
+	struct amdgpu_xcp *xcp = to_xcp(kobj);
+
+	if (!xcp || !xcp->valid)
+		return 0;
+
+	return attr->mode;
+}
+
+static struct kobj_attribute xcp_sysfs_metrics = __ATTR_RO(xcp_metrics);
+
+static struct attribute *amdgpu_xcp_attrs[] = {
+	&xcp_sysfs_metrics.attr,
+	NULL,
+};
+
+static const struct attribute_group amdgpu_xcp_attrs_group = {
+	.attrs = amdgpu_xcp_attrs,
+	.is_visible = amdgpu_xcp_attrs_is_visible
+};
+
+static const struct kobj_type xcp_sysfs_ktype = {
+	.sysfs_ops = &kobj_sysfs_ops,
+};
+
+static void amdgpu_xcp_sysfs_entries_fini(struct amdgpu_xcp_mgr *xcp_mgr, int n)
+{
+	struct amdgpu_xcp *xcp;
+
+	for (n--; n >= 0; n--) {
+		xcp = &xcp_mgr->xcp[n];
+		if (!xcp->ddev || !xcp->valid)
+			continue;
+		sysfs_remove_group(&xcp->kobj, &amdgpu_xcp_attrs_group);
+		kobject_put(&xcp->kobj);
+	}
+}
+
+static void amdgpu_xcp_sysfs_entries_init(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	struct amdgpu_xcp *xcp;
+	int i, r;
+
+	for (i = 0; i < MAX_XCP; i++) {
+		/* Redirect all IOCTLs to the primary device */
+		xcp = &xcp_mgr->xcp[i];
+		if (!xcp->ddev)
+			break;
+		r = kobject_init_and_add(&xcp->kobj, &xcp_sysfs_ktype,
+					 &xcp->ddev->dev->kobj, "xcp");
+		if (r)
+			goto out;
+
+		r = sysfs_create_group(&xcp->kobj, &amdgpu_xcp_attrs_group);
+		if (r)
+			goto out;
+	}
+
+	return;
+out:
+	kobject_put(&xcp->kobj);
+}
+
+static void amdgpu_xcp_sysfs_entries_update(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	struct amdgpu_xcp *xcp;
+	int i;
+
+	for (i = 0; i < MAX_XCP; i++) {
+		/* Redirect all IOCTLs to the primary device */
+		xcp = &xcp_mgr->xcp[i];
+		if (!xcp->ddev)
+			continue;
+		sysfs_update_group(&xcp->kobj, &amdgpu_xcp_attrs_group);
+	}
+
+	return;
+}
+
+void amdgpu_xcp_sysfs_init(struct amdgpu_device *adev)
+{
+	if (!adev->xcp_mgr)
+		return;
+
+	amdgpu_xcp_cfg_sysfs_init(adev);
+
+	return;
+}
+
+void amdgpu_xcp_sysfs_fini(struct amdgpu_device *adev)
+{
+	if (!adev->xcp_mgr)
+		return;
+	amdgpu_xcp_sysfs_entries_fini(adev->xcp_mgr, MAX_XCP);
+	amdgpu_xcp_cfg_sysfs_fini(adev);
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.h
new file mode 100644
index 000000000000..1928d9e224fc
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xcp.h
@@ -0,0 +1,217 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef AMDGPU_XCP_H
+#define AMDGPU_XCP_H
+
+#include <linux/pci.h>
+#include <linux/xarray.h>
+
+#include "amdgpu_ctx.h"
+
+#define MAX_XCP 8
+
+#define AMDGPU_XCP_MODE_NONE -1
+#define AMDGPU_XCP_MODE_TRANS -2
+
+#define AMDGPU_XCP_FL_NONE 0
+#define AMDGPU_XCP_FL_LOCKED (1 << 0)
+
+#define AMDGPU_XCP_NO_PARTITION (~0)
+
+#define AMDGPU_XCP_OPS_KFD	(1 << 0)
+
+struct amdgpu_fpriv;
+
+enum AMDGPU_XCP_IP_BLOCK {
+	AMDGPU_XCP_GFXHUB,
+	AMDGPU_XCP_GFX,
+	AMDGPU_XCP_SDMA,
+	AMDGPU_XCP_VCN,
+	AMDGPU_XCP_MAX_BLOCKS
+};
+
+enum AMDGPU_XCP_STATE {
+	AMDGPU_XCP_PREPARE_SUSPEND,
+	AMDGPU_XCP_SUSPEND,
+	AMDGPU_XCP_PREPARE_RESUME,
+	AMDGPU_XCP_RESUME,
+};
+
+enum amdgpu_xcp_res_id {
+	AMDGPU_XCP_RES_XCC,
+	AMDGPU_XCP_RES_DMA,
+	AMDGPU_XCP_RES_DEC,
+	AMDGPU_XCP_RES_JPEG,
+	AMDGPU_XCP_RES_MAX,
+};
+
+struct amdgpu_xcp_res_details {
+	enum amdgpu_xcp_res_id id;
+	u8 num_inst;
+	u8 num_shared;
+	struct kobject kobj;
+};
+
+struct amdgpu_xcp_cfg {
+	u8 mode;
+	struct amdgpu_xcp_res_details xcp_res[AMDGPU_XCP_RES_MAX];
+	u8 num_res;
+	struct amdgpu_xcp_mgr *xcp_mgr;
+	struct kobject kobj;
+	u16 compatible_nps_modes;
+};
+
+struct amdgpu_xcp_ip_funcs {
+	int (*prepare_suspend)(void *handle, uint32_t inst_mask);
+	int (*suspend)(void *handle, uint32_t inst_mask);
+	int (*prepare_resume)(void *handle, uint32_t inst_mask);
+	int (*resume)(void *handle, uint32_t inst_mask);
+};
+
+struct amdgpu_xcp_ip {
+	struct amdgpu_xcp_ip_funcs *ip_funcs;
+	uint32_t inst_mask;
+
+	enum AMDGPU_XCP_IP_BLOCK ip_id;
+	bool valid;
+};
+
+struct amdgpu_xcp {
+	struct amdgpu_xcp_ip ip[AMDGPU_XCP_MAX_BLOCKS];
+
+	uint8_t id;
+	uint8_t mem_id;
+	bool valid;
+	atomic_t	ref_cnt;
+	struct drm_device *ddev;
+	struct drm_device *rdev;
+	struct drm_device *pdev;
+	struct drm_driver *driver;
+	struct drm_vma_offset_manager *vma_offset_manager;
+	struct amdgpu_sched	gpu_sched[AMDGPU_HW_IP_NUM][AMDGPU_RING_PRIO_MAX];
+	struct amdgpu_xcp_mgr *xcp_mgr;
+	struct kobject kobj;
+	uint64_t unique_id;
+};
+
+struct amdgpu_xcp_mgr {
+	struct amdgpu_device *adev;
+	struct mutex xcp_lock;
+	struct amdgpu_xcp_mgr_funcs *funcs;
+
+	struct amdgpu_xcp xcp[MAX_XCP];
+	uint8_t num_xcps;
+	int8_t mode;
+
+	 /* Used to determine KFD memory size limits per XCP */
+	unsigned int num_xcp_per_mem_partition;
+	struct amdgpu_xcp_cfg *xcp_cfg;
+	uint32_t supp_xcp_modes;
+	uint32_t avail_xcp_modes;
+};
+
+struct amdgpu_xcp_mgr_funcs {
+	int (*switch_partition_mode)(struct amdgpu_xcp_mgr *xcp_mgr, int mode,
+				     int *num_xcps);
+	int (*query_partition_mode)(struct amdgpu_xcp_mgr *xcp_mgr);
+	int (*get_ip_details)(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id,
+			      enum AMDGPU_XCP_IP_BLOCK ip_id,
+			      struct amdgpu_xcp_ip *ip);
+	int (*get_xcp_mem_id)(struct amdgpu_xcp_mgr *xcp_mgr,
+			      struct amdgpu_xcp *xcp, uint8_t *mem_id);
+	int (*get_xcp_res_info)(struct amdgpu_xcp_mgr *xcp_mgr,
+				int mode,
+				struct amdgpu_xcp_cfg *xcp_cfg);
+	int (*prepare_suspend)(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id);
+	int (*suspend)(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id);
+	int (*prepare_resume)(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id);
+	int (*resume)(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id);
+};
+
+int amdgpu_xcp_prepare_suspend(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id);
+int amdgpu_xcp_suspend(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id);
+int amdgpu_xcp_prepare_resume(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id);
+int amdgpu_xcp_resume(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id);
+
+int amdgpu_xcp_mgr_init(struct amdgpu_device *adev, int init_mode,
+			int init_xcps, struct amdgpu_xcp_mgr_funcs *xcp_funcs);
+int amdgpu_xcp_init(struct amdgpu_xcp_mgr *xcp_mgr, int num_xcps, int mode);
+int amdgpu_xcp_query_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr, u32 flags);
+int amdgpu_xcp_switch_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr, int mode);
+int amdgpu_xcp_restore_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr);
+int amdgpu_xcp_get_partition(struct amdgpu_xcp_mgr *xcp_mgr,
+			     enum AMDGPU_XCP_IP_BLOCK ip, int instance);
+
+int amdgpu_xcp_get_inst_details(struct amdgpu_xcp *xcp,
+				enum AMDGPU_XCP_IP_BLOCK ip,
+				uint32_t *inst_mask);
+
+int amdgpu_xcp_dev_register(struct amdgpu_device *adev,
+				const struct pci_device_id *ent);
+void amdgpu_xcp_dev_unplug(struct amdgpu_device *adev);
+int amdgpu_xcp_open_device(struct amdgpu_device *adev,
+			   struct amdgpu_fpriv *fpriv,
+			   struct drm_file *file_priv);
+void amdgpu_xcp_release_sched(struct amdgpu_device *adev,
+			      struct amdgpu_ctx_entity *entity);
+int amdgpu_xcp_select_scheds(struct amdgpu_device *adev,
+			     u32 hw_ip, u32 hw_prio,
+			     struct amdgpu_fpriv *fpriv,
+			     unsigned int *num_scheds,
+			     struct drm_gpu_scheduler ***scheds);
+void amdgpu_xcp_update_supported_modes(struct amdgpu_xcp_mgr *xcp_mgr);
+int amdgpu_xcp_update_partition_sched_list(struct amdgpu_device *adev);
+int amdgpu_xcp_pre_partition_switch(struct amdgpu_xcp_mgr *xcp_mgr, u32 flags);
+int amdgpu_xcp_post_partition_switch(struct amdgpu_xcp_mgr *xcp_mgr, u32 flags);
+void amdgpu_xcp_sysfs_init(struct amdgpu_device *adev);
+void amdgpu_xcp_sysfs_fini(struct amdgpu_device *adev);
+
+static inline int amdgpu_xcp_get_num_xcp(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	if (!xcp_mgr)
+		return 1;
+	else
+		return xcp_mgr->num_xcps;
+}
+
+static inline struct amdgpu_xcp *
+amdgpu_get_next_xcp(struct amdgpu_xcp_mgr *xcp_mgr, int *from)
+{
+	if (!xcp_mgr)
+		return NULL;
+
+	while (*from < MAX_XCP) {
+		if (xcp_mgr->xcp[*from].valid)
+			return &xcp_mgr->xcp[*from];
+		++(*from);
+	}
+
+	return NULL;
+}
+
+#define for_each_xcp(xcp_mgr, xcp, i)                            \
+	for (i = 0, xcp = amdgpu_get_next_xcp(xcp_mgr, &i); xcp; \
+	     ++i, xcp = amdgpu_get_next_xcp(xcp_mgr, &i))
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c
index 439925477fb8..aad530c46a9f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c
@@ -40,6 +40,11 @@
 #define smnPCS_GOPX1_PCS_ERROR_STATUS    0x12200210
 #define smnPCS_GOPX1_PCS_ERROR_NONCORRECTABLE_MASK      0x12200218
 
+#define XGMI_STATE_DISABLE                      0xD1
+#define XGMI_STATE_LS0                          0x81
+#define XGMI_LINK_ACTIVE			1
+#define XGMI_LINK_INACTIVE			0
+
 static DEFINE_MUTEX(xgmi_mutex);
 
 #define AMDGPU_MAX_XGMI_DEVICE_PER_HIVE		4
@@ -103,6 +108,53 @@ static const int walf_pcs_err_noncorrectable_mask_reg_aldebaran[] = {
 	smnPCS_GOPX1_PCS_ERROR_NONCORRECTABLE_MASK + 0x100000
 };
 
+static const int xgmi3x16_pcs_err_status_reg_v6_4[] = {
+	smnPCS_XGMI3X16_PCS_ERROR_STATUS,
+	smnPCS_XGMI3X16_PCS_ERROR_STATUS + 0x100000
+};
+
+static const int xgmi3x16_pcs_err_noncorrectable_mask_reg_v6_4[] = {
+	smnPCS_XGMI3X16_PCS_ERROR_NONCORRECTABLE_MASK,
+	smnPCS_XGMI3X16_PCS_ERROR_NONCORRECTABLE_MASK + 0x100000
+};
+
+static const u64 xgmi_v6_4_0_mca_base_array[] = {
+	0x11a09200,
+	0x11b09200,
+};
+
+static const char *xgmi_v6_4_0_ras_error_code_ext[32] = {
+	[0x00] = "XGMI PCS DataLossErr",
+	[0x01] = "XGMI PCS TrainingErr",
+	[0x02] = "XGMI PCS FlowCtrlAckErr",
+	[0x03] = "XGMI PCS RxFifoUnderflowErr",
+	[0x04] = "XGMI PCS RxFifoOverflowErr",
+	[0x05] = "XGMI PCS CRCErr",
+	[0x06] = "XGMI PCS BERExceededErr",
+	[0x07] = "XGMI PCS TxMetaDataErr",
+	[0x08] = "XGMI PCS ReplayBufParityErr",
+	[0x09] = "XGMI PCS DataParityErr",
+	[0x0a] = "XGMI PCS ReplayFifoOverflowErr",
+	[0x0b] = "XGMI PCS ReplayFifoUnderflowErr",
+	[0x0c] = "XGMI PCS ElasticFifoOverflowErr",
+	[0x0d] = "XGMI PCS DeskewErr",
+	[0x0e] = "XGMI PCS FlowCtrlCRCErr",
+	[0x0f] = "XGMI PCS DataStartupLimitErr",
+	[0x10] = "XGMI PCS FCInitTimeoutErr",
+	[0x11] = "XGMI PCS RecoveryTimeoutErr",
+	[0x12] = "XGMI PCS ReadySerialTimeoutErr",
+	[0x13] = "XGMI PCS ReadySerialAttemptErr",
+	[0x14] = "XGMI PCS RecoveryAttemptErr",
+	[0x15] = "XGMI PCS RecoveryRelockAttemptErr",
+	[0x16] = "XGMI PCS ReplayAttemptErr",
+	[0x17] = "XGMI PCS SyncHdrErr",
+	[0x18] = "XGMI PCS TxReplayTimeoutErr",
+	[0x19] = "XGMI PCS RxReplayTimeoutErr",
+	[0x1a] = "XGMI PCS LinkSubTxTimeoutErr",
+	[0x1b] = "XGMI PCS LinkSubRxTimeoutErr",
+	[0x1c] = "XGMI PCS RxCMDPktErr",
+};
+
 static const struct amdgpu_pcs_ras_field xgmi_pcs_ras_fields[] = {
 	{"XGMI PCS DataLossErr",
 	 SOC15_REG_FIELD(XGMI0_PCS_GOPX16_PCS_ERROR_STATUS, DataLossErr)},
@@ -242,6 +294,82 @@ static const struct amdgpu_pcs_ras_field xgmi3x16_pcs_ras_fields[] = {
 	 SOC15_REG_FIELD(PCS_XGMI3X16_PCS_ERROR_STATUS, RxCMDPktErr)},
 };
 
+int amdgpu_xgmi_get_ext_link(struct amdgpu_device *adev, int link_num)
+{
+	int link_map_6_4_x[8] = { 0, 3, 1, 2, 7, 6, 4, 5 };
+
+	if (adev->gmc.xgmi.num_physical_nodes <= 1)
+		return -EINVAL;
+
+	switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) {
+	case IP_VERSION(6, 4, 0):
+	case IP_VERSION(6, 4, 1):
+		if (link_num < ARRAY_SIZE(link_map_6_4_x))
+			return link_map_6_4_x[link_num];
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return -EINVAL;
+}
+
+static u32 xgmi_v6_4_get_link_status(struct amdgpu_device *adev, int global_link_num)
+{
+	const u32 smn_xgmi_6_4_pcs_state_hist1[2] = { 0x11a00070, 0x11b00070 };
+	const u32 smn_xgmi_6_4_1_pcs_state_hist1[2] = { 0x12100070,
+							0x11b00070 };
+	u32 i, n;
+	u64 addr;
+
+	switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) {
+	case IP_VERSION(6, 4, 0):
+		n = ARRAY_SIZE(smn_xgmi_6_4_pcs_state_hist1);
+		addr = smn_xgmi_6_4_pcs_state_hist1[global_link_num % n];
+		break;
+	case IP_VERSION(6, 4, 1):
+		n = ARRAY_SIZE(smn_xgmi_6_4_1_pcs_state_hist1);
+		addr = smn_xgmi_6_4_1_pcs_state_hist1[global_link_num % n];
+		break;
+	default:
+		return U32_MAX;
+	}
+
+	i = global_link_num / n;
+
+	if (!(adev->aid_mask & BIT(i)))
+		return U32_MAX;
+
+	addr += adev->asic_funcs->encode_ext_smn_addressing(i);
+
+	return RREG32_PCIE_EXT(addr);
+}
+
+int amdgpu_get_xgmi_link_status(struct amdgpu_device *adev, int global_link_num)
+{
+	u32 xgmi_state_reg_val;
+
+	if (adev->gmc.xgmi.num_physical_nodes <= 1)
+		return -EINVAL;
+
+	switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) {
+	case IP_VERSION(6, 4, 0):
+	case IP_VERSION(6, 4, 1):
+		xgmi_state_reg_val = xgmi_v6_4_get_link_status(adev, global_link_num);
+		break;
+	default:
+		return -EOPNOTSUPP;
+	}
+
+	if ((xgmi_state_reg_val & 0xFF) == XGMI_STATE_DISABLE)
+		return -ENOLINK;
+
+	if ((xgmi_state_reg_val & 0xFF) == XGMI_STATE_LS0)
+		return XGMI_LINK_ACTIVE;
+
+	return XGMI_LINK_INACTIVE;
+}
+
 /**
  * DOC: AMDGPU XGMI Support
  *
@@ -325,6 +453,82 @@ static ssize_t amdgpu_xgmi_show_device_id(struct device *dev,
 
 }
 
+static ssize_t amdgpu_xgmi_show_physical_id(struct device *dev,
+				     struct device_attribute *attr,
+				     char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+
+	return sysfs_emit(buf, "%u\n", adev->gmc.xgmi.physical_node_id);
+
+}
+
+static ssize_t amdgpu_xgmi_show_num_hops(struct device *dev,
+					struct device_attribute *attr,
+					char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	struct psp_xgmi_topology_info *top = &adev->psp.xgmi_context.top_info;
+	int i;
+
+	for (i = 0; i < top->num_nodes; i++)
+		sprintf(buf + 3 * i, "%02x ", top->nodes[i].num_hops);
+
+	return sysfs_emit(buf, "%s\n", buf);
+}
+
+static ssize_t amdgpu_xgmi_show_num_links(struct device *dev,
+					struct device_attribute *attr,
+					char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	struct psp_xgmi_topology_info *top = &adev->psp.xgmi_context.top_info;
+	int i;
+
+	for (i = 0; i < top->num_nodes; i++)
+		sprintf(buf + 3 * i, "%02x ", top->nodes[i].num_links);
+
+	return sysfs_emit(buf, "%s\n", buf);
+}
+
+static ssize_t amdgpu_xgmi_show_connected_port_num(struct device *dev,
+					struct device_attribute *attr,
+					char *buf)
+{
+	struct drm_device *ddev = dev_get_drvdata(dev);
+	struct amdgpu_device *adev = drm_to_adev(ddev);
+	struct psp_xgmi_topology_info *top = &adev->psp.xgmi_context.top_info;
+	int i, j, size = 0;
+	int current_node;
+	/*
+	 * get the node id in the sysfs for the current socket and show
+	 * it in the port num info output in the sysfs for easy reading.
+	 * it is NOT the one retrieved from xgmi ta.
+	 */
+	for (i = 0; i < top->num_nodes; i++) {
+		if (top->nodes[i].node_id == adev->gmc.xgmi.node_id) {
+			current_node = i;
+			break;
+		}
+	}
+
+	if (i == top->num_nodes)
+		return -EINVAL;
+
+	for (i = 0; i < top->num_nodes; i++) {
+		for (j = 0; j < top->nodes[i].num_links; j++)
+			/* node id in sysfs starts from 1 rather than 0 so +1 here */
+			size += sysfs_emit_at(buf, size, "%02x:%02x ->  %02x:%02x\n", current_node + 1,
+					      top->nodes[i].port_num[j].src_xgmi_port_num, i + 1,
+					      top->nodes[i].port_num[j].dst_xgmi_port_num);
+	}
+
+	return size;
+}
+
 #define AMDGPU_XGMI_SET_FICAA(o)	((o) | 0x456801)
 static ssize_t amdgpu_xgmi_show_error(struct device *dev,
 				      struct device_attribute *attr,
@@ -360,7 +564,11 @@ static ssize_t amdgpu_xgmi_show_error(struct device *dev,
 
 
 static DEVICE_ATTR(xgmi_device_id, S_IRUGO, amdgpu_xgmi_show_device_id, NULL);
+static DEVICE_ATTR(xgmi_physical_id, 0444, amdgpu_xgmi_show_physical_id, NULL);
 static DEVICE_ATTR(xgmi_error, S_IRUGO, amdgpu_xgmi_show_error, NULL);
+static DEVICE_ATTR(xgmi_num_hops, S_IRUGO, amdgpu_xgmi_show_num_hops, NULL);
+static DEVICE_ATTR(xgmi_num_links, S_IRUGO, amdgpu_xgmi_show_num_links, NULL);
+static DEVICE_ATTR(xgmi_port_num, S_IRUGO, amdgpu_xgmi_show_connected_port_num, NULL);
 
 static int amdgpu_xgmi_sysfs_add_dev_info(struct amdgpu_device *adev,
 					 struct amdgpu_hive_info *hive)
@@ -375,11 +583,33 @@ static int amdgpu_xgmi_sysfs_add_dev_info(struct amdgpu_device *adev,
 		return ret;
 	}
 
+	ret = device_create_file(adev->dev, &dev_attr_xgmi_physical_id);
+	if (ret) {
+		dev_err(adev->dev, "XGMI: Failed to create device file xgmi_physical_id\n");
+		return ret;
+	}
+
 	/* Create xgmi error file */
 	ret = device_create_file(adev->dev, &dev_attr_xgmi_error);
 	if (ret)
 		pr_err("failed to create xgmi_error\n");
 
+	/* Create xgmi num hops file */
+	ret = device_create_file(adev->dev, &dev_attr_xgmi_num_hops);
+	if (ret)
+		pr_err("failed to create xgmi_num_hops\n");
+
+	/* Create xgmi num links file */
+	ret = device_create_file(adev->dev, &dev_attr_xgmi_num_links);
+	if (ret)
+		pr_err("failed to create xgmi_num_links\n");
+
+	/* Create xgmi port num file if supported */
+	if (adev->psp.xgmi_context.xgmi_ta_caps & EXTEND_PEER_LINK_INFO_CMD_FLAG) {
+		ret = device_create_file(adev->dev, &dev_attr_xgmi_port_num);
+		if (ret)
+			dev_err(adev->dev, "failed to create xgmi_port_num\n");
+	}
 
 	/* Create sysfs link to hive info folder on the first device */
 	if (hive->kobj.parent != (&adev->dev->kobj)) {
@@ -407,6 +637,12 @@ remove_link:
 
 remove_file:
 	device_remove_file(adev->dev, &dev_attr_xgmi_device_id);
+	device_remove_file(adev->dev, &dev_attr_xgmi_physical_id);
+	device_remove_file(adev->dev, &dev_attr_xgmi_error);
+	device_remove_file(adev->dev, &dev_attr_xgmi_num_hops);
+	device_remove_file(adev->dev, &dev_attr_xgmi_num_links);
+	if (adev->psp.xgmi_context.xgmi_ta_caps & EXTEND_PEER_LINK_INFO_CMD_FLAG)
+		device_remove_file(adev->dev, &dev_attr_xgmi_port_num);
 
 success:
 	return ret;
@@ -419,7 +655,12 @@ static void amdgpu_xgmi_sysfs_rem_dev_info(struct amdgpu_device *adev,
 	memset(node, 0, sizeof(node));
 
 	device_remove_file(adev->dev, &dev_attr_xgmi_device_id);
+	device_remove_file(adev->dev, &dev_attr_xgmi_physical_id);
 	device_remove_file(adev->dev, &dev_attr_xgmi_error);
+	device_remove_file(adev->dev, &dev_attr_xgmi_num_hops);
+	device_remove_file(adev->dev, &dev_attr_xgmi_num_links);
+	if (adev->psp.xgmi_context.xgmi_ta_caps & EXTEND_PEER_LINK_INFO_CMD_FLAG)
+		device_remove_file(adev->dev, &dev_attr_xgmi_port_num);
 
 	if (hive->kobj.parent != (&adev->dev->kobj))
 		sysfs_remove_link(&adev->dev->kobj,"xgmi_hive_info");
@@ -454,6 +695,7 @@ struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev)
 	hive = kzalloc(sizeof(*hive), GFP_KERNEL);
 	if (!hive) {
 		dev_err(adev->dev, "XGMI: allocation failed\n");
+		ret = -ENOMEM;
 		hive = NULL;
 		goto pro_end;
 	}
@@ -506,6 +748,7 @@ struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev)
 	task_barrier_init(&hive->tb);
 	hive->pstate = AMDGPU_XGMI_PSTATE_UNKNOWN;
 	hive->hi_req_gpu = NULL;
+	atomic_set(&hive->requested_nps_mode, UNKNOWN_MEMORY_PARTITION_MODE);
 
 	/*
 	 * hive pstate on boot is high in vega20 so we have to go to low
@@ -615,28 +858,88 @@ int amdgpu_xgmi_update_topology(struct amdgpu_hive_info *hive, struct amdgpu_dev
  * num_hops[2:0] = number of hops
  */
 int amdgpu_xgmi_get_hops_count(struct amdgpu_device *adev,
-		struct amdgpu_device *peer_adev)
+			       struct amdgpu_device *peer_adev)
 {
 	struct psp_xgmi_topology_info *top = &adev->psp.xgmi_context.top_info;
 	uint8_t num_hops_mask = 0x7;
 	int i;
 
+	if (!adev->gmc.xgmi.supported)
+		return 0;
+
 	for (i = 0 ; i < top->num_nodes; ++i)
 		if (top->nodes[i].node_id == peer_adev->gmc.xgmi.node_id)
 			return top->nodes[i].num_hops & num_hops_mask;
-	return	-EINVAL;
+
+	dev_err(adev->dev, "Failed to get xgmi hops count for peer %d.\n",
+		peer_adev->gmc.xgmi.physical_node_id);
+
+	return 0;
 }
 
-int amdgpu_xgmi_get_num_links(struct amdgpu_device *adev,
-		struct amdgpu_device *peer_adev)
+int amdgpu_xgmi_get_bandwidth(struct amdgpu_device *adev, struct amdgpu_device *peer_adev,
+			      enum amdgpu_xgmi_bw_mode bw_mode, enum amdgpu_xgmi_bw_unit bw_unit,
+			      uint32_t *min_bw, uint32_t *max_bw)
+{
+	bool peer_mode = bw_mode == AMDGPU_XGMI_BW_MODE_PER_PEER;
+	int unit_scale = bw_unit == AMDGPU_XGMI_BW_UNIT_MBYTES ? 1000 : 1;
+	int num_lanes = adev->gmc.xgmi.max_width;
+	int speed = adev->gmc.xgmi.max_speed;
+	int num_links = !peer_mode ? 1 : -1;
+
+	if (!(min_bw && max_bw))
+		return -EINVAL;
+
+	*min_bw = 0;
+	*max_bw = 0;
+
+	if (!adev->gmc.xgmi.supported)
+		return -ENODATA;
+
+	if (peer_mode && !peer_adev)
+		return -EINVAL;
+
+	if (peer_mode) {
+		struct psp_xgmi_topology_info *top = &adev->psp.xgmi_context.top_info;
+		int i;
+
+		for (i = 0 ; i < top->num_nodes; ++i) {
+			if (top->nodes[i].node_id != peer_adev->gmc.xgmi.node_id)
+				continue;
+
+			num_links =  top->nodes[i].num_links;
+			break;
+		}
+	}
+
+	if (num_links == -1) {
+		dev_err(adev->dev, "Failed to get number of xgmi links for peer %d.\n",
+			peer_adev->gmc.xgmi.physical_node_id);
+	} else if (num_links) {
+		int per_link_bw = (speed * num_lanes * unit_scale)/BITS_PER_BYTE;
+
+		*min_bw = per_link_bw;
+		*max_bw = num_links * per_link_bw;
+	}
+
+	return 0;
+}
+
+bool amdgpu_xgmi_get_is_sharing_enabled(struct amdgpu_device *adev,
+					struct amdgpu_device *peer_adev)
 {
 	struct psp_xgmi_topology_info *top = &adev->psp.xgmi_context.top_info;
 	int i;
 
+	/* Sharing should always be enabled for non-SRIOV. */
+	if (!amdgpu_sriov_vf(adev))
+		return true;
+
 	for (i = 0 ; i < top->num_nodes; ++i)
 		if (top->nodes[i].node_id == peer_adev->gmc.xgmi.node_id)
-			return top->nodes[i].num_links;
-	return	-EINVAL;
+			return !!top->nodes[i].is_sharing_enabled;
+
+	return false;
 }
 
 /*
@@ -677,8 +980,7 @@ int amdgpu_xgmi_add_device(struct amdgpu_device *adev)
 	if (!adev->gmc.xgmi.supported)
 		return 0;
 
-	if (!adev->gmc.xgmi.pending_reset &&
-	    amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_PSP)) {
+	if (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_PSP)) {
 		ret = psp_xgmi_initialize(&adev->psp, false, true);
 		if (ret) {
 			dev_err(adev->dev,
@@ -724,8 +1026,7 @@ int amdgpu_xgmi_add_device(struct amdgpu_device *adev)
 
 	task_barrier_add_task(&hive->tb);
 
-	if (!adev->gmc.xgmi.pending_reset &&
-	    amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_PSP)) {
+	if (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_PSP)) {
 		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
 			/* update node list for other device in the hive */
 			if (tmp_adev != adev) {
@@ -739,18 +1040,33 @@ int amdgpu_xgmi_add_device(struct amdgpu_device *adev)
 				goto exit_unlock;
 		}
 
-		/* get latest topology info for each device from psp */
-		list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
-			ret = psp_xgmi_get_topology_info(&tmp_adev->psp, count,
-					&tmp_adev->psp.xgmi_context.top_info, false);
+		if (amdgpu_sriov_vf(adev) &&
+			adev->psp.xgmi_context.xgmi_ta_caps & EXTEND_PEER_LINK_INFO_CMD_FLAG) {
+			/* only get topology for VF being init if it can support full duplex */
+			ret = psp_xgmi_get_topology_info(&adev->psp, count,
+						&adev->psp.xgmi_context.top_info, false);
 			if (ret) {
-				dev_err(tmp_adev->dev,
+				dev_err(adev->dev,
 					"XGMI: Get topology failure on device %llx, hive %llx, ret %d",
-					tmp_adev->gmc.xgmi.node_id,
-					tmp_adev->gmc.xgmi.hive_id, ret);
-				/* To do : continue with some node failed or disable the whole hive */
+					adev->gmc.xgmi.node_id,
+					adev->gmc.xgmi.hive_id, ret);
+				/* To do: continue with some node failed or disable the whole hive*/
 				goto exit_unlock;
 			}
+		} else {
+			/* get latest topology info for each device from psp */
+			list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
+				ret = psp_xgmi_get_topology_info(&tmp_adev->psp, count,
+					&tmp_adev->psp.xgmi_context.top_info, false);
+				if (ret) {
+					dev_err(tmp_adev->dev,
+						"XGMI: Get topology failure on device %llx, hive %llx, ret %d",
+						tmp_adev->gmc.xgmi.node_id,
+						tmp_adev->gmc.xgmi.hive_id, ret);
+					/* To do : continue with some node failed or disable the whole hive */
+					goto exit_unlock;
+				}
+			}
 		}
 
 		/* get topology again for hives that support extended data */
@@ -782,7 +1098,7 @@ int amdgpu_xgmi_add_device(struct amdgpu_device *adev)
 		}
 	}
 
-	if (!ret && !adev->gmc.xgmi.pending_reset)
+	if (!ret)
 		ret = amdgpu_xgmi_sysfs_add_dev_info(adev, hive);
 
 exit_unlock:
@@ -835,15 +1151,91 @@ int amdgpu_xgmi_remove_device(struct amdgpu_device *adev)
 	return 0;
 }
 
+static int xgmi_v6_4_0_aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank,
+				       enum aca_smu_type type, void *data)
+{
+	struct amdgpu_device *adev = handle->adev;
+	struct aca_bank_info info;
+	const char *error_str;
+	u64 status, count;
+	int ret, ext_error_code;
+
+	ret = aca_bank_info_decode(bank, &info);
+	if (ret)
+		return ret;
+
+	status = bank->regs[ACA_REG_IDX_STATUS];
+	ext_error_code = ACA_REG__STATUS__ERRORCODEEXT(status);
+
+	error_str = ext_error_code < ARRAY_SIZE(xgmi_v6_4_0_ras_error_code_ext) ?
+		xgmi_v6_4_0_ras_error_code_ext[ext_error_code] : NULL;
+	if (error_str)
+		dev_info(adev->dev, "%s detected\n", error_str);
+
+	count = ACA_REG__MISC0__ERRCNT(bank->regs[ACA_REG_IDX_MISC0]);
+
+	switch (type) {
+	case ACA_SMU_TYPE_UE:
+		if (ext_error_code != 0 && ext_error_code != 9)
+			count = 0ULL;
+
+		bank->aca_err_type = ACA_ERROR_TYPE_UE;
+		ret = aca_error_cache_log_bank_error(handle, &info, ACA_ERROR_TYPE_UE, count);
+		break;
+	case ACA_SMU_TYPE_CE:
+		count = ext_error_code == 6 ? count : 0ULL;
+		bank->aca_err_type = ACA_ERROR_TYPE_CE;
+		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type, count);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return ret;
+}
+
+static const struct aca_bank_ops xgmi_v6_4_0_aca_bank_ops = {
+	.aca_bank_parser = xgmi_v6_4_0_aca_bank_parser,
+};
+
+static const struct aca_info xgmi_v6_4_0_aca_info = {
+	.hwip = ACA_HWIP_TYPE_PCS_XGMI,
+	.mask = ACA_ERROR_UE_MASK | ACA_ERROR_CE_MASK,
+	.bank_ops = &xgmi_v6_4_0_aca_bank_ops,
+};
+
 static int amdgpu_xgmi_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
 {
+	int r;
+
 	if (!adev->gmc.xgmi.supported ||
 	    adev->gmc.xgmi.num_physical_nodes == 0)
 		return 0;
 
-	adev->gmc.xgmi.ras->ras_block.hw_ops->reset_ras_error_count(adev);
+	amdgpu_ras_reset_error_count(adev, AMDGPU_RAS_BLOCK__XGMI_WAFL);
+
+	r = amdgpu_ras_block_late_init(adev, ras_block);
+	if (r)
+		return r;
+
+	switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) {
+	case IP_VERSION(6, 4, 0):
+	case IP_VERSION(6, 4, 1):
+		r = amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__XGMI_WAFL,
+					&xgmi_v6_4_0_aca_info, NULL);
+		if (r)
+			goto late_fini;
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+
+late_fini:
+	amdgpu_ras_block_late_fini(adev, ras_block);
 
-	return amdgpu_ras_block_late_init(adev, ras_block);
+	return r;
 }
 
 uint64_t amdgpu_xgmi_get_relative_phy_addr(struct amdgpu_device *adev,
@@ -859,7 +1251,7 @@ static void pcs_clear_status(struct amdgpu_device *adev, uint32_t pcs_status_reg
 	WREG32_PCIE(pcs_status_reg, 0);
 }
 
-static void amdgpu_xgmi_reset_ras_error_count(struct amdgpu_device *adev)
+static void amdgpu_xgmi_legacy_reset_ras_error_count(struct amdgpu_device *adev)
 {
 	uint32_t i;
 
@@ -885,6 +1277,51 @@ static void amdgpu_xgmi_reset_ras_error_count(struct amdgpu_device *adev)
 	default:
 		break;
 	}
+
+	switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) {
+	case IP_VERSION(6, 4, 0):
+	case IP_VERSION(6, 4, 1):
+		for (i = 0; i < ARRAY_SIZE(xgmi3x16_pcs_err_status_reg_v6_4); i++)
+			pcs_clear_status(adev,
+					xgmi3x16_pcs_err_status_reg_v6_4[i]);
+		break;
+	default:
+		break;
+	}
+}
+
+static void __xgmi_v6_4_0_reset_error_count(struct amdgpu_device *adev, int xgmi_inst, u64 mca_base)
+{
+	WREG64_MCA(xgmi_inst, mca_base, ACA_REG_IDX_STATUS, 0ULL);
+}
+
+static void xgmi_v6_4_0_reset_error_count(struct amdgpu_device *adev, int xgmi_inst)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(xgmi_v6_4_0_mca_base_array); i++)
+		__xgmi_v6_4_0_reset_error_count(adev, xgmi_inst, xgmi_v6_4_0_mca_base_array[i]);
+}
+
+static void xgmi_v6_4_0_reset_ras_error_count(struct amdgpu_device *adev)
+{
+	int i;
+
+	for_each_inst(i, adev->aid_mask)
+		xgmi_v6_4_0_reset_error_count(adev, i);
+}
+
+static void amdgpu_xgmi_reset_ras_error_count(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) {
+	case IP_VERSION(6, 4, 0):
+	case IP_VERSION(6, 4, 1):
+		xgmi_v6_4_0_reset_ras_error_count(adev);
+		break;
+	default:
+		amdgpu_xgmi_legacy_reset_ras_error_count(adev);
+		break;
+	}
 }
 
 static int amdgpu_xgmi_query_pcs_error_status(struct amdgpu_device *adev,
@@ -901,7 +1338,12 @@ static int amdgpu_xgmi_query_pcs_error_status(struct amdgpu_device *adev,
 	uint32_t field_array_size = 0;
 
 	if (is_xgmi_pcs) {
-		if (adev->ip_versions[XGMI_HWIP][0] == IP_VERSION(6, 1, 0)) {
+		if (amdgpu_ip_version(adev, XGMI_HWIP, 0) ==
+		    IP_VERSION(6, 1, 0) ||
+		    amdgpu_ip_version(adev, XGMI_HWIP, 0) ==
+		    IP_VERSION(6, 4, 0) ||
+		    amdgpu_ip_version(adev, XGMI_HWIP, 0) ==
+		    IP_VERSION(6, 4, 1)) {
 			pcs_ras_fields = &xgmi3x16_pcs_ras_fields[0];
 			field_array_size = ARRAY_SIZE(xgmi3x16_pcs_ras_fields);
 		} else {
@@ -935,11 +1377,11 @@ static int amdgpu_xgmi_query_pcs_error_status(struct amdgpu_device *adev,
 	return 0;
 }
 
-static void amdgpu_xgmi_query_ras_error_count(struct amdgpu_device *adev,
-					     void *ras_error_status)
+static void amdgpu_xgmi_legacy_query_ras_error_count(struct amdgpu_device *adev,
+						     void *ras_error_status)
 {
 	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
-	int i;
+	int i, supported = 1;
 	uint32_t data, mask_data = 0;
 	uint32_t ue_cnt = 0, ce_cnt = 0;
 
@@ -1003,41 +1445,146 @@ static void amdgpu_xgmi_query_ras_error_count(struct amdgpu_device *adev,
 		}
 		break;
 	default:
-		dev_warn(adev->dev, "XGMI RAS error query not supported");
+		supported = 0;
+		break;
+	}
+
+	switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) {
+	case IP_VERSION(6, 4, 0):
+	case IP_VERSION(6, 4, 1):
+		/* check xgmi3x16 pcs error */
+		for (i = 0; i < ARRAY_SIZE(xgmi3x16_pcs_err_status_reg_v6_4); i++) {
+			data = RREG32_PCIE(xgmi3x16_pcs_err_status_reg_v6_4[i]);
+			mask_data =
+				RREG32_PCIE(xgmi3x16_pcs_err_noncorrectable_mask_reg_v6_4[i]);
+			if (data)
+				amdgpu_xgmi_query_pcs_error_status(adev, data,
+						mask_data, &ue_cnt, &ce_cnt, true, true);
+		}
+		break;
+	default:
+		if (!supported)
+			dev_warn(adev->dev, "XGMI RAS error query not supported");
 		break;
 	}
 
-	adev->gmc.xgmi.ras->ras_block.hw_ops->reset_ras_error_count(adev);
+	amdgpu_ras_reset_error_count(adev, AMDGPU_RAS_BLOCK__XGMI_WAFL);
 
 	err_data->ue_count += ue_cnt;
 	err_data->ce_count += ce_cnt;
 }
 
+static enum aca_error_type xgmi_v6_4_0_pcs_mca_get_error_type(struct amdgpu_device *adev, u64 status)
+{
+	const char *error_str;
+	int ext_error_code;
+
+	ext_error_code = ACA_REG__STATUS__ERRORCODEEXT(status);
+
+	error_str = ext_error_code < ARRAY_SIZE(xgmi_v6_4_0_ras_error_code_ext) ?
+		xgmi_v6_4_0_ras_error_code_ext[ext_error_code] : NULL;
+	if (error_str)
+		dev_info(adev->dev, "%s detected\n", error_str);
+
+	switch (ext_error_code) {
+	case 0:
+		return ACA_ERROR_TYPE_UE;
+	case 6:
+		return ACA_ERROR_TYPE_CE;
+	default:
+		return -EINVAL;
+	}
+
+	return -EINVAL;
+}
+
+static void __xgmi_v6_4_0_query_error_count(struct amdgpu_device *adev, struct amdgpu_smuio_mcm_config_info *mcm_info,
+					    u64 mca_base, struct ras_err_data *err_data)
+{
+	int xgmi_inst = mcm_info->die_id;
+	u64 status = 0;
+
+	status = RREG64_MCA(xgmi_inst, mca_base, ACA_REG_IDX_STATUS);
+	if (!ACA_REG__STATUS__VAL(status))
+		return;
+
+	switch (xgmi_v6_4_0_pcs_mca_get_error_type(adev, status)) {
+	case ACA_ERROR_TYPE_UE:
+		amdgpu_ras_error_statistic_ue_count(err_data, mcm_info, 1ULL);
+		break;
+	case ACA_ERROR_TYPE_CE:
+		amdgpu_ras_error_statistic_ce_count(err_data, mcm_info, 1ULL);
+		break;
+	default:
+		break;
+	}
+
+	WREG64_MCA(xgmi_inst, mca_base, ACA_REG_IDX_STATUS, 0ULL);
+}
+
+static void xgmi_v6_4_0_query_error_count(struct amdgpu_device *adev, int xgmi_inst, struct ras_err_data *err_data)
+{
+	struct amdgpu_smuio_mcm_config_info mcm_info = {
+		.socket_id = adev->smuio.funcs->get_socket_id(adev),
+		.die_id = xgmi_inst,
+	};
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(xgmi_v6_4_0_mca_base_array); i++)
+		__xgmi_v6_4_0_query_error_count(adev, &mcm_info, xgmi_v6_4_0_mca_base_array[i], err_data);
+}
+
+static void xgmi_v6_4_0_query_ras_error_count(struct amdgpu_device *adev, void *ras_error_status)
+{
+	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
+	int i;
+
+	for_each_inst(i, adev->aid_mask)
+		xgmi_v6_4_0_query_error_count(adev, i, err_data);
+}
+
+static void amdgpu_xgmi_query_ras_error_count(struct amdgpu_device *adev,
+					      void *ras_error_status)
+{
+	switch (amdgpu_ip_version(adev, XGMI_HWIP, 0)) {
+	case IP_VERSION(6, 4, 0):
+	case IP_VERSION(6, 4, 1):
+		xgmi_v6_4_0_query_ras_error_count(adev, ras_error_status);
+		break;
+	default:
+		amdgpu_xgmi_legacy_query_ras_error_count(adev, ras_error_status);
+		break;
+	}
+}
+
 /* Trigger XGMI/WAFL error */
-static int amdgpu_ras_error_inject_xgmi(struct amdgpu_device *adev,  void *inject_if)
+static int amdgpu_ras_error_inject_xgmi(struct amdgpu_device *adev,
+			void *inject_if, uint32_t instance_mask)
 {
-	int ret = 0;
+	int ret1, ret2;
 	struct ta_ras_trigger_error_input *block_info =
 				(struct ta_ras_trigger_error_input *)inject_if;
 
 	if (amdgpu_dpm_set_df_cstate(adev, DF_CSTATE_DISALLOW))
 		dev_warn(adev->dev, "Failed to disallow df cstate");
 
-	if (amdgpu_dpm_allow_xgmi_power_down(adev, false))
+	ret1 = amdgpu_dpm_set_pm_policy(adev, PP_PM_POLICY_XGMI_PLPD, XGMI_PLPD_DISALLOW);
+	if (ret1 && ret1 != -EOPNOTSUPP)
 		dev_warn(adev->dev, "Failed to disallow XGMI power down");
 
-	ret = psp_ras_trigger_error(&adev->psp, block_info);
+	ret2 = psp_ras_trigger_error(&adev->psp, block_info, instance_mask);
 
 	if (amdgpu_ras_intr_triggered())
-		return ret;
+		return ret2;
 
-	if (amdgpu_dpm_allow_xgmi_power_down(adev, true))
+	ret1 = amdgpu_dpm_set_pm_policy(adev, PP_PM_POLICY_XGMI_PLPD, XGMI_PLPD_DEFAULT);
+	if (ret1 && ret1 != -EOPNOTSUPP)
 		dev_warn(adev->dev, "Failed to allow XGMI power down");
 
 	if (amdgpu_dpm_set_df_cstate(adev, DF_CSTATE_ALLOW))
 		dev_warn(adev->dev, "Failed to allow df cstate");
 
-	return ret;
+	return ret2;
 }
 
 struct amdgpu_ras_block_hw_ops  xgmi_ras_hw_ops = {
@@ -1075,3 +1622,157 @@ int amdgpu_xgmi_ras_sw_init(struct amdgpu_device *adev)
 
 	return 0;
 }
+
+static void amdgpu_xgmi_reset_on_init_work(struct work_struct *work)
+{
+	struct amdgpu_hive_info *hive =
+		container_of(work, struct amdgpu_hive_info, reset_on_init_work);
+	struct amdgpu_reset_context reset_context;
+	struct amdgpu_device *tmp_adev;
+	struct list_head device_list;
+	int r;
+
+	mutex_lock(&hive->hive_lock);
+
+	INIT_LIST_HEAD(&device_list);
+	list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head)
+		list_add_tail(&tmp_adev->reset_list, &device_list);
+
+	tmp_adev = list_first_entry(&device_list, struct amdgpu_device,
+				    reset_list);
+	amdgpu_device_lock_reset_domain(tmp_adev->reset_domain);
+
+	reset_context.method = AMD_RESET_METHOD_ON_INIT;
+	reset_context.reset_req_dev = tmp_adev;
+	reset_context.hive = hive;
+	reset_context.reset_device_list = &device_list;
+	set_bit(AMDGPU_NEED_FULL_RESET, &reset_context.flags);
+	set_bit(AMDGPU_SKIP_COREDUMP, &reset_context.flags);
+
+	amdgpu_reset_do_xgmi_reset_on_init(&reset_context);
+	mutex_unlock(&hive->hive_lock);
+	amdgpu_device_unlock_reset_domain(tmp_adev->reset_domain);
+
+	list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
+		r = amdgpu_ras_init_badpage_info(tmp_adev);
+		if (r && r != -EHWPOISON)
+			dev_err(tmp_adev->dev,
+				"error during bad page data initialization");
+	}
+}
+
+static void amdgpu_xgmi_schedule_reset_on_init(struct amdgpu_hive_info *hive)
+{
+	INIT_WORK(&hive->reset_on_init_work, amdgpu_xgmi_reset_on_init_work);
+	amdgpu_reset_domain_schedule(hive->reset_domain,
+				     &hive->reset_on_init_work);
+}
+
+int amdgpu_xgmi_reset_on_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_hive_info *hive;
+	bool reset_scheduled;
+	int num_devs;
+
+	hive = amdgpu_get_xgmi_hive(adev);
+	if (!hive)
+		return -EINVAL;
+
+	mutex_lock(&hive->hive_lock);
+	num_devs = atomic_read(&hive->number_devices);
+	reset_scheduled = false;
+	if (num_devs == adev->gmc.xgmi.num_physical_nodes) {
+		amdgpu_xgmi_schedule_reset_on_init(hive);
+		reset_scheduled = true;
+	}
+
+	mutex_unlock(&hive->hive_lock);
+	amdgpu_put_xgmi_hive(hive);
+
+	if (reset_scheduled)
+		flush_work(&hive->reset_on_init_work);
+
+	return 0;
+}
+
+int amdgpu_xgmi_request_nps_change(struct amdgpu_device *adev,
+				   struct amdgpu_hive_info *hive,
+				   int req_nps_mode)
+{
+	struct amdgpu_device *tmp_adev;
+	int cur_nps_mode, r;
+
+	/* This is expected to be called only during unload of driver. The
+	 * request needs to be placed only once for all devices in the hive. If
+	 * one of them fail, revert the request for previous successful devices.
+	 * After placing the request, make hive mode as UNKNOWN so that other
+	 * devices don't request anymore.
+	 */
+	mutex_lock(&hive->hive_lock);
+	if (atomic_read(&hive->requested_nps_mode) ==
+	    UNKNOWN_MEMORY_PARTITION_MODE) {
+		dev_dbg(adev->dev, "Unexpected entry for hive NPS change");
+		mutex_unlock(&hive->hive_lock);
+		return 0;
+	}
+	list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) {
+		r = adev->gmc.gmc_funcs->request_mem_partition_mode(
+			tmp_adev, req_nps_mode);
+		if (r)
+			break;
+	}
+	if (r) {
+		/* Request back current mode if one of the requests failed */
+		cur_nps_mode =
+			adev->gmc.gmc_funcs->query_mem_partition_mode(tmp_adev);
+		list_for_each_entry_continue_reverse(
+			tmp_adev, &hive->device_list, gmc.xgmi.head)
+			adev->gmc.gmc_funcs->request_mem_partition_mode(
+				tmp_adev, cur_nps_mode);
+	}
+	/* Set to UNKNOWN so that other devices don't request anymore */
+	atomic_set(&hive->requested_nps_mode, UNKNOWN_MEMORY_PARTITION_MODE);
+	mutex_unlock(&hive->hive_lock);
+
+	return r;
+}
+
+bool amdgpu_xgmi_same_hive(struct amdgpu_device *adev,
+			   struct amdgpu_device *bo_adev)
+{
+	return (amdgpu_use_xgmi_p2p && adev != bo_adev &&
+		adev->gmc.xgmi.hive_id &&
+		adev->gmc.xgmi.hive_id == bo_adev->gmc.xgmi.hive_id);
+}
+
+void amdgpu_xgmi_early_init(struct amdgpu_device *adev)
+{
+	if (!adev->gmc.xgmi.supported)
+		return;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(9, 4, 0):
+	case IP_VERSION(9, 4, 1):
+	case IP_VERSION(9, 4, 2):
+		/* 25 GT/s */
+		adev->gmc.xgmi.max_speed = 25;
+		adev->gmc.xgmi.max_width = 16;
+		break;
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
+		/* 32 GT/s */
+		adev->gmc.xgmi.max_speed = 32;
+		adev->gmc.xgmi.max_width = 16;
+		break;
+	default:
+		break;
+	}
+}
+
+void amgpu_xgmi_set_max_speed_width(struct amdgpu_device *adev,
+				    uint16_t max_speed, uint8_t max_width)
+{
+	adev->gmc.xgmi.max_speed = max_speed;
+	adev->gmc.xgmi.max_width = max_width;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h
index 86fbf56938f4..5f36aff17e79 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.h
@@ -23,7 +23,6 @@
 #define __AMDGPU_XGMI_H__
 
 #include <drm/task_barrier.h>
-#include "amdgpu_psp.h"
 #include "amdgpu_ras.h"
 
 struct amdgpu_hive_info {
@@ -43,7 +42,10 @@ struct amdgpu_hive_info {
 	} pstate;
 
 	struct amdgpu_reset_domain *reset_domain;
-	uint32_t device_remove_count;
+	atomic_t ras_recovery;
+	struct ras_event_manager event_mgr;
+	struct work_struct reset_on_init_work;
+	atomic_t requested_nps_mode;
 };
 
 struct amdgpu_pcs_ras_field {
@@ -52,27 +54,80 @@ struct amdgpu_pcs_ras_field {
 	uint32_t pcs_err_shift;
 };
 
-extern struct amdgpu_xgmi_ras  xgmi_ras;
+/**
+ * Bandwidth range reporting comes in two modes.
+ *
+ * PER_LINK - range for any xgmi link
+ * PER_PEER - range of max of single xgmi link to max of multiple links based on source peer
+ */
+enum amdgpu_xgmi_bw_mode {
+	AMDGPU_XGMI_BW_MODE_PER_LINK = 0,
+	AMDGPU_XGMI_BW_MODE_PER_PEER
+};
+
+enum amdgpu_xgmi_bw_unit {
+	AMDGPU_XGMI_BW_UNIT_GBYTES = 0,
+	AMDGPU_XGMI_BW_UNIT_MBYTES
+};
+
+struct amdgpu_xgmi_ras {
+	struct amdgpu_ras_block_object ras_block;
+};
+extern struct amdgpu_xgmi_ras xgmi_ras;
+
+struct amdgpu_xgmi {
+	/* from psp */
+	u64 node_id;
+	u64 hive_id;
+	/* fixed per family */
+	u64 node_segment_size;
+	/* physical node (0-3) */
+	unsigned physical_node_id;
+	/* number of nodes (0-4) */
+	unsigned num_physical_nodes;
+	/* gpu list in the same hive */
+	struct list_head head;
+	bool supported;
+	struct ras_common_if *ras_if;
+	bool connected_to_cpu;
+	struct amdgpu_xgmi_ras *ras;
+	uint16_t max_speed;
+	uint8_t max_width;
+};
+
 struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev);
 void amdgpu_put_xgmi_hive(struct amdgpu_hive_info *hive);
 int amdgpu_xgmi_update_topology(struct amdgpu_hive_info *hive, struct amdgpu_device *adev);
 int amdgpu_xgmi_add_device(struct amdgpu_device *adev);
 int amdgpu_xgmi_remove_device(struct amdgpu_device *adev);
 int amdgpu_xgmi_set_pstate(struct amdgpu_device *adev, int pstate);
-int amdgpu_xgmi_get_hops_count(struct amdgpu_device *adev,
-		struct amdgpu_device *peer_adev);
-int amdgpu_xgmi_get_num_links(struct amdgpu_device *adev,
-		struct amdgpu_device *peer_adev);
+int amdgpu_xgmi_get_hops_count(struct amdgpu_device *adev, struct amdgpu_device *peer_adev);
+int amdgpu_xgmi_get_bandwidth(struct amdgpu_device *adev, struct amdgpu_device *peer_adev,
+			      enum amdgpu_xgmi_bw_mode bw_mode, enum amdgpu_xgmi_bw_unit bw_unit,
+			      uint32_t *min_bw, uint32_t *max_bw);
+bool amdgpu_xgmi_get_is_sharing_enabled(struct amdgpu_device *adev,
+					struct amdgpu_device *peer_adev);
 uint64_t amdgpu_xgmi_get_relative_phy_addr(struct amdgpu_device *adev,
 					   uint64_t addr);
-static inline bool amdgpu_xgmi_same_hive(struct amdgpu_device *adev,
-		struct amdgpu_device *bo_adev)
-{
-	return (amdgpu_use_xgmi_p2p &&
-		adev != bo_adev &&
-		adev->gmc.xgmi.hive_id &&
-		adev->gmc.xgmi.hive_id == bo_adev->gmc.xgmi.hive_id);
-}
+bool amdgpu_xgmi_same_hive(struct amdgpu_device *adev,
+			   struct amdgpu_device *bo_adev);
 int amdgpu_xgmi_ras_sw_init(struct amdgpu_device *adev);
+int amdgpu_xgmi_reset_on_init(struct amdgpu_device *adev);
+
+int amdgpu_xgmi_request_nps_change(struct amdgpu_device *adev,
+				   struct amdgpu_hive_info *hive,
+				   int req_nps_mode);
+int amdgpu_get_xgmi_link_status(struct amdgpu_device *adev,
+				int global_link_num);
+int amdgpu_xgmi_get_ext_link(struct amdgpu_device *adev, int link_num);
+
+void amdgpu_xgmi_early_init(struct amdgpu_device *adev);
+uint32_t amdgpu_xgmi_get_max_bandwidth(struct amdgpu_device *adev);
+
+void amgpu_xgmi_set_max_speed_width(struct amdgpu_device *adev,
+				    uint16_t max_speed, uint8_t max_width);
+
+/* Cleanup macro for use with __free(xgmi_put_hive) */
+DEFINE_FREE(xgmi_put_hive, struct amdgpu_hive_info *, if (_T) amdgpu_put_xgmi_hive(_T))
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h
index 24d42d24e6a0..3cdb1e0eca37 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgv_sriovmsg.h
@@ -23,22 +23,84 @@
 #ifndef AMDGV_SRIOV_MSG__H_
 #define AMDGV_SRIOV_MSG__H_
 
-/* unit in kilobytes */
-#define AMD_SRIOV_MSG_VBIOS_OFFSET	     0
-#define AMD_SRIOV_MSG_VBIOS_SIZE_KB	     64
-#define AMD_SRIOV_MSG_DATAEXCHANGE_OFFSET_KB AMD_SRIOV_MSG_VBIOS_SIZE_KB
-#define AMD_SRIOV_MSG_DATAEXCHANGE_SIZE_KB   4
+#define AMD_SRIOV_MSG_SIZE_KB                           1
 
 /*
- * layout
- * 0           64KB        65KB        66KB
- * |   VBIOS   |   PF2VF   |   VF2PF   |   Bad Page   | ...
- * |   64KB    |   1KB     |   1KB     |
+ * layout v1
+ * 0           64KB        65KB        66KB           68KB                   132KB
+ * |   VBIOS   |   PF2VF   |   VF2PF   |   Bad Page   | RAS Telemetry Region | ...
+ * |   64KB    |   1KB     |   1KB     |   2KB        | 64KB                 | ...
  */
-#define AMD_SRIOV_MSG_SIZE_KB                   1
-#define AMD_SRIOV_MSG_PF2VF_OFFSET_KB           AMD_SRIOV_MSG_DATAEXCHANGE_OFFSET_KB
-#define AMD_SRIOV_MSG_VF2PF_OFFSET_KB           (AMD_SRIOV_MSG_PF2VF_OFFSET_KB + AMD_SRIOV_MSG_SIZE_KB)
-#define AMD_SRIOV_MSG_BAD_PAGE_OFFSET_KB        (AMD_SRIOV_MSG_VF2PF_OFFSET_KB + AMD_SRIOV_MSG_SIZE_KB)
+
+/*
+ * layout v2 (offsets are dynamically allocated and the offsets below are examples)
+ * 0           1KB         64KB        65KB        66KB           68KB                   132KB
+ * |  INITD_H  |   VBIOS   |   PF2VF   |   VF2PF   |   Bad Page   | RAS Telemetry Region | ...
+ * |   1KB     |   64KB    |   1KB     |   1KB     |   2KB        | 64KB                 | ...
+ *
+ * Note: PF2VF + VF2PF + Bad Page = DataExchange region (allocated contiguously)
+ */
+
+/* v1 layout sizes */
+#define AMD_SRIOV_MSG_VBIOS_SIZE_KB_V1			64
+#define AMD_SRIOV_MSG_PF2VF_SIZE_KB_V1			1
+#define AMD_SRIOV_MSG_VF2PF_SIZE_KB_V1			1
+#define AMD_SRIOV_MSG_BAD_PAGE_SIZE_KB_V1		2
+#define AMD_SRIOV_MSG_RAS_TELEMETRY_SIZE_KB_V1		64
+#define AMD_SRIOV_MSG_DATAEXCHANGE_SIZE_KB_V1		\
+	(AMD_SRIOV_MSG_PF2VF_SIZE_KB_V1 + AMD_SRIOV_MSG_VF2PF_SIZE_KB_V1 + \
+	 AMD_SRIOV_MSG_BAD_PAGE_SIZE_KB_V1)
+
+/* v1 offsets */
+#define AMD_SRIOV_MSG_VBIOS_OFFSET_V1			0
+#define AMD_SRIOV_MSG_DATAEXCHANGE_OFFSET_KB_V1		AMD_SRIOV_MSG_VBIOS_SIZE_KB_V1
+#define AMD_SRIOV_MSG_TMR_OFFSET_KB			2048
+#define AMD_SRIOV_MSG_PF2VF_OFFSET_KB_V1		AMD_SRIOV_MSG_DATAEXCHANGE_OFFSET_KB_V1
+#define AMD_SRIOV_MSG_VF2PF_OFFSET_KB_V1		\
+	(AMD_SRIOV_MSG_PF2VF_OFFSET_KB_V1 + AMD_SRIOV_MSG_SIZE_KB)
+#define AMD_SRIOV_MSG_BAD_PAGE_OFFSET_KB_V1		\
+	(AMD_SRIOV_MSG_VF2PF_OFFSET_KB_V1 + AMD_SRIOV_MSG_SIZE_KB)
+#define AMD_SRIOV_MSG_RAS_TELEMETRY_OFFSET_KB_V1	\
+	(AMD_SRIOV_MSG_BAD_PAGE_OFFSET_KB_V1 + AMD_SRIOV_MSG_BAD_PAGE_SIZE_KB_V1)
+#define AMD_SRIOV_MSG_INIT_DATA_TOT_SIZE_KB_V1		\
+	(AMD_SRIOV_MSG_VBIOS_SIZE_KB_V1 + AMD_SRIOV_MSG_DATAEXCHANGE_SIZE_KB_V1 + \
+	 AMD_SRIOV_MSG_RAS_TELEMETRY_SIZE_KB_V1)
+
+enum amd_sriov_crit_region_version {
+	GPU_CRIT_REGION_V1 = 1,
+	GPU_CRIT_REGION_V2 = 2,
+};
+
+/* v2 layout offset enum (in order of allocation) */
+enum amd_sriov_msg_table_id_enum {
+	AMD_SRIOV_MSG_IPD_TABLE_ID = 0,
+	AMD_SRIOV_MSG_VBIOS_IMG_TABLE_ID,
+	AMD_SRIOV_MSG_RAS_TELEMETRY_TABLE_ID,
+	AMD_SRIOV_MSG_DATAEXCHANGE_TABLE_ID,
+	AMD_SRIOV_MSG_BAD_PAGE_INFO_TABLE_ID,
+	AMD_SRIOV_MSG_INITD_H_TABLE_ID,
+	AMD_SRIOV_MSG_MAX_TABLE_ID,
+};
+
+struct amd_sriov_msg_init_data_header {
+	char     signature[4];  /* "INDA"  */
+	uint32_t version;
+	uint32_t checksum;
+	uint32_t initdata_offset; /* 0 */
+	uint32_t initdata_size_in_kb; /* 5MB */
+	uint32_t valid_tables;
+	uint32_t vbios_img_offset;
+	uint32_t vbios_img_size_in_kb;
+	uint32_t dataexchange_offset;
+	uint32_t dataexchange_size_in_kb;
+	uint32_t ras_tele_info_offset;
+	uint32_t ras_tele_info_size_in_kb;
+	uint32_t ip_discovery_offset;
+	uint32_t ip_discovery_size_in_kb;
+	uint32_t bad_page_info_offset;
+	uint32_t bad_page_size_in_kb;
+	uint32_t reserved[8];
+};
 
 /*
  * PF2VF history log:
@@ -70,7 +132,6 @@ enum amd_sriov_ucode_engine_id {
 	AMD_SRIOV_UCODE_ID_RLC_SRLS,
 	AMD_SRIOV_UCODE_ID_MEC,
 	AMD_SRIOV_UCODE_ID_MEC2,
-	AMD_SRIOV_UCODE_ID_IMU,
 	AMD_SRIOV_UCODE_ID_SOS,
 	AMD_SRIOV_UCODE_ID_ASD,
 	AMD_SRIOV_UCODE_ID_TA_RAS,
@@ -87,28 +148,64 @@ enum amd_sriov_ucode_engine_id {
 
 union amd_sriov_msg_feature_flags {
 	struct {
-		uint32_t error_log_collect : 1;
-		uint32_t host_load_ucodes  : 1;
-		uint32_t host_flr_vramlost : 1;
-		uint32_t mm_bw_management  : 1;
-		uint32_t pp_one_vf_mode	   : 1;
-		uint32_t reg_indirect_acc  : 1;
-		uint32_t av1_support       : 1;
-		uint32_t reserved	   : 25;
+		uint32_t error_log_collect	: 1;
+		uint32_t host_load_ucodes	: 1;
+		uint32_t host_flr_vramlost	: 1;
+		uint32_t mm_bw_management	: 1;
+		uint32_t pp_one_vf_mode		: 1;
+		uint32_t reg_indirect_acc	: 1;
+		uint32_t av1_support		: 1;
+		uint32_t vcn_rb_decouple	: 1;
+		uint32_t mes_info_dump_enable	: 1;
+		uint32_t ras_caps		: 1;
+		uint32_t ras_telemetry		: 1;
+		uint32_t ras_cper		: 1;
+		uint32_t xgmi_ta_ext_peer_link	: 1;
+		uint32_t reserved		: 19;
 	} flags;
 	uint32_t all;
 };
 
 union amd_sriov_reg_access_flags {
 	struct {
-		uint32_t vf_reg_access_ih 	 : 1;
-		uint32_t vf_reg_access_mmhub : 1;
-		uint32_t vf_reg_access_gc 	 : 1;
-		uint32_t reserved	         : 29;
+		uint32_t vf_reg_access_ih		: 1;
+		uint32_t vf_reg_access_mmhub		: 1;
+		uint32_t vf_reg_access_gc		: 1;
+		uint32_t vf_reg_access_l1_tlb_cntl	: 1;
+		uint32_t vf_reg_access_sq_config	: 1;
+		uint32_t reserved			: 27;
 	} flags;
 	uint32_t all;
 };
 
+union amd_sriov_ras_caps {
+	struct {
+		uint64_t block_umc			: 1;
+		uint64_t block_sdma			: 1;
+		uint64_t block_gfx			: 1;
+		uint64_t block_mmhub			: 1;
+		uint64_t block_athub			: 1;
+		uint64_t block_pcie_bif			: 1;
+		uint64_t block_hdp			: 1;
+		uint64_t block_xgmi_wafl		: 1;
+		uint64_t block_df			: 1;
+		uint64_t block_smn			: 1;
+		uint64_t block_sem			: 1;
+		uint64_t block_mp0			: 1;
+		uint64_t block_mp1			: 1;
+		uint64_t block_fuse			: 1;
+		uint64_t block_mca			: 1;
+		uint64_t block_vcn			: 1;
+		uint64_t block_jpeg			: 1;
+		uint64_t block_ih			: 1;
+		uint64_t block_mpio			: 1;
+		uint64_t block_mmsch			: 1;
+		uint64_t poison_propogation_mode	: 1;
+		uint64_t reserved			: 43;
+	} bits;
+	uint64_t all;
+};
+
 union amd_sriov_msg_os_info {
 	struct {
 		uint32_t windows  : 1;
@@ -157,7 +254,7 @@ struct amd_sriov_msg_pf2vf_info_header {
 	uint32_t reserved[2];
 };
 
-#define AMD_SRIOV_MSG_PF2VF_INFO_FILLED_SIZE (48)
+#define AMD_SRIOV_MSG_PF2VF_INFO_FILLED_SIZE (55)
 struct amd_sriov_msg_pf2vf_info {
 	/* header contains size and version */
 	struct amd_sriov_msg_pf2vf_info_header header;
@@ -208,9 +305,17 @@ struct amd_sriov_msg_pf2vf_info {
 	struct amd_sriov_msg_uuid_info uuid_info;
 	/* PCIE atomic ops support flag */
 	uint32_t pcie_atomic_ops_support_flags;
+	/* Portion of GPU memory occupied by VF.  MAX value is 65535, but set to uint32_t to maintain alignment with reserved size */
+	uint32_t gpu_capacity;
+	/* vf bdf on host pci tree for debug only */
+	uint32_t bdf_on_host;
+	uint32_t more_bp;	//Reserved for future use.
+	union amd_sriov_ras_caps ras_en_caps;
+	union amd_sriov_ras_caps ras_telemetry_en_caps;
+
 	/* reserved */
 	uint32_t reserved[256 - AMD_SRIOV_MSG_PF2VF_INFO_FILLED_SIZE];
-};
+} __packed;
 
 struct amd_sriov_msg_vf2pf_info_header {
 	/* the total structure size in byte */
@@ -221,7 +326,7 @@ struct amd_sriov_msg_vf2pf_info_header {
 	uint32_t reserved[2];
 };
 
-#define AMD_SRIOV_MSG_VF2PF_INFO_FILLED_SIZE (70)
+#define AMD_SRIOV_MSG_VF2PF_INFO_FILLED_SIZE (73)
 struct amd_sriov_msg_vf2pf_info {
 	/* header contains size and version */
 	struct amd_sriov_msg_vf2pf_info_header header;
@@ -265,10 +370,12 @@ struct amd_sriov_msg_vf2pf_info {
 		uint32_t version;
 	} ucode_info[AMD_SRIOV_MSG_RESERVE_UCODE];
 	uint64_t dummy_page_addr;
-
+	/* FB allocated for guest MES to record UQ info */
+	uint64_t mes_info_addr;
+	uint32_t mes_info_size;
 	/* reserved */
 	uint32_t reserved[256 - AMD_SRIOV_MSG_VF2PF_INFO_FILLED_SIZE];
-};
+} __packed;
 
 /* mailbox message send from guest to host  */
 enum amd_sriov_mailbox_request_message {
@@ -278,22 +385,98 @@ enum amd_sriov_mailbox_request_message {
 	MB_REQ_MSG_REL_GPU_FINI_ACCESS,
 	MB_REQ_MSG_REQ_GPU_RESET_ACCESS,
 	MB_REQ_MSG_REQ_GPU_INIT_DATA,
+	MB_REQ_MSG_PSP_VF_CMD_RELAY,
 
 	MB_REQ_MSG_LOG_VF_ERROR = 200,
+	MB_REQ_MSG_READY_TO_RESET = 201,
+	MB_REQ_MSG_RAS_POISON = 202,
+	MB_REQ_RAS_ERROR_COUNT = 203,
+	MB_REQ_RAS_CPER_DUMP = 204,
+	MB_REQ_RAS_BAD_PAGES = 205,
 };
 
 /* mailbox message send from host to guest  */
 enum amd_sriov_mailbox_response_message {
-	MB_RES_MSG_CLR_MSG_BUF = 0,
-	MB_RES_MSG_READY_TO_ACCESS_GPU = 1,
-	MB_RES_MSG_FLR_NOTIFICATION,
-	MB_RES_MSG_FLR_NOTIFICATION_COMPLETION,
-	MB_RES_MSG_SUCCESS,
-	MB_RES_MSG_FAIL,
-	MB_RES_MSG_QUERY_ALIVE,
-	MB_RES_MSG_GPU_INIT_DATA_READY,
-
-	MB_RES_MSG_TEXT_MESSAGE = 255
+	MB_RES_MSG_CLR_MSG_BUF			= 0,
+	MB_RES_MSG_READY_TO_ACCESS_GPU		= 1,
+	MB_RES_MSG_FLR_NOTIFICATION		= 2,
+	MB_RES_MSG_FLR_NOTIFICATION_COMPLETION  = 3,
+	MB_RES_MSG_SUCCESS			= 4,
+	MB_RES_MSG_FAIL				= 5,
+	MB_RES_MSG_QUERY_ALIVE			= 6,
+	MB_RES_MSG_GPU_INIT_DATA_READY		= 7,
+	MB_RES_MSG_RAS_POISON_READY		= 8,
+	MB_RES_MSG_PF_SOFT_FLR_NOTIFICATION	= 9,
+	MB_RES_MSG_GPU_RMA			= 10,
+	MB_RES_MSG_RAS_ERROR_COUNT_READY	= 11,
+	MB_REQ_RAS_CPER_DUMP_READY		= 14,
+	MB_RES_MSG_RAS_BAD_PAGES_READY		= 15,
+	MB_RES_MSG_RAS_BAD_PAGES_NOTIFICATION	= 16,
+	MB_RES_MSG_UNRECOV_ERR_NOTIFICATION	= 17,
+	MB_RES_MSG_TEXT_MESSAGE			= 255
+};
+
+enum amd_sriov_ras_telemetry_gpu_block {
+	RAS_TELEMETRY_GPU_BLOCK_UMC		= 0,
+	RAS_TELEMETRY_GPU_BLOCK_SDMA		= 1,
+	RAS_TELEMETRY_GPU_BLOCK_GFX		= 2,
+	RAS_TELEMETRY_GPU_BLOCK_MMHUB		= 3,
+	RAS_TELEMETRY_GPU_BLOCK_ATHUB		= 4,
+	RAS_TELEMETRY_GPU_BLOCK_PCIE_BIF	= 5,
+	RAS_TELEMETRY_GPU_BLOCK_HDP		= 6,
+	RAS_TELEMETRY_GPU_BLOCK_XGMI_WAFL	= 7,
+	RAS_TELEMETRY_GPU_BLOCK_DF		= 8,
+	RAS_TELEMETRY_GPU_BLOCK_SMN		= 9,
+	RAS_TELEMETRY_GPU_BLOCK_SEM		= 10,
+	RAS_TELEMETRY_GPU_BLOCK_MP0		= 11,
+	RAS_TELEMETRY_GPU_BLOCK_MP1		= 12,
+	RAS_TELEMETRY_GPU_BLOCK_FUSE		= 13,
+	RAS_TELEMETRY_GPU_BLOCK_MCA		= 14,
+	RAS_TELEMETRY_GPU_BLOCK_VCN		= 15,
+	RAS_TELEMETRY_GPU_BLOCK_JPEG		= 16,
+	RAS_TELEMETRY_GPU_BLOCK_IH		= 17,
+	RAS_TELEMETRY_GPU_BLOCK_MPIO		= 18,
+	RAS_TELEMETRY_GPU_BLOCK_COUNT		= 19,
+};
+
+struct amd_sriov_ras_telemetry_header {
+	uint32_t checksum;
+	uint32_t used_size;
+	uint32_t reserved[2];
+};
+
+struct amd_sriov_ras_telemetry_error_count {
+	struct {
+		uint32_t ce_count;
+		uint32_t ue_count;
+		uint32_t de_count;
+		uint32_t ce_overflow_count;
+		uint32_t ue_overflow_count;
+		uint32_t de_overflow_count;
+		uint32_t reserved[6];
+	} block[RAS_TELEMETRY_GPU_BLOCK_COUNT];
+};
+
+struct amd_sriov_ras_cper_dump {
+	uint32_t more;
+	uint64_t overflow_count;
+	uint64_t count;
+	uint64_t wptr;
+	uint32_t buf[];
+};
+
+struct amd_sriov_ras_chk_criti {
+	uint32_t hit;
+};
+
+struct amdsriov_ras_telemetry {
+	struct amd_sriov_ras_telemetry_header header;
+
+	union {
+		struct amd_sriov_ras_telemetry_error_count error_count;
+		struct amd_sriov_ras_cper_dump cper_dump;
+		struct amd_sriov_ras_chk_criti chk_criti;
+	} body;
 };
 
 /* version data stored in MAILBOX_MSGBUF_RCV_DW1 for future expansion */
diff --git a/drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c b/drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c
new file mode 100644
index 000000000000..f9e2edf5260b
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/aqua_vanjaram.c
@@ -0,0 +1,986 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "soc15.h"
+
+#include "soc15_common.h"
+#include "amdgpu_reg_state.h"
+#include "amdgpu_xcp.h"
+#include "gfx_v9_4_3.h"
+#include "gfxhub_v1_2.h"
+#include "sdma_v4_4_2.h"
+#include "amdgpu_ip.h"
+
+#define XCP_INST_MASK(num_inst, xcp_id)                                        \
+	(num_inst ? GENMASK(num_inst - 1, 0) << (xcp_id * num_inst) : 0)
+
+void aqua_vanjaram_doorbell_index_init(struct amdgpu_device *adev)
+{
+	int i;
+
+	adev->doorbell_index.kiq = AMDGPU_DOORBELL_LAYOUT1_KIQ_START;
+
+	adev->doorbell_index.mec_ring0 = AMDGPU_DOORBELL_LAYOUT1_MEC_RING_START;
+
+	adev->doorbell_index.userqueue_start = AMDGPU_DOORBELL_LAYOUT1_USERQUEUE_START;
+	adev->doorbell_index.userqueue_end = AMDGPU_DOORBELL_LAYOUT1_USERQUEUE_END;
+	adev->doorbell_index.xcc_doorbell_range = AMDGPU_DOORBELL_LAYOUT1_XCC_RANGE;
+
+	adev->doorbell_index.sdma_doorbell_range = 20;
+	for (i = 0; i < adev->sdma.num_instances; i++)
+		adev->doorbell_index.sdma_engine[i] =
+			AMDGPU_DOORBELL_LAYOUT1_sDMA_ENGINE_START +
+			i * (adev->doorbell_index.sdma_doorbell_range >> 1);
+
+	adev->doorbell_index.ih = AMDGPU_DOORBELL_LAYOUT1_IH;
+	adev->doorbell_index.vcn.vcn_ring0_1 = AMDGPU_DOORBELL_LAYOUT1_VCN_START;
+
+	adev->doorbell_index.first_non_cp = AMDGPU_DOORBELL_LAYOUT1_FIRST_NON_CP;
+	adev->doorbell_index.last_non_cp = AMDGPU_DOORBELL_LAYOUT1_LAST_NON_CP;
+
+	adev->doorbell_index.max_assignment = AMDGPU_DOORBELL_LAYOUT1_MAX_ASSIGNMENT << 1;
+}
+
+/* Fixed pattern for smn addressing on different AIDs:
+ *   bit[34]: indicate cross AID access
+ *   bit[33:32]: indicate target AID id
+ * AID id range is 0 ~ 3 as maximum AID number is 4.
+ */
+u64 aqua_vanjaram_encode_ext_smn_addressing(int ext_id)
+{
+	u64 ext_offset;
+
+	/* local routing and bit[34:32] will be zeros */
+	if (ext_id == 0)
+		return 0;
+
+	/* Initiated from host, accessing to all non-zero aids are cross traffic */
+	ext_offset = ((u64)(ext_id & 0x3) << 32) | (1ULL << 34);
+
+	return ext_offset;
+}
+
+static enum amdgpu_gfx_partition
+__aqua_vanjaram_calc_xcp_mode(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	int num_xcc, num_xcc_per_xcp = 0, mode = 0;
+
+	num_xcc = NUM_XCC(xcp_mgr->adev->gfx.xcc_mask);
+	if (adev->gfx.funcs->get_xccs_per_xcp)
+		num_xcc_per_xcp = adev->gfx.funcs->get_xccs_per_xcp(adev);
+	if ((num_xcc_per_xcp) && (num_xcc % num_xcc_per_xcp == 0))
+		mode = num_xcc / num_xcc_per_xcp;
+
+	if (num_xcc_per_xcp == 1)
+		return AMDGPU_CPX_PARTITION_MODE;
+
+	switch (mode) {
+	case 1:
+		return AMDGPU_SPX_PARTITION_MODE;
+	case 2:
+		return AMDGPU_DPX_PARTITION_MODE;
+	case 3:
+		return AMDGPU_TPX_PARTITION_MODE;
+	case 4:
+		return AMDGPU_QPX_PARTITION_MODE;
+	default:
+		return AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE;
+	}
+
+	return AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE;
+}
+
+static int aqua_vanjaram_query_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	enum amdgpu_gfx_partition derv_mode,
+		mode = AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE;
+	struct amdgpu_device *adev = xcp_mgr->adev;
+
+	derv_mode = __aqua_vanjaram_calc_xcp_mode(xcp_mgr);
+
+	if (amdgpu_sriov_vf(adev))
+		return derv_mode;
+
+	if (adev->nbio.funcs->get_compute_partition_mode) {
+		mode = adev->nbio.funcs->get_compute_partition_mode(adev);
+		if (mode != derv_mode) {
+			dev_warn(
+				adev->dev,
+				"Mismatch in compute partition mode - reported : %d derived : %d",
+				mode, derv_mode);
+			if (derv_mode == AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE)
+				amdgpu_device_bus_status_check(adev);
+		}
+	}
+
+	return mode;
+}
+
+static int __aqua_vanjaram_get_xcc_per_xcp(struct amdgpu_xcp_mgr *xcp_mgr, int mode)
+{
+	int num_xcc, num_xcc_per_xcp = 0;
+
+	num_xcc = NUM_XCC(xcp_mgr->adev->gfx.xcc_mask);
+
+	switch (mode) {
+	case AMDGPU_SPX_PARTITION_MODE:
+		num_xcc_per_xcp = num_xcc;
+		break;
+	case AMDGPU_DPX_PARTITION_MODE:
+		num_xcc_per_xcp = num_xcc / 2;
+		break;
+	case AMDGPU_TPX_PARTITION_MODE:
+		num_xcc_per_xcp = num_xcc / 3;
+		break;
+	case AMDGPU_QPX_PARTITION_MODE:
+		num_xcc_per_xcp = num_xcc / 4;
+		break;
+	case AMDGPU_CPX_PARTITION_MODE:
+		num_xcc_per_xcp = 1;
+		break;
+	}
+
+	return num_xcc_per_xcp;
+}
+
+static int __aqua_vanjaram_get_xcp_ip_info(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id,
+				    enum AMDGPU_XCP_IP_BLOCK ip_id,
+				    struct amdgpu_xcp_ip *ip)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	int num_sdma, num_vcn, num_shared_vcn, num_xcp;
+	int num_xcc_xcp, num_sdma_xcp, num_vcn_xcp;
+
+	num_sdma = adev->sdma.num_instances;
+	num_vcn = adev->vcn.num_vcn_inst;
+	num_shared_vcn = 1;
+
+	num_xcc_xcp = adev->gfx.num_xcc_per_xcp;
+	num_xcp = NUM_XCC(adev->gfx.xcc_mask) / num_xcc_xcp;
+
+	switch (xcp_mgr->mode) {
+	case AMDGPU_SPX_PARTITION_MODE:
+	case AMDGPU_DPX_PARTITION_MODE:
+	case AMDGPU_TPX_PARTITION_MODE:
+	case AMDGPU_QPX_PARTITION_MODE:
+	case AMDGPU_CPX_PARTITION_MODE:
+		num_sdma_xcp = DIV_ROUND_UP(num_sdma, num_xcp);
+		num_vcn_xcp = DIV_ROUND_UP(num_vcn, num_xcp);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (num_vcn && num_xcp > num_vcn)
+		num_shared_vcn = num_xcp / num_vcn;
+
+	switch (ip_id) {
+	case AMDGPU_XCP_GFXHUB:
+		ip->inst_mask = XCP_INST_MASK(num_xcc_xcp, xcp_id);
+		ip->ip_funcs = &gfxhub_v1_2_xcp_funcs;
+		break;
+	case AMDGPU_XCP_GFX:
+		ip->inst_mask = XCP_INST_MASK(num_xcc_xcp, xcp_id);
+		ip->ip_funcs = &gfx_v9_4_3_xcp_funcs;
+		break;
+	case AMDGPU_XCP_SDMA:
+		ip->inst_mask = XCP_INST_MASK(num_sdma_xcp, xcp_id);
+		ip->ip_funcs = &sdma_v4_4_2_xcp_funcs;
+		break;
+	case AMDGPU_XCP_VCN:
+		ip->inst_mask =
+			XCP_INST_MASK(num_vcn_xcp, xcp_id / num_shared_vcn);
+		/* TODO : Assign IP funcs */
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	ip->ip_id = ip_id;
+
+	return 0;
+}
+
+static int __aqua_vanjaram_get_px_mode_info(struct amdgpu_xcp_mgr *xcp_mgr,
+					    int px_mode, int *num_xcp,
+					    uint16_t *nps_modes)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	uint32_t gc_ver = amdgpu_ip_version(adev, GC_HWIP, 0);
+
+	if (!num_xcp || !nps_modes || !(xcp_mgr->supp_xcp_modes & BIT(px_mode)))
+		return -EINVAL;
+
+	switch (px_mode) {
+	case AMDGPU_SPX_PARTITION_MODE:
+		*num_xcp = 1;
+		*nps_modes = BIT(AMDGPU_NPS1_PARTITION_MODE);
+		break;
+	case AMDGPU_DPX_PARTITION_MODE:
+		*num_xcp = 2;
+		*nps_modes = BIT(AMDGPU_NPS1_PARTITION_MODE) |
+			     BIT(AMDGPU_NPS2_PARTITION_MODE);
+		break;
+	case AMDGPU_TPX_PARTITION_MODE:
+		*num_xcp = 3;
+		*nps_modes = BIT(AMDGPU_NPS1_PARTITION_MODE) |
+			     BIT(AMDGPU_NPS4_PARTITION_MODE);
+		break;
+	case AMDGPU_QPX_PARTITION_MODE:
+		*num_xcp = 4;
+		*nps_modes = BIT(AMDGPU_NPS1_PARTITION_MODE) |
+			     BIT(AMDGPU_NPS4_PARTITION_MODE);
+		if (gc_ver == IP_VERSION(9, 5, 0))
+			*nps_modes |= BIT(AMDGPU_NPS2_PARTITION_MODE);
+		break;
+	case AMDGPU_CPX_PARTITION_MODE:
+		*num_xcp = NUM_XCC(adev->gfx.xcc_mask);
+		*nps_modes = BIT(AMDGPU_NPS1_PARTITION_MODE) |
+			     BIT(AMDGPU_NPS4_PARTITION_MODE);
+		if (gc_ver == IP_VERSION(9, 5, 0))
+			*nps_modes |= BIT(AMDGPU_NPS2_PARTITION_MODE);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int aqua_vanjaram_get_xcp_res_info(struct amdgpu_xcp_mgr *xcp_mgr,
+					  int mode,
+					  struct amdgpu_xcp_cfg *xcp_cfg)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	int max_res[AMDGPU_XCP_RES_MAX] = {};
+	bool res_lt_xcp;
+	int num_xcp, i, r;
+	u16 nps_modes;
+
+	if (!(xcp_mgr->supp_xcp_modes & BIT(mode)))
+		return -EINVAL;
+
+	max_res[AMDGPU_XCP_RES_XCC] = NUM_XCC(adev->gfx.xcc_mask);
+	max_res[AMDGPU_XCP_RES_DMA] = adev->sdma.num_instances;
+	max_res[AMDGPU_XCP_RES_DEC] = adev->vcn.num_vcn_inst;
+	max_res[AMDGPU_XCP_RES_JPEG] = adev->jpeg.num_jpeg_inst;
+
+	r = __aqua_vanjaram_get_px_mode_info(xcp_mgr, mode, &num_xcp, &nps_modes);
+	if (r)
+		return r;
+
+	xcp_cfg->compatible_nps_modes =
+		(adev->gmc.supported_nps_modes & nps_modes);
+	xcp_cfg->num_res = ARRAY_SIZE(max_res);
+
+	for (i = 0; i < xcp_cfg->num_res; i++) {
+		res_lt_xcp = max_res[i] < num_xcp;
+		xcp_cfg->xcp_res[i].id = i;
+		xcp_cfg->xcp_res[i].num_inst =
+			res_lt_xcp ? 1 : max_res[i] / num_xcp;
+		xcp_cfg->xcp_res[i].num_inst =
+			i == AMDGPU_XCP_RES_JPEG ?
+			xcp_cfg->xcp_res[i].num_inst *
+			adev->jpeg.num_jpeg_rings : xcp_cfg->xcp_res[i].num_inst;
+		xcp_cfg->xcp_res[i].num_shared =
+			res_lt_xcp ? num_xcp / max_res[i] : 1;
+	}
+
+	return 0;
+}
+
+static enum amdgpu_gfx_partition
+__aqua_vanjaram_get_auto_mode(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	int num_xcc;
+
+	num_xcc = NUM_XCC(xcp_mgr->adev->gfx.xcc_mask);
+
+	if (adev->gmc.num_mem_partitions == 1)
+		return AMDGPU_SPX_PARTITION_MODE;
+
+	if (adev->gmc.num_mem_partitions == num_xcc)
+		return AMDGPU_CPX_PARTITION_MODE;
+
+	if (adev->gmc.num_mem_partitions == num_xcc / 2)
+		return (adev->flags & AMD_IS_APU) ? AMDGPU_TPX_PARTITION_MODE :
+						    AMDGPU_CPX_PARTITION_MODE;
+
+	if (adev->gmc.num_mem_partitions == 2 && !(adev->flags & AMD_IS_APU))
+		return AMDGPU_DPX_PARTITION_MODE;
+
+	return AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE;
+}
+
+static bool __aqua_vanjaram_is_valid_mode(struct amdgpu_xcp_mgr *xcp_mgr,
+					  enum amdgpu_gfx_partition mode)
+{
+	struct amdgpu_device *adev = xcp_mgr->adev;
+	int num_xcc, num_xccs_per_xcp, r;
+	int num_xcp, nps_mode;
+	u16 supp_nps_modes;
+	bool comp_mode;
+
+	nps_mode = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+	r = __aqua_vanjaram_get_px_mode_info(xcp_mgr, mode, &num_xcp,
+					       &supp_nps_modes);
+	if (r)
+		return false;
+
+	comp_mode = !!(BIT(nps_mode) & supp_nps_modes);
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	switch (mode) {
+	case AMDGPU_SPX_PARTITION_MODE:
+		return comp_mode && num_xcc > 0;
+	case AMDGPU_DPX_PARTITION_MODE:
+		return comp_mode && (num_xcc % 4) == 0;
+	case AMDGPU_TPX_PARTITION_MODE:
+		return comp_mode && ((num_xcc % 3) == 0);
+	case AMDGPU_QPX_PARTITION_MODE:
+		num_xccs_per_xcp = num_xcc / 4;
+		return comp_mode && (num_xccs_per_xcp >= 2);
+	case AMDGPU_CPX_PARTITION_MODE:
+		return comp_mode && (num_xcc > 1);
+	default:
+		return false;
+	}
+
+	return false;
+}
+
+static void __aqua_vanjaram_update_available_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr)
+{
+	int mode;
+
+	xcp_mgr->avail_xcp_modes = 0;
+
+	for_each_inst(mode, xcp_mgr->supp_xcp_modes) {
+		if (__aqua_vanjaram_is_valid_mode(xcp_mgr, mode))
+			xcp_mgr->avail_xcp_modes |= BIT(mode);
+	}
+}
+
+static int aqua_vanjaram_switch_partition_mode(struct amdgpu_xcp_mgr *xcp_mgr,
+					       int mode, int *num_xcps)
+{
+	int num_xcc_per_xcp, num_xcc, ret;
+	struct amdgpu_device *adev;
+	u32 flags = 0;
+
+	adev = xcp_mgr->adev;
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+
+	if (mode == AMDGPU_AUTO_COMPUTE_PARTITION_MODE) {
+		mode = __aqua_vanjaram_get_auto_mode(xcp_mgr);
+		if (mode == AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE) {
+			dev_err(adev->dev,
+				"Invalid config, no compatible compute partition mode found, available memory partitions: %d",
+				adev->gmc.num_mem_partitions);
+			return -EINVAL;
+		}
+	} else if (!__aqua_vanjaram_is_valid_mode(xcp_mgr, mode)) {
+		dev_err(adev->dev,
+			"Invalid compute partition mode requested, requested: %s, available memory partitions: %d",
+			amdgpu_gfx_compute_mode_desc(mode), adev->gmc.num_mem_partitions);
+		return -EINVAL;
+	}
+
+	if (adev->kfd.init_complete && !amdgpu_in_reset(adev) &&
+		!adev->in_suspend)
+		flags |= AMDGPU_XCP_OPS_KFD;
+
+	if (flags & AMDGPU_XCP_OPS_KFD) {
+		ret = amdgpu_amdkfd_check_and_lock_kfd(adev);
+		if (ret)
+			goto out;
+	}
+
+	ret = amdgpu_xcp_pre_partition_switch(xcp_mgr, flags);
+	if (ret)
+		goto unlock;
+
+	num_xcc_per_xcp = __aqua_vanjaram_get_xcc_per_xcp(xcp_mgr, mode);
+	if (adev->gfx.funcs->switch_partition_mode)
+		adev->gfx.funcs->switch_partition_mode(xcp_mgr->adev,
+						       num_xcc_per_xcp);
+
+	/* Init info about new xcps */
+	*num_xcps = num_xcc / num_xcc_per_xcp;
+	amdgpu_xcp_init(xcp_mgr, *num_xcps, mode);
+
+	ret = amdgpu_xcp_post_partition_switch(xcp_mgr, flags);
+	if (!ret)
+		__aqua_vanjaram_update_available_partition_mode(xcp_mgr);
+unlock:
+	if (flags & AMDGPU_XCP_OPS_KFD)
+		amdgpu_amdkfd_unlock_kfd(adev);
+out:
+	return ret;
+}
+
+static int __aqua_vanjaram_get_xcp_mem_id(struct amdgpu_device *adev,
+					  int xcc_id, uint8_t *mem_id)
+{
+	/* memory/spatial modes validation check is already done */
+	*mem_id = xcc_id / adev->gfx.num_xcc_per_xcp;
+	*mem_id /= adev->xcp_mgr->num_xcp_per_mem_partition;
+
+	return 0;
+}
+
+static int aqua_vanjaram_get_xcp_mem_id(struct amdgpu_xcp_mgr *xcp_mgr,
+					struct amdgpu_xcp *xcp, uint8_t *mem_id)
+{
+	struct amdgpu_numa_info numa_info;
+	struct amdgpu_device *adev;
+	uint32_t xcc_mask;
+	int r, i, xcc_id;
+
+	adev = xcp_mgr->adev;
+	/* TODO: BIOS is not returning the right info now
+	 * Check on this later
+	 */
+	/*
+	if (adev->gmc.gmc_funcs->query_mem_partition_mode)
+		mode = adev->gmc.gmc_funcs->query_mem_partition_mode(adev);
+	*/
+	if (adev->gmc.num_mem_partitions == 1) {
+		/* Only one range */
+		*mem_id = 0;
+		return 0;
+	}
+
+	r = amdgpu_xcp_get_inst_details(xcp, AMDGPU_XCP_GFX, &xcc_mask);
+	if (r || !xcc_mask)
+		return -EINVAL;
+
+	xcc_id = ffs(xcc_mask) - 1;
+	if (!adev->gmc.is_app_apu)
+		return __aqua_vanjaram_get_xcp_mem_id(adev, xcc_id, mem_id);
+
+	r = amdgpu_acpi_get_mem_info(adev, xcc_id, &numa_info);
+
+	if (r)
+		return r;
+
+	r = -EINVAL;
+	for (i = 0; i < adev->gmc.num_mem_partitions; ++i) {
+		if (adev->gmc.mem_partitions[i].numa.node == numa_info.nid) {
+			*mem_id = i;
+			r = 0;
+			break;
+		}
+	}
+
+	return r;
+}
+
+static int aqua_vanjaram_get_xcp_ip_details(struct amdgpu_xcp_mgr *xcp_mgr, int xcp_id,
+				     enum AMDGPU_XCP_IP_BLOCK ip_id,
+				     struct amdgpu_xcp_ip *ip)
+{
+	if (!ip)
+		return -EINVAL;
+
+	return __aqua_vanjaram_get_xcp_ip_info(xcp_mgr, xcp_id, ip_id, ip);
+}
+
+struct amdgpu_xcp_mgr_funcs aqua_vanjaram_xcp_funcs = {
+	.switch_partition_mode = &aqua_vanjaram_switch_partition_mode,
+	.query_partition_mode = &aqua_vanjaram_query_partition_mode,
+	.get_ip_details = &aqua_vanjaram_get_xcp_ip_details,
+	.get_xcp_res_info = &aqua_vanjaram_get_xcp_res_info,
+	.get_xcp_mem_id = &aqua_vanjaram_get_xcp_mem_id,
+};
+
+static int aqua_vanjaram_xcp_mgr_init(struct amdgpu_device *adev)
+{
+	int ret;
+
+	if (amdgpu_sriov_vf(adev))
+		aqua_vanjaram_xcp_funcs.switch_partition_mode = NULL;
+
+	ret = amdgpu_xcp_mgr_init(adev, AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE, 1,
+				  &aqua_vanjaram_xcp_funcs);
+	if (ret)
+		return ret;
+
+	amdgpu_xcp_update_supported_modes(adev->xcp_mgr);
+	/* TODO: Default memory node affinity init */
+
+	return ret;
+}
+
+int aqua_vanjaram_init_soc_config(struct amdgpu_device *adev)
+{
+	u32 mask, avail_inst, inst_mask = adev->sdma.sdma_mask;
+	int ret, i;
+
+	/* generally 1 AID supports 4 instances */
+	adev->sdma.num_inst_per_aid = 4;
+	adev->sdma.num_instances = NUM_SDMA(adev->sdma.sdma_mask);
+
+	adev->aid_mask = i = 1;
+	inst_mask >>= adev->sdma.num_inst_per_aid;
+
+	for (mask = (1 << adev->sdma.num_inst_per_aid) - 1; inst_mask;
+	     inst_mask >>= adev->sdma.num_inst_per_aid, ++i) {
+		avail_inst = inst_mask & mask;
+		if (avail_inst == mask || avail_inst == 0x3 ||
+		    avail_inst == 0xc)
+			adev->aid_mask |= (1 << i);
+	}
+
+	/* Harvest config is not used for aqua vanjaram. VCN and JPEGs will be
+	 * addressed based on logical instance ids.
+	 */
+	adev->vcn.harvest_config = 0;
+	adev->vcn.num_inst_per_aid = 1;
+	adev->vcn.num_vcn_inst = hweight32(adev->vcn.inst_mask);
+	adev->jpeg.harvest_config = 0;
+	adev->jpeg.num_inst_per_aid = 1;
+	adev->jpeg.num_jpeg_inst = hweight32(adev->jpeg.inst_mask);
+
+	ret = aqua_vanjaram_xcp_mgr_init(adev);
+	if (ret)
+		return ret;
+
+	amdgpu_ip_map_init(adev);
+
+	return 0;
+}
+
+static void aqua_read_smn(struct amdgpu_device *adev,
+			  struct amdgpu_smn_reg_data *regdata,
+			  uint64_t smn_addr)
+{
+	regdata->addr = smn_addr;
+	regdata->value = RREG32_PCIE(smn_addr);
+}
+
+struct aqua_reg_list {
+	uint64_t start_addr;
+	uint32_t num_regs;
+	uint32_t incrx;
+};
+
+#define DW_ADDR_INCR	4
+
+static void aqua_read_smn_ext(struct amdgpu_device *adev,
+			      struct amdgpu_smn_reg_data *regdata,
+			      uint64_t smn_addr, int i)
+{
+	regdata->addr =
+		smn_addr + adev->asic_funcs->encode_ext_smn_addressing(i);
+	regdata->value = RREG32_PCIE_EXT(regdata->addr);
+}
+
+#define smnreg_0x1A340218	0x1A340218
+#define smnreg_0x1A3402E4	0x1A3402E4
+#define smnreg_0x1A340294	0x1A340294
+#define smreg_0x1A380088	0x1A380088
+
+#define NUM_PCIE_SMN_REGS	14
+
+static struct aqua_reg_list pcie_reg_addrs[] = {
+	{ smnreg_0x1A340218, 1, 0 },
+	{ smnreg_0x1A3402E4, 1, 0 },
+	{ smnreg_0x1A340294, 6, DW_ADDR_INCR },
+	{ smreg_0x1A380088, 6, DW_ADDR_INCR },
+};
+
+static ssize_t aqua_vanjaram_read_pcie_state(struct amdgpu_device *adev,
+					     void *buf, size_t max_size)
+{
+	struct amdgpu_reg_state_pcie_v1_0 *pcie_reg_state;
+	uint32_t start_addr, incrx, num_regs, szbuf;
+	struct amdgpu_regs_pcie_v1_0 *pcie_regs;
+	struct amdgpu_smn_reg_data *reg_data;
+	struct pci_dev *us_pdev, *ds_pdev;
+	int aer_cap, r, n;
+
+	if (!buf || !max_size)
+		return -EINVAL;
+
+	pcie_reg_state = (struct amdgpu_reg_state_pcie_v1_0 *)buf;
+
+	szbuf = sizeof(*pcie_reg_state) +
+		amdgpu_reginst_size(1, sizeof(*pcie_regs), NUM_PCIE_SMN_REGS);
+	/* Only one instance of pcie regs */
+	if (max_size < szbuf)
+		return -EOVERFLOW;
+
+	pcie_regs = (struct amdgpu_regs_pcie_v1_0 *)((uint8_t *)buf +
+						     sizeof(*pcie_reg_state));
+	pcie_regs->inst_header.instance = 0;
+	pcie_regs->inst_header.state = AMDGPU_INST_S_OK;
+	pcie_regs->inst_header.num_smn_regs = NUM_PCIE_SMN_REGS;
+
+	reg_data = pcie_regs->smn_reg_values;
+
+	for (r = 0; r < ARRAY_SIZE(pcie_reg_addrs); r++) {
+		start_addr = pcie_reg_addrs[r].start_addr;
+		incrx = pcie_reg_addrs[r].incrx;
+		num_regs = pcie_reg_addrs[r].num_regs;
+		for (n = 0; n < num_regs; n++) {
+			aqua_read_smn(adev, reg_data, start_addr + n * incrx);
+			++reg_data;
+		}
+	}
+
+	ds_pdev = pci_upstream_bridge(adev->pdev);
+	us_pdev = pci_upstream_bridge(ds_pdev);
+
+	pcie_capability_read_word(us_pdev, PCI_EXP_DEVSTA,
+				  &pcie_regs->device_status);
+	pcie_capability_read_word(us_pdev, PCI_EXP_LNKSTA,
+				  &pcie_regs->link_status);
+
+	aer_cap = pci_find_ext_capability(us_pdev, PCI_EXT_CAP_ID_ERR);
+	if (aer_cap) {
+		pci_read_config_dword(us_pdev, aer_cap + PCI_ERR_COR_STATUS,
+				      &pcie_regs->pcie_corr_err_status);
+		pci_read_config_dword(us_pdev, aer_cap + PCI_ERR_UNCOR_STATUS,
+				      &pcie_regs->pcie_uncorr_err_status);
+	}
+
+	pci_read_config_dword(us_pdev, PCI_PRIMARY_BUS,
+			      &pcie_regs->sub_bus_number_latency);
+
+	pcie_reg_state->common_header.structure_size = szbuf;
+	pcie_reg_state->common_header.format_revision = 1;
+	pcie_reg_state->common_header.content_revision = 0;
+	pcie_reg_state->common_header.state_type = AMDGPU_REG_STATE_TYPE_PCIE;
+	pcie_reg_state->common_header.num_instances = 1;
+
+	return pcie_reg_state->common_header.structure_size;
+}
+
+#define smnreg_0x11A00050	0x11A00050
+#define smnreg_0x11A00180	0x11A00180
+#define smnreg_0x11A00070	0x11A00070
+#define smnreg_0x11A00200	0x11A00200
+#define smnreg_0x11A0020C	0x11A0020C
+#define smnreg_0x11A00210	0x11A00210
+#define smnreg_0x11A00108	0x11A00108
+
+#define XGMI_LINK_REG(smnreg, l) ((smnreg) | (l << 20))
+
+#define NUM_XGMI_SMN_REGS 25
+
+static struct aqua_reg_list xgmi_reg_addrs[] = {
+	{ smnreg_0x11A00050, 1, 0 },
+	{ smnreg_0x11A00180, 16, DW_ADDR_INCR },
+	{ smnreg_0x11A00070, 4, DW_ADDR_INCR },
+	{ smnreg_0x11A00200, 1, 0 },
+	{ smnreg_0x11A0020C, 1, 0 },
+	{ smnreg_0x11A00210, 1, 0 },
+	{ smnreg_0x11A00108, 1, 0 },
+};
+
+static ssize_t aqua_vanjaram_read_xgmi_state(struct amdgpu_device *adev,
+					     void *buf, size_t max_size)
+{
+	struct amdgpu_reg_state_xgmi_v1_0 *xgmi_reg_state;
+	uint32_t start_addr, incrx, num_regs, szbuf;
+	struct amdgpu_regs_xgmi_v1_0 *xgmi_regs;
+	struct amdgpu_smn_reg_data *reg_data;
+	const int max_xgmi_instances = 8;
+	int inst = 0, i, j, r, n;
+	const int xgmi_inst = 2;
+	void *p;
+
+	if (!buf || !max_size)
+		return -EINVAL;
+
+	xgmi_reg_state = (struct amdgpu_reg_state_xgmi_v1_0 *)buf;
+
+	szbuf = sizeof(*xgmi_reg_state) +
+		amdgpu_reginst_size(max_xgmi_instances, sizeof(*xgmi_regs),
+				    NUM_XGMI_SMN_REGS);
+	/* Only one instance of pcie regs */
+	if (max_size < szbuf)
+		return -EOVERFLOW;
+
+	p = &xgmi_reg_state->xgmi_state_regs[0];
+	for_each_inst(i, adev->aid_mask) {
+		for (j = 0; j < xgmi_inst; ++j) {
+			xgmi_regs = (struct amdgpu_regs_xgmi_v1_0 *)p;
+			xgmi_regs->inst_header.instance = inst++;
+
+			xgmi_regs->inst_header.state = AMDGPU_INST_S_OK;
+			xgmi_regs->inst_header.num_smn_regs = NUM_XGMI_SMN_REGS;
+
+			reg_data = xgmi_regs->smn_reg_values;
+
+			for (r = 0; r < ARRAY_SIZE(xgmi_reg_addrs); r++) {
+				start_addr = xgmi_reg_addrs[r].start_addr;
+				incrx = xgmi_reg_addrs[r].incrx;
+				num_regs = xgmi_reg_addrs[r].num_regs;
+
+				for (n = 0; n < num_regs; n++) {
+					aqua_read_smn_ext(
+						adev, reg_data,
+						XGMI_LINK_REG(start_addr, j) +
+							n * incrx,
+						i);
+					++reg_data;
+				}
+			}
+			p = reg_data;
+		}
+	}
+
+	xgmi_reg_state->common_header.structure_size = szbuf;
+	xgmi_reg_state->common_header.format_revision = 1;
+	xgmi_reg_state->common_header.content_revision = 0;
+	xgmi_reg_state->common_header.state_type = AMDGPU_REG_STATE_TYPE_XGMI;
+	xgmi_reg_state->common_header.num_instances = max_xgmi_instances;
+
+	return xgmi_reg_state->common_header.structure_size;
+}
+
+#define smnreg_0x11C00070	0x11C00070
+#define smnreg_0x11C00210	0x11C00210
+
+static struct aqua_reg_list wafl_reg_addrs[] = {
+	{ smnreg_0x11C00070, 4, DW_ADDR_INCR },
+	{ smnreg_0x11C00210, 1, 0 },
+};
+
+#define WAFL_LINK_REG(smnreg, l) ((smnreg) | (l << 20))
+
+#define NUM_WAFL_SMN_REGS 5
+
+static ssize_t aqua_vanjaram_read_wafl_state(struct amdgpu_device *adev,
+					     void *buf, size_t max_size)
+{
+	struct amdgpu_reg_state_wafl_v1_0 *wafl_reg_state;
+	uint32_t start_addr, incrx, num_regs, szbuf;
+	struct amdgpu_regs_wafl_v1_0 *wafl_regs;
+	struct amdgpu_smn_reg_data *reg_data;
+	const int max_wafl_instances = 8;
+	int inst = 0, i, j, r, n;
+	const int wafl_inst = 2;
+	void *p;
+
+	if (!buf || !max_size)
+		return -EINVAL;
+
+	wafl_reg_state = (struct amdgpu_reg_state_wafl_v1_0 *)buf;
+
+	szbuf = sizeof(*wafl_reg_state) +
+		amdgpu_reginst_size(max_wafl_instances, sizeof(*wafl_regs),
+				    NUM_WAFL_SMN_REGS);
+
+	if (max_size < szbuf)
+		return -EOVERFLOW;
+
+	p = &wafl_reg_state->wafl_state_regs[0];
+	for_each_inst(i, adev->aid_mask) {
+		for (j = 0; j < wafl_inst; ++j) {
+			wafl_regs = (struct amdgpu_regs_wafl_v1_0 *)p;
+			wafl_regs->inst_header.instance = inst++;
+
+			wafl_regs->inst_header.state = AMDGPU_INST_S_OK;
+			wafl_regs->inst_header.num_smn_regs = NUM_WAFL_SMN_REGS;
+
+			reg_data = wafl_regs->smn_reg_values;
+
+			for (r = 0; r < ARRAY_SIZE(wafl_reg_addrs); r++) {
+				start_addr = wafl_reg_addrs[r].start_addr;
+				incrx = wafl_reg_addrs[r].incrx;
+				num_regs = wafl_reg_addrs[r].num_regs;
+				for (n = 0; n < num_regs; n++) {
+					aqua_read_smn_ext(
+						adev, reg_data,
+						WAFL_LINK_REG(start_addr, j) +
+							n * incrx,
+						i);
+					++reg_data;
+				}
+			}
+			p = reg_data;
+		}
+	}
+
+	wafl_reg_state->common_header.structure_size = szbuf;
+	wafl_reg_state->common_header.format_revision = 1;
+	wafl_reg_state->common_header.content_revision = 0;
+	wafl_reg_state->common_header.state_type = AMDGPU_REG_STATE_TYPE_WAFL;
+	wafl_reg_state->common_header.num_instances = max_wafl_instances;
+
+	return wafl_reg_state->common_header.structure_size;
+}
+
+#define smnreg_0x1B311060 0x1B311060
+#define smnreg_0x1B411060 0x1B411060
+#define smnreg_0x1B511060 0x1B511060
+#define smnreg_0x1B611060 0x1B611060
+
+#define smnreg_0x1C307120 0x1C307120
+#define smnreg_0x1C317120 0x1C317120
+
+#define smnreg_0x1C320830 0x1C320830
+#define smnreg_0x1C380830 0x1C380830
+#define smnreg_0x1C3D0830 0x1C3D0830
+#define smnreg_0x1C420830 0x1C420830
+
+#define smnreg_0x1C320100 0x1C320100
+#define smnreg_0x1C380100 0x1C380100
+#define smnreg_0x1C3D0100 0x1C3D0100
+#define smnreg_0x1C420100 0x1C420100
+
+#define smnreg_0x1B310500 0x1B310500
+#define smnreg_0x1C300400 0x1C300400
+
+#define USR_CAKE_INCR 0x11000
+#define USR_LINK_INCR 0x100000
+#define USR_CP_INCR 0x10000
+
+#define NUM_USR_SMN_REGS	20
+
+struct aqua_reg_list usr_reg_addrs[] = {
+	{ smnreg_0x1B311060, 4, DW_ADDR_INCR },
+	{ smnreg_0x1B411060, 4, DW_ADDR_INCR },
+	{ smnreg_0x1B511060, 4, DW_ADDR_INCR },
+	{ smnreg_0x1B611060, 4, DW_ADDR_INCR },
+	{ smnreg_0x1C307120, 2, DW_ADDR_INCR },
+	{ smnreg_0x1C317120, 2, DW_ADDR_INCR },
+};
+
+#define NUM_USR1_SMN_REGS	46
+struct aqua_reg_list usr1_reg_addrs[] = {
+	{ smnreg_0x1C320830, 6, USR_CAKE_INCR },
+	{ smnreg_0x1C380830, 5, USR_CAKE_INCR },
+	{ smnreg_0x1C3D0830, 5, USR_CAKE_INCR },
+	{ smnreg_0x1C420830, 4, USR_CAKE_INCR },
+	{ smnreg_0x1C320100, 6, USR_CAKE_INCR },
+	{ smnreg_0x1C380100, 5, USR_CAKE_INCR },
+	{ smnreg_0x1C3D0100, 5, USR_CAKE_INCR },
+	{ smnreg_0x1C420100, 4, USR_CAKE_INCR },
+	{ smnreg_0x1B310500, 4, USR_LINK_INCR },
+	{ smnreg_0x1C300400, 2, USR_CP_INCR },
+};
+
+static ssize_t aqua_vanjaram_read_usr_state(struct amdgpu_device *adev,
+					    void *buf, size_t max_size,
+					    int reg_state)
+{
+	uint32_t start_addr, incrx, num_regs, szbuf, num_smn;
+	struct amdgpu_reg_state_usr_v1_0 *usr_reg_state;
+	struct amdgpu_regs_usr_v1_0 *usr_regs;
+	struct amdgpu_smn_reg_data *reg_data;
+	const int max_usr_instances = 4;
+	struct aqua_reg_list *reg_addrs;
+	int inst = 0, i, n, r, arr_size;
+	void *p;
+
+	if (!buf || !max_size)
+		return -EINVAL;
+
+	switch (reg_state) {
+	case AMDGPU_REG_STATE_TYPE_USR:
+		arr_size = ARRAY_SIZE(usr_reg_addrs);
+		reg_addrs = usr_reg_addrs;
+		num_smn = NUM_USR_SMN_REGS;
+		break;
+	case AMDGPU_REG_STATE_TYPE_USR_1:
+		arr_size = ARRAY_SIZE(usr1_reg_addrs);
+		reg_addrs = usr1_reg_addrs;
+		num_smn = NUM_USR1_SMN_REGS;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	usr_reg_state = (struct amdgpu_reg_state_usr_v1_0 *)buf;
+
+	szbuf = sizeof(*usr_reg_state) + amdgpu_reginst_size(max_usr_instances,
+							     sizeof(*usr_regs),
+							     num_smn);
+	if (max_size < szbuf)
+		return -EOVERFLOW;
+
+	p = &usr_reg_state->usr_state_regs[0];
+	for_each_inst(i, adev->aid_mask) {
+		usr_regs = (struct amdgpu_regs_usr_v1_0 *)p;
+		usr_regs->inst_header.instance = inst++;
+		usr_regs->inst_header.state = AMDGPU_INST_S_OK;
+		usr_regs->inst_header.num_smn_regs = num_smn;
+		reg_data = usr_regs->smn_reg_values;
+
+		for (r = 0; r < arr_size; r++) {
+			start_addr = reg_addrs[r].start_addr;
+			incrx = reg_addrs[r].incrx;
+			num_regs = reg_addrs[r].num_regs;
+			for (n = 0; n < num_regs; n++) {
+				aqua_read_smn_ext(adev, reg_data,
+						  start_addr + n * incrx, i);
+				reg_data++;
+			}
+		}
+		p = reg_data;
+	}
+
+	usr_reg_state->common_header.structure_size = szbuf;
+	usr_reg_state->common_header.format_revision = 1;
+	usr_reg_state->common_header.content_revision = 0;
+	usr_reg_state->common_header.state_type = AMDGPU_REG_STATE_TYPE_USR;
+	usr_reg_state->common_header.num_instances = max_usr_instances;
+
+	return usr_reg_state->common_header.structure_size;
+}
+
+ssize_t aqua_vanjaram_get_reg_state(struct amdgpu_device *adev,
+				    enum amdgpu_reg_state reg_state, void *buf,
+				    size_t max_size)
+{
+	ssize_t size;
+
+	switch (reg_state) {
+	case AMDGPU_REG_STATE_TYPE_PCIE:
+		size = aqua_vanjaram_read_pcie_state(adev, buf, max_size);
+		break;
+	case AMDGPU_REG_STATE_TYPE_XGMI:
+		size = aqua_vanjaram_read_xgmi_state(adev, buf, max_size);
+		break;
+	case AMDGPU_REG_STATE_TYPE_WAFL:
+		size = aqua_vanjaram_read_wafl_state(adev, buf, max_size);
+		break;
+	case AMDGPU_REG_STATE_TYPE_USR:
+		size = aqua_vanjaram_read_usr_state(adev, buf, max_size,
+						    AMDGPU_REG_STATE_TYPE_USR);
+		break;
+	case AMDGPU_REG_STATE_TYPE_USR_1:
+		size = aqua_vanjaram_read_usr_state(
+			adev, buf, max_size, AMDGPU_REG_STATE_TYPE_USR_1);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return size;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/athub_v1_0.c
index a13c443ea10f..42f4e163e251 100644
--- a/drivers/gpu/drm/amd/amdgpu/athub_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/athub_v1_0.c
@@ -68,7 +68,7 @@ int athub_v1_0_set_clockgating(struct amdgpu_device *adev,
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[ATHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
 	case IP_VERSION(9, 0, 0):
 	case IP_VERSION(9, 1, 0):
 	case IP_VERSION(9, 2, 0):
diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v2_0.c b/drivers/gpu/drm/amd/amdgpu/athub_v2_0.c
index a9521c98e7f7..5a122f50a6e7 100644
--- a/drivers/gpu/drm/amd/amdgpu/athub_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/athub_v2_0.c
@@ -77,7 +77,7 @@ int athub_v2_0_set_clockgating(struct amdgpu_device *adev,
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[ATHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
 	case IP_VERSION(1, 3, 1):
 	case IP_VERSION(2, 0, 0):
 	case IP_VERSION(2, 0, 2):
diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v2_1.c b/drivers/gpu/drm/amd/amdgpu/athub_v2_1.c
index 78508ae6a670..e143fcc46148 100644
--- a/drivers/gpu/drm/amd/amdgpu/athub_v2_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/athub_v2_1.c
@@ -70,7 +70,7 @@ int athub_v2_1_set_clockgating(struct amdgpu_device *adev,
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[ATHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
 	case IP_VERSION(2, 1, 0):
 	case IP_VERSION(2, 1, 1):
 	case IP_VERSION(2, 1, 2):
diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v3_0.c b/drivers/gpu/drm/amd/amdgpu/athub_v3_0.c
index f0e235f98afb..d1bba9c64e16 100644
--- a/drivers/gpu/drm/amd/amdgpu/athub_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/athub_v3_0.c
@@ -30,16 +30,21 @@
 
 #define regATHUB_MISC_CNTL_V3_0_1			0x00d7
 #define regATHUB_MISC_CNTL_V3_0_1_BASE_IDX		0
+#define regATHUB_MISC_CNTL_V3_3_0			0x00d8
+#define regATHUB_MISC_CNTL_V3_3_0_BASE_IDX		0
 
 
 static uint32_t athub_v3_0_get_cg_cntl(struct amdgpu_device *adev)
 {
 	uint32_t data;
 
-	switch (adev->ip_versions[ATHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
 	case IP_VERSION(3, 0, 1):
 		data = RREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL_V3_0_1);
 		break;
+	case IP_VERSION(3, 3, 0):
+		data = RREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL_V3_3_0);
+		break;
 	default:
 		data = RREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL);
 		break;
@@ -49,10 +54,13 @@ static uint32_t athub_v3_0_get_cg_cntl(struct amdgpu_device *adev)
 
 static void athub_v3_0_set_cg_cntl(struct amdgpu_device *adev, uint32_t data)
 {
-	switch (adev->ip_versions[ATHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
 	case IP_VERSION(3, 0, 1):
 		WREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL_V3_0_1, data);
 		break;
+	case IP_VERSION(3, 3, 0):
+		WREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL_V3_3_0, data);
+		break;
 	default:
 		WREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL, data);
 		break;
@@ -99,10 +107,11 @@ int athub_v3_0_set_clockgating(struct amdgpu_device *adev,
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[ATHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
 	case IP_VERSION(3, 0, 0):
 	case IP_VERSION(3, 0, 1):
 	case IP_VERSION(3, 0, 2):
+	case IP_VERSION(3, 3, 0):
 		athub_v3_0_update_medium_grain_clock_gating(adev,
 				state == AMD_CG_STATE_GATE);
 		athub_v3_0_update_medium_grain_light_sleep(adev,
diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v4_1_0.c b/drivers/gpu/drm/amd/amdgpu/athub_v4_1_0.c
new file mode 100644
index 000000000000..8a0773b80864
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/athub_v4_1_0.c
@@ -0,0 +1,122 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+#include "athub_v4_1_0.h"
+#include "athub/athub_4_1_0_offset.h"
+#include "athub/athub_4_1_0_sh_mask.h"
+#include "soc15_common.h"
+
+static uint32_t athub_v4_1_0_get_cg_cntl(struct amdgpu_device *adev)
+{
+	uint32_t data;
+
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
+	case IP_VERSION(4, 1, 0):
+		data = RREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL);
+		break;
+	default:
+		data = 0;
+		break;
+	}
+	return data;
+}
+
+static void athub_v4_1_0_set_cg_cntl(struct amdgpu_device *adev, uint32_t data)
+{
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
+	case IP_VERSION(4, 1, 0):
+		WREG32_SOC15(ATHUB, 0, regATHUB_MISC_CNTL, data);
+		break;
+	default:
+		break;
+	}
+}
+
+static void
+athub_v4_1_0_update_medium_grain_clock_gating(struct amdgpu_device *adev,
+					      bool enable)
+{
+	uint32_t def, data;
+
+	def = data = athub_v4_1_0_get_cg_cntl(adev);
+
+	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_ATHUB_MGCG))
+		data |= ATHUB_MISC_CNTL__CG_ENABLE_MASK;
+	else
+		data &= ~ATHUB_MISC_CNTL__CG_ENABLE_MASK;
+
+	if (def != data)
+		athub_v4_1_0_set_cg_cntl(adev, data);
+}
+
+static void
+athub_v4_1_0_update_medium_grain_light_sleep(struct amdgpu_device *adev,
+					     bool enable)
+{
+	uint32_t def, data;
+
+	def = data = athub_v4_1_0_get_cg_cntl(adev);
+
+	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_ATHUB_LS))
+		data |= ATHUB_MISC_CNTL__CG_MEM_LS_ENABLE_MASK;
+	else
+		data &= ~ATHUB_MISC_CNTL__CG_MEM_LS_ENABLE_MASK;
+
+	if (def != data)
+		athub_v4_1_0_set_cg_cntl(adev, data);
+}
+
+int athub_v4_1_0_set_clockgating(struct amdgpu_device *adev,
+				 enum amd_clockgating_state state)
+{
+	if (amdgpu_sriov_vf(adev))
+		return 0;
+
+	switch (amdgpu_ip_version(adev, ATHUB_HWIP, 0)) {
+	case IP_VERSION(4, 1, 0):
+		athub_v4_1_0_update_medium_grain_clock_gating(adev,
+				state == AMD_CG_STATE_GATE);
+		athub_v4_1_0_update_medium_grain_light_sleep(adev,
+				state == AMD_CG_STATE_GATE);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+void athub_v4_1_0_get_clockgating(struct amdgpu_device *adev, u64 *flags)
+{
+	int data;
+
+	/* AMD_CG_SUPPORT_ATHUB_MGCG */
+	data = athub_v4_1_0_get_cg_cntl(adev);
+	if (data & ATHUB_MISC_CNTL__CG_ENABLE_MASK)
+		*flags |= AMD_CG_SUPPORT_ATHUB_MGCG;
+
+	/* AMD_CG_SUPPORT_ATHUB_LS */
+	if (data & ATHUB_MISC_CNTL__CG_MEM_LS_ENABLE_MASK)
+		*flags |= AMD_CG_SUPPORT_ATHUB_LS;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/athub_v4_1_0.h b/drivers/gpu/drm/amd/amdgpu/athub_v4_1_0.h
new file mode 100644
index 000000000000..4d18d0998fa8
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/athub_v4_1_0.h
@@ -0,0 +1,30 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __ATHUB_V4_1_0_H__
+#define __ATHUB_V4_1_0_H__
+
+int athub_v4_1_0_set_clockgating(struct amdgpu_device *adev,
+				 enum amd_clockgating_state state);
+void athub_v4_1_0_get_clockgating(struct amdgpu_device *adev, u64 *flags);
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/atom.c b/drivers/gpu/drm/amd/amdgpu/atom.c
index 1c5d9388ad0b..7a063e44d429 100644
--- a/drivers/gpu/drm/amd/amdgpu/atom.c
+++ b/drivers/gpu/drm/amd/amdgpu/atom.c
@@ -27,7 +27,7 @@
 #include <linux/slab.h>
 #include <linux/string_helpers.h>
 
-#include <asm/unaligned.h>
+#include <linux/unaligned.h>
 
 #include <drm/drm_util.h>
 
@@ -62,6 +62,7 @@
 typedef struct {
 	struct atom_context *ctx;
 	uint32_t *ps, *ws;
+	int ps_size, ws_size;
 	int ps_shift;
 	uint16_t start;
 	unsigned last_jump;
@@ -70,8 +71,8 @@ typedef struct {
 } atom_exec_context;
 
 int amdgpu_atom_debug;
-static int amdgpu_atom_execute_table_locked(struct atom_context *ctx, int index, uint32_t *params);
-int amdgpu_atom_execute_table(struct atom_context *ctx, int index, uint32_t *params);
+static int amdgpu_atom_execute_table_locked(struct atom_context *ctx, int index, uint32_t *params, int params_size);
+int amdgpu_atom_execute_table(struct atom_context *ctx, int index, uint32_t *params, int params_size);
 
 static uint32_t atom_arg_mask[8] =
 	{ 0xFFFFFFFF, 0xFFFF, 0xFFFF00, 0xFFFF0000, 0xFF, 0xFF00, 0xFF0000,
@@ -223,7 +224,10 @@ static uint32_t atom_get_src_int(atom_exec_context *ctx, uint8_t attr,
 		(*ptr)++;
 		/* get_unaligned_le32 avoids unaligned accesses from atombios
 		 * tables, noticed on a DEC Alpha. */
-		val = get_unaligned_le32((u32 *)&ctx->ps[idx]);
+		if (idx < ctx->ps_size)
+			val = get_unaligned_le32((u32 *)&ctx->ps[idx]);
+		else
+			pr_info("PS index out of range: %i > %i\n", idx, ctx->ps_size);
 		if (print)
 			DEBUG("PS[0x%02X,0x%04X]", idx, val);
 		break;
@@ -261,7 +265,10 @@ static uint32_t atom_get_src_int(atom_exec_context *ctx, uint8_t attr,
 			val = gctx->reg_block;
 			break;
 		default:
-			val = ctx->ws[idx];
+			if (idx < ctx->ws_size)
+				val = ctx->ws[idx];
+			else
+				pr_info("WS index out of range: %i > %i\n", idx, ctx->ws_size);
 		}
 		break;
 	case ATOM_ARG_ID:
@@ -294,7 +301,7 @@ static uint32_t atom_get_src_int(atom_exec_context *ctx, uint8_t attr,
 			(*ptr) += 4;
 			if (print)
 				DEBUG("IMM 0x%08X\n", val);
-			return val;
+			break;
 		case ATOM_SRC_WORD0:
 		case ATOM_SRC_WORD8:
 		case ATOM_SRC_WORD16:
@@ -302,7 +309,7 @@ static uint32_t atom_get_src_int(atom_exec_context *ctx, uint8_t attr,
 			(*ptr) += 2;
 			if (print)
 				DEBUG("IMM 0x%04X\n", val);
-			return val;
+			break;
 		case ATOM_SRC_BYTE0:
 		case ATOM_SRC_BYTE8:
 		case ATOM_SRC_BYTE16:
@@ -311,9 +318,9 @@ static uint32_t atom_get_src_int(atom_exec_context *ctx, uint8_t attr,
 			(*ptr)++;
 			if (print)
 				DEBUG("IMM 0x%02X\n", val);
-			return val;
+			break;
 		}
-		return 0;
+		return val;
 	case ATOM_ARG_PLL:
 		idx = U8(*ptr);
 		(*ptr)++;
@@ -395,7 +402,6 @@ static void atom_skip_src_int(atom_exec_context *ctx, uint8_t attr, int *ptr)
 			(*ptr)++;
 			return;
 		}
-		return;
 	}
 }
 
@@ -496,6 +502,10 @@ static void atom_put_dst(atom_exec_context *ctx, int arg, uint8_t attr,
 		idx = U8(*ptr);
 		(*ptr)++;
 		DEBUG("PS[0x%02X]", idx);
+		if (idx >= ctx->ps_size) {
+			pr_info("PS index out of range: %i > %i\n", idx, ctx->ps_size);
+			return;
+		}
 		ctx->ps[idx] = cpu_to_le32(val);
 		break;
 	case ATOM_ARG_WS:
@@ -528,6 +538,10 @@ static void atom_put_dst(atom_exec_context *ctx, int arg, uint8_t attr,
 			gctx->reg_block = val;
 			break;
 		default:
+			if (idx >= ctx->ws_size) {
+				pr_info("WS index out of range: %i > %i\n", idx, ctx->ws_size);
+				return;
+			}
 			ctx->ws[idx] = val;
 		}
 		break;
@@ -625,7 +639,7 @@ static void atom_op_calltable(atom_exec_context *ctx, int *ptr, int arg)
 	else
 		SDEBUG("   table: %d\n", idx);
 	if (U16(ctx->ctx->cmd_table + 4 + 2 * idx))
-		r = amdgpu_atom_execute_table_locked(ctx->ctx, idx, ctx->ps + ctx->ps_shift);
+		r = amdgpu_atom_execute_table_locked(ctx->ctx, idx, ctx->ps + ctx->ps_shift, ctx->ps_size - ctx->ps_shift);
 	if (r) {
 		ctx->abort = true;
 	}
@@ -1204,7 +1218,7 @@ static struct {
 	atom_op_div32, ATOM_ARG_WS},
 };
 
-static int amdgpu_atom_execute_table_locked(struct atom_context *ctx, int index, uint32_t *params)
+static int amdgpu_atom_execute_table_locked(struct atom_context *ctx, int index, uint32_t *params, int params_size)
 {
 	int base = CU16(ctx->cmd_table + 4 + 2 * index);
 	int len, ws, ps, ptr;
@@ -1226,12 +1240,21 @@ static int amdgpu_atom_execute_table_locked(struct atom_context *ctx, int index,
 	ectx.ps_shift = ps / 4;
 	ectx.start = base;
 	ectx.ps = params;
+	ectx.ps_size = params_size;
 	ectx.abort = false;
 	ectx.last_jump = 0;
-	if (ws)
+	ectx.last_jump_jiffies = 0;
+	if (ws) {
 		ectx.ws = kcalloc(4, ws, GFP_KERNEL);
-	else
+		if (!ectx.ws) {
+			ret = -ENOMEM;
+			goto free;
+		}
+		ectx.ws_size = ws;
+	} else {
 		ectx.ws = NULL;
+		ectx.ws_size = 0;
+	}
 
 	debug_depth++;
 	while (1) {
@@ -1265,7 +1288,7 @@ free:
 	return ret;
 }
 
-int amdgpu_atom_execute_table(struct atom_context *ctx, int index, uint32_t *params)
+int amdgpu_atom_execute_table(struct atom_context *ctx, int index, uint32_t *params, int params_size)
 {
 	int r;
 
@@ -1281,7 +1304,7 @@ int amdgpu_atom_execute_table(struct atom_context *ctx, int index, uint32_t *par
 	/* reset divmul */
 	ctx->divmul[0] = 0;
 	ctx->divmul[1] = 0;
-	r = amdgpu_atom_execute_table_locked(ctx, index, params);
+	r = amdgpu_atom_execute_table_locked(ctx, index, params, params_size);
 	mutex_unlock(&ctx->mutex);
 	return r;
 }
@@ -1425,6 +1448,7 @@ static void atom_get_vbios_pn(struct atom_context *ctx)
 		if (vbios_str == NULL)
 			vbios_str += sizeof(BIOS_ATOM_PREFIX) - 1;
 	}
+	OPTIMIZER_HIDE_VAR(vbios_str);
 	if (vbios_str != NULL && *vbios_str == 0)
 		vbios_str++;
 
@@ -1438,14 +1462,33 @@ static void atom_get_vbios_pn(struct atom_context *ctx)
 
 		ctx->vbios_pn[count] = 0;
 	}
+
+	pr_info("ATOM BIOS: %s\n", ctx->vbios_pn);
 }
 
 static void atom_get_vbios_version(struct atom_context *ctx)
 {
+	unsigned short start = 3, end;
 	unsigned char *vbios_ver;
+	unsigned char *p_rom;
+
+	p_rom = ctx->bios;
+	/* Search from strings offset if it's present */
+	start = *(unsigned short *)(p_rom +
+				    OFFSET_TO_GET_ATOMBIOS_STRING_START);
+
+	/* Search till atom rom header start point */
+	end = *(unsigned short *)(p_rom + OFFSET_TO_ATOM_ROM_HEADER_POINTER);
+
+	/* Use hardcoded offsets, if the offsets are not populated */
+	if (end <= start) {
+		start = 3;
+		end = 1024;
+	}
 
 	/* find anchor ATOMBIOSBK-AMD */
-	vbios_ver = atom_find_str_in_rom(ctx, BIOS_VERSION_PREFIX, 3, 1024, 64);
+	vbios_ver =
+		atom_find_str_in_rom(ctx, BIOS_VERSION_PREFIX, start, end, 64);
 	if (vbios_ver != NULL) {
 		/* skip ATOMBIOSBK-AMD VER */
 		vbios_ver += 18;
@@ -1455,16 +1498,36 @@ static void atom_get_vbios_version(struct atom_context *ctx)
 	}
 }
 
+static void atom_get_vbios_build(struct atom_context *ctx)
+{
+	unsigned char *atom_rom_hdr;
+	unsigned char *str;
+	uint16_t base, len;
+
+	base = CU16(ATOM_ROM_TABLE_PTR);
+	atom_rom_hdr = CSTR(base);
+
+	str = CSTR(CU16(base + ATOM_ROM_CFG_PTR));
+	/* Skip config string */
+	while (str < atom_rom_hdr && *str++)
+		;
+	/* Skip change list string */
+	while (str < atom_rom_hdr && *str++)
+		;
+
+	len = min(atom_rom_hdr - str, STRLEN_NORMAL);
+	if (len)
+		strscpy(ctx->build_num, str, len);
+}
+
 struct atom_context *amdgpu_atom_parse(struct card_info *card, void *bios)
 {
 	int base;
 	struct atom_context *ctx =
 	    kzalloc(sizeof(struct atom_context), GFP_KERNEL);
-	char *str;
 	struct _ATOM_ROM_HEADER *atom_rom_header;
 	struct _ATOM_MASTER_DATA_TABLE *master_table;
 	struct _ATOM_FIRMWARE_INFO *atom_fw_info;
-	u16 idx;
 
 	if (!ctx)
 		return NULL;
@@ -1502,16 +1565,6 @@ struct atom_context *amdgpu_atom_parse(struct card_info *card, void *bios)
 		return NULL;
 	}
 
-	idx = CU16(ATOM_ROM_PART_NUMBER_PTR);
-	if (idx == 0)
-		idx = 0x80;
-
-	str = CSTR(idx);
-	if (*str != '\0') {
-		pr_info("ATOM BIOS: %s\n", str);
-		strlcpy(ctx->vbios_version, str, sizeof(ctx->vbios_version));
-	}
-
 	atom_rom_header = (struct _ATOM_ROM_HEADER *)CSTR(base);
 	if (atom_rom_header->usMasterDataTableOffset != 0) {
 		master_table = (struct _ATOM_MASTER_DATA_TABLE *)
@@ -1527,6 +1580,7 @@ struct atom_context *amdgpu_atom_parse(struct card_info *card, void *bios)
 	atom_get_vbios_pn(ctx);
 	atom_get_vbios_date(ctx);
 	atom_get_vbios_version(ctx);
+	atom_get_vbios_build(ctx);
 
 	return ctx;
 }
@@ -1546,7 +1600,7 @@ int amdgpu_atom_asic_init(struct atom_context *ctx)
 
 	if (!CU16(ctx->cmd_table + 4 + 2 * ATOM_CMD_INIT))
 		return 1;
-	ret = amdgpu_atom_execute_table(ctx, ATOM_CMD_INIT, ps);
+	ret = amdgpu_atom_execute_table(ctx, ATOM_CMD_INIT, ps, 16);
 	if (ret)
 		return ret;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/atom.h b/drivers/gpu/drm/amd/amdgpu/atom.h
index 0c1839824520..825ff28731f5 100644
--- a/drivers/gpu/drm/amd/amdgpu/atom.h
+++ b/drivers/gpu/drm/amd/amdgpu/atom.h
@@ -33,11 +33,11 @@ struct drm_device;
 #define ATOM_ATI_MAGIC_PTR	0x30
 #define ATOM_ATI_MAGIC		" 761295520"
 #define ATOM_ROM_TABLE_PTR	0x48
-#define ATOM_ROM_PART_NUMBER_PTR	0x6E
 
 #define ATOM_ROM_MAGIC		"ATOM"
 #define ATOM_ROM_MAGIC_PTR	4
 
+#define ATOM_ROM_CFG_PTR 0xC
 #define ATOM_ROM_MSG_PTR	0x10
 #define ATOM_ROM_CMD_PTR	0x1E
 #define ATOM_ROM_DATA_PTR	0x20
@@ -118,12 +118,15 @@ struct drm_device;
 
 struct card_info {
 	struct drm_device *dev;
-	void (* reg_write)(struct card_info *, uint32_t, uint32_t);   /*  filled by driver */
-	uint32_t (* reg_read)(struct card_info *, uint32_t);          /*  filled by driver */
-	void (* mc_write)(struct card_info *, uint32_t, uint32_t);   /*  filled by driver */
-	uint32_t (* mc_read)(struct card_info *, uint32_t);          /*  filled by driver */
-	void (* pll_write)(struct card_info *, uint32_t, uint32_t);   /*  filled by driver */
-	uint32_t (* pll_read)(struct card_info *, uint32_t);          /*  filled by driver */
+	void (*reg_write)(struct card_info *info,
+			  u32 reg, uint32_t val);   /*  filled by driver */
+	uint32_t (*reg_read)(struct card_info *info, uint32_t reg);          /*  filled by driver */
+	void (*mc_write)(struct card_info *info,
+			 u32 reg, uint32_t val);   /*  filled by driver */
+	uint32_t (*mc_read)(struct card_info *info, uint32_t reg);          /*  filled by driver */
+	void (*pll_write)(struct card_info *info,
+			  u32 reg, uint32_t val);   /*  filled by driver */
+	uint32_t (*pll_read)(struct card_info *info, uint32_t reg);          /*  filled by driver */
 };
 
 struct atom_context {
@@ -143,21 +146,21 @@ struct atom_context {
 	int io_mode;
 	uint32_t *scratch;
 	int scratch_size_bytes;
-	char vbios_version[20];
 
 	uint8_t name[STRLEN_LONG];
 	uint8_t vbios_pn[STRLEN_LONG];
 	uint32_t version;
 	uint8_t vbios_ver_str[STRLEN_NORMAL];
 	uint8_t date[STRLEN_NORMAL];
+	uint8_t build_num[STRLEN_NORMAL];
 };
 
 extern int amdgpu_atom_debug;
 
-struct atom_context *amdgpu_atom_parse(struct card_info *, void *);
-int amdgpu_atom_execute_table(struct atom_context *, int, uint32_t *);
-int amdgpu_atom_asic_init(struct atom_context *);
-void amdgpu_atom_destroy(struct atom_context *);
+struct atom_context *amdgpu_atom_parse(struct card_info *card, void *bios);
+int amdgpu_atom_execute_table(struct atom_context *ctx, int index, uint32_t *params, int params_size);
+int amdgpu_atom_asic_init(struct atom_context *ctx);
+void amdgpu_atom_destroy(struct atom_context *ctx);
 bool amdgpu_atom_parse_data_header(struct atom_context *ctx, int index, uint16_t *size,
 			    uint8_t *frev, uint8_t *crev, uint16_t *data_start);
 bool amdgpu_atom_parse_cmd_header(struct atom_context *ctx, int index,
diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_crtc.c b/drivers/gpu/drm/amd/amdgpu/atombios_crtc.c
index 10098fdd33fc..3dfc28840a7d 100644
--- a/drivers/gpu/drm/amd/amdgpu/atombios_crtc.c
+++ b/drivers/gpu/drm/amd/amdgpu/atombios_crtc.c
@@ -77,7 +77,7 @@ void amdgpu_atombios_crtc_overscan_setup(struct drm_crtc *crtc,
 		args.usOverscanTop = cpu_to_le16(amdgpu_crtc->v_border);
 		break;
 	}
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 void amdgpu_atombios_crtc_scaler_setup(struct drm_crtc *crtc)
@@ -106,7 +106,7 @@ void amdgpu_atombios_crtc_scaler_setup(struct drm_crtc *crtc)
 		args.ucEnable = ATOM_SCALER_DISABLE;
 		break;
 	}
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 void amdgpu_atombios_crtc_lock(struct drm_crtc *crtc, int lock)
@@ -123,7 +123,7 @@ void amdgpu_atombios_crtc_lock(struct drm_crtc *crtc, int lock)
 	args.ucCRTC = amdgpu_crtc->crtc_id;
 	args.ucEnable = lock;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 void amdgpu_atombios_crtc_enable(struct drm_crtc *crtc, int state)
@@ -139,7 +139,7 @@ void amdgpu_atombios_crtc_enable(struct drm_crtc *crtc, int state)
 	args.ucCRTC = amdgpu_crtc->crtc_id;
 	args.ucEnable = state;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 void amdgpu_atombios_crtc_blank(struct drm_crtc *crtc, int state)
@@ -155,7 +155,7 @@ void amdgpu_atombios_crtc_blank(struct drm_crtc *crtc, int state)
 	args.ucCRTC = amdgpu_crtc->crtc_id;
 	args.ucBlanking = state;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 void amdgpu_atombios_crtc_powergate(struct drm_crtc *crtc, int state)
@@ -171,7 +171,7 @@ void amdgpu_atombios_crtc_powergate(struct drm_crtc *crtc, int state)
 	args.ucDispPipeId = amdgpu_crtc->crtc_id;
 	args.ucEnable = state;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 void amdgpu_atombios_crtc_powergate_init(struct amdgpu_device *adev)
@@ -183,7 +183,7 @@ void amdgpu_atombios_crtc_powergate_init(struct amdgpu_device *adev)
 
 	args.ucEnable = ATOM_INIT;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 void amdgpu_atombios_crtc_set_dtd_timing(struct drm_crtc *crtc,
@@ -228,7 +228,7 @@ void amdgpu_atombios_crtc_set_dtd_timing(struct drm_crtc *crtc,
 	args.susModeMiscInfo.usAccess = cpu_to_le16(misc);
 	args.ucCRTC = amdgpu_crtc->crtc_id;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 union atom_enable_ss {
@@ -293,7 +293,7 @@ static void amdgpu_atombios_crtc_program_ss(struct amdgpu_device *adev,
 	args.v3.usSpreadSpectrumStep = cpu_to_le16(ss->step);
 	args.v3.ucEnable = enable;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 union adjust_pixel_clock {
@@ -395,7 +395,7 @@ static u32 amdgpu_atombios_crtc_adjust_pll(struct drm_crtc *crtc,
 					ADJUST_DISPLAY_CONFIG_SS_ENABLE;
 
 			amdgpu_atom_execute_table(adev->mode_info.atom_context,
-					   index, (uint32_t *)&args);
+					   index, (uint32_t *)&args, sizeof(args));
 			adjusted_clock = le16_to_cpu(args.v1.usPixelClock) * 10;
 			break;
 		case 3:
@@ -428,7 +428,7 @@ static u32 amdgpu_atombios_crtc_adjust_pll(struct drm_crtc *crtc,
 				args.v3.sInput.ucExtTransmitterID = 0;
 
 			amdgpu_atom_execute_table(adev->mode_info.atom_context,
-					   index, (uint32_t *)&args);
+					   index, (uint32_t *)&args, sizeof(args));
 			adjusted_clock = le32_to_cpu(args.v3.sOutput.ulDispPllFreq) * 10;
 			if (args.v3.sOutput.ucRefDiv) {
 				amdgpu_crtc->pll_flags |= AMDGPU_PLL_USE_FRAC_FB_DIV;
@@ -514,7 +514,7 @@ void amdgpu_atombios_crtc_set_disp_eng_pll(struct amdgpu_device *adev,
 		DRM_ERROR("Unknown table version %d %d\n", frev, crev);
 		return;
 	}
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 union set_dce_clock {
@@ -544,7 +544,7 @@ u32 amdgpu_atombios_crtc_set_dce_clock(struct amdgpu_device *adev,
 			args.v2_1.asParam.ulDCEClkFreq = cpu_to_le32(freq); /* 10kHz units */
 			args.v2_1.asParam.ucDCEClkType = clk_type;
 			args.v2_1.asParam.ucDCEClkSrc = clk_src;
-			amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+			amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 			ret_freq = le32_to_cpu(args.v2_1.asParam.ulDCEClkFreq) * 10;
 			break;
 		default:
@@ -740,7 +740,7 @@ void amdgpu_atombios_crtc_program_pll(struct drm_crtc *crtc,
 		return;
 	}
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 int amdgpu_atombios_crtc_prepare_pll(struct drm_crtc *crtc,
diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_dp.c b/drivers/gpu/drm/amd/amdgpu/atombios_dp.c
index 87c41e0e9b7c..492813ab1b54 100644
--- a/drivers/gpu/drm/amd/amdgpu/atombios_dp.c
+++ b/drivers/gpu/drm/amd/amdgpu/atombios_dp.c
@@ -83,7 +83,7 @@ static int amdgpu_atombios_dp_process_aux_ch(struct amdgpu_i2c_chan *chan,
 	args.v2.ucDelay = delay / 10;
 	args.v2.ucHPD_ID = chan->rec.hpd;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 
 	*ack = args.v2.ucReplyStatus;
 
@@ -301,7 +301,7 @@ static u8 amdgpu_atombios_dp_encoder_service(struct amdgpu_device *adev,
 	args.ucLaneNum = lane_num;
 	args.ucStatus = 0;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 	return args.ucStatus;
 }
 
@@ -430,7 +430,7 @@ void amdgpu_atombios_dp_set_link_config(struct drm_connector *connector,
 }
 
 int amdgpu_atombios_dp_mode_valid_helper(struct drm_connector *connector,
-				  struct drm_display_mode *mode)
+				  const struct drm_display_mode *mode)
 {
 	struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
 	struct amdgpu_connector_atom_dig *dig_connector;
@@ -458,8 +458,8 @@ bool amdgpu_atombios_dp_needs_link_train(struct amdgpu_connector *amdgpu_connect
 	u8 link_status[DP_LINK_STATUS_SIZE];
 	struct amdgpu_connector_atom_dig *dig = amdgpu_connector->con_priv;
 
-	if (drm_dp_dpcd_read_link_status(&amdgpu_connector->ddc_bus->aux, link_status)
-	    <= 0)
+	if (drm_dp_dpcd_read_link_status(&amdgpu_connector->ddc_bus->aux,
+					 link_status) < 0)
 		return false;
 	if (drm_dp_channel_eq_ok(link_status, dig->dp_lane_count))
 		return false;
@@ -616,7 +616,7 @@ amdgpu_atombios_dp_link_train_cr(struct amdgpu_atombios_dp_link_train_info *dp_i
 		drm_dp_link_train_clock_recovery_delay(dp_info->aux, dp_info->dpcd);
 
 		if (drm_dp_dpcd_read_link_status(dp_info->aux,
-						 dp_info->link_status) <= 0) {
+						 dp_info->link_status) < 0) {
 			DRM_ERROR("displayport link status failed\n");
 			break;
 		}
@@ -681,7 +681,7 @@ amdgpu_atombios_dp_link_train_ce(struct amdgpu_atombios_dp_link_train_info *dp_i
 		drm_dp_link_train_channel_eq_delay(dp_info->aux, dp_info->dpcd);
 
 		if (drm_dp_dpcd_read_link_status(dp_info->aux,
-						 dp_info->link_status) <= 0) {
+						 dp_info->link_status) < 0) {
 			DRM_ERROR("displayport link status failed\n");
 			break;
 		}
diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_dp.h b/drivers/gpu/drm/amd/amdgpu/atombios_dp.h
index f59d85eaddf0..3e24acf8133f 100644
--- a/drivers/gpu/drm/amd/amdgpu/atombios_dp.h
+++ b/drivers/gpu/drm/amd/amdgpu/atombios_dp.h
@@ -32,7 +32,7 @@ int amdgpu_atombios_dp_get_panel_mode(struct drm_encoder *encoder,
 void amdgpu_atombios_dp_set_link_config(struct drm_connector *connector,
 				 const struct drm_display_mode *mode);
 int amdgpu_atombios_dp_mode_valid_helper(struct drm_connector *connector,
-				  struct drm_display_mode *mode);
+				  const struct drm_display_mode *mode);
 bool amdgpu_atombios_dp_needs_link_train(struct amdgpu_connector *amdgpu_connector);
 void amdgpu_atombios_dp_set_rx_power_state(struct drm_connector *connector,
 				    u8 power_state);
diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c b/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c
index d95b2dc78063..a51f3414b65d 100644
--- a/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c
+++ b/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c
@@ -28,6 +28,7 @@
 
 #include <acpi/video.h>
 
+#include <drm/drm_edid.h>
 #include <drm/amdgpu_drm.h>
 #include "amdgpu.h"
 #include "amdgpu_connectors.h"
@@ -214,7 +215,7 @@ void amdgpu_atombios_encoder_init_backlight(struct amdgpu_encoder *amdgpu_encode
 	dig->bl_dev = bd;
 
 	bd->props.brightness = amdgpu_atombios_encoder_get_backlight_brightness(bd);
-	bd->props.power = FB_BLANK_UNBLANK;
+	bd->props.power = BACKLIGHT_POWER_ON;
 	backlight_update_status(bd);
 
 	DRM_INFO("amdgpu atom DIG backlight initialized\n");
@@ -228,7 +229,6 @@ error:
 register_acpi_backlight:
 	/* Try registering an ACPI video backlight device instead. */
 	acpi_video_register_backlight();
-	return;
 }
 
 void
@@ -335,7 +335,7 @@ amdgpu_atombios_encoder_setup_dac(struct drm_encoder *encoder, int action)
 	args.ucDacStandard = ATOM_DAC1_PS2;
 	args.usPixelClock = cpu_to_le16(amdgpu_encoder->pixel_clock / 10);
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 
 }
 
@@ -432,7 +432,7 @@ amdgpu_atombios_encoder_setup_dvo(struct drm_encoder *encoder, int action)
 		break;
 	}
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 int amdgpu_atombios_encoder_get_encoder_mode(struct drm_encoder *encoder)
@@ -732,7 +732,7 @@ amdgpu_atombios_encoder_setup_dig_encoder(struct drm_encoder *encoder,
 		break;
 	}
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 
 }
 
@@ -1136,7 +1136,7 @@ amdgpu_atombios_encoder_setup_dig_transmitter(struct drm_encoder *encoder, int a
 		break;
 	}
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 bool
@@ -1164,7 +1164,7 @@ amdgpu_atombios_encoder_set_edp_panel_power(struct drm_connector *connector,
 
 	args.v1.ucAction = action;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 
 	/* wait for the panel to power up */
 	if (action == ATOM_TRANSMITTER_ACTION_POWER_ON) {
@@ -1288,7 +1288,7 @@ amdgpu_atombios_encoder_setup_external_encoder(struct drm_encoder *encoder,
 		DRM_ERROR("Unknown table version: %d, %d\n", frev, crev);
 		return;
 	}
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 static void
@@ -1633,7 +1633,7 @@ amdgpu_atombios_encoder_set_crtc_source(struct drm_encoder *encoder)
 		return;
 	}
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
 
 /* This only needs to be called once at startup */
@@ -1706,7 +1706,7 @@ amdgpu_atombios_encoder_dac_load_detect(struct drm_encoder *encoder,
 				args.sDacload.ucMisc = DAC_LOAD_MISC_YPrPb;
 		}
 
-		amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+		amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 
 		return true;
 	} else
@@ -2064,27 +2064,25 @@ amdgpu_atombios_encoder_get_lcd_info(struct amdgpu_encoder *encoder)
 				case LCD_FAKE_EDID_PATCH_RECORD_TYPE:
 					fake_edid_record = (ATOM_FAKE_EDID_PATCH_RECORD *)record;
 					if (fake_edid_record->ucFakeEDIDLength) {
-						struct edid *edid;
-						int edid_size =
-							max((int)EDID_LENGTH, (int)fake_edid_record->ucFakeEDIDLength);
-						edid = kmalloc(edid_size, GFP_KERNEL);
-						if (edid) {
-							memcpy((u8 *)edid, (u8 *)&fake_edid_record->ucFakeEDIDString[0],
-							       fake_edid_record->ucFakeEDIDLength);
-
-							if (drm_edid_is_valid(edid)) {
-								adev->mode_info.bios_hardcoded_edid = edid;
-								adev->mode_info.bios_hardcoded_edid_size = edid_size;
-							} else
-								kfree(edid);
-						}
+						const struct drm_edid *edid;
+						int edid_size;
+
+						if (fake_edid_record->ucFakeEDIDLength == 128)
+							edid_size = fake_edid_record->ucFakeEDIDLength;
+						else
+							edid_size = fake_edid_record->ucFakeEDIDLength * 128;
+						edid = drm_edid_alloc(fake_edid_record->ucFakeEDIDString, edid_size);
+						if (drm_edid_valid(edid))
+							adev->mode_info.bios_hardcoded_edid = edid;
+						else
+							drm_edid_free(edid);
+						record += struct_size(fake_edid_record,
+								      ucFakeEDIDString,
+								      edid_size);
+					} else {
+						/* empty fake edid record must be 3 bytes long */
+						record += sizeof(ATOM_FAKE_EDID_PATCH_RECORD) + 1;
 					}
-					record += fake_edid_record->ucFakeEDIDLength ?
-						  struct_size(fake_edid_record,
-							      ucFakeEDIDString,
-							      fake_edid_record->ucFakeEDIDLength) :
-						  /* empty fake edid record must be 3 bytes long */
-						  sizeof(ATOM_FAKE_EDID_PATCH_RECORD) + 1;
 					break;
 				case LCD_PANEL_RESOLUTION_RECORD_TYPE:
 					panel_res_record = (ATOM_PANEL_RESOLUTION_PATCH_RECORD *)record;
diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_i2c.c b/drivers/gpu/drm/amd/amdgpu/atombios_i2c.c
index af0335535f82..a6501114322f 100644
--- a/drivers/gpu/drm/amd/amdgpu/atombios_i2c.c
+++ b/drivers/gpu/drm/amd/amdgpu/atombios_i2c.c
@@ -86,7 +86,7 @@ static int amdgpu_atombios_i2c_process_i2c_ch(struct amdgpu_i2c_chan *chan,
 	args.ucSlaveAddr = slave_addr << 1;
 	args.ucLineNumber = chan->rec.i2c_id;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 
 	/* error */
 	if (args.ucStatus != HW_ASSISTED_I2C_STATUS_SUCCESS) {
@@ -172,5 +172,5 @@ void amdgpu_atombios_i2c_channel_trans(struct amdgpu_device *adev, u8 slave_addr
 	args.ucSlaveAddr = slave_addr;
 	args.ucLineNumber = line_number;
 
-	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args);
+	amdgpu_atom_execute_table(adev->mode_info.atom_context, index, (uint32_t *)&args, sizeof(args));
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/cik.c b/drivers/gpu/drm/amd/amdgpu/cik.c
index de6d10390ab2..9cd63b4177bf 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik.c
+++ b/drivers/gpu/drm/amd/amdgpu/cik.c
@@ -1141,12 +1141,12 @@ static uint32_t cik_get_register_value(struct amdgpu_device *adev,
 
 		mutex_lock(&adev->grbm_idx_mutex);
 		if (se_num != 0xffffffff || sh_num != 0xffffffff)
-			amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff);
+			amdgpu_gfx_select_se_sh(adev, se_num, sh_num, 0xffffffff, 0);
 
 		val = RREG32(reg_offset);
 
 		if (se_num != 0xffffffff || sh_num != 0xffffffff)
-			amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+			amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 		mutex_unlock(&adev->grbm_idx_mutex);
 		return val;
 	} else {
@@ -1375,14 +1375,14 @@ static int cik_asic_pci_config_reset(struct amdgpu_device *adev)
 	return r;
 }
 
-static bool cik_asic_supports_baco(struct amdgpu_device *adev)
+static int cik_asic_supports_baco(struct amdgpu_device *adev)
 {
 	switch (adev->asic_type) {
 	case CHIP_BONAIRE:
 	case CHIP_HAWAII:
 		return amdgpu_dpm_is_baco_supported(adev);
 	default:
-		return false;
+		return 0;
 	}
 }
 
@@ -1574,17 +1574,8 @@ static void cik_pcie_gen3_enable(struct amdgpu_device *adev)
 			u16 bridge_cfg2, gpu_cfg2;
 			u32 max_lw, current_lw, tmp;
 
-			pcie_capability_read_word(root, PCI_EXP_LNKCTL,
-						  &bridge_cfg);
-			pcie_capability_read_word(adev->pdev, PCI_EXP_LNKCTL,
-						  &gpu_cfg);
-
-			tmp16 = bridge_cfg | PCI_EXP_LNKCTL_HAWD;
-			pcie_capability_write_word(root, PCI_EXP_LNKCTL, tmp16);
-
-			tmp16 = gpu_cfg | PCI_EXP_LNKCTL_HAWD;
-			pcie_capability_write_word(adev->pdev, PCI_EXP_LNKCTL,
-						   tmp16);
+			pcie_capability_set_word(root, PCI_EXP_LNKCTL, PCI_EXP_LNKCTL_HAWD);
+			pcie_capability_set_word(adev->pdev, PCI_EXP_LNKCTL, PCI_EXP_LNKCTL_HAWD);
 
 			tmp = RREG32_PCIE(ixPCIE_LC_STATUS1);
 			max_lw = (tmp & PCIE_LC_STATUS1__LC_DETECTED_LINK_WIDTH_MASK) >>
@@ -1637,45 +1628,28 @@ static void cik_pcie_gen3_enable(struct amdgpu_device *adev)
 				msleep(100);
 
 				/* linkctl */
-				pcie_capability_read_word(root, PCI_EXP_LNKCTL,
-							  &tmp16);
-				tmp16 &= ~PCI_EXP_LNKCTL_HAWD;
-				tmp16 |= (bridge_cfg & PCI_EXP_LNKCTL_HAWD);
-				pcie_capability_write_word(root, PCI_EXP_LNKCTL,
-							   tmp16);
-
-				pcie_capability_read_word(adev->pdev,
-							  PCI_EXP_LNKCTL,
-							  &tmp16);
-				tmp16 &= ~PCI_EXP_LNKCTL_HAWD;
-				tmp16 |= (gpu_cfg & PCI_EXP_LNKCTL_HAWD);
-				pcie_capability_write_word(adev->pdev,
-							   PCI_EXP_LNKCTL,
-							   tmp16);
+				pcie_capability_clear_and_set_word(root, PCI_EXP_LNKCTL,
+								   PCI_EXP_LNKCTL_HAWD,
+								   bridge_cfg &
+								   PCI_EXP_LNKCTL_HAWD);
+				pcie_capability_clear_and_set_word(adev->pdev, PCI_EXP_LNKCTL,
+								   PCI_EXP_LNKCTL_HAWD,
+								   gpu_cfg &
+								   PCI_EXP_LNKCTL_HAWD);
 
 				/* linkctl2 */
-				pcie_capability_read_word(root, PCI_EXP_LNKCTL2,
-							  &tmp16);
-				tmp16 &= ~(PCI_EXP_LNKCTL2_ENTER_COMP |
-					   PCI_EXP_LNKCTL2_TX_MARGIN);
-				tmp16 |= (bridge_cfg2 &
-					  (PCI_EXP_LNKCTL2_ENTER_COMP |
-					   PCI_EXP_LNKCTL2_TX_MARGIN));
-				pcie_capability_write_word(root,
-							   PCI_EXP_LNKCTL2,
-							   tmp16);
-
-				pcie_capability_read_word(adev->pdev,
-							  PCI_EXP_LNKCTL2,
-							  &tmp16);
-				tmp16 &= ~(PCI_EXP_LNKCTL2_ENTER_COMP |
-					   PCI_EXP_LNKCTL2_TX_MARGIN);
-				tmp16 |= (gpu_cfg2 &
-					  (PCI_EXP_LNKCTL2_ENTER_COMP |
-					   PCI_EXP_LNKCTL2_TX_MARGIN));
-				pcie_capability_write_word(adev->pdev,
-							   PCI_EXP_LNKCTL2,
-							   tmp16);
+				pcie_capability_clear_and_set_word(root, PCI_EXP_LNKCTL2,
+								   PCI_EXP_LNKCTL2_ENTER_COMP |
+								   PCI_EXP_LNKCTL2_TX_MARGIN,
+								   bridge_cfg2 &
+								   (PCI_EXP_LNKCTL2_ENTER_COMP |
+								    PCI_EXP_LNKCTL2_TX_MARGIN));
+				pcie_capability_clear_and_set_word(adev->pdev, PCI_EXP_LNKCTL2,
+								   PCI_EXP_LNKCTL2_ENTER_COMP |
+								   PCI_EXP_LNKCTL2_TX_MARGIN,
+								   gpu_cfg2 &
+								   (PCI_EXP_LNKCTL2_ENTER_COMP |
+								    PCI_EXP_LNKCTL2_TX_MARGIN));
 
 				tmp = RREG32_PCIE(ixPCIE_LC_CNTL4);
 				tmp &= ~PCIE_LC_CNTL4__LC_SET_QUIESCE_MASK;
@@ -1690,16 +1664,15 @@ static void cik_pcie_gen3_enable(struct amdgpu_device *adev)
 	speed_cntl &= ~PCIE_LC_SPEED_CNTL__LC_FORCE_DIS_SW_SPEED_CHANGE_MASK;
 	WREG32_PCIE(ixPCIE_LC_SPEED_CNTL, speed_cntl);
 
-	pcie_capability_read_word(adev->pdev, PCI_EXP_LNKCTL2, &tmp16);
-	tmp16 &= ~PCI_EXP_LNKCTL2_TLS;
-
+	tmp16 = 0;
 	if (adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN3)
 		tmp16 |= PCI_EXP_LNKCTL2_TLS_8_0GT; /* gen3 */
 	else if (adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN2)
 		tmp16 |= PCI_EXP_LNKCTL2_TLS_5_0GT; /* gen2 */
 	else
 		tmp16 |= PCI_EXP_LNKCTL2_TLS_2_5GT; /* gen1 */
-	pcie_capability_write_word(adev->pdev, PCI_EXP_LNKCTL2, tmp16);
+	pcie_capability_clear_and_set_word(adev->pdev, PCI_EXP_LNKCTL2,
+					   PCI_EXP_LNKCTL2_TLS, tmp16);
 
 	speed_cntl = RREG32_PCIE(ixPCIE_LC_SPEED_CNTL);
 	speed_cntl |= PCIE_LC_SPEED_CNTL__LC_INITIATE_LINK_SPEED_CHANGE_MASK;
@@ -1725,10 +1698,6 @@ static void cik_program_aspm(struct amdgpu_device *adev)
 	if (pci_is_root_bus(adev->pdev->bus))
 		return;
 
-	/* XXX double check APUs */
-	if (adev->flags & AMD_IS_APU)
-		return;
-
 	orig = data = RREG32_PCIE(ixPCIE_LC_N_FTS_CNTL);
 	data &= ~PCIE_LC_N_FTS_CNTL__LC_XMIT_N_FTS_MASK;
 	data |= (0x24 << PCIE_LC_N_FTS_CNTL__LC_XMIT_N_FTS__SHIFT) |
@@ -2016,9 +1985,9 @@ static const struct amdgpu_asic_funcs cik_asic_funcs =
 	.query_video_codecs = &cik_query_video_codecs,
 };
 
-static int cik_common_early_init(void *handle)
+static int cik_common_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->smc_rreg = &cik_smc_rreg;
 	adev->smc_wreg = &cik_smc_wreg;
@@ -2155,19 +2124,9 @@ static int cik_common_early_init(void *handle)
 	return 0;
 }
 
-static int cik_common_sw_init(void *handle)
-{
-	return 0;
-}
-
-static int cik_common_sw_fini(void *handle)
-{
-	return 0;
-}
-
-static int cik_common_hw_init(void *handle)
+static int cik_common_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* move the golden regs per IP block */
 	cik_init_golden_registers(adev);
@@ -2179,48 +2138,36 @@ static int cik_common_hw_init(void *handle)
 	return 0;
 }
 
-static int cik_common_hw_fini(void *handle)
+static int cik_common_hw_fini(struct amdgpu_ip_block *ip_block)
 {
 	return 0;
 }
 
-static int cik_common_suspend(void *handle)
+static int cik_common_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return cik_common_hw_fini(adev);
-}
-
-static int cik_common_resume(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return cik_common_hw_init(adev);
+	return cik_common_hw_init(ip_block);
 }
 
-static bool cik_common_is_idle(void *handle)
+static bool cik_common_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	return true;
 }
 
-static int cik_common_wait_for_idle(void *handle)
-{
-	return 0;
-}
 
-static int cik_common_soft_reset(void *handle)
+
+static int cik_common_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	/* XXX hard reset?? */
 	return 0;
 }
 
-static int cik_common_set_clockgating_state(void *handle,
+static int cik_common_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					    enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int cik_common_set_powergating_state(void *handle,
+static int cik_common_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					    enum amd_powergating_state state)
 {
 	return 0;
@@ -2229,15 +2176,10 @@ static int cik_common_set_powergating_state(void *handle,
 static const struct amd_ip_funcs cik_common_ip_funcs = {
 	.name = "cik_common",
 	.early_init = cik_common_early_init,
-	.late_init = NULL,
-	.sw_init = cik_common_sw_init,
-	.sw_fini = cik_common_sw_fini,
 	.hw_init = cik_common_hw_init,
 	.hw_fini = cik_common_hw_fini,
-	.suspend = cik_common_suspend,
 	.resume = cik_common_resume,
 	.is_idle = cik_common_is_idle,
-	.wait_for_idle = cik_common_wait_for_idle,
 	.soft_reset = cik_common_soft_reset,
 	.set_clockgating_state = cik_common_set_clockgating_state,
 	.set_powergating_state = cik_common_set_powergating_state,
diff --git a/drivers/gpu/drm/amd/amdgpu/cik_ih.c b/drivers/gpu/drm/amd/amdgpu/cik_ih.c
index df385ffc9768..876a3256dba4 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/cik_ih.c
@@ -156,6 +156,9 @@ static int cik_ih_irq_init(struct amdgpu_device *adev)
 	/* enable irqs */
 	cik_ih_enable_interrupts(adev);
 
+	if (adev->irq.ih_soft.ring_size)
+		adev->irq.ih_soft.enabled = true;
+
 	return 0;
 }
 
@@ -192,6 +195,9 @@ static u32 cik_ih_get_wptr(struct amdgpu_device *adev,
 
 	wptr = le32_to_cpu(*ih->wptr_cpu);
 
+	if (ih == &adev->irq.ih_soft)
+		goto out;
+
 	if (wptr & IH_RB_WPTR__RB_OVERFLOW_MASK) {
 		wptr &= ~IH_RB_WPTR__RB_OVERFLOW_MASK;
 		/* When a ring buffer overflow happen start parsing interrupt
@@ -204,7 +210,15 @@ static u32 cik_ih_get_wptr(struct amdgpu_device *adev,
 		tmp = RREG32(mmIH_RB_CNTL);
 		tmp |= IH_RB_CNTL__WPTR_OVERFLOW_CLEAR_MASK;
 		WREG32(mmIH_RB_CNTL, tmp);
+
+		/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+		 * can be detected.
+		 */
+		tmp &= ~IH_RB_CNTL__WPTR_OVERFLOW_CLEAR_MASK;
+		WREG32(mmIH_RB_CNTL, tmp);
 	}
+
+out:
 	return (wptr & ih->ptr_mask);
 }
 
@@ -277,9 +291,9 @@ static void cik_ih_set_rptr(struct amdgpu_device *adev,
 	WREG32(mmIH_RB_RPTR, ih->rptr);
 }
 
-static int cik_ih_early_init(void *handle)
+static int cik_ih_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	ret = amdgpu_irq_add_domain(adev);
@@ -291,23 +305,27 @@ static int cik_ih_early_init(void *handle)
 	return 0;
 }
 
-static int cik_ih_sw_init(void *handle)
+static int cik_ih_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, false);
 	if (r)
 		return r;
 
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, IH_SW_RING_SIZE, true);
+	if (r)
+		return r;
+
 	r = amdgpu_irq_init(adev);
 
 	return r;
 }
 
-static int cik_ih_sw_fini(void *handle)
+static int cik_ih_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_fini_sw(adev);
 	amdgpu_irq_remove_domain(adev);
@@ -315,39 +333,33 @@ static int cik_ih_sw_fini(void *handle)
 	return 0;
 }
 
-static int cik_ih_hw_init(void *handle)
+static int cik_ih_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return cik_ih_irq_init(adev);
 }
 
-static int cik_ih_hw_fini(void *handle)
+static int cik_ih_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	cik_ih_irq_disable(adev);
+	cik_ih_irq_disable(ip_block->adev);
 
 	return 0;
 }
 
-static int cik_ih_suspend(void *handle)
+static int cik_ih_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return cik_ih_hw_fini(adev);
+	return cik_ih_hw_fini(ip_block);
 }
 
-static int cik_ih_resume(void *handle)
+static int cik_ih_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return cik_ih_hw_init(adev);
+	return cik_ih_hw_init(ip_block);
 }
 
-static bool cik_ih_is_idle(void *handle)
+static bool cik_ih_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (tmp & SRBM_STATUS__IH_BUSY_MASK)
@@ -356,11 +368,11 @@ static bool cik_ih_is_idle(void *handle)
 	return true;
 }
 
-static int cik_ih_wait_for_idle(void *handle)
+static int cik_ih_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		/* read MC_STATUS */
@@ -372,9 +384,9 @@ static int cik_ih_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int cik_ih_soft_reset(void *handle)
+static int cik_ih_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	u32 srbm_soft_reset = 0;
 	u32 tmp = RREG32(mmSRBM_STATUS);
@@ -402,13 +414,13 @@ static int cik_ih_soft_reset(void *handle)
 	return 0;
 }
 
-static int cik_ih_set_clockgating_state(void *handle,
+static int cik_ih_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int cik_ih_set_powergating_state(void *handle,
+static int cik_ih_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -417,7 +429,6 @@ static int cik_ih_set_powergating_state(void *handle,
 static const struct amd_ip_funcs cik_ih_ip_funcs = {
 	.name = "cik_ih",
 	.early_init = cik_ih_early_init,
-	.late_init = NULL,
 	.sw_init = cik_ih_sw_init,
 	.sw_fini = cik_ih_sw_fini,
 	.hw_init = cik_ih_hw_init,
@@ -442,8 +453,7 @@ static void cik_ih_set_interrupt_funcs(struct amdgpu_device *adev)
 	adev->irq.ih_funcs = &cik_ih_funcs;
 }
 
-const struct amdgpu_ip_block_version cik_ih_ip_block =
-{
+const struct amdgpu_ip_block_version cik_ih_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_IH,
 	.major = 2,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
index 67d16236b216..9e8715b4739d 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
+++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
@@ -54,7 +54,9 @@ static void cik_sdma_set_ring_funcs(struct amdgpu_device *adev);
 static void cik_sdma_set_irq_funcs(struct amdgpu_device *adev);
 static void cik_sdma_set_buffer_funcs(struct amdgpu_device *adev);
 static void cik_sdma_set_vm_pte_funcs(struct amdgpu_device *adev);
-static int cik_sdma_soft_reset(void *handle);
+static int cik_sdma_soft_reset(struct amdgpu_ip_block *ip_block);
+
+u32 amdgpu_cik_gpu_check_soft_reset(struct amdgpu_device *adev);
 
 MODULE_FIRMWARE("amdgpu/bonaire_sdma.bin");
 MODULE_FIRMWARE("amdgpu/bonaire_sdma1.bin");
@@ -67,9 +69,6 @@ MODULE_FIRMWARE("amdgpu/kabini_sdma1.bin");
 MODULE_FIRMWARE("amdgpu/mullins_sdma.bin");
 MODULE_FIRMWARE("amdgpu/mullins_sdma1.bin");
 
-u32 amdgpu_cik_gpu_check_soft_reset(struct amdgpu_device *adev);
-
-
 static void cik_sdma_free_microcode(struct amdgpu_device *adev)
 {
 	int i;
@@ -107,7 +106,6 @@ static void cik_sdma_free_microcode(struct amdgpu_device *adev)
 static int cik_sdma_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err = 0, i;
 
 	DRM_DEBUG("\n");
@@ -133,16 +131,20 @@ static int cik_sdma_init_microcode(struct amdgpu_device *adev)
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		if (i == 0)
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma.bin", chip_name);
+			err = amdgpu_ucode_request(adev, &adev->sdma.instance[i].fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_sdma.bin", chip_name);
 		else
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma1.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->sdma.instance[i].fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->sdma.instance[i].fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_sdma1.bin", chip_name);
 		if (err)
 			goto out;
 	}
 out:
 	if (err) {
-		pr_err("cik_sdma: Failed to load firmware \"%s\"\n", fw_name);
+		pr_err("cik_sdma: Failed to load firmware \"%s_sdma%s.bin\"\n",
+		       chip_name, i == 0 ? "" : "1");
 		for (i = 0; i < adev->sdma.num_instances; i++)
 			amdgpu_ucode_release(&adev->sdma.instance[i].fw);
 	}
@@ -308,8 +310,6 @@ static void cik_sdma_gfx_stop(struct amdgpu_device *adev)
 	u32 rb_cntl;
 	int i;
 
-	amdgpu_sdma_unset_buffer_funcs_helper(adev);
-
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		rb_cntl = RREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i]);
 		rb_cntl &= ~SDMA0_GFX_RB_CNTL__RB_ENABLE_MASK;
@@ -489,8 +489,6 @@ static int cik_sdma_gfx_resume(struct amdgpu_device *adev)
 #endif
 		/* enable DMA IBs */
 		WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
-
-		ring->sched.ready = true;
 	}
 
 	cik_sdma_enable(adev, true);
@@ -500,9 +498,6 @@ static int cik_sdma_gfx_resume(struct amdgpu_device *adev)
 		r = amdgpu_ring_test_helper(ring);
 		if (r)
 			return r;
-
-		if (adev->mman.buffer_funcs_ring == ring)
-			amdgpu_ttm_set_buffer_funcs_status(adev, true);
 	}
 
 	return 0;
@@ -702,7 +697,7 @@ static int cik_sdma_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 err1:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err0:
 	amdgpu_device_wb_free(adev, index);
@@ -924,12 +919,17 @@ static void cik_enable_sdma_mgls(struct amdgpu_device *adev,
 	}
 }
 
-static int cik_sdma_early_init(void *handle)
+static int cik_sdma_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
 
 	adev->sdma.num_instances = SDMA_MAX_INSTANCE;
 
+	r = cik_sdma_init_microcode(adev);
+	if (r)
+		return r;
+
 	cik_sdma_set_ring_funcs(adev);
 	cik_sdma_set_irq_funcs(adev);
 	cik_sdma_set_buffer_funcs(adev);
@@ -938,18 +938,12 @@ static int cik_sdma_early_init(void *handle)
 	return 0;
 }
 
-static int cik_sdma_sw_init(void *handle)
+static int cik_sdma_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r, i;
 
-	r = cik_sdma_init_microcode(adev);
-	if (r) {
-		DRM_ERROR("Failed to load sdma firmware!\n");
-		return r;
-	}
-
 	/* SDMA trap event */
 	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 224,
 			      &adev->sdma.trap_irq);
@@ -984,9 +978,9 @@ static int cik_sdma_sw_init(void *handle)
 	return r;
 }
 
-static int cik_sdma_sw_fini(void *handle)
+static int cik_sdma_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
 	for (i = 0; i < adev->sdma.num_instances; i++)
@@ -996,21 +990,16 @@ static int cik_sdma_sw_fini(void *handle)
 	return 0;
 }
 
-static int cik_sdma_hw_init(void *handle)
+static int cik_sdma_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	r = cik_sdma_start(adev);
-	if (r)
-		return r;
-
-	return r;
+	return cik_sdma_start(adev);
 }
 
-static int cik_sdma_hw_fini(void *handle)
+static int cik_sdma_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	cik_ctx_switch_enable(adev, false);
 	cik_sdma_enable(adev, false);
@@ -1018,25 +1007,21 @@ static int cik_sdma_hw_fini(void *handle)
 	return 0;
 }
 
-static int cik_sdma_suspend(void *handle)
+static int cik_sdma_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return cik_sdma_hw_fini(adev);
+	return cik_sdma_hw_fini(ip_block);
 }
 
-static int cik_sdma_resume(void *handle)
+static int cik_sdma_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	cik_sdma_soft_reset(ip_block);
 
-	cik_sdma_soft_reset(handle);
-
-	return cik_sdma_hw_init(adev);
+	return cik_sdma_hw_init(ip_block);
 }
 
-static bool cik_sdma_is_idle(void *handle)
+static bool cik_sdma_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS2);
 
 	if (tmp & (SRBM_STATUS2__SDMA_BUSY_MASK |
@@ -1046,27 +1031,23 @@ static bool cik_sdma_is_idle(void *handle)
 	return true;
 }
 
-static int cik_sdma_wait_for_idle(void *handle)
+static int cik_sdma_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
-	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(mmSRBM_STATUS2) & (SRBM_STATUS2__SDMA_BUSY_MASK |
-				SRBM_STATUS2__SDMA1_BUSY_MASK);
-
-		if (!tmp)
+		if (cik_sdma_is_idle(ip_block))
 			return 0;
 		udelay(1);
 	}
 	return -ETIMEDOUT;
 }
 
-static int cik_sdma_soft_reset(void *handle)
+static int cik_sdma_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 srbm_soft_reset = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp;
 
 	/* sdma0 */
@@ -1200,11 +1181,11 @@ static int cik_sdma_process_illegal_inst_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int cik_sdma_set_clockgating_state(void *handle,
+static int cik_sdma_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	bool gate = false;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_CG_STATE_GATE)
 		gate = true;
@@ -1215,7 +1196,7 @@ static int cik_sdma_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int cik_sdma_set_powergating_state(void *handle,
+static int cik_sdma_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -1224,7 +1205,6 @@ static int cik_sdma_set_powergating_state(void *handle,
 static const struct amd_ip_funcs cik_sdma_ip_funcs = {
 	.name = "cik_sdma",
 	.early_init = cik_sdma_early_init,
-	.late_init = NULL,
 	.sw_init = cik_sdma_sw_init,
 	.sw_fini = cik_sdma_sw_fini,
 	.hw_init = cik_sdma_hw_init,
@@ -1298,7 +1278,7 @@ static void cik_sdma_set_irq_funcs(struct amdgpu_device *adev)
  * @src_offset: src GPU address
  * @dst_offset: dst GPU address
  * @byte_count: number of bytes to xfer
- * @tmz: is this a secure operation
+ * @copy_flags: unused
  *
  * Copy GPU buffers using the DMA engine (CIK).
  * Used by the amdgpu ttm implementation to move pages if
@@ -1308,7 +1288,7 @@ static void cik_sdma_emit_copy_buffer(struct amdgpu_ib *ib,
 				      uint64_t src_offset,
 				      uint64_t dst_offset,
 				      uint32_t byte_count,
-				      bool tmz)
+				      uint32_t copy_flags)
 {
 	ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_COPY, SDMA_COPY_SUB_OPCODE_LINEAR, 0);
 	ib->ptr[ib->length_dw++] = byte_count;
diff --git a/drivers/gpu/drm/amd/amdgpu/cikd.h b/drivers/gpu/drm/amd/amdgpu/cikd.h
index 55982c0064b5..8aca4f2734f2 100644
--- a/drivers/gpu/drm/amd/amdgpu/cikd.h
+++ b/drivers/gpu/drm/amd/amdgpu/cikd.h
@@ -51,8 +51,14 @@
 #define HPD4_REGISTER_OFFSET                 (0x1813 - 0x1807)
 #define HPD5_REGISTER_OFFSET                 (0x1816 - 0x1807)
 
-#define BONAIRE_GB_ADDR_CONFIG_GOLDEN        0x12010001
-#define HAWAII_GB_ADDR_CONFIG_GOLDEN         0x12011003
+/* audio endpt instance offsets */
+#define AUD0_REGISTER_OFFSET                 (0x1780 - 0x1780)
+#define AUD1_REGISTER_OFFSET                 (0x1786 - 0x1780)
+#define AUD2_REGISTER_OFFSET                 (0x178c - 0x1780)
+#define AUD3_REGISTER_OFFSET                 (0x1792 - 0x1780)
+#define AUD4_REGISTER_OFFSET                 (0x1798 - 0x1780)
+#define AUD5_REGISTER_OFFSET                 (0x179d - 0x1780)
+#define AUD6_REGISTER_OFFSET                 (0x17a4 - 0x1780)
 
 #define		PIPEID(x)					((x) << 0)
 #define		MEID(x)						((x) << 2)
@@ -364,6 +370,7 @@
 		 * 1 - Stream
 		 * 2 - Bypass
 		 */
+#define		EOP_EXEC				(1 << 28) /* For Trailing Fence */
 #define		DATA_SEL(x)                             ((x) << 29)
 		/* 0 - discard
 		 * 1 - send low 32bit data
diff --git a/drivers/gpu/drm/amd/amdgpu/clearstate_gfx12.h b/drivers/gpu/drm/amd/amdgpu/clearstate_gfx12.h
new file mode 100644
index 000000000000..2f6c9d11d5ae
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/clearstate_gfx12.h
@@ -0,0 +1,121 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __CLEARSTATE_GFX12_H_
+#define __CLEARSTATE_GFX12_H_
+
+static const unsigned int gfx12_SECT_CONTEXT_def_1[] = {
+0x00000000, //mmSC_MEM_TEMPORAL
+0x00000000, //mmSC_MEM_SPEC_READ
+0x00000000, //mmPA_SC_VPORT_0_TL
+0x00000000, //mmPA_SC_VPORT_0_BR
+0x00000000, //mmPA_SC_VPORT_1_TL
+0x00000000, //mmPA_SC_VPORT_1_BR
+0x00000000, //mmPA_SC_VPORT_2_TL
+0x00000000, //mmPA_SC_VPORT_2_BR
+0x00000000, //mmPA_SC_VPORT_3_TL
+0x00000000, //mmPA_SC_VPORT_3_BR
+0x00000000, //mmPA_SC_VPORT_4_TL
+0x00000000, //mmPA_SC_VPORT_4_BR
+0x00000000, //mmPA_SC_VPORT_5_TL
+0x00000000, //mmPA_SC_VPORT_5_BR
+0x00000000, //mmPA_SC_VPORT_6_TL
+0x00000000, //mmPA_SC_VPORT_6_BR
+0x00000000, //mmPA_SC_VPORT_7_TL
+0x00000000, //mmPA_SC_VPORT_7_BR
+0x00000000, //mmPA_SC_VPORT_8_TL
+0x00000000, //mmPA_SC_VPORT_8_BR
+0x00000000, //mmPA_SC_VPORT_9_TL
+0x00000000, //mmPA_SC_VPORT_9_BR
+0x00000000, //mmPA_SC_VPORT_10_TL
+0x00000000, //mmPA_SC_VPORT_10_BR
+0x00000000, //mmPA_SC_VPORT_11_TL
+0x00000000, //mmPA_SC_VPORT_11_BR
+0x00000000, //mmPA_SC_VPORT_12_TL
+0x00000000, //mmPA_SC_VPORT_12_BR
+0x00000000, //mmPA_SC_VPORT_13_TL
+0x00000000, //mmPA_SC_VPORT_13_BR
+0x00000000, //mmPA_SC_VPORT_14_TL
+0x00000000, //mmPA_SC_VPORT_14_BR
+0x00000000, //mmPA_SC_VPORT_15_TL
+0x00000000, //mmPA_SC_VPORT_15_BR
+};
+
+static const unsigned int gfx12_SECT_CONTEXT_def_2[] = {
+0x00000000, //mmPA_CL_PROG_NEAR_CLIP_Z
+0x00000000, //mmPA_RATE_CNTL
+};
+
+static const unsigned int gfx12_SECT_CONTEXT_def_3[] = {
+0x00000000, //mmCP_PERFMON_CNTX_CNTL
+};
+
+static const unsigned int gfx12_SECT_CONTEXT_def_4[] = {
+0x00000000, //mmCONTEXT_RESERVED_REG0
+0x00000000, //mmCONTEXT_RESERVED_REG1
+0x00000000, //mmPA_SC_CLIPRECT_0_EXT
+0x00000000, //mmPA_SC_CLIPRECT_1_EXT
+0x00000000, //mmPA_SC_CLIPRECT_2_EXT
+0x00000000, //mmPA_SC_CLIPRECT_3_EXT
+};
+
+static const unsigned int gfx12_SECT_CONTEXT_def_5[] = {
+0x00000000, //mmPA_SC_HIZ_INFO
+0x00000000, //mmPA_SC_HIS_INFO
+0x00000000, //mmPA_SC_HIZ_BASE
+0x00000000, //mmPA_SC_HIZ_BASE_EXT
+0x00000000, //mmPA_SC_HIZ_SIZE_XY
+0x00000000, //mmPA_SC_HIS_BASE
+0x00000000, //mmPA_SC_HIS_BASE_EXT
+0x00000000, //mmPA_SC_HIS_SIZE_XY
+0x00000000, //mmPA_SC_BINNER_OUTPUT_TIMEOUT_CNTL
+0x00000000, //mmPA_SC_BINNER_DYNAMIC_BATCH_LIMIT
+0x00000000, //mmPA_SC_HISZ_CONTROL
+};
+
+static const unsigned int gfx12_SECT_CONTEXT_def_6[] = {
+0x00000000, //mmCB_MEM0_INFO
+0x00000000, //mmCB_MEM1_INFO
+0x00000000, //mmCB_MEM2_INFO
+0x00000000, //mmCB_MEM3_INFO
+0x00000000, //mmCB_MEM4_INFO
+0x00000000, //mmCB_MEM5_INFO
+0x00000000, //mmCB_MEM6_INFO
+0x00000000, //mmCB_MEM7_INFO
+};
+
+static const struct cs_extent_def gfx12_SECT_CONTEXT_defs[] = {
+    {gfx12_SECT_CONTEXT_def_1, 0x0000a03e, 34 },
+    {gfx12_SECT_CONTEXT_def_2, 0x0000a0cc, 2 },
+    {gfx12_SECT_CONTEXT_def_3, 0x0000a0d8, 1 },
+    {gfx12_SECT_CONTEXT_def_4, 0x0000a0db, 6 },
+    {gfx12_SECT_CONTEXT_def_5, 0x0000a2e5, 11 },
+    {gfx12_SECT_CONTEXT_def_6, 0x0000a3c0, 8 },
+    { 0, 0, 0 }
+};
+
+static const struct cs_section_def gfx12_cs_data[] = {
+    { gfx12_SECT_CONTEXT_defs, SECT_CONTEXT },
+    { 0, SECT_NONE }
+};
+
+#endif /* __CLEARSTATE_GFX12_H_ */
diff --git a/drivers/gpu/drm/amd/amdgpu/clearstate_gfx9.h b/drivers/gpu/drm/amd/amdgpu/clearstate_gfx9.h
index 567a904804bc..9c85ca6358c1 100644
--- a/drivers/gpu/drm/amd/amdgpu/clearstate_gfx9.h
+++ b/drivers/gpu/drm/amd/amdgpu/clearstate_gfx9.h
@@ -21,8 +21,7 @@
  *
  */
 
-static const unsigned int gfx9_SECT_CONTEXT_def_1[] =
-{
+static const unsigned int gfx9_SECT_CONTEXT_def_1[] = {
     0x00000000, // DB_RENDER_CONTROL
     0x00000000, // DB_COUNT_CONTROL
     0x00000000, // DB_DEPTH_VIEW
@@ -236,8 +235,7 @@ static const unsigned int gfx9_SECT_CONTEXT_def_1[] =
     0x00000000, // PA_SC_VPORT_ZMIN_15
     0x3f800000, // PA_SC_VPORT_ZMAX_15
 };
-static const unsigned int gfx9_SECT_CONTEXT_def_2[] =
-{
+static const unsigned int gfx9_SECT_CONTEXT_def_2[] = {
     0x00000000, // PA_SC_SCREEN_EXTENT_CONTROL
     0x00000000, // PA_SC_TILE_STEERING_OVERRIDE
     0x00000000, // CP_PERFMON_CNTX_CNTL
@@ -521,15 +519,13 @@ static const unsigned int gfx9_SECT_CONTEXT_def_2[] =
     0x00000000, // CB_MRT6_EPITCH
     0x00000000, // CB_MRT7_EPITCH
 };
-static const unsigned int gfx9_SECT_CONTEXT_def_3[] =
-{
+static const unsigned int gfx9_SECT_CONTEXT_def_3[] = {
     0x00000000, // PA_CL_POINT_X_RAD
     0x00000000, // PA_CL_POINT_Y_RAD
     0x00000000, // PA_CL_POINT_SIZE
     0x00000000, // PA_CL_POINT_CULL_RAD
 };
-static const unsigned int gfx9_SECT_CONTEXT_def_4[] =
-{
+static const unsigned int gfx9_SECT_CONTEXT_def_4[] = {
     0x00000000, // DB_DEPTH_CONTROL
     0x00000000, // DB_EQAA
     0x00000000, // CB_COLOR_CONTROL
@@ -688,17 +684,14 @@ static const unsigned int gfx9_SECT_CONTEXT_def_4[] =
     0x00000000, // VGT_GS_OUT_PRIM_TYPE
     0x00000000, // IA_ENHANCE
 };
-static const unsigned int gfx9_SECT_CONTEXT_def_5[] =
-{
+static const unsigned int gfx9_SECT_CONTEXT_def_5[] = {
     0x00000000, // WD_ENHANCE
     0x00000000, // VGT_PRIMITIVEID_EN
 };
-static const unsigned int gfx9_SECT_CONTEXT_def_6[] =
-{
+static const unsigned int gfx9_SECT_CONTEXT_def_6[] = {
     0x00000000, // VGT_PRIMITIVEID_RESET
 };
-static const unsigned int gfx9_SECT_CONTEXT_def_7[] =
-{
+static const unsigned int gfx9_SECT_CONTEXT_def_7[] = {
     0x00000000, // VGT_GS_MAX_PRIMS_PER_SUBGROUP
     0x00000000, // VGT_DRAW_PAYLOAD_CNTL
     0, // HOLE
@@ -766,8 +759,7 @@ static const unsigned int gfx9_SECT_CONTEXT_def_7[] =
     0x00000000, // VGT_STRMOUT_CONFIG
     0x00000000, // VGT_STRMOUT_BUFFER_CONFIG
 };
-static const unsigned int gfx9_SECT_CONTEXT_def_8[] =
-{
+static const unsigned int gfx9_SECT_CONTEXT_def_8[] = {
     0x00000000, // PA_SC_CENTROID_PRIORITY_0
     0x00000000, // PA_SC_CENTROID_PRIORITY_1
     0x00001000, // PA_SC_LINE_CNTL
@@ -924,8 +916,7 @@ static const unsigned int gfx9_SECT_CONTEXT_def_8[] =
     0x00000000, // CB_COLOR7_DCC_BASE
     0x00000000, // CB_COLOR7_DCC_BASE_EXT
 };
-static const struct cs_extent_def gfx9_SECT_CONTEXT_defs[] =
-{
+static const struct cs_extent_def gfx9_SECT_CONTEXT_defs[] = {
     {gfx9_SECT_CONTEXT_def_1, 0x0000a000, 212 },
     {gfx9_SECT_CONTEXT_def_2, 0x0000a0d6, 282 },
     {gfx9_SECT_CONTEXT_def_3, 0x0000a1f5, 4 },
diff --git a/drivers/gpu/drm/amd/amdgpu/clearstate_si.h b/drivers/gpu/drm/amd/amdgpu/clearstate_si.h
index 66e39cdb5cb0..5fd96ddd7f0f 100644
--- a/drivers/gpu/drm/amd/amdgpu/clearstate_si.h
+++ b/drivers/gpu/drm/amd/amdgpu/clearstate_si.h
@@ -21,8 +21,7 @@
  *
  */
 
-static const u32 si_SECT_CONTEXT_def_1[] =
-{
+static const u32 si_SECT_CONTEXT_def_1[] = {
     0x00000000, // DB_RENDER_CONTROL
     0x00000000, // DB_COUNT_CONTROL
     0x00000000, // DB_DEPTH_VIEW
@@ -236,8 +235,7 @@ static const u32 si_SECT_CONTEXT_def_1[] =
     0x00000000, // PA_SC_VPORT_ZMIN_15
     0x3f800000, // PA_SC_VPORT_ZMAX_15
 };
-static const u32 si_SECT_CONTEXT_def_2[] =
-{
+static const u32 si_SECT_CONTEXT_def_2[] = {
     0x00000000, // CP_PERFMON_CNTX_CNTL
     0x00000000, // CP_RINGID
     0x00000000, // CP_VMID
@@ -511,8 +509,7 @@ static const u32 si_SECT_CONTEXT_def_2[] =
     0x00000000, // CB_BLEND6_CONTROL
     0x00000000, // CB_BLEND7_CONTROL
 };
-static const u32 si_SECT_CONTEXT_def_3[] =
-{
+static const u32 si_SECT_CONTEXT_def_3[] = {
     0x00000000, // PA_CL_POINT_X_RAD
     0x00000000, // PA_CL_POINT_Y_RAD
     0x00000000, // PA_CL_POINT_SIZE
@@ -520,8 +517,7 @@ static const u32 si_SECT_CONTEXT_def_3[] =
     0x00000000, // VGT_DMA_BASE_HI
     0x00000000, // VGT_DMA_BASE
 };
-static const u32 si_SECT_CONTEXT_def_4[] =
-{
+static const u32 si_SECT_CONTEXT_def_4[] = {
     0x00000000, // DB_DEPTH_CONTROL
     0x00000000, // DB_EQAA
     0x00000000, // CB_COLOR_CONTROL
@@ -680,16 +676,13 @@ static const u32 si_SECT_CONTEXT_def_4[] =
     0x00000000, // VGT_GS_OUT_PRIM_TYPE
     0x00000000, // IA_ENHANCE
 };
-static const u32 si_SECT_CONTEXT_def_5[] =
-{
+static const u32 si_SECT_CONTEXT_def_5[] = {
     0x00000000, // VGT_PRIMITIVEID_EN
 };
-static const u32 si_SECT_CONTEXT_def_6[] =
-{
+static const u32 si_SECT_CONTEXT_def_6[] = {
     0x00000000, // VGT_PRIMITIVEID_RESET
 };
-static const u32 si_SECT_CONTEXT_def_7[] =
-{
+static const u32 si_SECT_CONTEXT_def_7[] = {
     0x00000000, // VGT_MULTI_PRIM_IB_RESET_EN
     0, // HOLE
     0, // HOLE
@@ -924,8 +917,7 @@ static const u32 si_SECT_CONTEXT_def_7[] =
     0x00000000, // CB_COLOR7_CLEAR_WORD0
     0x00000000, // CB_COLOR7_CLEAR_WORD1
 };
-static const struct cs_extent_def si_SECT_CONTEXT_defs[] =
-{
+static const struct cs_extent_def si_SECT_CONTEXT_defs[] = {
     {si_SECT_CONTEXT_def_1, 0x0000a000, 212 },
     {si_SECT_CONTEXT_def_2, 0x0000a0d8, 272 },
     {si_SECT_CONTEXT_def_3, 0x0000a1f5, 6 },
diff --git a/drivers/gpu/drm/amd/amdgpu/cyan_skillfish_reg_init.c b/drivers/gpu/drm/amd/amdgpu/cyan_skillfish_reg_init.c
new file mode 100644
index 000000000000..ed1e25661706
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/cyan_skillfish_reg_init.c
@@ -0,0 +1,56 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright 2018 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "nv.h"
+
+#include "soc15_common.h"
+#include "soc15_hw_ip.h"
+#include "cyan_skillfish_ip_offset.h"
+
+int cyan_skillfish_reg_base_init(struct amdgpu_device *adev)
+{
+	/* HW has more IP blocks,  only initialized the blocke needed by driver */
+	uint32_t i;
+
+	adev->gfx.xcc_mask = 1;
+	for (i = 0 ; i < MAX_INSTANCE ; ++i) {
+		adev->reg_offset[GC_HWIP][i] = (uint32_t *)(&(GC_BASE.instance[i]));
+		adev->reg_offset[HDP_HWIP][i] = (uint32_t *)(&(HDP_BASE.instance[i]));
+		adev->reg_offset[MMHUB_HWIP][i] = (uint32_t *)(&(MMHUB_BASE.instance[i]));
+		adev->reg_offset[ATHUB_HWIP][i] = (uint32_t *)(&(ATHUB_BASE.instance[i]));
+		adev->reg_offset[NBIO_HWIP][i] = (uint32_t *)(&(NBIO_BASE.instance[i]));
+		adev->reg_offset[MP0_HWIP][i] = (uint32_t *)(&(MP0_BASE.instance[i]));
+		adev->reg_offset[MP1_HWIP][i] = (uint32_t *)(&(MP1_BASE.instance[i]));
+		adev->reg_offset[VCN_HWIP][i] = (uint32_t *)(&(UVD0_BASE.instance[i]));
+		adev->reg_offset[DF_HWIP][i] = (uint32_t *)(&(DF_BASE.instance[i]));
+		adev->reg_offset[DCE_HWIP][i] = (uint32_t *)(&(DMU_BASE.instance[i]));
+		adev->reg_offset[OSSSYS_HWIP][i] = (uint32_t *)(&(OSSSYS_BASE.instance[i]));
+		adev->reg_offset[SDMA0_HWIP][i] = (uint32_t *)(&(GC_BASE.instance[i]));
+		adev->reg_offset[SDMA1_HWIP][i] = (uint32_t *)(&(GC_BASE.instance[i]));
+		adev->reg_offset[SMUIO_HWIP][i] = (uint32_t *)(&(SMUIO_BASE.instance[i]));
+		adev->reg_offset[THM_HWIP][i] = (uint32_t *)(&(THM_BASE.instance[i]));
+		adev->reg_offset[CLK_HWIP][i] = (uint32_t *)(&(CLK_BASE.instance[i]));
+	}
+	return 0;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/cz_ih.c b/drivers/gpu/drm/amd/amdgpu/cz_ih.c
index b8c47e0cf37a..bc7a2e06ab5f 100644
--- a/drivers/gpu/drm/amd/amdgpu/cz_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/cz_ih.c
@@ -157,6 +157,9 @@ static int cz_ih_irq_init(struct amdgpu_device *adev)
 	/* enable interrupts */
 	cz_ih_enable_interrupts(adev);
 
+	if (adev->irq.ih_soft.ring_size)
+		adev->irq.ih_soft.enabled = true;
+
 	return 0;
 }
 
@@ -194,6 +197,9 @@ static u32 cz_ih_get_wptr(struct amdgpu_device *adev,
 
 	wptr = le32_to_cpu(*ih->wptr_cpu);
 
+	if (ih == &adev->irq.ih_soft)
+		goto out;
+
 	if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
 		goto out;
 
@@ -216,6 +222,11 @@ static u32 cz_ih_get_wptr(struct amdgpu_device *adev,
 	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
 	WREG32(mmIH_RB_CNTL, tmp);
 
+	/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+	 * can be detected.
+	 */
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	WREG32(mmIH_RB_CNTL, tmp);
 
 out:
 	return (wptr & ih->ptr_mask);
@@ -269,9 +280,9 @@ static void cz_ih_set_rptr(struct amdgpu_device *adev,
 	WREG32(mmIH_RB_RPTR, ih->rptr);
 }
 
-static int cz_ih_early_init(void *handle)
+static int cz_ih_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	ret = amdgpu_irq_add_domain(adev);
@@ -283,23 +294,27 @@ static int cz_ih_early_init(void *handle)
 	return 0;
 }
 
-static int cz_ih_sw_init(void *handle)
+static int cz_ih_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, false);
 	if (r)
 		return r;
 
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, IH_SW_RING_SIZE, true);
+	if (r)
+		return r;
+
 	r = amdgpu_irq_init(adev);
 
 	return r;
 }
 
-static int cz_ih_sw_fini(void *handle)
+static int cz_ih_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_fini_sw(adev);
 	amdgpu_irq_remove_domain(adev);
@@ -307,10 +322,10 @@ static int cz_ih_sw_fini(void *handle)
 	return 0;
 }
 
-static int cz_ih_hw_init(void *handle)
+static int cz_ih_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = cz_ih_irq_init(adev);
 	if (r)
@@ -319,32 +334,26 @@ static int cz_ih_hw_init(void *handle)
 	return 0;
 }
 
-static int cz_ih_hw_fini(void *handle)
+static int cz_ih_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	cz_ih_irq_disable(adev);
+	cz_ih_irq_disable(ip_block->adev);
 
 	return 0;
 }
 
-static int cz_ih_suspend(void *handle)
+static int cz_ih_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return cz_ih_hw_fini(adev);
+	return cz_ih_hw_fini(ip_block);
 }
 
-static int cz_ih_resume(void *handle)
+static int cz_ih_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return cz_ih_hw_init(adev);
+	return cz_ih_hw_init(ip_block);
 }
 
-static bool cz_ih_is_idle(void *handle)
+static bool cz_ih_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (REG_GET_FIELD(tmp, SRBM_STATUS, IH_BUSY))
@@ -353,11 +362,11 @@ static bool cz_ih_is_idle(void *handle)
 	return true;
 }
 
-static int cz_ih_wait_for_idle(void *handle)
+static int cz_ih_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		/* read MC_STATUS */
@@ -369,10 +378,10 @@ static int cz_ih_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int cz_ih_soft_reset(void *handle)
+static int cz_ih_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 srbm_soft_reset = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (tmp & SRBM_STATUS__IH_BUSY_MASK)
@@ -399,14 +408,14 @@ static int cz_ih_soft_reset(void *handle)
 	return 0;
 }
 
-static int cz_ih_set_clockgating_state(void *handle,
+static int cz_ih_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	// TODO
 	return 0;
 }
 
-static int cz_ih_set_powergating_state(void *handle,
+static int cz_ih_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	// TODO
@@ -416,7 +425,6 @@ static int cz_ih_set_powergating_state(void *handle,
 static const struct amd_ip_funcs cz_ih_ip_funcs = {
 	.name = "cz_ih",
 	.early_init = cz_ih_early_init,
-	.late_init = NULL,
 	.sw_init = cz_ih_sw_init,
 	.sw_fini = cz_ih_sw_fini,
 	.hw_init = cz_ih_hw_init,
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c
index 9a24ed463abd..72ca6538b2e4 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c
@@ -21,6 +21,7 @@
  *
  */
 
+#include <drm/drm_edid.h>
 #include <drm/drm_fourcc.h>
 #include <drm/drm_modeset_helper.h>
 #include <drm/drm_modeset_helper_vtables.h>
@@ -51,9 +52,9 @@
 
 static void dce_v10_0_set_display_funcs(struct amdgpu_device *adev);
 static void dce_v10_0_set_irq_funcs(struct amdgpu_device *adev);
+static void dce_v10_0_hpd_int_ack(struct amdgpu_device *adev, int hpd);
 
-static const u32 crtc_offsets[] =
-{
+static const u32 crtc_offsets[] = {
 	CRTC0_REGISTER_OFFSET,
 	CRTC1_REGISTER_OFFSET,
 	CRTC2_REGISTER_OFFSET,
@@ -63,8 +64,7 @@ static const u32 crtc_offsets[] =
 	CRTC6_REGISTER_OFFSET
 };
 
-static const u32 hpd_offsets[] =
-{
+static const u32 hpd_offsets[] = {
 	HPD0_REGISTER_OFFSET,
 	HPD1_REGISTER_OFFSET,
 	HPD2_REGISTER_OFFSET,
@@ -121,30 +121,26 @@ static const struct {
 	.hpd = DISP_INTERRUPT_STATUS_CONTINUE5__DC_HPD6_INTERRUPT_MASK
 } };
 
-static const u32 golden_settings_tonga_a11[] =
-{
+static const u32 golden_settings_tonga_a11[] = {
 	mmDCI_CLK_CNTL, 0x00000080, 0x00000000,
 	mmFBC_DEBUG_COMP, 0x000000f0, 0x00000070,
 	mmFBC_MISC, 0x1f311fff, 0x12300000,
 	mmHDMI_CONTROL, 0x31000111, 0x00000011,
 };
 
-static const u32 tonga_mgcg_cgcg_init[] =
-{
+static const u32 tonga_mgcg_cgcg_init[] = {
 	mmXDMA_CLOCK_GATING_CNTL, 0xffffffff, 0x00000100,
 	mmXDMA_MEM_POWER_CNTL, 0x00000101, 0x00000000,
 };
 
-static const u32 golden_settings_fiji_a10[] =
-{
+static const u32 golden_settings_fiji_a10[] = {
 	mmDCI_CLK_CNTL, 0x00000080, 0x00000000,
 	mmFBC_DEBUG_COMP, 0x000000f0, 0x00000070,
 	mmFBC_MISC, 0x1f311fff, 0x12300000,
 	mmHDMI_CONTROL, 0x31000111, 0x00000011,
 };
 
-static const u32 fiji_mgcg_cgcg_init[] =
-{
+static const u32 fiji_mgcg_cgcg_init[] = {
 	mmXDMA_CLOCK_GATING_CNTL, 0xffffffff, 0x00000100,
 	mmXDMA_MEM_POWER_CNTL, 0x00000101, 0x00000000,
 };
@@ -369,6 +365,7 @@ static void dce_v10_0_hpd_init(struct amdgpu_device *adev)
 				    AMDGPU_HPD_DISCONNECT_INT_DELAY_IN_MS);
 		WREG32(mmDC_HPD_TOGGLE_FILT_CNTL + hpd_offsets[amdgpu_connector->hpd.hpd], tmp);
 
+		dce_v10_0_hpd_int_ack(adev, amdgpu_connector->hpd.hpd);
 		dce_v10_0_hpd_set_polarity(adev, amdgpu_connector->hpd.hpd);
 		amdgpu_irq_get(adev, &adev->hpd_irq,
 			       amdgpu_connector->hpd.hpd);
@@ -1042,7 +1039,7 @@ static void dce_v10_0_program_watermarks(struct amdgpu_device *adev,
 					    (u32)mode->clock);
 		line_time = (u32) div_u64((u64)mode->crtc_htotal * 1000000,
 					  (u32)mode->clock);
-		line_time = min(line_time, (u32)65535);
+		line_time = min_t(u32, line_time, 65535);
 
 		/* watermark for high clocks */
 		if (adev->pm.dpm_enabled) {
@@ -1072,7 +1069,7 @@ static void dce_v10_0_program_watermarks(struct amdgpu_device *adev,
 		wm_high.num_heads = num_heads;
 
 		/* set for high clocks */
-		latency_watermark_a = min(dce_v10_0_latency_watermark(&wm_high), (u32)65535);
+		latency_watermark_a = min_t(u32, dce_v10_0_latency_watermark(&wm_high), 65535);
 
 		/* possibly force display priority to high */
 		/* should really do this at mode validation time... */
@@ -1111,7 +1108,7 @@ static void dce_v10_0_program_watermarks(struct amdgpu_device *adev,
 		wm_low.num_heads = num_heads;
 
 		/* set for low clocks */
-		latency_watermark_b = min(dce_v10_0_latency_watermark(&wm_low), (u32)65535);
+		latency_watermark_b = min_t(u32, dce_v10_0_latency_watermark(&wm_low), 65535);
 
 		/* possibly force display priority to high */
 		/* should really do this at mode validation time... */
@@ -1144,8 +1141,7 @@ static void dce_v10_0_program_watermarks(struct amdgpu_device *adev,
 
 	/* save values for DPM */
 	amdgpu_crtc->line_time = line_time;
-	amdgpu_crtc->wm_high = latency_watermark_a;
-	amdgpu_crtc->wm_low = latency_watermark_b;
+
 	/* Save number of lines the linebuffer leads before the scanout */
 	amdgpu_crtc->lb_vblank_lead_lines = lb_vblank_lead_lines;
 }
@@ -1302,7 +1298,7 @@ static void dce_v10_0_audio_write_speaker_allocation(struct drm_encoder *encoder
 		return;
 	}
 
-	sad_count = drm_edid_to_speaker_allocation(amdgpu_connector_edid(connector), &sadb);
+	sad_count = drm_edid_to_speaker_allocation(amdgpu_connector->edid, &sadb);
 	if (sad_count < 0) {
 		DRM_ERROR("Couldn't read Speaker Allocation Data Block: %d\n", sad_count);
 		sad_count = 0;
@@ -1372,7 +1368,7 @@ static void dce_v10_0_audio_write_sad_regs(struct drm_encoder *encoder)
 		return;
 	}
 
-	sad_count = drm_edid_to_sad(amdgpu_connector_edid(connector), &sads);
+	sad_count = drm_edid_to_sad(amdgpu_connector->edid, &sads);
 	if (sad_count < 0)
 		DRM_ERROR("Couldn't read SADs: %d\n", sad_count);
 	if (sad_count <= 0)
@@ -1425,8 +1421,7 @@ static void dce_v10_0_audio_enable(struct amdgpu_device *adev,
 			   enable ? AZALIA_F0_CODEC_PIN_CONTROL_HOT_PLUG_CONTROL__AUDIO_ENABLED_MASK : 0);
 }
 
-static const u32 pin_offsets[] =
-{
+static const u32 pin_offsets[] = {
 	AUD0_REGISTER_OFFSET,
 	AUD1_REGISTER_OFFSET,
 	AUD2_REGISTER_OFFSET,
@@ -1466,17 +1461,12 @@ static int dce_v10_0_audio_init(struct amdgpu_device *adev)
 
 static void dce_v10_0_audio_fini(struct amdgpu_device *adev)
 {
-	int i;
-
 	if (!amdgpu_audio)
 		return;
 
 	if (!adev->mode_info.audio.enabled)
 		return;
 
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++)
-		dce_v10_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
-
 	adev->mode_info.audio.enabled = false;
 }
 
@@ -1811,8 +1801,7 @@ static void dce_v10_0_afmt_fini(struct amdgpu_device *adev)
 	}
 }
 
-static const u32 vga_control_regs[6] =
-{
+static const u32 vga_control_regs[6] = {
 	mmD1VGA_CONTROL,
 	mmD2VGA_CONTROL,
 	mmD3VGA_CONTROL,
@@ -1886,6 +1875,7 @@ static int dce_v10_0_crtc_do_set_base(struct drm_crtc *crtc,
 		return r;
 
 	if (!atomic) {
+		abo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 		r = amdgpu_bo_pin(abo, AMDGPU_GEM_DOMAIN_VRAM);
 		if (unlikely(r != 0)) {
 			amdgpu_bo_unreserve(abo);
@@ -2406,6 +2396,7 @@ static int dce_v10_0_crtc_cursor_set2(struct drm_crtc *crtc,
 		return ret;
 	}
 
+	aobj->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 	ret = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM);
 	amdgpu_bo_unreserve(aobj);
 	if (ret) {
@@ -2690,6 +2681,32 @@ static const struct drm_crtc_helper_funcs dce_v10_0_crtc_helper_funcs = {
 	.get_scanout_position = amdgpu_crtc_get_scanout_position,
 };
 
+static void dce_v10_0_panic_flush(struct drm_plane *plane)
+{
+	struct drm_framebuffer *fb;
+	struct amdgpu_crtc *amdgpu_crtc;
+	struct amdgpu_device *adev;
+	uint32_t fb_format;
+
+	if (!plane->fb)
+		return;
+
+	fb = plane->fb;
+	amdgpu_crtc = to_amdgpu_crtc(plane->crtc);
+	adev = drm_to_adev(fb->dev);
+
+	/* Disable DC tiling */
+	fb_format = RREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset);
+	fb_format &= ~GRPH_CONTROL__GRPH_ARRAY_MODE_MASK;
+	WREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset, fb_format);
+
+}
+
+static const struct drm_plane_helper_funcs dce_v10_0_drm_primary_plane_helper_funcs = {
+	.get_scanout_buffer = amdgpu_display_get_scanout_buffer,
+	.panic_flush = dce_v10_0_panic_flush,
+};
+
 static int dce_v10_0_crtc_init(struct amdgpu_device *adev, int index)
 {
 	struct amdgpu_crtc *amdgpu_crtc;
@@ -2737,13 +2754,14 @@ static int dce_v10_0_crtc_init(struct amdgpu_device *adev, int index)
 	amdgpu_crtc->encoder = NULL;
 	amdgpu_crtc->connector = NULL;
 	drm_crtc_helper_add(&amdgpu_crtc->base, &dce_v10_0_crtc_helper_funcs);
+	drm_plane_helper_add(amdgpu_crtc->base.primary, &dce_v10_0_drm_primary_plane_helper_funcs);
 
 	return 0;
 }
 
-static int dce_v10_0_early_init(void *handle)
+static int dce_v10_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->audio_endpt_rreg = &dce_v10_0_audio_endpt_rreg;
 	adev->audio_endpt_wreg = &dce_v10_0_audio_endpt_wreg;
@@ -2768,10 +2786,10 @@ static int dce_v10_0_early_init(void *handle)
 	return 0;
 }
 
-static int dce_v10_0_sw_init(void *handle)
+static int dce_v10_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->mode_info.num_crtc; i++) {
 		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + 1, &adev->crtc_irq);
@@ -2847,11 +2865,11 @@ static int dce_v10_0_sw_init(void *handle)
 	return 0;
 }
 
-static int dce_v10_0_sw_fini(void *handle)
+static int dce_v10_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	kfree(adev->mode_info.bios_hardcoded_edid);
+	drm_edid_free(adev->mode_info.bios_hardcoded_edid);
 
 	drm_kms_helper_poll_fini(adev_to_drm(adev));
 
@@ -2865,10 +2883,10 @@ static int dce_v10_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int dce_v10_0_hw_init(void *handle)
+static int dce_v10_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	dce_v10_0_init_golden_registers(adev);
 
@@ -2890,10 +2908,10 @@ static int dce_v10_0_hw_init(void *handle)
 	return 0;
 }
 
-static int dce_v10_0_hw_fini(void *handle)
+static int dce_v10_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	dce_v10_0_hpd_fini(adev);
 
@@ -2908,9 +2926,9 @@ static int dce_v10_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int dce_v10_0_suspend(void *handle)
+static int dce_v10_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_display_suspend_helper(adev);
@@ -2920,18 +2938,18 @@ static int dce_v10_0_suspend(void *handle)
 	adev->mode_info.bl_level =
 		amdgpu_atombios_encoder_get_backlight_level_from_reg(adev);
 
-	return dce_v10_0_hw_fini(handle);
+	return dce_v10_0_hw_fini(ip_block);
 }
 
-static int dce_v10_0_resume(void *handle)
+static int dce_v10_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	amdgpu_atombios_encoder_set_backlight_level_to_reg(adev,
 							   adev->mode_info.bl_level);
 
-	ret = dce_v10_0_hw_init(handle);
+	ret = dce_v10_0_hw_init(ip_block);
 
 	/* turn on the BL */
 	if (adev->mode_info.bl_encoder) {
@@ -2946,27 +2964,22 @@ static int dce_v10_0_resume(void *handle)
 	return amdgpu_display_resume_helper(adev);
 }
 
-static bool dce_v10_0_is_idle(void *handle)
+static bool dce_v10_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	return true;
 }
 
-static int dce_v10_0_wait_for_idle(void *handle)
+static bool dce_v10_0_check_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	return 0;
-}
-
-static bool dce_v10_0_check_soft_reset(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return dce_v10_0_is_display_hung(adev);
 }
 
-static int dce_v10_0_soft_reset(void *handle)
+static int dce_v10_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 srbm_soft_reset = 0, tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (dce_v10_0_is_display_hung(adev))
 		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_DC_MASK;
@@ -3056,7 +3069,7 @@ static int dce_v10_0_set_hpd_irq_state(struct amdgpu_device *adev,
 	u32 tmp;
 
 	if (hpd >= adev->mode_info.num_hpd) {
-		DRM_DEBUG("invalid hdp %d\n", hpd);
+		DRM_DEBUG("invalid hpd %d\n", hpd);
 		return 0;
 	}
 
@@ -3208,7 +3221,7 @@ static void dce_v10_0_hpd_int_ack(struct amdgpu_device *adev,
 	u32 tmp;
 
 	if (hpd >= adev->mode_info.num_hpd) {
-		DRM_DEBUG("invalid hdp %d\n", hpd);
+		DRM_DEBUG("invalid hpd %d\n", hpd);
 		return;
 	}
 
@@ -3310,13 +3323,13 @@ static int dce_v10_0_hpd_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int dce_v10_0_set_clockgating_state(void *handle,
+static int dce_v10_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int dce_v10_0_set_powergating_state(void *handle,
+static int dce_v10_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -3325,7 +3338,6 @@ static int dce_v10_0_set_powergating_state(void *handle,
 static const struct amd_ip_funcs dce_v10_0_ip_funcs = {
 	.name = "dce_v10_0",
 	.early_init = dce_v10_0_early_init,
-	.late_init = NULL,
 	.sw_init = dce_v10_0_sw_init,
 	.sw_fini = dce_v10_0_sw_fini,
 	.hw_init = dce_v10_0_hw_init,
@@ -3333,7 +3345,6 @@ static const struct amd_ip_funcs dce_v10_0_ip_funcs = {
 	.suspend = dce_v10_0_suspend,
 	.resume = dce_v10_0_resume,
 	.is_idle = dce_v10_0_is_idle,
-	.wait_for_idle = dce_v10_0_wait_for_idle,
 	.check_soft_reset = dce_v10_0_check_soft_reset,
 	.soft_reset = dce_v10_0_soft_reset,
 	.set_clockgating_state = dce_v10_0_set_clockgating_state,
@@ -3651,8 +3662,7 @@ static void dce_v10_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->hpd_irq.funcs = &dce_v10_0_hpd_irq_funcs;
 }
 
-const struct amdgpu_ip_block_version dce_v10_0_ip_block =
-{
+const struct amdgpu_ip_block_version dce_v10_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 10,
 	.minor = 0,
@@ -3660,8 +3670,7 @@ const struct amdgpu_ip_block_version dce_v10_0_ip_block =
 	.funcs = &dce_v10_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version dce_v10_1_ip_block =
-{
+const struct amdgpu_ip_block_version dce_v10_1_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 10,
 	.minor = 1,
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c
deleted file mode 100644
index c14b70350a51..000000000000
--- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c
+++ /dev/null
@@ -1,3799 +0,0 @@
-/*
- * Copyright 2014 Advanced Micro Devices, Inc.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
- * OTHER DEALINGS IN THE SOFTWARE.
- *
- */
-
-#include <drm/drm_fourcc.h>
-#include <drm/drm_modeset_helper.h>
-#include <drm/drm_modeset_helper_vtables.h>
-#include <drm/drm_vblank.h>
-
-#include "amdgpu.h"
-#include "amdgpu_pm.h"
-#include "amdgpu_i2c.h"
-#include "vid.h"
-#include "atom.h"
-#include "amdgpu_atombios.h"
-#include "atombios_crtc.h"
-#include "atombios_encoders.h"
-#include "amdgpu_pll.h"
-#include "amdgpu_connectors.h"
-#include "amdgpu_display.h"
-#include "dce_v11_0.h"
-
-#include "dce/dce_11_0_d.h"
-#include "dce/dce_11_0_sh_mask.h"
-#include "dce/dce_11_0_enum.h"
-#include "oss/oss_3_0_d.h"
-#include "oss/oss_3_0_sh_mask.h"
-#include "gmc/gmc_8_1_d.h"
-#include "gmc/gmc_8_1_sh_mask.h"
-
-#include "ivsrcid/ivsrcid_vislands30.h"
-
-static void dce_v11_0_set_display_funcs(struct amdgpu_device *adev);
-static void dce_v11_0_set_irq_funcs(struct amdgpu_device *adev);
-
-static const u32 crtc_offsets[] =
-{
-	CRTC0_REGISTER_OFFSET,
-	CRTC1_REGISTER_OFFSET,
-	CRTC2_REGISTER_OFFSET,
-	CRTC3_REGISTER_OFFSET,
-	CRTC4_REGISTER_OFFSET,
-	CRTC5_REGISTER_OFFSET,
-	CRTC6_REGISTER_OFFSET
-};
-
-static const u32 hpd_offsets[] =
-{
-	HPD0_REGISTER_OFFSET,
-	HPD1_REGISTER_OFFSET,
-	HPD2_REGISTER_OFFSET,
-	HPD3_REGISTER_OFFSET,
-	HPD4_REGISTER_OFFSET,
-	HPD5_REGISTER_OFFSET
-};
-
-static const uint32_t dig_offsets[] = {
-	DIG0_REGISTER_OFFSET,
-	DIG1_REGISTER_OFFSET,
-	DIG2_REGISTER_OFFSET,
-	DIG3_REGISTER_OFFSET,
-	DIG4_REGISTER_OFFSET,
-	DIG5_REGISTER_OFFSET,
-	DIG6_REGISTER_OFFSET,
-	DIG7_REGISTER_OFFSET,
-	DIG8_REGISTER_OFFSET
-};
-
-static const struct {
-	uint32_t        reg;
-	uint32_t        vblank;
-	uint32_t        vline;
-	uint32_t        hpd;
-
-} interrupt_status_offsets[] = { {
-	.reg = mmDISP_INTERRUPT_STATUS,
-	.vblank = DISP_INTERRUPT_STATUS__LB_D1_VBLANK_INTERRUPT_MASK,
-	.vline = DISP_INTERRUPT_STATUS__LB_D1_VLINE_INTERRUPT_MASK,
-	.hpd = DISP_INTERRUPT_STATUS__DC_HPD1_INTERRUPT_MASK
-}, {
-	.reg = mmDISP_INTERRUPT_STATUS_CONTINUE,
-	.vblank = DISP_INTERRUPT_STATUS_CONTINUE__LB_D2_VBLANK_INTERRUPT_MASK,
-	.vline = DISP_INTERRUPT_STATUS_CONTINUE__LB_D2_VLINE_INTERRUPT_MASK,
-	.hpd = DISP_INTERRUPT_STATUS_CONTINUE__DC_HPD2_INTERRUPT_MASK
-}, {
-	.reg = mmDISP_INTERRUPT_STATUS_CONTINUE2,
-	.vblank = DISP_INTERRUPT_STATUS_CONTINUE2__LB_D3_VBLANK_INTERRUPT_MASK,
-	.vline = DISP_INTERRUPT_STATUS_CONTINUE2__LB_D3_VLINE_INTERRUPT_MASK,
-	.hpd = DISP_INTERRUPT_STATUS_CONTINUE2__DC_HPD3_INTERRUPT_MASK
-}, {
-	.reg = mmDISP_INTERRUPT_STATUS_CONTINUE3,
-	.vblank = DISP_INTERRUPT_STATUS_CONTINUE3__LB_D4_VBLANK_INTERRUPT_MASK,
-	.vline = DISP_INTERRUPT_STATUS_CONTINUE3__LB_D4_VLINE_INTERRUPT_MASK,
-	.hpd = DISP_INTERRUPT_STATUS_CONTINUE3__DC_HPD4_INTERRUPT_MASK
-}, {
-	.reg = mmDISP_INTERRUPT_STATUS_CONTINUE4,
-	.vblank = DISP_INTERRUPT_STATUS_CONTINUE4__LB_D5_VBLANK_INTERRUPT_MASK,
-	.vline = DISP_INTERRUPT_STATUS_CONTINUE4__LB_D5_VLINE_INTERRUPT_MASK,
-	.hpd = DISP_INTERRUPT_STATUS_CONTINUE4__DC_HPD5_INTERRUPT_MASK
-}, {
-	.reg = mmDISP_INTERRUPT_STATUS_CONTINUE5,
-	.vblank = DISP_INTERRUPT_STATUS_CONTINUE5__LB_D6_VBLANK_INTERRUPT_MASK,
-	.vline = DISP_INTERRUPT_STATUS_CONTINUE5__LB_D6_VLINE_INTERRUPT_MASK,
-	.hpd = DISP_INTERRUPT_STATUS_CONTINUE5__DC_HPD6_INTERRUPT_MASK
-} };
-
-static const u32 cz_golden_settings_a11[] =
-{
-	mmCRTC_DOUBLE_BUFFER_CONTROL, 0x00010101, 0x00010000,
-	mmFBC_MISC, 0x1f311fff, 0x14300000,
-};
-
-static const u32 cz_mgcg_cgcg_init[] =
-{
-	mmXDMA_CLOCK_GATING_CNTL, 0xffffffff, 0x00000100,
-	mmXDMA_MEM_POWER_CNTL, 0x00000101, 0x00000000,
-};
-
-static const u32 stoney_golden_settings_a11[] =
-{
-	mmCRTC_DOUBLE_BUFFER_CONTROL, 0x00010101, 0x00010000,
-	mmFBC_MISC, 0x1f311fff, 0x14302000,
-};
-
-static const u32 polaris11_golden_settings_a11[] =
-{
-	mmDCI_CLK_CNTL, 0x00000080, 0x00000000,
-	mmFBC_DEBUG_COMP, 0x000000f0, 0x00000070,
-	mmFBC_DEBUG1, 0xffffffff, 0x00000008,
-	mmFBC_MISC, 0x9f313fff, 0x14302008,
-	mmHDMI_CONTROL, 0x313f031f, 0x00000011,
-};
-
-static const u32 polaris10_golden_settings_a11[] =
-{
-	mmDCI_CLK_CNTL, 0x00000080, 0x00000000,
-	mmFBC_DEBUG_COMP, 0x000000f0, 0x00000070,
-	mmFBC_MISC, 0x9f313fff, 0x14302008,
-	mmHDMI_CONTROL, 0x313f031f, 0x00000011,
-};
-
-static void dce_v11_0_init_golden_registers(struct amdgpu_device *adev)
-{
-	switch (adev->asic_type) {
-	case CHIP_CARRIZO:
-		amdgpu_device_program_register_sequence(adev,
-							cz_mgcg_cgcg_init,
-							ARRAY_SIZE(cz_mgcg_cgcg_init));
-		amdgpu_device_program_register_sequence(adev,
-							cz_golden_settings_a11,
-							ARRAY_SIZE(cz_golden_settings_a11));
-		break;
-	case CHIP_STONEY:
-		amdgpu_device_program_register_sequence(adev,
-							stoney_golden_settings_a11,
-							ARRAY_SIZE(stoney_golden_settings_a11));
-		break;
-	case CHIP_POLARIS11:
-	case CHIP_POLARIS12:
-		amdgpu_device_program_register_sequence(adev,
-							polaris11_golden_settings_a11,
-							ARRAY_SIZE(polaris11_golden_settings_a11));
-		break;
-	case CHIP_POLARIS10:
-	case CHIP_VEGAM:
-		amdgpu_device_program_register_sequence(adev,
-							polaris10_golden_settings_a11,
-							ARRAY_SIZE(polaris10_golden_settings_a11));
-		break;
-	default:
-		break;
-	}
-}
-
-static u32 dce_v11_0_audio_endpt_rreg(struct amdgpu_device *adev,
-				     u32 block_offset, u32 reg)
-{
-	unsigned long flags;
-	u32 r;
-
-	spin_lock_irqsave(&adev->audio_endpt_idx_lock, flags);
-	WREG32(mmAZALIA_F0_CODEC_ENDPOINT_INDEX + block_offset, reg);
-	r = RREG32(mmAZALIA_F0_CODEC_ENDPOINT_DATA + block_offset);
-	spin_unlock_irqrestore(&adev->audio_endpt_idx_lock, flags);
-
-	return r;
-}
-
-static void dce_v11_0_audio_endpt_wreg(struct amdgpu_device *adev,
-				      u32 block_offset, u32 reg, u32 v)
-{
-	unsigned long flags;
-
-	spin_lock_irqsave(&adev->audio_endpt_idx_lock, flags);
-	WREG32(mmAZALIA_F0_CODEC_ENDPOINT_INDEX + block_offset, reg);
-	WREG32(mmAZALIA_F0_CODEC_ENDPOINT_DATA + block_offset, v);
-	spin_unlock_irqrestore(&adev->audio_endpt_idx_lock, flags);
-}
-
-static u32 dce_v11_0_vblank_get_counter(struct amdgpu_device *adev, int crtc)
-{
-	if (crtc < 0 || crtc >= adev->mode_info.num_crtc)
-		return 0;
-	else
-		return RREG32(mmCRTC_STATUS_FRAME_COUNT + crtc_offsets[crtc]);
-}
-
-static void dce_v11_0_pageflip_interrupt_init(struct amdgpu_device *adev)
-{
-	unsigned i;
-
-	/* Enable pflip interrupts */
-	for (i = 0; i < adev->mode_info.num_crtc; i++)
-		amdgpu_irq_get(adev, &adev->pageflip_irq, i);
-}
-
-static void dce_v11_0_pageflip_interrupt_fini(struct amdgpu_device *adev)
-{
-	unsigned i;
-
-	/* Disable pflip interrupts */
-	for (i = 0; i < adev->mode_info.num_crtc; i++)
-		amdgpu_irq_put(adev, &adev->pageflip_irq, i);
-}
-
-/**
- * dce_v11_0_page_flip - pageflip callback.
- *
- * @adev: amdgpu_device pointer
- * @crtc_id: crtc to cleanup pageflip on
- * @crtc_base: new address of the crtc (GPU MC address)
- * @async: asynchronous flip
- *
- * Triggers the actual pageflip by updating the primary
- * surface base address.
- */
-static void dce_v11_0_page_flip(struct amdgpu_device *adev,
-				int crtc_id, u64 crtc_base, bool async)
-{
-	struct amdgpu_crtc *amdgpu_crtc = adev->mode_info.crtcs[crtc_id];
-	struct drm_framebuffer *fb = amdgpu_crtc->base.primary->fb;
-	u32 tmp;
-
-	/* flip immediate for async, default is vsync */
-	tmp = RREG32(mmGRPH_FLIP_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, GRPH_FLIP_CONTROL,
-			    GRPH_SURFACE_UPDATE_IMMEDIATE_EN, async ? 1 : 0);
-	WREG32(mmGRPH_FLIP_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-	/* update pitch */
-	WREG32(mmGRPH_PITCH + amdgpu_crtc->crtc_offset,
-	       fb->pitches[0] / fb->format->cpp[0]);
-	/* update the scanout addresses */
-	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS_HIGH + amdgpu_crtc->crtc_offset,
-	       upper_32_bits(crtc_base));
-	/* writing to the low address triggers the update */
-	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
-	       lower_32_bits(crtc_base));
-	/* post the write */
-	RREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset);
-}
-
-static int dce_v11_0_crtc_get_scanoutpos(struct amdgpu_device *adev, int crtc,
-					u32 *vbl, u32 *position)
-{
-	if ((crtc < 0) || (crtc >= adev->mode_info.num_crtc))
-		return -EINVAL;
-
-	*vbl = RREG32(mmCRTC_V_BLANK_START_END + crtc_offsets[crtc]);
-	*position = RREG32(mmCRTC_STATUS_POSITION + crtc_offsets[crtc]);
-
-	return 0;
-}
-
-/**
- * dce_v11_0_hpd_sense - hpd sense callback.
- *
- * @adev: amdgpu_device pointer
- * @hpd: hpd (hotplug detect) pin
- *
- * Checks if a digital monitor is connected (evergreen+).
- * Returns true if connected, false if not connected.
- */
-static bool dce_v11_0_hpd_sense(struct amdgpu_device *adev,
-			       enum amdgpu_hpd_id hpd)
-{
-	bool connected = false;
-
-	if (hpd >= adev->mode_info.num_hpd)
-		return connected;
-
-	if (RREG32(mmDC_HPD_INT_STATUS + hpd_offsets[hpd]) &
-	    DC_HPD_INT_STATUS__DC_HPD_SENSE_MASK)
-		connected = true;
-
-	return connected;
-}
-
-/**
- * dce_v11_0_hpd_set_polarity - hpd set polarity callback.
- *
- * @adev: amdgpu_device pointer
- * @hpd: hpd (hotplug detect) pin
- *
- * Set the polarity of the hpd pin (evergreen+).
- */
-static void dce_v11_0_hpd_set_polarity(struct amdgpu_device *adev,
-				      enum amdgpu_hpd_id hpd)
-{
-	u32 tmp;
-	bool connected = dce_v11_0_hpd_sense(adev, hpd);
-
-	if (hpd >= adev->mode_info.num_hpd)
-		return;
-
-	tmp = RREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[hpd]);
-	if (connected)
-		tmp = REG_SET_FIELD(tmp, DC_HPD_INT_CONTROL, DC_HPD_INT_POLARITY, 0);
-	else
-		tmp = REG_SET_FIELD(tmp, DC_HPD_INT_CONTROL, DC_HPD_INT_POLARITY, 1);
-	WREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[hpd], tmp);
-}
-
-/**
- * dce_v11_0_hpd_init - hpd setup callback.
- *
- * @adev: amdgpu_device pointer
- *
- * Setup the hpd pins used by the card (evergreen+).
- * Enable the pin, set the polarity, and enable the hpd interrupts.
- */
-static void dce_v11_0_hpd_init(struct amdgpu_device *adev)
-{
-	struct drm_device *dev = adev_to_drm(adev);
-	struct drm_connector *connector;
-	struct drm_connector_list_iter iter;
-	u32 tmp;
-
-	drm_connector_list_iter_begin(dev, &iter);
-	drm_for_each_connector_iter(connector, &iter) {
-		struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
-
-		if (amdgpu_connector->hpd.hpd >= adev->mode_info.num_hpd)
-			continue;
-
-		if (connector->connector_type == DRM_MODE_CONNECTOR_eDP ||
-		    connector->connector_type == DRM_MODE_CONNECTOR_LVDS) {
-			/* don't try to enable hpd on eDP or LVDS avoid breaking the
-			 * aux dp channel on imac and help (but not completely fix)
-			 * https://bugzilla.redhat.com/show_bug.cgi?id=726143
-			 * also avoid interrupt storms during dpms.
-			 */
-			tmp = RREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd]);
-			tmp = REG_SET_FIELD(tmp, DC_HPD_INT_CONTROL, DC_HPD_INT_EN, 0);
-			WREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd], tmp);
-			continue;
-		}
-
-		tmp = RREG32(mmDC_HPD_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd]);
-		tmp = REG_SET_FIELD(tmp, DC_HPD_CONTROL, DC_HPD_EN, 1);
-		WREG32(mmDC_HPD_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd], tmp);
-
-		tmp = RREG32(mmDC_HPD_TOGGLE_FILT_CNTL + hpd_offsets[amdgpu_connector->hpd.hpd]);
-		tmp = REG_SET_FIELD(tmp, DC_HPD_TOGGLE_FILT_CNTL,
-				    DC_HPD_CONNECT_INT_DELAY,
-				    AMDGPU_HPD_CONNECT_INT_DELAY_IN_MS);
-		tmp = REG_SET_FIELD(tmp, DC_HPD_TOGGLE_FILT_CNTL,
-				    DC_HPD_DISCONNECT_INT_DELAY,
-				    AMDGPU_HPD_DISCONNECT_INT_DELAY_IN_MS);
-		WREG32(mmDC_HPD_TOGGLE_FILT_CNTL + hpd_offsets[amdgpu_connector->hpd.hpd], tmp);
-
-		dce_v11_0_hpd_set_polarity(adev, amdgpu_connector->hpd.hpd);
-		amdgpu_irq_get(adev, &adev->hpd_irq, amdgpu_connector->hpd.hpd);
-	}
-	drm_connector_list_iter_end(&iter);
-}
-
-/**
- * dce_v11_0_hpd_fini - hpd tear down callback.
- *
- * @adev: amdgpu_device pointer
- *
- * Tear down the hpd pins used by the card (evergreen+).
- * Disable the hpd interrupts.
- */
-static void dce_v11_0_hpd_fini(struct amdgpu_device *adev)
-{
-	struct drm_device *dev = adev_to_drm(adev);
-	struct drm_connector *connector;
-	struct drm_connector_list_iter iter;
-	u32 tmp;
-
-	drm_connector_list_iter_begin(dev, &iter);
-	drm_for_each_connector_iter(connector, &iter) {
-		struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
-
-		if (amdgpu_connector->hpd.hpd >= adev->mode_info.num_hpd)
-			continue;
-
-		tmp = RREG32(mmDC_HPD_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd]);
-		tmp = REG_SET_FIELD(tmp, DC_HPD_CONTROL, DC_HPD_EN, 0);
-		WREG32(mmDC_HPD_CONTROL + hpd_offsets[amdgpu_connector->hpd.hpd], tmp);
-
-		amdgpu_irq_put(adev, &adev->hpd_irq, amdgpu_connector->hpd.hpd);
-	}
-	drm_connector_list_iter_end(&iter);
-}
-
-static u32 dce_v11_0_hpd_get_gpio_reg(struct amdgpu_device *adev)
-{
-	return mmDC_GPIO_HPD_A;
-}
-
-static bool dce_v11_0_is_display_hung(struct amdgpu_device *adev)
-{
-	u32 crtc_hung = 0;
-	u32 crtc_status[6];
-	u32 i, j, tmp;
-
-	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		tmp = RREG32(mmCRTC_CONTROL + crtc_offsets[i]);
-		if (REG_GET_FIELD(tmp, CRTC_CONTROL, CRTC_MASTER_EN)) {
-			crtc_status[i] = RREG32(mmCRTC_STATUS_HV_COUNT + crtc_offsets[i]);
-			crtc_hung |= (1 << i);
-		}
-	}
-
-	for (j = 0; j < 10; j++) {
-		for (i = 0; i < adev->mode_info.num_crtc; i++) {
-			if (crtc_hung & (1 << i)) {
-				tmp = RREG32(mmCRTC_STATUS_HV_COUNT + crtc_offsets[i]);
-				if (tmp != crtc_status[i])
-					crtc_hung &= ~(1 << i);
-			}
-		}
-		if (crtc_hung == 0)
-			return false;
-		udelay(100);
-	}
-
-	return true;
-}
-
-static void dce_v11_0_set_vga_render_state(struct amdgpu_device *adev,
-					   bool render)
-{
-	u32 tmp;
-
-	/* Lockout access through VGA aperture*/
-	tmp = RREG32(mmVGA_HDP_CONTROL);
-	if (render)
-		tmp = REG_SET_FIELD(tmp, VGA_HDP_CONTROL, VGA_MEMORY_DISABLE, 0);
-	else
-		tmp = REG_SET_FIELD(tmp, VGA_HDP_CONTROL, VGA_MEMORY_DISABLE, 1);
-	WREG32(mmVGA_HDP_CONTROL, tmp);
-
-	/* disable VGA render */
-	tmp = RREG32(mmVGA_RENDER_CONTROL);
-	if (render)
-		tmp = REG_SET_FIELD(tmp, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 1);
-	else
-		tmp = REG_SET_FIELD(tmp, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0);
-	WREG32(mmVGA_RENDER_CONTROL, tmp);
-}
-
-static int dce_v11_0_get_num_crtc (struct amdgpu_device *adev)
-{
-	int num_crtc = 0;
-
-	switch (adev->asic_type) {
-	case CHIP_CARRIZO:
-		num_crtc = 3;
-		break;
-	case CHIP_STONEY:
-		num_crtc = 2;
-		break;
-	case CHIP_POLARIS10:
-	case CHIP_VEGAM:
-		num_crtc = 6;
-		break;
-	case CHIP_POLARIS11:
-	case CHIP_POLARIS12:
-		num_crtc = 5;
-		break;
-	default:
-		num_crtc = 0;
-	}
-	return num_crtc;
-}
-
-void dce_v11_0_disable_dce(struct amdgpu_device *adev)
-{
-	/*Disable VGA render and enabled crtc, if has DCE engine*/
-	if (amdgpu_atombios_has_dce_engine_info(adev)) {
-		u32 tmp;
-		int crtc_enabled, i;
-
-		dce_v11_0_set_vga_render_state(adev, false);
-
-		/*Disable crtc*/
-		for (i = 0; i < dce_v11_0_get_num_crtc(adev); i++) {
-			crtc_enabled = REG_GET_FIELD(RREG32(mmCRTC_CONTROL + crtc_offsets[i]),
-									 CRTC_CONTROL, CRTC_MASTER_EN);
-			if (crtc_enabled) {
-				WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 1);
-				tmp = RREG32(mmCRTC_CONTROL + crtc_offsets[i]);
-				tmp = REG_SET_FIELD(tmp, CRTC_CONTROL, CRTC_MASTER_EN, 0);
-				WREG32(mmCRTC_CONTROL + crtc_offsets[i], tmp);
-				WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 0);
-			}
-		}
-	}
-}
-
-static void dce_v11_0_program_fmt(struct drm_encoder *encoder)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(encoder->crtc);
-	struct drm_connector *connector = amdgpu_get_connector_for_encoder(encoder);
-	int bpc = 0;
-	u32 tmp = 0;
-	enum amdgpu_connector_dither dither = AMDGPU_FMT_DITHER_DISABLE;
-
-	if (connector) {
-		struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
-		bpc = amdgpu_connector_get_monitor_bpc(connector);
-		dither = amdgpu_connector->dither;
-	}
-
-	/* LVDS/eDP FMT is set up by atom */
-	if (amdgpu_encoder->devices & ATOM_DEVICE_LCD_SUPPORT)
-		return;
-
-	/* not needed for analog */
-	if ((amdgpu_encoder->encoder_id == ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC1) ||
-	    (amdgpu_encoder->encoder_id == ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC2))
-		return;
-
-	if (bpc == 0)
-		return;
-
-	switch (bpc) {
-	case 6:
-		if (dither == AMDGPU_FMT_DITHER_ENABLE) {
-			/* XXX sort out optimal dither settings */
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_FRAME_RANDOM_ENABLE, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_HIGHPASS_RANDOM_ENABLE, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_EN, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_DEPTH, 0);
-		} else {
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_EN, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_DEPTH, 0);
-		}
-		break;
-	case 8:
-		if (dither == AMDGPU_FMT_DITHER_ENABLE) {
-			/* XXX sort out optimal dither settings */
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_FRAME_RANDOM_ENABLE, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_HIGHPASS_RANDOM_ENABLE, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_RGB_RANDOM_ENABLE, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_EN, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_DEPTH, 1);
-		} else {
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_EN, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_DEPTH, 1);
-		}
-		break;
-	case 10:
-		if (dither == AMDGPU_FMT_DITHER_ENABLE) {
-			/* XXX sort out optimal dither settings */
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_FRAME_RANDOM_ENABLE, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_HIGHPASS_RANDOM_ENABLE, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_RGB_RANDOM_ENABLE, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_EN, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_SPATIAL_DITHER_DEPTH, 2);
-		} else {
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_EN, 1);
-			tmp = REG_SET_FIELD(tmp, FMT_BIT_DEPTH_CONTROL, FMT_TRUNCATE_DEPTH, 2);
-		}
-		break;
-	default:
-		/* not needed */
-		break;
-	}
-
-	WREG32(mmFMT_BIT_DEPTH_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-}
-
-
-/* display watermark setup */
-/**
- * dce_v11_0_line_buffer_adjust - Set up the line buffer
- *
- * @adev: amdgpu_device pointer
- * @amdgpu_crtc: the selected display controller
- * @mode: the current display mode on the selected display
- * controller
- *
- * Setup up the line buffer allocation for
- * the selected display controller (CIK).
- * Returns the line buffer size in pixels.
- */
-static u32 dce_v11_0_line_buffer_adjust(struct amdgpu_device *adev,
-				       struct amdgpu_crtc *amdgpu_crtc,
-				       struct drm_display_mode *mode)
-{
-	u32 tmp, buffer_alloc, i, mem_cfg;
-	u32 pipe_offset = amdgpu_crtc->crtc_id;
-	/*
-	 * Line Buffer Setup
-	 * There are 6 line buffers, one for each display controllers.
-	 * There are 3 partitions per LB. Select the number of partitions
-	 * to enable based on the display width.  For display widths larger
-	 * than 4096, you need use to use 2 display controllers and combine
-	 * them using the stereo blender.
-	 */
-	if (amdgpu_crtc->base.enabled && mode) {
-		if (mode->crtc_hdisplay < 1920) {
-			mem_cfg = 1;
-			buffer_alloc = 2;
-		} else if (mode->crtc_hdisplay < 2560) {
-			mem_cfg = 2;
-			buffer_alloc = 2;
-		} else if (mode->crtc_hdisplay < 4096) {
-			mem_cfg = 0;
-			buffer_alloc = (adev->flags & AMD_IS_APU) ? 2 : 4;
-		} else {
-			DRM_DEBUG_KMS("Mode too big for LB!\n");
-			mem_cfg = 0;
-			buffer_alloc = (adev->flags & AMD_IS_APU) ? 2 : 4;
-		}
-	} else {
-		mem_cfg = 1;
-		buffer_alloc = 0;
-	}
-
-	tmp = RREG32(mmLB_MEMORY_CTRL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, LB_MEMORY_CTRL, LB_MEMORY_CONFIG, mem_cfg);
-	WREG32(mmLB_MEMORY_CTRL + amdgpu_crtc->crtc_offset, tmp);
-
-	tmp = RREG32(mmPIPE0_DMIF_BUFFER_CONTROL + pipe_offset);
-	tmp = REG_SET_FIELD(tmp, PIPE0_DMIF_BUFFER_CONTROL, DMIF_BUFFERS_ALLOCATED, buffer_alloc);
-	WREG32(mmPIPE0_DMIF_BUFFER_CONTROL + pipe_offset, tmp);
-
-	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32(mmPIPE0_DMIF_BUFFER_CONTROL + pipe_offset);
-		if (REG_GET_FIELD(tmp, PIPE0_DMIF_BUFFER_CONTROL, DMIF_BUFFERS_ALLOCATION_COMPLETED))
-			break;
-		udelay(1);
-	}
-
-	if (amdgpu_crtc->base.enabled && mode) {
-		switch (mem_cfg) {
-		case 0:
-		default:
-			return 4096 * 2;
-		case 1:
-			return 1920 * 2;
-		case 2:
-			return 2560 * 2;
-		}
-	}
-
-	/* controller not enabled, so no lb used */
-	return 0;
-}
-
-/**
- * cik_get_number_of_dram_channels - get the number of dram channels
- *
- * @adev: amdgpu_device pointer
- *
- * Look up the number of video ram channels (CIK).
- * Used for display watermark bandwidth calculations
- * Returns the number of dram channels
- */
-static u32 cik_get_number_of_dram_channels(struct amdgpu_device *adev)
-{
-	u32 tmp = RREG32(mmMC_SHARED_CHMAP);
-
-	switch (REG_GET_FIELD(tmp, MC_SHARED_CHMAP, NOOFCHAN)) {
-	case 0:
-	default:
-		return 1;
-	case 1:
-		return 2;
-	case 2:
-		return 4;
-	case 3:
-		return 8;
-	case 4:
-		return 3;
-	case 5:
-		return 6;
-	case 6:
-		return 10;
-	case 7:
-		return 12;
-	case 8:
-		return 16;
-	}
-}
-
-struct dce10_wm_params {
-	u32 dram_channels; /* number of dram channels */
-	u32 yclk;          /* bandwidth per dram data pin in kHz */
-	u32 sclk;          /* engine clock in kHz */
-	u32 disp_clk;      /* display clock in kHz */
-	u32 src_width;     /* viewport width */
-	u32 active_time;   /* active display time in ns */
-	u32 blank_time;    /* blank time in ns */
-	bool interlaced;    /* mode is interlaced */
-	fixed20_12 vsc;    /* vertical scale ratio */
-	u32 num_heads;     /* number of active crtcs */
-	u32 bytes_per_pixel; /* bytes per pixel display + overlay */
-	u32 lb_size;       /* line buffer allocated to pipe */
-	u32 vtaps;         /* vertical scaler taps */
-};
-
-/**
- * dce_v11_0_dram_bandwidth - get the dram bandwidth
- *
- * @wm: watermark calculation data
- *
- * Calculate the raw dram bandwidth (CIK).
- * Used for display watermark bandwidth calculations
- * Returns the dram bandwidth in MBytes/s
- */
-static u32 dce_v11_0_dram_bandwidth(struct dce10_wm_params *wm)
-{
-	/* Calculate raw DRAM Bandwidth */
-	fixed20_12 dram_efficiency; /* 0.7 */
-	fixed20_12 yclk, dram_channels, bandwidth;
-	fixed20_12 a;
-
-	a.full = dfixed_const(1000);
-	yclk.full = dfixed_const(wm->yclk);
-	yclk.full = dfixed_div(yclk, a);
-	dram_channels.full = dfixed_const(wm->dram_channels * 4);
-	a.full = dfixed_const(10);
-	dram_efficiency.full = dfixed_const(7);
-	dram_efficiency.full = dfixed_div(dram_efficiency, a);
-	bandwidth.full = dfixed_mul(dram_channels, yclk);
-	bandwidth.full = dfixed_mul(bandwidth, dram_efficiency);
-
-	return dfixed_trunc(bandwidth);
-}
-
-/**
- * dce_v11_0_dram_bandwidth_for_display - get the dram bandwidth for display
- *
- * @wm: watermark calculation data
- *
- * Calculate the dram bandwidth used for display (CIK).
- * Used for display watermark bandwidth calculations
- * Returns the dram bandwidth for display in MBytes/s
- */
-static u32 dce_v11_0_dram_bandwidth_for_display(struct dce10_wm_params *wm)
-{
-	/* Calculate DRAM Bandwidth and the part allocated to display. */
-	fixed20_12 disp_dram_allocation; /* 0.3 to 0.7 */
-	fixed20_12 yclk, dram_channels, bandwidth;
-	fixed20_12 a;
-
-	a.full = dfixed_const(1000);
-	yclk.full = dfixed_const(wm->yclk);
-	yclk.full = dfixed_div(yclk, a);
-	dram_channels.full = dfixed_const(wm->dram_channels * 4);
-	a.full = dfixed_const(10);
-	disp_dram_allocation.full = dfixed_const(3); /* XXX worse case value 0.3 */
-	disp_dram_allocation.full = dfixed_div(disp_dram_allocation, a);
-	bandwidth.full = dfixed_mul(dram_channels, yclk);
-	bandwidth.full = dfixed_mul(bandwidth, disp_dram_allocation);
-
-	return dfixed_trunc(bandwidth);
-}
-
-/**
- * dce_v11_0_data_return_bandwidth - get the data return bandwidth
- *
- * @wm: watermark calculation data
- *
- * Calculate the data return bandwidth used for display (CIK).
- * Used for display watermark bandwidth calculations
- * Returns the data return bandwidth in MBytes/s
- */
-static u32 dce_v11_0_data_return_bandwidth(struct dce10_wm_params *wm)
-{
-	/* Calculate the display Data return Bandwidth */
-	fixed20_12 return_efficiency; /* 0.8 */
-	fixed20_12 sclk, bandwidth;
-	fixed20_12 a;
-
-	a.full = dfixed_const(1000);
-	sclk.full = dfixed_const(wm->sclk);
-	sclk.full = dfixed_div(sclk, a);
-	a.full = dfixed_const(10);
-	return_efficiency.full = dfixed_const(8);
-	return_efficiency.full = dfixed_div(return_efficiency, a);
-	a.full = dfixed_const(32);
-	bandwidth.full = dfixed_mul(a, sclk);
-	bandwidth.full = dfixed_mul(bandwidth, return_efficiency);
-
-	return dfixed_trunc(bandwidth);
-}
-
-/**
- * dce_v11_0_dmif_request_bandwidth - get the dmif bandwidth
- *
- * @wm: watermark calculation data
- *
- * Calculate the dmif bandwidth used for display (CIK).
- * Used for display watermark bandwidth calculations
- * Returns the dmif bandwidth in MBytes/s
- */
-static u32 dce_v11_0_dmif_request_bandwidth(struct dce10_wm_params *wm)
-{
-	/* Calculate the DMIF Request Bandwidth */
-	fixed20_12 disp_clk_request_efficiency; /* 0.8 */
-	fixed20_12 disp_clk, bandwidth;
-	fixed20_12 a, b;
-
-	a.full = dfixed_const(1000);
-	disp_clk.full = dfixed_const(wm->disp_clk);
-	disp_clk.full = dfixed_div(disp_clk, a);
-	a.full = dfixed_const(32);
-	b.full = dfixed_mul(a, disp_clk);
-
-	a.full = dfixed_const(10);
-	disp_clk_request_efficiency.full = dfixed_const(8);
-	disp_clk_request_efficiency.full = dfixed_div(disp_clk_request_efficiency, a);
-
-	bandwidth.full = dfixed_mul(b, disp_clk_request_efficiency);
-
-	return dfixed_trunc(bandwidth);
-}
-
-/**
- * dce_v11_0_available_bandwidth - get the min available bandwidth
- *
- * @wm: watermark calculation data
- *
- * Calculate the min available bandwidth used for display (CIK).
- * Used for display watermark bandwidth calculations
- * Returns the min available bandwidth in MBytes/s
- */
-static u32 dce_v11_0_available_bandwidth(struct dce10_wm_params *wm)
-{
-	/* Calculate the Available bandwidth. Display can use this temporarily but not in average. */
-	u32 dram_bandwidth = dce_v11_0_dram_bandwidth(wm);
-	u32 data_return_bandwidth = dce_v11_0_data_return_bandwidth(wm);
-	u32 dmif_req_bandwidth = dce_v11_0_dmif_request_bandwidth(wm);
-
-	return min(dram_bandwidth, min(data_return_bandwidth, dmif_req_bandwidth));
-}
-
-/**
- * dce_v11_0_average_bandwidth - get the average available bandwidth
- *
- * @wm: watermark calculation data
- *
- * Calculate the average available bandwidth used for display (CIK).
- * Used for display watermark bandwidth calculations
- * Returns the average available bandwidth in MBytes/s
- */
-static u32 dce_v11_0_average_bandwidth(struct dce10_wm_params *wm)
-{
-	/* Calculate the display mode Average Bandwidth
-	 * DisplayMode should contain the source and destination dimensions,
-	 * timing, etc.
-	 */
-	fixed20_12 bpp;
-	fixed20_12 line_time;
-	fixed20_12 src_width;
-	fixed20_12 bandwidth;
-	fixed20_12 a;
-
-	a.full = dfixed_const(1000);
-	line_time.full = dfixed_const(wm->active_time + wm->blank_time);
-	line_time.full = dfixed_div(line_time, a);
-	bpp.full = dfixed_const(wm->bytes_per_pixel);
-	src_width.full = dfixed_const(wm->src_width);
-	bandwidth.full = dfixed_mul(src_width, bpp);
-	bandwidth.full = dfixed_mul(bandwidth, wm->vsc);
-	bandwidth.full = dfixed_div(bandwidth, line_time);
-
-	return dfixed_trunc(bandwidth);
-}
-
-/**
- * dce_v11_0_latency_watermark - get the latency watermark
- *
- * @wm: watermark calculation data
- *
- * Calculate the latency watermark (CIK).
- * Used for display watermark bandwidth calculations
- * Returns the latency watermark in ns
- */
-static u32 dce_v11_0_latency_watermark(struct dce10_wm_params *wm)
-{
-	/* First calculate the latency in ns */
-	u32 mc_latency = 2000; /* 2000 ns. */
-	u32 available_bandwidth = dce_v11_0_available_bandwidth(wm);
-	u32 worst_chunk_return_time = (512 * 8 * 1000) / available_bandwidth;
-	u32 cursor_line_pair_return_time = (128 * 4 * 1000) / available_bandwidth;
-	u32 dc_latency = 40000000 / wm->disp_clk; /* dc pipe latency */
-	u32 other_heads_data_return_time = ((wm->num_heads + 1) * worst_chunk_return_time) +
-		(wm->num_heads * cursor_line_pair_return_time);
-	u32 latency = mc_latency + other_heads_data_return_time + dc_latency;
-	u32 max_src_lines_per_dst_line, lb_fill_bw, line_fill_time;
-	u32 tmp, dmif_size = 12288;
-	fixed20_12 a, b, c;
-
-	if (wm->num_heads == 0)
-		return 0;
-
-	a.full = dfixed_const(2);
-	b.full = dfixed_const(1);
-	if ((wm->vsc.full > a.full) ||
-	    ((wm->vsc.full > b.full) && (wm->vtaps >= 3)) ||
-	    (wm->vtaps >= 5) ||
-	    ((wm->vsc.full >= a.full) && wm->interlaced))
-		max_src_lines_per_dst_line = 4;
-	else
-		max_src_lines_per_dst_line = 2;
-
-	a.full = dfixed_const(available_bandwidth);
-	b.full = dfixed_const(wm->num_heads);
-	a.full = dfixed_div(a, b);
-	tmp = div_u64((u64) dmif_size * (u64) wm->disp_clk, mc_latency + 512);
-	tmp = min(dfixed_trunc(a), tmp);
-
-	lb_fill_bw = min(tmp, wm->disp_clk * wm->bytes_per_pixel / 1000);
-
-	a.full = dfixed_const(max_src_lines_per_dst_line * wm->src_width * wm->bytes_per_pixel);
-	b.full = dfixed_const(1000);
-	c.full = dfixed_const(lb_fill_bw);
-	b.full = dfixed_div(c, b);
-	a.full = dfixed_div(a, b);
-	line_fill_time = dfixed_trunc(a);
-
-	if (line_fill_time < wm->active_time)
-		return latency;
-	else
-		return latency + (line_fill_time - wm->active_time);
-
-}
-
-/**
- * dce_v11_0_average_bandwidth_vs_dram_bandwidth_for_display - check
- * average and available dram bandwidth
- *
- * @wm: watermark calculation data
- *
- * Check if the display average bandwidth fits in the display
- * dram bandwidth (CIK).
- * Used for display watermark bandwidth calculations
- * Returns true if the display fits, false if not.
- */
-static bool dce_v11_0_average_bandwidth_vs_dram_bandwidth_for_display(struct dce10_wm_params *wm)
-{
-	if (dce_v11_0_average_bandwidth(wm) <=
-	    (dce_v11_0_dram_bandwidth_for_display(wm) / wm->num_heads))
-		return true;
-	else
-		return false;
-}
-
-/**
- * dce_v11_0_average_bandwidth_vs_available_bandwidth - check
- * average and available bandwidth
- *
- * @wm: watermark calculation data
- *
- * Check if the display average bandwidth fits in the display
- * available bandwidth (CIK).
- * Used for display watermark bandwidth calculations
- * Returns true if the display fits, false if not.
- */
-static bool dce_v11_0_average_bandwidth_vs_available_bandwidth(struct dce10_wm_params *wm)
-{
-	if (dce_v11_0_average_bandwidth(wm) <=
-	    (dce_v11_0_available_bandwidth(wm) / wm->num_heads))
-		return true;
-	else
-		return false;
-}
-
-/**
- * dce_v11_0_check_latency_hiding - check latency hiding
- *
- * @wm: watermark calculation data
- *
- * Check latency hiding (CIK).
- * Used for display watermark bandwidth calculations
- * Returns true if the display fits, false if not.
- */
-static bool dce_v11_0_check_latency_hiding(struct dce10_wm_params *wm)
-{
-	u32 lb_partitions = wm->lb_size / wm->src_width;
-	u32 line_time = wm->active_time + wm->blank_time;
-	u32 latency_tolerant_lines;
-	u32 latency_hiding;
-	fixed20_12 a;
-
-	a.full = dfixed_const(1);
-	if (wm->vsc.full > a.full)
-		latency_tolerant_lines = 1;
-	else {
-		if (lb_partitions <= (wm->vtaps + 1))
-			latency_tolerant_lines = 1;
-		else
-			latency_tolerant_lines = 2;
-	}
-
-	latency_hiding = (latency_tolerant_lines * line_time + wm->blank_time);
-
-	if (dce_v11_0_latency_watermark(wm) <= latency_hiding)
-		return true;
-	else
-		return false;
-}
-
-/**
- * dce_v11_0_program_watermarks - program display watermarks
- *
- * @adev: amdgpu_device pointer
- * @amdgpu_crtc: the selected display controller
- * @lb_size: line buffer size
- * @num_heads: number of display controllers in use
- *
- * Calculate and program the display watermarks for the
- * selected display controller (CIK).
- */
-static void dce_v11_0_program_watermarks(struct amdgpu_device *adev,
-					struct amdgpu_crtc *amdgpu_crtc,
-					u32 lb_size, u32 num_heads)
-{
-	struct drm_display_mode *mode = &amdgpu_crtc->base.mode;
-	struct dce10_wm_params wm_low, wm_high;
-	u32 active_time;
-	u32 line_time = 0;
-	u32 latency_watermark_a = 0, latency_watermark_b = 0;
-	u32 tmp, wm_mask, lb_vblank_lead_lines = 0;
-
-	if (amdgpu_crtc->base.enabled && num_heads && mode) {
-		active_time = (u32) div_u64((u64)mode->crtc_hdisplay * 1000000,
-					    (u32)mode->clock);
-		line_time = (u32) div_u64((u64)mode->crtc_htotal * 1000000,
-					  (u32)mode->clock);
-		line_time = min(line_time, (u32)65535);
-
-		/* watermark for high clocks */
-		if (adev->pm.dpm_enabled) {
-			wm_high.yclk =
-				amdgpu_dpm_get_mclk(adev, false) * 10;
-			wm_high.sclk =
-				amdgpu_dpm_get_sclk(adev, false) * 10;
-		} else {
-			wm_high.yclk = adev->pm.current_mclk * 10;
-			wm_high.sclk = adev->pm.current_sclk * 10;
-		}
-
-		wm_high.disp_clk = mode->clock;
-		wm_high.src_width = mode->crtc_hdisplay;
-		wm_high.active_time = active_time;
-		wm_high.blank_time = line_time - wm_high.active_time;
-		wm_high.interlaced = false;
-		if (mode->flags & DRM_MODE_FLAG_INTERLACE)
-			wm_high.interlaced = true;
-		wm_high.vsc = amdgpu_crtc->vsc;
-		wm_high.vtaps = 1;
-		if (amdgpu_crtc->rmx_type != RMX_OFF)
-			wm_high.vtaps = 2;
-		wm_high.bytes_per_pixel = 4; /* XXX: get this from fb config */
-		wm_high.lb_size = lb_size;
-		wm_high.dram_channels = cik_get_number_of_dram_channels(adev);
-		wm_high.num_heads = num_heads;
-
-		/* set for high clocks */
-		latency_watermark_a = min(dce_v11_0_latency_watermark(&wm_high), (u32)65535);
-
-		/* possibly force display priority to high */
-		/* should really do this at mode validation time... */
-		if (!dce_v11_0_average_bandwidth_vs_dram_bandwidth_for_display(&wm_high) ||
-		    !dce_v11_0_average_bandwidth_vs_available_bandwidth(&wm_high) ||
-		    !dce_v11_0_check_latency_hiding(&wm_high) ||
-		    (adev->mode_info.disp_priority == 2)) {
-			DRM_DEBUG_KMS("force priority to high\n");
-		}
-
-		/* watermark for low clocks */
-		if (adev->pm.dpm_enabled) {
-			wm_low.yclk =
-				amdgpu_dpm_get_mclk(adev, true) * 10;
-			wm_low.sclk =
-				amdgpu_dpm_get_sclk(adev, true) * 10;
-		} else {
-			wm_low.yclk = adev->pm.current_mclk * 10;
-			wm_low.sclk = adev->pm.current_sclk * 10;
-		}
-
-		wm_low.disp_clk = mode->clock;
-		wm_low.src_width = mode->crtc_hdisplay;
-		wm_low.active_time = active_time;
-		wm_low.blank_time = line_time - wm_low.active_time;
-		wm_low.interlaced = false;
-		if (mode->flags & DRM_MODE_FLAG_INTERLACE)
-			wm_low.interlaced = true;
-		wm_low.vsc = amdgpu_crtc->vsc;
-		wm_low.vtaps = 1;
-		if (amdgpu_crtc->rmx_type != RMX_OFF)
-			wm_low.vtaps = 2;
-		wm_low.bytes_per_pixel = 4; /* XXX: get this from fb config */
-		wm_low.lb_size = lb_size;
-		wm_low.dram_channels = cik_get_number_of_dram_channels(adev);
-		wm_low.num_heads = num_heads;
-
-		/* set for low clocks */
-		latency_watermark_b = min(dce_v11_0_latency_watermark(&wm_low), (u32)65535);
-
-		/* possibly force display priority to high */
-		/* should really do this at mode validation time... */
-		if (!dce_v11_0_average_bandwidth_vs_dram_bandwidth_for_display(&wm_low) ||
-		    !dce_v11_0_average_bandwidth_vs_available_bandwidth(&wm_low) ||
-		    !dce_v11_0_check_latency_hiding(&wm_low) ||
-		    (adev->mode_info.disp_priority == 2)) {
-			DRM_DEBUG_KMS("force priority to high\n");
-		}
-		lb_vblank_lead_lines = DIV_ROUND_UP(lb_size, mode->crtc_hdisplay);
-	}
-
-	/* select wm A */
-	wm_mask = RREG32(mmDPG_WATERMARK_MASK_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(wm_mask, DPG_WATERMARK_MASK_CONTROL, URGENCY_WATERMARK_MASK, 1);
-	WREG32(mmDPG_WATERMARK_MASK_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-	tmp = RREG32(mmDPG_PIPE_URGENCY_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, DPG_PIPE_URGENCY_CONTROL, URGENCY_LOW_WATERMARK, latency_watermark_a);
-	tmp = REG_SET_FIELD(tmp, DPG_PIPE_URGENCY_CONTROL, URGENCY_HIGH_WATERMARK, line_time);
-	WREG32(mmDPG_PIPE_URGENCY_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-	/* select wm B */
-	tmp = REG_SET_FIELD(wm_mask, DPG_WATERMARK_MASK_CONTROL, URGENCY_WATERMARK_MASK, 2);
-	WREG32(mmDPG_WATERMARK_MASK_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-	tmp = RREG32(mmDPG_PIPE_URGENCY_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, DPG_PIPE_URGENCY_CONTROL, URGENCY_LOW_WATERMARK, latency_watermark_b);
-	tmp = REG_SET_FIELD(tmp, DPG_PIPE_URGENCY_CONTROL, URGENCY_HIGH_WATERMARK, line_time);
-	WREG32(mmDPG_PIPE_URGENCY_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-	/* restore original selection */
-	WREG32(mmDPG_WATERMARK_MASK_CONTROL + amdgpu_crtc->crtc_offset, wm_mask);
-
-	/* save values for DPM */
-	amdgpu_crtc->line_time = line_time;
-	amdgpu_crtc->wm_high = latency_watermark_a;
-	amdgpu_crtc->wm_low = latency_watermark_b;
-	/* Save number of lines the linebuffer leads before the scanout */
-	amdgpu_crtc->lb_vblank_lead_lines = lb_vblank_lead_lines;
-}
-
-/**
- * dce_v11_0_bandwidth_update - program display watermarks
- *
- * @adev: amdgpu_device pointer
- *
- * Calculate and program the display watermarks and line
- * buffer allocation (CIK).
- */
-static void dce_v11_0_bandwidth_update(struct amdgpu_device *adev)
-{
-	struct drm_display_mode *mode = NULL;
-	u32 num_heads = 0, lb_size;
-	int i;
-
-	amdgpu_display_update_priority(adev);
-
-	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		if (adev->mode_info.crtcs[i]->base.enabled)
-			num_heads++;
-	}
-	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		mode = &adev->mode_info.crtcs[i]->base.mode;
-		lb_size = dce_v11_0_line_buffer_adjust(adev, adev->mode_info.crtcs[i], mode);
-		dce_v11_0_program_watermarks(adev, adev->mode_info.crtcs[i],
-					    lb_size, num_heads);
-	}
-}
-
-static void dce_v11_0_audio_get_connected_pins(struct amdgpu_device *adev)
-{
-	int i;
-	u32 offset, tmp;
-
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++) {
-		offset = adev->mode_info.audio.pin[i].offset;
-		tmp = RREG32_AUDIO_ENDPT(offset,
-					 ixAZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_CONFIGURATION_DEFAULT);
-		if (((tmp &
-		AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_CONFIGURATION_DEFAULT__PORT_CONNECTIVITY_MASK) >>
-		AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_CONFIGURATION_DEFAULT__PORT_CONNECTIVITY__SHIFT) == 1)
-			adev->mode_info.audio.pin[i].connected = false;
-		else
-			adev->mode_info.audio.pin[i].connected = true;
-	}
-}
-
-static struct amdgpu_audio_pin *dce_v11_0_audio_get_pin(struct amdgpu_device *adev)
-{
-	int i;
-
-	dce_v11_0_audio_get_connected_pins(adev);
-
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++) {
-		if (adev->mode_info.audio.pin[i].connected)
-			return &adev->mode_info.audio.pin[i];
-	}
-	DRM_ERROR("No connected audio pins found!\n");
-	return NULL;
-}
-
-static void dce_v11_0_afmt_audio_select_pin(struct drm_encoder *encoder)
-{
-	struct amdgpu_device *adev = drm_to_adev(encoder->dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-	u32 tmp;
-
-	if (!dig || !dig->afmt || !dig->afmt->pin)
-		return;
-
-	tmp = RREG32(mmAFMT_AUDIO_SRC_CONTROL + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, AFMT_AUDIO_SRC_CONTROL, AFMT_AUDIO_SRC_SELECT, dig->afmt->pin->id);
-	WREG32(mmAFMT_AUDIO_SRC_CONTROL + dig->afmt->offset, tmp);
-}
-
-static void dce_v11_0_audio_write_latency_fields(struct drm_encoder *encoder,
-						struct drm_display_mode *mode)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-	struct drm_connector *connector;
-	struct drm_connector_list_iter iter;
-	struct amdgpu_connector *amdgpu_connector = NULL;
-	u32 tmp;
-	int interlace = 0;
-
-	if (!dig || !dig->afmt || !dig->afmt->pin)
-		return;
-
-	drm_connector_list_iter_begin(dev, &iter);
-	drm_for_each_connector_iter(connector, &iter) {
-		if (connector->encoder == encoder) {
-			amdgpu_connector = to_amdgpu_connector(connector);
-			break;
-		}
-	}
-	drm_connector_list_iter_end(&iter);
-
-	if (!amdgpu_connector) {
-		DRM_ERROR("Couldn't find encoder's connector\n");
-		return;
-	}
-
-	if (mode->flags & DRM_MODE_FLAG_INTERLACE)
-		interlace = 1;
-	if (connector->latency_present[interlace]) {
-		tmp = REG_SET_FIELD(0, AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC,
-				    VIDEO_LIPSYNC, connector->video_latency[interlace]);
-		tmp = REG_SET_FIELD(0, AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC,
-				    AUDIO_LIPSYNC, connector->audio_latency[interlace]);
-	} else {
-		tmp = REG_SET_FIELD(0, AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC,
-				    VIDEO_LIPSYNC, 0);
-		tmp = REG_SET_FIELD(0, AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC,
-				    AUDIO_LIPSYNC, 0);
-	}
-	WREG32_AUDIO_ENDPT(dig->afmt->pin->offset,
-			   ixAZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC, tmp);
-}
-
-static void dce_v11_0_audio_write_speaker_allocation(struct drm_encoder *encoder)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-	struct drm_connector *connector;
-	struct drm_connector_list_iter iter;
-	struct amdgpu_connector *amdgpu_connector = NULL;
-	u32 tmp;
-	u8 *sadb = NULL;
-	int sad_count;
-
-	if (!dig || !dig->afmt || !dig->afmt->pin)
-		return;
-
-	drm_connector_list_iter_begin(dev, &iter);
-	drm_for_each_connector_iter(connector, &iter) {
-		if (connector->encoder == encoder) {
-			amdgpu_connector = to_amdgpu_connector(connector);
-			break;
-		}
-	}
-	drm_connector_list_iter_end(&iter);
-
-	if (!amdgpu_connector) {
-		DRM_ERROR("Couldn't find encoder's connector\n");
-		return;
-	}
-
-	sad_count = drm_edid_to_speaker_allocation(amdgpu_connector_edid(connector), &sadb);
-	if (sad_count < 0) {
-		DRM_ERROR("Couldn't read Speaker Allocation Data Block: %d\n", sad_count);
-		sad_count = 0;
-	}
-
-	/* program the speaker allocation */
-	tmp = RREG32_AUDIO_ENDPT(dig->afmt->pin->offset,
-				 ixAZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER);
-	tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER,
-			    DP_CONNECTION, 0);
-	/* set HDMI mode */
-	tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER,
-			    HDMI_CONNECTION, 1);
-	if (sad_count)
-		tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER,
-				    SPEAKER_ALLOCATION, sadb[0]);
-	else
-		tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER,
-				    SPEAKER_ALLOCATION, 5); /* stereo */
-	WREG32_AUDIO_ENDPT(dig->afmt->pin->offset,
-			   ixAZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER, tmp);
-
-	kfree(sadb);
-}
-
-static void dce_v11_0_audio_write_sad_regs(struct drm_encoder *encoder)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-	struct drm_connector *connector;
-	struct drm_connector_list_iter iter;
-	struct amdgpu_connector *amdgpu_connector = NULL;
-	struct cea_sad *sads;
-	int i, sad_count;
-
-	static const u16 eld_reg_to_type[][2] = {
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0, HDMI_AUDIO_CODING_TYPE_PCM },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR1, HDMI_AUDIO_CODING_TYPE_AC3 },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR2, HDMI_AUDIO_CODING_TYPE_MPEG1 },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR3, HDMI_AUDIO_CODING_TYPE_MP3 },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR4, HDMI_AUDIO_CODING_TYPE_MPEG2 },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR5, HDMI_AUDIO_CODING_TYPE_AAC_LC },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR6, HDMI_AUDIO_CODING_TYPE_DTS },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR7, HDMI_AUDIO_CODING_TYPE_ATRAC },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR9, HDMI_AUDIO_CODING_TYPE_EAC3 },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR10, HDMI_AUDIO_CODING_TYPE_DTS_HD },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR11, HDMI_AUDIO_CODING_TYPE_MLP },
-		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR13, HDMI_AUDIO_CODING_TYPE_WMA_PRO },
-	};
-
-	if (!dig || !dig->afmt || !dig->afmt->pin)
-		return;
-
-	drm_connector_list_iter_begin(dev, &iter);
-	drm_for_each_connector_iter(connector, &iter) {
-		if (connector->encoder == encoder) {
-			amdgpu_connector = to_amdgpu_connector(connector);
-			break;
-		}
-	}
-	drm_connector_list_iter_end(&iter);
-
-	if (!amdgpu_connector) {
-		DRM_ERROR("Couldn't find encoder's connector\n");
-		return;
-	}
-
-	sad_count = drm_edid_to_sad(amdgpu_connector_edid(connector), &sads);
-	if (sad_count < 0)
-		DRM_ERROR("Couldn't read SADs: %d\n", sad_count);
-	if (sad_count <= 0)
-		return;
-	BUG_ON(!sads);
-
-	for (i = 0; i < ARRAY_SIZE(eld_reg_to_type); i++) {
-		u32 tmp = 0;
-		u8 stereo_freqs = 0;
-		int max_channels = -1;
-		int j;
-
-		for (j = 0; j < sad_count; j++) {
-			struct cea_sad *sad = &sads[j];
-
-			if (sad->format == eld_reg_to_type[i][1]) {
-				if (sad->channels > max_channels) {
-					tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0,
-							    MAX_CHANNELS, sad->channels);
-					tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0,
-							    DESCRIPTOR_BYTE_2, sad->byte2);
-					tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0,
-							    SUPPORTED_FREQUENCIES, sad->freq);
-					max_channels = sad->channels;
-				}
-
-				if (sad->format == HDMI_AUDIO_CODING_TYPE_PCM)
-					stereo_freqs |= sad->freq;
-				else
-					break;
-			}
-		}
-
-		tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0,
-				    SUPPORTED_FREQUENCIES_STEREO, stereo_freqs);
-		WREG32_AUDIO_ENDPT(dig->afmt->pin->offset, eld_reg_to_type[i][0], tmp);
-	}
-
-	kfree(sads);
-}
-
-static void dce_v11_0_audio_enable(struct amdgpu_device *adev,
-				  struct amdgpu_audio_pin *pin,
-				  bool enable)
-{
-	if (!pin)
-		return;
-
-	WREG32_AUDIO_ENDPT(pin->offset, ixAZALIA_F0_CODEC_PIN_CONTROL_HOT_PLUG_CONTROL,
-			   enable ? AZALIA_F0_CODEC_PIN_CONTROL_HOT_PLUG_CONTROL__AUDIO_ENABLED_MASK : 0);
-}
-
-static const u32 pin_offsets[] =
-{
-	AUD0_REGISTER_OFFSET,
-	AUD1_REGISTER_OFFSET,
-	AUD2_REGISTER_OFFSET,
-	AUD3_REGISTER_OFFSET,
-	AUD4_REGISTER_OFFSET,
-	AUD5_REGISTER_OFFSET,
-	AUD6_REGISTER_OFFSET,
-	AUD7_REGISTER_OFFSET,
-};
-
-static int dce_v11_0_audio_init(struct amdgpu_device *adev)
-{
-	int i;
-
-	if (!amdgpu_audio)
-		return 0;
-
-	adev->mode_info.audio.enabled = true;
-
-	switch (adev->asic_type) {
-	case CHIP_CARRIZO:
-	case CHIP_STONEY:
-		adev->mode_info.audio.num_pins = 7;
-		break;
-	case CHIP_POLARIS10:
-	case CHIP_VEGAM:
-		adev->mode_info.audio.num_pins = 8;
-		break;
-	case CHIP_POLARIS11:
-	case CHIP_POLARIS12:
-		adev->mode_info.audio.num_pins = 6;
-		break;
-	default:
-		return -EINVAL;
-	}
-
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++) {
-		adev->mode_info.audio.pin[i].channels = -1;
-		adev->mode_info.audio.pin[i].rate = -1;
-		adev->mode_info.audio.pin[i].bits_per_sample = -1;
-		adev->mode_info.audio.pin[i].status_bits = 0;
-		adev->mode_info.audio.pin[i].category_code = 0;
-		adev->mode_info.audio.pin[i].connected = false;
-		adev->mode_info.audio.pin[i].offset = pin_offsets[i];
-		adev->mode_info.audio.pin[i].id = i;
-		/* disable audio.  it will be set up later */
-		/* XXX remove once we switch to ip funcs */
-		dce_v11_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
-	}
-
-	return 0;
-}
-
-static void dce_v11_0_audio_fini(struct amdgpu_device *adev)
-{
-	int i;
-
-	if (!amdgpu_audio)
-		return;
-
-	if (!adev->mode_info.audio.enabled)
-		return;
-
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++)
-		dce_v11_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
-
-	adev->mode_info.audio.enabled = false;
-}
-
-/*
- * update the N and CTS parameters for a given pixel clock rate
- */
-static void dce_v11_0_afmt_update_ACR(struct drm_encoder *encoder, uint32_t clock)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_afmt_acr acr = amdgpu_afmt_acr(clock);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-	u32 tmp;
-
-	tmp = RREG32(mmHDMI_ACR_32_0 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_ACR_32_0, HDMI_ACR_CTS_32, acr.cts_32khz);
-	WREG32(mmHDMI_ACR_32_0 + dig->afmt->offset, tmp);
-	tmp = RREG32(mmHDMI_ACR_32_1 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_ACR_32_1, HDMI_ACR_N_32, acr.n_32khz);
-	WREG32(mmHDMI_ACR_32_1 + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmHDMI_ACR_44_0 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_ACR_44_0, HDMI_ACR_CTS_44, acr.cts_44_1khz);
-	WREG32(mmHDMI_ACR_44_0 + dig->afmt->offset, tmp);
-	tmp = RREG32(mmHDMI_ACR_44_1 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_ACR_44_1, HDMI_ACR_N_44, acr.n_44_1khz);
-	WREG32(mmHDMI_ACR_44_1 + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmHDMI_ACR_48_0 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_ACR_48_0, HDMI_ACR_CTS_48, acr.cts_48khz);
-	WREG32(mmHDMI_ACR_48_0 + dig->afmt->offset, tmp);
-	tmp = RREG32(mmHDMI_ACR_48_1 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_ACR_48_1, HDMI_ACR_N_48, acr.n_48khz);
-	WREG32(mmHDMI_ACR_48_1 + dig->afmt->offset, tmp);
-
-}
-
-/*
- * build a HDMI Video Info Frame
- */
-static void dce_v11_0_afmt_update_avi_infoframe(struct drm_encoder *encoder,
-					       void *buffer, size_t size)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-	uint8_t *frame = buffer + 3;
-	uint8_t *header = buffer;
-
-	WREG32(mmAFMT_AVI_INFO0 + dig->afmt->offset,
-		frame[0x0] | (frame[0x1] << 8) | (frame[0x2] << 16) | (frame[0x3] << 24));
-	WREG32(mmAFMT_AVI_INFO1 + dig->afmt->offset,
-		frame[0x4] | (frame[0x5] << 8) | (frame[0x6] << 16) | (frame[0x7] << 24));
-	WREG32(mmAFMT_AVI_INFO2 + dig->afmt->offset,
-		frame[0x8] | (frame[0x9] << 8) | (frame[0xA] << 16) | (frame[0xB] << 24));
-	WREG32(mmAFMT_AVI_INFO3 + dig->afmt->offset,
-		frame[0xC] | (frame[0xD] << 8) | (header[1] << 24));
-}
-
-static void dce_v11_0_audio_set_dto(struct drm_encoder *encoder, u32 clock)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(encoder->crtc);
-	u32 dto_phase = 24 * 1000;
-	u32 dto_modulo = clock;
-	u32 tmp;
-
-	if (!dig || !dig->afmt)
-		return;
-
-	/* XXX two dtos; generally use dto0 for hdmi */
-	/* Express [24MHz / target pixel clock] as an exact rational
-	 * number (coefficient of two integer numbers.  DCCG_AUDIO_DTOx_PHASE
-	 * is the numerator, DCCG_AUDIO_DTOx_MODULE is the denominator
-	 */
-	tmp = RREG32(mmDCCG_AUDIO_DTO_SOURCE);
-	tmp = REG_SET_FIELD(tmp, DCCG_AUDIO_DTO_SOURCE, DCCG_AUDIO_DTO0_SOURCE_SEL,
-			    amdgpu_crtc->crtc_id);
-	WREG32(mmDCCG_AUDIO_DTO_SOURCE, tmp);
-	WREG32(mmDCCG_AUDIO_DTO0_PHASE, dto_phase);
-	WREG32(mmDCCG_AUDIO_DTO0_MODULE, dto_modulo);
-}
-
-/*
- * update the info frames with the data from the current display mode
- */
-static void dce_v11_0_afmt_setmode(struct drm_encoder *encoder,
-				  struct drm_display_mode *mode)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-	struct drm_connector *connector = amdgpu_get_connector_for_encoder(encoder);
-	u8 buffer[HDMI_INFOFRAME_HEADER_SIZE + HDMI_AVI_INFOFRAME_SIZE];
-	struct hdmi_avi_infoframe frame;
-	ssize_t err;
-	u32 tmp;
-	int bpc = 8;
-
-	if (!dig || !dig->afmt)
-		return;
-
-	/* Silent, r600_hdmi_enable will raise WARN for us */
-	if (!dig->afmt->enabled)
-		return;
-
-	/* hdmi deep color mode general control packets setup, if bpc > 8 */
-	if (encoder->crtc) {
-		struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(encoder->crtc);
-		bpc = amdgpu_crtc->bpc;
-	}
-
-	/* disable audio prior to setting up hw */
-	dig->afmt->pin = dce_v11_0_audio_get_pin(adev);
-	dce_v11_0_audio_enable(adev, dig->afmt->pin, false);
-
-	dce_v11_0_audio_set_dto(encoder, mode->clock);
-
-	tmp = RREG32(mmHDMI_VBI_PACKET_CONTROL + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_VBI_PACKET_CONTROL, HDMI_NULL_SEND, 1);
-	WREG32(mmHDMI_VBI_PACKET_CONTROL + dig->afmt->offset, tmp); /* send null packets when required */
-
-	WREG32(mmAFMT_AUDIO_CRC_CONTROL + dig->afmt->offset, 0x1000);
-
-	tmp = RREG32(mmHDMI_CONTROL + dig->afmt->offset);
-	switch (bpc) {
-	case 0:
-	case 6:
-	case 8:
-	case 16:
-	default:
-		tmp = REG_SET_FIELD(tmp, HDMI_CONTROL, HDMI_DEEP_COLOR_ENABLE, 0);
-		tmp = REG_SET_FIELD(tmp, HDMI_CONTROL, HDMI_DEEP_COLOR_DEPTH, 0);
-		DRM_DEBUG("%s: Disabling hdmi deep color for %d bpc.\n",
-			  connector->name, bpc);
-		break;
-	case 10:
-		tmp = REG_SET_FIELD(tmp, HDMI_CONTROL, HDMI_DEEP_COLOR_ENABLE, 1);
-		tmp = REG_SET_FIELD(tmp, HDMI_CONTROL, HDMI_DEEP_COLOR_DEPTH, 1);
-		DRM_DEBUG("%s: Enabling hdmi deep color 30 for 10 bpc.\n",
-			  connector->name);
-		break;
-	case 12:
-		tmp = REG_SET_FIELD(tmp, HDMI_CONTROL, HDMI_DEEP_COLOR_ENABLE, 1);
-		tmp = REG_SET_FIELD(tmp, HDMI_CONTROL, HDMI_DEEP_COLOR_DEPTH, 2);
-		DRM_DEBUG("%s: Enabling hdmi deep color 36 for 12 bpc.\n",
-			  connector->name);
-		break;
-	}
-	WREG32(mmHDMI_CONTROL + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmHDMI_VBI_PACKET_CONTROL + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_VBI_PACKET_CONTROL, HDMI_NULL_SEND, 1); /* send null packets when required */
-	tmp = REG_SET_FIELD(tmp, HDMI_VBI_PACKET_CONTROL, HDMI_GC_SEND, 1); /* send general control packets */
-	tmp = REG_SET_FIELD(tmp, HDMI_VBI_PACKET_CONTROL, HDMI_GC_CONT, 1); /* send general control packets every frame */
-	WREG32(mmHDMI_VBI_PACKET_CONTROL + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmHDMI_INFOFRAME_CONTROL0 + dig->afmt->offset);
-	/* enable audio info frames (frames won't be set until audio is enabled) */
-	tmp = REG_SET_FIELD(tmp, HDMI_INFOFRAME_CONTROL0, HDMI_AUDIO_INFO_SEND, 1);
-	/* required for audio info values to be updated */
-	tmp = REG_SET_FIELD(tmp, HDMI_INFOFRAME_CONTROL0, HDMI_AUDIO_INFO_CONT, 1);
-	WREG32(mmHDMI_INFOFRAME_CONTROL0 + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmAFMT_INFOFRAME_CONTROL0 + dig->afmt->offset);
-	/* required for audio info values to be updated */
-	tmp = REG_SET_FIELD(tmp, AFMT_INFOFRAME_CONTROL0, AFMT_AUDIO_INFO_UPDATE, 1);
-	WREG32(mmAFMT_INFOFRAME_CONTROL0 + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmHDMI_INFOFRAME_CONTROL1 + dig->afmt->offset);
-	/* anything other than 0 */
-	tmp = REG_SET_FIELD(tmp, HDMI_INFOFRAME_CONTROL1, HDMI_AUDIO_INFO_LINE, 2);
-	WREG32(mmHDMI_INFOFRAME_CONTROL1 + dig->afmt->offset, tmp);
-
-	WREG32(mmHDMI_GC + dig->afmt->offset, 0); /* unset HDMI_GC_AVMUTE */
-
-	tmp = RREG32(mmHDMI_AUDIO_PACKET_CONTROL + dig->afmt->offset);
-	/* set the default audio delay */
-	tmp = REG_SET_FIELD(tmp, HDMI_AUDIO_PACKET_CONTROL, HDMI_AUDIO_DELAY_EN, 1);
-	/* should be suffient for all audio modes and small enough for all hblanks */
-	tmp = REG_SET_FIELD(tmp, HDMI_AUDIO_PACKET_CONTROL, HDMI_AUDIO_PACKETS_PER_LINE, 3);
-	WREG32(mmHDMI_AUDIO_PACKET_CONTROL + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmAFMT_AUDIO_PACKET_CONTROL + dig->afmt->offset);
-	/* allow 60958 channel status fields to be updated */
-	tmp = REG_SET_FIELD(tmp, AFMT_AUDIO_PACKET_CONTROL, AFMT_60958_CS_UPDATE, 1);
-	WREG32(mmAFMT_AUDIO_PACKET_CONTROL + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmHDMI_ACR_PACKET_CONTROL + dig->afmt->offset);
-	if (bpc > 8)
-		/* clear SW CTS value */
-		tmp = REG_SET_FIELD(tmp, HDMI_ACR_PACKET_CONTROL, HDMI_ACR_SOURCE, 0);
-	else
-		/* select SW CTS value */
-		tmp = REG_SET_FIELD(tmp, HDMI_ACR_PACKET_CONTROL, HDMI_ACR_SOURCE, 1);
-	/* allow hw to sent ACR packets when required */
-	tmp = REG_SET_FIELD(tmp, HDMI_ACR_PACKET_CONTROL, HDMI_ACR_AUTO_SEND, 1);
-	WREG32(mmHDMI_ACR_PACKET_CONTROL + dig->afmt->offset, tmp);
-
-	dce_v11_0_afmt_update_ACR(encoder, mode->clock);
-
-	tmp = RREG32(mmAFMT_60958_0 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, AFMT_60958_0, AFMT_60958_CS_CHANNEL_NUMBER_L, 1);
-	WREG32(mmAFMT_60958_0 + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmAFMT_60958_1 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, AFMT_60958_1, AFMT_60958_CS_CHANNEL_NUMBER_R, 2);
-	WREG32(mmAFMT_60958_1 + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmAFMT_60958_2 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_2, 3);
-	tmp = REG_SET_FIELD(tmp, AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_3, 4);
-	tmp = REG_SET_FIELD(tmp, AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_4, 5);
-	tmp = REG_SET_FIELD(tmp, AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_5, 6);
-	tmp = REG_SET_FIELD(tmp, AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_6, 7);
-	tmp = REG_SET_FIELD(tmp, AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_7, 8);
-	WREG32(mmAFMT_60958_2 + dig->afmt->offset, tmp);
-
-	dce_v11_0_audio_write_speaker_allocation(encoder);
-
-	WREG32(mmAFMT_AUDIO_PACKET_CONTROL2 + dig->afmt->offset,
-	       (0xff << AFMT_AUDIO_PACKET_CONTROL2__AFMT_AUDIO_CHANNEL_ENABLE__SHIFT));
-
-	dce_v11_0_afmt_audio_select_pin(encoder);
-	dce_v11_0_audio_write_sad_regs(encoder);
-	dce_v11_0_audio_write_latency_fields(encoder, mode);
-
-	err = drm_hdmi_avi_infoframe_from_display_mode(&frame, connector, mode);
-	if (err < 0) {
-		DRM_ERROR("failed to setup AVI infoframe: %zd\n", err);
-		return;
-	}
-
-	err = hdmi_avi_infoframe_pack(&frame, buffer, sizeof(buffer));
-	if (err < 0) {
-		DRM_ERROR("failed to pack AVI infoframe: %zd\n", err);
-		return;
-	}
-
-	dce_v11_0_afmt_update_avi_infoframe(encoder, buffer, sizeof(buffer));
-
-	tmp = RREG32(mmHDMI_INFOFRAME_CONTROL0 + dig->afmt->offset);
-	/* enable AVI info frames */
-	tmp = REG_SET_FIELD(tmp, HDMI_INFOFRAME_CONTROL0, HDMI_AVI_INFO_SEND, 1);
-	/* required for audio info values to be updated */
-	tmp = REG_SET_FIELD(tmp, HDMI_INFOFRAME_CONTROL0, HDMI_AVI_INFO_CONT, 1);
-	WREG32(mmHDMI_INFOFRAME_CONTROL0 + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmHDMI_INFOFRAME_CONTROL1 + dig->afmt->offset);
-	tmp = REG_SET_FIELD(tmp, HDMI_INFOFRAME_CONTROL1, HDMI_AVI_INFO_LINE, 2);
-	WREG32(mmHDMI_INFOFRAME_CONTROL1 + dig->afmt->offset, tmp);
-
-	tmp = RREG32(mmAFMT_AUDIO_PACKET_CONTROL + dig->afmt->offset);
-	/* send audio packets */
-	tmp = REG_SET_FIELD(tmp, AFMT_AUDIO_PACKET_CONTROL, AFMT_AUDIO_SAMPLE_SEND, 1);
-	WREG32(mmAFMT_AUDIO_PACKET_CONTROL + dig->afmt->offset, tmp);
-
-	WREG32(mmAFMT_RAMP_CONTROL0 + dig->afmt->offset, 0x00FFFFFF);
-	WREG32(mmAFMT_RAMP_CONTROL1 + dig->afmt->offset, 0x007FFFFF);
-	WREG32(mmAFMT_RAMP_CONTROL2 + dig->afmt->offset, 0x00000001);
-	WREG32(mmAFMT_RAMP_CONTROL3 + dig->afmt->offset, 0x00000001);
-
-	/* enable audio after to setting up hw */
-	dce_v11_0_audio_enable(adev, dig->afmt->pin, true);
-}
-
-static void dce_v11_0_afmt_enable(struct drm_encoder *encoder, bool enable)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-
-	if (!dig || !dig->afmt)
-		return;
-
-	/* Silent, r600_hdmi_enable will raise WARN for us */
-	if (enable && dig->afmt->enabled)
-		return;
-	if (!enable && !dig->afmt->enabled)
-		return;
-
-	if (!enable && dig->afmt->pin) {
-		dce_v11_0_audio_enable(adev, dig->afmt->pin, false);
-		dig->afmt->pin = NULL;
-	}
-
-	dig->afmt->enabled = enable;
-
-	DRM_DEBUG("%sabling AFMT interface @ 0x%04X for encoder 0x%x\n",
-		  enable ? "En" : "Dis", dig->afmt->offset, amdgpu_encoder->encoder_id);
-}
-
-static int dce_v11_0_afmt_init(struct amdgpu_device *adev)
-{
-	int i;
-
-	for (i = 0; i < adev->mode_info.num_dig; i++)
-		adev->mode_info.afmt[i] = NULL;
-
-	/* DCE11 has audio blocks tied to DIG encoders */
-	for (i = 0; i < adev->mode_info.num_dig; i++) {
-		adev->mode_info.afmt[i] = kzalloc(sizeof(struct amdgpu_afmt), GFP_KERNEL);
-		if (adev->mode_info.afmt[i]) {
-			adev->mode_info.afmt[i]->offset = dig_offsets[i];
-			adev->mode_info.afmt[i]->id = i;
-		} else {
-			int j;
-			for (j = 0; j < i; j++) {
-				kfree(adev->mode_info.afmt[j]);
-				adev->mode_info.afmt[j] = NULL;
-			}
-			return -ENOMEM;
-		}
-	}
-	return 0;
-}
-
-static void dce_v11_0_afmt_fini(struct amdgpu_device *adev)
-{
-	int i;
-
-	for (i = 0; i < adev->mode_info.num_dig; i++) {
-		kfree(adev->mode_info.afmt[i]);
-		adev->mode_info.afmt[i] = NULL;
-	}
-}
-
-static const u32 vga_control_regs[6] =
-{
-	mmD1VGA_CONTROL,
-	mmD2VGA_CONTROL,
-	mmD3VGA_CONTROL,
-	mmD4VGA_CONTROL,
-	mmD5VGA_CONTROL,
-	mmD6VGA_CONTROL,
-};
-
-static void dce_v11_0_vga_enable(struct drm_crtc *crtc, bool enable)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	u32 vga_control;
-
-	vga_control = RREG32(vga_control_regs[amdgpu_crtc->crtc_id]) & ~1;
-	if (enable)
-		WREG32(vga_control_regs[amdgpu_crtc->crtc_id], vga_control | 1);
-	else
-		WREG32(vga_control_regs[amdgpu_crtc->crtc_id], vga_control);
-}
-
-static void dce_v11_0_grph_enable(struct drm_crtc *crtc, bool enable)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-
-	if (enable)
-		WREG32(mmGRPH_ENABLE + amdgpu_crtc->crtc_offset, 1);
-	else
-		WREG32(mmGRPH_ENABLE + amdgpu_crtc->crtc_offset, 0);
-}
-
-static int dce_v11_0_crtc_do_set_base(struct drm_crtc *crtc,
-				     struct drm_framebuffer *fb,
-				     int x, int y, int atomic)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct drm_framebuffer *target_fb;
-	struct drm_gem_object *obj;
-	struct amdgpu_bo *abo;
-	uint64_t fb_location, tiling_flags;
-	uint32_t fb_format, fb_pitch_pixels;
-	u32 fb_swap = REG_SET_FIELD(0, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP, ENDIAN_NONE);
-	u32 pipe_config;
-	u32 tmp, viewport_w, viewport_h;
-	int r;
-	bool bypass_lut = false;
-
-	/* no fb bound */
-	if (!atomic && !crtc->primary->fb) {
-		DRM_DEBUG_KMS("No FB bound\n");
-		return 0;
-	}
-
-	if (atomic)
-		target_fb = fb;
-	else
-		target_fb = crtc->primary->fb;
-
-	/* If atomic, assume fb object is pinned & idle & fenced and
-	 * just update base pointers
-	 */
-	obj = target_fb->obj[0];
-	abo = gem_to_amdgpu_bo(obj);
-	r = amdgpu_bo_reserve(abo, false);
-	if (unlikely(r != 0))
-		return r;
-
-	if (!atomic) {
-		r = amdgpu_bo_pin(abo, AMDGPU_GEM_DOMAIN_VRAM);
-		if (unlikely(r != 0)) {
-			amdgpu_bo_unreserve(abo);
-			return -EINVAL;
-		}
-	}
-	fb_location = amdgpu_bo_gpu_offset(abo);
-
-	amdgpu_bo_get_tiling_flags(abo, &tiling_flags);
-	amdgpu_bo_unreserve(abo);
-
-	pipe_config = AMDGPU_TILING_GET(tiling_flags, PIPE_CONFIG);
-
-	switch (target_fb->format->format) {
-	case DRM_FORMAT_C8:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 0);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 0);
-		break;
-	case DRM_FORMAT_XRGB4444:
-	case DRM_FORMAT_ARGB4444:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 1);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 2);
-#ifdef __BIG_ENDIAN
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP,
-					ENDIAN_8IN16);
-#endif
-		break;
-	case DRM_FORMAT_XRGB1555:
-	case DRM_FORMAT_ARGB1555:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 1);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 0);
-#ifdef __BIG_ENDIAN
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP,
-					ENDIAN_8IN16);
-#endif
-		break;
-	case DRM_FORMAT_BGRX5551:
-	case DRM_FORMAT_BGRA5551:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 1);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 5);
-#ifdef __BIG_ENDIAN
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP,
-					ENDIAN_8IN16);
-#endif
-		break;
-	case DRM_FORMAT_RGB565:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 1);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 1);
-#ifdef __BIG_ENDIAN
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP,
-					ENDIAN_8IN16);
-#endif
-		break;
-	case DRM_FORMAT_XRGB8888:
-	case DRM_FORMAT_ARGB8888:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 2);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 0);
-#ifdef __BIG_ENDIAN
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP,
-					ENDIAN_8IN32);
-#endif
-		break;
-	case DRM_FORMAT_XRGB2101010:
-	case DRM_FORMAT_ARGB2101010:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 2);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 1);
-#ifdef __BIG_ENDIAN
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP,
-					ENDIAN_8IN32);
-#endif
-		/* Greater 8 bpc fb needs to bypass hw-lut to retain precision */
-		bypass_lut = true;
-		break;
-	case DRM_FORMAT_BGRX1010102:
-	case DRM_FORMAT_BGRA1010102:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 2);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 4);
-#ifdef __BIG_ENDIAN
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP,
-					ENDIAN_8IN32);
-#endif
-		/* Greater 8 bpc fb needs to bypass hw-lut to retain precision */
-		bypass_lut = true;
-		break;
-	case DRM_FORMAT_XBGR8888:
-	case DRM_FORMAT_ABGR8888:
-		fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 2);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 0);
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_RED_CROSSBAR, 2);
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_BLUE_CROSSBAR, 2);
-#ifdef __BIG_ENDIAN
-		fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP,
-					ENDIAN_8IN32);
-#endif
-		break;
-	default:
-		DRM_ERROR("Unsupported screen format %p4cc\n",
-			  &target_fb->format->format);
-		return -EINVAL;
-	}
-
-	if (AMDGPU_TILING_GET(tiling_flags, ARRAY_MODE) == ARRAY_2D_TILED_THIN1) {
-		unsigned bankw, bankh, mtaspect, tile_split, num_banks;
-
-		bankw = AMDGPU_TILING_GET(tiling_flags, BANK_WIDTH);
-		bankh = AMDGPU_TILING_GET(tiling_flags, BANK_HEIGHT);
-		mtaspect = AMDGPU_TILING_GET(tiling_flags, MACRO_TILE_ASPECT);
-		tile_split = AMDGPU_TILING_GET(tiling_flags, TILE_SPLIT);
-		num_banks = AMDGPU_TILING_GET(tiling_flags, NUM_BANKS);
-
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_NUM_BANKS, num_banks);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_ARRAY_MODE,
-					  ARRAY_2D_TILED_THIN1);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_TILE_SPLIT,
-					  tile_split);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_BANK_WIDTH, bankw);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_BANK_HEIGHT, bankh);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_MACRO_TILE_ASPECT,
-					  mtaspect);
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_MICRO_TILE_MODE,
-					  ADDR_SURF_MICRO_TILING_DISPLAY);
-	} else if (AMDGPU_TILING_GET(tiling_flags, ARRAY_MODE) == ARRAY_1D_TILED_THIN1) {
-		fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_ARRAY_MODE,
-					  ARRAY_1D_TILED_THIN1);
-	}
-
-	fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_PIPE_CONFIG,
-				  pipe_config);
-
-	dce_v11_0_vga_enable(crtc, false);
-
-	/* Make sure surface address is updated at vertical blank rather than
-	 * horizontal blank
-	 */
-	tmp = RREG32(mmGRPH_FLIP_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, GRPH_FLIP_CONTROL,
-			    GRPH_SURFACE_UPDATE_H_RETRACE_EN, 0);
-	WREG32(mmGRPH_FLIP_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-
-	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS_HIGH + amdgpu_crtc->crtc_offset,
-	       upper_32_bits(fb_location));
-	WREG32(mmGRPH_SECONDARY_SURFACE_ADDRESS_HIGH + amdgpu_crtc->crtc_offset,
-	       upper_32_bits(fb_location));
-	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
-	       (u32)fb_location & GRPH_PRIMARY_SURFACE_ADDRESS__GRPH_PRIMARY_SURFACE_ADDRESS_MASK);
-	WREG32(mmGRPH_SECONDARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
-	       (u32) fb_location & GRPH_SECONDARY_SURFACE_ADDRESS__GRPH_SECONDARY_SURFACE_ADDRESS_MASK);
-	WREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset, fb_format);
-	WREG32(mmGRPH_SWAP_CNTL + amdgpu_crtc->crtc_offset, fb_swap);
-
-	/*
-	 * The LUT only has 256 slots for indexing by a 8 bpc fb. Bypass the LUT
-	 * for > 8 bpc scanout to avoid truncation of fb indices to 8 msb's, to
-	 * retain the full precision throughout the pipeline.
-	 */
-	tmp = RREG32(mmGRPH_LUT_10BIT_BYPASS + amdgpu_crtc->crtc_offset);
-	if (bypass_lut)
-		tmp = REG_SET_FIELD(tmp, GRPH_LUT_10BIT_BYPASS, GRPH_LUT_10BIT_BYPASS_EN, 1);
-	else
-		tmp = REG_SET_FIELD(tmp, GRPH_LUT_10BIT_BYPASS, GRPH_LUT_10BIT_BYPASS_EN, 0);
-	WREG32(mmGRPH_LUT_10BIT_BYPASS + amdgpu_crtc->crtc_offset, tmp);
-
-	if (bypass_lut)
-		DRM_DEBUG_KMS("Bypassing hardware LUT due to 10 bit fb scanout.\n");
-
-	WREG32(mmGRPH_SURFACE_OFFSET_X + amdgpu_crtc->crtc_offset, 0);
-	WREG32(mmGRPH_SURFACE_OFFSET_Y + amdgpu_crtc->crtc_offset, 0);
-	WREG32(mmGRPH_X_START + amdgpu_crtc->crtc_offset, 0);
-	WREG32(mmGRPH_Y_START + amdgpu_crtc->crtc_offset, 0);
-	WREG32(mmGRPH_X_END + amdgpu_crtc->crtc_offset, target_fb->width);
-	WREG32(mmGRPH_Y_END + amdgpu_crtc->crtc_offset, target_fb->height);
-
-	fb_pitch_pixels = target_fb->pitches[0] / target_fb->format->cpp[0];
-	WREG32(mmGRPH_PITCH + amdgpu_crtc->crtc_offset, fb_pitch_pixels);
-
-	dce_v11_0_grph_enable(crtc, true);
-
-	WREG32(mmLB_DESKTOP_HEIGHT + amdgpu_crtc->crtc_offset,
-	       target_fb->height);
-
-	x &= ~3;
-	y &= ~1;
-	WREG32(mmVIEWPORT_START + amdgpu_crtc->crtc_offset,
-	       (x << 16) | y);
-	viewport_w = crtc->mode.hdisplay;
-	viewport_h = (crtc->mode.vdisplay + 1) & ~1;
-	WREG32(mmVIEWPORT_SIZE + amdgpu_crtc->crtc_offset,
-	       (viewport_w << 16) | viewport_h);
-
-	/* set pageflip to happen anywhere in vblank interval */
-	WREG32(mmCRTC_MASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 0);
-
-	if (!atomic && fb && fb != crtc->primary->fb) {
-		abo = gem_to_amdgpu_bo(fb->obj[0]);
-		r = amdgpu_bo_reserve(abo, true);
-		if (unlikely(r != 0))
-			return r;
-		amdgpu_bo_unpin(abo);
-		amdgpu_bo_unreserve(abo);
-	}
-
-	/* Bytes per pixel may have changed */
-	dce_v11_0_bandwidth_update(adev);
-
-	return 0;
-}
-
-static void dce_v11_0_set_interleave(struct drm_crtc *crtc,
-				     struct drm_display_mode *mode)
-{
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	u32 tmp;
-
-	tmp = RREG32(mmLB_DATA_FORMAT + amdgpu_crtc->crtc_offset);
-	if (mode->flags & DRM_MODE_FLAG_INTERLACE)
-		tmp = REG_SET_FIELD(tmp, LB_DATA_FORMAT, INTERLEAVE_EN, 1);
-	else
-		tmp = REG_SET_FIELD(tmp, LB_DATA_FORMAT, INTERLEAVE_EN, 0);
-	WREG32(mmLB_DATA_FORMAT + amdgpu_crtc->crtc_offset, tmp);
-}
-
-static void dce_v11_0_crtc_load_lut(struct drm_crtc *crtc)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	u16 *r, *g, *b;
-	int i;
-	u32 tmp;
-
-	DRM_DEBUG_KMS("%d\n", amdgpu_crtc->crtc_id);
-
-	tmp = RREG32(mmINPUT_CSC_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, INPUT_CSC_CONTROL, INPUT_CSC_GRPH_MODE, 0);
-	WREG32(mmINPUT_CSC_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-
-	tmp = RREG32(mmPRESCALE_GRPH_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, PRESCALE_GRPH_CONTROL, GRPH_PRESCALE_BYPASS, 1);
-	WREG32(mmPRESCALE_GRPH_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-
-	tmp = RREG32(mmINPUT_GAMMA_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, INPUT_GAMMA_CONTROL, GRPH_INPUT_GAMMA_MODE, 0);
-	WREG32(mmINPUT_GAMMA_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-
-	WREG32(mmDC_LUT_CONTROL + amdgpu_crtc->crtc_offset, 0);
-
-	WREG32(mmDC_LUT_BLACK_OFFSET_BLUE + amdgpu_crtc->crtc_offset, 0);
-	WREG32(mmDC_LUT_BLACK_OFFSET_GREEN + amdgpu_crtc->crtc_offset, 0);
-	WREG32(mmDC_LUT_BLACK_OFFSET_RED + amdgpu_crtc->crtc_offset, 0);
-
-	WREG32(mmDC_LUT_WHITE_OFFSET_BLUE + amdgpu_crtc->crtc_offset, 0xffff);
-	WREG32(mmDC_LUT_WHITE_OFFSET_GREEN + amdgpu_crtc->crtc_offset, 0xffff);
-	WREG32(mmDC_LUT_WHITE_OFFSET_RED + amdgpu_crtc->crtc_offset, 0xffff);
-
-	WREG32(mmDC_LUT_RW_MODE + amdgpu_crtc->crtc_offset, 0);
-	WREG32(mmDC_LUT_WRITE_EN_MASK + amdgpu_crtc->crtc_offset, 0x00000007);
-
-	WREG32(mmDC_LUT_RW_INDEX + amdgpu_crtc->crtc_offset, 0);
-	r = crtc->gamma_store;
-	g = r + crtc->gamma_size;
-	b = g + crtc->gamma_size;
-	for (i = 0; i < 256; i++) {
-		WREG32(mmDC_LUT_30_COLOR + amdgpu_crtc->crtc_offset,
-		       ((*r++ & 0xffc0) << 14) |
-		       ((*g++ & 0xffc0) << 4) |
-		       (*b++ >> 6));
-	}
-
-	tmp = RREG32(mmDEGAMMA_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, DEGAMMA_CONTROL, GRPH_DEGAMMA_MODE, 0);
-	tmp = REG_SET_FIELD(tmp, DEGAMMA_CONTROL, CURSOR_DEGAMMA_MODE, 0);
-	tmp = REG_SET_FIELD(tmp, DEGAMMA_CONTROL, CURSOR2_DEGAMMA_MODE, 0);
-	WREG32(mmDEGAMMA_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-
-	tmp = RREG32(mmGAMUT_REMAP_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, GAMUT_REMAP_CONTROL, GRPH_GAMUT_REMAP_MODE, 0);
-	WREG32(mmGAMUT_REMAP_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-
-	tmp = RREG32(mmREGAMMA_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, REGAMMA_CONTROL, GRPH_REGAMMA_MODE, 0);
-	WREG32(mmREGAMMA_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-
-	tmp = RREG32(mmOUTPUT_CSC_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, OUTPUT_CSC_CONTROL, OUTPUT_CSC_GRPH_MODE, 0);
-	WREG32(mmOUTPUT_CSC_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-
-	/* XXX match this to the depth of the crtc fmt block, move to modeset? */
-	WREG32(mmDENORM_CONTROL + amdgpu_crtc->crtc_offset, 0);
-	/* XXX this only needs to be programmed once per crtc at startup,
-	 * not sure where the best place for it is
-	 */
-	tmp = RREG32(mmALPHA_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, ALPHA_CONTROL, CURSOR_ALPHA_BLND_ENA, 1);
-	WREG32(mmALPHA_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-}
-
-static int dce_v11_0_pick_dig_encoder(struct drm_encoder *encoder)
-{
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-
-	switch (amdgpu_encoder->encoder_id) {
-	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
-		if (dig->linkb)
-			return 1;
-		else
-			return 0;
-	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
-		if (dig->linkb)
-			return 3;
-		else
-			return 2;
-	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
-		if (dig->linkb)
-			return 5;
-		else
-			return 4;
-	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3:
-		return 6;
-	default:
-		DRM_ERROR("invalid encoder_id: 0x%x\n", amdgpu_encoder->encoder_id);
-		return 0;
-	}
-}
-
-/**
- * dce_v11_0_pick_pll - Allocate a PPLL for use by the crtc.
- *
- * @crtc: drm crtc
- *
- * Returns the PPLL (Pixel PLL) to be used by the crtc.  For DP monitors
- * a single PPLL can be used for all DP crtcs/encoders.  For non-DP
- * monitors a dedicated PPLL must be used.  If a particular board has
- * an external DP PLL, return ATOM_PPLL_INVALID to skip PLL programming
- * as there is no need to program the PLL itself.  If we are not able to
- * allocate a PLL, return ATOM_PPLL_INVALID to skip PLL programming to
- * avoid messing up an existing monitor.
- *
- * Asic specific PLL information
- *
- * DCE 10.x
- * Tonga
- * - PPLL1, PPLL2 are available for all UNIPHY (both DP and non-DP)
- * CI
- * - PPLL0, PPLL1, PPLL2 are available for all UNIPHY (both DP and non-DP) and DAC
- *
- */
-static u32 dce_v11_0_pick_pll(struct drm_crtc *crtc)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	u32 pll_in_use;
-	int pll;
-
-	if ((adev->asic_type == CHIP_POLARIS10) ||
-	    (adev->asic_type == CHIP_POLARIS11) ||
-	    (adev->asic_type == CHIP_POLARIS12) ||
-	    (adev->asic_type == CHIP_VEGAM)) {
-		struct amdgpu_encoder *amdgpu_encoder =
-			to_amdgpu_encoder(amdgpu_crtc->encoder);
-		struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-
-		if (ENCODER_MODE_IS_DP(amdgpu_atombios_encoder_get_encoder_mode(amdgpu_crtc->encoder)))
-			return ATOM_DP_DTO;
-
-		switch (amdgpu_encoder->encoder_id) {
-		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
-			if (dig->linkb)
-				return ATOM_COMBOPHY_PLL1;
-			else
-				return ATOM_COMBOPHY_PLL0;
-		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
-			if (dig->linkb)
-				return ATOM_COMBOPHY_PLL3;
-			else
-				return ATOM_COMBOPHY_PLL2;
-		case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
-			if (dig->linkb)
-				return ATOM_COMBOPHY_PLL5;
-			else
-				return ATOM_COMBOPHY_PLL4;
-		default:
-			DRM_ERROR("invalid encoder_id: 0x%x\n", amdgpu_encoder->encoder_id);
-			return ATOM_PPLL_INVALID;
-		}
-	}
-
-	if (ENCODER_MODE_IS_DP(amdgpu_atombios_encoder_get_encoder_mode(amdgpu_crtc->encoder))) {
-		if (adev->clock.dp_extclk)
-			/* skip PPLL programming if using ext clock */
-			return ATOM_PPLL_INVALID;
-		else {
-			/* use the same PPLL for all DP monitors */
-			pll = amdgpu_pll_get_shared_dp_ppll(crtc);
-			if (pll != ATOM_PPLL_INVALID)
-				return pll;
-		}
-	} else {
-		/* use the same PPLL for all monitors with the same clock */
-		pll = amdgpu_pll_get_shared_nondp_ppll(crtc);
-		if (pll != ATOM_PPLL_INVALID)
-			return pll;
-	}
-
-	/* XXX need to determine what plls are available on each DCE11 part */
-	pll_in_use = amdgpu_pll_get_use_mask(crtc);
-	if (adev->flags & AMD_IS_APU) {
-		if (!(pll_in_use & (1 << ATOM_PPLL1)))
-			return ATOM_PPLL1;
-		if (!(pll_in_use & (1 << ATOM_PPLL0)))
-			return ATOM_PPLL0;
-		DRM_ERROR("unable to allocate a PPLL\n");
-		return ATOM_PPLL_INVALID;
-	} else {
-		if (!(pll_in_use & (1 << ATOM_PPLL2)))
-			return ATOM_PPLL2;
-		if (!(pll_in_use & (1 << ATOM_PPLL1)))
-			return ATOM_PPLL1;
-		if (!(pll_in_use & (1 << ATOM_PPLL0)))
-			return ATOM_PPLL0;
-		DRM_ERROR("unable to allocate a PPLL\n");
-		return ATOM_PPLL_INVALID;
-	}
-	return ATOM_PPLL_INVALID;
-}
-
-static void dce_v11_0_lock_cursor(struct drm_crtc *crtc, bool lock)
-{
-	struct amdgpu_device *adev = drm_to_adev(crtc->dev);
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	uint32_t cur_lock;
-
-	cur_lock = RREG32(mmCUR_UPDATE + amdgpu_crtc->crtc_offset);
-	if (lock)
-		cur_lock = REG_SET_FIELD(cur_lock, CUR_UPDATE, CURSOR_UPDATE_LOCK, 1);
-	else
-		cur_lock = REG_SET_FIELD(cur_lock, CUR_UPDATE, CURSOR_UPDATE_LOCK, 0);
-	WREG32(mmCUR_UPDATE + amdgpu_crtc->crtc_offset, cur_lock);
-}
-
-static void dce_v11_0_hide_cursor(struct drm_crtc *crtc)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct amdgpu_device *adev = drm_to_adev(crtc->dev);
-	u32 tmp;
-
-	tmp = RREG32(mmCUR_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, CUR_CONTROL, CURSOR_EN, 0);
-	WREG32(mmCUR_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-}
-
-static void dce_v11_0_show_cursor(struct drm_crtc *crtc)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct amdgpu_device *adev = drm_to_adev(crtc->dev);
-	u32 tmp;
-
-	WREG32(mmCUR_SURFACE_ADDRESS_HIGH + amdgpu_crtc->crtc_offset,
-	       upper_32_bits(amdgpu_crtc->cursor_addr));
-	WREG32(mmCUR_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
-	       lower_32_bits(amdgpu_crtc->cursor_addr));
-
-	tmp = RREG32(mmCUR_CONTROL + amdgpu_crtc->crtc_offset);
-	tmp = REG_SET_FIELD(tmp, CUR_CONTROL, CURSOR_EN, 1);
-	tmp = REG_SET_FIELD(tmp, CUR_CONTROL, CURSOR_MODE, 2);
-	WREG32(mmCUR_CONTROL + amdgpu_crtc->crtc_offset, tmp);
-}
-
-static int dce_v11_0_cursor_move_locked(struct drm_crtc *crtc,
-					int x, int y)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct amdgpu_device *adev = drm_to_adev(crtc->dev);
-	int xorigin = 0, yorigin = 0;
-
-	amdgpu_crtc->cursor_x = x;
-	amdgpu_crtc->cursor_y = y;
-
-	/* avivo cursor are offset into the total surface */
-	x += crtc->x;
-	y += crtc->y;
-	DRM_DEBUG("x %d y %d c->x %d c->y %d\n", x, y, crtc->x, crtc->y);
-
-	if (x < 0) {
-		xorigin = min(-x, amdgpu_crtc->max_cursor_width - 1);
-		x = 0;
-	}
-	if (y < 0) {
-		yorigin = min(-y, amdgpu_crtc->max_cursor_height - 1);
-		y = 0;
-	}
-
-	WREG32(mmCUR_POSITION + amdgpu_crtc->crtc_offset, (x << 16) | y);
-	WREG32(mmCUR_HOT_SPOT + amdgpu_crtc->crtc_offset, (xorigin << 16) | yorigin);
-	WREG32(mmCUR_SIZE + amdgpu_crtc->crtc_offset,
-	       ((amdgpu_crtc->cursor_width - 1) << 16) | (amdgpu_crtc->cursor_height - 1));
-
-	return 0;
-}
-
-static int dce_v11_0_crtc_cursor_move(struct drm_crtc *crtc,
-				      int x, int y)
-{
-	int ret;
-
-	dce_v11_0_lock_cursor(crtc, true);
-	ret = dce_v11_0_cursor_move_locked(crtc, x, y);
-	dce_v11_0_lock_cursor(crtc, false);
-
-	return ret;
-}
-
-static int dce_v11_0_crtc_cursor_set2(struct drm_crtc *crtc,
-				      struct drm_file *file_priv,
-				      uint32_t handle,
-				      uint32_t width,
-				      uint32_t height,
-				      int32_t hot_x,
-				      int32_t hot_y)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_gem_object *obj;
-	struct amdgpu_bo *aobj;
-	int ret;
-
-	if (!handle) {
-		/* turn off cursor */
-		dce_v11_0_hide_cursor(crtc);
-		obj = NULL;
-		goto unpin;
-	}
-
-	if ((width > amdgpu_crtc->max_cursor_width) ||
-	    (height > amdgpu_crtc->max_cursor_height)) {
-		DRM_ERROR("bad cursor width or height %d x %d\n", width, height);
-		return -EINVAL;
-	}
-
-	obj = drm_gem_object_lookup(file_priv, handle);
-	if (!obj) {
-		DRM_ERROR("Cannot find cursor object %x for crtc %d\n", handle, amdgpu_crtc->crtc_id);
-		return -ENOENT;
-	}
-
-	aobj = gem_to_amdgpu_bo(obj);
-	ret = amdgpu_bo_reserve(aobj, false);
-	if (ret != 0) {
-		drm_gem_object_put(obj);
-		return ret;
-	}
-
-	ret = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM);
-	amdgpu_bo_unreserve(aobj);
-	if (ret) {
-		DRM_ERROR("Failed to pin new cursor BO (%d)\n", ret);
-		drm_gem_object_put(obj);
-		return ret;
-	}
-	amdgpu_crtc->cursor_addr = amdgpu_bo_gpu_offset(aobj);
-
-	dce_v11_0_lock_cursor(crtc, true);
-
-	if (width != amdgpu_crtc->cursor_width ||
-	    height != amdgpu_crtc->cursor_height ||
-	    hot_x != amdgpu_crtc->cursor_hot_x ||
-	    hot_y != amdgpu_crtc->cursor_hot_y) {
-		int x, y;
-
-		x = amdgpu_crtc->cursor_x + amdgpu_crtc->cursor_hot_x - hot_x;
-		y = amdgpu_crtc->cursor_y + amdgpu_crtc->cursor_hot_y - hot_y;
-
-		dce_v11_0_cursor_move_locked(crtc, x, y);
-
-		amdgpu_crtc->cursor_width = width;
-		amdgpu_crtc->cursor_height = height;
-		amdgpu_crtc->cursor_hot_x = hot_x;
-		amdgpu_crtc->cursor_hot_y = hot_y;
-	}
-
-	dce_v11_0_show_cursor(crtc);
-	dce_v11_0_lock_cursor(crtc, false);
-
-unpin:
-	if (amdgpu_crtc->cursor_bo) {
-		struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
-		ret = amdgpu_bo_reserve(aobj, true);
-		if (likely(ret == 0)) {
-			amdgpu_bo_unpin(aobj);
-			amdgpu_bo_unreserve(aobj);
-		}
-		drm_gem_object_put(amdgpu_crtc->cursor_bo);
-	}
-
-	amdgpu_crtc->cursor_bo = obj;
-	return 0;
-}
-
-static void dce_v11_0_cursor_reset(struct drm_crtc *crtc)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-
-	if (amdgpu_crtc->cursor_bo) {
-		dce_v11_0_lock_cursor(crtc, true);
-
-		dce_v11_0_cursor_move_locked(crtc, amdgpu_crtc->cursor_x,
-					     amdgpu_crtc->cursor_y);
-
-		dce_v11_0_show_cursor(crtc);
-
-		dce_v11_0_lock_cursor(crtc, false);
-	}
-}
-
-static int dce_v11_0_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green,
-				    u16 *blue, uint32_t size,
-				    struct drm_modeset_acquire_ctx *ctx)
-{
-	dce_v11_0_crtc_load_lut(crtc);
-
-	return 0;
-}
-
-static void dce_v11_0_crtc_destroy(struct drm_crtc *crtc)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-
-	drm_crtc_cleanup(crtc);
-	kfree(amdgpu_crtc);
-}
-
-static const struct drm_crtc_funcs dce_v11_0_crtc_funcs = {
-	.cursor_set2 = dce_v11_0_crtc_cursor_set2,
-	.cursor_move = dce_v11_0_crtc_cursor_move,
-	.gamma_set = dce_v11_0_crtc_gamma_set,
-	.set_config = amdgpu_display_crtc_set_config,
-	.destroy = dce_v11_0_crtc_destroy,
-	.page_flip_target = amdgpu_display_crtc_page_flip_target,
-	.get_vblank_counter = amdgpu_get_vblank_counter_kms,
-	.enable_vblank = amdgpu_enable_vblank_kms,
-	.disable_vblank = amdgpu_disable_vblank_kms,
-	.get_vblank_timestamp = drm_crtc_vblank_helper_get_vblank_timestamp,
-};
-
-static void dce_v11_0_crtc_dpms(struct drm_crtc *crtc, int mode)
-{
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	unsigned type;
-
-	switch (mode) {
-	case DRM_MODE_DPMS_ON:
-		amdgpu_crtc->enabled = true;
-		amdgpu_atombios_crtc_enable(crtc, ATOM_ENABLE);
-		dce_v11_0_vga_enable(crtc, true);
-		amdgpu_atombios_crtc_blank(crtc, ATOM_DISABLE);
-		dce_v11_0_vga_enable(crtc, false);
-		/* Make sure VBLANK and PFLIP interrupts are still enabled */
-		type = amdgpu_display_crtc_idx_to_irq_type(adev,
-						amdgpu_crtc->crtc_id);
-		amdgpu_irq_update(adev, &adev->crtc_irq, type);
-		amdgpu_irq_update(adev, &adev->pageflip_irq, type);
-		drm_crtc_vblank_on(crtc);
-		dce_v11_0_crtc_load_lut(crtc);
-		break;
-	case DRM_MODE_DPMS_STANDBY:
-	case DRM_MODE_DPMS_SUSPEND:
-	case DRM_MODE_DPMS_OFF:
-		drm_crtc_vblank_off(crtc);
-		if (amdgpu_crtc->enabled) {
-			dce_v11_0_vga_enable(crtc, true);
-			amdgpu_atombios_crtc_blank(crtc, ATOM_ENABLE);
-			dce_v11_0_vga_enable(crtc, false);
-		}
-		amdgpu_atombios_crtc_enable(crtc, ATOM_DISABLE);
-		amdgpu_crtc->enabled = false;
-		break;
-	}
-	/* adjust pm to dpms */
-	amdgpu_dpm_compute_clocks(adev);
-}
-
-static void dce_v11_0_crtc_prepare(struct drm_crtc *crtc)
-{
-	/* disable crtc pair power gating before programming */
-	amdgpu_atombios_crtc_powergate(crtc, ATOM_DISABLE);
-	amdgpu_atombios_crtc_lock(crtc, ATOM_ENABLE);
-	dce_v11_0_crtc_dpms(crtc, DRM_MODE_DPMS_OFF);
-}
-
-static void dce_v11_0_crtc_commit(struct drm_crtc *crtc)
-{
-	dce_v11_0_crtc_dpms(crtc, DRM_MODE_DPMS_ON);
-	amdgpu_atombios_crtc_lock(crtc, ATOM_DISABLE);
-}
-
-static void dce_v11_0_crtc_disable(struct drm_crtc *crtc)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-	struct amdgpu_atom_ss ss;
-	int i;
-
-	dce_v11_0_crtc_dpms(crtc, DRM_MODE_DPMS_OFF);
-	if (crtc->primary->fb) {
-		int r;
-		struct amdgpu_bo *abo;
-
-		abo = gem_to_amdgpu_bo(crtc->primary->fb->obj[0]);
-		r = amdgpu_bo_reserve(abo, true);
-		if (unlikely(r))
-			DRM_ERROR("failed to reserve abo before unpin\n");
-		else {
-			amdgpu_bo_unpin(abo);
-			amdgpu_bo_unreserve(abo);
-		}
-	}
-	/* disable the GRPH */
-	dce_v11_0_grph_enable(crtc, false);
-
-	amdgpu_atombios_crtc_powergate(crtc, ATOM_ENABLE);
-
-	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		if (adev->mode_info.crtcs[i] &&
-		    adev->mode_info.crtcs[i]->enabled &&
-		    i != amdgpu_crtc->crtc_id &&
-		    amdgpu_crtc->pll_id == adev->mode_info.crtcs[i]->pll_id) {
-			/* one other crtc is using this pll don't turn
-			 * off the pll
-			 */
-			goto done;
-		}
-	}
-
-	switch (amdgpu_crtc->pll_id) {
-	case ATOM_PPLL0:
-	case ATOM_PPLL1:
-	case ATOM_PPLL2:
-		/* disable the ppll */
-		amdgpu_atombios_crtc_program_pll(crtc, amdgpu_crtc->crtc_id, amdgpu_crtc->pll_id,
-						 0, 0, ATOM_DISABLE, 0, 0, 0, 0, 0, false, &ss);
-		break;
-	case ATOM_COMBOPHY_PLL0:
-	case ATOM_COMBOPHY_PLL1:
-	case ATOM_COMBOPHY_PLL2:
-	case ATOM_COMBOPHY_PLL3:
-	case ATOM_COMBOPHY_PLL4:
-	case ATOM_COMBOPHY_PLL5:
-		/* disable the ppll */
-		amdgpu_atombios_crtc_program_pll(crtc, ATOM_CRTC_INVALID, amdgpu_crtc->pll_id,
-						 0, 0, ATOM_DISABLE, 0, 0, 0, 0, 0, false, &ss);
-		break;
-	default:
-		break;
-	}
-done:
-	amdgpu_crtc->pll_id = ATOM_PPLL_INVALID;
-	amdgpu_crtc->adjusted_clock = 0;
-	amdgpu_crtc->encoder = NULL;
-	amdgpu_crtc->connector = NULL;
-}
-
-static int dce_v11_0_crtc_mode_set(struct drm_crtc *crtc,
-				  struct drm_display_mode *mode,
-				  struct drm_display_mode *adjusted_mode,
-				  int x, int y, struct drm_framebuffer *old_fb)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-
-	if (!amdgpu_crtc->adjusted_clock)
-		return -EINVAL;
-
-	if ((adev->asic_type == CHIP_POLARIS10) ||
-	    (adev->asic_type == CHIP_POLARIS11) ||
-	    (adev->asic_type == CHIP_POLARIS12) ||
-	    (adev->asic_type == CHIP_VEGAM)) {
-		struct amdgpu_encoder *amdgpu_encoder =
-			to_amdgpu_encoder(amdgpu_crtc->encoder);
-		int encoder_mode =
-			amdgpu_atombios_encoder_get_encoder_mode(amdgpu_crtc->encoder);
-
-		/* SetPixelClock calculates the plls and ss values now */
-		amdgpu_atombios_crtc_program_pll(crtc, amdgpu_crtc->crtc_id,
-						 amdgpu_crtc->pll_id,
-						 encoder_mode, amdgpu_encoder->encoder_id,
-						 adjusted_mode->clock, 0, 0, 0, 0,
-						 amdgpu_crtc->bpc, amdgpu_crtc->ss_enabled, &amdgpu_crtc->ss);
-	} else {
-		amdgpu_atombios_crtc_set_pll(crtc, adjusted_mode);
-	}
-	amdgpu_atombios_crtc_set_dtd_timing(crtc, adjusted_mode);
-	dce_v11_0_crtc_do_set_base(crtc, old_fb, x, y, 0);
-	amdgpu_atombios_crtc_overscan_setup(crtc, mode, adjusted_mode);
-	amdgpu_atombios_crtc_scaler_setup(crtc);
-	dce_v11_0_cursor_reset(crtc);
-	/* update the hw version fpr dpm */
-	amdgpu_crtc->hw_mode = *adjusted_mode;
-
-	return 0;
-}
-
-static bool dce_v11_0_crtc_mode_fixup(struct drm_crtc *crtc,
-				     const struct drm_display_mode *mode,
-				     struct drm_display_mode *adjusted_mode)
-{
-	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
-	struct drm_device *dev = crtc->dev;
-	struct drm_encoder *encoder;
-
-	/* assign the encoder to the amdgpu crtc to avoid repeated lookups later */
-	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
-		if (encoder->crtc == crtc) {
-			amdgpu_crtc->encoder = encoder;
-			amdgpu_crtc->connector = amdgpu_get_connector_for_encoder(encoder);
-			break;
-		}
-	}
-	if ((amdgpu_crtc->encoder == NULL) || (amdgpu_crtc->connector == NULL)) {
-		amdgpu_crtc->encoder = NULL;
-		amdgpu_crtc->connector = NULL;
-		return false;
-	}
-	if (!amdgpu_display_crtc_scaling_mode_fixup(crtc, mode, adjusted_mode))
-		return false;
-	if (amdgpu_atombios_crtc_prepare_pll(crtc, adjusted_mode))
-		return false;
-	/* pick pll */
-	amdgpu_crtc->pll_id = dce_v11_0_pick_pll(crtc);
-	/* if we can't get a PPLL for a non-DP encoder, fail */
-	if ((amdgpu_crtc->pll_id == ATOM_PPLL_INVALID) &&
-	    !ENCODER_MODE_IS_DP(amdgpu_atombios_encoder_get_encoder_mode(amdgpu_crtc->encoder)))
-		return false;
-
-	return true;
-}
-
-static int dce_v11_0_crtc_set_base(struct drm_crtc *crtc, int x, int y,
-				  struct drm_framebuffer *old_fb)
-{
-	return dce_v11_0_crtc_do_set_base(crtc, old_fb, x, y, 0);
-}
-
-static int dce_v11_0_crtc_set_base_atomic(struct drm_crtc *crtc,
-					 struct drm_framebuffer *fb,
-					 int x, int y, enum mode_set_atomic state)
-{
-	return dce_v11_0_crtc_do_set_base(crtc, fb, x, y, 1);
-}
-
-static const struct drm_crtc_helper_funcs dce_v11_0_crtc_helper_funcs = {
-	.dpms = dce_v11_0_crtc_dpms,
-	.mode_fixup = dce_v11_0_crtc_mode_fixup,
-	.mode_set = dce_v11_0_crtc_mode_set,
-	.mode_set_base = dce_v11_0_crtc_set_base,
-	.mode_set_base_atomic = dce_v11_0_crtc_set_base_atomic,
-	.prepare = dce_v11_0_crtc_prepare,
-	.commit = dce_v11_0_crtc_commit,
-	.disable = dce_v11_0_crtc_disable,
-	.get_scanout_position = amdgpu_crtc_get_scanout_position,
-};
-
-static int dce_v11_0_crtc_init(struct amdgpu_device *adev, int index)
-{
-	struct amdgpu_crtc *amdgpu_crtc;
-
-	amdgpu_crtc = kzalloc(sizeof(struct amdgpu_crtc) +
-			      (AMDGPUFB_CONN_LIMIT * sizeof(struct drm_connector *)), GFP_KERNEL);
-	if (amdgpu_crtc == NULL)
-		return -ENOMEM;
-
-	drm_crtc_init(adev_to_drm(adev), &amdgpu_crtc->base, &dce_v11_0_crtc_funcs);
-
-	drm_mode_crtc_set_gamma_size(&amdgpu_crtc->base, 256);
-	amdgpu_crtc->crtc_id = index;
-	adev->mode_info.crtcs[index] = amdgpu_crtc;
-
-	amdgpu_crtc->max_cursor_width = 128;
-	amdgpu_crtc->max_cursor_height = 128;
-	adev_to_drm(adev)->mode_config.cursor_width = amdgpu_crtc->max_cursor_width;
-	adev_to_drm(adev)->mode_config.cursor_height = amdgpu_crtc->max_cursor_height;
-
-	switch (amdgpu_crtc->crtc_id) {
-	case 0:
-	default:
-		amdgpu_crtc->crtc_offset = CRTC0_REGISTER_OFFSET;
-		break;
-	case 1:
-		amdgpu_crtc->crtc_offset = CRTC1_REGISTER_OFFSET;
-		break;
-	case 2:
-		amdgpu_crtc->crtc_offset = CRTC2_REGISTER_OFFSET;
-		break;
-	case 3:
-		amdgpu_crtc->crtc_offset = CRTC3_REGISTER_OFFSET;
-		break;
-	case 4:
-		amdgpu_crtc->crtc_offset = CRTC4_REGISTER_OFFSET;
-		break;
-	case 5:
-		amdgpu_crtc->crtc_offset = CRTC5_REGISTER_OFFSET;
-		break;
-	}
-
-	amdgpu_crtc->pll_id = ATOM_PPLL_INVALID;
-	amdgpu_crtc->adjusted_clock = 0;
-	amdgpu_crtc->encoder = NULL;
-	amdgpu_crtc->connector = NULL;
-	drm_crtc_helper_add(&amdgpu_crtc->base, &dce_v11_0_crtc_helper_funcs);
-
-	return 0;
-}
-
-static int dce_v11_0_early_init(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	adev->audio_endpt_rreg = &dce_v11_0_audio_endpt_rreg;
-	adev->audio_endpt_wreg = &dce_v11_0_audio_endpt_wreg;
-
-	dce_v11_0_set_display_funcs(adev);
-
-	adev->mode_info.num_crtc = dce_v11_0_get_num_crtc(adev);
-
-	switch (adev->asic_type) {
-	case CHIP_CARRIZO:
-		adev->mode_info.num_hpd = 6;
-		adev->mode_info.num_dig = 9;
-		break;
-	case CHIP_STONEY:
-		adev->mode_info.num_hpd = 6;
-		adev->mode_info.num_dig = 9;
-		break;
-	case CHIP_POLARIS10:
-	case CHIP_VEGAM:
-		adev->mode_info.num_hpd = 6;
-		adev->mode_info.num_dig = 6;
-		break;
-	case CHIP_POLARIS11:
-	case CHIP_POLARIS12:
-		adev->mode_info.num_hpd = 5;
-		adev->mode_info.num_dig = 5;
-		break;
-	default:
-		/* FIXME: not supported yet */
-		return -EINVAL;
-	}
-
-	dce_v11_0_set_irq_funcs(adev);
-
-	return 0;
-}
-
-static int dce_v11_0_sw_init(void *handle)
-{
-	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + 1, &adev->crtc_irq);
-		if (r)
-			return r;
-	}
-
-	for (i = VISLANDS30_IV_SRCID_D1_GRPH_PFLIP; i < 20; i += 2) {
-		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i, &adev->pageflip_irq);
-		if (r)
-			return r;
-	}
-
-	/* HPD hotplug */
-	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_HOTPLUG_DETECT_A, &adev->hpd_irq);
-	if (r)
-		return r;
-
-	adev_to_drm(adev)->mode_config.funcs = &amdgpu_mode_funcs;
-
-	adev_to_drm(adev)->mode_config.async_page_flip = true;
-
-	adev_to_drm(adev)->mode_config.max_width = 16384;
-	adev_to_drm(adev)->mode_config.max_height = 16384;
-
-	adev_to_drm(adev)->mode_config.preferred_depth = 24;
-	adev_to_drm(adev)->mode_config.prefer_shadow = 1;
-
-	adev_to_drm(adev)->mode_config.fb_modifiers_not_supported = true;
-
-	r = amdgpu_display_modeset_create_props(adev);
-	if (r)
-		return r;
-
-	adev_to_drm(adev)->mode_config.max_width = 16384;
-	adev_to_drm(adev)->mode_config.max_height = 16384;
-
-
-	/* allocate crtcs */
-	for (i = 0; i < adev->mode_info.num_crtc; i++) {
-		r = dce_v11_0_crtc_init(adev, i);
-		if (r)
-			return r;
-	}
-
-	if (amdgpu_atombios_get_connector_info_from_object_table(adev))
-		amdgpu_display_print_display_setup(adev_to_drm(adev));
-	else
-		return -EINVAL;
-
-	/* setup afmt */
-	r = dce_v11_0_afmt_init(adev);
-	if (r)
-		return r;
-
-	r = dce_v11_0_audio_init(adev);
-	if (r)
-		return r;
-
-	/* Disable vblank IRQs aggressively for power-saving */
-	/* XXX: can this be enabled for DC? */
-	adev_to_drm(adev)->vblank_disable_immediate = true;
-
-	r = drm_vblank_init(adev_to_drm(adev), adev->mode_info.num_crtc);
-	if (r)
-		return r;
-
-	INIT_DELAYED_WORK(&adev->hotplug_work,
-		  amdgpu_display_hotplug_work_func);
-
-	drm_kms_helper_poll_init(adev_to_drm(adev));
-
-	adev->mode_info.mode_config_initialized = true;
-	return 0;
-}
-
-static int dce_v11_0_sw_fini(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	kfree(adev->mode_info.bios_hardcoded_edid);
-
-	drm_kms_helper_poll_fini(adev_to_drm(adev));
-
-	dce_v11_0_audio_fini(adev);
-
-	dce_v11_0_afmt_fini(adev);
-
-	drm_mode_config_cleanup(adev_to_drm(adev));
-	adev->mode_info.mode_config_initialized = false;
-
-	return 0;
-}
-
-static int dce_v11_0_hw_init(void *handle)
-{
-	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	dce_v11_0_init_golden_registers(adev);
-
-	/* disable vga render */
-	dce_v11_0_set_vga_render_state(adev, false);
-	/* init dig PHYs, disp eng pll */
-	amdgpu_atombios_crtc_powergate_init(adev);
-	amdgpu_atombios_encoder_init_dig(adev);
-	if ((adev->asic_type == CHIP_POLARIS10) ||
-	    (adev->asic_type == CHIP_POLARIS11) ||
-	    (adev->asic_type == CHIP_POLARIS12) ||
-	    (adev->asic_type == CHIP_VEGAM)) {
-		amdgpu_atombios_crtc_set_dce_clock(adev, adev->clock.default_dispclk,
-						   DCE_CLOCK_TYPE_DISPCLK, ATOM_GCK_DFS);
-		amdgpu_atombios_crtc_set_dce_clock(adev, 0,
-						   DCE_CLOCK_TYPE_DPREFCLK, ATOM_GCK_DFS);
-	} else {
-		amdgpu_atombios_crtc_set_disp_eng_pll(adev, adev->clock.default_dispclk);
-	}
-
-	/* initialize hpd */
-	dce_v11_0_hpd_init(adev);
-
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++) {
-		dce_v11_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
-	}
-
-	dce_v11_0_pageflip_interrupt_init(adev);
-
-	return 0;
-}
-
-static int dce_v11_0_hw_fini(void *handle)
-{
-	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	dce_v11_0_hpd_fini(adev);
-
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++) {
-		dce_v11_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
-	}
-
-	dce_v11_0_pageflip_interrupt_fini(adev);
-
-	flush_delayed_work(&adev->hotplug_work);
-
-	return 0;
-}
-
-static int dce_v11_0_suspend(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int r;
-
-	r = amdgpu_display_suspend_helper(adev);
-	if (r)
-		return r;
-
-	adev->mode_info.bl_level =
-		amdgpu_atombios_encoder_get_backlight_level_from_reg(adev);
-
-	return dce_v11_0_hw_fini(handle);
-}
-
-static int dce_v11_0_resume(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int ret;
-
-	amdgpu_atombios_encoder_set_backlight_level_to_reg(adev,
-							   adev->mode_info.bl_level);
-
-	ret = dce_v11_0_hw_init(handle);
-
-	/* turn on the BL */
-	if (adev->mode_info.bl_encoder) {
-		u8 bl_level = amdgpu_display_backlight_get_level(adev,
-								  adev->mode_info.bl_encoder);
-		amdgpu_display_backlight_set_level(adev, adev->mode_info.bl_encoder,
-						    bl_level);
-	}
-	if (ret)
-		return ret;
-
-	return amdgpu_display_resume_helper(adev);
-}
-
-static bool dce_v11_0_is_idle(void *handle)
-{
-	return true;
-}
-
-static int dce_v11_0_wait_for_idle(void *handle)
-{
-	return 0;
-}
-
-static int dce_v11_0_soft_reset(void *handle)
-{
-	u32 srbm_soft_reset = 0, tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	if (dce_v11_0_is_display_hung(adev))
-		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_DC_MASK;
-
-	if (srbm_soft_reset) {
-		tmp = RREG32(mmSRBM_SOFT_RESET);
-		tmp |= srbm_soft_reset;
-		dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
-		WREG32(mmSRBM_SOFT_RESET, tmp);
-		tmp = RREG32(mmSRBM_SOFT_RESET);
-
-		udelay(50);
-
-		tmp &= ~srbm_soft_reset;
-		WREG32(mmSRBM_SOFT_RESET, tmp);
-		tmp = RREG32(mmSRBM_SOFT_RESET);
-
-		/* Wait a little for things to settle down */
-		udelay(50);
-	}
-	return 0;
-}
-
-static void dce_v11_0_set_crtc_vblank_interrupt_state(struct amdgpu_device *adev,
-						     int crtc,
-						     enum amdgpu_interrupt_state state)
-{
-	u32 lb_interrupt_mask;
-
-	if (crtc >= adev->mode_info.num_crtc) {
-		DRM_DEBUG("invalid crtc %d\n", crtc);
-		return;
-	}
-
-	switch (state) {
-	case AMDGPU_IRQ_STATE_DISABLE:
-		lb_interrupt_mask = RREG32(mmLB_INTERRUPT_MASK + crtc_offsets[crtc]);
-		lb_interrupt_mask = REG_SET_FIELD(lb_interrupt_mask, LB_INTERRUPT_MASK,
-						  VBLANK_INTERRUPT_MASK, 0);
-		WREG32(mmLB_INTERRUPT_MASK + crtc_offsets[crtc], lb_interrupt_mask);
-		break;
-	case AMDGPU_IRQ_STATE_ENABLE:
-		lb_interrupt_mask = RREG32(mmLB_INTERRUPT_MASK + crtc_offsets[crtc]);
-		lb_interrupt_mask = REG_SET_FIELD(lb_interrupt_mask, LB_INTERRUPT_MASK,
-						  VBLANK_INTERRUPT_MASK, 1);
-		WREG32(mmLB_INTERRUPT_MASK + crtc_offsets[crtc], lb_interrupt_mask);
-		break;
-	default:
-		break;
-	}
-}
-
-static void dce_v11_0_set_crtc_vline_interrupt_state(struct amdgpu_device *adev,
-						    int crtc,
-						    enum amdgpu_interrupt_state state)
-{
-	u32 lb_interrupt_mask;
-
-	if (crtc >= adev->mode_info.num_crtc) {
-		DRM_DEBUG("invalid crtc %d\n", crtc);
-		return;
-	}
-
-	switch (state) {
-	case AMDGPU_IRQ_STATE_DISABLE:
-		lb_interrupt_mask = RREG32(mmLB_INTERRUPT_MASK + crtc_offsets[crtc]);
-		lb_interrupt_mask = REG_SET_FIELD(lb_interrupt_mask, LB_INTERRUPT_MASK,
-						  VLINE_INTERRUPT_MASK, 0);
-		WREG32(mmLB_INTERRUPT_MASK + crtc_offsets[crtc], lb_interrupt_mask);
-		break;
-	case AMDGPU_IRQ_STATE_ENABLE:
-		lb_interrupt_mask = RREG32(mmLB_INTERRUPT_MASK + crtc_offsets[crtc]);
-		lb_interrupt_mask = REG_SET_FIELD(lb_interrupt_mask, LB_INTERRUPT_MASK,
-						  VLINE_INTERRUPT_MASK, 1);
-		WREG32(mmLB_INTERRUPT_MASK + crtc_offsets[crtc], lb_interrupt_mask);
-		break;
-	default:
-		break;
-	}
-}
-
-static int dce_v11_0_set_hpd_irq_state(struct amdgpu_device *adev,
-					struct amdgpu_irq_src *source,
-					unsigned hpd,
-					enum amdgpu_interrupt_state state)
-{
-	u32 tmp;
-
-	if (hpd >= adev->mode_info.num_hpd) {
-		DRM_DEBUG("invalid hdp %d\n", hpd);
-		return 0;
-	}
-
-	switch (state) {
-	case AMDGPU_IRQ_STATE_DISABLE:
-		tmp = RREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[hpd]);
-		tmp = REG_SET_FIELD(tmp, DC_HPD_INT_CONTROL, DC_HPD_INT_EN, 0);
-		WREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[hpd], tmp);
-		break;
-	case AMDGPU_IRQ_STATE_ENABLE:
-		tmp = RREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[hpd]);
-		tmp = REG_SET_FIELD(tmp, DC_HPD_INT_CONTROL, DC_HPD_INT_EN, 1);
-		WREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[hpd], tmp);
-		break;
-	default:
-		break;
-	}
-
-	return 0;
-}
-
-static int dce_v11_0_set_crtc_irq_state(struct amdgpu_device *adev,
-					struct amdgpu_irq_src *source,
-					unsigned type,
-					enum amdgpu_interrupt_state state)
-{
-	switch (type) {
-	case AMDGPU_CRTC_IRQ_VBLANK1:
-		dce_v11_0_set_crtc_vblank_interrupt_state(adev, 0, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VBLANK2:
-		dce_v11_0_set_crtc_vblank_interrupt_state(adev, 1, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VBLANK3:
-		dce_v11_0_set_crtc_vblank_interrupt_state(adev, 2, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VBLANK4:
-		dce_v11_0_set_crtc_vblank_interrupt_state(adev, 3, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VBLANK5:
-		dce_v11_0_set_crtc_vblank_interrupt_state(adev, 4, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VBLANK6:
-		dce_v11_0_set_crtc_vblank_interrupt_state(adev, 5, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VLINE1:
-		dce_v11_0_set_crtc_vline_interrupt_state(adev, 0, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VLINE2:
-		dce_v11_0_set_crtc_vline_interrupt_state(adev, 1, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VLINE3:
-		dce_v11_0_set_crtc_vline_interrupt_state(adev, 2, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VLINE4:
-		dce_v11_0_set_crtc_vline_interrupt_state(adev, 3, state);
-		break;
-	case AMDGPU_CRTC_IRQ_VLINE5:
-		dce_v11_0_set_crtc_vline_interrupt_state(adev, 4, state);
-		break;
-	 case AMDGPU_CRTC_IRQ_VLINE6:
-		dce_v11_0_set_crtc_vline_interrupt_state(adev, 5, state);
-		break;
-	default:
-		break;
-	}
-	return 0;
-}
-
-static int dce_v11_0_set_pageflip_irq_state(struct amdgpu_device *adev,
-					    struct amdgpu_irq_src *src,
-					    unsigned type,
-					    enum amdgpu_interrupt_state state)
-{
-	u32 reg;
-
-	if (type >= adev->mode_info.num_crtc) {
-		DRM_ERROR("invalid pageflip crtc %d\n", type);
-		return -EINVAL;
-	}
-
-	reg = RREG32(mmGRPH_INTERRUPT_CONTROL + crtc_offsets[type]);
-	if (state == AMDGPU_IRQ_STATE_DISABLE)
-		WREG32(mmGRPH_INTERRUPT_CONTROL + crtc_offsets[type],
-		       reg & ~GRPH_INTERRUPT_CONTROL__GRPH_PFLIP_INT_MASK_MASK);
-	else
-		WREG32(mmGRPH_INTERRUPT_CONTROL + crtc_offsets[type],
-		       reg | GRPH_INTERRUPT_CONTROL__GRPH_PFLIP_INT_MASK_MASK);
-
-	return 0;
-}
-
-static int dce_v11_0_pageflip_irq(struct amdgpu_device *adev,
-				  struct amdgpu_irq_src *source,
-				  struct amdgpu_iv_entry *entry)
-{
-	unsigned long flags;
-	unsigned crtc_id;
-	struct amdgpu_crtc *amdgpu_crtc;
-	struct amdgpu_flip_work *works;
-
-	crtc_id = (entry->src_id - 8) >> 1;
-	amdgpu_crtc = adev->mode_info.crtcs[crtc_id];
-
-	if (crtc_id >= adev->mode_info.num_crtc) {
-		DRM_ERROR("invalid pageflip crtc %d\n", crtc_id);
-		return -EINVAL;
-	}
-
-	if (RREG32(mmGRPH_INTERRUPT_STATUS + crtc_offsets[crtc_id]) &
-	    GRPH_INTERRUPT_STATUS__GRPH_PFLIP_INT_OCCURRED_MASK)
-		WREG32(mmGRPH_INTERRUPT_STATUS + crtc_offsets[crtc_id],
-		       GRPH_INTERRUPT_STATUS__GRPH_PFLIP_INT_CLEAR_MASK);
-
-	/* IRQ could occur when in initial stage */
-	if(amdgpu_crtc == NULL)
-		return 0;
-
-	spin_lock_irqsave(&adev_to_drm(adev)->event_lock, flags);
-	works = amdgpu_crtc->pflip_works;
-	if (amdgpu_crtc->pflip_status != AMDGPU_FLIP_SUBMITTED){
-		DRM_DEBUG_DRIVER("amdgpu_crtc->pflip_status = %d != "
-						 "AMDGPU_FLIP_SUBMITTED(%d)\n",
-						 amdgpu_crtc->pflip_status,
-						 AMDGPU_FLIP_SUBMITTED);
-		spin_unlock_irqrestore(&adev_to_drm(adev)->event_lock, flags);
-		return 0;
-	}
-
-	/* page flip completed. clean up */
-	amdgpu_crtc->pflip_status = AMDGPU_FLIP_NONE;
-	amdgpu_crtc->pflip_works = NULL;
-
-	/* wakeup usersapce */
-	if(works->event)
-		drm_crtc_send_vblank_event(&amdgpu_crtc->base, works->event);
-
-	spin_unlock_irqrestore(&adev_to_drm(adev)->event_lock, flags);
-
-	drm_crtc_vblank_put(&amdgpu_crtc->base);
-	schedule_work(&works->unpin_work);
-
-	return 0;
-}
-
-static void dce_v11_0_hpd_int_ack(struct amdgpu_device *adev,
-				  int hpd)
-{
-	u32 tmp;
-
-	if (hpd >= adev->mode_info.num_hpd) {
-		DRM_DEBUG("invalid hdp %d\n", hpd);
-		return;
-	}
-
-	tmp = RREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[hpd]);
-	tmp = REG_SET_FIELD(tmp, DC_HPD_INT_CONTROL, DC_HPD_INT_ACK, 1);
-	WREG32(mmDC_HPD_INT_CONTROL + hpd_offsets[hpd], tmp);
-}
-
-static void dce_v11_0_crtc_vblank_int_ack(struct amdgpu_device *adev,
-					  int crtc)
-{
-	u32 tmp;
-
-	if (crtc < 0 || crtc >= adev->mode_info.num_crtc) {
-		DRM_DEBUG("invalid crtc %d\n", crtc);
-		return;
-	}
-
-	tmp = RREG32(mmLB_VBLANK_STATUS + crtc_offsets[crtc]);
-	tmp = REG_SET_FIELD(tmp, LB_VBLANK_STATUS, VBLANK_ACK, 1);
-	WREG32(mmLB_VBLANK_STATUS + crtc_offsets[crtc], tmp);
-}
-
-static void dce_v11_0_crtc_vline_int_ack(struct amdgpu_device *adev,
-					 int crtc)
-{
-	u32 tmp;
-
-	if (crtc < 0 || crtc >= adev->mode_info.num_crtc) {
-		DRM_DEBUG("invalid crtc %d\n", crtc);
-		return;
-	}
-
-	tmp = RREG32(mmLB_VLINE_STATUS + crtc_offsets[crtc]);
-	tmp = REG_SET_FIELD(tmp, LB_VLINE_STATUS, VLINE_ACK, 1);
-	WREG32(mmLB_VLINE_STATUS + crtc_offsets[crtc], tmp);
-}
-
-static int dce_v11_0_crtc_irq(struct amdgpu_device *adev,
-				struct amdgpu_irq_src *source,
-				struct amdgpu_iv_entry *entry)
-{
-	unsigned crtc = entry->src_id - 1;
-	uint32_t disp_int = RREG32(interrupt_status_offsets[crtc].reg);
-	unsigned int irq_type = amdgpu_display_crtc_idx_to_irq_type(adev,
-								    crtc);
-
-	switch (entry->src_data[0]) {
-	case 0: /* vblank */
-		if (disp_int & interrupt_status_offsets[crtc].vblank)
-			dce_v11_0_crtc_vblank_int_ack(adev, crtc);
-		else
-			DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
-
-		if (amdgpu_irq_enabled(adev, source, irq_type)) {
-			drm_handle_vblank(adev_to_drm(adev), crtc);
-		}
-		DRM_DEBUG("IH: D%d vblank\n", crtc + 1);
-
-		break;
-	case 1: /* vline */
-		if (disp_int & interrupt_status_offsets[crtc].vline)
-			dce_v11_0_crtc_vline_int_ack(adev, crtc);
-		else
-			DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
-
-		DRM_DEBUG("IH: D%d vline\n", crtc + 1);
-
-		break;
-	default:
-		DRM_DEBUG("Unhandled interrupt: %d %d\n", entry->src_id, entry->src_data[0]);
-		break;
-	}
-
-	return 0;
-}
-
-static int dce_v11_0_hpd_irq(struct amdgpu_device *adev,
-			     struct amdgpu_irq_src *source,
-			     struct amdgpu_iv_entry *entry)
-{
-	uint32_t disp_int, mask;
-	unsigned hpd;
-
-	if (entry->src_data[0] >= adev->mode_info.num_hpd) {
-		DRM_DEBUG("Unhandled interrupt: %d %d\n", entry->src_id, entry->src_data[0]);
-		return 0;
-	}
-
-	hpd = entry->src_data[0];
-	disp_int = RREG32(interrupt_status_offsets[hpd].reg);
-	mask = interrupt_status_offsets[hpd].hpd;
-
-	if (disp_int & mask) {
-		dce_v11_0_hpd_int_ack(adev, hpd);
-		schedule_delayed_work(&adev->hotplug_work, 0);
-		DRM_DEBUG("IH: HPD%d\n", hpd + 1);
-	}
-
-	return 0;
-}
-
-static int dce_v11_0_set_clockgating_state(void *handle,
-					  enum amd_clockgating_state state)
-{
-	return 0;
-}
-
-static int dce_v11_0_set_powergating_state(void *handle,
-					  enum amd_powergating_state state)
-{
-	return 0;
-}
-
-static const struct amd_ip_funcs dce_v11_0_ip_funcs = {
-	.name = "dce_v11_0",
-	.early_init = dce_v11_0_early_init,
-	.late_init = NULL,
-	.sw_init = dce_v11_0_sw_init,
-	.sw_fini = dce_v11_0_sw_fini,
-	.hw_init = dce_v11_0_hw_init,
-	.hw_fini = dce_v11_0_hw_fini,
-	.suspend = dce_v11_0_suspend,
-	.resume = dce_v11_0_resume,
-	.is_idle = dce_v11_0_is_idle,
-	.wait_for_idle = dce_v11_0_wait_for_idle,
-	.soft_reset = dce_v11_0_soft_reset,
-	.set_clockgating_state = dce_v11_0_set_clockgating_state,
-	.set_powergating_state = dce_v11_0_set_powergating_state,
-};
-
-static void
-dce_v11_0_encoder_mode_set(struct drm_encoder *encoder,
-			  struct drm_display_mode *mode,
-			  struct drm_display_mode *adjusted_mode)
-{
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-
-	amdgpu_encoder->pixel_clock = adjusted_mode->clock;
-
-	/* need to call this here rather than in prepare() since we need some crtc info */
-	amdgpu_atombios_encoder_dpms(encoder, DRM_MODE_DPMS_OFF);
-
-	/* set scaler clears this on some chips */
-	dce_v11_0_set_interleave(encoder->crtc, mode);
-
-	if (amdgpu_atombios_encoder_get_encoder_mode(encoder) == ATOM_ENCODER_MODE_HDMI) {
-		dce_v11_0_afmt_enable(encoder, true);
-		dce_v11_0_afmt_setmode(encoder, adjusted_mode);
-	}
-}
-
-static void dce_v11_0_encoder_prepare(struct drm_encoder *encoder)
-{
-	struct amdgpu_device *adev = drm_to_adev(encoder->dev);
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct drm_connector *connector = amdgpu_get_connector_for_encoder(encoder);
-
-	if ((amdgpu_encoder->active_device &
-	     (ATOM_DEVICE_DFP_SUPPORT | ATOM_DEVICE_LCD_SUPPORT)) ||
-	    (amdgpu_encoder_get_dp_bridge_encoder_id(encoder) !=
-	     ENCODER_OBJECT_ID_NONE)) {
-		struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
-		if (dig) {
-			dig->dig_encoder = dce_v11_0_pick_dig_encoder(encoder);
-			if (amdgpu_encoder->active_device & ATOM_DEVICE_DFP_SUPPORT)
-				dig->afmt = adev->mode_info.afmt[dig->dig_encoder];
-		}
-	}
-
-	amdgpu_atombios_scratch_regs_lock(adev, true);
-
-	if (connector) {
-		struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
-
-		/* select the clock/data port if it uses a router */
-		if (amdgpu_connector->router.cd_valid)
-			amdgpu_i2c_router_select_cd_port(amdgpu_connector);
-
-		/* turn eDP panel on for mode set */
-		if (connector->connector_type == DRM_MODE_CONNECTOR_eDP)
-			amdgpu_atombios_encoder_set_edp_panel_power(connector,
-							     ATOM_TRANSMITTER_ACTION_POWER_ON);
-	}
-
-	/* this is needed for the pll/ss setup to work correctly in some cases */
-	amdgpu_atombios_encoder_set_crtc_source(encoder);
-	/* set up the FMT blocks */
-	dce_v11_0_program_fmt(encoder);
-}
-
-static void dce_v11_0_encoder_commit(struct drm_encoder *encoder)
-{
-	struct drm_device *dev = encoder->dev;
-	struct amdgpu_device *adev = drm_to_adev(dev);
-
-	/* need to call this here as we need the crtc set up */
-	amdgpu_atombios_encoder_dpms(encoder, DRM_MODE_DPMS_ON);
-	amdgpu_atombios_scratch_regs_lock(adev, false);
-}
-
-static void dce_v11_0_encoder_disable(struct drm_encoder *encoder)
-{
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	struct amdgpu_encoder_atom_dig *dig;
-
-	amdgpu_atombios_encoder_dpms(encoder, DRM_MODE_DPMS_OFF);
-
-	if (amdgpu_atombios_encoder_is_digital(encoder)) {
-		if (amdgpu_atombios_encoder_get_encoder_mode(encoder) == ATOM_ENCODER_MODE_HDMI)
-			dce_v11_0_afmt_enable(encoder, false);
-		dig = amdgpu_encoder->enc_priv;
-		dig->dig_encoder = -1;
-	}
-	amdgpu_encoder->active_device = 0;
-}
-
-/* these are handled by the primary encoders */
-static void dce_v11_0_ext_prepare(struct drm_encoder *encoder)
-{
-
-}
-
-static void dce_v11_0_ext_commit(struct drm_encoder *encoder)
-{
-
-}
-
-static void
-dce_v11_0_ext_mode_set(struct drm_encoder *encoder,
-		      struct drm_display_mode *mode,
-		      struct drm_display_mode *adjusted_mode)
-{
-
-}
-
-static void dce_v11_0_ext_disable(struct drm_encoder *encoder)
-{
-
-}
-
-static void
-dce_v11_0_ext_dpms(struct drm_encoder *encoder, int mode)
-{
-
-}
-
-static const struct drm_encoder_helper_funcs dce_v11_0_ext_helper_funcs = {
-	.dpms = dce_v11_0_ext_dpms,
-	.prepare = dce_v11_0_ext_prepare,
-	.mode_set = dce_v11_0_ext_mode_set,
-	.commit = dce_v11_0_ext_commit,
-	.disable = dce_v11_0_ext_disable,
-	/* no detect for TMDS/LVDS yet */
-};
-
-static const struct drm_encoder_helper_funcs dce_v11_0_dig_helper_funcs = {
-	.dpms = amdgpu_atombios_encoder_dpms,
-	.mode_fixup = amdgpu_atombios_encoder_mode_fixup,
-	.prepare = dce_v11_0_encoder_prepare,
-	.mode_set = dce_v11_0_encoder_mode_set,
-	.commit = dce_v11_0_encoder_commit,
-	.disable = dce_v11_0_encoder_disable,
-	.detect = amdgpu_atombios_encoder_dig_detect,
-};
-
-static const struct drm_encoder_helper_funcs dce_v11_0_dac_helper_funcs = {
-	.dpms = amdgpu_atombios_encoder_dpms,
-	.mode_fixup = amdgpu_atombios_encoder_mode_fixup,
-	.prepare = dce_v11_0_encoder_prepare,
-	.mode_set = dce_v11_0_encoder_mode_set,
-	.commit = dce_v11_0_encoder_commit,
-	.detect = amdgpu_atombios_encoder_dac_detect,
-};
-
-static void dce_v11_0_encoder_destroy(struct drm_encoder *encoder)
-{
-	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
-	if (amdgpu_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT))
-		amdgpu_atombios_encoder_fini_backlight(amdgpu_encoder);
-	kfree(amdgpu_encoder->enc_priv);
-	drm_encoder_cleanup(encoder);
-	kfree(amdgpu_encoder);
-}
-
-static const struct drm_encoder_funcs dce_v11_0_encoder_funcs = {
-	.destroy = dce_v11_0_encoder_destroy,
-};
-
-static void dce_v11_0_encoder_add(struct amdgpu_device *adev,
-				 uint32_t encoder_enum,
-				 uint32_t supported_device,
-				 u16 caps)
-{
-	struct drm_device *dev = adev_to_drm(adev);
-	struct drm_encoder *encoder;
-	struct amdgpu_encoder *amdgpu_encoder;
-
-	/* see if we already added it */
-	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
-		amdgpu_encoder = to_amdgpu_encoder(encoder);
-		if (amdgpu_encoder->encoder_enum == encoder_enum) {
-			amdgpu_encoder->devices |= supported_device;
-			return;
-		}
-
-	}
-
-	/* add a new one */
-	amdgpu_encoder = kzalloc(sizeof(struct amdgpu_encoder), GFP_KERNEL);
-	if (!amdgpu_encoder)
-		return;
-
-	encoder = &amdgpu_encoder->base;
-	switch (adev->mode_info.num_crtc) {
-	case 1:
-		encoder->possible_crtcs = 0x1;
-		break;
-	case 2:
-	default:
-		encoder->possible_crtcs = 0x3;
-		break;
-	case 3:
-		encoder->possible_crtcs = 0x7;
-		break;
-	case 4:
-		encoder->possible_crtcs = 0xf;
-		break;
-	case 5:
-		encoder->possible_crtcs = 0x1f;
-		break;
-	case 6:
-		encoder->possible_crtcs = 0x3f;
-		break;
-	}
-
-	amdgpu_encoder->enc_priv = NULL;
-
-	amdgpu_encoder->encoder_enum = encoder_enum;
-	amdgpu_encoder->encoder_id = (encoder_enum & OBJECT_ID_MASK) >> OBJECT_ID_SHIFT;
-	amdgpu_encoder->devices = supported_device;
-	amdgpu_encoder->rmx_type = RMX_OFF;
-	amdgpu_encoder->underscan_type = UNDERSCAN_OFF;
-	amdgpu_encoder->is_ext_encoder = false;
-	amdgpu_encoder->caps = caps;
-
-	switch (amdgpu_encoder->encoder_id) {
-	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC1:
-	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC2:
-		drm_encoder_init(dev, encoder, &dce_v11_0_encoder_funcs,
-				 DRM_MODE_ENCODER_DAC, NULL);
-		drm_encoder_helper_add(encoder, &dce_v11_0_dac_helper_funcs);
-		break;
-	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DVO1:
-	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
-	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
-	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
-	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3:
-		if (amdgpu_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT)) {
-			amdgpu_encoder->rmx_type = RMX_FULL;
-			drm_encoder_init(dev, encoder, &dce_v11_0_encoder_funcs,
-					 DRM_MODE_ENCODER_LVDS, NULL);
-			amdgpu_encoder->enc_priv = amdgpu_atombios_encoder_get_lcd_info(amdgpu_encoder);
-		} else if (amdgpu_encoder->devices & (ATOM_DEVICE_CRT_SUPPORT)) {
-			drm_encoder_init(dev, encoder, &dce_v11_0_encoder_funcs,
-					 DRM_MODE_ENCODER_DAC, NULL);
-			amdgpu_encoder->enc_priv = amdgpu_atombios_encoder_get_dig_info(amdgpu_encoder);
-		} else {
-			drm_encoder_init(dev, encoder, &dce_v11_0_encoder_funcs,
-					 DRM_MODE_ENCODER_TMDS, NULL);
-			amdgpu_encoder->enc_priv = amdgpu_atombios_encoder_get_dig_info(amdgpu_encoder);
-		}
-		drm_encoder_helper_add(encoder, &dce_v11_0_dig_helper_funcs);
-		break;
-	case ENCODER_OBJECT_ID_SI170B:
-	case ENCODER_OBJECT_ID_CH7303:
-	case ENCODER_OBJECT_ID_EXTERNAL_SDVOA:
-	case ENCODER_OBJECT_ID_EXTERNAL_SDVOB:
-	case ENCODER_OBJECT_ID_TITFP513:
-	case ENCODER_OBJECT_ID_VT1623:
-	case ENCODER_OBJECT_ID_HDMI_SI1930:
-	case ENCODER_OBJECT_ID_TRAVIS:
-	case ENCODER_OBJECT_ID_NUTMEG:
-		/* these are handled by the primary encoders */
-		amdgpu_encoder->is_ext_encoder = true;
-		if (amdgpu_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT))
-			drm_encoder_init(dev, encoder, &dce_v11_0_encoder_funcs,
-					 DRM_MODE_ENCODER_LVDS, NULL);
-		else if (amdgpu_encoder->devices & (ATOM_DEVICE_CRT_SUPPORT))
-			drm_encoder_init(dev, encoder, &dce_v11_0_encoder_funcs,
-					 DRM_MODE_ENCODER_DAC, NULL);
-		else
-			drm_encoder_init(dev, encoder, &dce_v11_0_encoder_funcs,
-					 DRM_MODE_ENCODER_TMDS, NULL);
-		drm_encoder_helper_add(encoder, &dce_v11_0_ext_helper_funcs);
-		break;
-	}
-}
-
-static const struct amdgpu_display_funcs dce_v11_0_display_funcs = {
-	.bandwidth_update = &dce_v11_0_bandwidth_update,
-	.vblank_get_counter = &dce_v11_0_vblank_get_counter,
-	.backlight_set_level = &amdgpu_atombios_encoder_set_backlight_level,
-	.backlight_get_level = &amdgpu_atombios_encoder_get_backlight_level,
-	.hpd_sense = &dce_v11_0_hpd_sense,
-	.hpd_set_polarity = &dce_v11_0_hpd_set_polarity,
-	.hpd_get_gpio_reg = &dce_v11_0_hpd_get_gpio_reg,
-	.page_flip = &dce_v11_0_page_flip,
-	.page_flip_get_scanoutpos = &dce_v11_0_crtc_get_scanoutpos,
-	.add_encoder = &dce_v11_0_encoder_add,
-	.add_connector = &amdgpu_connector_add,
-};
-
-static void dce_v11_0_set_display_funcs(struct amdgpu_device *adev)
-{
-	adev->mode_info.funcs = &dce_v11_0_display_funcs;
-}
-
-static const struct amdgpu_irq_src_funcs dce_v11_0_crtc_irq_funcs = {
-	.set = dce_v11_0_set_crtc_irq_state,
-	.process = dce_v11_0_crtc_irq,
-};
-
-static const struct amdgpu_irq_src_funcs dce_v11_0_pageflip_irq_funcs = {
-	.set = dce_v11_0_set_pageflip_irq_state,
-	.process = dce_v11_0_pageflip_irq,
-};
-
-static const struct amdgpu_irq_src_funcs dce_v11_0_hpd_irq_funcs = {
-	.set = dce_v11_0_set_hpd_irq_state,
-	.process = dce_v11_0_hpd_irq,
-};
-
-static void dce_v11_0_set_irq_funcs(struct amdgpu_device *adev)
-{
-	if (adev->mode_info.num_crtc > 0)
-		adev->crtc_irq.num_types = AMDGPU_CRTC_IRQ_VLINE1 + adev->mode_info.num_crtc;
-	else
-		adev->crtc_irq.num_types = 0;
-	adev->crtc_irq.funcs = &dce_v11_0_crtc_irq_funcs;
-
-	adev->pageflip_irq.num_types = adev->mode_info.num_crtc;
-	adev->pageflip_irq.funcs = &dce_v11_0_pageflip_irq_funcs;
-
-	adev->hpd_irq.num_types = adev->mode_info.num_hpd;
-	adev->hpd_irq.funcs = &dce_v11_0_hpd_irq_funcs;
-}
-
-const struct amdgpu_ip_block_version dce_v11_0_ip_block =
-{
-	.type = AMD_IP_BLOCK_TYPE_DCE,
-	.major = 11,
-	.minor = 0,
-	.rev = 0,
-	.funcs = &dce_v11_0_ip_funcs,
-};
-
-const struct amdgpu_ip_block_version dce_v11_2_ip_block =
-{
-	.type = AMD_IP_BLOCK_TYPE_DCE,
-	.major = 11,
-	.minor = 2,
-	.rev = 0,
-	.funcs = &dce_v11_0_ip_funcs,
-};
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c
index 7f85ba5b726f..acc887a58518 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c
@@ -23,6 +23,7 @@
 
 #include <linux/pci.h>
 
+#include <drm/drm_edid.h>
 #include <drm/drm_fourcc.h>
 #include <drm/drm_modeset_helper.h>
 #include <drm/drm_modeset_helper_vtables.h>
@@ -39,18 +40,25 @@
 #include "amdgpu_connectors.h"
 #include "amdgpu_display.h"
 
+#include "dce_v6_0.h"
+#include "sid.h"
+
 #include "bif/bif_3_0_d.h"
 #include "bif/bif_3_0_sh_mask.h"
+
 #include "oss/oss_1_0_d.h"
 #include "oss/oss_1_0_sh_mask.h"
+
 #include "gca/gfx_6_0_d.h"
 #include "gca/gfx_6_0_sh_mask.h"
+#include "gca/gfx_7_2_enum.h"
+
 #include "gmc/gmc_6_0_d.h"
 #include "gmc/gmc_6_0_sh_mask.h"
+
 #include "dce/dce_6_0_d.h"
 #include "dce/dce_6_0_sh_mask.h"
-#include "gca/gfx_7_2_enum.h"
-#include "dce_v6_0.h"
+
 #include "si_enums.h"
 
 static void dce_v6_0_set_display_funcs(struct amdgpu_device *adev);
@@ -58,31 +66,31 @@ static void dce_v6_0_set_irq_funcs(struct amdgpu_device *adev);
 
 static const u32 crtc_offsets[6] =
 {
-	SI_CRTC0_REGISTER_OFFSET,
-	SI_CRTC1_REGISTER_OFFSET,
-	SI_CRTC2_REGISTER_OFFSET,
-	SI_CRTC3_REGISTER_OFFSET,
-	SI_CRTC4_REGISTER_OFFSET,
-	SI_CRTC5_REGISTER_OFFSET
+	CRTC0_REGISTER_OFFSET,
+	CRTC1_REGISTER_OFFSET,
+	CRTC2_REGISTER_OFFSET,
+	CRTC3_REGISTER_OFFSET,
+	CRTC4_REGISTER_OFFSET,
+	CRTC5_REGISTER_OFFSET
 };
 
 static const u32 hpd_offsets[] =
 {
-	mmDC_HPD1_INT_STATUS - mmDC_HPD1_INT_STATUS,
-	mmDC_HPD2_INT_STATUS - mmDC_HPD1_INT_STATUS,
-	mmDC_HPD3_INT_STATUS - mmDC_HPD1_INT_STATUS,
-	mmDC_HPD4_INT_STATUS - mmDC_HPD1_INT_STATUS,
-	mmDC_HPD5_INT_STATUS - mmDC_HPD1_INT_STATUS,
-	mmDC_HPD6_INT_STATUS - mmDC_HPD1_INT_STATUS,
+	HPD0_REGISTER_OFFSET,
+	HPD1_REGISTER_OFFSET,
+	HPD2_REGISTER_OFFSET,
+	HPD3_REGISTER_OFFSET,
+	HPD4_REGISTER_OFFSET,
+	HPD5_REGISTER_OFFSET
 };
 
 static const uint32_t dig_offsets[] = {
-	SI_CRTC0_REGISTER_OFFSET,
-	SI_CRTC1_REGISTER_OFFSET,
-	SI_CRTC2_REGISTER_OFFSET,
-	SI_CRTC3_REGISTER_OFFSET,
-	SI_CRTC4_REGISTER_OFFSET,
-	SI_CRTC5_REGISTER_OFFSET,
+	CRTC0_REGISTER_OFFSET,
+	CRTC1_REGISTER_OFFSET,
+	CRTC2_REGISTER_OFFSET,
+	CRTC3_REGISTER_OFFSET,
+	CRTC4_REGISTER_OFFSET,
+	CRTC5_REGISTER_OFFSET,
 	(0x13830 - 0x7030) >> 2,
 };
 
@@ -205,9 +213,9 @@ static void dce_v6_0_page_flip(struct amdgpu_device *adev,
 	/* update the scanout addresses */
 	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS_HIGH + amdgpu_crtc->crtc_offset,
 	       upper_32_bits(crtc_base));
+	/* writing to the low address triggers the update */
 	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
 	       (u32)crtc_base);
-
 	/* post the write */
 	RREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset);
 }
@@ -217,11 +225,11 @@ static int dce_v6_0_crtc_get_scanoutpos(struct amdgpu_device *adev, int crtc,
 {
 	if ((crtc < 0) || (crtc >= adev->mode_info.num_crtc))
 		return -EINVAL;
+
 	*vbl = RREG32(mmCRTC_V_BLANK_START_END + crtc_offsets[crtc]);
 	*position = RREG32(mmCRTC_STATUS_POSITION + crtc_offsets[crtc]);
 
 	return 0;
-
 }
 
 /**
@@ -241,7 +249,8 @@ static bool dce_v6_0_hpd_sense(struct amdgpu_device *adev,
 	if (hpd >= adev->mode_info.num_hpd)
 		return connected;
 
-	if (RREG32(mmDC_HPD1_INT_STATUS + hpd_offsets[hpd]) & DC_HPD1_INT_STATUS__DC_HPD1_SENSE_MASK)
+	if (RREG32(mmDC_HPD1_INT_STATUS + hpd_offsets[hpd]) &
+	    DC_HPD1_INT_STATUS__DC_HPD1_SENSE_MASK)
 		connected = true;
 
 	return connected;
@@ -272,6 +281,21 @@ static void dce_v6_0_hpd_set_polarity(struct amdgpu_device *adev,
 	WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd], tmp);
 }
 
+static void dce_v6_0_hpd_int_ack(struct amdgpu_device *adev,
+				 int hpd)
+{
+	u32 tmp;
+
+	if (hpd >= adev->mode_info.num_hpd) {
+		DRM_DEBUG("invalid hpd %d\n", hpd);
+		return;
+	}
+
+	tmp = RREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd]);
+	tmp |= DC_HPD1_INT_CONTROL__DC_HPD1_INT_ACK_MASK;
+	WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd], tmp);
+}
+
 /**
  * dce_v6_0_hpd_init - hpd setup callback.
  *
@@ -311,6 +335,7 @@ static void dce_v6_0_hpd_init(struct amdgpu_device *adev)
 			continue;
 		}
 
+		dce_v6_0_hpd_int_ack(adev, amdgpu_connector->hpd.hpd);
 		dce_v6_0_hpd_set_polarity(adev, amdgpu_connector->hpd.hpd);
 		amdgpu_irq_get(adev, &adev->hpd_irq, amdgpu_connector->hpd.hpd);
 	}
@@ -353,13 +378,41 @@ static u32 dce_v6_0_hpd_get_gpio_reg(struct amdgpu_device *adev)
 	return mmDC_GPIO_HPD_A;
 }
 
+static bool dce_v6_0_is_display_hung(struct amdgpu_device *adev)
+{
+	u32 crtc_hung = 0;
+	u32 crtc_status[6];
+	u32 i, j, tmp;
+
+	for (i = 0; i < adev->mode_info.num_crtc; i++) {
+		if (RREG32(mmCRTC_CONTROL + crtc_offsets[i]) & CRTC_CONTROL__CRTC_MASTER_EN_MASK) {
+			crtc_status[i] = RREG32(mmCRTC_STATUS_HV_COUNT + crtc_offsets[i]);
+			crtc_hung |= (1 << i);
+		}
+	}
+
+	for (j = 0; j < 10; j++) {
+		for (i = 0; i < adev->mode_info.num_crtc; i++) {
+			if (crtc_hung & (1 << i)) {
+				tmp = RREG32(mmCRTC_STATUS_HV_COUNT + crtc_offsets[i]);
+				if (tmp != crtc_status[i])
+					crtc_hung &= ~(1 << i);
+			}
+		}
+		if (crtc_hung == 0)
+			return false;
+		udelay(100);
+	}
+
+	return true;
+}
+
 static void dce_v6_0_set_vga_render_state(struct amdgpu_device *adev,
 					  bool render)
 {
 	if (!render)
 		WREG32(mmVGA_RENDER_CONTROL,
-			RREG32(mmVGA_RENDER_CONTROL) & VGA_VSTATUS_CNTL);
-
+		       RREG32(mmVGA_RENDER_CONTROL) & ~VGA_RENDER_CONTROL__VGA_VSTATUS_CNTL_MASK);
 }
 
 static int dce_v6_0_get_num_crtc(struct amdgpu_device *adev)
@@ -402,7 +455,6 @@ void dce_v6_0_disable_dce(struct amdgpu_device *adev)
 
 static void dce_v6_0_program_fmt(struct drm_encoder *encoder)
 {
-
 	struct drm_device *dev = encoder->dev;
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
@@ -845,7 +897,7 @@ static void dce_v6_0_program_watermarks(struct amdgpu_device *adev,
 					    (u32)mode->clock);
 		line_time = (u32) div_u64((u64)mode->crtc_htotal * 1000000,
 					  (u32)mode->clock);
-		line_time = min(line_time, (u32)65535);
+		line_time = min_t(u32, line_time, 65535);
 		priority_a_cnt = 0;
 		priority_b_cnt = 0;
 
@@ -878,8 +930,8 @@ static void dce_v6_0_program_watermarks(struct amdgpu_device *adev,
 		wm_high.dram_channels = dram_channels;
 		wm_high.num_heads = num_heads;
 
-		if (adev->pm.dpm_enabled) {
 		/* watermark for low clocks */
+		if (adev->pm.dpm_enabled) {
 			wm_low.yclk =
 				amdgpu_dpm_get_mclk(adev, true) * 10;
 			wm_low.sclk =
@@ -906,9 +958,9 @@ static void dce_v6_0_program_watermarks(struct amdgpu_device *adev,
 		wm_low.num_heads = num_heads;
 
 		/* set for high clocks */
-		latency_watermark_a = min(dce_v6_0_latency_watermark(&wm_high), (u32)65535);
+		latency_watermark_a = min_t(u32, dce_v6_0_latency_watermark(&wm_high), 65535);
 		/* set for low clocks */
-		latency_watermark_b = min(dce_v6_0_latency_watermark(&wm_low), (u32)65535);
+		latency_watermark_b = min_t(u32, dce_v6_0_latency_watermark(&wm_low), 65535);
 
 		/* possibly force display priority to high */
 		/* should really do this at mode validation time... */
@@ -959,16 +1011,16 @@ static void dce_v6_0_program_watermarks(struct amdgpu_device *adev,
 	/* select wm A */
 	arb_control3 = RREG32(mmDPG_PIPE_ARBITRATION_CONTROL3 + amdgpu_crtc->crtc_offset);
 	tmp = arb_control3;
-	tmp &= ~LATENCY_WATERMARK_MASK(3);
-	tmp |= LATENCY_WATERMARK_MASK(1);
+	tmp &= ~(3 << DPG_PIPE_ARBITRATION_CONTROL3__URGENCY_WATERMARK_MASK__SHIFT);
+	tmp |= (1 << DPG_PIPE_ARBITRATION_CONTROL3__URGENCY_WATERMARK_MASK__SHIFT);
 	WREG32(mmDPG_PIPE_ARBITRATION_CONTROL3 + amdgpu_crtc->crtc_offset, tmp);
 	WREG32(mmDPG_PIPE_URGENCY_CONTROL + amdgpu_crtc->crtc_offset,
 	       ((latency_watermark_a << DPG_PIPE_URGENCY_CONTROL__URGENCY_LOW_WATERMARK__SHIFT)  |
 		(line_time << DPG_PIPE_URGENCY_CONTROL__URGENCY_HIGH_WATERMARK__SHIFT)));
 	/* select wm B */
 	tmp = RREG32(mmDPG_PIPE_ARBITRATION_CONTROL3 + amdgpu_crtc->crtc_offset);
-	tmp &= ~LATENCY_WATERMARK_MASK(3);
-	tmp |= LATENCY_WATERMARK_MASK(2);
+	tmp &= ~(3 << DPG_PIPE_ARBITRATION_CONTROL3__URGENCY_WATERMARK_MASK__SHIFT);
+	tmp |= (2 << DPG_PIPE_ARBITRATION_CONTROL3__URGENCY_WATERMARK_MASK__SHIFT);
 	WREG32(mmDPG_PIPE_ARBITRATION_CONTROL3 + amdgpu_crtc->crtc_offset, tmp);
 	WREG32(mmDPG_PIPE_URGENCY_CONTROL + amdgpu_crtc->crtc_offset,
 	       ((latency_watermark_b << DPG_PIPE_URGENCY_CONTROL__URGENCY_LOW_WATERMARK__SHIFT) |
@@ -982,13 +1034,26 @@ static void dce_v6_0_program_watermarks(struct amdgpu_device *adev,
 
 	/* save values for DPM */
 	amdgpu_crtc->line_time = line_time;
-	amdgpu_crtc->wm_high = latency_watermark_a;
 
 	/* Save number of lines the linebuffer leads before the scanout */
 	amdgpu_crtc->lb_vblank_lead_lines = lb_vblank_lead_lines;
 }
 
 /* watermark setup */
+/**
+ * dce_v6_0_line_buffer_adjust - Set up the line buffer
+ *
+ * @adev: amdgpu_device pointer
+ * @amdgpu_crtc: the selected display controller
+ * @mode: the current display mode on the selected display
+ * controller
+ * @other_mode: the display mode of another display controller
+ *              that may be sharing the line buffer
+ *
+ * Setup up the line buffer allocation for
+ * the selected display controller (CIK).
+ * Returns the line buffer size in pixels.
+ */
 static u32 dce_v6_0_line_buffer_adjust(struct amdgpu_device *adev,
 				   struct amdgpu_crtc *amdgpu_crtc,
 				   struct drm_display_mode *mode,
@@ -1023,7 +1088,7 @@ static u32 dce_v6_0_line_buffer_adjust(struct amdgpu_device *adev,
 	}
 
 	WREG32(mmDC_LB_MEMORY_SPLIT + amdgpu_crtc->crtc_offset,
-	       DC_LB_MEMORY_CONFIG(tmp));
+	       (tmp << DC_LB_MEMORY_SPLIT__DC_LB_MEMORY_CONFIG__SHIFT));
 
 	WREG32(mmPIPE0_DMIF_BUFFER_CONTROL + pipe_offset,
 	       (buffer_alloc << PIPE0_DMIF_BUFFER_CONTROL__DMIF_BUFFERS_ALLOCATED__SHIFT));
@@ -1200,7 +1265,7 @@ static void dce_v6_0_audio_write_speaker_allocation(struct drm_encoder *encoder)
 		return;
 	}
 
-	sad_count = drm_edid_to_speaker_allocation(amdgpu_connector_edid(connector), &sadb);
+	sad_count = drm_edid_to_speaker_allocation(amdgpu_connector->edid, &sadb);
 	if (sad_count < 0) {
 		DRM_ERROR("Couldn't read Speaker Allocation Data Block: %d\n", sad_count);
 		sad_count = 0;
@@ -1240,6 +1305,7 @@ static void dce_v6_0_audio_write_sad_regs(struct drm_encoder *encoder)
 	struct amdgpu_device *adev = drm_to_adev(dev);
 	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
 	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
+	u32 offset;
 	struct drm_connector *connector;
 	struct drm_connector_list_iter iter;
 	struct amdgpu_connector *amdgpu_connector = NULL;
@@ -1261,6 +1327,11 @@ static void dce_v6_0_audio_write_sad_regs(struct drm_encoder *encoder)
 		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR13, HDMI_AUDIO_CODING_TYPE_WMA_PRO },
 	};
 
+	if (!dig || !dig->afmt || !dig->afmt->pin)
+		return;
+
+	offset = dig->afmt->pin->offset;
+
 	drm_connector_list_iter_begin(dev, &iter);
 	drm_for_each_connector_iter(connector, &iter) {
 		if (connector->encoder == encoder) {
@@ -1275,14 +1346,14 @@ static void dce_v6_0_audio_write_sad_regs(struct drm_encoder *encoder)
 		return;
 	}
 
-	sad_count = drm_edid_to_sad(amdgpu_connector_edid(connector), &sads);
+	sad_count = drm_edid_to_sad(amdgpu_connector->edid, &sads);
 	if (sad_count < 0)
 		DRM_ERROR("Couldn't read SADs: %d\n", sad_count);
 	if (sad_count <= 0)
 		return;
 
 	for (i = 0; i < ARRAY_SIZE(eld_reg_to_type); i++) {
-		u32 tmp = 0;
+		u32 value = 0;
 		u8 stereo_freqs = 0;
 		int max_channels = -1;
 		int j;
@@ -1292,12 +1363,12 @@ static void dce_v6_0_audio_write_sad_regs(struct drm_encoder *encoder)
 
 			if (sad->format == eld_reg_to_type[i][1]) {
 				if (sad->channels > max_channels) {
-					tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0,
-							MAX_CHANNELS, sad->channels);
-					tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0,
-							DESCRIPTOR_BYTE_2, sad->byte2);
-					tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0,
-							SUPPORTED_FREQUENCIES, sad->freq);
+					value = (sad->channels <<
+						AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__MAX_CHANNELS__SHIFT) |
+					       (sad->byte2 <<
+						AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__DESCRIPTOR_BYTE_2__SHIFT) |
+					       (sad->freq <<
+						AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__SUPPORTED_FREQUENCIES__SHIFT);
 					max_channels = sad->channels;
 				}
 
@@ -1308,13 +1379,13 @@ static void dce_v6_0_audio_write_sad_regs(struct drm_encoder *encoder)
 			}
 		}
 
-		tmp = REG_SET_FIELD(tmp, AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0,
-				SUPPORTED_FREQUENCIES_STEREO, stereo_freqs);
-		WREG32_AUDIO_ENDPT(dig->afmt->pin->offset, eld_reg_to_type[i][0], tmp);
+		value |= (stereo_freqs <<
+			AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__SUPPORTED_FREQUENCIES_STEREO__SHIFT);
+
+		WREG32_AUDIO_ENDPT(offset, eld_reg_to_type[i][0], value);
 	}
 
 	kfree(sads);
-
 }
 
 static void dce_v6_0_audio_enable(struct amdgpu_device *adev,
@@ -1330,13 +1401,13 @@ static void dce_v6_0_audio_enable(struct amdgpu_device *adev,
 
 static const u32 pin_offsets[7] =
 {
-	(0x1780 - 0x1780),
-	(0x1786 - 0x1780),
-	(0x178c - 0x1780),
-	(0x1792 - 0x1780),
-	(0x1798 - 0x1780),
-	(0x179d - 0x1780),
-	(0x17a4 - 0x1780),
+	AUD0_REGISTER_OFFSET,
+	AUD1_REGISTER_OFFSET,
+	AUD2_REGISTER_OFFSET,
+	AUD3_REGISTER_OFFSET,
+	AUD4_REGISTER_OFFSET,
+	AUD5_REGISTER_OFFSET,
+	AUD6_REGISTER_OFFSET,
 };
 
 static int dce_v6_0_audio_init(struct amdgpu_device *adev)
@@ -1369,6 +1440,8 @@ static int dce_v6_0_audio_init(struct amdgpu_device *adev)
 		adev->mode_info.audio.pin[i].connected = false;
 		adev->mode_info.audio.pin[i].offset = pin_offsets[i];
 		adev->mode_info.audio.pin[i].id = i;
+		/* disable audio.  it will be set up later */
+		/* XXX remove once we switch to ip funcs */
 		dce_v6_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
 	}
 
@@ -1377,17 +1450,12 @@ static int dce_v6_0_audio_init(struct amdgpu_device *adev)
 
 static void dce_v6_0_audio_fini(struct amdgpu_device *adev)
 {
-	int i;
-
 	if (!amdgpu_audio)
 		return;
 
 	if (!adev->mode_info.audio.enabled)
 		return;
 
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++)
-		dce_v6_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
-
 	adev->mode_info.audio.enabled = false;
 }
 
@@ -1818,7 +1886,7 @@ static int dce_v6_0_crtc_do_set_base(struct drm_crtc *crtc,
 	struct amdgpu_bo *abo;
 	uint64_t fb_location, tiling_flags;
 	uint32_t fb_format, fb_pitch_pixels, pipe_config;
-	u32 fb_swap = GRPH_ENDIAN_SWAP(GRPH_ENDIAN_NONE);
+	u32 fb_swap = (GRPH_ENDIAN_NONE << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 	u32 viewport_w, viewport_h;
 	int r;
 	bool bypass_lut = false;
@@ -1844,6 +1912,7 @@ static int dce_v6_0_crtc_do_set_base(struct drm_crtc *crtc,
 		return r;
 
 	if (!atomic) {
+		abo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 		r = amdgpu_bo_pin(abo, AMDGPU_GEM_DOMAIN_VRAM);
 		if (unlikely(r != 0)) {
 			amdgpu_bo_unreserve(abo);
@@ -1857,76 +1926,76 @@ static int dce_v6_0_crtc_do_set_base(struct drm_crtc *crtc,
 
 	switch (target_fb->format->format) {
 	case DRM_FORMAT_C8:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_8BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_INDEXED));
+		fb_format = ((GRPH_DEPTH_8BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_INDEXED << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 		break;
 	case DRM_FORMAT_XRGB4444:
 	case DRM_FORMAT_ARGB4444:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_16BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_ARGB4444));
+		fb_format = ((GRPH_DEPTH_16BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB4444 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 #ifdef __BIG_ENDIAN
-		fb_swap = GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN16);
+		fb_swap = (GRPH_ENDIAN_8IN16 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
 		break;
 	case DRM_FORMAT_XRGB1555:
 	case DRM_FORMAT_ARGB1555:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_16BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_ARGB1555));
+		fb_format = ((GRPH_DEPTH_16BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB1555 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 #ifdef __BIG_ENDIAN
-		fb_swap = GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN16);
+		fb_swap = (GRPH_ENDIAN_8IN16 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
 		break;
 	case DRM_FORMAT_BGRX5551:
 	case DRM_FORMAT_BGRA5551:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_16BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_BGRA5551));
+		fb_format = ((GRPH_DEPTH_16BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_BGRA5551 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 #ifdef __BIG_ENDIAN
-		fb_swap = GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN16);
+		fb_swap = (GRPH_ENDIAN_8IN16 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
 		break;
 	case DRM_FORMAT_RGB565:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_16BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_ARGB565));
+		fb_format = ((GRPH_DEPTH_16BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB565 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 #ifdef __BIG_ENDIAN
-		fb_swap = GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN16);
+		fb_swap = (GRPH_ENDIAN_8IN16 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
 		break;
 	case DRM_FORMAT_XRGB8888:
 	case DRM_FORMAT_ARGB8888:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_32BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_ARGB8888));
+		fb_format = ((GRPH_DEPTH_32BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB8888 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 #ifdef __BIG_ENDIAN
-		fb_swap = GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN32);
+		fb_swap = (GRPH_ENDIAN_8IN32 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
 		break;
 	case DRM_FORMAT_XRGB2101010:
 	case DRM_FORMAT_ARGB2101010:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_32BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_ARGB2101010));
+		fb_format = ((GRPH_DEPTH_32BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB2101010 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 #ifdef __BIG_ENDIAN
-		fb_swap = GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN32);
+		fb_swap = (GRPH_ENDIAN_8IN32 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
 		/* Greater 8 bpc fb needs to bypass hw-lut to retain precision */
 		bypass_lut = true;
 		break;
 	case DRM_FORMAT_BGRX1010102:
 	case DRM_FORMAT_BGRA1010102:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_32BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_BGRA1010102));
+		fb_format = ((GRPH_DEPTH_32BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_BGRA1010102 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 #ifdef __BIG_ENDIAN
-		fb_swap = GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN32);
+		fb_swap = (GRPH_ENDIAN_8IN32 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
 		/* Greater 8 bpc fb needs to bypass hw-lut to retain precision */
 		bypass_lut = true;
 		break;
 	case DRM_FORMAT_XBGR8888:
 	case DRM_FORMAT_ABGR8888:
-		fb_format = (GRPH_DEPTH(GRPH_DEPTH_32BPP) |
-			     GRPH_FORMAT(GRPH_FORMAT_ARGB8888));
-		fb_swap = (GRPH_RED_CROSSBAR(GRPH_RED_SEL_B) |
-			   GRPH_BLUE_CROSSBAR(GRPH_BLUE_SEL_R));
+		fb_format = ((GRPH_DEPTH_32BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB8888 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
+		fb_swap = ((GRPH_RED_SEL_B << GRPH_SWAP_CNTL__GRPH_RED_CROSSBAR__SHIFT) |
+			   (GRPH_BLUE_SEL_R << GRPH_SWAP_CNTL__GRPH_BLUE_CROSSBAR__SHIFT));
 #ifdef __BIG_ENDIAN
-		fb_swap |= GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN32);
+		fb_swap |= (GRPH_ENDIAN_8IN32 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
 		break;
 	default:
@@ -1944,18 +2013,18 @@ static int dce_v6_0_crtc_do_set_base(struct drm_crtc *crtc,
 		tile_split = AMDGPU_TILING_GET(tiling_flags, TILE_SPLIT);
 		num_banks = AMDGPU_TILING_GET(tiling_flags, NUM_BANKS);
 
-		fb_format |= GRPH_NUM_BANKS(num_banks);
-		fb_format |= GRPH_ARRAY_MODE(GRPH_ARRAY_2D_TILED_THIN1);
-		fb_format |= GRPH_TILE_SPLIT(tile_split);
-		fb_format |= GRPH_BANK_WIDTH(bankw);
-		fb_format |= GRPH_BANK_HEIGHT(bankh);
-		fb_format |= GRPH_MACRO_TILE_ASPECT(mtaspect);
+		fb_format |= (num_banks << GRPH_CONTROL__GRPH_NUM_BANKS__SHIFT);
+		fb_format |= (GRPH_ARRAY_2D_TILED_THIN1 << GRPH_CONTROL__GRPH_ARRAY_MODE__SHIFT);
+		fb_format |= (tile_split << GRPH_CONTROL__GRPH_TILE_SPLIT__SHIFT);
+		fb_format |= (bankw << GRPH_CONTROL__GRPH_BANK_WIDTH__SHIFT);
+		fb_format |= (bankh << GRPH_CONTROL__GRPH_BANK_HEIGHT__SHIFT);
+		fb_format |= (mtaspect << GRPH_CONTROL__GRPH_MACRO_TILE_ASPECT__SHIFT);
 	} else if (AMDGPU_TILING_GET(tiling_flags, ARRAY_MODE) == ARRAY_1D_TILED_THIN1) {
-		fb_format |= GRPH_ARRAY_MODE(GRPH_ARRAY_1D_TILED_THIN1);
+		fb_format |= (GRPH_ARRAY_1D_TILED_THIN1 << GRPH_CONTROL__GRPH_ARRAY_MODE__SHIFT);
 	}
 
 	pipe_config = AMDGPU_TILING_GET(tiling_flags, PIPE_CONFIG);
-	fb_format |= GRPH_PIPE_CONFIG(pipe_config);
+	fb_format |= (pipe_config << GRPH_CONTROL__GRPH_PIPE_CONFIG__SHIFT);
 
 	dce_v6_0_vga_enable(crtc, false);
 
@@ -1971,7 +2040,7 @@ static int dce_v6_0_crtc_do_set_base(struct drm_crtc *crtc,
 	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
 	       (u32)fb_location & GRPH_PRIMARY_SURFACE_ADDRESS__GRPH_PRIMARY_SURFACE_ADDRESS_MASK);
 	WREG32(mmGRPH_SECONDARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
-	       (u32) fb_location & GRPH_PRIMARY_SURFACE_ADDRESS__GRPH_PRIMARY_SURFACE_ADDRESS_MASK);
+	       (u32) fb_location & GRPH_SECONDARY_SURFACE_ADDRESS__GRPH_SECONDARY_SURFACE_ADDRESS_MASK);
 	WREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset, fb_format);
 	WREG32(mmGRPH_SWAP_CNTL + amdgpu_crtc->crtc_offset, fb_swap);
 
@@ -2039,14 +2108,13 @@ static void dce_v6_0_set_interleave(struct drm_crtc *crtc,
 
 	if (mode->flags & DRM_MODE_FLAG_INTERLACE)
 		WREG32(mmDATA_FORMAT + amdgpu_crtc->crtc_offset,
-		       INTERLEAVE_EN);
+			DATA_FORMAT__INTERLEAVE_EN_MASK);
 	else
 		WREG32(mmDATA_FORMAT + amdgpu_crtc->crtc_offset, 0);
 }
 
 static void dce_v6_0_crtc_load_lut(struct drm_crtc *crtc)
 {
-
 	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
 	struct drm_device *dev = crtc->dev;
 	struct amdgpu_device *adev = drm_to_adev(dev);
@@ -2056,15 +2124,15 @@ static void dce_v6_0_crtc_load_lut(struct drm_crtc *crtc)
 	DRM_DEBUG_KMS("%d\n", amdgpu_crtc->crtc_id);
 
 	WREG32(mmINPUT_CSC_CONTROL + amdgpu_crtc->crtc_offset,
-	       ((0 << INPUT_CSC_CONTROL__INPUT_CSC_GRPH_MODE__SHIFT) |
-		(0 << INPUT_CSC_CONTROL__INPUT_CSC_OVL_MODE__SHIFT)));
+	       ((INPUT_CSC_BYPASS << INPUT_CSC_CONTROL__INPUT_CSC_GRPH_MODE__SHIFT) |
+		(INPUT_CSC_BYPASS << INPUT_CSC_CONTROL__INPUT_CSC_OVL_MODE__SHIFT)));
 	WREG32(mmPRESCALE_GRPH_CONTROL + amdgpu_crtc->crtc_offset,
 	       PRESCALE_GRPH_CONTROL__GRPH_PRESCALE_BYPASS_MASK);
 	WREG32(mmPRESCALE_OVL_CONTROL + amdgpu_crtc->crtc_offset,
 	       PRESCALE_OVL_CONTROL__OVL_PRESCALE_BYPASS_MASK);
 	WREG32(mmINPUT_GAMMA_CONTROL + amdgpu_crtc->crtc_offset,
-	       ((0 << INPUT_GAMMA_CONTROL__GRPH_INPUT_GAMMA_MODE__SHIFT) |
-		(0 << INPUT_GAMMA_CONTROL__OVL_INPUT_GAMMA_MODE__SHIFT)));
+	       ((INPUT_GAMMA_USE_LUT << INPUT_GAMMA_CONTROL__GRPH_INPUT_GAMMA_MODE__SHIFT) |
+		(INPUT_GAMMA_USE_LUT << INPUT_GAMMA_CONTROL__OVL_INPUT_GAMMA_MODE__SHIFT)));
 
 	WREG32(mmDC_LUT_CONTROL + amdgpu_crtc->crtc_offset, 0);
 
@@ -2091,19 +2159,19 @@ static void dce_v6_0_crtc_load_lut(struct drm_crtc *crtc)
 	}
 
 	WREG32(mmDEGAMMA_CONTROL + amdgpu_crtc->crtc_offset,
-	       ((0 << DEGAMMA_CONTROL__GRPH_DEGAMMA_MODE__SHIFT) |
-		(0 << DEGAMMA_CONTROL__OVL_DEGAMMA_MODE__SHIFT) |
-		ICON_DEGAMMA_MODE(0) |
-		(0 << DEGAMMA_CONTROL__CURSOR_DEGAMMA_MODE__SHIFT)));
+	       ((DEGAMMA_BYPASS << DEGAMMA_CONTROL__GRPH_DEGAMMA_MODE__SHIFT) |
+		(DEGAMMA_BYPASS << DEGAMMA_CONTROL__OVL_DEGAMMA_MODE__SHIFT) |
+		(DEGAMMA_BYPASS << DEGAMMA_CONTROL__ICON_DEGAMMA_MODE__SHIFT) |
+		(DEGAMMA_BYPASS << DEGAMMA_CONTROL__CURSOR_DEGAMMA_MODE__SHIFT)));
 	WREG32(mmGAMUT_REMAP_CONTROL + amdgpu_crtc->crtc_offset,
-	       ((0 << GAMUT_REMAP_CONTROL__GRPH_GAMUT_REMAP_MODE__SHIFT) |
-		(0 << GAMUT_REMAP_CONTROL__OVL_GAMUT_REMAP_MODE__SHIFT)));
+	       ((GAMUT_REMAP_BYPASS << GAMUT_REMAP_CONTROL__GRPH_GAMUT_REMAP_MODE__SHIFT) |
+		(GAMUT_REMAP_BYPASS << GAMUT_REMAP_CONTROL__OVL_GAMUT_REMAP_MODE__SHIFT)));
 	WREG32(mmREGAMMA_CONTROL + amdgpu_crtc->crtc_offset,
-	       ((0 << REGAMMA_CONTROL__GRPH_REGAMMA_MODE__SHIFT) |
-		(0 << REGAMMA_CONTROL__OVL_REGAMMA_MODE__SHIFT)));
+	       ((REGAMMA_BYPASS << REGAMMA_CONTROL__GRPH_REGAMMA_MODE__SHIFT) |
+		(REGAMMA_BYPASS << REGAMMA_CONTROL__OVL_REGAMMA_MODE__SHIFT)));
 	WREG32(mmOUTPUT_CSC_CONTROL + amdgpu_crtc->crtc_offset,
-	       ((0 << OUTPUT_CSC_CONTROL__OUTPUT_CSC_GRPH_MODE__SHIFT) |
-		(0 << OUTPUT_CSC_CONTROL__OUTPUT_CSC_OVL_MODE__SHIFT)));
+	       ((OUTPUT_CSC_BYPASS << OUTPUT_CSC_CONTROL__OUTPUT_CSC_GRPH_MODE__SHIFT) |
+		(OUTPUT_CSC_BYPASS << OUTPUT_CSC_CONTROL__OUTPUT_CSC_OVL_MODE__SHIFT)));
 	/* XXX match this to the depth of the crtc fmt block, move to modeset? */
 	WREG32(0x1a50 + amdgpu_crtc->crtc_offset, 0);
 
@@ -2198,8 +2266,6 @@ static void dce_v6_0_hide_cursor(struct drm_crtc *crtc)
 	WREG32(mmCUR_CONTROL + amdgpu_crtc->crtc_offset,
 	       (CURSOR_24_8_PRE_MULT << CUR_CONTROL__CURSOR_MODE__SHIFT) |
 	       (CURSOR_URGENT_1_2 << CUR_CONTROL__CURSOR_URGENT_CONTROL__SHIFT));
-
-
 }
 
 static void dce_v6_0_show_cursor(struct drm_crtc *crtc)
@@ -2216,7 +2282,6 @@ static void dce_v6_0_show_cursor(struct drm_crtc *crtc)
 	       CUR_CONTROL__CURSOR_EN_MASK |
 	       (CURSOR_24_8_PRE_MULT << CUR_CONTROL__CURSOR_MODE__SHIFT) |
 	       (CURSOR_URGENT_1_2 << CUR_CONTROL__CURSOR_URGENT_CONTROL__SHIFT));
-
 }
 
 static int dce_v6_0_cursor_move_locked(struct drm_crtc *crtc,
@@ -2304,6 +2369,7 @@ static int dce_v6_0_crtc_cursor_set2(struct drm_crtc *crtc,
 		return ret;
 	}
 
+	aobj->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 	ret = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM);
 	amdgpu_bo_unreserve(aobj);
 	if (ret) {
@@ -2526,7 +2592,6 @@ static bool dce_v6_0_crtc_mode_fixup(struct drm_crtc *crtc,
 				     const struct drm_display_mode *mode,
 				     struct drm_display_mode *adjusted_mode)
 {
-
 	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
 	struct drm_device *dev = crtc->dev;
 	struct drm_encoder *encoder;
@@ -2583,6 +2648,32 @@ static const struct drm_crtc_helper_funcs dce_v6_0_crtc_helper_funcs = {
 	.get_scanout_position = amdgpu_crtc_get_scanout_position,
 };
 
+static void dce_v6_0_panic_flush(struct drm_plane *plane)
+{
+	struct drm_framebuffer *fb;
+	struct amdgpu_crtc *amdgpu_crtc;
+	struct amdgpu_device *adev;
+	uint32_t fb_format;
+
+	if (!plane->fb)
+		return;
+
+	fb = plane->fb;
+	amdgpu_crtc = to_amdgpu_crtc(plane->crtc);
+	adev = drm_to_adev(fb->dev);
+
+	/* Disable DC tiling */
+	fb_format = RREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset);
+	fb_format &= ~GRPH_CONTROL__GRPH_ARRAY_MODE_MASK;
+	WREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset, fb_format);
+
+}
+
+static const struct drm_plane_helper_funcs dce_v6_0_drm_primary_plane_helper_funcs = {
+	.get_scanout_buffer = amdgpu_display_get_scanout_buffer,
+	.panic_flush = dce_v6_0_panic_flush,
+};
+
 static int dce_v6_0_crtc_init(struct amdgpu_device *adev, int index)
 {
 	struct amdgpu_crtc *amdgpu_crtc;
@@ -2610,13 +2701,14 @@ static int dce_v6_0_crtc_init(struct amdgpu_device *adev, int index)
 	amdgpu_crtc->encoder = NULL;
 	amdgpu_crtc->connector = NULL;
 	drm_crtc_helper_add(&amdgpu_crtc->base, &dce_v6_0_crtc_helper_funcs);
+	drm_plane_helper_add(amdgpu_crtc->base.primary, &dce_v6_0_drm_primary_plane_helper_funcs);
 
 	return 0;
 }
 
-static int dce_v6_0_early_init(void *handle)
+static int dce_v6_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->audio_endpt_rreg = &dce_v6_0_audio_endpt_rreg;
 	adev->audio_endpt_wreg = &dce_v6_0_audio_endpt_wreg;
@@ -2645,11 +2737,10 @@ static int dce_v6_0_early_init(void *handle)
 	return 0;
 }
 
-static int dce_v6_0_sw_init(void *handle)
+static int dce_v6_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, i;
-	bool ret;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->mode_info.num_crtc; i++) {
 		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + 1, &adev->crtc_irq);
@@ -2692,8 +2783,7 @@ static int dce_v6_0_sw_init(void *handle)
 			return r;
 	}
 
-	ret = amdgpu_atombios_get_connector_info_from_object_table(adev);
-	if (ret)
+	if (amdgpu_atombios_get_connector_info_from_object_table(adev))
 		amdgpu_display_print_display_setup(adev_to_drm(adev));
 	else
 		return -EINVAL;
@@ -2724,11 +2814,11 @@ static int dce_v6_0_sw_init(void *handle)
 	return r;
 }
 
-static int dce_v6_0_sw_fini(void *handle)
+static int dce_v6_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	kfree(adev->mode_info.bios_hardcoded_edid);
+	drm_edid_free(adev->mode_info.bios_hardcoded_edid);
 
 	drm_kms_helper_poll_fini(adev_to_drm(adev));
 
@@ -2741,10 +2831,10 @@ static int dce_v6_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int dce_v6_0_hw_init(void *handle)
+static int dce_v6_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* disable vga render */
 	dce_v6_0_set_vga_render_state(adev, false);
@@ -2764,10 +2854,10 @@ static int dce_v6_0_hw_init(void *handle)
 	return 0;
 }
 
-static int dce_v6_0_hw_fini(void *handle)
+static int dce_v6_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	dce_v6_0_hpd_fini(adev);
 
@@ -2782,9 +2872,9 @@ static int dce_v6_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int dce_v6_0_suspend(void *handle)
+static int dce_v6_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_display_suspend_helper(adev);
@@ -2793,18 +2883,18 @@ static int dce_v6_0_suspend(void *handle)
 	adev->mode_info.bl_level =
 		amdgpu_atombios_encoder_get_backlight_level_from_reg(adev);
 
-	return dce_v6_0_hw_fini(handle);
+	return dce_v6_0_hw_fini(ip_block);
 }
 
-static int dce_v6_0_resume(void *handle)
+static int dce_v6_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	amdgpu_atombios_encoder_set_backlight_level_to_reg(adev,
 							   adev->mode_info.bl_level);
 
-	ret = dce_v6_0_hw_init(handle);
+	ret = dce_v6_0_hw_init(ip_block);
 
 	/* turn on the BL */
 	if (adev->mode_info.bl_encoder) {
@@ -2819,19 +2909,35 @@ static int dce_v6_0_resume(void *handle)
 	return amdgpu_display_resume_helper(adev);
 }
 
-static bool dce_v6_0_is_idle(void *handle)
+static bool dce_v6_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	return true;
 }
 
-static int dce_v6_0_wait_for_idle(void *handle)
+static int dce_v6_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	return 0;
-}
+	u32 srbm_soft_reset = 0, tmp;
+	struct amdgpu_device *adev = ip_block->adev;
 
-static int dce_v6_0_soft_reset(void *handle)
-{
-	DRM_INFO("xxxx: dce_v6_0_soft_reset --- no impl!!\n");
+	if (dce_v6_0_is_display_hung(adev))
+		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_DC_MASK;
+
+	if (srbm_soft_reset) {
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+		tmp |= srbm_soft_reset;
+		dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32(mmSRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+
+		udelay(50);
+
+		tmp &= ~srbm_soft_reset;
+		WREG32(mmSRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+
+		/* Wait a little for things to settle down */
+		udelay(50);
+	}
 	return 0;
 }
 
@@ -2848,22 +2954,22 @@ static void dce_v6_0_set_crtc_vblank_interrupt_state(struct amdgpu_device *adev,
 
 	switch (crtc) {
 	case 0:
-		reg_block = SI_CRTC0_REGISTER_OFFSET;
+		reg_block = CRTC0_REGISTER_OFFSET;
 		break;
 	case 1:
-		reg_block = SI_CRTC1_REGISTER_OFFSET;
+		reg_block = CRTC1_REGISTER_OFFSET;
 		break;
 	case 2:
-		reg_block = SI_CRTC2_REGISTER_OFFSET;
+		reg_block = CRTC2_REGISTER_OFFSET;
 		break;
 	case 3:
-		reg_block = SI_CRTC3_REGISTER_OFFSET;
+		reg_block = CRTC3_REGISTER_OFFSET;
 		break;
 	case 4:
-		reg_block = SI_CRTC4_REGISTER_OFFSET;
+		reg_block = CRTC4_REGISTER_OFFSET;
 		break;
 	case 5:
-		reg_block = SI_CRTC5_REGISTER_OFFSET;
+		reg_block = CRTC5_REGISTER_OFFSET;
 		break;
 	default:
 		DRM_DEBUG("invalid crtc %d\n", crtc);
@@ -2873,12 +2979,12 @@ static void dce_v6_0_set_crtc_vblank_interrupt_state(struct amdgpu_device *adev,
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
 		interrupt_mask = RREG32(mmINT_MASK + reg_block);
-		interrupt_mask &= ~VBLANK_INT_MASK;
+		interrupt_mask &= ~INT_MASK__VBLANK_INT_MASK;
 		WREG32(mmINT_MASK + reg_block, interrupt_mask);
 		break;
 	case AMDGPU_IRQ_STATE_ENABLE:
 		interrupt_mask = RREG32(mmINT_MASK + reg_block);
-		interrupt_mask |= VBLANK_INT_MASK;
+		interrupt_mask |= INT_MASK__VBLANK_INT_MASK;
 		WREG32(mmINT_MASK + reg_block, interrupt_mask);
 		break;
 	default:
@@ -2893,28 +2999,28 @@ static void dce_v6_0_set_crtc_vline_interrupt_state(struct amdgpu_device *adev,
 
 }
 
-static int dce_v6_0_set_hpd_interrupt_state(struct amdgpu_device *adev,
+static int dce_v6_0_set_hpd_irq_state(struct amdgpu_device *adev,
 					    struct amdgpu_irq_src *src,
-					    unsigned type,
+					    unsigned hpd,
 					    enum amdgpu_interrupt_state state)
 {
 	u32 dc_hpd_int_cntl;
 
-	if (type >= adev->mode_info.num_hpd) {
-		DRM_DEBUG("invalid hdp %d\n", type);
+	if (hpd >= adev->mode_info.num_hpd) {
+		DRM_DEBUG("invalid hpd %d\n", hpd);
 		return 0;
 	}
 
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
-		dc_hpd_int_cntl = RREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[type]);
-		dc_hpd_int_cntl &= ~DC_HPDx_INT_EN;
-		WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[type], dc_hpd_int_cntl);
+		dc_hpd_int_cntl = RREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd]);
+		dc_hpd_int_cntl &= ~DC_HPD1_INT_CONTROL__DC_HPD1_INT_EN_MASK;
+		WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd], dc_hpd_int_cntl);
 		break;
 	case AMDGPU_IRQ_STATE_ENABLE:
-		dc_hpd_int_cntl = RREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[type]);
-		dc_hpd_int_cntl |= DC_HPDx_INT_EN;
-		WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[type], dc_hpd_int_cntl);
+		dc_hpd_int_cntl = RREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd]);
+		dc_hpd_int_cntl |= DC_HPD1_INT_CONTROL__DC_HPD1_INT_EN_MASK;
+		WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd], dc_hpd_int_cntl);
 		break;
 	default:
 		break;
@@ -2923,7 +3029,7 @@ static int dce_v6_0_set_hpd_interrupt_state(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int dce_v6_0_set_crtc_interrupt_state(struct amdgpu_device *adev,
+static int dce_v6_0_set_crtc_irq_state(struct amdgpu_device *adev,
 					     struct amdgpu_irq_src *src,
 					     unsigned type,
 					     enum amdgpu_interrupt_state state)
@@ -2983,7 +3089,7 @@ static int dce_v6_0_crtc_irq(struct amdgpu_device *adev,
 	switch (entry->src_data[0]) {
 	case 0: /* vblank */
 		if (disp_int & interrupt_status_offsets[crtc].vblank)
-			WREG32(mmVBLANK_STATUS + crtc_offsets[crtc], VBLANK_ACK);
+			WREG32(mmVBLANK_STATUS + crtc_offsets[crtc], VBLANK_STATUS__VBLANK_ACK_MASK);
 		else
 			DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
 
@@ -2994,7 +3100,7 @@ static int dce_v6_0_crtc_irq(struct amdgpu_device *adev,
 		break;
 	case 1: /* vline */
 		if (disp_int & interrupt_status_offsets[crtc].vline)
-			WREG32(mmVLINE_STATUS + crtc_offsets[crtc], VLINE_ACK);
+			WREG32(mmVLINE_STATUS + crtc_offsets[crtc], VLINE_STATUS__VLINE_ACK_MASK);
 		else
 			DRM_DEBUG("IH: IH event w/o asserted irq bit?\n");
 
@@ -3008,7 +3114,7 @@ static int dce_v6_0_crtc_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int dce_v6_0_set_pageflip_interrupt_state(struct amdgpu_device *adev,
+static int dce_v6_0_set_pageflip_irq_state(struct amdgpu_device *adev,
 						 struct amdgpu_irq_src *src,
 						 unsigned type,
 						 enum amdgpu_interrupt_state state)
@@ -3059,7 +3165,7 @@ static int dce_v6_0_pageflip_irq(struct amdgpu_device *adev,
 
 	spin_lock_irqsave(&adev_to_drm(adev)->event_lock, flags);
 	works = amdgpu_crtc->pflip_works;
-	if (amdgpu_crtc->pflip_status != AMDGPU_FLIP_SUBMITTED){
+	if (amdgpu_crtc->pflip_status != AMDGPU_FLIP_SUBMITTED) {
 		DRM_DEBUG_DRIVER("amdgpu_crtc->pflip_status = %d != "
 						"AMDGPU_FLIP_SUBMITTED(%d)\n",
 						amdgpu_crtc->pflip_status,
@@ -3088,7 +3194,7 @@ static int dce_v6_0_hpd_irq(struct amdgpu_device *adev,
 			    struct amdgpu_irq_src *source,
 			    struct amdgpu_iv_entry *entry)
 {
-	uint32_t disp_int, mask, tmp;
+	uint32_t disp_int, mask;
 	unsigned hpd;
 
 	if (entry->src_data[0] >= adev->mode_info.num_hpd) {
@@ -3101,24 +3207,21 @@ static int dce_v6_0_hpd_irq(struct amdgpu_device *adev,
 	mask = interrupt_status_offsets[hpd].hpd;
 
 	if (disp_int & mask) {
-		tmp = RREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd]);
-		tmp |= DC_HPD1_INT_CONTROL__DC_HPD1_INT_ACK_MASK;
-		WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd], tmp);
+		dce_v6_0_hpd_int_ack(adev, hpd);
 		schedule_delayed_work(&adev->hotplug_work, 0);
 		DRM_DEBUG("IH: HPD%d\n", hpd + 1);
 	}
 
 	return 0;
-
 }
 
-static int dce_v6_0_set_clockgating_state(void *handle,
+static int dce_v6_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int dce_v6_0_set_powergating_state(void *handle,
+static int dce_v6_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -3127,7 +3230,6 @@ static int dce_v6_0_set_powergating_state(void *handle,
 static const struct amd_ip_funcs dce_v6_0_ip_funcs = {
 	.name = "dce_v6_0",
 	.early_init = dce_v6_0_early_init,
-	.late_init = NULL,
 	.sw_init = dce_v6_0_sw_init,
 	.sw_fini = dce_v6_0_sw_fini,
 	.hw_init = dce_v6_0_hw_init,
@@ -3135,18 +3237,15 @@ static const struct amd_ip_funcs dce_v6_0_ip_funcs = {
 	.suspend = dce_v6_0_suspend,
 	.resume = dce_v6_0_resume,
 	.is_idle = dce_v6_0_is_idle,
-	.wait_for_idle = dce_v6_0_wait_for_idle,
 	.soft_reset = dce_v6_0_soft_reset,
 	.set_clockgating_state = dce_v6_0_set_clockgating_state,
 	.set_powergating_state = dce_v6_0_set_powergating_state,
 };
 
-static void
-dce_v6_0_encoder_mode_set(struct drm_encoder *encoder,
+static void dce_v6_0_encoder_mode_set(struct drm_encoder *encoder,
 			  struct drm_display_mode *mode,
 			  struct drm_display_mode *adjusted_mode)
 {
-
 	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
 	int em = amdgpu_atombios_encoder_get_encoder_mode(encoder);
 
@@ -3166,7 +3265,6 @@ dce_v6_0_encoder_mode_set(struct drm_encoder *encoder,
 
 static void dce_v6_0_encoder_prepare(struct drm_encoder *encoder)
 {
-
 	struct amdgpu_device *adev = drm_to_adev(encoder->dev);
 	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
 	struct drm_connector *connector = amdgpu_get_connector_for_encoder(encoder);
@@ -3206,7 +3304,6 @@ static void dce_v6_0_encoder_prepare(struct drm_encoder *encoder)
 
 static void dce_v6_0_encoder_commit(struct drm_encoder *encoder)
 {
-
 	struct drm_device *dev = encoder->dev;
 	struct amdgpu_device *adev = drm_to_adev(dev);
 
@@ -3217,7 +3314,6 @@ static void dce_v6_0_encoder_commit(struct drm_encoder *encoder)
 
 static void dce_v6_0_encoder_disable(struct drm_encoder *encoder)
 {
-
 	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
 	struct amdgpu_encoder_atom_dig *dig;
 	int em = amdgpu_atombios_encoder_get_encoder_mode(encoder);
@@ -3244,8 +3340,7 @@ static void dce_v6_0_ext_commit(struct drm_encoder *encoder)
 
 }
 
-static void
-dce_v6_0_ext_mode_set(struct drm_encoder *encoder,
+static void dce_v6_0_ext_mode_set(struct drm_encoder *encoder,
 		      struct drm_display_mode *mode,
 		      struct drm_display_mode *adjusted_mode)
 {
@@ -3257,8 +3352,7 @@ static void dce_v6_0_ext_disable(struct drm_encoder *encoder)
 
 }
 
-static void
-dce_v6_0_ext_dpms(struct drm_encoder *encoder, int mode)
+static void dce_v6_0_ext_dpms(struct drm_encoder *encoder, int mode)
 {
 
 }
@@ -3329,7 +3423,6 @@ static void dce_v6_0_encoder_add(struct amdgpu_device *adev,
 			amdgpu_encoder->devices |= supported_device;
 			return;
 		}
-
 	}
 
 	/* add a new one */
@@ -3436,17 +3529,17 @@ static void dce_v6_0_set_display_funcs(struct amdgpu_device *adev)
 }
 
 static const struct amdgpu_irq_src_funcs dce_v6_0_crtc_irq_funcs = {
-	.set = dce_v6_0_set_crtc_interrupt_state,
+	.set = dce_v6_0_set_crtc_irq_state,
 	.process = dce_v6_0_crtc_irq,
 };
 
 static const struct amdgpu_irq_src_funcs dce_v6_0_pageflip_irq_funcs = {
-	.set = dce_v6_0_set_pageflip_interrupt_state,
+	.set = dce_v6_0_set_pageflip_irq_state,
 	.process = dce_v6_0_pageflip_irq,
 };
 
 static const struct amdgpu_irq_src_funcs dce_v6_0_hpd_irq_funcs = {
-	.set = dce_v6_0_set_hpd_interrupt_state,
+	.set = dce_v6_0_set_hpd_irq_state,
 	.process = dce_v6_0_hpd_irq,
 };
 
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c
index d421a268c9ff..2ccd6aad8dd6 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c
@@ -21,6 +21,7 @@
  *
  */
 
+#include <drm/drm_edid.h>
 #include <drm/drm_fourcc.h>
 #include <drm/drm_modeset_helper.h>
 #include <drm/drm_modeset_helper_vtables.h>
@@ -53,8 +54,7 @@
 static void dce_v8_0_set_display_funcs(struct amdgpu_device *adev);
 static void dce_v8_0_set_irq_funcs(struct amdgpu_device *adev);
 
-static const u32 crtc_offsets[6] =
-{
+static const u32 crtc_offsets[6] = {
 	CRTC0_REGISTER_OFFSET,
 	CRTC1_REGISTER_OFFSET,
 	CRTC2_REGISTER_OFFSET,
@@ -63,8 +63,7 @@ static const u32 crtc_offsets[6] =
 	CRTC5_REGISTER_OFFSET
 };
 
-static const u32 hpd_offsets[] =
-{
+static const u32 hpd_offsets[] = {
 	HPD0_REGISTER_OFFSET,
 	HPD1_REGISTER_OFFSET,
 	HPD2_REGISTER_OFFSET,
@@ -266,6 +265,21 @@ static void dce_v8_0_hpd_set_polarity(struct amdgpu_device *adev,
 	WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd], tmp);
 }
 
+static void dce_v8_0_hpd_int_ack(struct amdgpu_device *adev,
+				 int hpd)
+{
+	u32 tmp;
+
+	if (hpd >= adev->mode_info.num_hpd) {
+		DRM_DEBUG("invalid hpd %d\n", hpd);
+		return;
+	}
+
+	tmp = RREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd]);
+	tmp |= DC_HPD1_INT_CONTROL__DC_HPD1_INT_ACK_MASK;
+	WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd], tmp);
+}
+
 /**
  * dce_v8_0_hpd_init - hpd setup callback.
  *
@@ -305,6 +319,7 @@ static void dce_v8_0_hpd_init(struct amdgpu_device *adev)
 			continue;
 		}
 
+		dce_v8_0_hpd_int_ack(adev, amdgpu_connector->hpd.hpd);
 		dce_v8_0_hpd_set_polarity(adev, amdgpu_connector->hpd.hpd);
 		amdgpu_irq_get(adev, &adev->hpd_irq, amdgpu_connector->hpd.hpd);
 	}
@@ -977,7 +992,7 @@ static void dce_v8_0_program_watermarks(struct amdgpu_device *adev,
 					    (u32)mode->clock);
 		line_time = (u32) div_u64((u64)mode->crtc_htotal * 1000000,
 					  (u32)mode->clock);
-		line_time = min(line_time, (u32)65535);
+		line_time = min_t(u32, line_time, 65535);
 
 		/* watermark for high clocks */
 		if (adev->pm.dpm_enabled) {
@@ -1007,7 +1022,7 @@ static void dce_v8_0_program_watermarks(struct amdgpu_device *adev,
 		wm_high.num_heads = num_heads;
 
 		/* set for high clocks */
-		latency_watermark_a = min(dce_v8_0_latency_watermark(&wm_high), (u32)65535);
+		latency_watermark_a = min_t(u32, dce_v8_0_latency_watermark(&wm_high), 65535);
 
 		/* possibly force display priority to high */
 		/* should really do this at mode validation time... */
@@ -1046,7 +1061,7 @@ static void dce_v8_0_program_watermarks(struct amdgpu_device *adev,
 		wm_low.num_heads = num_heads;
 
 		/* set for low clocks */
-		latency_watermark_b = min(dce_v8_0_latency_watermark(&wm_low), (u32)65535);
+		latency_watermark_b = min_t(u32, dce_v8_0_latency_watermark(&wm_low), 65535);
 
 		/* possibly force display priority to high */
 		/* should really do this at mode validation time... */
@@ -1081,8 +1096,7 @@ static void dce_v8_0_program_watermarks(struct amdgpu_device *adev,
 
 	/* save values for DPM */
 	amdgpu_crtc->line_time = line_time;
-	amdgpu_crtc->wm_high = latency_watermark_a;
-	amdgpu_crtc->wm_low = latency_watermark_b;
+
 	/* Save number of lines the linebuffer leads before the scanout */
 	amdgpu_crtc->lb_vblank_lead_lines = lb_vblank_lead_lines;
 }
@@ -1257,7 +1271,7 @@ static void dce_v8_0_audio_write_speaker_allocation(struct drm_encoder *encoder)
 		return;
 	}
 
-	sad_count = drm_edid_to_speaker_allocation(amdgpu_connector_edid(connector), &sadb);
+	sad_count = drm_edid_to_speaker_allocation(amdgpu_connector->edid, &sadb);
 	if (sad_count < 0) {
 		DRM_ERROR("Couldn't read Speaker Allocation Data Block: %d\n", sad_count);
 		sad_count = 0;
@@ -1325,7 +1339,7 @@ static void dce_v8_0_audio_write_sad_regs(struct drm_encoder *encoder)
 		return;
 	}
 
-	sad_count = drm_edid_to_sad(amdgpu_connector_edid(connector), &sads);
+	sad_count = drm_edid_to_sad(amdgpu_connector->edid, &sads);
 	if (sad_count < 0)
 		DRM_ERROR("Couldn't read SADs: %d\n", sad_count);
 	if (sad_count <= 0)
@@ -1345,9 +1359,9 @@ static void dce_v8_0_audio_write_sad_regs(struct drm_encoder *encoder)
 				if (sad->channels > max_channels) {
 					value = (sad->channels <<
 						 AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__MAX_CHANNELS__SHIFT) |
-					        (sad->byte2 <<
+						(sad->byte2 <<
 						 AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__DESCRIPTOR_BYTE_2__SHIFT) |
-					        (sad->freq <<
+						(sad->freq <<
 						 AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__SUPPORTED_FREQUENCIES__SHIFT);
 					max_channels = sad->channels;
 				}
@@ -1379,15 +1393,14 @@ static void dce_v8_0_audio_enable(struct amdgpu_device *adev,
 		enable ? AZALIA_F0_CODEC_PIN_CONTROL_HOT_PLUG_CONTROL__AUDIO_ENABLED_MASK : 0);
 }
 
-static const u32 pin_offsets[7] =
-{
-	(0x1780 - 0x1780),
-	(0x1786 - 0x1780),
-	(0x178c - 0x1780),
-	(0x1792 - 0x1780),
-	(0x1798 - 0x1780),
-	(0x179d - 0x1780),
-	(0x17a4 - 0x1780),
+static const u32 pin_offsets[7] = {
+	AUD0_REGISTER_OFFSET,
+	AUD1_REGISTER_OFFSET,
+	AUD2_REGISTER_OFFSET,
+	AUD3_REGISTER_OFFSET,
+	AUD4_REGISTER_OFFSET,
+	AUD5_REGISTER_OFFSET,
+	AUD6_REGISTER_OFFSET,
 };
 
 static int dce_v8_0_audio_init(struct amdgpu_device *adev)
@@ -1429,17 +1442,12 @@ static int dce_v8_0_audio_init(struct amdgpu_device *adev)
 
 static void dce_v8_0_audio_fini(struct amdgpu_device *adev)
 {
-	int i;
-
 	if (!amdgpu_audio)
 		return;
 
 	if (!adev->mode_info.audio.enabled)
 		return;
 
-	for (i = 0; i < adev->mode_info.audio.num_pins; i++)
-		dce_v8_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
-
 	adev->mode_info.audio.enabled = false;
 }
 
@@ -1740,8 +1748,7 @@ static void dce_v8_0_afmt_fini(struct amdgpu_device *adev)
 	}
 }
 
-static const u32 vga_control_regs[6] =
-{
+static const u32 vga_control_regs[6] = {
 	mmD1VGA_CONTROL,
 	mmD2VGA_CONTROL,
 	mmD3VGA_CONTROL,
@@ -1815,6 +1822,7 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc,
 		return r;
 
 	if (!atomic) {
+		abo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 		r = amdgpu_bo_pin(abo, AMDGPU_GEM_DOMAIN_VRAM);
 		if (unlikely(r != 0)) {
 			amdgpu_bo_unreserve(abo);
@@ -1895,9 +1903,9 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc,
 	case DRM_FORMAT_XBGR8888:
 	case DRM_FORMAT_ABGR8888:
 		fb_format = ((GRPH_DEPTH_32BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
-		             (GRPH_FORMAT_ARGB8888 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
+				(GRPH_FORMAT_ARGB8888 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
 		fb_swap = ((GRPH_RED_SEL_B << GRPH_SWAP_CNTL__GRPH_RED_CROSSBAR__SHIFT) |
-		           (GRPH_BLUE_SEL_R << GRPH_SWAP_CNTL__GRPH_BLUE_CROSSBAR__SHIFT));
+			(GRPH_BLUE_SEL_R << GRPH_SWAP_CNTL__GRPH_BLUE_CROSSBAR__SHIFT));
 #ifdef __BIG_ENDIAN
 		fb_swap |= (GRPH_ENDIAN_8IN32 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
 #endif
@@ -2307,6 +2315,7 @@ static int dce_v8_0_crtc_cursor_set2(struct drm_crtc *crtc,
 		return ret;
 	}
 
+	aobj->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS;
 	ret = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM);
 	amdgpu_bo_unreserve(aobj);
 	if (ret) {
@@ -2598,6 +2607,31 @@ static const struct drm_crtc_helper_funcs dce_v8_0_crtc_helper_funcs = {
 	.get_scanout_position = amdgpu_crtc_get_scanout_position,
 };
 
+static void dce_v8_0_panic_flush(struct drm_plane *plane)
+{
+	struct drm_framebuffer *fb;
+	struct amdgpu_crtc *amdgpu_crtc;
+	struct amdgpu_device *adev;
+	uint32_t fb_format;
+
+	if (!plane->fb)
+		return;
+
+	fb = plane->fb;
+	amdgpu_crtc = to_amdgpu_crtc(plane->crtc);
+	adev = drm_to_adev(fb->dev);
+
+	/* Disable DC tiling */
+	fb_format = RREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset);
+	fb_format &= ~GRPH_CONTROL__GRPH_ARRAY_MODE_MASK;
+	WREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset, fb_format);
+}
+
+static const struct drm_plane_helper_funcs dce_v8_0_drm_primary_plane_helper_funcs = {
+	.get_scanout_buffer = amdgpu_display_get_scanout_buffer,
+	.panic_flush = dce_v8_0_panic_flush,
+};
+
 static int dce_v8_0_crtc_init(struct amdgpu_device *adev, int index)
 {
 	struct amdgpu_crtc *amdgpu_crtc;
@@ -2625,13 +2659,14 @@ static int dce_v8_0_crtc_init(struct amdgpu_device *adev, int index)
 	amdgpu_crtc->encoder = NULL;
 	amdgpu_crtc->connector = NULL;
 	drm_crtc_helper_add(&amdgpu_crtc->base, &dce_v8_0_crtc_helper_funcs);
+	drm_plane_helper_add(amdgpu_crtc->base.primary, &dce_v8_0_drm_primary_plane_helper_funcs);
 
 	return 0;
 }
 
-static int dce_v8_0_early_init(void *handle)
+static int dce_v8_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->audio_endpt_rreg = &dce_v8_0_audio_endpt_rreg;
 	adev->audio_endpt_wreg = &dce_v8_0_audio_endpt_wreg;
@@ -2665,10 +2700,10 @@ static int dce_v8_0_early_init(void *handle)
 	return 0;
 }
 
-static int dce_v8_0_sw_init(void *handle)
+static int dce_v8_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->mode_info.num_crtc; i++) {
 		r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + 1, &adev->crtc_irq);
@@ -2749,11 +2784,11 @@ static int dce_v8_0_sw_init(void *handle)
 	return 0;
 }
 
-static int dce_v8_0_sw_fini(void *handle)
+static int dce_v8_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	kfree(adev->mode_info.bios_hardcoded_edid);
+	drm_edid_free(adev->mode_info.bios_hardcoded_edid);
 
 	drm_kms_helper_poll_fini(adev_to_drm(adev));
 
@@ -2767,10 +2802,10 @@ static int dce_v8_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int dce_v8_0_hw_init(void *handle)
+static int dce_v8_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* disable vga render */
 	dce_v8_0_set_vga_render_state(adev, false);
@@ -2790,10 +2825,10 @@ static int dce_v8_0_hw_init(void *handle)
 	return 0;
 }
 
-static int dce_v8_0_hw_fini(void *handle)
+static int dce_v8_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	dce_v8_0_hpd_fini(adev);
 
@@ -2808,9 +2843,9 @@ static int dce_v8_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int dce_v8_0_suspend(void *handle)
+static int dce_v8_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_display_suspend_helper(adev);
@@ -2820,18 +2855,18 @@ static int dce_v8_0_suspend(void *handle)
 	adev->mode_info.bl_level =
 		amdgpu_atombios_encoder_get_backlight_level_from_reg(adev);
 
-	return dce_v8_0_hw_fini(handle);
+	return dce_v8_0_hw_fini(ip_block);
 }
 
-static int dce_v8_0_resume(void *handle)
+static int dce_v8_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	amdgpu_atombios_encoder_set_backlight_level_to_reg(adev,
 							   adev->mode_info.bl_level);
 
-	ret = dce_v8_0_hw_init(handle);
+	ret = dce_v8_0_hw_init(ip_block);
 
 	/* turn on the BL */
 	if (adev->mode_info.bl_encoder) {
@@ -2846,20 +2881,15 @@ static int dce_v8_0_resume(void *handle)
 	return amdgpu_display_resume_helper(adev);
 }
 
-static bool dce_v8_0_is_idle(void *handle)
+static bool dce_v8_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	return true;
 }
 
-static int dce_v8_0_wait_for_idle(void *handle)
-{
-	return 0;
-}
-
-static int dce_v8_0_soft_reset(void *handle)
+static int dce_v8_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 srbm_soft_reset = 0, tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (dce_v8_0_is_display_hung(adev))
 		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_DC_MASK;
@@ -2985,7 +3015,7 @@ static void dce_v8_0_set_crtc_vline_interrupt_state(struct amdgpu_device *adev,
 	}
 }
 
-static int dce_v8_0_set_hpd_interrupt_state(struct amdgpu_device *adev,
+static int dce_v8_0_set_hpd_irq_state(struct amdgpu_device *adev,
 					    struct amdgpu_irq_src *src,
 					    unsigned type,
 					    enum amdgpu_interrupt_state state)
@@ -2993,7 +3023,7 @@ static int dce_v8_0_set_hpd_interrupt_state(struct amdgpu_device *adev,
 	u32 dc_hpd_int_cntl;
 
 	if (type >= adev->mode_info.num_hpd) {
-		DRM_DEBUG("invalid hdp %d\n", type);
+		DRM_DEBUG("invalid hpd %d\n", type);
 		return 0;
 	}
 
@@ -3015,7 +3045,7 @@ static int dce_v8_0_set_hpd_interrupt_state(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int dce_v8_0_set_crtc_interrupt_state(struct amdgpu_device *adev,
+static int dce_v8_0_set_crtc_irq_state(struct amdgpu_device *adev,
 					     struct amdgpu_irq_src *src,
 					     unsigned type,
 					     enum amdgpu_interrupt_state state)
@@ -3100,7 +3130,7 @@ static int dce_v8_0_crtc_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int dce_v8_0_set_pageflip_interrupt_state(struct amdgpu_device *adev,
+static int dce_v8_0_set_pageflip_irq_state(struct amdgpu_device *adev,
 						 struct amdgpu_irq_src *src,
 						 unsigned type,
 						 enum amdgpu_interrupt_state state)
@@ -3151,7 +3181,7 @@ static int dce_v8_0_pageflip_irq(struct amdgpu_device *adev,
 
 	spin_lock_irqsave(&adev_to_drm(adev)->event_lock, flags);
 	works = amdgpu_crtc->pflip_works;
-	if (amdgpu_crtc->pflip_status != AMDGPU_FLIP_SUBMITTED){
+	if (amdgpu_crtc->pflip_status != AMDGPU_FLIP_SUBMITTED) {
 		DRM_DEBUG_DRIVER("amdgpu_crtc->pflip_status = %d != "
 						"AMDGPU_FLIP_SUBMITTED(%d)\n",
 						amdgpu_crtc->pflip_status,
@@ -3180,7 +3210,7 @@ static int dce_v8_0_hpd_irq(struct amdgpu_device *adev,
 			    struct amdgpu_irq_src *source,
 			    struct amdgpu_iv_entry *entry)
 {
-	uint32_t disp_int, mask, tmp;
+	uint32_t disp_int, mask;
 	unsigned hpd;
 
 	if (entry->src_data[0] >= adev->mode_info.num_hpd) {
@@ -3193,9 +3223,7 @@ static int dce_v8_0_hpd_irq(struct amdgpu_device *adev,
 	mask = interrupt_status_offsets[hpd].hpd;
 
 	if (disp_int & mask) {
-		tmp = RREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd]);
-		tmp |= DC_HPD1_INT_CONTROL__DC_HPD1_INT_ACK_MASK;
-		WREG32(mmDC_HPD1_INT_CONTROL + hpd_offsets[hpd], tmp);
+		dce_v8_0_hpd_int_ack(adev, hpd);
 		schedule_delayed_work(&adev->hotplug_work, 0);
 		DRM_DEBUG("IH: HPD%d\n", hpd + 1);
 	}
@@ -3204,13 +3232,13 @@ static int dce_v8_0_hpd_irq(struct amdgpu_device *adev,
 
 }
 
-static int dce_v8_0_set_clockgating_state(void *handle,
+static int dce_v8_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int dce_v8_0_set_powergating_state(void *handle,
+static int dce_v8_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -3219,7 +3247,6 @@ static int dce_v8_0_set_powergating_state(void *handle,
 static const struct amd_ip_funcs dce_v8_0_ip_funcs = {
 	.name = "dce_v8_0",
 	.early_init = dce_v8_0_early_init,
-	.late_init = NULL,
 	.sw_init = dce_v8_0_sw_init,
 	.sw_fini = dce_v8_0_sw_fini,
 	.hw_init = dce_v8_0_hw_init,
@@ -3227,7 +3254,6 @@ static const struct amd_ip_funcs dce_v8_0_ip_funcs = {
 	.suspend = dce_v8_0_suspend,
 	.resume = dce_v8_0_resume,
 	.is_idle = dce_v8_0_is_idle,
-	.wait_for_idle = dce_v8_0_wait_for_idle,
 	.soft_reset = dce_v8_0_soft_reset,
 	.set_clockgating_state = dce_v8_0_set_clockgating_state,
 	.set_powergating_state = dce_v8_0_set_powergating_state,
@@ -3515,17 +3541,17 @@ static void dce_v8_0_set_display_funcs(struct amdgpu_device *adev)
 }
 
 static const struct amdgpu_irq_src_funcs dce_v8_0_crtc_irq_funcs = {
-	.set = dce_v8_0_set_crtc_interrupt_state,
+	.set = dce_v8_0_set_crtc_irq_state,
 	.process = dce_v8_0_crtc_irq,
 };
 
 static const struct amdgpu_irq_src_funcs dce_v8_0_pageflip_irq_funcs = {
-	.set = dce_v8_0_set_pageflip_interrupt_state,
+	.set = dce_v8_0_set_pageflip_irq_state,
 	.process = dce_v8_0_pageflip_irq,
 };
 
 static const struct amdgpu_irq_src_funcs dce_v8_0_hpd_irq_funcs = {
-	.set = dce_v8_0_set_hpd_interrupt_state,
+	.set = dce_v8_0_set_hpd_irq_state,
 	.process = dce_v8_0_hpd_irq,
 };
 
@@ -3544,8 +3570,7 @@ static void dce_v8_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->hpd_irq.funcs = &dce_v8_0_hpd_irq_funcs;
 }
 
-const struct amdgpu_ip_block_version dce_v8_0_ip_block =
-{
+const struct amdgpu_ip_block_version dce_v8_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 8,
 	.minor = 0,
@@ -3553,8 +3578,7 @@ const struct amdgpu_ip_block_version dce_v8_0_ip_block =
 	.funcs = &dce_v8_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version dce_v8_1_ip_block =
-{
+const struct amdgpu_ip_block_version dce_v8_1_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 8,
 	.minor = 1,
@@ -3562,8 +3586,7 @@ const struct amdgpu_ip_block_version dce_v8_1_ip_block =
 	.funcs = &dce_v8_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version dce_v8_2_ip_block =
-{
+const struct amdgpu_ip_block_version dce_v8_2_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 8,
 	.minor = 2,
@@ -3571,8 +3594,7 @@ const struct amdgpu_ip_block_version dce_v8_2_ip_block =
 	.funcs = &dce_v8_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version dce_v8_3_ip_block =
-{
+const struct amdgpu_ip_block_version dce_v8_3_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 8,
 	.minor = 3,
@@ -3580,8 +3602,7 @@ const struct amdgpu_ip_block_version dce_v8_3_ip_block =
 	.funcs = &dce_v8_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version dce_v8_5_ip_block =
-{
+const struct amdgpu_ip_block_version dce_v8_5_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_DCE,
 	.major = 8,
 	.minor = 5,
diff --git a/drivers/gpu/drm/amd/amdgpu/df_v1_7.c b/drivers/gpu/drm/amd/amdgpu/df_v1_7.c
index 5dfab80ffff2..cd298556f7a6 100644
--- a/drivers/gpu/drm/amd/amdgpu/df_v1_7.c
+++ b/drivers/gpu/drm/amd/amdgpu/df_v1_7.c
@@ -70,6 +70,8 @@ static u32 df_v1_7_get_hbm_channel_number(struct amdgpu_device *adev)
 	int fb_channel_number;
 
 	fb_channel_number = adev->df.funcs->get_fb_channel_number(adev);
+	if (fb_channel_number >= ARRAY_SIZE(df_v1_7_channel_number))
+		fb_channel_number = 0;
 
 	return df_v1_7_channel_number[fb_channel_number];
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/df_v3_6.c b/drivers/gpu/drm/amd/amdgpu/df_v3_6.c
index 483a441b46aa..621aeca53880 100644
--- a/drivers/gpu/drm/amd/amdgpu/df_v3_6.c
+++ b/drivers/gpu/drm/amd/amdgpu/df_v3_6.c
@@ -254,8 +254,8 @@ static void df_v3_6_sw_init(struct amdgpu_device *adev)
 
 static void df_v3_6_sw_fini(struct amdgpu_device *adev)
 {
-
-	device_remove_file(adev->dev, &dev_attr_df_cntr_avail);
+	if (adev->dev->kobj.sd)
+		device_remove_file(adev->dev, &dev_attr_df_cntr_avail);
 
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/df_v4_15.c b/drivers/gpu/drm/amd/amdgpu/df_v4_15.c
new file mode 100644
index 000000000000..2a573e33908b
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/df_v4_15.c
@@ -0,0 +1,45 @@
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "df_v4_15.h"
+
+#include "df/df_4_15_offset.h"
+#include "df/df_4_15_sh_mask.h"
+
+static void df_v4_15_hw_init(struct amdgpu_device *adev)
+{
+	if (adev->have_atomics_support) {
+		uint32_t tmp;
+		uint32_t dis_lcl_proc = (1 <<  1 |
+					1 <<  2 |
+					1 << 13);
+
+		tmp = RREG32_SOC15(DF, 0, regNCSConfigurationRegister1);
+		tmp |= (dis_lcl_proc << NCSConfigurationRegister1__DisIntAtomicsLclProcessing__SHIFT);
+		WREG32_SOC15(DF, 0, regNCSConfigurationRegister1, tmp);
+	}
+}
+
+const struct amdgpu_df_funcs df_v4_15_funcs = {
+	.hw_init = df_v4_15_hw_init
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/df_v4_15.h b/drivers/gpu/drm/amd/amdgpu/df_v4_15.h
new file mode 100644
index 000000000000..dddf2422112a
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/df_v4_15.h
@@ -0,0 +1,30 @@
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __DF_V4_15_H__
+#define __DF_V4_15_H__
+
+extern const struct amdgpu_df_funcs df_v4_15_funcs;
+
+#endif /* __DF_V4_15_H__ */
+
diff --git a/drivers/gpu/drm/amd/amdgpu/df_v4_6_2.c b/drivers/gpu/drm/amd/amdgpu/df_v4_6_2.c
new file mode 100644
index 000000000000..a47960a0babd
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/df_v4_6_2.c
@@ -0,0 +1,34 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "df_v4_6_2.h"
+
+static bool df_v4_6_2_query_ras_poison_mode(struct amdgpu_device *adev)
+{
+	/* return true since related regs are inaccessible */
+	return true;
+}
+
+const struct amdgpu_df_funcs df_v4_6_2_funcs = {
+	.query_ras_poison_mode = df_v4_6_2_query_ras_poison_mode,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.h b/drivers/gpu/drm/amd/amdgpu/df_v4_6_2.h
index 0d878ca3acba..3bc3e6d216e2 100644
--- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.h
+++ b/drivers/gpu/drm/amd/amdgpu/df_v4_6_2.h
@@ -1,5 +1,5 @@
 /*
- * Copyright 2014 Advanced Micro Devices, Inc.
+ * Copyright 2023 Advanced Micro Devices, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -21,12 +21,11 @@
  *
  */
 
-#ifndef __DCE_V11_0_H__
-#define __DCE_V11_0_H__
+#ifndef __DF_V4_6_2_H__
+#define __DF_V4_6_2_H__
 
-extern const struct amdgpu_ip_block_version dce_v11_0_ip_block;
-extern const struct amdgpu_ip_block_version dce_v11_2_ip_block;
+#include "soc15_common.h"
 
-void dce_v11_0_disable_dce(struct amdgpu_device *adev);
+extern const struct amdgpu_df_funcs df_v4_6_2_funcs;
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
index 516409989235..d75b9940f248 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c
@@ -40,11 +40,11 @@
 #include "ivsrcid/gfx/irqsrcs_gfx_10_1.h"
 
 #include "soc15.h"
-#include "soc15d.h"
 #include "soc15_common.h"
 #include "clearstate_gfx10.h"
 #include "v10_structs.h"
 #include "gfx_v10_0.h"
+#include "gfx_v10_0_cleaner_shader.h"
 #include "nbio_v2_3.h"
 
 /*
@@ -102,6 +102,11 @@
 #define mmGCR_GENERAL_CNTL_Sienna_Cichlid			0x1580
 #define mmGCR_GENERAL_CNTL_Sienna_Cichlid_BASE_IDX	0
 
+#define mmGOLDEN_TSC_COUNT_UPPER_Cyan_Skillfish                0x0105
+#define mmGOLDEN_TSC_COUNT_UPPER_Cyan_Skillfish_BASE_IDX       1
+#define mmGOLDEN_TSC_COUNT_LOWER_Cyan_Skillfish                0x0106
+#define mmGOLDEN_TSC_COUNT_LOWER_Cyan_Skillfish_BASE_IDX       1
+
 #define mmGOLDEN_TSC_COUNT_UPPER_Vangogh                0x0025
 #define mmGOLDEN_TSC_COUNT_UPPER_Vangogh_BASE_IDX       1
 #define mmGOLDEN_TSC_COUNT_LOWER_Vangogh                0x0026
@@ -271,8 +276,211 @@ MODULE_FIRMWARE("amdgpu/gc_10_3_7_mec.bin");
 MODULE_FIRMWARE("amdgpu/gc_10_3_7_mec2.bin");
 MODULE_FIRMWARE("amdgpu/gc_10_3_7_rlc.bin");
 
-static const struct soc15_reg_golden golden_settings_gc_10_1[] =
-{
+static const struct amdgpu_hwip_reg_entry gc_reg_list_10_1[] = {
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS3),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_STALLED_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_BUSY_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_BUSY_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HPD_STATUS0),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB0_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB0_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB0_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB1_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB1_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB1_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB2_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB2_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB2_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCPF_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCPC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCPG_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGDS_PROTECTION_FAULT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGDS_VM_PROTECTION_FAULT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmIA_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmIA_UTCL1_STATUS_2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmPA_CL_CNTL_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRMI_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSQC_DCACHE_UTCL0_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSQC_ICACHE_UTCL0_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSQG_UTCL0_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmTCP_UTCL0_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmWD_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGCVM_L2_PROTECTION_FAULT_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGCVM_L2_PROTECTION_FAULT_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_DEBUG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MES_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC1_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC2_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MES_DEBUG_INTERRUPT_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MES_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_COMMAND),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_MESSAGE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_ARGUMENT_1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_ARGUMENT_2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_ARGUMENT_3),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_ARGUMENT_4),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSMU_RLC_RESPONSE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SAFE_MODE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_SAFE_MODE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_RLCS_GPM_STAT_2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SPP_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_RLCS_BOOTLOAD_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_INT_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_GPM_GENERAL_6),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_GPM_DEBUG_INST_A),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_GPM_DEBUG_INST_B),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_GPM_DEBUG_INST_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_LX6_CORE_PDEBUG_INST),
+	/* cp header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MES_HEADER_DUMP),
+	/* SE status registers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE0),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE3)
+};
+
+static const struct amdgpu_hwip_reg_entry gc_cp_reg_list_10[] = {
+	/* compute registers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_VMID),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PERSISTENT_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PIPE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_QUEUE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_QUANTUM),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_DEQUEUE_REQUEST),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_EVENTS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_BASE_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CNTL_STACK_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CNTL_STACK_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_WG_STATE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_GDS_RESOURCE_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_WPTR_MEM),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_SUSPEND_CNTL_STACK_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_SUSPEND_CNTL_STACK_DW_CNT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_SUSPEND_WG_STATE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_DEQUEUE_STATUS),
+	/* cp header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+};
+
+static const struct amdgpu_hwip_reg_entry gc_gfx_queue_reg_list_10[] = {
+	/* gfx queue registers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_ACTIVE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_QUEUE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_CSMD_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_DEQUEUE_REQUEST),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_MAPPED),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_QUE_MGR_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_HQ_CONTROL0),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_HQ_STATUS0),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_CE_WPTR_POLL_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_CE_WPTR_POLL_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_CE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_CE_CSMD_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_CE_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_HQD_CE_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_MQD_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_MQD_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_WPTR_POLL_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_WPTR_POLL_ADDR_HI),
+	/* gfx header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+};
+
+static const struct soc15_reg_golden golden_settings_gc_10_1[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCB_HW_CONTROL_4, 0xffffffff, 0x00400014),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_CPF_CLK_CTRL, 0xfcff8fff, 0xf8000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_CLK_CTRL, 0xcd000000, 0x0d000100),
@@ -315,13 +523,11 @@ static const struct soc15_reg_golden golden_settings_gc_10_1[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0x00c00000, 0x00c00000)
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_0_nv10[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_0_nv10[] = {
 	/* Pending on emulation bring up */
 };
 
-static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_0_nv10[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_0_nv10[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xe0000000, 0x0),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xffffff, 0x0),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_SPM_GLB_SAMPLEDELAY_IND_ADDR, 0xFFFFFFFF, 0x28),
@@ -1376,8 +1582,7 @@ static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_0_nv10[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xFFFFFFFF, 0xe0000000)
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_1_1[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_1_1[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCB_HW_CONTROL_4, 0xffffffff, 0x003c0014),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_GS_NGG_CLK_CTRL, 0xffff8fff, 0xffff8100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_IA_CLK_CTRL, 0xffff0fff, 0xffff0100),
@@ -1418,8 +1623,7 @@ static const struct soc15_reg_golden golden_settings_gc_10_1_1[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0x00c00000, 0x00c00000),
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_1_2[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_1_2[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCB_HW_CONTROL_4, 0x003e001f, 0x003c0014),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_GS_NGG_CLK_CTRL, 0xffff8fff, 0xffff8100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_IA_CLK_CTRL, 0xffff0fff, 0xffff0100),
@@ -1464,13 +1668,11 @@ static const struct soc15_reg_golden golden_settings_gc_10_1_2[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0xffffffff, 0x00c00000)
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_1_nv14[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_1_nv14[] = {
 	/* Pending on emulation bring up */
 };
 
-static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_1_nv14[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_1_nv14[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xE0000000L, 0x0),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xffffff, 0x0),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_SPM_GLB_SAMPLEDELAY_IND_ADDR, 0xFFFFFFFF, 0x28),
@@ -2093,13 +2295,11 @@ static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_1_nv14[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xFFFFFFFF, 0xe0000000)
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_1_2_nv12[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_1_2_nv12[] = {
 	/* Pending on emulation bring up */
 };
 
-static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_1_2_nv12[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_1_2_nv12[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xe0000000L, 0x0),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xffffff, 0x0),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_SPM_GLB_SAMPLEDELAY_IND_ADDR, 0xFFFFFFFF, 0x28),
@@ -3154,8 +3354,7 @@ static const struct soc15_reg_golden golden_settings_gc_rlc_spm_10_1_2_nv12[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_GFX_INDEX, 0xFFFFFFFF, 0xe0000000)
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_3[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_3[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_CS_CLK_CTRL, 0x78000000, 0x78000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_PS_CLK_CTRL, 0xff7f0fff, 0x78000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_RA0_CLK_CTRL, 0xff7f0fff, 0x30000100),
@@ -3164,7 +3363,7 @@ static const struct soc15_reg_golden golden_settings_gc_10_3[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG3, 0xffffffff, 0x00000280),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG4, 0xffffffff, 0x00800000),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_EXCEPTION_CONTROL, 0x7fff0f1f, 0x00b80000),
-	SOC15_REG_GOLDEN_VALUE(GC, 0 ,mmGCEA_SDP_TAG_RESERVE0, 0xffffffff, 0x10100100),
+	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGCEA_SDP_TAG_RESERVE0, 0xffffffff, 0x10100100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGCEA_SDP_TAG_RESERVE1, 0xffffffff, 0x17000088),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGCR_GENERAL_CNTL_Sienna_Cichlid, 0x1ff1ffff, 0x00000500),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmGCUTCL2_CGTT_CLK_CTRL_Sienna_Cichlid, 0xff000000, 0xff008080),
@@ -3201,13 +3400,11 @@ static const struct soc15_reg_golden golden_settings_gc_10_3[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0xffbfffff, 0x00a00000)
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_3_sienna_cichlid[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_3_sienna_cichlid[] = {
 	/* Pending on emulation bring up */
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_3_2[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_3_2[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_CS_CLK_CTRL, 0xff7f0fff, 0x78000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_PS_CLK_CTRL, 0xff7f0fff, 0x78000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_RA0_CLK_CTRL, 0xff7f0fff, 0x30000100),
@@ -3254,8 +3451,7 @@ static const struct soc15_reg_golden golden_settings_gc_10_3_2[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmLDS_CONFIG,  0x00000020, 0x00000020),
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_3_vangogh[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_3_vangogh[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_RA0_CLK_CTRL, 0xff7f0fff, 0x30000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_RA1_CLK_CTRL, 0xff7f0fff, 0x7e000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCH_PIPE_STEER, 0x000000ff, 0x000000e4),
@@ -3285,8 +3481,7 @@ static const struct soc15_reg_golden golden_settings_gc_10_3_vangogh[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmLDS_CONFIG,  0x00000020, 0x00000020),
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_3_3[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_3_3[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_CS_CLK_CTRL, 0xff7f0fff, 0x78000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCH_PIPE_STEER, 0x000000ff, 0x000000e4),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPF_GCR_CNTL, 0x0007ffff, 0x0000c200),
@@ -3309,8 +3504,7 @@ static const struct soc15_reg_golden golden_settings_gc_10_3_3[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0xffffffff, 0x00100000)
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_3_4[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_3_4[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_CS_CLK_CTRL, 0x78000000, 0x78000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_RA0_CLK_CTRL, 0x30000000, 0x30000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_RA1_CLK_CTRL, 0x7e000000, 0x7e000100),
@@ -3380,7 +3574,7 @@ static const struct soc15_reg_golden golden_settings_gc_10_3_5[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER7_SELECT, 0xf0f001ff, 0x00000000),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER8_SELECT, 0xf0f001ff, 0x00000000),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQ_PERFCOUNTER9_SELECT, 0xf0f001ff, 0x00000000),
-	SOC15_REG_GOLDEN_VALUE(GC, 0, mmTA_CNTL_AUX,0xfff7ffff, 0x01030000),
+	SOC15_REG_GOLDEN_VALUE(GC, 0, mmTA_CNTL_AUX, 0xfff7ffff, 0x01030000),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0xffbfffff, 0x00a00000)
 };
 
@@ -3421,8 +3615,7 @@ static const struct soc15_reg_golden golden_settings_gc_10_0_cyan_skillfish[] =
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmUTCL1_CTRL, 0x00800000, 0x00800000)
 };
 
-static const struct soc15_reg_golden golden_settings_gc_10_3_6[] =
-{
+static const struct soc15_reg_golden golden_settings_gc_10_3_6[] = {
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCGTT_SPI_CS_CLK_CTRL, 0xff7f0fff, 0x78000100),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCH_PIPE_STEER, 0x000000ff, 0x00000044),
 	SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPF_GCR_CNTL, 0x0007ffff, 0x0000c200),
@@ -3490,7 +3683,7 @@ static int gfx_v10_0_get_cu_info(struct amdgpu_device *adev,
 				 struct amdgpu_cu_info *cu_info);
 static uint64_t gfx_v10_0_get_gpu_clock_counter(struct amdgpu_device *adev);
 static void gfx_v10_0_select_se_sh(struct amdgpu_device *adev, u32 se_num,
-				   u32 sh_num, u32 instance);
+				   u32 sh_num, u32 instance, int xcc_id);
 static u32 gfx_v10_0_get_wgp_active_bitmap_per_sh(struct amdgpu_device *adev);
 
 static int gfx_v10_0_rlc_backdoor_autoload_buffer_init(struct amdgpu_device *adev);
@@ -3506,16 +3699,26 @@ static void gfx_v10_3_set_power_brake_sequence(struct amdgpu_device *adev);
 static void gfx_v10_0_ring_invalidate_tlbs(struct amdgpu_ring *ring,
 					   uint16_t pasid, uint32_t flush_type,
 					   bool all_hub, uint8_t dst_sel);
+static void gfx_v10_0_update_spm_vmid_internal(struct amdgpu_device *adev,
+					       unsigned int vmid);
 
+static int gfx_v10_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_powergating_state state);
 static void gfx10_kiq_set_resources(struct amdgpu_ring *kiq_ring, uint64_t queue_mask)
 {
+	struct amdgpu_device *adev = kiq_ring->adev;
+	u64 shader_mc_addr;
+
+	/* Cleaner shader MC address */
+	shader_mc_addr = adev->gfx.cleaner_shader_gpu_addr >> 8;
+
 	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_RESOURCES, 6));
 	amdgpu_ring_write(kiq_ring, PACKET3_SET_RESOURCES_VMID_MASK(0) |
 			  PACKET3_SET_RESOURCES_QUEUE_TYPE(0));	/* vmid_mask:0 queue_type:0 (KIQ) */
 	amdgpu_ring_write(kiq_ring, lower_32_bits(queue_mask));	/* queue mask lo */
 	amdgpu_ring_write(kiq_ring, upper_32_bits(queue_mask));	/* queue mask hi */
-	amdgpu_ring_write(kiq_ring, 0);	/* gws mask lo */
-	amdgpu_ring_write(kiq_ring, 0);	/* gws mask hi */
+	amdgpu_ring_write(kiq_ring, lower_32_bits(shader_mc_addr)); /* cleaner shader addr lo */
+	amdgpu_ring_write(kiq_ring, upper_32_bits(shader_mc_addr)); /* cleaner shader addr hi */
 	amdgpu_ring_write(kiq_ring, 0);	/* oac mask */
 	amdgpu_ring_write(kiq_ring, 0);	/* gds heap base:0, gds heap size:0 */
 }
@@ -3565,14 +3768,8 @@ static void gfx10_kiq_unmap_queues(struct amdgpu_ring *kiq_ring,
 				   enum amdgpu_unmap_queues_action action,
 				   u64 gpu_addr, u64 seq)
 {
-	struct amdgpu_device *adev = kiq_ring->adev;
 	uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0;
 
-	if (adev->enable_mes && !adev->gfx.kiq.ring.sched.ready) {
-		amdgpu_mes_unmap_legacy_queue(adev, ring, action, gpu_addr, seq);
-		return;
-	}
-
 	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4));
 	amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
 			  PACKET3_UNMAP_QUEUES_ACTION(action) |
@@ -3621,12 +3818,65 @@ static void gfx10_kiq_invalidate_tlbs(struct amdgpu_ring *kiq_ring,
 	gfx_v10_0_ring_invalidate_tlbs(kiq_ring, pasid, flush_type, all_hub, 1);
 }
 
+static void gfx_v10_0_kiq_reset_hw_queue(struct amdgpu_ring *kiq_ring, uint32_t queue_type,
+					uint32_t me_id, uint32_t pipe_id, uint32_t queue_id,
+					uint32_t xcc_id, uint32_t vmid)
+{
+	struct amdgpu_device *adev = kiq_ring->adev;
+	unsigned i;
+	uint32_t tmp;
+
+	/* enter save mode */
+	amdgpu_gfx_rlc_enter_safe_mode(adev, xcc_id);
+	mutex_lock(&adev->srbm_mutex);
+	nv_grbm_select(adev, me_id, pipe_id, queue_id, 0);
+
+	if (queue_type == AMDGPU_RING_TYPE_COMPUTE) {
+		WREG32_SOC15(GC, 0, mmCP_HQD_DEQUEUE_REQUEST, 0x2);
+		WREG32_SOC15(GC, 0, mmSPI_COMPUTE_QUEUE_RESET, 0x1);
+		/* wait till dequeue take effects */
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (!(RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+		if (i >= adev->usec_timeout)
+			dev_err(adev->dev, "fail to wait on hqd deactive\n");
+	} else if (queue_type == AMDGPU_RING_TYPE_GFX) {
+		WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX,
+			     (uint32_t)(0x1 << GRBM_GFX_INDEX__SE_BROADCAST_WRITES__SHIFT));
+		tmp = REG_SET_FIELD(0, CP_VMID_RESET, RESET_REQUEST, 1 << vmid);
+		if (pipe_id == 0)
+			tmp = REG_SET_FIELD(tmp, CP_VMID_RESET, PIPE0_QUEUES, 1 << queue_id);
+		else
+			tmp = REG_SET_FIELD(tmp, CP_VMID_RESET, PIPE1_QUEUES, 1 << queue_id);
+		WREG32_SOC15(GC, 0, mmCP_VMID_RESET, tmp);
+
+		/* wait till dequeue take effects */
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (!(RREG32_SOC15(GC, 0, mmCP_GFX_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+		if (i >= adev->usec_timeout)
+			dev_err(adev->dev, "failed to wait on gfx hqd deactivate\n");
+	} else {
+		dev_err(adev->dev, "reset queue_type(%d) not supported\n", queue_type);
+	}
+
+	nv_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	/* exit safe mode */
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
+}
+
 static const struct kiq_pm4_funcs gfx_v10_0_kiq_pm4_funcs = {
 	.kiq_set_resources = gfx10_kiq_set_resources,
 	.kiq_map_queues = gfx10_kiq_map_queues,
 	.kiq_unmap_queues = gfx10_kiq_unmap_queues,
 	.kiq_query_status = gfx10_kiq_query_status,
 	.kiq_invalidate_tlbs = gfx10_kiq_invalidate_tlbs,
+	.kiq_reset_hw_queue = gfx_v10_0_kiq_reset_hw_queue,
 	.set_resources_size = 8,
 	.map_queues_size = 7,
 	.unmap_queues_size = 6,
@@ -3636,12 +3886,12 @@ static const struct kiq_pm4_funcs gfx_v10_0_kiq_pm4_funcs = {
 
 static void gfx_v10_0_set_kiq_pm4_funcs(struct amdgpu_device *adev)
 {
-	adev->gfx.kiq.pmf = &gfx_v10_0_kiq_pm4_funcs;
+	adev->gfx.kiq[0].pmf = &gfx_v10_0_kiq_pm4_funcs;
 }
 
 static void gfx_v10_0_init_spm_golden_registers(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 		soc15_program_register_sequence(adev,
 						golden_settings_gc_rlc_spm_10_0_nv10,
@@ -3664,7 +3914,10 @@ static void gfx_v10_0_init_spm_golden_registers(struct amdgpu_device *adev)
 
 static void gfx_v10_0_init_golden_registers(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 		soc15_program_register_sequence(adev,
 						golden_settings_gc_10_1,
@@ -3714,8 +3967,8 @@ static void gfx_v10_0_init_golden_registers(struct amdgpu_device *adev)
 		break;
 	case IP_VERSION(10, 3, 4):
 		soc15_program_register_sequence(adev,
-                                                golden_settings_gc_10_3_4,
-                                                (const u32)ARRAY_SIZE(golden_settings_gc_10_3_4));
+						golden_settings_gc_10_3_4,
+						(const u32)ARRAY_SIZE(golden_settings_gc_10_3_4));
 		break;
 	case IP_VERSION(10, 3, 5):
 		soc15_program_register_sequence(adev,
@@ -3782,7 +4035,7 @@ static int gfx_v10_0_ring_test_ring(struct amdgpu_ring *ring)
 	struct amdgpu_device *adev = ring->adev;
 	uint32_t scratch = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0);
 	uint32_t tmp = 0;
-	unsigned i;
+	unsigned int i;
 	int r;
 
 	WREG32(scratch, 0xCAFEDEAD);
@@ -3820,40 +4073,25 @@ static int gfx_v10_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	struct amdgpu_device *adev = ring->adev;
 	struct amdgpu_ib ib;
 	struct dma_fence *f = NULL;
-	unsigned index;
+	unsigned int index;
 	uint64_t gpu_addr;
-	volatile uint32_t *cpu_ptr;
+	uint32_t *cpu_ptr;
 	long r;
 
 	memset(&ib, 0, sizeof(ib));
 
-	if (ring->is_mes_queue) {
-		uint32_t padding, offset;
-
-		offset = amdgpu_mes_ctx_get_offs(ring, AMDGPU_MES_CTX_IB_OFFS);
-		padding = amdgpu_mes_ctx_get_offs(ring,
-						  AMDGPU_MES_CTX_PADDING_OFFS);
-
-		ib.gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		ib.ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-
-		gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, padding);
-		cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, padding);
-		*cpu_ptr = cpu_to_le32(0xCAFEDEAD);
-	} else {
-		r = amdgpu_device_wb_get(adev, &index);
-		if (r)
-			return r;
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r)
+		return r;
 
-		gpu_addr = adev->wb.gpu_addr + (index * 4);
-		adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
-		cpu_ptr = &adev->wb.wb[index];
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
+	cpu_ptr = &adev->wb.wb[index];
 
-		r = amdgpu_ib_get(adev, NULL, 20, AMDGPU_IB_POOL_DIRECT, &ib);
-		if (r) {
-			DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
-			goto err1;
-		}
+	r = amdgpu_ib_get(adev, NULL, 20, AMDGPU_IB_POOL_DIRECT, &ib);
+	if (r) {
+		DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
+		goto err1;
 	}
 
 	ib.ptr[0] = PACKET3(PACKET3_WRITE_DATA, 3);
@@ -3880,12 +4118,10 @@ static int gfx_v10_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	else
 		r = -EINVAL;
 err2:
-	if (!ring->is_mes_queue)
-		amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err1:
-	if (!ring->is_mes_queue)
-		amdgpu_device_wb_free(adev, index);
+	amdgpu_device_wb_free(adev, index);
 	return r;
 }
 
@@ -3905,7 +4141,7 @@ static void gfx_v10_0_check_fw_write_wait(struct amdgpu_device *adev)
 {
 	adev->gfx.cp_fw_write_wait = false;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 2):
 	case IP_VERSION(10, 1, 1):
@@ -3951,12 +4187,12 @@ static bool gfx_v10_0_navi10_gfxoff_should_enable(struct amdgpu_device *adev)
 		break;
 	}
 
-	return ret ;
+	return ret;
 }
 
 static void gfx_v10_0_check_gfxoff_flag(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 		if (!gfx_v10_0_navi10_gfxoff_should_enable(adev))
 			adev->pm.pp_feature &= ~PP_GFXOFF_MASK;
@@ -3968,7 +4204,7 @@ static void gfx_v10_0_check_gfxoff_flag(struct amdgpu_device *adev)
 
 static int gfx_v10_0_init_microcode(struct amdgpu_device *adev)
 {
-	char fw_name[40];
+	char fw_name[53];
 	char ucode_prefix[30];
 	const char *wks = "";
 	int err;
@@ -3978,41 +4214,41 @@ static int gfx_v10_0_init_microcode(struct amdgpu_device *adev)
 
 	DRM_DEBUG("\n");
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 1) &&
-	   (!(adev->pdev->device == 0x7340 && adev->pdev->revision != 0x00)))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 1) &&
+	    (!(adev->pdev->device == 0x7340 && adev->pdev->revision != 0x00)))
 		wks = "_wks";
 	amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix));
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp%s.bin", ucode_prefix, wks);
-	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_pfp%s.bin", ucode_prefix, wks);
 	if (err)
 		goto out;
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_PFP);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me%s.bin", ucode_prefix, wks);
-	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_me%s.bin", ucode_prefix, wks);
 	if (err)
 		goto out;
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_ME);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce%s.bin", ucode_prefix, wks);
-	err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_ce%s.bin", ucode_prefix, wks);
 	if (err)
 		goto out;
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_CE);
 
 	if (!amdgpu_sriov_vf(adev)) {
 		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", ucode_prefix);
-		err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, fw_name);
-		/* don't check this.  There are apparently firmwares in the wild with
-		 * incorrect size in the header
-		 */
-		if (err == -ENODEV)
-			goto out;
+		err = request_firmware(&adev->gfx.rlc_fw, fw_name, adev->dev);
 		if (err)
-			dev_dbg(adev->dev,
-				"gfx10: amdgpu_ucode_request() failed \"%s\"\n",
-				fw_name);
+			goto out;
+
+		/* don't validate this firmware. There are apparently firmwares
+		 * in the wild with incorrect size in the header
+		 */
 		rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
 		version_major = le16_to_cpu(rlc_hdr->header.header_version_major);
 		version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor);
@@ -4021,15 +4257,17 @@ static int gfx_v10_0_init_microcode(struct amdgpu_device *adev)
 			goto out;
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec%s.bin", ucode_prefix, wks);
-	err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_mec%s.bin", ucode_prefix, wks);
 	if (err)
 		goto out;
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1);
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1_JT);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2%s.bin", ucode_prefix, wks);
-	err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_mec2%s.bin", ucode_prefix, wks);
 	if (!err) {
 		amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2);
 		amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2_JT);
@@ -4037,8 +4275,6 @@ static int gfx_v10_0_init_microcode(struct amdgpu_device *adev)
 		err = 0;
 		adev->gfx.mec2_fw = NULL;
 	}
-	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2);
-	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2_JT);
 
 	gfx_v10_0_check_fw_write_wait(adev);
 out:
@@ -4086,12 +4322,9 @@ static u32 gfx_v10_0_get_csb_size(struct amdgpu_device *adev)
 	return count;
 }
 
-static void gfx_v10_0_get_csb_buffer(struct amdgpu_device *adev,
-				    volatile u32 *buffer)
+static void gfx_v10_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer)
 {
-	u32 count = 0, i;
-	const struct cs_section_def *sect = NULL;
-	const struct cs_extent_def *ext = NULL;
+	u32 count = 0;
 	int ctx_reg_offset;
 
 	if (adev->gfx.rlc.cs_data == NULL)
@@ -4099,39 +4332,15 @@ static void gfx_v10_0_get_csb_buffer(struct amdgpu_device *adev,
 	if (buffer == NULL)
 		return;
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1));
-	buffer[count++] = cpu_to_le32(0x80000000);
-	buffer[count++] = cpu_to_le32(0x80000000);
-
-	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
-		for (ext = sect->section; ext->extent != NULL; ++ext) {
-			if (sect->id == SECT_CONTEXT) {
-				buffer[count++] =
-					cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
-				buffer[count++] = cpu_to_le32(ext->reg_index -
-						PACKET3_SET_CONTEXT_REG_START);
-				for (i = 0; i < ext->reg_count; i++)
-					buffer[count++] = cpu_to_le32(ext->extent[i]);
-			} else {
-				return;
-			}
-		}
-	}
+	count = amdgpu_gfx_csb_preamble_start(buffer);
+	count = amdgpu_gfx_csb_data_parser(adev, buffer, count);
 
-	ctx_reg_offset =
-		SOC15_REG_OFFSET(GC, 0, mmPA_SC_TILE_STEERING_OVERRIDE) - PACKET3_SET_CONTEXT_REG_START;
+	ctx_reg_offset = SOC15_REG_OFFSET(GC, 0, mmPA_SC_TILE_STEERING_OVERRIDE) - PACKET3_SET_CONTEXT_REG_START;
 	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, 1));
 	buffer[count++] = cpu_to_le32(ctx_reg_offset);
 	buffer[count++] = cpu_to_le32(adev->gfx.config.pa_sc_tile_steering_override);
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0));
-	buffer[count++] = cpu_to_le32(0);
+	amdgpu_gfx_csb_preamble_end(buffer, count);
 }
 
 static void gfx_v10_0_rlc_fini(struct amdgpu_device *adev)
@@ -4151,22 +4360,22 @@ static void gfx_v10_0_init_rlcg_reg_access_ctrl(struct amdgpu_device *adev)
 {
 	struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl;
 
-	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl;
+	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl[0];
 	reg_access_ctrl->scratch_reg0 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0);
 	reg_access_ctrl->scratch_reg1 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG1);
 	reg_access_ctrl->scratch_reg2 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG2);
 	reg_access_ctrl->scratch_reg3 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG3);
 	reg_access_ctrl->grbm_cntl = SOC15_REG_OFFSET(GC, 0, mmGRBM_GFX_CNTL);
 	reg_access_ctrl->grbm_idx = SOC15_REG_OFFSET(GC, 0, mmGRBM_GFX_INDEX);
-	switch (adev->ip_versions[GC_HWIP][0]) {
-		case IP_VERSION(10, 3, 0):
-			reg_access_ctrl->spare_int =
-				SOC15_REG_OFFSET(GC, 0, mmRLC_SPARE_INT_0_Sienna_Cichlid);
-			break;
-		default:
-			reg_access_ctrl->spare_int =
-				SOC15_REG_OFFSET(GC, 0, mmRLC_SPARE_INT);
-			break;
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(10, 3, 0):
+		reg_access_ctrl->spare_int =
+			SOC15_REG_OFFSET(GC, 0, mmRLC_SPARE_INT_0_Sienna_Cichlid);
+		break;
+	default:
+		reg_access_ctrl->spare_int =
+			SOC15_REG_OFFSET(GC, 0, mmRLC_SPARE_INT);
+		break;
 	}
 	adev->gfx.rlc.rlcg_reg_access_supported = true;
 }
@@ -4187,11 +4396,6 @@ static int gfx_v10_0_rlc_init(struct amdgpu_device *adev)
 			return r;
 	}
 
-	/* init spm vmid with 0xf */
-	if (adev->gfx.rlc.funcs->update_spm_vmid)
-		adev->gfx.rlc.funcs->update_spm_vmid(adev, 0xf);
-
-
 	return 0;
 }
 
@@ -4213,13 +4417,13 @@ static int gfx_v10_0_mec_init(struct amdgpu_device *adev)
 	int r;
 	u32 *hpd;
 	const __le32 *fw_data = NULL;
-	unsigned fw_size;
+	unsigned int fw_size;
 	u32 *fw = NULL;
 	size_t mec_hpd_size;
 
 	const struct gfx_firmware_header_v1_0 *mec_hdr = NULL;
 
-	bitmap_zero(adev->gfx.mec.queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
+	bitmap_zero(adev->gfx.mec_bitmap[0].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
 
 	/* take ownership of the relevant compute queues */
 	amdgpu_gfx_compute_queue_acquire(adev);
@@ -4291,11 +4495,12 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t wave,
 		*(out++) = RREG32_SOC15(GC, 0, mmSQ_IND_DATA);
 }
 
-static void gfx_v10_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
+static void gfx_v10_0_read_wave_data(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
 {
 	/* in gfx10 the SIMD_ID is specified as part of the INSTANCE
 	 * field when performing a select_se_sh so it should be
-	 * zero here */
+	 * zero here
+	 */
 	WARN_ON(simd != 0);
 
 	/* type 2 wave data */
@@ -4318,7 +4523,7 @@ static void gfx_v10_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd,
 	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_MODE);
 }
 
-static void gfx_v10_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v10_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				     uint32_t wave, uint32_t start,
 				     uint32_t size, uint32_t *dst)
 {
@@ -4329,7 +4534,7 @@ static void gfx_v10_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
 		dst);
 }
 
-static void gfx_v10_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v10_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				      uint32_t wave, uint32_t thread,
 				      uint32_t start, uint32_t size,
 				      uint32_t *dst)
@@ -4340,7 +4545,7 @@ static void gfx_v10_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd,
 }
 
 static void gfx_v10_0_select_me_pipe_q(struct amdgpu_device *adev,
-				       u32 me, u32 pipe, u32 q, u32 vm)
+				       u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
 {
 	nv_grbm_select(adev, me, pipe, q, vm);
 }
@@ -4376,7 +4581,7 @@ static void gfx_v10_0_gpu_early_init(struct amdgpu_device *adev)
 {
 	u32 gb_addr_config;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 1):
 	case IP_VERSION(10, 1, 2):
@@ -4461,6 +4666,7 @@ static int gfx_v10_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id,
 		ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
 	else
 		ring->doorbell_index = adev->doorbell_index.gfx_ring1 << 1;
+	ring->vm_hub = AMDGPU_GFXHUB(0);
 	sprintf(ring->name, "gfx_%d.%d.%d", ring->me, ring->pipe, ring->queue);
 
 	irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + ring->pipe;
@@ -4473,7 +4679,7 @@ static int gfx_v10_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id,
 static int gfx_v10_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 				       int mec, int pipe, int queue)
 {
-	unsigned irq_type;
+	unsigned int irq_type;
 	struct amdgpu_ring *ring;
 	unsigned int hw_prio;
 
@@ -4489,6 +4695,7 @@ static int gfx_v10_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 	ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
 	ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
 				+ (ring_id * GFX10_MEC_HPD_SIZE);
+	ring->vm_hub = AMDGPU_GFXHUB(0);
 	sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
 
 	irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
@@ -4501,13 +4708,57 @@ static int gfx_v10_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 			     hw_prio, NULL);
 }
 
-static int gfx_v10_0_sw_init(void *handle)
+static void gfx_v10_0_alloc_ip_dump(struct amdgpu_device *adev)
+{
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_10_1);
+	uint32_t *ptr;
+	uint32_t inst;
+
+	ptr = kcalloc(reg_count, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for GFX IP Dump\n");
+		adev->gfx.ip_dump_core = NULL;
+	} else {
+		adev->gfx.ip_dump_core = ptr;
+	}
+
+	/* Allocate memory for compute queue registers for all the instances */
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_10);
+	inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec *
+		adev->gfx.mec.num_queue_per_pipe;
+
+	ptr = kcalloc(reg_count * inst, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for Compute Queues IP Dump\n");
+		adev->gfx.ip_dump_compute_queues = NULL;
+	} else {
+		adev->gfx.ip_dump_compute_queues = ptr;
+	}
+
+	/* Allocate memory for gfx queue registers for all the instances */
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_10);
+	inst = adev->gfx.me.num_me * adev->gfx.me.num_pipe_per_me *
+		adev->gfx.me.num_queue_per_pipe;
+
+	ptr = kcalloc(reg_count * inst, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for GFX Queues IP Dump\n");
+		adev->gfx.ip_dump_gfx_queues = NULL;
+	} else {
+		adev->gfx.ip_dump_gfx_queues = ptr;
+	}
+}
+
+static int gfx_v10_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int i, j, k, r, ring_id = 0;
-	struct amdgpu_kiq *kiq;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	int xcc_id = 0;
+	struct amdgpu_device *adev = ip_block->adev;
+	int num_queue_per_pipe = 1; /* we only enable 1 KGQ per pipe */
+
+	INIT_DELAYED_WORK(&adev->gfx.idle_work, amdgpu_gfx_profile_idle_work_handler);
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 1):
 	case IP_VERSION(10, 1, 2):
@@ -4515,7 +4766,7 @@ static int gfx_v10_0_sw_init(void *handle)
 	case IP_VERSION(10, 1, 4):
 		adev->gfx.me.num_me = 1;
 		adev->gfx.me.num_pipe_per_me = 1;
-		adev->gfx.me.num_queue_per_pipe = 1;
+		adev->gfx.me.num_queue_per_pipe = 8;
 		adev->gfx.mec.num_mec = 2;
 		adev->gfx.mec.num_pipe_per_mec = 4;
 		adev->gfx.mec.num_queue_per_pipe = 8;
@@ -4529,8 +4780,8 @@ static int gfx_v10_0_sw_init(void *handle)
 	case IP_VERSION(10, 3, 3):
 	case IP_VERSION(10, 3, 7):
 		adev->gfx.me.num_me = 1;
-		adev->gfx.me.num_pipe_per_me = 1;
-		adev->gfx.me.num_queue_per_pipe = 1;
+		adev->gfx.me.num_pipe_per_me = 2;
+		adev->gfx.me.num_queue_per_pipe = 2;
 		adev->gfx.mec.num_mec = 2;
 		adev->gfx.mec.num_pipe_per_mec = 4;
 		adev->gfx.mec.num_queue_per_pipe = 4;
@@ -4544,11 +4795,79 @@ static int gfx_v10_0_sw_init(void *handle)
 		adev->gfx.mec.num_queue_per_pipe = 8;
 		break;
 	}
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(10, 1, 10):
+	case IP_VERSION(10, 1, 1):
+	case IP_VERSION(10, 1, 2):
+		adev->gfx.cleaner_shader_ptr = gfx_10_1_10_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_10_1_10_cleaner_shader_hex);
+		if (adev->gfx.me_fw_version >= 101 &&
+		    adev->gfx.pfp_fw_version  >= 158 &&
+		    adev->gfx.mec_fw_version >= 151) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	case IP_VERSION(10, 3, 0):
+	case IP_VERSION(10, 3, 1):
+	case IP_VERSION(10, 3, 2):
+	case IP_VERSION(10, 3, 3):
+	case IP_VERSION(10, 3, 4):
+	case IP_VERSION(10, 3, 5):
+		adev->gfx.cleaner_shader_ptr = gfx_10_3_0_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_10_3_0_cleaner_shader_hex);
+		if (adev->gfx.me_fw_version >= 64 &&
+		    adev->gfx.pfp_fw_version >= 100 &&
+		    adev->gfx.mec_fw_version >= 122) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	case IP_VERSION(10, 3, 6):
+		adev->gfx.cleaner_shader_ptr = gfx_10_3_0_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_10_3_0_cleaner_shader_hex);
+		if (adev->gfx.me_fw_version >= 14 &&
+		    adev->gfx.pfp_fw_version >= 17 &&
+		    adev->gfx.mec_fw_version >= 24) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	case IP_VERSION(10, 3, 7):
+		adev->gfx.cleaner_shader_ptr = gfx_10_3_0_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_10_3_0_cleaner_shader_hex);
+		if (adev->gfx.me_fw_version >= 4 &&
+		    adev->gfx.pfp_fw_version >= 9 &&
+		    adev->gfx.mec_fw_version >= 12) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	default:
+		adev->gfx.enable_cleaner_shader = false;
+		break;
+	}
 
 	/* KIQ event */
 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP,
 			      GFX_10_1__SRCID__CP_IB2_INTERRUPT_PKT,
-			      &adev->gfx.kiq.irq);
+			      &adev->gfx.kiq[0].irq);
 	if (r)
 		return r;
 
@@ -4559,6 +4878,13 @@ static int gfx_v10_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	/* Bad opcode Event */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP,
+			      GFX_10_1__SRCID__CP_BAD_OPCODE_ERROR,
+			      &adev->gfx.bad_op_irq);
+	if (r)
+		return r;
+
 	/* Privileged reg */
 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_10_1__SRCID__CP_PRIV_REG_FAULT,
 			      &adev->gfx.priv_reg_irq);
@@ -4593,7 +4919,7 @@ static int gfx_v10_0_sw_init(void *handle)
 
 	/* set up the gfx ring */
 	for (i = 0; i < adev->gfx.me.num_me; i++) {
-		for (j = 0; j < adev->gfx.me.num_queue_per_pipe; j++) {
+		for (j = 0; j < num_queue_per_pipe; j++) {
 			for (k = 0; k < adev->gfx.me.num_pipe_per_me; k++) {
 				if (!amdgpu_gfx_is_me_queue_enabled(adev, i, k, j))
 					continue;
@@ -4612,8 +4938,8 @@ static int gfx_v10_0_sw_init(void *handle)
 	for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
 		for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
 			for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
-				if (!amdgpu_gfx_is_mec_queue_enabled(adev, i, k,
-								     j))
+				if (!amdgpu_gfx_is_mec_queue_enabled(adev, 0, i,
+								     k, j))
 					continue;
 
 				r = gfx_v10_0_compute_ring_init(adev, ring_id,
@@ -4626,20 +4952,27 @@ static int gfx_v10_0_sw_init(void *handle)
 		}
 	}
 
-	if (!adev->enable_mes_kiq) {
-		r = amdgpu_gfx_kiq_init(adev, GFX10_MEC_HPD_SIZE);
-		if (r) {
-			DRM_ERROR("Failed to init KIQ BOs!\n");
-			return r;
-		}
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+	if (!amdgpu_sriov_vf(adev) &&
+	    !adev->debug_disable_gpu_ring_reset) {
+		adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+		adev->gfx.gfx_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+	}
 
-		kiq = &adev->gfx.kiq;
-		r = amdgpu_gfx_kiq_init_ring(adev, &kiq->ring, &kiq->irq);
-		if (r)
-			return r;
+	r = amdgpu_gfx_kiq_init(adev, GFX10_MEC_HPD_SIZE, 0);
+	if (r) {
+		DRM_ERROR("Failed to init KIQ BOs!\n");
+		return r;
 	}
 
-	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v10_compute_mqd));
+	r = amdgpu_gfx_kiq_init_ring(adev, xcc_id);
+	if (r)
+		return r;
+
+	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v10_compute_mqd), 0);
 	if (r)
 		return r;
 
@@ -4654,6 +4987,12 @@ static int gfx_v10_0_sw_init(void *handle)
 
 	gfx_v10_0_gpu_early_init(adev);
 
+	gfx_v10_0_alloc_ip_dump(adev);
+
+	r = amdgpu_gfx_sysfs_init(adev);
+	if (r)
+		return r;
+
 	return 0;
 }
 
@@ -4678,22 +5017,22 @@ static void gfx_v10_0_me_fini(struct amdgpu_device *adev)
 			      (void **)&adev->gfx.me.me_fw_ptr);
 }
 
-static int gfx_v10_0_sw_fini(void *handle)
+static int gfx_v10_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.gfx_ring[i]);
 	for (i = 0; i < adev->gfx.num_compute_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
 
-	amdgpu_gfx_mqd_sw_fini(adev);
+	amdgpu_gfx_mqd_sw_fini(adev, 0);
 
-	if (!adev->enable_mes_kiq) {
-		amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring);
-		amdgpu_gfx_kiq_fini(adev);
-	}
+	amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq[0].ring);
+	amdgpu_gfx_kiq_fini(adev, 0);
+
+	amdgpu_gfx_cleaner_shader_sw_fini(adev);
 
 	gfx_v10_0_pfp_fini(adev);
 	gfx_v10_0_ce_fini(adev);
@@ -4705,12 +5044,17 @@ static int gfx_v10_0_sw_fini(void *handle)
 		gfx_v10_0_rlc_backdoor_autoload_buffer_fini(adev);
 
 	gfx_v10_0_free_microcode(adev);
+	amdgpu_gfx_sysfs_fini(adev);
+
+	kfree(adev->gfx.ip_dump_core);
+	kfree(adev->gfx.ip_dump_compute_queues);
+	kfree(adev->gfx.ip_dump_gfx_queues);
 
 	return 0;
 }
 
 static void gfx_v10_0_select_se_sh(struct amdgpu_device *adev, u32 se_num,
-				   u32 sh_num, u32 instance)
+				   u32 sh_num, u32 instance, int xcc_id)
 {
 	u32 data;
 
@@ -4765,18 +5109,21 @@ static void gfx_v10_0_setup_rb(struct amdgpu_device *adev)
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
 			bitmap = i * adev->gfx.config.max_sh_per_se + j;
-			if (((adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 0)) ||
-				(adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 3)) ||
-				(adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 6))) &&
+			if (((amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			      IP_VERSION(10, 3, 0)) ||
+			     (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			      IP_VERSION(10, 3, 3)) ||
+			     (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			      IP_VERSION(10, 3, 6))) &&
 			    ((gfx_v10_3_get_disabled_sa(adev) >> bitmap) & 1))
 				continue;
-			gfx_v10_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v10_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			data = gfx_v10_0_get_rb_active_bitmap(adev);
 			active_rbs |= data << ((i * adev->gfx.config.max_sh_per_se + j) *
 					       rb_bitmap_width_per_sh);
 		}
 	}
-	gfx_v10_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v10_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	adev->gfx.config.backend_enable_mask = active_rbs;
@@ -4793,8 +5140,9 @@ static u32 gfx_v10_0_init_pa_sc_tile_steering_override(struct amdgpu_device *ade
 	uint32_t pa_sc_tile_steering_override;
 
 	/* for ASICs that integrates GFX v10.3
-	 * pa_sc_tile_steering_override should be set to 0 */
-	if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0))
+	 * pa_sc_tile_steering_override should be set to 0
+	 */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 3, 0))
 		return 0;
 
 	/* init num_sc */
@@ -4823,6 +5171,29 @@ static u32 gfx_v10_0_init_pa_sc_tile_steering_override(struct amdgpu_device *ade
 
 #define DEFAULT_SH_MEM_BASES	(0x6000)
 
+static void gfx_v10_0_debug_trap_config_init(struct amdgpu_device *adev,
+				uint32_t first_vmid,
+				uint32_t last_vmid)
+{
+	uint32_t data;
+	uint32_t trap_config_vmid_mask = 0;
+	int i;
+
+	/* Calculate trap config vmid mask */
+	for (i = first_vmid; i < last_vmid; i++)
+		trap_config_vmid_mask |= (1 << i);
+
+	data = REG_SET_FIELD(0, SPI_GDBG_TRAP_CONFIG,
+			VMID_SEL, trap_config_vmid_mask);
+	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_CONFIG,
+			TRAP_EN, 1);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_CONFIG), data);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA0), 0);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA1), 0);
+}
+
 static void gfx_v10_0_init_compute_vmid(struct amdgpu_device *adev)
 {
 	int i;
@@ -4846,14 +5217,19 @@ static void gfx_v10_0_init_compute_vmid(struct amdgpu_device *adev)
 	nv_grbm_select(adev, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
 
-	/* Initialize all compute VMIDs to have no GDS, GWS, or OA
-	   access. These should be enabled by FW for target VMIDs. */
+	/*
+	 * Initialize all compute VMIDs to have no GDS, GWS, or OA
+	 * access. These should be enabled by FW for target VMIDs.
+	 */
 	for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) {
 		WREG32_SOC15_OFFSET(GC, 0, mmGDS_VMID0_BASE, 2 * i, 0);
 		WREG32_SOC15_OFFSET(GC, 0, mmGDS_VMID0_SIZE, 2 * i, 0);
 		WREG32_SOC15_OFFSET(GC, 0, mmGDS_GWS_VMID0, i, 0);
 		WREG32_SOC15_OFFSET(GC, 0, mmGDS_OA_VMID0, i, 0);
 	}
+
+	gfx_v10_0_debug_trap_config_init(adev, adev->vm_manager.first_kfd_vmid,
+					AMDGPU_NUM_VMID);
 }
 
 static void gfx_v10_0_init_gds_vmid(struct amdgpu_device *adev)
@@ -4905,7 +5281,7 @@ static void gfx_v10_0_tcp_harvest(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v10_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v10_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			wgp_active_bitmap = gfx_v10_0_get_wgp_active_bitmap_per_sh(adev);
 			/*
 			 * Set corresponding TCP bits for the inactive WGPs in
@@ -4938,7 +5314,7 @@ static void gfx_v10_0_tcp_harvest(struct amdgpu_device *adev)
 		}
 	}
 
-	gfx_v10_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v10_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 }
 
@@ -4947,7 +5323,7 @@ static void gfx_v10_0_get_tcc_info(struct amdgpu_device *adev)
 	/* TCCs are global (not instanced). */
 	uint32_t tcc_disable;
 
-	if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0)) {
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 3, 0)) {
 		tcc_disable = RREG32_SOC15(GC, 0, mmCGTS_TCC_DISABLE_gc_10_3) |
 			      RREG32_SOC15(GC, 0, mmCGTS_USER_TCC_DISABLE_gc_10_3);
 	} else {
@@ -4965,7 +5341,8 @@ static void gfx_v10_0_constants_init(struct amdgpu_device *adev)
 	u32 tmp;
 	int i;
 
-	WREG32_FIELD15(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff);
+	if (!amdgpu_sriov_vf(adev))
+		WREG32_FIELD15(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff);
 
 	gfx_v10_0_setup_rb(adev);
 	gfx_v10_0_get_cu_info(adev, &adev->gfx.cu_info);
@@ -4976,7 +5353,7 @@ static void gfx_v10_0_constants_init(struct amdgpu_device *adev)
 	/* XXX SH_MEM regs */
 	/* where to put LDS, scratch, GPUVM in FSA64 space */
 	mutex_lock(&adev->srbm_mutex);
-	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) {
+	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
 		nv_grbm_select(adev, 0, 0, 0, i);
 		/* CP and shaders */
 		WREG32_SOC15(GC, 0, mmSH_MEM_CONFIG, DEFAULT_SH_MEM_CONFIG);
@@ -4997,26 +5374,74 @@ static void gfx_v10_0_constants_init(struct amdgpu_device *adev)
 
 }
 
+static u32 gfx_v10_0_get_cpg_int_cntl(struct amdgpu_device *adev,
+				      int me, int pipe)
+{
+	if (me != 0)
+		return 0;
+
+	switch (pipe) {
+	case 0:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_INT_CNTL_RING0);
+	case 1:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_INT_CNTL_RING1);
+	default:
+		return 0;
+	}
+}
+
+static u32 gfx_v10_0_get_cpc_int_cntl(struct amdgpu_device *adev,
+				      int me, int pipe)
+{
+	/*
+	 * amdgpu controls only the first MEC. That's why this function only
+	 * handles the setting of interrupts for this specific MEC. All other
+	 * pipes' interrupts are set by amdkfd.
+	 */
+	if (me != 1)
+		return 0;
+
+	switch (pipe) {
+	case 0:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE0_INT_CNTL);
+	case 1:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE1_INT_CNTL);
+	case 2:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE2_INT_CNTL);
+	case 3:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE3_INT_CNTL);
+	default:
+		return 0;
+	}
+}
+
 static void gfx_v10_0_enable_gui_idle_interrupt(struct amdgpu_device *adev,
 					       bool enable)
 {
-	u32 tmp;
+	u32 tmp, cp_int_cntl_reg;
+	int i, j;
 
 	if (amdgpu_sriov_vf(adev))
 		return;
 
-	tmp = RREG32_SOC15(GC, 0, mmCP_INT_CNTL_RING0);
-
-	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE,
-			    enable ? 1 : 0);
-	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE,
-			    enable ? 1 : 0);
-	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE,
-			    enable ? 1 : 0);
-	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE,
-			    enable ? 1 : 0);
-
-	WREG32_SOC15(GC, 0, mmCP_INT_CNTL_RING0, tmp);
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			cp_int_cntl_reg = gfx_v10_0_get_cpg_int_cntl(adev, i, j);
+
+			if (cp_int_cntl_reg) {
+				tmp = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE,
+						    enable ? 1 : 0);
+				WREG32_SOC15_IP(GC, cp_int_cntl_reg, tmp);
+			}
+		}
+	}
 }
 
 static int gfx_v10_0_init_csb(struct amdgpu_device *adev)
@@ -5024,7 +5449,7 @@ static int gfx_v10_0_init_csb(struct amdgpu_device *adev)
 	adev->gfx.rlc.funcs->get_csb_buffer(adev, adev->gfx.rlc.cs_ptr);
 
 	/* csib */
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 2)) {
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 2)) {
 		WREG32_SOC15_RLC(GC, 0, mmRLC_CSIB_ADDR_HI,
 				adev->gfx.rlc.clear_state_gpu_addr >> 32);
 		WREG32_SOC15_RLC(GC, 0, mmRLC_CSIB_ADDR_LO,
@@ -5080,8 +5505,10 @@ static void gfx_v10_0_rlc_smu_handshake_cntl(struct amdgpu_device *adev,
 
 static void gfx_v10_0_rlc_start(struct amdgpu_device *adev)
 {
-	/* TODO: enable rlc & smu handshake until smu
-	 * and gfxoff feature works as expected */
+	/*
+	 * TODO: enable rlc & smu handshake until smu
+	 * and gfxoff feature works as expected
+	 */
 	if (!(amdgpu_pp_feature_mask & PP_GFXOFF_MASK))
 		gfx_v10_0_rlc_smu_handshake_cntl(adev, false);
 
@@ -5104,7 +5531,7 @@ static int gfx_v10_0_rlc_load_microcode(struct amdgpu_device *adev)
 {
 	const struct rlc_firmware_header_v2_0 *hdr;
 	const __le32 *fw_data;
-	unsigned i, fw_size;
+	unsigned int i, fw_size;
 
 	if (!adev->gfx.rlc_fw)
 		return -EINVAL;
@@ -5141,6 +5568,8 @@ static int gfx_v10_0_rlc_resume(struct amdgpu_device *adev)
 
 		gfx_v10_0_init_csb(adev);
 
+		gfx_v10_0_update_spm_vmid_internal(adev, 0xf);
+
 		if (!amdgpu_sriov_vf(adev)) /* enable RLC SRM */
 			gfx_v10_0_rlc_enable_srm(adev);
 	} else {
@@ -5171,6 +5600,8 @@ static int gfx_v10_0_rlc_resume(struct amdgpu_device *adev)
 
 		gfx_v10_0_init_csb(adev);
 
+		gfx_v10_0_update_spm_vmid_internal(adev, 0xf);
+
 		adev->gfx.rlc.funcs->start(adev);
 
 		if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) {
@@ -5179,6 +5610,7 @@ static int gfx_v10_0_rlc_resume(struct amdgpu_device *adev)
 				return r;
 		}
 	}
+
 	return 0;
 }
 
@@ -5646,13 +6078,12 @@ static int gfx_v10_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable)
 	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_HALT, enable ? 0 : 1);
 	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_HALT, enable ? 0 : 1);
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 2)) {
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 2))
 		WREG32_SOC15_RLC(GC, 0, mmCP_ME_CNTL, tmp);
-	} else {
+	else
 		WREG32_SOC15(GC, 0, mmCP_ME_CNTL, tmp);
-	}
 
-	if (adev->job_hang && !enable)
+	if (amdgpu_in_reset(adev) && !enable)
 		return 0;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
@@ -5672,7 +6103,7 @@ static int gfx_v10_0_cp_gfx_load_pfp_microcode(struct amdgpu_device *adev)
 	int r;
 	const struct gfx_firmware_header_v1_0 *pfp_hdr;
 	const __le32 *fw_data;
-	unsigned i, fw_size;
+	unsigned int i, fw_size;
 	uint32_t tmp;
 	uint32_t usec_timeout = 50000;  /* wait for 50ms */
 
@@ -5721,7 +6152,7 @@ static int gfx_v10_0_cp_gfx_load_pfp_microcode(struct amdgpu_device *adev)
 	}
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	tmp = RREG32_SOC15(GC, 0, mmCP_PFP_IC_BASE_CNTL);
 	tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, VMID, 0);
@@ -5750,7 +6181,7 @@ static int gfx_v10_0_cp_gfx_load_ce_microcode(struct amdgpu_device *adev)
 	int r;
 	const struct gfx_firmware_header_v1_0 *ce_hdr;
 	const __le32 *fw_data;
-	unsigned i, fw_size;
+	unsigned int i, fw_size;
 	uint32_t tmp;
 	uint32_t usec_timeout = 50000;  /* wait for 50ms */
 
@@ -5799,7 +6230,7 @@ static int gfx_v10_0_cp_gfx_load_ce_microcode(struct amdgpu_device *adev)
 	}
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	tmp = RREG32_SOC15(GC, 0, mmCP_CE_IC_BASE_CNTL);
 	tmp = REG_SET_FIELD(tmp, CP_CE_IC_BASE_CNTL, VMID, 0);
@@ -5827,7 +6258,7 @@ static int gfx_v10_0_cp_gfx_load_me_microcode(struct amdgpu_device *adev)
 	int r;
 	const struct gfx_firmware_header_v1_0 *me_hdr;
 	const __le32 *fw_data;
-	unsigned i, fw_size;
+	unsigned int i, fw_size;
 	uint32_t tmp;
 	uint32_t usec_timeout = 50000;  /* wait for 50ms */
 
@@ -5876,7 +6307,7 @@ static int gfx_v10_0_cp_gfx_load_me_microcode(struct amdgpu_device *adev)
 	}
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	tmp = RREG32_SOC15(GC, 0, mmCP_ME_IC_BASE_CNTL);
 	tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, VMID, 0);
@@ -6038,7 +6469,7 @@ static void gfx_v10_0_cp_gfx_set_doorbell(struct amdgpu_device *adev,
 		}
 		WREG32_SOC15(GC, 0, mmCP_RB_DOORBELL_CONTROL, tmp);
 	}
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 	case IP_VERSION(10, 3, 2):
 	case IP_VERSION(10, 3, 1):
@@ -6071,7 +6502,6 @@ static int gfx_v10_0_cp_gfx_resume(struct amdgpu_device *adev)
 	u32 tmp;
 	u32 rb_bufsz;
 	u64 rb_addr, rptr_addr, wptr_gpu_addr;
-	u32 i;
 
 	/* Set the write pointer delay */
 	WREG32_SOC15(GC, 0, mmCP_RB_WPTR_DELAY, 0);
@@ -6098,7 +6528,7 @@ static int gfx_v10_0_cp_gfx_resume(struct amdgpu_device *adev)
 	WREG32_SOC15(GC, 0, mmCP_RB0_WPTR, lower_32_bits(ring->wptr));
 	WREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI, upper_32_bits(ring->wptr));
 
-	/* set the wb address wether it's enabled or not */
+	/* set the wb address whether it's enabled or not */
 	rptr_addr = ring->rptr_gpu_addr;
 	WREG32_SOC15(GC, 0, mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr));
 	WREG32_SOC15(GC, 0, mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) &
@@ -6136,7 +6566,7 @@ static int gfx_v10_0_cp_gfx_resume(struct amdgpu_device *adev)
 		ring->wptr = 0;
 		WREG32_SOC15(GC, 0, mmCP_RB1_WPTR, lower_32_bits(ring->wptr));
 		WREG32_SOC15(GC, 0, mmCP_RB1_WPTR_HI, upper_32_bits(ring->wptr));
-		/* Set the wb address wether it's enabled or not */
+		/* Set the wb address whether it's enabled or not */
 		rptr_addr = ring->rptr_gpu_addr;
 		WREG32_SOC15(GC, 0, mmCP_RB1_RPTR_ADDR, lower_32_bits(rptr_addr));
 		WREG32_SOC15(GC, 0, mmCP_RB1_RPTR_ADDR_HI, upper_32_bits(rptr_addr) &
@@ -6166,18 +6596,13 @@ static int gfx_v10_0_cp_gfx_resume(struct amdgpu_device *adev)
 	/* start the ring */
 	gfx_v10_0_cp_gfx_start(adev);
 
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
-		ring = &adev->gfx.gfx_ring[i];
-		ring->sched.ready = true;
-	}
-
 	return 0;
 }
 
 static void gfx_v10_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
 {
 	if (enable) {
-		switch (adev->ip_versions[GC_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 		case IP_VERSION(10, 3, 0):
 		case IP_VERSION(10, 3, 2):
 		case IP_VERSION(10, 3, 1):
@@ -6193,7 +6618,7 @@ static void gfx_v10_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
 			break;
 		}
 	} else {
-		switch (adev->ip_versions[GC_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 		case IP_VERSION(10, 3, 0):
 		case IP_VERSION(10, 3, 2):
 		case IP_VERSION(10, 3, 1):
@@ -6212,7 +6637,7 @@ static void gfx_v10_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
 				      CP_MEC_CNTL__MEC_ME2_HALT_MASK));
 			break;
 		}
-		adev->gfx.kiq.ring.sched.ready = false;
+		adev->gfx.kiq[0].ring.sched.ready = false;
 	}
 	udelay(50);
 }
@@ -6221,7 +6646,7 @@ static int gfx_v10_0_cp_compute_load_microcode(struct amdgpu_device *adev)
 {
 	const struct gfx_firmware_header_v1_0 *mec_hdr;
 	const __le32 *fw_data;
-	unsigned i;
+	unsigned int i;
 	u32 tmp;
 	u32 usec_timeout = 50000; /* Wait for 50 ms */
 
@@ -6257,7 +6682,7 @@ static int gfx_v10_0_cp_compute_load_microcode(struct amdgpu_device *adev)
 	}
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	tmp = RREG32_SOC15(GC, 0, mmCP_CPC_IC_BASE_CNTL);
 	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, CACHE_POLICY, 0);
@@ -6293,7 +6718,7 @@ static void gfx_v10_0_kiq_setting(struct amdgpu_ring *ring)
 	struct amdgpu_device *adev = ring->adev;
 
 	/* tell RLC which is KIQ queue */
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 	case IP_VERSION(10, 3, 2):
 	case IP_VERSION(10, 3, 1):
@@ -6305,17 +6730,13 @@ static void gfx_v10_0_kiq_setting(struct amdgpu_ring *ring)
 		tmp = RREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS_Sienna_Cichlid);
 		tmp &= 0xffffff00;
 		tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
-		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS_Sienna_Cichlid, tmp);
-		tmp |= 0x80;
-		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS_Sienna_Cichlid, tmp);
+		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS_Sienna_Cichlid, tmp | 0x80);
 		break;
 	default:
 		tmp = RREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS);
 		tmp &= 0xffffff00;
 		tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
-		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS, tmp);
-		tmp |= 0x80;
-		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS, tmp);
+		WREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS, tmp | 0x80);
 		break;
 	}
 }
@@ -6421,62 +6842,13 @@ static int gfx_v10_0_gfx_mqd_init(struct amdgpu_device *adev, void *m,
 	return 0;
 }
 
-#ifdef BRING_UP_DEBUG
-static int gfx_v10_0_gfx_queue_init_register(struct amdgpu_ring *ring)
-{
-	struct amdgpu_device *adev = ring->adev;
-	struct v10_gfx_mqd *mqd = ring->mqd_ptr;
-
-	/* set mmCP_GFX_HQD_WPTR/_HI to 0 */
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_WPTR, mqd->cp_gfx_hqd_wptr);
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_WPTR_HI, mqd->cp_gfx_hqd_wptr_hi);
-
-	/* set GFX_MQD_BASE */
-	WREG32_SOC15(GC, 0, mmCP_MQD_BASE_ADDR, mqd->cp_mqd_base_addr);
-	WREG32_SOC15(GC, 0, mmCP_MQD_BASE_ADDR_HI, mqd->cp_mqd_base_addr_hi);
-
-	/* set GFX_MQD_CONTROL */
-	WREG32_SOC15(GC, 0, mmCP_GFX_MQD_CONTROL, mqd->cp_gfx_mqd_control);
-
-	/* set GFX_HQD_VMID to 0 */
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_VMID, mqd->cp_gfx_hqd_vmid);
-
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_QUEUE_PRIORITY,
-			mqd->cp_gfx_hqd_queue_priority);
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_QUANTUM, mqd->cp_gfx_hqd_quantum);
-
-	/* set GFX_HQD_BASE, similar as CP_RB_BASE */
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_BASE, mqd->cp_gfx_hqd_base);
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_BASE_HI, mqd->cp_gfx_hqd_base_hi);
-
-	/* set GFX_HQD_RPTR_ADDR, similar as CP_RB_RPTR */
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_RPTR_ADDR, mqd->cp_gfx_hqd_rptr_addr);
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_RPTR_ADDR_HI, mqd->cp_gfx_hqd_rptr_addr_hi);
-
-	/* set GFX_HQD_CNTL, similar as CP_RB_CNTL */
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_CNTL, mqd->cp_gfx_hqd_cntl);
-
-	/* set RB_WPTR_POLL_ADDR */
-	WREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_ADDR_LO, mqd->cp_rb_wptr_poll_addr_lo);
-	WREG32_SOC15(GC, 0, mmCP_RB_WPTR_POLL_ADDR_HI, mqd->cp_rb_wptr_poll_addr_hi);
-
-	/* set RB_DOORBELL_CONTROL */
-	WREG32_SOC15(GC, 0, mmCP_RB_DOORBELL_CONTROL, mqd->cp_rb_doorbell_control);
-
-	/* active the queue */
-	WREG32_SOC15(GC, 0, mmCP_GFX_HQD_ACTIVE, mqd->cp_gfx_hqd_active);
-
-	return 0;
-}
-#endif
-
-static int gfx_v10_0_gfx_init_queue(struct amdgpu_ring *ring)
+static int gfx_v10_0_kgq_init_queue(struct amdgpu_ring *ring, bool reset)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v10_gfx_mqd *mqd = ring->mqd_ptr;
 	int mqd_idx = ring - &adev->gfx.gfx_ring[0];
 
-	if (!amdgpu_in_reset(adev) && !adev->in_suspend) {
+	if (!reset && !amdgpu_in_reset(adev) && !adev->in_suspend) {
 		memset((void *)mqd, 0, sizeof(*mqd));
 		mutex_lock(&adev->srbm_mutex);
 		nv_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
@@ -6490,96 +6862,45 @@ static int gfx_v10_0_gfx_init_queue(struct amdgpu_ring *ring)
 		if (ring->doorbell_index == adev->doorbell_index.gfx_ring0 << 1)
 			gfx_v10_0_cp_gfx_set_doorbell(adev, ring);
 
-#ifdef BRING_UP_DEBUG
-		gfx_v10_0_gfx_queue_init_register(ring);
-#endif
 		nv_grbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 		if (adev->gfx.me.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.me.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
-	} else if (amdgpu_in_reset(adev)) {
-		/* reset mqd with the backup copy */
-		if (adev->gfx.me.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.me.mqd_backup[mqd_idx], sizeof(*mqd));
-		/* reset the ring */
-		ring->wptr = 0;
-		*ring->wptr_cpu_addr = 0;
-		amdgpu_ring_clear_ring(ring);
-#ifdef BRING_UP_DEBUG
+			memcpy_fromio(adev->gfx.me.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+	} else {
 		mutex_lock(&adev->srbm_mutex);
 		nv_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
-		gfx_v10_0_gfx_queue_init_register(ring);
+		if (ring->doorbell_index == adev->doorbell_index.gfx_ring0 << 1)
+			gfx_v10_0_cp_gfx_set_doorbell(adev, ring);
+
 		nv_grbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
-#endif
-	} else {
+		/* restore mqd with the backup copy */
+		if (adev->gfx.me.mqd_backup[mqd_idx])
+			memcpy_toio(mqd, adev->gfx.me.mqd_backup[mqd_idx], sizeof(*mqd));
+		/* reset the ring */
+		ring->wptr = 0;
+		*ring->wptr_cpu_addr = 0;
 		amdgpu_ring_clear_ring(ring);
 	}
 
 	return 0;
 }
 
-#ifndef BRING_UP_DEBUG
-static int gfx_v10_0_kiq_enable_kgq(struct amdgpu_device *adev)
-{
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
-	int r, i;
-
-	if (!kiq->pmf || !kiq->pmf->kiq_map_queues)
-		return -EINVAL;
-
-	r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size *
-					adev->gfx.num_gfx_rings);
-	if (r) {
-		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
-		return r;
-	}
-
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
-		kiq->pmf->kiq_map_queues(kiq_ring, &adev->gfx.gfx_ring[i]);
-
-	return amdgpu_ring_test_helper(kiq_ring);
-}
-#endif
-
 static int gfx_v10_0_cp_async_gfx_ring_resume(struct amdgpu_device *adev)
 {
 	int r, i;
-	struct amdgpu_ring *ring;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
-		ring = &adev->gfx.gfx_ring[i];
-
-		r = amdgpu_bo_reserve(ring->mqd_obj, false);
-		if (unlikely(r != 0))
-			goto done;
-
-		r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
-		if (!r) {
-			r = gfx_v10_0_gfx_init_queue(ring);
-			amdgpu_bo_kunmap(ring->mqd_obj);
-			ring->mqd_ptr = NULL;
-		}
-		amdgpu_bo_unreserve(ring->mqd_obj);
+		r = gfx_v10_0_kgq_init_queue(&adev->gfx.gfx_ring[i], false);
 		if (r)
-			goto done;
+			return r;
 	}
-#ifndef BRING_UP_DEBUG
-	r = gfx_v10_0_kiq_enable_kgq(adev);
-	if (r)
-		goto done;
-#endif
-	r = gfx_v10_0_cp_gfx_start(adev);
+
+	r = amdgpu_gfx_enable_kgq(adev, 0);
 	if (r)
-		goto done;
+		return r;
 
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
-		ring = &adev->gfx.gfx_ring[i];
-		ring->sched.ready = true;
-	}
-done:
-	return r;
+	return gfx_v10_0_cp_gfx_start(adev);
 }
 
 static int gfx_v10_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
@@ -6656,8 +6977,9 @@ static int gfx_v10_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
 #ifdef __BIG_ENDIAN
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, ENDIAN_SWAP, 1);
 #endif
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 0);
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TUNNEL_DISPATCH, 0);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 1);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TUNNEL_DISPATCH,
+			    prop->allow_tunneling);
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1);
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1);
 	mqd->cp_hqd_pq_control = tmp;
@@ -6810,14 +7132,13 @@ static int gfx_v10_0_kiq_init_queue(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v10_compute_mqd *mqd = ring->mqd_ptr;
-	int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS;
 
 	gfx_v10_0_kiq_setting(ring);
 
 	if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
 		/* reset MQD to a clean status */
-		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
+		if (adev->gfx.kiq[0].mqd_backup)
+			memcpy_toio(mqd, adev->gfx.kiq[0].mqd_backup, sizeof(*mqd));
 
 		/* reset ring buffer */
 		ring->wptr = 0;
@@ -6839,20 +7160,20 @@ static int gfx_v10_0_kiq_init_queue(struct amdgpu_ring *ring)
 		nv_grbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 
-		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+		if (adev->gfx.kiq[0].mqd_backup)
+			memcpy_fromio(adev->gfx.kiq[0].mqd_backup, mqd, sizeof(*mqd));
 	}
 
 	return 0;
 }
 
-static int gfx_v10_0_kcq_init_queue(struct amdgpu_ring *ring)
+static int gfx_v10_0_kcq_init_queue(struct amdgpu_ring *ring, bool restore)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v10_compute_mqd *mqd = ring->mqd_ptr;
 	int mqd_idx = ring - &adev->gfx.compute_ring[0];
 
-	if (!amdgpu_in_reset(adev) && !adev->in_suspend) {
+	if (!restore && !amdgpu_in_reset(adev) && !adev->in_suspend) {
 		memset((void *)mqd, 0, sizeof(*mqd));
 		mutex_lock(&adev->srbm_mutex);
 		nv_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
@@ -6861,18 +7182,15 @@ static int gfx_v10_0_kcq_init_queue(struct amdgpu_ring *ring)
 		mutex_unlock(&adev->srbm_mutex);
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
-	} else if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
-		/* reset MQD to a clean status */
+			memcpy_fromio(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+	} else {
+		/* restore MQD to a clean status */
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
-
+			memcpy_toio(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
 		/* reset ring buffer */
 		ring->wptr = 0;
 		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, 0);
 		amdgpu_ring_clear_ring(ring);
-	} else {
-		amdgpu_ring_clear_ring(ring);
 	}
 
 	return 0;
@@ -6880,54 +7198,24 @@ static int gfx_v10_0_kcq_init_queue(struct amdgpu_ring *ring)
 
 static int gfx_v10_0_kiq_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring;
-	int r;
-
-	ring = &adev->gfx.kiq.ring;
-
-	r = amdgpu_bo_reserve(ring->mqd_obj, false);
-	if (unlikely(r != 0))
-		return r;
-
-	r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
-	if (unlikely(r != 0))
-		return r;
-
-	gfx_v10_0_kiq_init_queue(ring);
-	amdgpu_bo_kunmap(ring->mqd_obj);
-	ring->mqd_ptr = NULL;
-	amdgpu_bo_unreserve(ring->mqd_obj);
-	ring->sched.ready = true;
+	gfx_v10_0_kiq_init_queue(&adev->gfx.kiq[0].ring);
 	return 0;
 }
 
 static int gfx_v10_0_kcq_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring = NULL;
-	int r = 0, i;
+	int i, r;
 
 	gfx_v10_0_cp_compute_enable(adev, true);
 
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-		ring = &adev->gfx.compute_ring[i];
-
-		r = amdgpu_bo_reserve(ring->mqd_obj, false);
-		if (unlikely(r != 0))
-			goto done;
-		r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
-		if (!r) {
-			r = gfx_v10_0_kcq_init_queue(ring);
-			amdgpu_bo_kunmap(ring->mqd_obj);
-			ring->mqd_ptr = NULL;
-		}
-		amdgpu_bo_unreserve(ring->mqd_obj);
+		r = gfx_v10_0_kcq_init_queue(&adev->gfx.compute_ring[i],
+					     false);
 		if (r)
-			goto done;
+			return r;
 	}
 
-	r = amdgpu_gfx_enable_kcq(adev);
-done:
-	return r;
+	return amdgpu_gfx_enable_kcq(adev, 0);
 }
 
 static int gfx_v10_0_cp_resume(struct amdgpu_device *adev)
@@ -6949,10 +7237,7 @@ static int gfx_v10_0_cp_resume(struct amdgpu_device *adev)
 			return r;
 	}
 
-	if (adev->enable_mes_kiq && adev->mes.kiq_hw_init)
-		r = amdgpu_mes_kiq_hw_init(adev);
-	else
-		r = gfx_v10_0_kiq_resume(adev);
+	r = gfx_v10_0_kiq_resume(adev);
 	if (r)
 		return r;
 
@@ -6997,9 +7282,11 @@ static bool gfx_v10_0_check_grbm_cam_remapping(struct amdgpu_device *adev)
 {
 	uint32_t data, pattern = 0xDEADBEEF;
 
-	/* check if mmVGT_ESGS_RING_SIZE_UMD
-	 * has been remapped to mmVGT_ESGS_RING_SIZE */
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	/*
+	 * check if mmVGT_ESGS_RING_SIZE_UMD
+	 * has been remapped to mmVGT_ESGS_RING_SIZE
+	 */
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 	case IP_VERSION(10, 3, 2):
 	case IP_VERSION(10, 3, 4):
@@ -7009,12 +7296,10 @@ static bool gfx_v10_0_check_grbm_cam_remapping(struct amdgpu_device *adev)
 		WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_UMD, pattern);
 
 		if (RREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_Sienna_Cichlid) == pattern) {
-			WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_UMD , data);
+			WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_UMD, data);
 			return true;
-		} else {
-			WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_Sienna_Cichlid, data);
-			return false;
 		}
+		WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_Sienna_Cichlid, data);
 		break;
 	case IP_VERSION(10, 3, 1):
 	case IP_VERSION(10, 3, 3):
@@ -7029,12 +7314,12 @@ static bool gfx_v10_0_check_grbm_cam_remapping(struct amdgpu_device *adev)
 		if (RREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE) == pattern) {
 			WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE_UMD, data);
 			return true;
-		} else {
-			WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE, data);
-			return false;
 		}
+		WREG32_SOC15(GC, 0, mmVGT_ESGS_RING_SIZE, data);
 		break;
 	}
+
+	return false;
 }
 
 static void gfx_v10_0_setup_grbm_cam_remapping(struct amdgpu_device *adev)
@@ -7044,11 +7329,13 @@ static void gfx_v10_0_setup_grbm_cam_remapping(struct amdgpu_device *adev)
 	if (amdgpu_sriov_vf(adev))
 		return;
 
-	/* initialize cam_index to 0
-	 * index will auto-inc after each data writting */
+	/*
+	 * Initialize cam_index to 0
+	 * index will auto-inc after each data writing
+	 */
 	WREG32_SOC15(GC, 0, mmGRBM_CAM_INDEX, 0);
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 	case IP_VERSION(10, 3, 2):
 	case IP_VERSION(10, 3, 1):
@@ -7175,6 +7462,7 @@ static void gfx_v10_0_setup_grbm_cam_remapping(struct amdgpu_device *adev)
 static void gfx_v10_0_disable_gpa_mode(struct amdgpu_device *adev)
 {
 	uint32_t data;
+
 	data = RREG32_SOC15(GC, 0, mmCPC_PSP_DEBUG);
 	data |= CPC_PSP_DEBUG__GPA_OVERRIDE_MASK;
 	WREG32_SOC15(GC, 0, mmCPC_PSP_DEBUG, data);
@@ -7184,25 +7472,26 @@ static void gfx_v10_0_disable_gpa_mode(struct amdgpu_device *adev)
 	WREG32_SOC15(GC, 0, mmCPG_PSP_DEBUG, data);
 }
 
-static int gfx_v10_0_hw_init(void *handle)
+static int gfx_v10_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (!amdgpu_emu_mode)
 		gfx_v10_0_init_golden_registers(adev);
 
+	amdgpu_gfx_cleaner_shader_init(adev, adev->gfx.cleaner_shader_size,
+				       adev->gfx.cleaner_shader_ptr);
+
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
 		/**
 		 * For gfx 10, rlc firmware loading relies on smu firmware is
 		 * loaded firstly, so in direct type, it has to load smc ucode
 		 * here before rlc.
 		 */
-		if (!(adev->flags & AMD_IS_APU)) {
-			r = amdgpu_pm_load_smu_firmware(adev, NULL);
-			if (r)
-				return r;
-		}
+		r = amdgpu_pm_load_smu_firmware(adev, NULL);
+		if (r)
+			return r;
 		gfx_v10_0_disable_gpa_mode(adev);
 	}
 
@@ -7220,65 +7509,48 @@ static int gfx_v10_0_hw_init(void *handle)
 	 * init golden registers and rlc resume may override some registers,
 	 * reconfig them here
 	 */
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 10) ||
-	    adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 1) ||
-	    adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 2))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 10) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 1) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 2))
 		gfx_v10_0_tcp_harvest(adev);
 
 	r = gfx_v10_0_cp_resume(adev);
 	if (r)
 		return r;
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 0))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 3, 0))
 		gfx_v10_3_program_pbb_mode(adev);
 
-	if (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(10, 3, 0))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 3, 0) && !amdgpu_sriov_vf(adev))
 		gfx_v10_3_set_power_brake_sequence(adev);
 
 	return r;
 }
 
-#ifndef BRING_UP_DEBUG
-static int gfx_v10_0_kiq_disable_kgq(struct amdgpu_device *adev)
+static int gfx_v10_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-	struct amdgpu_ring *kiq_ring = &kiq->ring;
-	int i;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
-		return -EINVAL;
-
-	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size *
-					adev->gfx.num_gfx_rings))
-		return -ENOMEM;
-
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
-		kiq->pmf->kiq_unmap_queues(kiq_ring, &adev->gfx.gfx_ring[i],
-					   PREEMPT_QUEUES, 0, 0);
-	if (!adev->job_hang)
-		return amdgpu_ring_test_helper(kiq_ring);
-	else
-		return 0;
-}
-#endif
-
-static int gfx_v10_0_hw_fini(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int r;
+	cancel_delayed_work_sync(&adev->gfx.idle_work);
 
 	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
+	amdgpu_irq_put(adev, &adev->gfx.bad_op_irq, 0);
+
+	/* WA added for Vangogh asic fixing the SMU suspend failure
+	 * It needs to set power gating again during gfxoff control
+	 * otherwise the gfxoff disallowing will be failed to set.
+	 */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 3, 1))
+		gfx_v10_0_set_powergating_state(ip_block, AMD_PG_STATE_UNGATE);
 
 	if (!adev->no_hw_access) {
-#ifndef BRING_UP_DEBUG
 		if (amdgpu_async_gfx_ring) {
-			r = gfx_v10_0_kiq_disable_kgq(adev);
-			if (r)
+			if (amdgpu_gfx_disable_kgq(adev, 0))
 				DRM_ERROR("KGQ disable failed\n");
 		}
-#endif
-		if (amdgpu_gfx_disable_kcq(adev))
+
+		if (amdgpu_gfx_disable_kcq(adev, 0))
 			DRM_ERROR("KCQ disable failed\n");
 	}
 
@@ -7295,19 +7567,19 @@ static int gfx_v10_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int gfx_v10_0_suspend(void *handle)
+static int gfx_v10_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	return gfx_v10_0_hw_fini(handle);
+	return gfx_v10_0_hw_fini(ip_block);
 }
 
-static int gfx_v10_0_resume(void *handle)
+static int gfx_v10_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	return gfx_v10_0_hw_init(handle);
+	return gfx_v10_0_hw_init(ip_block);
 }
 
-static bool gfx_v10_0_is_idle(void *handle)
+static bool gfx_v10_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (REG_GET_FIELD(RREG32_SOC15(GC, 0, mmGRBM_STATUS),
 				GRBM_STATUS, GUI_ACTIVE))
@@ -7316,11 +7588,11 @@ static bool gfx_v10_0_is_idle(void *handle)
 		return true;
 }
 
-static int gfx_v10_0_wait_for_idle(void *handle)
+static int gfx_v10_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	unsigned i;
+	unsigned int i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		/* read MC_STATUS */
@@ -7334,11 +7606,11 @@ static int gfx_v10_0_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int gfx_v10_0_soft_reset(void *handle)
+static int gfx_v10_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 grbm_soft_reset = 0;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* GRBM_STATUS */
 	tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS);
@@ -7363,7 +7635,7 @@ static int gfx_v10_0_soft_reset(void *handle)
 
 	/* GRBM_STATUS2 */
 	tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS2);
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 	case IP_VERSION(10, 3, 2):
 	case IP_VERSION(10, 3, 1):
@@ -7396,19 +7668,17 @@ static int gfx_v10_0_soft_reset(void *handle)
 		/* Disable MEC parsing/prefetching */
 		gfx_v10_0_cp_compute_enable(adev, false);
 
-		if (grbm_soft_reset) {
-			tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
-			tmp |= grbm_soft_reset;
-			dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp);
-			WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
-			tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
+		tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
+		tmp |= grbm_soft_reset;
+		dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
+		tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
 
-			udelay(50);
+		udelay(50);
 
-			tmp &= ~grbm_soft_reset;
-			WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
-			tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
-		}
+		tmp &= ~grbm_soft_reset;
+		WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
+		tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
 
 		/* Wait a little for things to settle down */
 		udelay(50);
@@ -7420,7 +7690,23 @@ static uint64_t gfx_v10_0_get_gpu_clock_counter(struct amdgpu_device *adev)
 {
 	uint64_t clock, clock_lo, clock_hi, hi_check;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(10, 1, 3):
+	case IP_VERSION(10, 1, 4):
+		preempt_disable();
+		clock_hi = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Cyan_Skillfish);
+		clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Cyan_Skillfish);
+		hi_check = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Cyan_Skillfish);
+		/* The SMUIO TSC clock frequency is 100MHz, which sets 32-bit carry over
+		 * roughly every 42 seconds.
+		 */
+		if (hi_check != clock_hi) {
+			clock_lo = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_LOWER_Cyan_Skillfish);
+			clock_hi = hi_check;
+		}
+		preempt_enable();
+		clock = clock_lo | (clock_hi << 32ULL);
+		break;
 	case IP_VERSION(10, 3, 1):
 	case IP_VERSION(10, 3, 3):
 	case IP_VERSION(10, 3, 7):
@@ -7501,13 +7787,13 @@ static void gfx_v10_0_ring_emit_gds_switch(struct amdgpu_ring *ring,
 				    (1 << (oa_size + oa_base)) - (1 << oa_base));
 }
 
-static int gfx_v10_0_early_init(void *handle)
+static int gfx_v10_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->gfx.funcs = &gfx_v10_0_gfx_funcs;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 1):
 	case IP_VERSION(10, 1, 2):
@@ -7545,9 +7831,9 @@ static int gfx_v10_0_early_init(void *handle)
 	return gfx_v10_0_init_microcode(adev);
 }
 
-static int gfx_v10_0_late_init(void *handle)
+static int gfx_v10_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0);
@@ -7558,6 +7844,10 @@ static int gfx_v10_0_late_init(void *handle)
 	if (r)
 		return r;
 
+	r = amdgpu_irq_get(adev, &adev->gfx.bad_op_irq, 0);
+	if (r)
+		return r;
+
 	return 0;
 }
 
@@ -7570,15 +7860,15 @@ static bool gfx_v10_0_is_rlc_enabled(struct amdgpu_device *adev)
 	return (REG_GET_FIELD(rlc_cntl, RLC_CNTL, RLC_ENABLE_F32)) ? true : false;
 }
 
-static void gfx_v10_0_set_safe_mode(struct amdgpu_device *adev)
+static void gfx_v10_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	uint32_t data;
-	unsigned i;
+	unsigned int i;
 
 	data = RLC_SAFE_MODE__CMD_MASK;
 	data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT);
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 	case IP_VERSION(10, 3, 2):
 	case IP_VERSION(10, 3, 1):
@@ -7611,12 +7901,12 @@ static void gfx_v10_0_set_safe_mode(struct amdgpu_device *adev)
 	}
 }
 
-static void gfx_v10_0_unset_safe_mode(struct amdgpu_device *adev)
+static void gfx_v10_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	uint32_t data;
 
 	data = RLC_SAFE_MODE__CMD_MASK;
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 	case IP_VERSION(10, 3, 2):
 	case IP_VERSION(10, 3, 1):
@@ -7927,7 +8217,7 @@ static void gfx_v10_0_apply_medium_grain_clock_gating_workaround(struct amdgpu_d
 		mmCGTS_SA1_QUAD1_SM_CTRL_REG
 	};
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 2)) {
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 2)) {
 		for (i = 0; i < ARRAY_SIZE(tcp_ctrl_regs_nv12); i++) {
 			reg_idx = adev->reg_offset[GC_HWIP][0][mmCGTS_SA0_WGP00_CU0_TCP_CTRL_REG_BASE_IDX] +
 				  tcp_ctrl_regs_nv12[i];
@@ -7958,7 +8248,7 @@ static void gfx_v10_0_apply_medium_grain_clock_gating_workaround(struct amdgpu_d
 static int gfx_v10_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 					    bool enable)
 {
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 
 	if (enable) {
 		/* enable FGCG firstly*/
@@ -7972,9 +8262,12 @@ static int gfx_v10_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 		/* ===  CGCG + CGLS === */
 		gfx_v10_0_update_coarse_grain_clock_gating(adev, enable);
 
-		if ((adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 10)) ||
-		    (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 1)) ||
-		    (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 2)))
+		if ((amdgpu_ip_version(adev, GC_HWIP, 0) ==
+		     IP_VERSION(10, 1, 10)) ||
+		    (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+		     IP_VERSION(10, 1, 1)) ||
+		    (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+		     IP_VERSION(10, 1, 2)))
 			gfx_v10_0_apply_medium_grain_clock_gating_workaround(adev);
 	} else {
 		/* CGCG/CGLS should be disabled before MGCG/MGLS
@@ -7997,31 +8290,39 @@ static int gfx_v10_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 	     AMD_CG_SUPPORT_GFX_3D_CGLS))
 		gfx_v10_0_enable_gui_idle_interrupt(adev, enable);
 
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 
 	return 0;
 }
 
-static void gfx_v10_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid)
+static void gfx_v10_0_update_spm_vmid_internal(struct amdgpu_device *adev,
+					       unsigned int vmid)
 {
-	u32 reg, data;
-
-	amdgpu_gfx_off_ctrl(adev, false);
+	u32 reg, pre_data, data;
 
-	/* not for *_SOC15 */
 	reg = SOC15_REG_OFFSET(GC, 0, mmRLC_SPM_MC_CNTL);
-	if (amdgpu_sriov_is_pp_one_vf(adev))
-		data = RREG32_NO_KIQ(reg);
+	/* not for *_SOC15 */
+	if (amdgpu_sriov_is_pp_one_vf(adev) && !amdgpu_sriov_runtime(adev))
+		pre_data = RREG32_NO_KIQ(reg);
 	else
-		data = RREG32_SOC15(GC, 0, mmRLC_SPM_MC_CNTL);
+		pre_data = RREG32(reg);
 
-	data &= ~RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK;
+	data = pre_data & (~RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK);
 	data |= (vmid & RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK) << RLC_SPM_MC_CNTL__RLC_SPM_VMID__SHIFT;
 
-	if (amdgpu_sriov_is_pp_one_vf(adev))
-		WREG32_SOC15_NO_KIQ(GC, 0, mmRLC_SPM_MC_CNTL, data);
-	else
-		WREG32_SOC15(GC, 0, mmRLC_SPM_MC_CNTL, data);
+	if (pre_data != data) {
+		if (amdgpu_sriov_is_pp_one_vf(adev) && !amdgpu_sriov_runtime(adev)) {
+			WREG32_SOC15_NO_KIQ(GC, 0, mmRLC_SPM_MC_CNTL, data);
+		} else
+			WREG32_SOC15(GC, 0, mmRLC_SPM_MC_CNTL, data);
+	}
+}
+
+static void gfx_v10_0_update_spm_vmid(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned int vmid)
+{
+	amdgpu_gfx_off_ctrl(adev, false);
+
+	gfx_v10_0_update_spm_vmid_internal(adev, vmid);
 
 	amdgpu_gfx_off_ctrl(adev, true);
 }
@@ -8075,7 +8376,7 @@ static void gfx_v10_cntl_power_gating(struct amdgpu_device *adev, bool enable)
 	 * Power/performance team will optimize it and might give a new value later.
 	 */
 	if (enable && (adev->pg_flags & AMD_PG_SUPPORT_GFX_PG)) {
-		switch (adev->ip_versions[GC_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 		case IP_VERSION(10, 3, 1):
 		case IP_VERSION(10, 3, 3):
 		case IP_VERSION(10, 3, 6):
@@ -8091,11 +8392,11 @@ static void gfx_v10_cntl_power_gating(struct amdgpu_device *adev, bool enable)
 
 static void gfx_v10_cntl_pg(struct amdgpu_device *adev, bool enable)
 {
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 
 	gfx_v10_cntl_power_gating(adev, enable);
 
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 }
 
 static const struct amdgpu_rlc_funcs gfx_v10_0_rlc_funcs = {
@@ -8127,16 +8428,16 @@ static const struct amdgpu_rlc_funcs gfx_v10_0_rlc_funcs_sriov = {
 	.is_rlcg_access_range = gfx_v10_0_is_rlcg_access_range,
 };
 
-static int gfx_v10_0_set_powergating_state(void *handle,
+static int gfx_v10_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_PG_STATE_GATE);
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 1):
 	case IP_VERSION(10, 1, 2):
@@ -8150,8 +8451,14 @@ static int gfx_v10_0_set_powergating_state(void *handle,
 	case IP_VERSION(10, 3, 3):
 	case IP_VERSION(10, 3, 6):
 	case IP_VERSION(10, 3, 7):
+		if (!enable)
+			amdgpu_gfx_off_ctrl(adev, false);
+
 		gfx_v10_cntl_pg(adev, enable);
-		amdgpu_gfx_off_ctrl(adev, enable);
+
+		if (enable)
+			amdgpu_gfx_off_ctrl(adev, true);
+
 		break;
 	default:
 		break;
@@ -8159,15 +8466,15 @@ static int gfx_v10_0_set_powergating_state(void *handle,
 	return 0;
 }
 
-static int gfx_v10_0_set_clockgating_state(void *handle,
+static int gfx_v10_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 1):
 	case IP_VERSION(10, 1, 2):
@@ -8188,9 +8495,9 @@ static int gfx_v10_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static void gfx_v10_0_get_clockgating_state(void *handle, u64 *flags)
+static void gfx_v10_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	/* AMD_CG_SUPPORT_GFX_FGCG */
@@ -8257,45 +8564,17 @@ static u64 gfx_v10_0_ring_get_wptr_gfx(struct amdgpu_ring *ring)
 static void gfx_v10_0_ring_set_wptr_gfx(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t *wptr_saved;
-	uint32_t *is_queue_unmap;
-	uint64_t aggregated_db_index;
-	uint32_t mqd_size = adev->mqds[AMDGPU_HW_IP_GFX].mqd_size;
-	uint64_t wptr_tmp;
-
-	if (ring->is_mes_queue) {
-		wptr_saved = (uint32_t *)(ring->mqd_ptr + mqd_size);
-		is_queue_unmap = (uint32_t *)(ring->mqd_ptr + mqd_size +
-					      sizeof(uint32_t));
-		aggregated_db_index =
-			amdgpu_mes_get_aggregated_doorbell_index(adev,
-			AMDGPU_MES_PRIORITY_LEVEL_NORMAL);
-
-		wptr_tmp = ring->wptr & ring->buf_mask;
-		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, wptr_tmp);
-		*wptr_saved = wptr_tmp;
-		/* assume doorbell always being used by mes mapped queue */
-		if (*is_queue_unmap) {
-			WDOORBELL64(aggregated_db_index, wptr_tmp);
-			WDOORBELL64(ring->doorbell_index, wptr_tmp);
-		} else {
-			WDOORBELL64(ring->doorbell_index, wptr_tmp);
 
-			if (*is_queue_unmap)
-				WDOORBELL64(aggregated_db_index, wptr_tmp);
-		}
+	if (ring->use_doorbell) {
+		/* XXX check if swapping is necessary on BE */
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
+			     ring->wptr);
+		WDOORBELL64(ring->doorbell_index, ring->wptr);
 	} else {
-		if (ring->use_doorbell) {
-			/* XXX check if swapping is necessary on BE */
-			atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
-				     ring->wptr);
-			WDOORBELL64(ring->doorbell_index, ring->wptr);
-		} else {
-			WREG32_SOC15(GC, 0, mmCP_RB0_WPTR,
-				     lower_32_bits(ring->wptr));
-			WREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI,
-				     upper_32_bits(ring->wptr));
-		}
+		WREG32_SOC15(GC, 0, mmCP_RB0_WPTR,
+			     lower_32_bits(ring->wptr));
+		WREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI,
+			     upper_32_bits(ring->wptr));
 	}
 }
 
@@ -8320,42 +8599,13 @@ static u64 gfx_v10_0_ring_get_wptr_compute(struct amdgpu_ring *ring)
 static void gfx_v10_0_ring_set_wptr_compute(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t *wptr_saved;
-	uint32_t *is_queue_unmap;
-	uint64_t aggregated_db_index;
-	uint32_t mqd_size = adev->mqds[AMDGPU_HW_IP_COMPUTE].mqd_size;
-	uint64_t wptr_tmp;
-
-	if (ring->is_mes_queue) {
-		wptr_saved = (uint32_t *)(ring->mqd_ptr + mqd_size);
-		is_queue_unmap = (uint32_t *)(ring->mqd_ptr + mqd_size +
-					      sizeof(uint32_t));
-		aggregated_db_index =
-			amdgpu_mes_get_aggregated_doorbell_index(adev,
-			AMDGPU_MES_PRIORITY_LEVEL_NORMAL);
-
-		wptr_tmp = ring->wptr & ring->buf_mask;
-		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, wptr_tmp);
-		*wptr_saved = wptr_tmp;
-		/* assume doorbell always used by mes mapped queue */
-		if (*is_queue_unmap) {
-			WDOORBELL64(aggregated_db_index, wptr_tmp);
-			WDOORBELL64(ring->doorbell_index, wptr_tmp);
-		} else {
-			WDOORBELL64(ring->doorbell_index, wptr_tmp);
 
-			if (*is_queue_unmap)
-				WDOORBELL64(aggregated_db_index, wptr_tmp);
-		}
+	if (ring->use_doorbell) {
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
+			     ring->wptr);
+		WDOORBELL64(ring->doorbell_index, ring->wptr);
 	} else {
-		/* XXX check if swapping is necessary on BE */
-		if (ring->use_doorbell) {
-			atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
-				     ring->wptr);
-			WDOORBELL64(ring->doorbell_index, ring->wptr);
-		} else {
-			BUG(); /* only DOORBELL method supported on gfx10 now */
-		}
+		BUG(); /* only DOORBELL method supported on gfx10 now */
 	}
 }
 
@@ -8378,7 +8628,7 @@ static void gfx_v10_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 		}
 		reg_mem_engine = 0;
 	} else {
-		ref_and_mask = nbio_hf_reg->ref_and_mask_cp0;
+		ref_and_mask = nbio_hf_reg->ref_and_mask_cp0 << ring->pipe;
 		reg_mem_engine = 1; /* pfp */
 	}
 
@@ -8393,7 +8643,7 @@ static void gfx_v10_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
 				       struct amdgpu_ib *ib,
 				       uint32_t flags)
 {
-	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
+	unsigned int vmid = AMDGPU_JOB_GET_VMID(job);
 	u32 header, control = 0;
 
 	if (ib->flags & AMDGPU_IB_FLAG_CE)
@@ -8403,7 +8653,7 @@ static void gfx_v10_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
 
 	control |= ib->length_dw | (vmid << 24);
 
-	if (amdgpu_mcbp && (ib->flags & AMDGPU_IB_FLAG_PREEMPT)) {
+	if (ring->adev->gfx.mcbp && (ib->flags & AMDGPU_IB_FLAG_PREEMPT)) {
 		control |= INDIRECT_BUFFER_PRE_ENB(1);
 
 		if (flags & AMDGPU_IB_PREEMPTED)
@@ -8414,10 +8664,6 @@ static void gfx_v10_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
 				    (!amdgpu_sriov_vf(ring->adev) && flags & AMDGPU_IB_PREEMPTED) ? true : false);
 	}
 
-	if (ring->is_mes_queue)
-		/* inherit vmid from mqd */
-		control |= 0x400000;
-
 	amdgpu_ring_write(ring, header);
 	BUG_ON(ib->gpu_addr & 0x3); /* Dword align */
 	amdgpu_ring_write(ring,
@@ -8434,13 +8680,9 @@ static void gfx_v10_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
 					   struct amdgpu_ib *ib,
 					   uint32_t flags)
 {
-	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
+	unsigned int vmid = AMDGPU_JOB_GET_VMID(job);
 	u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24);
 
-	if (ring->is_mes_queue)
-		/* inherit vmid from mqd */
-		control |= 0x40000000;
-
 	/* Currently, there is a high possibility to get wave ID mismatch
 	 * between ME and GDS, leading to a hw deadlock, because ME generates
 	 * different wave IDs than the GDS expects. This situation happens
@@ -8469,7 +8711,7 @@ static void gfx_v10_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
 }
 
 static void gfx_v10_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
-				     u64 seq, unsigned flags)
+				     u64 seq, unsigned int flags)
 {
 	bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT;
 	bool int_sel = flags & AMDGPU_FENCE_FLAG_INT;
@@ -8498,8 +8740,7 @@ static void gfx_v10_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
 	amdgpu_ring_write(ring, upper_32_bits(addr));
 	amdgpu_ring_write(ring, lower_32_bits(seq));
 	amdgpu_ring_write(ring, upper_32_bits(seq));
-	amdgpu_ring_write(ring, ring->is_mes_queue ?
-			 (ring->hw_queue_id | AMDGPU_FENCE_MES_QUEUE_FLAG) : 0);
+	amdgpu_ring_write(ring, 0);
 }
 
 static void gfx_v10_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
@@ -8525,12 +8766,9 @@ static void gfx_v10_0_ring_invalidate_tlbs(struct amdgpu_ring *ring,
 }
 
 static void gfx_v10_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
-					 unsigned vmid, uint64_t pd_addr)
+					 unsigned int vmid, uint64_t pd_addr)
 {
-	if (ring->is_mes_queue)
-		gfx_v10_0_ring_invalidate_tlbs(ring, 0, 0, false, 0);
-	else
-		amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
+	amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
 
 	/* compute doesn't have PFP */
 	if (ring->funcs->type == AMDGPU_RING_TYPE_GFX) {
@@ -8578,7 +8816,7 @@ static void gfx_v10_0_ring_emit_cntxcntl(struct amdgpu_ring *ring,
 {
 	uint32_t dw2 = 0;
 
-	if (amdgpu_mcbp)
+	if (ring->adev->gfx.mcbp)
 		gfx_v10_0_ring_emit_ce_meta(ring,
 				    (!amdgpu_sriov_vf(ring->adev) && flags & AMDGPU_IB_PREEMPTED) ? true : false);
 
@@ -8607,38 +8845,28 @@ static void gfx_v10_0_ring_emit_cntxcntl(struct amdgpu_ring *ring,
 	amdgpu_ring_write(ring, 0);
 }
 
-static unsigned gfx_v10_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned int gfx_v10_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring,
+						       uint64_t addr)
 {
-	unsigned ret;
+	unsigned int ret;
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, 0); /* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	/* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, 0);
 	ret = ring->wptr & ring->buf_mask;
-	amdgpu_ring_write(ring, 0x55aa55aa); /* patch dummy value later */
+	/* patch dummy value later */
+	amdgpu_ring_write(ring, 0);
 
 	return ret;
 }
 
-static void gfx_v10_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigned offset)
-{
-	unsigned cur;
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
-
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (likely(cur > offset))
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->buf_mask + 1) - offset + cur;
-}
-
 static int gfx_v10_0_ring_preempt_ib(struct amdgpu_ring *ring)
 {
 	int i, r = 0;
 	struct amdgpu_device *adev = ring->adev;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
 	struct amdgpu_ring *kiq_ring = &kiq->ring;
 	unsigned long flags;
 
@@ -8691,19 +8919,9 @@ static void gfx_v10_0_ring_emit_ce_meta(struct amdgpu_ring *ring, bool resume)
 
 	cnt = (sizeof(ce_payload) >> 2) + 4 - 2;
 
-	if (ring->is_mes_queue) {
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  gfx[0].gfx_meta_data) +
-			offsetof(struct v10_gfx_meta_data, ce_payload);
-		ce_payload_gpu_addr =
-			amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		ce_payload_cpu_addr =
-			amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-	} else {
-		offset = offsetof(struct v10_gfx_meta_data, ce_payload);
-		ce_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
-		ce_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
-	}
+	offset = offsetof(struct v10_gfx_meta_data, ce_payload);
+	ce_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
+	ce_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, cnt));
 	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(2) |
@@ -8729,28 +8947,13 @@ static void gfx_v10_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume)
 	void *de_payload_cpu_addr;
 	int cnt;
 
-	if (ring->is_mes_queue) {
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  gfx[0].gfx_meta_data) +
-			offsetof(struct v10_gfx_meta_data, de_payload);
-		de_payload_gpu_addr =
-			amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		de_payload_cpu_addr =
-			amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  gfx[0].gds_backup) +
-			offsetof(struct v10_gfx_meta_data, de_payload);
-		gds_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-	} else {
-		offset = offsetof(struct v10_gfx_meta_data, de_payload);
-		de_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
-		de_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
+	offset = offsetof(struct v10_gfx_meta_data, de_payload);
+	de_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
+	de_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
 
-		gds_addr = ALIGN(amdgpu_csa_vaddr(ring->adev) +
-				 AMDGPU_CSA_SIZE - adev->gds.gds_size,
-				 PAGE_SIZE);
-	}
+	gds_addr = ALIGN(amdgpu_csa_vaddr(ring->adev) +
+			 AMDGPU_CSA_SIZE - adev->gds.gds_size,
+			 PAGE_SIZE);
 
 	de_payload.gds_backup_addrlo = lower_32_bits(gds_addr);
 	de_payload.gds_backup_addrhi = upper_32_bits(gds_addr);
@@ -8845,19 +9048,6 @@ static void gfx_v10_0_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring,
 							   ref, mask);
 }
 
-static void gfx_v10_0_ring_soft_recovery(struct amdgpu_ring *ring,
-					 unsigned vmid)
-{
-	struct amdgpu_device *adev = ring->adev;
-	uint32_t value = 0;
-
-	value = REG_SET_FIELD(value, SQ_CMD, CMD, 0x03);
-	value = REG_SET_FIELD(value, SQ_CMD, MODE, 0x01);
-	value = REG_SET_FIELD(value, SQ_CMD, CHECK_VMID, 1);
-	value = REG_SET_FIELD(value, SQ_CMD, VM_ID, vmid);
-	WREG32_SOC15(GC, 0, mmSQ_CMD, value);
-}
-
 static void
 gfx_v10_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev,
 				      uint32_t me, uint32_t pipe,
@@ -8955,7 +9145,7 @@ static void gfx_v10_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev
 
 static int gfx_v10_0_set_eop_interrupt_state(struct amdgpu_device *adev,
 					    struct amdgpu_irq_src *src,
-					    unsigned type,
+					    unsigned int type,
 					    enum amdgpu_interrupt_state state)
 {
 	switch (type) {
@@ -9002,49 +9192,34 @@ static int gfx_v10_0_eop_irq(struct amdgpu_device *adev,
 	int i;
 	u8 me_id, pipe_id, queue_id;
 	struct amdgpu_ring *ring;
-	uint32_t mes_queue_id = entry->src_data[0];
 
 	DRM_DEBUG("IH: CP EOP\n");
 
-	if (adev->enable_mes && (mes_queue_id & AMDGPU_FENCE_MES_QUEUE_FLAG)) {
-		struct amdgpu_mes_queue *queue;
-
-		mes_queue_id &= AMDGPU_FENCE_MES_QUEUE_ID_MASK;
-
-		spin_lock(&adev->mes.queue_id_lock);
-		queue = idr_find(&adev->mes.queue_id_idr, mes_queue_id);
-		if (queue) {
-			DRM_DEBUG("process mes queue id = %d\n", mes_queue_id);
-			amdgpu_fence_process(queue->ring);
-		}
-		spin_unlock(&adev->mes.queue_id_lock);
-	} else {
-		me_id = (entry->ring_id & 0x0c) >> 2;
-		pipe_id = (entry->ring_id & 0x03) >> 0;
-		queue_id = (entry->ring_id & 0x70) >> 4;
+	me_id = (entry->ring_id & 0x0c) >> 2;
+	pipe_id = (entry->ring_id & 0x03) >> 0;
+	queue_id = (entry->ring_id & 0x70) >> 4;
 
-		switch (me_id) {
-		case 0:
-			if (pipe_id == 0)
-				amdgpu_fence_process(&adev->gfx.gfx_ring[0]);
-			else
-				amdgpu_fence_process(&adev->gfx.gfx_ring[1]);
-			break;
-		case 1:
-		case 2:
-			for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-				ring = &adev->gfx.compute_ring[i];
-				/* Per-queue interrupt is supported for MEC starting from VI.
-				 * The interrupt can only be enabled/disabled per pipe instead
-				 * of per queue.
-				 */
-				if ((ring->me == me_id) &&
-				    (ring->pipe == pipe_id) &&
-				    (ring->queue == queue_id))
-					amdgpu_fence_process(ring);
-			}
-			break;
+	switch (me_id) {
+	case 0:
+		if (pipe_id == 0)
+			amdgpu_fence_process(&adev->gfx.gfx_ring[0]);
+		else
+			amdgpu_fence_process(&adev->gfx.gfx_ring[1]);
+		break;
+	case 1:
+	case 2:
+		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+			ring = &adev->gfx.compute_ring[i];
+			/* Per-queue interrupt is supported for MEC starting from VI.
+			 * The interrupt can only be enabled/disabled per pipe instead
+			 * of per queue.
+			 */
+			if ((ring->me == me_id) &&
+			    (ring->pipe == pipe_id) &&
+			    (ring->queue == queue_id))
+				amdgpu_fence_process(ring);
 		}
+		break;
 	}
 
 	return 0;
@@ -9052,15 +9227,42 @@ static int gfx_v10_0_eop_irq(struct amdgpu_device *adev,
 
 static int gfx_v10_0_set_priv_reg_fault_state(struct amdgpu_device *adev,
 					      struct amdgpu_irq_src *source,
-					      unsigned type,
+					      unsigned int type,
 					      enum amdgpu_interrupt_state state)
 {
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
 	case AMDGPU_IRQ_STATE_ENABLE:
-		WREG32_FIELD15(GC, 0, CP_INT_CNTL_RING0,
-			       PRIV_REG_INT_ENABLE,
-			       state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v10_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    PRIV_REG_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				/* MECs start at 1 */
+				cp_int_cntl_reg = gfx_v10_0_get_cpc_int_cntl(adev, i + 1, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+								    PRIV_REG_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
 		break;
 	default:
 		break;
@@ -9069,17 +9271,75 @@ static int gfx_v10_0_set_priv_reg_fault_state(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int gfx_v10_0_set_bad_op_fault_state(struct amdgpu_device *adev,
+					    struct amdgpu_irq_src *source,
+					    unsigned type,
+					    enum amdgpu_interrupt_state state)
+{
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v10_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    OPCODE_ERROR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				/* MECs start at 1 */
+				cp_int_cntl_reg = gfx_v10_0_get_cpc_int_cntl(adev, i + 1, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+								    OPCODE_ERROR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
 static int gfx_v10_0_set_priv_inst_fault_state(struct amdgpu_device *adev,
 					       struct amdgpu_irq_src *source,
-					       unsigned type,
+					       unsigned int type,
 					       enum amdgpu_interrupt_state state)
 {
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
 	case AMDGPU_IRQ_STATE_ENABLE:
-		WREG32_FIELD15(GC, 0, CP_INT_CNTL_RING0,
-			       PRIV_INSTR_INT_ENABLE,
-			       state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v10_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    PRIV_INSTR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
 		break;
 	default:
 		break;
@@ -9103,8 +9363,8 @@ static void gfx_v10_0_handle_priv_fault(struct amdgpu_device *adev,
 	case 0:
 		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
 			ring = &adev->gfx.gfx_ring[i];
-			/* we only enabled 1 gfx queue per pipe for now */
-			if (ring->me == me_id && ring->pipe == pipe_id)
+			if (ring->me == me_id && ring->pipe == pipe_id &&
+			    ring->queue == queue_id)
 				drm_sched_fault(&ring->sched);
 		}
 		break;
@@ -9131,6 +9391,15 @@ static int gfx_v10_0_priv_reg_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int gfx_v10_0_bad_op_irq(struct amdgpu_device *adev,
+				struct amdgpu_irq_src *source,
+				struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal opcode in command stream \n");
+	gfx_v10_0_handle_priv_fault(adev, entry);
+	return 0;
+}
+
 static int gfx_v10_0_priv_inst_irq(struct amdgpu_device *adev,
 				   struct amdgpu_irq_src *source,
 				   struct amdgpu_iv_entry *entry)
@@ -9146,7 +9415,7 @@ static int gfx_v10_0_kiq_set_interrupt_state(struct amdgpu_device *adev,
 					     enum amdgpu_interrupt_state state)
 {
 	uint32_t tmp, target;
-	struct amdgpu_ring *ring = &(adev->gfx.kiq.ring);
+	struct amdgpu_ring *ring = &(adev->gfx.kiq[0].ring);
 
 	if (ring->me == 1)
 		target = SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE0_INT_CNTL);
@@ -9190,7 +9459,7 @@ static int gfx_v10_0_kiq_irq(struct amdgpu_device *adev,
 			     struct amdgpu_iv_entry *entry)
 {
 	u8 me_id, pipe_id, queue_id;
-	struct amdgpu_ring *ring = &(adev->gfx.kiq.ring);
+	struct amdgpu_ring *ring = &(adev->gfx.kiq[0].ring);
 
 	me_id = (entry->ring_id & 0x0c) >> 2;
 	pipe_id = (entry->ring_id & 0x03) >> 0;
@@ -9225,6 +9494,319 @@ static void gfx_v10_0_emit_mem_sync(struct amdgpu_ring *ring)
 	amdgpu_ring_write(ring, gcr_cntl); /* GCR_CNTL */
 }
 
+static void gfx_v10_ring_insert_nop(struct amdgpu_ring *ring, uint32_t num_nop)
+{
+	/* Header itself is a NOP packet */
+	if (num_nop == 1) {
+		amdgpu_ring_write(ring, ring->funcs->nop);
+		return;
+	}
+
+	/* Max HW optimization till 0x3ffe, followed by remaining one NOP at a time*/
+	amdgpu_ring_write(ring, PACKET3(PACKET3_NOP, min(num_nop - 2, 0x3ffe)));
+
+	/* Header is at index 0, followed by num_nops - 1 NOP packet's */
+	amdgpu_ring_insert_nop(ring, num_nop - 1);
+}
+
+static int gfx_v10_0_reset_kgq(struct amdgpu_ring *ring,
+			       unsigned int vmid,
+			       struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	unsigned long flags;
+	u32 tmp;
+	u64 addr;
+	int r;
+
+	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
+		return -EINVAL;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+
+	if (amdgpu_ring_alloc(kiq_ring, 5 + 7 + 7)) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+
+	addr = amdgpu_bo_gpu_offset(ring->mqd_obj) +
+		offsetof(struct v10_gfx_mqd, cp_gfx_hqd_active);
+	tmp = REG_SET_FIELD(0, CP_VMID_RESET, RESET_REQUEST, 1 << vmid);
+	if (ring->pipe == 0)
+		tmp = REG_SET_FIELD(tmp, CP_VMID_RESET, PIPE0_QUEUES, 1 << ring->queue);
+	else
+		tmp = REG_SET_FIELD(tmp, CP_VMID_RESET, PIPE1_QUEUES, 1 << ring->queue);
+
+	gfx_v10_0_ring_emit_wreg(kiq_ring,
+				 SOC15_REG_OFFSET(GC, 0, mmCP_VMID_RESET), tmp);
+	gfx_v10_0_wait_reg_mem(kiq_ring, 0, 1, 0,
+			       lower_32_bits(addr), upper_32_bits(addr),
+			       0, 1, 0x20);
+	gfx_v10_0_ring_emit_reg_wait(kiq_ring,
+				     SOC15_REG_OFFSET(GC, 0, mmCP_VMID_RESET), 0, 0xffffffff);
+	amdgpu_ring_commit(kiq_ring);
+	r = amdgpu_ring_test_ring(kiq_ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+	if (r)
+		return r;
+
+	r = gfx_v10_0_kgq_init_queue(ring, true);
+	if (r) {
+		DRM_ERROR("fail to init kgq\n");
+		return r;
+	}
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+
+	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size)) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+	kiq->pmf->kiq_map_queues(kiq_ring, ring);
+	amdgpu_ring_commit(kiq_ring);
+	r = amdgpu_ring_test_ring(kiq_ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+	if (r)
+		return r;
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static int gfx_v10_0_reset_kcq(struct amdgpu_ring *ring,
+			       unsigned int vmid,
+			       struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	unsigned long flags;
+	int i, r;
+
+	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
+		return -EINVAL;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+
+	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size)) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+
+	kiq->pmf->kiq_unmap_queues(kiq_ring, ring, RESET_QUEUES,
+				   0, 0);
+	amdgpu_ring_commit(kiq_ring);
+	r = amdgpu_ring_test_ring(kiq_ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+	if (r)
+		return r;
+
+	/* make sure dequeue is complete*/
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+	mutex_lock(&adev->srbm_mutex);
+	nv_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (!(RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1))
+			break;
+		udelay(1);
+	}
+	if (i >= adev->usec_timeout)
+		r = -ETIMEDOUT;
+	nv_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
+	if (r) {
+		dev_err(adev->dev, "fail to wait on hqd deactivate\n");
+		return r;
+	}
+
+	r = gfx_v10_0_kcq_init_queue(ring, true);
+	if (r) {
+		dev_err(adev->dev, "fail to init kcq\n");
+		return r;
+	}
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size)) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+	kiq->pmf->kiq_map_queues(kiq_ring, ring);
+	amdgpu_ring_commit(kiq_ring);
+	r = amdgpu_ring_test_ring(kiq_ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+	if (r)
+		return r;
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static void gfx_v10_ip_print(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k, reg, index = 0;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_10_1);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	for (i = 0; i < reg_count; i++)
+		drm_printf(p, "%-50s \t 0x%08x\n",
+			   gc_reg_list_10_1[i].reg_name,
+			   adev->gfx.ip_dump_core[i]);
+
+	/* print compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_10);
+	drm_printf(p, "\nnum_mec: %d num_pipe: %d num_queue: %d\n",
+		   adev->gfx.mec.num_mec,
+		   adev->gfx.mec.num_pipe_per_mec,
+		   adev->gfx.mec.num_queue_per_pipe);
+
+	for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+		for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+			for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+				drm_printf(p, "\nmec %d, pipe %d, queue %d\n", i, j, k);
+				for (reg = 0; reg < reg_count; reg++) {
+					if (i && gc_cp_reg_list_10[reg].reg_offset == mmCP_MEC_ME1_HEADER_DUMP)
+						drm_printf(p, "%-50s \t 0x%08x\n",
+							   "mmCP_MEC_ME2_HEADER_DUMP",
+							   adev->gfx.ip_dump_compute_queues[index + reg]);
+					else
+						drm_printf(p, "%-50s \t 0x%08x\n",
+							   gc_cp_reg_list_10[reg].reg_name,
+							   adev->gfx.ip_dump_compute_queues[index + reg]);
+				}
+				index += reg_count;
+			}
+		}
+	}
+
+	/* print gfx queue registers for all instances */
+	if (!adev->gfx.ip_dump_gfx_queues)
+		return;
+
+	index = 0;
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_10);
+	drm_printf(p, "\nnum_me: %d num_pipe: %d num_queue: %d\n",
+		   adev->gfx.me.num_me,
+		   adev->gfx.me.num_pipe_per_me,
+		   adev->gfx.me.num_queue_per_pipe);
+
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			for (k = 0; k < adev->gfx.me.num_queue_per_pipe; k++) {
+				drm_printf(p, "\nme %d, pipe %d, queue %d\n", i, j, k);
+				for (reg = 0; reg < reg_count; reg++) {
+					drm_printf(p, "%-50s \t 0x%08x\n",
+						   gc_gfx_queue_reg_list_10[reg].reg_name,
+						   adev->gfx.ip_dump_gfx_queues[index + reg]);
+				}
+				index += reg_count;
+			}
+		}
+	}
+}
+
+static void gfx_v10_ip_dump(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k, reg, index = 0;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_10_1);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	amdgpu_gfx_off_ctrl(adev, false);
+	for (i = 0; i < reg_count; i++)
+		adev->gfx.ip_dump_core[i] = RREG32(SOC15_REG_ENTRY_OFFSET(gc_reg_list_10_1[i]));
+	amdgpu_gfx_off_ctrl(adev, true);
+
+	/* dump compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_10);
+	amdgpu_gfx_off_ctrl(adev, false);
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+		for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+			for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+				/* ME0 is for GFX so start from 1 for CP */
+				nv_grbm_select(adev, adev->gfx.me.num_me + i, j, k, 0);
+
+				for (reg = 0; reg < reg_count; reg++) {
+					if (i && gc_cp_reg_list_10[reg].reg_offset == mmCP_MEC_ME1_HEADER_DUMP)
+						adev->gfx.ip_dump_compute_queues[index + reg] =
+							RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEC_ME2_HEADER_DUMP));
+					else
+						adev->gfx.ip_dump_compute_queues[index + reg] =
+							RREG32(SOC15_REG_ENTRY_OFFSET(
+								       gc_cp_reg_list_10[reg]));
+				}
+				index += reg_count;
+			}
+		}
+	}
+	nv_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_off_ctrl(adev, true);
+
+	/* dump gfx queue registers for all instances */
+	if (!adev->gfx.ip_dump_gfx_queues)
+		return;
+
+	index = 0;
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_10);
+	amdgpu_gfx_off_ctrl(adev, false);
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			for (k = 0; k < adev->gfx.me.num_queue_per_pipe; k++) {
+				nv_grbm_select(adev, i, j, k, 0);
+
+				for (reg = 0; reg < reg_count; reg++) {
+					adev->gfx.ip_dump_gfx_queues[index + reg] =
+						RREG32(SOC15_REG_ENTRY_OFFSET(
+							gc_gfx_queue_reg_list_10[reg]));
+				}
+				index += reg_count;
+			}
+		}
+	}
+	nv_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_off_ctrl(adev, true);
+}
+
+static void gfx_v10_0_ring_emit_cleaner_shader(struct amdgpu_ring *ring)
+{
+	/* Emit the cleaner shader */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_RUN_CLEANER_SHADER, 0));
+	amdgpu_ring_write(ring, 0);  /* RESERVED field, programmed to zero */
+}
+
+static void gfx_v10_0_ring_begin_use(struct amdgpu_ring *ring)
+{
+	amdgpu_gfx_profile_ring_begin_use(ring);
+
+	amdgpu_gfx_enforce_isolation_ring_begin_use(ring);
+}
+
+static void gfx_v10_0_ring_end_use(struct amdgpu_ring *ring)
+{
+	amdgpu_gfx_profile_ring_end_use(ring);
+
+	amdgpu_gfx_enforce_isolation_ring_end_use(ring);
+}
+
 static const struct amd_ip_funcs gfx_v10_0_ip_funcs = {
 	.name = "gfx_v10_0",
 	.early_init = gfx_v10_0_early_init,
@@ -9241,6 +9823,8 @@ static const struct amd_ip_funcs gfx_v10_0_ip_funcs = {
 	.set_clockgating_state = gfx_v10_0_set_clockgating_state,
 	.set_powergating_state = gfx_v10_0_set_powergating_state,
 	.get_clockgating_state = gfx_v10_0_get_clockgating_state,
+	.dump_ip_state = gfx_v10_ip_dump,
+	.print_ip_state = gfx_v10_ip_print,
 };
 
 static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {
@@ -9249,7 +9833,6 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {
 	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
 	.support_64bit_ptrs = true,
 	.secure_submission_supported = true,
-	.vmhub = AMDGPU_GFXHUB_0,
 	.get_rptr = gfx_v10_0_ring_get_rptr_gfx,
 	.get_wptr = gfx_v10_0_ring_get_wptr_gfx,
 	.set_wptr = gfx_v10_0_ring_set_wptr_gfx,
@@ -9258,7 +9841,7 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {
 		7 + /* PIPELINE_SYNC */
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
-		2 + /* VM_FLUSH */
+		4 + /* VM_FLUSH */
 		8 + /* FENCE for VM_FLUSH */
 		20 + /* GDS switch */
 		4 + /* double SWITCH_BUFFER,
@@ -9274,7 +9857,8 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {
 		5 + /* HDP_INVL */
 		8 + 8 + /* FENCE x2 */
 		2 + /* SWITCH_BUFFER */
-		8, /* gfx_v10_0_emit_mem_sync */
+		8 + /* gfx_v10_0_emit_mem_sync */
+		2, /* gfx_v10_0_ring_emit_cleaner_shader */
 	.emit_ib_size =	4, /* gfx_v10_0_ring_emit_ib_gfx */
 	.emit_ib = gfx_v10_0_ring_emit_ib_gfx,
 	.emit_fence = gfx_v10_0_ring_emit_fence,
@@ -9284,19 +9868,21 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_gfx = {
 	.emit_hdp_flush = gfx_v10_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v10_0_ring_test_ring,
 	.test_ib = gfx_v10_0_ring_test_ib,
-	.insert_nop = amdgpu_ring_insert_nop,
+	.insert_nop = gfx_v10_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_switch_buffer = gfx_v10_0_ring_emit_sb,
 	.emit_cntxcntl = gfx_v10_0_ring_emit_cntxcntl,
 	.init_cond_exec = gfx_v10_0_ring_emit_init_cond_exec,
-	.patch_cond_exec = gfx_v10_0_ring_emit_patch_cond_exec,
 	.preempt_ib = gfx_v10_0_ring_preempt_ib,
 	.emit_frame_cntl = gfx_v10_0_ring_emit_frame_cntl,
 	.emit_wreg = gfx_v10_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v10_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v10_0_ring_emit_reg_write_reg_wait,
-	.soft_recovery = gfx_v10_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v10_0_emit_mem_sync,
+	.reset = gfx_v10_0_reset_kgq,
+	.emit_cleaner_shader = gfx_v10_0_ring_emit_cleaner_shader,
+	.begin_use = gfx_v10_0_ring_begin_use,
+	.end_use = gfx_v10_0_ring_end_use,
 };
 
 static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_compute = {
@@ -9304,7 +9890,6 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_compute = {
 	.align_mask = 0xff,
 	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
 	.support_64bit_ptrs = true,
-	.vmhub = AMDGPU_GFXHUB_0,
 	.get_rptr = gfx_v10_0_ring_get_rptr_compute,
 	.get_wptr = gfx_v10_0_ring_get_wptr_compute,
 	.set_wptr = gfx_v10_0_ring_set_wptr_compute,
@@ -9317,7 +9902,8 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_compute = {
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
 		2 + /* gfx_v10_0_ring_emit_vm_flush */
 		8 + 8 + 8 + /* gfx_v10_0_ring_emit_fence x3 for user fence, vm fence */
-		8, /* gfx_v10_0_emit_mem_sync */
+		8 + /* gfx_v10_0_emit_mem_sync */
+		2, /* gfx_v10_0_ring_emit_cleaner_shader */
 	.emit_ib_size =	7, /* gfx_v10_0_ring_emit_ib_compute */
 	.emit_ib = gfx_v10_0_ring_emit_ib_compute,
 	.emit_fence = gfx_v10_0_ring_emit_fence,
@@ -9327,12 +9913,16 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_compute = {
 	.emit_hdp_flush = gfx_v10_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v10_0_ring_test_ring,
 	.test_ib = gfx_v10_0_ring_test_ib,
-	.insert_nop = amdgpu_ring_insert_nop,
+	.insert_nop = gfx_v10_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_wreg = gfx_v10_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v10_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v10_0_ring_emit_reg_write_reg_wait,
 	.emit_mem_sync = gfx_v10_0_emit_mem_sync,
+	.reset = gfx_v10_0_reset_kcq,
+	.emit_cleaner_shader = gfx_v10_0_ring_emit_cleaner_shader,
+	.begin_use = gfx_v10_0_ring_begin_use,
+	.end_use = gfx_v10_0_ring_end_use,
 };
 
 static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_kiq = {
@@ -9340,7 +9930,6 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_kiq = {
 	.align_mask = 0xff,
 	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
 	.support_64bit_ptrs = true,
-	.vmhub = AMDGPU_GFXHUB_0,
 	.get_rptr = gfx_v10_0_ring_get_rptr_compute,
 	.get_wptr = gfx_v10_0_ring_get_wptr_compute,
 	.set_wptr = gfx_v10_0_ring_set_wptr_compute,
@@ -9351,7 +9940,6 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_kiq = {
 		7 + /* gfx_v10_0_ring_emit_pipeline_sync */
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
-		2 + /* gfx_v10_0_ring_emit_vm_flush */
 		8 + 8 + 8, /* gfx_v10_0_ring_emit_fence_kiq x3 for user fence, vm fence */
 	.emit_ib_size =	7, /* gfx_v10_0_ring_emit_ib_compute */
 	.emit_ib = gfx_v10_0_ring_emit_ib_compute,
@@ -9364,13 +9952,14 @@ static const struct amdgpu_ring_funcs gfx_v10_0_ring_funcs_kiq = {
 	.emit_wreg = gfx_v10_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v10_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v10_0_ring_emit_reg_write_reg_wait,
+	.emit_hdp_flush = gfx_v10_0_ring_emit_hdp_flush,
 };
 
 static void gfx_v10_0_set_ring_funcs(struct amdgpu_device *adev)
 {
 	int i;
 
-	adev->gfx.kiq.ring.funcs = &gfx_v10_0_ring_funcs_kiq;
+	adev->gfx.kiq[0].ring.funcs = &gfx_v10_0_ring_funcs_kiq;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 		adev->gfx.gfx_ring[i].funcs = &gfx_v10_0_ring_funcs_gfx;
@@ -9389,6 +9978,11 @@ static const struct amdgpu_irq_src_funcs gfx_v10_0_priv_reg_irq_funcs = {
 	.process = gfx_v10_0_priv_reg_irq,
 };
 
+static const struct amdgpu_irq_src_funcs gfx_v10_0_bad_op_irq_funcs = {
+	.set = gfx_v10_0_set_bad_op_fault_state,
+	.process = gfx_v10_0_bad_op_irq,
+};
+
 static const struct amdgpu_irq_src_funcs gfx_v10_0_priv_inst_irq_funcs = {
 	.set = gfx_v10_0_set_priv_inst_fault_state,
 	.process = gfx_v10_0_priv_inst_irq,
@@ -9404,19 +9998,22 @@ static void gfx_v10_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->gfx.eop_irq.num_types = AMDGPU_CP_IRQ_LAST;
 	adev->gfx.eop_irq.funcs = &gfx_v10_0_eop_irq_funcs;
 
-	adev->gfx.kiq.irq.num_types = AMDGPU_CP_KIQ_IRQ_LAST;
-	adev->gfx.kiq.irq.funcs = &gfx_v10_0_kiq_irq_funcs;
+	adev->gfx.kiq[0].irq.num_types = AMDGPU_CP_KIQ_IRQ_LAST;
+	adev->gfx.kiq[0].irq.funcs = &gfx_v10_0_kiq_irq_funcs;
 
 	adev->gfx.priv_reg_irq.num_types = 1;
 	adev->gfx.priv_reg_irq.funcs = &gfx_v10_0_priv_reg_irq_funcs;
 
+	adev->gfx.bad_op_irq.num_types = 1;
+	adev->gfx.bad_op_irq.funcs = &gfx_v10_0_bad_op_irq_funcs;
+
 	adev->gfx.priv_inst_irq.num_types = 1;
 	adev->gfx.priv_inst_irq.funcs = &gfx_v10_0_priv_inst_irq_funcs;
 }
 
 static void gfx_v10_0_set_rlc_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 1):
 	case IP_VERSION(10, 1, 3):
@@ -9441,7 +10038,7 @@ static void gfx_v10_0_set_rlc_funcs(struct amdgpu_device *adev)
 
 static void gfx_v10_0_set_gds_init(struct amdgpu_device *adev)
 {
-	unsigned total_cu = adev->gfx.config.max_cu_per_sh *
+	unsigned int total_cu = adev->gfx.config.max_cu_per_sh *
 			    adev->gfx.config.max_sh_per_se *
 			    adev->gfx.config.max_shader_engines;
 
@@ -9522,7 +10119,7 @@ static int gfx_v10_0_get_cu_info(struct amdgpu_device *adev,
 {
 	int i, j, k, counter, active_cu_number = 0;
 	u32 mask, bitmap, ao_bitmap, ao_cu_mask = 0;
-	unsigned disable_masks[4 * 2];
+	unsigned int disable_masks[4 * 2];
 
 	if (!adev || !cu_info)
 		return -EINVAL;
@@ -9533,21 +10130,25 @@ static int gfx_v10_0_get_cu_info(struct amdgpu_device *adev,
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
 			bitmap = i * adev->gfx.config.max_sh_per_se + j;
-			if (((adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 0)) ||
-			     (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 3)) ||
-			     (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 6)) ||
-			     (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 3, 7))) &&
+			if (((amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			      IP_VERSION(10, 3, 0)) ||
+			     (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			      IP_VERSION(10, 3, 3)) ||
+			     (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			      IP_VERSION(10, 3, 6)) ||
+			     (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			      IP_VERSION(10, 3, 7))) &&
 			    ((gfx_v10_3_get_disabled_sa(adev) >> bitmap) & 1))
 				continue;
 			mask = 1;
 			ao_bitmap = 0;
 			counter = 0;
-			gfx_v10_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v10_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			if (i < 4 && j < 2)
 				gfx_v10_0_set_user_wgp_inactive_bitmap_per_sh(
 					adev, disable_masks[i * 2 + j]);
 			bitmap = gfx_v10_0_get_cu_active_bitmap_per_sh(adev);
-			cu_info->bitmap[i][j] = bitmap;
+			cu_info->bitmap[0][i][j] = bitmap;
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
 				if (bitmap & mask) {
@@ -9563,7 +10164,7 @@ static int gfx_v10_0_get_cu_info(struct amdgpu_device *adev,
 			cu_info->ao_cu_bitmap[i][j] = ao_bitmap;
 		}
 	}
-	gfx_v10_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v10_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	cu_info->number = active_cu_number;
@@ -9639,8 +10240,7 @@ static void gfx_v10_3_set_power_brake_sequence(struct amdgpu_device *adev)
 		     (0x1 << DIDT_SQ_THROTTLE_CTRL__PWRBRK_STALL_EN__SHIFT));
 }
 
-const struct amdgpu_ip_block_version gfx_v10_0_ip_block =
-{
+const struct amdgpu_ip_block_version gfx_v10_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GFX,
 	.major = 10,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0_cleaner_shader.h b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0_cleaner_shader.h
new file mode 100644
index 000000000000..f67569ccf9f6
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0_cleaner_shader.h
@@ -0,0 +1,91 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/* Define the cleaner shader gfx_10_1_10 */
+static const u32 gfx_10_1_10_cleaner_shader_hex[] = {
+	0xb0804004, 0xbf8a0000,
+	0xbf068100, 0xbf840023,
+	0xbe8203b8, 0xbefc0380,
+	0x7e008480, 0x7e028480,
+	0x7e048480, 0x7e068480,
+	0x7e088480, 0x7e0a8480,
+	0x7e0c8480, 0x7e0e8480,
+	0xbefc0302, 0x80828802,
+	0xbf84fff5, 0xbe8203ff,
+	0x80000000, 0x87020102,
+	0xbf840012, 0xbefe03c1,
+	0xbeff03c1, 0xd7650001,
+	0x0001007f, 0xd7660001,
+	0x0002027e, 0x16020288,
+	0xbe8203bf, 0xbefc03c1,
+	0xd9382000, 0x00020201,
+	0xd9386040, 0x00040401,
+	0xd70f6a01, 0x000202ff,
+	0x00000400, 0x80828102,
+	0xbf84fff7, 0xbefc03ff,
+	0x00000068, 0xbe803000,
+	0xbe813000, 0xbe823000,
+	0xbe833000, 0x80fc847c,
+	0xbf84fffa, 0xbeea0480,
+	0xbeec0480, 0xbeee0480,
+	0xbef00480, 0xbef20480,
+	0xbef40480, 0xbef60480,
+	0xbef80480, 0xbefa0480,
+	0xbf810000, 0xbf9f0000,
+	0xbf9f0000, 0xbf9f0000,
+	0xbf9f0000, 0xbf9f0000,
+};
+
+/* Define the cleaner shader gfx_10_3_0 */
+static const u32 gfx_10_3_0_cleaner_shader_hex[] = {
+	0xb0804004, 0xbf8a0000,
+	0xbe8203b8, 0xbefc0380,
+	0x7e008480, 0x7e028480,
+	0x7e048480, 0x7e068480,
+	0x7e088480, 0x7e0a8480,
+	0x7e0c8480, 0x7e0e8480,
+	0xbefc0302, 0x80828802,
+	0xbf84fff5, 0xbe8203ff,
+	0x80000000, 0x87020002,
+	0xbf840012, 0xbefe03c1,
+	0xbeff03c1, 0xd7650001,
+	0x0001007f, 0xd7660001,
+	0x0002027e, 0x16020288,
+	0xbe8203bf, 0xbefc03c1,
+	0xd9382000, 0x00020201,
+	0xd9386040, 0x00040401,
+	0xd70f6a01, 0x000202ff,
+	0x00000400, 0x80828102,
+	0xbf84fff7, 0xbefc03ff,
+	0x00000068, 0xbe803080,
+	0xbe813080, 0xbe823080,
+	0xbe833080, 0x80fc847c,
+	0xbf84fffa, 0xbeea0480,
+	0xbeec0480, 0xbeee0480,
+	0xbef00480, 0xbef20480,
+	0xbef40480, 0xbef60480,
+	0xbef80480, 0xbefa0480,
+	0xbf810000, 0xbf9f0000,
+	0xbf9f0000, 0xbf9f0000,
+	0xbf9f0000, 0xbf9f0000,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_1_10_cleaner_shader.asm b/drivers/gpu/drm/amd/amdgpu/gfx_v10_1_10_cleaner_shader.asm
new file mode 100644
index 000000000000..54f7ed9e2801
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_1_10_cleaner_shader.asm
@@ -0,0 +1,125 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+// This shader is to clean LDS, SGPRs and VGPRs. It is  first 64 Dwords or 256 bytes of 256 Dwords cleaner shader.
+
+// GFX10.1 : Clear SGPRs, VGPRs and LDS
+//   Launch 32 waves per CU (16 per SIMD) as a workgroup (threadgroup) to fill every wave slot
+//   Waves are "wave32" and have 64 VGPRs each, which uses all 1024 VGPRs per SIMD
+//   Waves are launched in "CU" mode, and the workgroup shares 64KB of LDS (half of the WGP's LDS)
+//      It takes 2 workgroups to use all of LDS: one on each CU of the WGP
+//   Each wave clears SGPRs 0 - 107
+//   Each wave clears VGPRs 0 - 63
+//   The first wave of the workgroup clears its 64KB of LDS
+//   The shader starts with "S_BARRIER" to ensure SPI has launched all waves of the workgroup
+//       before any wave in the workgroup could end.  Without this, it is possible not all SGPRs get cleared.
+
+
+shader main
+  asic(GFX10.1)
+  type(CS)
+  wave_size(32)
+// Note: original source code from SQ team
+//
+// Create 32 waves in a threadgroup (CS waves)
+// Each allocates 64 VGPRs
+// The workgroup allocates all of LDS (64kbytes)
+//
+// Takes about 2500 clocks to run.
+//   (theorhetical fastest = 1024clks vgpr + 640lds = 1660 clks)
+//
+  S_BARRIER
+  s_cmp_eq_u32 s0, 1                                // Bit0 is set, sgpr0 is set then clear VGPRS and LDS as FW set COMPUTE_USER_DATA_0
+  s_cbranch_scc0  label_0023                        // Clean VGPRs and LDS if sgpr0 of wave is set, scc = (s0 == 1)
+
+  s_mov_b32     s2, 0x00000038  // Loop 64/8=8 times  (loop unrolled for performance)
+  s_mov_b32     m0, 0
+  //
+  // CLEAR VGPRs
+  //
+label_0005:
+  v_movreld_b32     v0, 0
+  v_movreld_b32     v1, 0
+  v_movreld_b32     v2, 0
+  v_movreld_b32     v3, 0
+  v_movreld_b32     v4, 0
+  v_movreld_b32     v5, 0
+  v_movreld_b32     v6, 0
+  v_movreld_b32     v7, 0
+  s_mov_b32         m0, s2
+  s_sub_u32     s2, s2, 8
+  s_cbranch_scc0  label_0005
+  //
+  s_mov_b32     s2, 0x80000000                       // Bit31 is first_wave
+  s_and_b32     s2, s2, s1                           // sgpr0 has tg_size (first_wave) term as in ucode only COMPUTE_PGM_RSRC2.tg_size_en is set
+  s_cbranch_scc0  label_0023                         // Clean LDS if its first wave of ThreadGroup/WorkGroup
+  // CLEAR LDS
+  //
+  s_mov_b32 exec_lo, 0xffffffff
+  s_mov_b32 exec_hi, 0xffffffff
+  v_mbcnt_lo_u32_b32  v1, exec_hi, 0          // Set V1 to thread-ID (0..63)
+  v_mbcnt_hi_u32_b32  v1, exec_lo, v1        // Set V1 to thread-ID (0..63)
+  v_mul_u32_u24  v1, 0x00000008, v1          // * 8, so each thread is a double-dword address (8byte)
+  s_mov_b32     s2, 0x00000003f                    // 64 loop iterations
+  s_mov_b32     m0, 0xffffffff
+  // Clear all of LDS space
+  // Each FirstWave of WorkGroup clears 64kbyte block
+
+label_001F:
+  ds_write2_b64  v1, v[2:3], v[2:3] offset1:32
+  ds_write2_b64  v1, v[4:5], v[4:5] offset0:64 offset1:96
+  v_add_co_u32     v1, vcc, 0x00000400, v1
+  s_sub_u32     s2, s2, 1
+  s_cbranch_scc0  label_001F
+
+  //
+  // CLEAR SGPRs
+  //
+label_0023:
+  s_mov_b32     m0, 0x00000068  // Loop 108/4=27 times  (loop unrolled for performance)
+label_sgpr_loop:
+  s_movreld_b32     s0, s0
+  s_movreld_b32     s1, s0
+  s_movreld_b32     s2, s0
+  s_movreld_b32     s3, s0
+  s_sub_u32         m0, m0, 4
+  s_cbranch_scc0  label_sgpr_loop
+
+  //clear vcc
+  s_mov_b64 vcc, 0          //clear vcc
+  //s_setreg_imm32_b32 hw_reg_shader_flat_scratch_lo, 0   //clear  flat scratch lo SGPR
+  //s_setreg_imm32_b32 hw_reg_shader_flat_scratch_hi, 0    //clear  flat scratch hi SGPR
+  s_mov_b64 ttmp0, 0        //Clear ttmp0 and ttmp1
+  s_mov_b64 ttmp2, 0        //Clear ttmp2 and ttmp3
+  s_mov_b64 ttmp4, 0        //Clear ttmp4 and ttmp5
+  s_mov_b64 ttmp6, 0        //Clear ttmp6 and ttmp7
+  s_mov_b64 ttmp8, 0        //Clear ttmp8 and ttmp9
+  s_mov_b64 ttmp10, 0       //Clear ttmp10 and ttmp11
+  s_mov_b64 ttmp12, 0       //Clear ttmp12 and ttmp13
+  s_mov_b64 ttmp14, 0       //Clear ttmp14 and ttmp15
+
+ s_endpgm
+
+end
+
+
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_3_0_cleaner_shader.asm b/drivers/gpu/drm/amd/amdgpu/gfx_v10_3_0_cleaner_shader.asm
new file mode 100644
index 000000000000..0e1c246166c0
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_3_0_cleaner_shader.asm
@@ -0,0 +1,124 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2025 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+// This shader is to clean LDS, SGPRs and VGPRs. It is  first 64 Dwords or 256 bytes of 192 Dwords cleaner shader.
+//To turn this shader program on for complitaion change this to main and lower shader main to main_1
+
+// GFX10.3 : Clear SGPRs, VGPRs and LDS
+//   Launch 32 waves per CU (16 per SIMD) as a workgroup (threadgroup) to fill every wave slot
+//   Waves are "wave32" and have 64 VGPRs each, which uses all 1024 VGPRs per SIMD
+//   Waves are launched in "CU" mode, and the workgroup shares 64KB of LDS (half of the WGP's LDS)
+//      It takes 2 workgroups to use all of LDS: one on each CU of the WGP
+//   Each wave clears SGPRs 0 - 107
+//   Each wave clears VGPRs 0 - 63
+//   The first wave of the workgroup clears its 64KB of LDS
+//   The shader starts with "S_BARRIER" to ensure SPI has launched all waves of the workgroup
+//       before any wave in the workgroup could end.  Without this, it is possible not all SGPRs get cleared.
+
+
+shader main
+  asic(GFX10)
+  type(CS)
+  wave_size(32)
+// Note: original source code from SQ team
+
+//
+// Create 32 waves in a threadgroup (CS waves)
+// Each allocates 64 VGPRs
+// The workgroup allocates all of LDS (64kbytes)
+//
+// Takes about 2500 clocks to run.
+//   (theorhetical fastest = 1024clks vgpr + 640lds = 1660 clks)
+//
+  S_BARRIER
+  s_mov_b32     s2, 0x00000038  // Loop 64/8=8 times  (loop unrolled for performance)
+  s_mov_b32     m0, 0
+  //
+  // CLEAR VGPRs
+  //
+label_0005:
+  v_movreld_b32     v0, 0
+  v_movreld_b32     v1, 0
+  v_movreld_b32     v2, 0
+  v_movreld_b32     v3, 0
+  v_movreld_b32     v4, 0
+  v_movreld_b32     v5, 0
+  v_movreld_b32     v6, 0
+  v_movreld_b32     v7, 0
+  s_mov_b32         m0, s2
+  s_sub_u32     s2, s2, 8
+  s_cbranch_scc0  label_0005
+  //
+  s_mov_b32     s2, 0x80000000                     // Bit31 is first_wave
+  s_and_b32     s2, s2, s0                                  // sgpr0 has tg_size (first_wave) term as in ucode only COMPUTE_PGM_RSRC2.tg_size_en is set
+  s_cbranch_scc0  label_0023                         // Clean LDS if its first wave of ThreadGroup/WorkGroup
+  // CLEAR LDS
+  //
+  s_mov_b32 exec_lo, 0xffffffff
+  s_mov_b32 exec_hi, 0xffffffff
+  v_mbcnt_lo_u32_b32  v1, exec_hi, 0          // Set V1 to thread-ID (0..63)
+  v_mbcnt_hi_u32_b32  v1, exec_lo, v1        // Set V1 to thread-ID (0..63)
+  v_mul_u32_u24  v1, 0x00000008, v1          // * 8, so each thread is a double-dword address (8byte)
+  s_mov_b32     s2, 0x00000003f                    // 64 loop iterations
+  s_mov_b32     m0, 0xffffffff
+  // Clear all of LDS space
+  // Each FirstWave of WorkGroup clears 64kbyte block
+
+label_001F:
+  ds_write2_b64  v1, v[2:3], v[2:3] offset1:32
+  ds_write2_b64  v1, v[4:5], v[4:5] offset0:64 offset1:96
+  v_add_co_u32     v1, vcc, 0x00000400, v1
+  s_sub_u32     s2, s2, 1
+  s_cbranch_scc0  label_001F
+
+  //
+  // CLEAR SGPRs
+  //
+label_0023:
+  s_mov_b32     m0, 0x00000068  // Loop 108/4=27 times  (loop unrolled for performance)
+label_sgpr_loop:
+  s_movreld_b32     s0, 0
+  s_movreld_b32     s1, 0
+  s_movreld_b32     s2, 0
+  s_movreld_b32     s3, 0
+  s_sub_u32         m0, m0, 4
+  s_cbranch_scc0  label_sgpr_loop
+
+  //clear vcc
+  s_mov_b32 flat_scratch_lo, 0   //clear  flat scratch lo SGPR
+  s_mov_b32 flat_scratch_hi, 0   //clear  flat scratch hi SGPR
+  s_mov_b64 vcc, 0          //clear vcc
+  s_mov_b64 ttmp0, 0        //Clear ttmp0 and ttmp1
+  s_mov_b64 ttmp2, 0        //Clear ttmp2 and ttmp3
+  s_mov_b64 ttmp4, 0        //Clear ttmp4 and ttmp5
+  s_mov_b64 ttmp6, 0        //Clear ttmp6 and ttmp7
+  s_mov_b64 ttmp8, 0        //Clear ttmp8 and ttmp9
+  s_mov_b64 ttmp10, 0       //Clear ttmp10 and ttmp11
+  s_mov_b64 ttmp12, 0       //Clear ttmp12 and ttmp13
+  s_mov_b64 ttmp14, 0       //Clear ttmp14 and ttmp15
+
+ s_endpgm
+
+end
+
+
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
index ecf8ceb53311..8a2ee2de390f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
@@ -29,7 +29,6 @@
 #include "amdgpu_gfx.h"
 #include "amdgpu_psp.h"
 #include "amdgpu_smu.h"
-#include "amdgpu_atomfirmware.h"
 #include "imu_v11_0.h"
 #include "soc21.h"
 #include "nvd.h"
@@ -42,13 +41,15 @@
 #include "ivsrcid/gfx/irqsrcs_gfx_11_0_0.h"
 
 #include "soc15.h"
-#include "soc15d.h"
 #include "clearstate_gfx11.h"
 #include "v11_structs.h"
 #include "gfx_v11_0.h"
+#include "gfx_v11_0_cleaner_shader.h"
 #include "gfx_v11_0_3.h"
 #include "nbio_v4_3.h"
 #include "mes_v11_0.h"
+#include "mes_userqueue.h"
+#include "amdgpu_userq_fence.h"
 
 #define GFX11_NUM_GFX_RINGS		1
 #define GFX11_MEC_HPD_SIZE	2048
@@ -60,11 +61,32 @@
 #define regCGTT_WD_CLK_CTRL_BASE_IDX	1
 #define regRLC_RLCS_BOOTLOAD_STATUS_gc_11_0_1	0x4e7e
 #define regRLC_RLCS_BOOTLOAD_STATUS_gc_11_0_1_BASE_IDX	1
+#define regPC_CONFIG_CNTL_1		0x194d
+#define regPC_CONFIG_CNTL_1_BASE_IDX	1
+
+#define regCP_GFX_MQD_CONTROL_DEFAULT                                             0x00000100
+#define regCP_GFX_HQD_VMID_DEFAULT                                                0x00000000
+#define regCP_GFX_HQD_QUEUE_PRIORITY_DEFAULT                                      0x00000000
+#define regCP_GFX_HQD_QUANTUM_DEFAULT                                             0x00000a01
+#define regCP_GFX_HQD_CNTL_DEFAULT                                                0x00a00000
+#define regCP_RB_DOORBELL_CONTROL_DEFAULT                                         0x00000000
+#define regCP_GFX_HQD_RPTR_DEFAULT                                                0x00000000
+
+#define regCP_HQD_EOP_CONTROL_DEFAULT                                             0x00000006
+#define regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT                                     0x00000000
+#define regCP_MQD_CONTROL_DEFAULT                                                 0x00000100
+#define regCP_HQD_PQ_CONTROL_DEFAULT                                              0x00308509
+#define regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT                                     0x00000000
+#define regCP_HQD_PQ_RPTR_DEFAULT                                                 0x00000000
+#define regCP_HQD_PERSISTENT_STATE_DEFAULT                                        0x0be05501
+#define regCP_HQD_IB_CONTROL_DEFAULT                                              0x00300000
 
 MODULE_FIRMWARE("amdgpu/gc_11_0_0_pfp.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_0_me.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_0_mec.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_0_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_0_0_rlc_kicker.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_0_0_rlc_1.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_0_toc.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_1_pfp.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_1_me.bin");
@@ -82,6 +104,200 @@ MODULE_FIRMWARE("amdgpu/gc_11_0_4_pfp.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_4_me.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_4_mec.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_4_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_0_pfp.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_0_me.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_0_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_0_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_1_pfp.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_1_me.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_1_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_1_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_2_pfp.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_2_me.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_2_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_2_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_3_pfp.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_3_me.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_3_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_3_rlc.bin");
+
+static const struct amdgpu_hwip_reg_entry gc_reg_list_11_0[] = {
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS2),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS3),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT3),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_BUSY_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_BUSY_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HPD_STATUS0),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB0_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB0_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB0_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB1_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB1_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB1_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB2_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB2_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB2_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB2_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPF_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPG_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regGDS_PROTECTION_FAULT),
+	SOC15_REG_ENTRY_STR(GC, 0, regGDS_VM_PROTECTION_FAULT),
+	SOC15_REG_ENTRY_STR(GC, 0, regIA_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regIA_UTCL1_STATUS_2),
+	SOC15_REG_ENTRY_STR(GC, 0, regPA_CL_CNTL_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regRMI_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSQC_CACHES),
+	SOC15_REG_ENTRY_STR(GC, 0, regSQG_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regWD_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regGCVM_L2_PROTECTION_FAULT_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_DEBUG),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC1_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_DEBUG_INTERRUPT_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_STATUS),
+	/* cp header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	/* SE status registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE0),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE1),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE2),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE3),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE4),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE5)
+};
+
+static const struct amdgpu_hwip_reg_entry gc_cp_reg_list_11[] = {
+	/* compute registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_VMID),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PERSISTENT_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PIPE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_QUEUE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_QUANTUM),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_DEQUEUE_REQUEST),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_EVENTS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_BASE_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CNTL_STACK_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CNTL_STACK_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_WG_STATE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_GDS_RESOURCE_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_WPTR_MEM),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_SUSPEND_CNTL_STACK_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_SUSPEND_CNTL_STACK_DW_CNT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_SUSPEND_WG_STATE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_DEQUEUE_STATUS),
+	/* cp header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+};
+
+static const struct amdgpu_hwip_reg_entry gc_gfx_queue_reg_list_11[] = {
+	/* gfx queue registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_ACTIVE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_VMID),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_QUEUE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_QUANTUM),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_CSMD_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_DEQUEUE_REQUEST),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_MAPPED),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_QUE_MGR_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_HQ_CONTROL0),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_HQ_STATUS0),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_MQD_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_MQD_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_WPTR_POLL_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_WPTR_POLL_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BUFSZ),
+	/* cp header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+};
+
+static const struct soc15_reg_golden golden_settings_gc_11_0[] = {
+	SOC15_REG_GOLDEN_VALUE(GC, 0, regTCP_CNTL, 0x20000000, 0x20000000)
+};
 
 static const struct soc15_reg_golden golden_settings_gc_11_0_1[] =
 {
@@ -112,7 +328,7 @@ static int gfx_v11_0_get_cu_info(struct amdgpu_device *adev,
                                  struct amdgpu_cu_info *cu_info);
 static uint64_t gfx_v11_0_get_gpu_clock_counter(struct amdgpu_device *adev);
 static void gfx_v11_0_select_se_sh(struct amdgpu_device *adev, u32 se_num,
-				   u32 sh_num, u32 instance);
+				   u32 sh_num, u32 instance, int xcc_id);
 static u32 gfx_v11_0_get_wgp_active_bitmap_per_sh(struct amdgpu_device *adev);
 
 static void gfx_v11_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume);
@@ -123,20 +339,27 @@ static int gfx_v11_0_wait_for_rlc_autoload_complete(struct amdgpu_device *adev);
 static void gfx_v11_0_ring_invalidate_tlbs(struct amdgpu_ring *ring,
 					   uint16_t pasid, uint32_t flush_type,
 					   bool all_hub, uint8_t dst_sel);
-static void gfx_v11_0_set_safe_mode(struct amdgpu_device *adev);
-static void gfx_v11_0_unset_safe_mode(struct amdgpu_device *adev);
+static void gfx_v11_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id);
+static void gfx_v11_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id);
 static void gfx_v11_0_update_perf_clk(struct amdgpu_device *adev,
 				      bool enable);
 
 static void gfx11_kiq_set_resources(struct amdgpu_ring *kiq_ring, uint64_t queue_mask)
 {
+	struct amdgpu_device *adev = kiq_ring->adev;
+	u64 shader_mc_addr;
+
+	/* Cleaner shader MC address */
+	shader_mc_addr = adev->gfx.cleaner_shader_gpu_addr >> 8;
+
 	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_RESOURCES, 6));
 	amdgpu_ring_write(kiq_ring, PACKET3_SET_RESOURCES_VMID_MASK(0) |
+			  PACKET3_SET_RESOURCES_UNMAP_LATENTY(0xa) | /* unmap_latency: 0xa (~ 1s) */
 			  PACKET3_SET_RESOURCES_QUEUE_TYPE(0));	/* vmid_mask:0 queue_type:0 (KIQ) */
 	amdgpu_ring_write(kiq_ring, lower_32_bits(queue_mask));	/* queue mask lo */
 	amdgpu_ring_write(kiq_ring, upper_32_bits(queue_mask));	/* queue mask hi */
-	amdgpu_ring_write(kiq_ring, 0);	/* gws mask lo */
-	amdgpu_ring_write(kiq_ring, 0);	/* gws mask hi */
+	amdgpu_ring_write(kiq_ring, lower_32_bits(shader_mc_addr)); /* cleaner shader addr lo */
+	amdgpu_ring_write(kiq_ring, upper_32_bits(shader_mc_addr)); /* cleaner shader addr hi */
 	amdgpu_ring_write(kiq_ring, 0);	/* oac mask */
 	amdgpu_ring_write(kiq_ring, 0);	/* gds heap base:0, gds heap size:0 */
 }
@@ -192,7 +415,7 @@ static void gfx11_kiq_unmap_queues(struct amdgpu_ring *kiq_ring,
 	struct amdgpu_device *adev = kiq_ring->adev;
 	uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0;
 
-	if (adev->enable_mes && !adev->gfx.kiq.ring.sched.ready) {
+	if (adev->enable_mes && !adev->gfx.kiq[0].ring.sched.ready) {
 		amdgpu_mes_unmap_legacy_queue(adev, ring, action, gpu_addr, seq);
 		return;
 	}
@@ -260,12 +483,15 @@ static const struct kiq_pm4_funcs gfx_v11_0_kiq_pm4_funcs = {
 
 static void gfx_v11_0_set_kiq_pm4_funcs(struct amdgpu_device *adev)
 {
-	adev->gfx.kiq.pmf = &gfx_v11_0_kiq_pm4_funcs;
+	adev->gfx.kiq[0].pmf = &gfx_v11_0_kiq_pm4_funcs;
 }
 
 static void gfx_v11_0_init_golden_registers(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 4):
 		soc15_program_register_sequence(adev,
@@ -275,6 +501,10 @@ static void gfx_v11_0_init_golden_registers(struct amdgpu_device *adev)
 	default:
 		break;
 	}
+	soc15_program_register_sequence(adev,
+					golden_settings_gc_11_0,
+					(const u32)ARRAY_SIZE(golden_settings_gc_11_0));
+
 }
 
 static void gfx_v11_0_write_data_to_reg(struct amdgpu_ring *ring, int eng_sel,
@@ -310,6 +540,21 @@ static void gfx_v11_0_wait_reg_mem(struct amdgpu_ring *ring, int eng_sel,
 	amdgpu_ring_write(ring, inv); /* poll interval */
 }
 
+static void gfx_v11_ring_insert_nop(struct amdgpu_ring *ring, uint32_t num_nop)
+{
+	/* Header itself is a NOP packet */
+	if (num_nop == 1) {
+		amdgpu_ring_write(ring, ring->funcs->nop);
+		return;
+	}
+
+	/* Max HW optimization till 0x3ffe, followed by remaining one NOP at a time*/
+	amdgpu_ring_write(ring, PACKET3(PACKET3_NOP, min(num_nop - 2, 0x3ffe)));
+
+	/* Header is at index 0, followed by num_nops - 1 NOP packet's */
+	amdgpu_ring_insert_nop(ring, num_nop - 1);
+}
+
 static int gfx_v11_0_ring_test_ring(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
@@ -358,7 +603,7 @@ static int gfx_v11_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	struct dma_fence *f = NULL;
 	unsigned index;
 	uint64_t gpu_addr;
-	volatile uint32_t *cpu_ptr;
+	uint32_t *cpu_ptr;
 	long r;
 
 	/* MES KIQ fw hasn't indirect buffer support for now */
@@ -368,33 +613,18 @@ static int gfx_v11_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 
 	memset(&ib, 0, sizeof(ib));
 
-	if (ring->is_mes_queue) {
-		uint32_t padding, offset;
-
-		offset = amdgpu_mes_ctx_get_offs(ring, AMDGPU_MES_CTX_IB_OFFS);
-		padding = amdgpu_mes_ctx_get_offs(ring,
-						  AMDGPU_MES_CTX_PADDING_OFFS);
-
-		ib.gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		ib.ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-
-		gpu_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, padding);
-		cpu_ptr = amdgpu_mes_ctx_get_offs_cpu_addr(ring, padding);
-		*cpu_ptr = cpu_to_le32(0xCAFEDEAD);
-	} else {
-		r = amdgpu_device_wb_get(adev, &index);
-		if (r)
-			return r;
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r)
+		return r;
 
-		gpu_addr = adev->wb.gpu_addr + (index * 4);
-		adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
-		cpu_ptr = &adev->wb.wb[index];
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
+	cpu_ptr = &adev->wb.wb[index];
 
-		r = amdgpu_ib_get(adev, NULL, 16, AMDGPU_IB_POOL_DIRECT, &ib);
-		if (r) {
-			DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
-			goto err1;
-		}
+	r = amdgpu_ib_get(adev, NULL, 20, AMDGPU_IB_POOL_DIRECT, &ib);
+	if (r) {
+		DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r);
+		goto err1;
 	}
 
 	ib.ptr[0] = PACKET3(PACKET3_WRITE_DATA, 3);
@@ -421,12 +651,10 @@ static int gfx_v11_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	else
 		r = -EINVAL;
 err2:
-	if (!ring->is_mes_queue)
-		amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err1:
-	if (!ring->is_mes_queue)
-		amdgpu_device_wb_free(adev, index);
+	amdgpu_device_wb_free(adev, index);
 	return r;
 }
 
@@ -444,10 +672,10 @@ static int gfx_v11_0_init_toc_microcode(struct amdgpu_device *adev, const char *
 {
 	const struct psp_firmware_header_v1_0 *toc_hdr;
 	int err = 0;
-	char fw_name[40];
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_toc.bin", ucode_prefix);
-	err = amdgpu_ucode_request(adev, &adev->psp.toc_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->psp.toc_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_toc.bin", ucode_prefix);
 	if (err)
 		goto out;
 
@@ -463,10 +691,30 @@ out:
 	return err;
 }
 
+static void gfx_v11_0_check_fw_cp_gfx_shadow(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(11, 0, 0):
+	case IP_VERSION(11, 0, 2):
+	case IP_VERSION(11, 0, 3):
+		if ((adev->gfx.me_fw_version >= 1505) &&
+		    (adev->gfx.pfp_fw_version >= 1600) &&
+		    (adev->gfx.mec_fw_version >= 512)) {
+			if (amdgpu_sriov_vf(adev))
+				adev->gfx.cp_gfx_shadow = true;
+			else
+				adev->gfx.cp_gfx_shadow = false;
+		}
+		break;
+	default:
+		adev->gfx.cp_gfx_shadow = false;
+		break;
+	}
+}
+
 static int gfx_v11_0_init_microcode(struct amdgpu_device *adev)
 {
-	char fw_name[40];
-	char ucode_prefix[30];
+	char ucode_prefix[25];
 	int err;
 	const struct rlc_firmware_header_v2_0 *rlc_hdr;
 	uint16_t version_major;
@@ -475,9 +723,9 @@ static int gfx_v11_0_init_microcode(struct amdgpu_device *adev)
 	DRM_DEBUG("\n");
 
 	amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix));
-
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", ucode_prefix);
-	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_pfp.bin", ucode_prefix);
 	if (err)
 		goto out;
 	/* check pfp fw hdr version to decide if enable rs64 for gfx11.*/
@@ -493,8 +741,9 @@ static int gfx_v11_0_init_microcode(struct amdgpu_device *adev)
 		amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_PFP);
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", ucode_prefix);
-	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_me.bin", ucode_prefix);
 	if (err)
 		goto out;
 	if (adev->gfx.rs64_enable) {
@@ -506,8 +755,19 @@ static int gfx_v11_0_init_microcode(struct amdgpu_device *adev)
 	}
 
 	if (!amdgpu_sriov_vf(adev)) {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", ucode_prefix);
-		err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, fw_name);
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 0, 0) &&
+		    adev->pdev->revision == 0xCE)
+			err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/gc_11_0_0_rlc_1.bin");
+		else if (amdgpu_is_kicker_fw(adev))
+			err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_rlc_kicker.bin", ucode_prefix);
+		else
+			err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_rlc.bin", ucode_prefix);
 		if (err)
 			goto out;
 		rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
@@ -518,8 +778,9 @@ static int gfx_v11_0_init_microcode(struct amdgpu_device *adev)
 			goto out;
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", ucode_prefix);
-	err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_mec.bin", ucode_prefix);
 	if (err)
 		goto out;
 	if (adev->gfx.rs64_enable) {
@@ -539,6 +800,15 @@ static int gfx_v11_0_init_microcode(struct amdgpu_device *adev)
 	/* only one MEC for gfx 11.0.0. */
 	adev->gfx.mec2_fw = NULL;
 
+	gfx_v11_0_check_fw_cp_gfx_shadow(adev);
+
+	if (adev->gfx.imu.funcs && adev->gfx.imu.funcs->init_microcode) {
+		err = adev->gfx.imu.funcs->init_microcode(adev);
+		if (err)
+			DRM_ERROR("Failed to init imu firmware!\n");
+		return err;
+	}
+
 out:
 	if (err) {
 		amdgpu_ucode_release(&adev->gfx.pfp_fw);
@@ -580,12 +850,9 @@ static u32 gfx_v11_0_get_csb_size(struct amdgpu_device *adev)
 	return count;
 }
 
-static void gfx_v11_0_get_csb_buffer(struct amdgpu_device *adev,
-				    volatile u32 *buffer)
+static void gfx_v11_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer)
 {
-	u32 count = 0, i;
-	const struct cs_section_def *sect = NULL;
-	const struct cs_extent_def *ext = NULL;
+	u32 count = 0;
 	int ctx_reg_offset;
 
 	if (adev->gfx.rlc.cs_data == NULL)
@@ -593,39 +860,15 @@ static void gfx_v11_0_get_csb_buffer(struct amdgpu_device *adev,
 	if (buffer == NULL)
 		return;
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1));
-	buffer[count++] = cpu_to_le32(0x80000000);
-	buffer[count++] = cpu_to_le32(0x80000000);
-
-	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
-		for (ext = sect->section; ext->extent != NULL; ++ext) {
-			if (sect->id == SECT_CONTEXT) {
-				buffer[count++] =
-					cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
-				buffer[count++] = cpu_to_le32(ext->reg_index -
-						PACKET3_SET_CONTEXT_REG_START);
-				for (i = 0; i < ext->reg_count; i++)
-					buffer[count++] = cpu_to_le32(ext->extent[i]);
-			} else {
-				return;
-			}
-		}
-	}
+	count = amdgpu_gfx_csb_preamble_start(buffer);
+	count = amdgpu_gfx_csb_data_parser(adev, buffer, count);
 
-	ctx_reg_offset =
-		SOC15_REG_OFFSET(GC, 0, regPA_SC_TILE_STEERING_OVERRIDE) - PACKET3_SET_CONTEXT_REG_START;
+	ctx_reg_offset = SOC15_REG_OFFSET(GC, 0, regPA_SC_TILE_STEERING_OVERRIDE) - PACKET3_SET_CONTEXT_REG_START;
 	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, 1));
 	buffer[count++] = cpu_to_le32(ctx_reg_offset);
 	buffer[count++] = cpu_to_le32(adev->gfx.config.pa_sc_tile_steering_override);
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0));
-	buffer[count++] = cpu_to_le32(0);
+	amdgpu_gfx_csb_preamble_end(buffer, count);
 }
 
 static void gfx_v11_0_rlc_fini(struct amdgpu_device *adev)
@@ -645,7 +888,7 @@ static void gfx_v11_0_init_rlcg_reg_access_ctrl(struct amdgpu_device *adev)
 {
 	struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl;
 
-	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl;
+	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl[0];
 	reg_access_ctrl->scratch_reg0 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG0);
 	reg_access_ctrl->scratch_reg1 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG1);
 	reg_access_ctrl->scratch_reg2 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG2);
@@ -674,7 +917,7 @@ static int gfx_v11_0_rlc_init(struct amdgpu_device *adev)
 
 	/* init spm vmid with 0xf */
 	if (adev->gfx.rlc.funcs->update_spm_vmid)
-		adev->gfx.rlc.funcs->update_spm_vmid(adev, 0xf);
+		adev->gfx.rlc.funcs->update_spm_vmid(adev, NULL, 0xf);
 
 	return 0;
 }
@@ -699,7 +942,7 @@ static int gfx_v11_0_mec_init(struct amdgpu_device *adev)
 	u32 *hpd;
 	size_t mec_hpd_size;
 
-	bitmap_zero(adev->gfx.mec.queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
+	bitmap_zero(adev->gfx.mec_bitmap[0].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
 
 	/* take ownership of the relevant compute queues */
 	amdgpu_gfx_compute_queue_acquire(adev);
@@ -747,7 +990,7 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t wave,
 		*(out++) = RREG32_SOC15(GC, 0, regSQ_IND_DATA);
 }
 
-static void gfx_v11_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
+static void gfx_v11_0_read_wave_data(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
 {
 	/* in gfx11 the SIMD_ID is specified as part of the INSTANCE
 	 * field when performing a select_se_sh so it should be
@@ -773,7 +1016,7 @@ static void gfx_v11_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd,
 	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_MODE);
 }
 
-static void gfx_v11_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v11_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				     uint32_t wave, uint32_t start,
 				     uint32_t size, uint32_t *dst)
 {
@@ -784,7 +1027,7 @@ static void gfx_v11_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
 		dst);
 }
 
-static void gfx_v11_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v11_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				      uint32_t wave, uint32_t thread,
 				      uint32_t start, uint32_t size,
 				      uint32_t *dst)
@@ -795,11 +1038,39 @@ static void gfx_v11_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd,
 }
 
 static void gfx_v11_0_select_me_pipe_q(struct amdgpu_device *adev,
-									  u32 me, u32 pipe, u32 q, u32 vm)
+					u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
 {
 	soc21_grbm_select(adev, me, pipe, q, vm);
 }
 
+/* all sizes are in bytes */
+#define MQD_SHADOW_BASE_SIZE      73728
+#define MQD_SHADOW_BASE_ALIGNMENT 256
+#define MQD_FWWORKAREA_SIZE       484
+#define MQD_FWWORKAREA_ALIGNMENT  256
+
+static void gfx_v11_0_get_gfx_shadow_info_nocheck(struct amdgpu_device *adev,
+					 struct amdgpu_gfx_shadow_info *shadow_info)
+{
+	shadow_info->shadow_size = MQD_SHADOW_BASE_SIZE;
+	shadow_info->shadow_alignment = MQD_SHADOW_BASE_ALIGNMENT;
+	shadow_info->csa_size = MQD_FWWORKAREA_SIZE;
+	shadow_info->csa_alignment = MQD_FWWORKAREA_ALIGNMENT;
+}
+
+static int gfx_v11_0_get_gfx_shadow_info(struct amdgpu_device *adev,
+					 struct amdgpu_gfx_shadow_info *shadow_info,
+					 bool skip_check)
+{
+	if (adev->gfx.cp_gfx_shadow || skip_check) {
+		gfx_v11_0_get_gfx_shadow_info_nocheck(adev, shadow_info);
+		return 0;
+	} else {
+		memset(shadow_info, 0, sizeof(struct amdgpu_gfx_shadow_info));
+		return -ENOTSUPP;
+	}
+}
+
 static const struct amdgpu_gfx_funcs gfx_v11_0_gfx_funcs = {
 	.get_gpu_clock_counter = &gfx_v11_0_get_gpu_clock_counter,
 	.select_se_sh = &gfx_v11_0_select_se_sh,
@@ -808,12 +1079,12 @@ static const struct amdgpu_gfx_funcs gfx_v11_0_gfx_funcs = {
 	.read_wave_vgprs = &gfx_v11_0_read_wave_vgprs,
 	.select_me_pipe_q = &gfx_v11_0_select_me_pipe_q,
 	.update_perfmon_mgcg = &gfx_v11_0_update_perf_clk,
+	.get_gfx_shadow_info = &gfx_v11_0_get_gfx_shadow_info,
 };
 
 static int gfx_v11_0_gpu_early_init(struct amdgpu_device *adev)
 {
-
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
 	case IP_VERSION(11, 0, 2):
 		adev->gfx.config.max_hw_contexts = 8;
@@ -832,6 +1103,10 @@ static int gfx_v11_0_gpu_early_init(struct amdgpu_device *adev)
 		break;
 	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 		adev->gfx.config.max_hw_contexts = 8;
 		adev->gfx.config.sc_prim_fifo_size_frontend = 0x20;
 		adev->gfx.config.sc_prim_fifo_size_backend = 0x100;
@@ -849,9 +1124,9 @@ static int gfx_v11_0_gpu_early_init(struct amdgpu_device *adev)
 static int gfx_v11_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id,
 				   int me, int pipe, int queue)
 {
-	int r;
 	struct amdgpu_ring *ring;
 	unsigned int irq_type;
+	unsigned int hw_prio;
 
 	ring = &adev->gfx.gfx_ring[ring_id];
 
@@ -861,19 +1136,23 @@ static int gfx_v11_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id,
 
 	ring->ring_obj = NULL;
 	ring->use_doorbell = true;
+	if (adev->gfx.disable_kq) {
+		ring->no_scheduler = true;
+		ring->no_user_submission = true;
+	}
 
 	if (!ring_id)
 		ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
 	else
 		ring->doorbell_index = adev->doorbell_index.gfx_ring1 << 1;
+	ring->vm_hub = AMDGPU_GFXHUB(0);
 	sprintf(ring->name, "gfx_%d.%d.%d", ring->me, ring->pipe, ring->queue);
 
 	irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + ring->pipe;
-	r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type,
-			     AMDGPU_RING_PRIO_DEFAULT, NULL);
-	if (r)
-		return r;
-	return 0;
+	hw_prio = amdgpu_gfx_is_high_priority_graphics_queue(adev, ring) ?
+		AMDGPU_GFX_PIPE_PRIO_HIGH : AMDGPU_GFX_PIPE_PRIO_NORMAL;
+	return amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type,
+				hw_prio, NULL);
 }
 
 static int gfx_v11_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
@@ -896,6 +1175,7 @@ static int gfx_v11_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 	ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
 	ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
 				+ (ring_id * GFX11_MEC_HPD_SIZE);
+	ring->vm_hub = AMDGPU_GFXHUB(0);
 	sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
 
 	irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
@@ -1249,30 +1529,69 @@ static int gfx_v11_0_rlc_backdoor_autoload_enable(struct amdgpu_device *adev)
 	return 0;
 }
 
-static int gfx_v11_0_sw_init(void *handle)
+static void gfx_v11_0_alloc_ip_dump(struct amdgpu_device *adev)
+{
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_11_0);
+	uint32_t *ptr;
+	uint32_t inst;
+
+	ptr = kcalloc(reg_count, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for GFX IP Dump\n");
+		adev->gfx.ip_dump_core = NULL;
+	} else {
+		adev->gfx.ip_dump_core = ptr;
+	}
+
+	/* Allocate memory for compute queue registers for all the instances */
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_11);
+	inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec *
+		adev->gfx.mec.num_queue_per_pipe;
+
+	ptr = kcalloc(reg_count * inst, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for Compute Queues IP Dump\n");
+		adev->gfx.ip_dump_compute_queues = NULL;
+	} else {
+		adev->gfx.ip_dump_compute_queues = ptr;
+	}
+
+	/* Allocate memory for gfx queue registers for all the instances */
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_11);
+	inst = adev->gfx.me.num_me * adev->gfx.me.num_pipe_per_me *
+		adev->gfx.me.num_queue_per_pipe;
+
+	ptr = kcalloc(reg_count * inst, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for GFX Queues IP Dump\n");
+		adev->gfx.ip_dump_gfx_queues = NULL;
+	} else {
+		adev->gfx.ip_dump_gfx_queues = ptr;
+	}
+}
+
+static int gfx_v11_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
-	int i, j, k, r, ring_id = 0;
-	struct amdgpu_kiq *kiq;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	int i, j, k, r, ring_id;
+	int xcc_id = 0;
+	struct amdgpu_device *adev = ip_block->adev;
+	int num_queue_per_pipe = 1; /* we only enable 1 KGQ per pipe */
 
-	adev->gfxhub.funcs->init(adev);
+	INIT_DELAYED_WORK(&adev->gfx.idle_work, amdgpu_gfx_profile_idle_work_handler);
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
+	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
-		adev->gfx.me.num_me = 1;
-		adev->gfx.me.num_pipe_per_me = 1;
-		adev->gfx.me.num_queue_per_pipe = 1;
-		adev->gfx.mec.num_mec = 2;
-		adev->gfx.mec.num_pipe_per_mec = 4;
-		adev->gfx.mec.num_queue_per_pipe = 4;
-		break;
-	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 		adev->gfx.me.num_me = 1;
 		adev->gfx.me.num_pipe_per_me = 1;
-		adev->gfx.me.num_queue_per_pipe = 1;
+		adev->gfx.me.num_queue_per_pipe = 2;
 		adev->gfx.mec.num_mec = 1;
 		adev->gfx.mec.num_pipe_per_mec = 4;
 		adev->gfx.mec.num_queue_per_pipe = 4;
@@ -1287,9 +1606,118 @@ static int gfx_v11_0_sw_init(void *handle)
 		break;
 	}
 
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(11, 0, 0):
+	case IP_VERSION(11, 0, 2):
+	case IP_VERSION(11, 0, 3):
+		if (!adev->gfx.disable_uq &&
+		    adev->gfx.me_fw_version  >= 2420 &&
+		    adev->gfx.pfp_fw_version >= 2580 &&
+		    adev->gfx.mec_fw_version >= 2650 &&
+		    adev->mes.fw_version[0] >= 120) {
+			adev->userq_funcs[AMDGPU_HW_IP_GFX] = &userq_mes_funcs;
+			adev->userq_funcs[AMDGPU_HW_IP_COMPUTE] = &userq_mes_funcs;
+		}
+		break;
+	case IP_VERSION(11, 0, 1):
+	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
+		/* add firmware version checks here */
+		if (0 && !adev->gfx.disable_uq) {
+			adev->userq_funcs[AMDGPU_HW_IP_GFX] = &userq_mes_funcs;
+			adev->userq_funcs[AMDGPU_HW_IP_COMPUTE] = &userq_mes_funcs;
+		}
+		break;
+	default:
+		break;
+	}
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(11, 0, 0):
+	case IP_VERSION(11, 0, 2):
+	case IP_VERSION(11, 0, 3):
+		adev->gfx.cleaner_shader_ptr = gfx_11_0_3_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_11_0_3_cleaner_shader_hex);
+		if (adev->gfx.me_fw_version  >= 2280 &&
+		    adev->gfx.pfp_fw_version >= 2370 &&
+		    adev->gfx.mec_fw_version >= 2450  &&
+		    adev->mes.fw_version[0] >= 99) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	case IP_VERSION(11, 0, 1):
+	case IP_VERSION(11, 0, 4):
+		adev->gfx.cleaner_shader_ptr = gfx_11_0_3_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_11_0_3_cleaner_shader_hex);
+		if (adev->gfx.pfp_fw_version >= 102 &&
+		    adev->gfx.mec_fw_version >= 66 &&
+		    adev->mes.fw_version[0] >= 128) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+		adev->gfx.cleaner_shader_ptr = gfx_11_0_3_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_11_0_3_cleaner_shader_hex);
+		if (adev->gfx.mec_fw_version >= 26 &&
+		    adev->mes.fw_version[0] >= 114) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	case IP_VERSION(11, 5, 2):
+		adev->gfx.cleaner_shader_ptr = gfx_11_0_3_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_11_0_3_cleaner_shader_hex);
+		if (adev->gfx.me_fw_version  >= 12 &&
+		    adev->gfx.pfp_fw_version >= 15 &&
+		    adev->gfx.mec_fw_version >= 15) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	case IP_VERSION(11, 5, 3):
+		adev->gfx.cleaner_shader_ptr = gfx_11_0_3_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_11_0_3_cleaner_shader_hex);
+		if (adev->gfx.me_fw_version  >= 7 &&
+		    adev->gfx.pfp_fw_version >= 8 &&
+		    adev->gfx.mec_fw_version >= 8) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	default:
+		adev->gfx.enable_cleaner_shader = false;
+		break;
+	}
+
 	/* Enable CG flag in one VF mode for enabling RLC safe mode enter/exit */
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(11, 0, 3) &&
-		amdgpu_sriov_is_pp_one_vf(adev))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 0, 3) &&
+	    amdgpu_sriov_is_pp_one_vf(adev))
 		adev->cg_flags = AMD_CG_SUPPORT_GFX_CGCG;
 
 	/* EOP Event */
@@ -1299,6 +1727,13 @@ static int gfx_v11_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	/* Bad opcode Event */
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP,
+			      GFX_11_0_0__SRCID__CP_BAD_OPCODE_ERROR,
+			      &adev->gfx.bad_op_irq);
+	if (r)
+		return r;
+
 	/* Privileged reg */
 	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP,
 			      GFX_11_0_0__SRCID__CP_PRIV_REG_FAULT,
@@ -1313,13 +1748,6 @@ static int gfx_v11_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	/* ECC error */
-	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP,
-				  GFX_11_0_0__SRCID__CP_ECC_ERROR,
-				  &adev->gfx.cp_ecc_error_irq);
-	if (r)
-		return r;
-
 	/* FED error */
 	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GFX,
 				  GFX_11_0_0__SRCID__RLC_GC_FED_INTERRUPT,
@@ -1329,14 +1757,6 @@ static int gfx_v11_0_sw_init(void *handle)
 
 	adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE;
 
-	if (adev->gfx.imu.funcs) {
-		if (adev->gfx.imu.funcs->init_microcode) {
-			r = adev->gfx.imu.funcs->init_microcode(adev);
-			if (r)
-				DRM_ERROR("Failed to load imu firmware!\n");
-		}
-	}
-
 	gfx_v11_0_me_init(adev);
 
 	r = gfx_v11_0_rlc_init(adev);
@@ -1351,55 +1771,84 @@ static int gfx_v11_0_sw_init(void *handle)
 		return r;
 	}
 
-	/* set up the gfx ring */
-	for (i = 0; i < adev->gfx.me.num_me; i++) {
-		for (j = 0; j < adev->gfx.me.num_queue_per_pipe; j++) {
-			for (k = 0; k < adev->gfx.me.num_pipe_per_me; k++) {
-				if (!amdgpu_gfx_is_me_queue_enabled(adev, i, k, j))
-					continue;
-
-				r = gfx_v11_0_gfx_ring_init(adev, ring_id,
-							    i, k, j);
-				if (r)
-					return r;
-				ring_id++;
+	if (adev->gfx.num_gfx_rings) {
+		ring_id = 0;
+		/* set up the gfx ring */
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < num_queue_per_pipe; j++) {
+				for (k = 0; k < adev->gfx.me.num_pipe_per_me; k++) {
+					if (!amdgpu_gfx_is_me_queue_enabled(adev, i, k, j))
+						continue;
+
+					r = gfx_v11_0_gfx_ring_init(adev, ring_id,
+								    i, k, j);
+					if (r)
+						return r;
+					ring_id++;
+				}
 			}
 		}
 	}
 
-	ring_id = 0;
-	/* set up the compute queues - allocate horizontally across pipes */
-	for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
-		for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
-			for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
-				if (!amdgpu_gfx_is_mec_queue_enabled(adev, i, k,
-								     j))
-					continue;
+	if (adev->gfx.num_compute_rings) {
+		ring_id = 0;
+		/* set up the compute queues - allocate horizontally across pipes */
+		for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
+			for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
+				for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
+					if (!amdgpu_gfx_is_mec_queue_enabled(adev, 0, i,
+									     k, j))
+						continue;
 
-				r = gfx_v11_0_compute_ring_init(adev, ring_id,
-								i, k, j);
-				if (r)
-					return r;
+					r = gfx_v11_0_compute_ring_init(adev, ring_id,
+									i, k, j);
+					if (r)
+						return r;
 
-				ring_id++;
+					ring_id++;
+				}
 			}
 		}
 	}
 
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(11, 0, 0):
+	case IP_VERSION(11, 0, 2):
+	case IP_VERSION(11, 0, 3):
+		if ((adev->gfx.me_fw_version >= 2280) &&
+		    (adev->gfx.mec_fw_version >= 2410) &&
+		    !amdgpu_sriov_vf(adev) &&
+		    !adev->debug_disable_gpu_ring_reset) {
+			adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+			adev->gfx.gfx_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+		}
+		break;
+	default:
+		if (!amdgpu_sriov_vf(adev) &&
+		    !adev->debug_disable_gpu_ring_reset) {
+			adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+			adev->gfx.gfx_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+		}
+		break;
+	}
+
 	if (!adev->enable_mes_kiq) {
-		r = amdgpu_gfx_kiq_init(adev, GFX11_MEC_HPD_SIZE);
+		r = amdgpu_gfx_kiq_init(adev, GFX11_MEC_HPD_SIZE, 0);
 		if (r) {
 			DRM_ERROR("Failed to init KIQ BOs!\n");
 			return r;
 		}
 
-		kiq = &adev->gfx.kiq;
-		r = amdgpu_gfx_kiq_init_ring(adev, &kiq->ring, &kiq->irq);
+		r = amdgpu_gfx_kiq_init_ring(adev, xcc_id);
 		if (r)
 			return r;
 	}
 
-	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v11_compute_mqd));
+	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v11_compute_mqd), 0);
 	if (r)
 		return r;
 
@@ -1419,6 +1868,12 @@ static int gfx_v11_0_sw_init(void *handle)
 		return -EINVAL;
 	}
 
+	gfx_v11_0_alloc_ip_dump(adev);
+
+	r = amdgpu_gfx_sysfs_init(adev);
+	if (r)
+		return r;
+
 	return 0;
 }
 
@@ -1451,23 +1906,25 @@ static void gfx_v11_0_rlc_autoload_buffer_fini(struct amdgpu_device *adev)
 			(void **)&adev->gfx.rlc.rlc_autoload_ptr);
 }
 
-static int gfx_v11_0_sw_fini(void *handle)
+static int gfx_v11_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.gfx_ring[i]);
 	for (i = 0; i < adev->gfx.num_compute_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
 
-	amdgpu_gfx_mqd_sw_fini(adev);
+	amdgpu_gfx_mqd_sw_fini(adev, 0);
 
 	if (!adev->enable_mes_kiq) {
-		amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring);
-		amdgpu_gfx_kiq_fini(adev);
+		amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq[0].ring);
+		amdgpu_gfx_kiq_fini(adev, 0);
 	}
 
+	amdgpu_gfx_cleaner_shader_sw_fini(adev);
+
 	gfx_v11_0_pfp_fini(adev);
 	gfx_v11_0_me_fini(adev);
 	gfx_v11_0_rlc_fini(adev);
@@ -1478,11 +1935,17 @@ static int gfx_v11_0_sw_fini(void *handle)
 
 	gfx_v11_0_free_microcode(adev);
 
+	amdgpu_gfx_sysfs_fini(adev);
+
+	kfree(adev->gfx.ip_dump_core);
+	kfree(adev->gfx.ip_dump_compute_queues);
+	kfree(adev->gfx.ip_dump_gfx_queues);
+
 	return 0;
 }
 
 static void gfx_v11_0_select_se_sh(struct amdgpu_device *adev, u32 se_num,
-				   u32 sh_num, u32 instance)
+				   u32 sh_num, u32 instance, int xcc_id)
 {
 	u32 data;
 
@@ -1547,6 +2010,7 @@ static u32 gfx_v11_0_get_rb_active_bitmap(struct amdgpu_device *adev)
 
 static void gfx_v11_0_setup_rb(struct amdgpu_device *adev)
 {
+	u32 rb_bitmap_per_sa;
 	u32 rb_bitmap_width_per_sa;
 	u32 max_sa;
 	u32 active_sa_bitmap;
@@ -1564,12 +2028,14 @@ static void gfx_v11_0_setup_rb(struct amdgpu_device *adev)
 		 adev->gfx.config.max_sh_per_se;
 	rb_bitmap_width_per_sa = adev->gfx.config.max_backends_per_se /
 				 adev->gfx.config.max_sh_per_se;
+	rb_bitmap_per_sa = amdgpu_gfx_create_bitmask(rb_bitmap_width_per_sa);
+
 	for (i = 0; i < max_sa; i++) {
 		if (active_sa_bitmap & (1 << i))
-			active_rb_bitmap |= (0x3 << (i * rb_bitmap_width_per_sa));
+			active_rb_bitmap |= (rb_bitmap_per_sa << (i * rb_bitmap_width_per_sa));
 	}
 
-	active_rb_bitmap |= global_active_rb_bitmap;
+	active_rb_bitmap &= global_active_rb_bitmap;
 	adev->gfx.config.backend_enable_mask = active_rb_bitmap;
 	adev->gfx.config.num_rbs = hweight32(active_rb_bitmap);
 }
@@ -1603,12 +2069,15 @@ static void gfx_v11_0_init_compute_vmid(struct amdgpu_device *adev)
 		/* Enable trap for each kfd vmid. */
 		data = RREG32_SOC15(GC, 0, regSPI_GDBG_PER_VMID_CNTL);
 		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+		WREG32_SOC15(GC, 0, regSPI_GDBG_PER_VMID_CNTL, data);
 	}
 	soc21_grbm_select(adev, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
 
-	/* Initialize all compute VMIDs to have no GDS, GWS, or OA
-	   acccess. These should be enabled by FW for target VMIDs. */
+	/*
+	 * Initialize all compute VMIDs to have no GDS, GWS, or OA
+	 * access. These should be enabled by FW for target VMIDs.
+	 */
 	for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) {
 		WREG32_SOC15_OFFSET(GC, 0, regGDS_VMID0_BASE, 2 * i, 0);
 		WREG32_SOC15_OFFSET(GC, 0, regGDS_VMID0_SIZE, 2 * i, 0);
@@ -1672,7 +2141,7 @@ static void gfx_v11_0_constants_init(struct amdgpu_device *adev)
 	/* XXX SH_MEM regs */
 	/* where to put LDS, scratch, GPUVM in FSA64 space */
 	mutex_lock(&adev->srbm_mutex);
-	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) {
+	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
 		soc21_grbm_select(adev, 0, 0, 0, i);
 		/* CP and shaders */
 		WREG32_SOC15(GC, 0, regSH_MEM_CONFIG, DEFAULT_SH_MEM_CONFIG);
@@ -1692,26 +2161,74 @@ static void gfx_v11_0_constants_init(struct amdgpu_device *adev)
 	gfx_v11_0_init_gds_vmid(adev);
 }
 
+static u32 gfx_v11_0_get_cpg_int_cntl(struct amdgpu_device *adev,
+				      int me, int pipe)
+{
+	if (me != 0)
+		return 0;
+
+	switch (pipe) {
+	case 0:
+		return SOC15_REG_OFFSET(GC, 0, regCP_INT_CNTL_RING0);
+	case 1:
+		return SOC15_REG_OFFSET(GC, 0, regCP_INT_CNTL_RING1);
+	default:
+		return 0;
+	}
+}
+
+static u32 gfx_v11_0_get_cpc_int_cntl(struct amdgpu_device *adev,
+				      int me, int pipe)
+{
+	/*
+	 * amdgpu controls only the first MEC. That's why this function only
+	 * handles the setting of interrupts for this specific MEC. All other
+	 * pipes' interrupts are set by amdkfd.
+	 */
+	if (me != 1)
+		return 0;
+
+	switch (pipe) {
+	case 0:
+		return SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE0_INT_CNTL);
+	case 1:
+		return SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE1_INT_CNTL);
+	case 2:
+		return SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE2_INT_CNTL);
+	case 3:
+		return SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE3_INT_CNTL);
+	default:
+		return 0;
+	}
+}
+
 static void gfx_v11_0_enable_gui_idle_interrupt(struct amdgpu_device *adev,
 					       bool enable)
 {
-	u32 tmp;
+	u32 tmp, cp_int_cntl_reg;
+	int i, j;
 
 	if (amdgpu_sriov_vf(adev))
 		return;
 
-	tmp = RREG32_SOC15(GC, 0, regCP_INT_CNTL_RING0);
-
-	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE,
-			    enable ? 1 : 0);
-	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE,
-			    enable ? 1 : 0);
-	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE,
-			    enable ? 1 : 0);
-	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE,
-			    enable ? 1 : 0);
-
-	WREG32_SOC15(GC, 0, regCP_INT_CNTL_RING0, tmp);
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			cp_int_cntl_reg = gfx_v11_0_get_cpg_int_cntl(adev, i, j);
+
+			if (cp_int_cntl_reg) {
+				tmp = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE,
+						    enable ? 1 : 0);
+				WREG32_SOC15_IP(GC, cp_int_cntl_reg, tmp);
+			}
+		}
+	}
 }
 
 static int gfx_v11_0_init_csb(struct amdgpu_device *adev)
@@ -1923,7 +2440,7 @@ static int gfx_v11_0_rlc_load_microcode(struct amdgpu_device *adev)
 			if (version_minor == 3)
 				gfx_v11_0_load_rlcp_rlcv_microcode(adev);
 		}
-		
+
 		return 0;
 	}
 
@@ -1993,7 +2510,7 @@ static int gfx_v11_0_config_me_cache(struct amdgpu_device *adev, uint64_t addr)
 	}
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_BASE_CNTL);
 	tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, VMID, 0);
@@ -2037,7 +2554,7 @@ static int gfx_v11_0_config_pfp_cache(struct amdgpu_device *adev, uint64_t addr)
 	}
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_CNTL);
 	tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, VMID, 0);
@@ -2082,7 +2599,7 @@ static int gfx_v11_0_config_mec_cache(struct amdgpu_device *adev, uint64_t addr)
 	}
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_CNTL);
 	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, CACHE_POLICY, 0);
@@ -2522,8 +3039,14 @@ static int gfx_v11_0_wait_for_rlc_autoload_complete(struct amdgpu_device *adev)
 	for (i = 0; i < adev->usec_timeout; i++) {
 		cp_status = RREG32_SOC15(GC, 0, regCP_STAT);
 
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(11, 0, 1) ||
-				adev->ip_versions[GC_HWIP][0] == IP_VERSION(11, 0, 4))
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			    IP_VERSION(11, 0, 1) ||
+		    amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			    IP_VERSION(11, 0, 4) ||
+		    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 5, 0) ||
+		    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 5, 1) ||
+		    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 5, 2) ||
+		    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(11, 5, 3))
 			bootload_status = RREG32_SOC15(GC, 0,
 					regRLC_RLCS_BOOTLOAD_STATUS_gc_11_0_1);
 		else
@@ -2712,7 +3235,7 @@ static int gfx_v11_0_cp_gfx_load_pfp_microcode_rs64(struct amdgpu_device *adev)
 	amdgpu_bo_unreserve(adev->gfx.pfp.pfp_fw_data_obj);
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_LO,
 		lower_32_bits(adev->gfx.pfp.pfp_fw_gpu_addr));
@@ -2930,7 +3453,7 @@ static int gfx_v11_0_cp_gfx_load_me_microcode_rs64(struct amdgpu_device *adev)
 	amdgpu_bo_unreserve(adev->gfx.me.me_fw_data_obj);
 
 	if (amdgpu_emu_mode == 1)
-		adev->hdp.funcs->flush_hdp(adev, NULL);
+		amdgpu_device_flush_hdp(adev, NULL);
 
 	WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_LO,
 		lower_32_bits(adev->gfx.me.me_fw_gpu_addr));
@@ -3193,7 +3716,6 @@ static int gfx_v11_0_cp_gfx_resume(struct amdgpu_device *adev)
 	u32 tmp;
 	u32 rb_bufsz;
 	u64 rb_addr, rptr_addr, wptr_gpu_addr;
-	u32 i;
 
 	/* Set the write pointer delay */
 	WREG32_SOC15(GC, 0, regCP_RB_WPTR_DELAY, 0);
@@ -3217,7 +3739,7 @@ static int gfx_v11_0_cp_gfx_resume(struct amdgpu_device *adev)
 	WREG32_SOC15(GC, 0, regCP_RB0_WPTR, lower_32_bits(ring->wptr));
 	WREG32_SOC15(GC, 0, regCP_RB0_WPTR_HI, upper_32_bits(ring->wptr));
 
-	/* set the wb address wether it's enabled or not */
+	/* set the wb address whether it's enabled or not */
 	rptr_addr = ring->rptr_gpu_addr;
 	WREG32_SOC15(GC, 0, regCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr));
 	WREG32_SOC15(GC, 0, regCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) &
@@ -3255,7 +3777,7 @@ static int gfx_v11_0_cp_gfx_resume(struct amdgpu_device *adev)
 		ring->wptr = 0;
 		WREG32_SOC15(GC, 0, regCP_RB1_WPTR, lower_32_bits(ring->wptr));
 		WREG32_SOC15(GC, 0, regCP_RB1_WPTR_HI, upper_32_bits(ring->wptr));
-		/* Set the wb address wether it's enabled or not */
+		/* Set the wb address whether it's enabled or not */
 		rptr_addr = ring->rptr_gpu_addr;
 		WREG32_SOC15(GC, 0, regCP_RB1_RPTR_ADDR, lower_32_bits(rptr_addr));
 		WREG32_SOC15(GC, 0, regCP_RB1_RPTR_ADDR_HI, upper_32_bits(rptr_addr) &
@@ -3285,11 +3807,6 @@ static int gfx_v11_0_cp_gfx_resume(struct amdgpu_device *adev)
 	/* start the ring */
 	gfx_v11_0_cp_gfx_start(adev);
 
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
-		ring = &adev->gfx.gfx_ring[i];
-		ring->sched.ready = true;
-	}
-
 	return 0;
 }
 
@@ -3335,8 +3852,6 @@ static void gfx_v11_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
 		WREG32_SOC15(GC, 0, regCP_MEC_CNTL, data);
 	}
 
-	adev->gfx.kiq.ring.sched.ready = enable;
-
 	udelay(50);
 }
 
@@ -3373,7 +3888,7 @@ static int gfx_v11_0_cp_compute_load_microcode(struct amdgpu_device *adev)
 	}
 
 	memcpy(fw, fw_data, fw_size);
-	
+
 	amdgpu_bo_kunmap(adev->gfx.mec.mec_fw_obj);
 	amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_obj);
 
@@ -3532,9 +4047,7 @@ static void gfx_v11_0_kiq_setting(struct amdgpu_ring *ring)
 	tmp = RREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS);
 	tmp &= 0xffffff00;
 	tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
-	WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp);
-	tmp |= 0x80;
-	WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp);
+	WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp | 0x80);
 }
 
 static void gfx_v11_0_cp_set_doorbell_range(struct amdgpu_device *adev)
@@ -3552,6 +4065,24 @@ static void gfx_v11_0_cp_set_doorbell_range(struct amdgpu_device *adev)
 		     (adev->doorbell_index.userqueue_end * 2) << 2);
 }
 
+static void gfx_v11_0_gfx_mqd_set_priority(struct amdgpu_device *adev,
+					   struct v11_gfx_mqd *mqd,
+					   struct amdgpu_mqd_prop *prop)
+{
+	bool priority = 0;
+	u32 tmp;
+
+	/* set up default queue priority level
+	 * 0x0 = low priority, 0x1 = high priority
+	 */
+	if (prop->hqd_pipe_priority == AMDGPU_GFX_PIPE_PRIO_HIGH)
+		priority = 1;
+
+	tmp = regCP_GFX_HQD_QUEUE_PRIORITY_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUEUE_PRIORITY, PRIORITY_LEVEL, priority);
+	mqd->cp_gfx_hqd_queue_priority = tmp;
+}
+
 static int gfx_v11_0_gfx_mqd_init(struct amdgpu_device *adev, void *m,
 				  struct amdgpu_mqd_prop *prop)
 {
@@ -3569,25 +4100,22 @@ static int gfx_v11_0_gfx_mqd_init(struct amdgpu_device *adev, void *m,
 	mqd->cp_mqd_base_addr_hi = upper_32_bits(prop->mqd_gpu_addr);
 
 	/* set up mqd control */
-	tmp = RREG32_SOC15(GC, 0, regCP_GFX_MQD_CONTROL);
+	tmp = regCP_GFX_MQD_CONTROL_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, VMID, 0);
 	tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, PRIV_STATE, 1);
 	tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, CACHE_POLICY, 0);
 	mqd->cp_gfx_mqd_control = tmp;
 
 	/* set up gfx_hqd_vimd with 0x0 to indicate the ring buffer's vmid */
-	tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_VMID);
+	tmp = regCP_GFX_HQD_VMID_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_VMID, VMID, 0);
 	mqd->cp_gfx_hqd_vmid = 0;
 
-	/* set up default queue priority level
-	 * 0x0 = low priority, 0x1 = high priority */
-	tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_QUEUE_PRIORITY);
-	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUEUE_PRIORITY, PRIORITY_LEVEL, 0);
-	mqd->cp_gfx_hqd_queue_priority = tmp;
+	/* set up gfx queue priority */
+	gfx_v11_0_gfx_mqd_set_priority(adev, mqd, prop);
 
 	/* set up time quantum */
-	tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_QUANTUM);
+	tmp = regCP_GFX_HQD_QUANTUM_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUANTUM, QUANTUM_EN, 1);
 	mqd->cp_gfx_hqd_quantum = tmp;
 
@@ -3609,16 +4137,20 @@ static int gfx_v11_0_gfx_mqd_init(struct amdgpu_device *adev, void *m,
 
 	/* set up the gfx_hqd_control, similar as CP_RB0_CNTL */
 	rb_bufsz = order_base_2(prop->queue_size / 4) - 1;
-	tmp = RREG32_SOC15(GC, 0, regCP_GFX_HQD_CNTL);
+	tmp = regCP_GFX_HQD_CNTL_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_BUFSZ, rb_bufsz);
 	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_BLKSZ, rb_bufsz - 2);
 #ifdef __BIG_ENDIAN
 	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, BUF_SWAP, 1);
 #endif
+	if (prop->tmz_queue)
+		tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, TMZ_MATCH, 1);
+	if (!prop->kernel_queue)
+		tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_NON_PRIV, 1);
 	mqd->cp_gfx_hqd_cntl = tmp;
 
 	/* set up cp_doorbell_control */
-	tmp = RREG32_SOC15(GC, 0, regCP_RB_DOORBELL_CONTROL);
+	tmp = regCP_RB_DOORBELL_CONTROL_DEFAULT;
 	if (prop->use_doorbell) {
 		tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL,
 				    DOORBELL_OFFSET, prop->doorbell_index);
@@ -3630,164 +4162,67 @@ static int gfx_v11_0_gfx_mqd_init(struct amdgpu_device *adev, void *m,
 	mqd->cp_rb_doorbell_control = tmp;
 
 	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
-	mqd->cp_gfx_hqd_rptr = RREG32_SOC15(GC, 0, regCP_GFX_HQD_RPTR);
+	mqd->cp_gfx_hqd_rptr = regCP_GFX_HQD_RPTR_DEFAULT;
 
 	/* active the queue */
 	mqd->cp_gfx_hqd_active = 1;
 
-	return 0;
-}
-
-#ifdef BRING_UP_DEBUG
-static int gfx_v11_0_gfx_queue_init_register(struct amdgpu_ring *ring)
-{
-	struct amdgpu_device *adev = ring->adev;
-	struct v11_gfx_mqd *mqd = ring->mqd_ptr;
-
-	/* set mmCP_GFX_HQD_WPTR/_HI to 0 */
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_WPTR, mqd->cp_gfx_hqd_wptr);
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_WPTR_HI, mqd->cp_gfx_hqd_wptr_hi);
-
-	/* set GFX_MQD_BASE */
-	WREG32_SOC15(GC, 0, regCP_MQD_BASE_ADDR, mqd->cp_mqd_base_addr);
-	WREG32_SOC15(GC, 0, regCP_MQD_BASE_ADDR_HI, mqd->cp_mqd_base_addr_hi);
-
-	/* set GFX_MQD_CONTROL */
-	WREG32_SOC15(GC, 0, regCP_GFX_MQD_CONTROL, mqd->cp_gfx_mqd_control);
-
-	/* set GFX_HQD_VMID to 0 */
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_VMID, mqd->cp_gfx_hqd_vmid);
-
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_QUEUE_PRIORITY,
-			mqd->cp_gfx_hqd_queue_priority);
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_QUANTUM, mqd->cp_gfx_hqd_quantum);
-
-	/* set GFX_HQD_BASE, similar as CP_RB_BASE */
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_BASE, mqd->cp_gfx_hqd_base);
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_BASE_HI, mqd->cp_gfx_hqd_base_hi);
-
-	/* set GFX_HQD_RPTR_ADDR, similar as CP_RB_RPTR */
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_RPTR_ADDR, mqd->cp_gfx_hqd_rptr_addr);
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_RPTR_ADDR_HI, mqd->cp_gfx_hqd_rptr_addr_hi);
-
-	/* set GFX_HQD_CNTL, similar as CP_RB_CNTL */
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_CNTL, mqd->cp_gfx_hqd_cntl);
-
-	/* set RB_WPTR_POLL_ADDR */
-	WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_ADDR_LO, mqd->cp_rb_wptr_poll_addr_lo);
-	WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_ADDR_HI, mqd->cp_rb_wptr_poll_addr_hi);
-
-	/* set RB_DOORBELL_CONTROL */
-	WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_CONTROL, mqd->cp_rb_doorbell_control);
-
-	/* active the queue */
-	WREG32_SOC15(GC, 0, regCP_GFX_HQD_ACTIVE, mqd->cp_gfx_hqd_active);
+	/* set gfx UQ items */
+	mqd->shadow_base_lo = lower_32_bits(prop->shadow_addr);
+	mqd->shadow_base_hi = upper_32_bits(prop->shadow_addr);
+	mqd->gds_bkup_base_lo = lower_32_bits(prop->gds_bkup_addr);
+	mqd->gds_bkup_base_hi = upper_32_bits(prop->gds_bkup_addr);
+	mqd->fw_work_area_base_lo = lower_32_bits(prop->csa_addr);
+	mqd->fw_work_area_base_hi = upper_32_bits(prop->csa_addr);
+	mqd->fence_address_lo = lower_32_bits(prop->fence_address);
+	mqd->fence_address_hi = upper_32_bits(prop->fence_address);
 
 	return 0;
 }
-#endif
 
-static int gfx_v11_0_gfx_init_queue(struct amdgpu_ring *ring)
+static int gfx_v11_0_kgq_init_queue(struct amdgpu_ring *ring, bool reset)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v11_gfx_mqd *mqd = ring->mqd_ptr;
 	int mqd_idx = ring - &adev->gfx.gfx_ring[0];
 
-	if (!amdgpu_in_reset(adev) && !adev->in_suspend) {
+	if (!reset && !amdgpu_in_reset(adev) && !adev->in_suspend) {
 		memset((void *)mqd, 0, sizeof(*mqd));
 		mutex_lock(&adev->srbm_mutex);
 		soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
 		amdgpu_ring_init_mqd(ring);
-#ifdef BRING_UP_DEBUG
-		gfx_v11_0_gfx_queue_init_register(ring);
-#endif
 		soc21_grbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 		if (adev->gfx.me.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.me.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
-	} else if (amdgpu_in_reset(adev)) {
-		/* reset mqd with the backup copy */
+			memcpy_fromio(adev->gfx.me.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+	} else {
+		/* restore mqd with the backup copy */
 		if (adev->gfx.me.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.me.mqd_backup[mqd_idx], sizeof(*mqd));
+			memcpy_toio(mqd, adev->gfx.me.mqd_backup[mqd_idx], sizeof(*mqd));
 		/* reset the ring */
 		ring->wptr = 0;
 		*ring->wptr_cpu_addr = 0;
 		amdgpu_ring_clear_ring(ring);
-#ifdef BRING_UP_DEBUG
-		mutex_lock(&adev->srbm_mutex);
-		soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
-		gfx_v11_0_gfx_queue_init_register(ring);
-		soc21_grbm_select(adev, 0, 0, 0, 0);
-		mutex_unlock(&adev->srbm_mutex);
-#endif
-	} else {
-		amdgpu_ring_clear_ring(ring);
 	}
 
 	return 0;
 }
 
-#ifndef BRING_UP_DEBUG
-static int gfx_v11_0_kiq_enable_kgq(struct amdgpu_device *adev)
-{
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
-	int r, i;
-
-	if (!kiq->pmf || !kiq->pmf->kiq_map_queues)
-		return -EINVAL;
-
-	r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size *
-					adev->gfx.num_gfx_rings);
-	if (r) {
-		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
-		return r;
-	}
-
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
-		kiq->pmf->kiq_map_queues(kiq_ring, &adev->gfx.gfx_ring[i]);
-
-	return amdgpu_ring_test_helper(kiq_ring);
-}
-#endif
-
 static int gfx_v11_0_cp_async_gfx_ring_resume(struct amdgpu_device *adev)
 {
 	int r, i;
-	struct amdgpu_ring *ring;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
-		ring = &adev->gfx.gfx_ring[i];
-
-		r = amdgpu_bo_reserve(ring->mqd_obj, false);
-		if (unlikely(r != 0))
-			goto done;
-
-		r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
-		if (!r) {
-			r = gfx_v11_0_gfx_init_queue(ring);
-			amdgpu_bo_kunmap(ring->mqd_obj);
-			ring->mqd_ptr = NULL;
-		}
-		amdgpu_bo_unreserve(ring->mqd_obj);
+		r = gfx_v11_0_kgq_init_queue(&adev->gfx.gfx_ring[i], false);
 		if (r)
-			goto done;
+			return r;
 	}
-#ifndef BRING_UP_DEBUG
-	r = gfx_v11_0_kiq_enable_kgq(adev);
-	if (r)
-		goto done;
-#endif
-	r = gfx_v11_0_cp_gfx_start(adev);
+
+	r = amdgpu_gfx_enable_kgq(adev, 0);
 	if (r)
-		goto done;
+		return r;
 
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
-		ring = &adev->gfx.gfx_ring[i];
-		ring->sched.ready = true;
-	}
-done:
-	return r;
+	return gfx_v11_0_cp_gfx_start(adev);
 }
 
 static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
@@ -3810,14 +4245,14 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
 	mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr);
 
 	/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
-	tmp = RREG32_SOC15(GC, 0, regCP_HQD_EOP_CONTROL);
+	tmp = regCP_HQD_EOP_CONTROL_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_HQD_EOP_CONTROL, EOP_SIZE,
 			(order_base_2(GFX11_MEC_HPD_SIZE / 4) - 1));
 
 	mqd->cp_hqd_eop_control = tmp;
 
 	/* enable doorbell? */
-	tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL);
+	tmp = regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT;
 
 	if (prop->use_doorbell) {
 		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
@@ -3846,7 +4281,7 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
 	mqd->cp_mqd_base_addr_hi = upper_32_bits(prop->mqd_gpu_addr);
 
 	/* set MQD vmid to 0 */
-	tmp = RREG32_SOC15(GC, 0, regCP_MQD_CONTROL);
+	tmp = regCP_MQD_CONTROL_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_MQD_CONTROL, VMID, 0);
 	mqd->cp_mqd_control = tmp;
 
@@ -3856,15 +4291,20 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
 	mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr);
 
 	/* set up the HQD, this is similar to CP_RB0_CNTL */
-	tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_CONTROL);
+	tmp = regCP_HQD_PQ_CONTROL_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE,
 			    (order_base_2(prop->queue_size / 4) - 1));
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE,
 			    (order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1));
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 0);
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TUNNEL_DISPATCH, 0);
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1);
-	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 1);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TUNNEL_DISPATCH,
+			    prop->allow_tunneling);
+	if (prop->kernel_queue) {
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1);
+	}
+	if (prop->tmz_queue)
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TMZ, 1);
 	mqd->cp_hqd_pq_control = tmp;
 
 	/* set the wb address whether it's enabled or not */
@@ -3881,7 +4321,7 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
 	tmp = 0;
 	/* enable the doorbell if requested */
 	if (prop->use_doorbell) {
-		tmp = RREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL);
+		tmp = regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT;
 		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
 				DOORBELL_OFFSET, prop->doorbell_index);
 
@@ -3896,17 +4336,17 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
 	mqd->cp_hqd_pq_doorbell_control = tmp;
 
 	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
-	mqd->cp_hqd_pq_rptr = RREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR);
+	mqd->cp_hqd_pq_rptr = regCP_HQD_PQ_RPTR_DEFAULT;
 
 	/* set the vmid for the queue */
 	mqd->cp_hqd_vmid = 0;
 
-	tmp = RREG32_SOC15(GC, 0, regCP_HQD_PERSISTENT_STATE);
+	tmp = regCP_HQD_PERSISTENT_STATE_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_HQD_PERSISTENT_STATE, PRELOAD_SIZE, 0x55);
 	mqd->cp_hqd_persistent_state = tmp;
 
 	/* set MIN_IB_AVAIL_SIZE */
-	tmp = RREG32_SOC15(GC, 0, regCP_HQD_IB_CONTROL);
+	tmp = regCP_HQD_IB_CONTROL_DEFAULT;
 	tmp = REG_SET_FIELD(tmp, CP_HQD_IB_CONTROL, MIN_IB_AVAIL_SIZE, 3);
 	mqd->cp_hqd_ib_control = tmp;
 
@@ -3916,6 +4356,10 @@ static int gfx_v11_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
 
 	mqd->cp_hqd_active = prop->hqd_active;
 
+	/* set UQ fenceaddress */
+	mqd->fence_address_lo = lower_32_bits(prop->fence_address);
+	mqd->fence_address_hi = upper_32_bits(prop->fence_address);
+
 	return 0;
 }
 
@@ -4033,14 +4477,13 @@ static int gfx_v11_0_kiq_init_queue(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v11_compute_mqd *mqd = ring->mqd_ptr;
-	int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS;
 
 	gfx_v11_0_kiq_setting(ring);
 
 	if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
 		/* reset MQD to a clean status */
-		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
+		if (adev->gfx.kiq[0].mqd_backup)
+			memcpy_toio(mqd, adev->gfx.kiq[0].mqd_backup, sizeof(*mqd));
 
 		/* reset ring buffer */
 		ring->wptr = 0;
@@ -4062,20 +4505,20 @@ static int gfx_v11_0_kiq_init_queue(struct amdgpu_ring *ring)
 		soc21_grbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 
-		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+		if (adev->gfx.kiq[0].mqd_backup)
+			memcpy_fromio(adev->gfx.kiq[0].mqd_backup, mqd, sizeof(*mqd));
 	}
 
 	return 0;
 }
 
-static int gfx_v11_0_kcq_init_queue(struct amdgpu_ring *ring)
+static int gfx_v11_0_kcq_init_queue(struct amdgpu_ring *ring, bool reset)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v11_compute_mqd *mqd = ring->mqd_ptr;
 	int mqd_idx = ring - &adev->gfx.compute_ring[0];
 
-	if (!amdgpu_in_reset(adev) && !adev->in_suspend) {
+	if (!reset && !amdgpu_in_reset(adev) && !adev->in_suspend) {
 		memset((void *)mqd, 0, sizeof(*mqd));
 		mutex_lock(&adev->srbm_mutex);
 		soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
@@ -4084,18 +4527,15 @@ static int gfx_v11_0_kcq_init_queue(struct amdgpu_ring *ring)
 		mutex_unlock(&adev->srbm_mutex);
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
-	} else if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
-		/* reset MQD to a clean status */
+			memcpy_fromio(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+	} else {
+		/* restore MQD to a clean status */
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
-
+			memcpy_toio(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
 		/* reset ring buffer */
 		ring->wptr = 0;
 		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, 0);
 		amdgpu_ring_clear_ring(ring);
-	} else {
-		amdgpu_ring_clear_ring(ring);
 	}
 
 	return 0;
@@ -4103,57 +4543,24 @@ static int gfx_v11_0_kcq_init_queue(struct amdgpu_ring *ring)
 
 static int gfx_v11_0_kiq_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring;
-	int r;
-
-	ring = &adev->gfx.kiq.ring;
-
-	r = amdgpu_bo_reserve(ring->mqd_obj, false);
-	if (unlikely(r != 0))
-		return r;
-
-	r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
-	if (unlikely(r != 0)) {
-		amdgpu_bo_unreserve(ring->mqd_obj);
-		return r;
-	}
-
-	gfx_v11_0_kiq_init_queue(ring);
-	amdgpu_bo_kunmap(ring->mqd_obj);
-	ring->mqd_ptr = NULL;
-	amdgpu_bo_unreserve(ring->mqd_obj);
-	ring->sched.ready = true;
+	gfx_v11_0_kiq_init_queue(&adev->gfx.kiq[0].ring);
 	return 0;
 }
 
 static int gfx_v11_0_kcq_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring = NULL;
-	int r = 0, i;
+	int i, r;
 
 	if (!amdgpu_async_gfx_ring)
 		gfx_v11_0_cp_compute_enable(adev, true);
 
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-		ring = &adev->gfx.compute_ring[i];
-
-		r = amdgpu_bo_reserve(ring->mqd_obj, false);
-		if (unlikely(r != 0))
-			goto done;
-		r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
-		if (!r) {
-			r = gfx_v11_0_kcq_init_queue(ring);
-			amdgpu_bo_kunmap(ring->mqd_obj);
-			ring->mqd_ptr = NULL;
-		}
-		amdgpu_bo_unreserve(ring->mqd_obj);
+		r = gfx_v11_0_kcq_init_queue(&adev->gfx.compute_ring[i], false);
 		if (r)
-			goto done;
+			return r;
 	}
 
-	r = amdgpu_gfx_enable_kcq(adev);
-done:
-	return r;
+	return amdgpu_gfx_enable_kcq(adev, 0);
 }
 
 static int gfx_v11_0_cp_resume(struct amdgpu_device *adev)
@@ -4206,11 +4613,23 @@ static int gfx_v11_0_cp_resume(struct amdgpu_device *adev)
 			return r;
 	}
 
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
-		ring = &adev->gfx.gfx_ring[i];
-		r = amdgpu_ring_test_helper(ring);
-		if (r)
-			return r;
+	if (adev->gfx.disable_kq) {
+		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+			ring = &adev->gfx.gfx_ring[i];
+			/* we don't want to set ring->ready */
+			r = amdgpu_ring_test_ring(ring);
+			if (r)
+				return r;
+		}
+		if (amdgpu_async_gfx_ring)
+			amdgpu_gfx_disable_kgq(adev, 0);
+	} else {
+		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+			ring = &adev->gfx.gfx_ring[i];
+			r = amdgpu_ring_test_helper(ring);
+			if (r)
+				return r;
+		}
 	}
 
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
@@ -4238,13 +4657,14 @@ static int gfx_v11_0_gfxhub_enable(struct amdgpu_device *adev)
 	if (r)
 		return r;
 
-	adev->hdp.funcs->flush_hdp(adev, NULL);
+	amdgpu_device_flush_hdp(adev, NULL);
 
-	value = (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS) ?
-		false : true;
+	value = amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS;
 
 	adev->gfxhub.funcs->set_fault_enable_default(adev, value);
-	amdgpu_gmc_flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB_0, 0);
+	/* TODO investigate why this and the hdp flush above is needed,
+	 * are we missing a flush somewhere else? */
+	adev->gmc.gmc_funcs->flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB(0), 0);
 
 	return 0;
 }
@@ -4317,21 +4737,24 @@ static void gfx_v11_0_disable_gpa_mode(struct amdgpu_device *adev)
 	WREG32_SOC15(GC, 0, regCPG_PSP_DEBUG, data);
 }
 
-static int gfx_v11_0_hw_init(void *handle)
+static int gfx_v11_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	amdgpu_gfx_cleaner_shader_init(adev, adev->gfx.cleaner_shader_size,
+				       adev->gfx.cleaner_shader_ptr);
 
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) {
 		if (adev->gfx.imu.funcs) {
 			/* RLC autoload sequence 1: Program rlc ram */
 			if (adev->gfx.imu.funcs->program_rlc_ram)
 				adev->gfx.imu.funcs->program_rlc_ram(adev);
+			/* rlc autoload firmware */
+			r = gfx_v11_0_rlc_backdoor_autoload_enable(adev);
+			if (r)
+				return r;
 		}
-		/* rlc autoload firmware */
-		r = gfx_v11_0_rlc_backdoor_autoload_enable(adev);
-		if (r)
-			return r;
 	} else {
 		if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
 			if (adev->gfx.imu.funcs && (amdgpu_dpm > 0)) {
@@ -4380,11 +4803,9 @@ static int gfx_v11_0_hw_init(void *handle)
 		 * loaded firstly, so in direct type, it has to load smc ucode
 		 * here before rlc.
 		 */
-		if (!(adev->flags & AMD_IS_APU)) {
-			r = amdgpu_pm_load_smu_firmware(adev, NULL);
-			if (r)
-				return r;
-		}
+		r = amdgpu_pm_load_smu_firmware(adev, NULL);
+		if (r)
+			return r;
 	}
 
 	gfx_v11_0_constants_init(adev);
@@ -4409,52 +4830,75 @@ static int gfx_v11_0_hw_init(void *handle)
 	if (r)
 		return r;
 
+	/* get IMU version from HW if it's not set */
+	if (!adev->gfx.imu_fw_version)
+		adev->gfx.imu_fw_version = RREG32_SOC15(GC, 0, regGFX_IMU_SCRATCH_0);
+
 	return r;
 }
 
-#ifndef BRING_UP_DEBUG
-static int gfx_v11_0_kiq_disable_kgq(struct amdgpu_device *adev)
+static int gfx_v11_0_set_userq_eop_interrupts(struct amdgpu_device *adev,
+					      bool enable)
 {
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-	struct amdgpu_ring *kiq_ring = &kiq->ring;
-	int i, r = 0;
-
-	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
-		return -EINVAL;
-
-	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size *
-					adev->gfx.num_gfx_rings))
-		return -ENOMEM;
-
-	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
-		kiq->pmf->kiq_unmap_queues(kiq_ring, &adev->gfx.gfx_ring[i],
-					   PREEMPT_QUEUES, 0, 0);
+	unsigned int irq_type;
+	int m, p, r;
+
+	if (adev->userq_funcs[AMDGPU_HW_IP_GFX]) {
+		for (m = 0; m < adev->gfx.me.num_me; m++) {
+			for (p = 0; p < adev->gfx.me.num_pipe_per_me; p++) {
+				irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + p;
+				if (enable)
+					r = amdgpu_irq_get(adev, &adev->gfx.eop_irq,
+							   irq_type);
+				else
+					r = amdgpu_irq_put(adev, &adev->gfx.eop_irq,
+							   irq_type);
+				if (r)
+					return r;
+			}
+		}
+	}
 
-	if (adev->gfx.kiq.ring.sched.ready)
-		r = amdgpu_ring_test_helper(kiq_ring);
+	if (adev->userq_funcs[AMDGPU_HW_IP_COMPUTE]) {
+		for (m = 0; m < adev->gfx.mec.num_mec; ++m) {
+			for (p = 0; p < adev->gfx.mec.num_pipe_per_mec; p++) {
+				irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
+					+ (m * adev->gfx.mec.num_pipe_per_mec)
+					+ p;
+				if (enable)
+					r = amdgpu_irq_get(adev, &adev->gfx.eop_irq,
+							   irq_type);
+				else
+					r = amdgpu_irq_put(adev, &adev->gfx.eop_irq,
+							   irq_type);
+				if (r)
+					return r;
+			}
+		}
+	}
 
-	return r;
+	return 0;
 }
-#endif
 
-static int gfx_v11_0_hw_fini(void *handle)
+static int gfx_v11_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	cancel_delayed_work_sync(&adev->gfx.idle_work);
 
-	amdgpu_irq_put(adev, &adev->gfx.cp_ecc_error_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
+	amdgpu_irq_put(adev, &adev->gfx.bad_op_irq, 0);
+	gfx_v11_0_set_userq_eop_interrupts(adev, false);
 
 	if (!adev->no_hw_access) {
-#ifndef BRING_UP_DEBUG
-		if (amdgpu_async_gfx_ring) {
-			r = gfx_v11_0_kiq_disable_kgq(adev);
-			if (r)
+		if (amdgpu_async_gfx_ring &&
+		    !adev->gfx.disable_kq) {
+			if (amdgpu_gfx_disable_kgq(adev, 0))
 				DRM_ERROR("KGQ disable failed\n");
 		}
-#endif
-		if (amdgpu_gfx_disable_kcq(adev))
+
+		if (amdgpu_gfx_disable_kcq(adev, 0))
 			DRM_ERROR("KCQ disable failed\n");
 
 		amdgpu_mes_kiq_hw_fini(adev);
@@ -4478,19 +4922,19 @@ static int gfx_v11_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int gfx_v11_0_suspend(void *handle)
+static int gfx_v11_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	return gfx_v11_0_hw_fini(handle);
+	return gfx_v11_0_hw_fini(ip_block);
 }
 
-static int gfx_v11_0_resume(void *handle)
+static int gfx_v11_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	return gfx_v11_0_hw_init(handle);
+	return gfx_v11_0_hw_init(ip_block);
 }
 
-static bool gfx_v11_0_is_idle(void *handle)
+static bool gfx_v11_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (REG_GET_FIELD(RREG32_SOC15(GC, 0, regGRBM_STATUS),
 				GRBM_STATUS, GUI_ACTIVE))
@@ -4499,11 +4943,11 @@ static bool gfx_v11_0_is_idle(void *handle)
 		return true;
 }
 
-static int gfx_v11_0_wait_for_idle(void *handle)
+static int gfx_v11_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		/* read MC_STATUS */
@@ -4517,12 +4961,46 @@ static int gfx_v11_0_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int gfx_v11_0_soft_reset(void *handle)
+int gfx_v11_0_request_gfx_index_mutex(struct amdgpu_device *adev,
+				      bool req)
+{
+	u32 i, tmp, val;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		/* Request with MeId=2, PipeId=0 */
+		tmp = REG_SET_FIELD(0, CP_GFX_INDEX_MUTEX, REQUEST, req);
+		tmp = REG_SET_FIELD(tmp, CP_GFX_INDEX_MUTEX, CLIENTID, 4);
+		WREG32_SOC15(GC, 0, regCP_GFX_INDEX_MUTEX, tmp);
+
+		val = RREG32_SOC15(GC, 0, regCP_GFX_INDEX_MUTEX);
+		if (req) {
+			if (val == tmp)
+				break;
+		} else {
+			tmp = REG_SET_FIELD(tmp, CP_GFX_INDEX_MUTEX,
+					    REQUEST, 1);
+
+			/* unlocked or locked by firmware */
+			if (val != tmp)
+				break;
+		}
+		udelay(1);
+	}
+
+	if (i >= adev->usec_timeout)
+		return -EINVAL;
+
+	return 0;
+}
+
+static int gfx_v11_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 grbm_soft_reset = 0;
 	u32 tmp;
-	int i, j, k;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	int r, i, j, k;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 
 	tmp = RREG32_SOC15(GC, 0, regCP_INT_CNTL);
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL, CMP_BUSY_INT_ENABLE, 0);
@@ -4531,16 +5009,11 @@ static int gfx_v11_0_soft_reset(void *handle)
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL, GFX_IDLE_INT_ENABLE, 0);
 	WREG32_SOC15(GC, 0, regCP_INT_CNTL, tmp);
 
-	gfx_v11_0_set_safe_mode(adev);
-
+	mutex_lock(&adev->srbm_mutex);
 	for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
 		for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
 			for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
-				tmp = RREG32_SOC15(GC, 0, regGRBM_GFX_CNTL);
-				tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, MEID, i);
-				tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, QUEUEID, j);
-				tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, PIPEID, k);
-				WREG32_SOC15(GC, 0, regGRBM_GFX_CNTL, tmp);
+				soc21_grbm_select(adev, i, k, j, 0);
 
 				WREG32_SOC15(GC, 0, regCP_HQD_DEQUEUE_REQUEST, 0x2);
 				WREG32_SOC15(GC, 0, regSPI_COMPUTE_QUEUE_RESET, 0x1);
@@ -4550,16 +5023,23 @@ static int gfx_v11_0_soft_reset(void *handle)
 	for (i = 0; i < adev->gfx.me.num_me; ++i) {
 		for (j = 0; j < adev->gfx.me.num_queue_per_pipe; j++) {
 			for (k = 0; k < adev->gfx.me.num_pipe_per_me; k++) {
-				tmp = RREG32_SOC15(GC, 0, regGRBM_GFX_CNTL);
-				tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, MEID, i);
-				tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, QUEUEID, j);
-				tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, PIPEID, k);
-				WREG32_SOC15(GC, 0, regGRBM_GFX_CNTL, tmp);
+				soc21_grbm_select(adev, i, k, j, 0);
 
 				WREG32_SOC15(GC, 0, regCP_GFX_HQD_DEQUEUE_REQUEST, 0x1);
 			}
 		}
 	}
+	soc21_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+
+	/* Try to acquire the gfx mutex before access to CP_VMID_RESET */
+	mutex_lock(&adev->gfx.reset_sem_mutex);
+	r = gfx_v11_0_request_gfx_index_mutex(adev, true);
+	if (r) {
+		mutex_unlock(&adev->gfx.reset_sem_mutex);
+		DRM_ERROR("Failed to acquire the gfx mutex during soft reset\n");
+		return r;
+	}
 
 	WREG32_SOC15(GC, 0, regCP_VMID_RESET, 0xfffffffe);
 
@@ -4569,6 +5049,14 @@ static int gfx_v11_0_soft_reset(void *handle)
 	RREG32_SOC15(GC, 0, regCP_VMID_RESET);
 	RREG32_SOC15(GC, 0, regCP_VMID_RESET);
 
+	/* release the gfx mutex */
+	r = gfx_v11_0_request_gfx_index_mutex(adev, false);
+	mutex_unlock(&adev->gfx.reset_sem_mutex);
+	if (r) {
+		DRM_ERROR("Failed to release the gfx mutex during soft reset\n");
+		return r;
+	}
+
 	for (i = 0; i < adev->usec_timeout; i++) {
 		if (!RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) &&
 		    !RREG32_SOC15(GC, 0, regCP_GFX_HQD_ACTIVE))
@@ -4631,15 +5119,15 @@ static int gfx_v11_0_soft_reset(void *handle)
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL, GFX_IDLE_INT_ENABLE, 1);
 	WREG32_SOC15(GC, 0, regCP_INT_CNTL, tmp);
 
-	gfx_v11_0_unset_safe_mode(adev);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 
 	return gfx_v11_0_cp_resume(adev);
 }
 
-static bool gfx_v11_0_check_soft_reset(void *handle)
+static bool gfx_v11_0_check_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	int i, r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	struct amdgpu_ring *ring;
 	long tmo = msecs_to_jiffies(1000);
 
@@ -4660,24 +5148,41 @@ static bool gfx_v11_0_check_soft_reset(void *handle)
 	return false;
 }
 
-static int gfx_v11_0_post_soft_reset(void *handle)
+static int gfx_v11_0_post_soft_reset(struct amdgpu_ip_block *ip_block)
 {
+	struct amdgpu_device *adev = ip_block->adev;
 	/**
 	 * GFX soft reset will impact MES, need resume MES when do GFX soft reset
 	 */
-	return amdgpu_mes_resume((struct amdgpu_device *)handle);
+	return amdgpu_mes_resume(adev);
 }
 
 static uint64_t gfx_v11_0_get_gpu_clock_counter(struct amdgpu_device *adev)
 {
 	uint64_t clock;
+	uint64_t clock_counter_lo, clock_counter_hi_pre, clock_counter_hi_after;
+
+	if (amdgpu_sriov_vf(adev)) {
+		amdgpu_gfx_off_ctrl(adev, false);
+		mutex_lock(&adev->gfx.gpu_clock_mutex);
+		clock_counter_hi_pre = (uint64_t)RREG32_SOC15(GC, 0, regCP_MES_MTIME_HI);
+		clock_counter_lo = (uint64_t)RREG32_SOC15(GC, 0, regCP_MES_MTIME_LO);
+		clock_counter_hi_after = (uint64_t)RREG32_SOC15(GC, 0, regCP_MES_MTIME_HI);
+		if (clock_counter_hi_pre != clock_counter_hi_after)
+			clock_counter_lo = (uint64_t)RREG32_SOC15(GC, 0, regCP_MES_MTIME_LO);
+		mutex_unlock(&adev->gfx.gpu_clock_mutex);
+		amdgpu_gfx_off_ctrl(adev, true);
+	} else {
+		preempt_disable();
+		clock_counter_hi_pre = (uint64_t)RREG32_SOC15(SMUIO, 0, regGOLDEN_TSC_COUNT_UPPER);
+		clock_counter_lo = (uint64_t)RREG32_SOC15(SMUIO, 0, regGOLDEN_TSC_COUNT_LOWER);
+		clock_counter_hi_after = (uint64_t)RREG32_SOC15(SMUIO, 0, regGOLDEN_TSC_COUNT_UPPER);
+		if (clock_counter_hi_pre != clock_counter_hi_after)
+			clock_counter_lo = (uint64_t)RREG32_SOC15(SMUIO, 0, regGOLDEN_TSC_COUNT_LOWER);
+		preempt_enable();
+	}
+	clock = clock_counter_lo | (clock_counter_hi_after << 32ULL);
 
-	amdgpu_gfx_off_ctrl(adev, false);
-	mutex_lock(&adev->gfx.gpu_clock_mutex);
-	clock = (uint64_t)RREG32_SOC15(SMUIO, 0, regGOLDEN_TSC_COUNT_LOWER) |
-		((uint64_t)RREG32_SOC15(SMUIO, 0, regGOLDEN_TSC_COUNT_UPPER) << 32ULL);
-	mutex_unlock(&adev->gfx.gpu_clock_mutex);
-	amdgpu_gfx_off_ctrl(adev, true);
 	return clock;
 }
 
@@ -4710,15 +5215,40 @@ static void gfx_v11_0_ring_emit_gds_switch(struct amdgpu_ring *ring,
 				    (1 << (oa_size + oa_base)) - (1 << oa_base));
 }
 
-static int gfx_v11_0_early_init(void *handle)
+static int gfx_v11_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	switch (amdgpu_user_queue) {
+	case -1:
+	case 0:
+	default:
+		adev->gfx.disable_kq = false;
+		adev->gfx.disable_uq = true;
+		break;
+	case 1:
+		adev->gfx.disable_kq = false;
+		adev->gfx.disable_uq = false;
+		break;
+	case 2:
+		adev->gfx.disable_kq = true;
+		adev->gfx.disable_uq = false;
+		break;
+	}
 
 	adev->gfx.funcs = &gfx_v11_0_gfx_funcs;
 
-	adev->gfx.num_gfx_rings = GFX11_NUM_GFX_RINGS;
-	adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
-					  AMDGPU_MAX_COMPUTE_RINGS);
+	if (adev->gfx.disable_kq) {
+		/* We need one GFX ring temporarily to set up
+		 * the clear state.
+		 */
+		adev->gfx.num_gfx_rings = 1;
+		adev->gfx.num_compute_rings = 0;
+	} else {
+		adev->gfx.num_gfx_rings = GFX11_NUM_GFX_RINGS;
+		adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
+						  AMDGPU_MAX_COMPUTE_RINGS);
+	}
 
 	gfx_v11_0_set_kiq_pm4_funcs(adev);
 	gfx_v11_0_set_ring_funcs(adev);
@@ -4733,29 +5263,9 @@ static int gfx_v11_0_early_init(void *handle)
 	return gfx_v11_0_init_microcode(adev);
 }
 
-static int gfx_v11_0_ras_late_init(void *handle)
-{
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-	struct ras_common_if *gfx_common_if;
-	int ret;
-
-	gfx_common_if = kzalloc(sizeof(struct ras_common_if), GFP_KERNEL);
-	if (!gfx_common_if)
-		return -ENOMEM;
-
-	gfx_common_if->block = AMDGPU_RAS_BLOCK__GFX;
-
-	ret = amdgpu_ras_feature_enable(adev, gfx_common_if, true);
-	if (ret)
-		dev_warn(adev->dev, "Failed to enable gfx11 ras feature\n");
-
-	kfree(gfx_common_if);
-	return 0;
-}
-
-static int gfx_v11_0_late_init(void *handle)
+static int gfx_v11_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0);
@@ -4766,11 +5276,13 @@ static int gfx_v11_0_late_init(void *handle)
 	if (r)
 		return r;
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(11, 0, 3)) {
-		r = gfx_v11_0_ras_late_init(handle);
-		if (r)
-			return r;
-	}
+	r = amdgpu_irq_get(adev, &adev->gfx.bad_op_irq, 0);
+	if (r)
+		return r;
+
+	r = gfx_v11_0_set_userq_eop_interrupts(adev, true);
+	if (r)
+		return r;
 
 	return 0;
 }
@@ -4784,7 +5296,7 @@ static bool gfx_v11_0_is_rlc_enabled(struct amdgpu_device *adev)
 	return (REG_GET_FIELD(rlc_cntl, RLC_CNTL, RLC_ENABLE_F32)) ? true : false;
 }
 
-static void gfx_v11_0_set_safe_mode(struct amdgpu_device *adev)
+static void gfx_v11_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	uint32_t data;
 	unsigned i;
@@ -4803,7 +5315,7 @@ static void gfx_v11_0_set_safe_mode(struct amdgpu_device *adev)
 	}
 }
 
-static void gfx_v11_0_unset_safe_mode(struct amdgpu_device *adev)
+static void gfx_v11_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	WREG32_SOC15(GC, 0, regRLC_SAFE_MODE, RLC_SAFE_MODE__CMD_MASK);
 }
@@ -5031,7 +5543,7 @@ static void gfx_v11_0_update_coarse_grain_clock_gating(struct amdgpu_device *ade
 static int gfx_v11_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 					    bool enable)
 {
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 
 	gfx_v11_0_update_coarse_grain_clock_gating(adev, enable);
 
@@ -5051,32 +5563,40 @@ static int gfx_v11_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 	     AMD_CG_SUPPORT_GFX_3D_CGLS))
 	        gfx_v11_0_enable_gui_idle_interrupt(adev, enable);
 
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 
 	return 0;
 }
 
-static void gfx_v11_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid)
+static void gfx_v11_0_update_spm_vmid(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned vmid)
 {
-	u32 reg, data;
+	u32 reg, pre_data, data;
 
 	amdgpu_gfx_off_ctrl(adev, false);
-
 	reg = SOC15_REG_OFFSET(GC, 0, regRLC_SPM_MC_CNTL);
-	if (amdgpu_sriov_is_pp_one_vf(adev))
-		data = RREG32_NO_KIQ(reg);
+	if (amdgpu_sriov_is_pp_one_vf(adev) && !amdgpu_sriov_runtime(adev))
+		pre_data = RREG32_NO_KIQ(reg);
 	else
-		data = RREG32(reg);
+		pre_data = RREG32(reg);
 
-	data &= ~RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK;
+	data = pre_data & (~RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK);
 	data |= (vmid & RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK) << RLC_SPM_MC_CNTL__RLC_SPM_VMID__SHIFT;
 
-	if (amdgpu_sriov_is_pp_one_vf(adev))
-		WREG32_SOC15_NO_KIQ(GC, 0, regRLC_SPM_MC_CNTL, data);
-	else
-		WREG32_SOC15(GC, 0, regRLC_SPM_MC_CNTL, data);
-
+	if (pre_data != data) {
+		if (amdgpu_sriov_is_pp_one_vf(adev) && !amdgpu_sriov_runtime(adev)) {
+			WREG32_SOC15_NO_KIQ(GC, 0, regRLC_SPM_MC_CNTL, data);
+		} else
+			WREG32_SOC15(GC, 0, regRLC_SPM_MC_CNTL, data);
+	}
 	amdgpu_gfx_off_ctrl(adev, true);
+
+	if (ring
+		&& amdgpu_sriov_is_pp_one_vf(adev)
+		&& (pre_data != data)
+		&& ((ring->funcs->type == AMDGPU_RING_TYPE_GFX)
+			|| (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE))) {
+		amdgpu_ring_emit_wreg(ring, reg, data);
+	}
 }
 
 static const struct amdgpu_rlc_funcs gfx_v11_0_rlc_funcs = {
@@ -5106,9 +5626,13 @@ static void gfx_v11_cntl_power_gating(struct amdgpu_device *adev, bool enable)
 
 	// Program RLC_PG_DELAY3 for CGPG hysteresis
 	if (enable && (adev->pg_flags & AMD_PG_SUPPORT_GFX_PG)) {
-		switch (adev->ip_versions[GC_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 		case IP_VERSION(11, 0, 1):
 		case IP_VERSION(11, 0, 4):
+		case IP_VERSION(11, 5, 0):
+		case IP_VERSION(11, 5, 1):
+		case IP_VERSION(11, 5, 2):
+		case IP_VERSION(11, 5, 3):
 			WREG32_SOC15(GC, 0, regRLC_PG_DELAY_3, RLC_PG_DELAY_3_DEFAULT_GC_11_0_1);
 			break;
 		default:
@@ -5119,23 +5643,23 @@ static void gfx_v11_cntl_power_gating(struct amdgpu_device *adev, bool enable)
 
 static void gfx_v11_cntl_pg(struct amdgpu_device *adev, bool enable)
 {
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 
 	gfx_v11_cntl_power_gating(adev, enable);
 
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 }
 
-static int gfx_v11_0_set_powergating_state(void *handle,
+static int gfx_v11_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_PG_STATE_GATE);
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
@@ -5143,8 +5667,18 @@ static int gfx_v11_0_set_powergating_state(void *handle,
 		break;
 	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
+		if (!enable)
+			amdgpu_gfx_off_ctrl(adev, false);
+
 		gfx_v11_cntl_pg(adev, enable);
-		amdgpu_gfx_off_ctrl(adev, enable);
+
+		if (enable)
+			amdgpu_gfx_off_ctrl(adev, true);
+
 		break;
 	default:
 		break;
@@ -5153,20 +5687,24 @@ static int gfx_v11_0_set_powergating_state(void *handle,
 	return 0;
 }
 
-static int gfx_v11_0_set_clockgating_state(void *handle,
+static int gfx_v11_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 	        return 0;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
 	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
 	case IP_VERSION(11, 0, 4):
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
 	        gfx_v11_0_update_gfx_clock_gating(adev,
 	                        state ==  AMD_CG_STATE_GATE);
 	        break;
@@ -5177,9 +5715,9 @@ static int gfx_v11_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static void gfx_v11_0_get_clockgating_state(void *handle, u64 *flags)
+static void gfx_v11_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	/* AMD_CG_SUPPORT_GFX_MGCG */
@@ -5243,45 +5781,17 @@ static u64 gfx_v11_0_ring_get_wptr_gfx(struct amdgpu_ring *ring)
 static void gfx_v11_0_ring_set_wptr_gfx(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t *wptr_saved;
-	uint32_t *is_queue_unmap;
-	uint64_t aggregated_db_index;
-	uint32_t mqd_size = adev->mqds[AMDGPU_HW_IP_GFX].mqd_size;
-	uint64_t wptr_tmp;
-
-	if (ring->is_mes_queue) {
-		wptr_saved = (uint32_t *)(ring->mqd_ptr + mqd_size);
-		is_queue_unmap = (uint32_t *)(ring->mqd_ptr + mqd_size +
-					      sizeof(uint32_t));
-		aggregated_db_index =
-			amdgpu_mes_get_aggregated_doorbell_index(adev,
-								 ring->hw_prio);
-
-		wptr_tmp = ring->wptr & ring->buf_mask;
-		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, wptr_tmp);
-		*wptr_saved = wptr_tmp;
-		/* assume doorbell always being used by mes mapped queue */
-		if (*is_queue_unmap) {
-			WDOORBELL64(aggregated_db_index, wptr_tmp);
-			WDOORBELL64(ring->doorbell_index, wptr_tmp);
-		} else {
-			WDOORBELL64(ring->doorbell_index, wptr_tmp);
 
-			if (*is_queue_unmap)
-				WDOORBELL64(aggregated_db_index, wptr_tmp);
-		}
+	if (ring->use_doorbell) {
+		/* XXX check if swapping is necessary on BE */
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
+			     ring->wptr);
+		WDOORBELL64(ring->doorbell_index, ring->wptr);
 	} else {
-		if (ring->use_doorbell) {
-			/* XXX check if swapping is necessary on BE */
-			atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
-				     ring->wptr);
-			WDOORBELL64(ring->doorbell_index, ring->wptr);
-		} else {
-			WREG32_SOC15(GC, 0, regCP_RB0_WPTR,
-				     lower_32_bits(ring->wptr));
-			WREG32_SOC15(GC, 0, regCP_RB0_WPTR_HI,
-				     upper_32_bits(ring->wptr));
-		}
+		WREG32_SOC15(GC, 0, regCP_RB0_WPTR,
+			     lower_32_bits(ring->wptr));
+		WREG32_SOC15(GC, 0, regCP_RB0_WPTR_HI,
+			     upper_32_bits(ring->wptr));
 	}
 }
 
@@ -5306,42 +5816,14 @@ static u64 gfx_v11_0_ring_get_wptr_compute(struct amdgpu_ring *ring)
 static void gfx_v11_0_ring_set_wptr_compute(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
-	uint32_t *wptr_saved;
-	uint32_t *is_queue_unmap;
-	uint64_t aggregated_db_index;
-	uint32_t mqd_size = adev->mqds[AMDGPU_HW_IP_COMPUTE].mqd_size;
-	uint64_t wptr_tmp;
-
-	if (ring->is_mes_queue) {
-		wptr_saved = (uint32_t *)(ring->mqd_ptr + mqd_size);
-		is_queue_unmap = (uint32_t *)(ring->mqd_ptr + mqd_size +
-					      sizeof(uint32_t));
-		aggregated_db_index =
-			amdgpu_mes_get_aggregated_doorbell_index(adev,
-								 ring->hw_prio);
-
-		wptr_tmp = ring->wptr & ring->buf_mask;
-		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, wptr_tmp);
-		*wptr_saved = wptr_tmp;
-		/* assume doorbell always used by mes mapped queue */
-		if (*is_queue_unmap) {
-			WDOORBELL64(aggregated_db_index, wptr_tmp);
-			WDOORBELL64(ring->doorbell_index, wptr_tmp);
-		} else {
-			WDOORBELL64(ring->doorbell_index, wptr_tmp);
 
-			if (*is_queue_unmap)
-				WDOORBELL64(aggregated_db_index, wptr_tmp);
-		}
+	/* XXX check if swapping is necessary on BE */
+	if (ring->use_doorbell) {
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
+			     ring->wptr);
+		WDOORBELL64(ring->doorbell_index, ring->wptr);
 	} else {
-		/* XXX check if swapping is necessary on BE */
-		if (ring->use_doorbell) {
-			atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
-				     ring->wptr);
-			WDOORBELL64(ring->doorbell_index, ring->wptr);
-		} else {
-			BUG(); /* only DOORBELL method supported on gfx11 now */
-		}
+		BUG(); /* only DOORBELL method supported on gfx11 now */
 	}
 }
 
@@ -5364,7 +5846,7 @@ static void gfx_v11_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
 		}
 		reg_mem_engine = 0;
 	} else {
-		ref_and_mask = nbio_hf_reg->ref_and_mask_cp0;
+		ref_and_mask = nbio_hf_reg->ref_and_mask_cp0 << ring->pipe;
 		reg_mem_engine = 1; /* pfp */
 	}
 
@@ -5382,27 +5864,21 @@ static void gfx_v11_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
 	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
 	u32 header, control = 0;
 
-	BUG_ON(ib->flags & AMDGPU_IB_FLAG_CE);
-
 	header = PACKET3(PACKET3_INDIRECT_BUFFER, 2);
 
 	control |= ib->length_dw | (vmid << 24);
 
-	if (amdgpu_mcbp && (ib->flags & AMDGPU_IB_FLAG_PREEMPT)) {
+	if (ring->adev->gfx.mcbp && (ib->flags & AMDGPU_IB_FLAG_PREEMPT)) {
 		control |= INDIRECT_BUFFER_PRE_ENB(1);
 
 		if (flags & AMDGPU_IB_PREEMPTED)
 			control |= INDIRECT_BUFFER_PRE_RESUME(1);
 
-		if (vmid)
+		if (vmid && !ring->adev->gfx.rs64_enable)
 			gfx_v11_0_ring_emit_de_meta(ring,
-				    (!amdgpu_sriov_vf(ring->adev) && flags & AMDGPU_IB_PREEMPTED) ? true : false);
+				!amdgpu_sriov_vf(ring->adev) && (flags & AMDGPU_IB_PREEMPTED));
 	}
 
-	if (ring->is_mes_queue)
-		/* inherit vmid from mqd */
-		control |= 0x400000;
-
 	amdgpu_ring_write(ring, header);
 	BUG_ON(ib->gpu_addr & 0x3); /* Dword align */
 	amdgpu_ring_write(ring,
@@ -5422,10 +5898,6 @@ static void gfx_v11_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
 	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
 	u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24);
 
-	if (ring->is_mes_queue)
-		/* inherit vmid from mqd */
-		control |= 0x40000000;
-
 	/* Currently, there is a high possibility to get wave ID mismatch
 	 * between ME and GDS, leading to a hw deadlock, because ME generates
 	 * different wave IDs than the GDS expects. This situation happens
@@ -5463,11 +5935,7 @@ static void gfx_v11_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
 	amdgpu_ring_write(ring, PACKET3(PACKET3_RELEASE_MEM, 6));
 	amdgpu_ring_write(ring, (PACKET3_RELEASE_MEM_GCR_SEQ |
 				 PACKET3_RELEASE_MEM_GCR_GL2_WB |
-				 PACKET3_RELEASE_MEM_GCR_GL2_INV |
-				 PACKET3_RELEASE_MEM_GCR_GL2_US |
-				 PACKET3_RELEASE_MEM_GCR_GL1_INV |
-				 PACKET3_RELEASE_MEM_GCR_GLV_INV |
-				 PACKET3_RELEASE_MEM_GCR_GLM_INV |
+				 PACKET3_RELEASE_MEM_GCR_GLM_INV | /* must be set with GLM_WB */
 				 PACKET3_RELEASE_MEM_GCR_GLM_WB |
 				 PACKET3_RELEASE_MEM_CACHE_POLICY(3) |
 				 PACKET3_RELEASE_MEM_EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) |
@@ -5487,8 +5955,7 @@ static void gfx_v11_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
 	amdgpu_ring_write(ring, upper_32_bits(addr));
 	amdgpu_ring_write(ring, lower_32_bits(seq));
 	amdgpu_ring_write(ring, upper_32_bits(seq));
-	amdgpu_ring_write(ring, ring->is_mes_queue ?
-			 (ring->hw_queue_id | AMDGPU_FENCE_MES_QUEUE_FLAG) : 0);
+	amdgpu_ring_write(ring, 0);
 }
 
 static void gfx_v11_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
@@ -5516,10 +5983,7 @@ static void gfx_v11_0_ring_invalidate_tlbs(struct amdgpu_ring *ring,
 static void gfx_v11_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 					 unsigned vmid, uint64_t pd_addr)
 {
-	if (ring->is_mes_queue)
-		gfx_v11_0_ring_invalidate_tlbs(ring, 0, 0, false, 0);
-	else
-		amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
+	amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
 
 	/* compute doesn't have PFP */
 	if (ring->funcs->type == AMDGPU_RING_TYPE_GFX) {
@@ -5527,6 +5991,12 @@ static void gfx_v11_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 		amdgpu_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0));
 		amdgpu_ring_write(ring, 0x0);
 	}
+
+	/* Make sure that we can't skip the SET_Q_MODE packets when the VM
+	 * changed in any way.
+	 */
+	ring->set_q_mode_offs = 0;
+	ring->set_q_mode_ptr = NULL;
 }
 
 static void gfx_v11_0_ring_emit_fence_kiq(struct amdgpu_ring *ring, u64 addr,
@@ -5576,41 +6046,125 @@ static void gfx_v11_0_ring_emit_cntxcntl(struct amdgpu_ring *ring,
 	amdgpu_ring_write(ring, 0);
 }
 
-static unsigned gfx_v11_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned gfx_v11_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring,
+						   uint64_t addr)
 {
 	unsigned ret;
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, 0); /* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	/* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, 0);
 	ret = ring->wptr & ring->buf_mask;
-	amdgpu_ring_write(ring, 0x55aa55aa); /* patch dummy value later */
+	/* patch dummy value later */
+	amdgpu_ring_write(ring, 0);
 
 	return ret;
 }
 
-static void gfx_v11_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigned offset)
+static void gfx_v11_0_ring_emit_gfx_shadow(struct amdgpu_ring *ring,
+					   u64 shadow_va, u64 csa_va,
+					   u64 gds_va, bool init_shadow,
+					   int vmid)
 {
-	unsigned cur;
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
+	struct amdgpu_device *adev = ring->adev;
+	unsigned int offs, end;
 
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (likely(cur > offset))
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->buf_mask + 1) - offset + cur;
+	if (!adev->gfx.cp_gfx_shadow || !ring->ring_obj)
+		return;
+
+	/*
+	 * The logic here isn't easy to understand because we need to keep state
+	 * accross multiple executions of the function as well as between the
+	 * CPU and GPU. The general idea is that the newly written GPU command
+	 * has a condition on the previous one and only executed if really
+	 * necessary.
+	 */
+
+	/*
+	 * The dw in the NOP controls if the next SET_Q_MODE packet should be
+	 * executed or not. Reserve 64bits just to be on the save side.
+	 */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_NOP, 1));
+	offs = ring->wptr & ring->buf_mask;
+
+	/*
+	 * We start with skipping the prefix SET_Q_MODE and always executing
+	 * the postfix SET_Q_MODE packet. This is changed below with a
+	 * WRITE_DATA command when the postfix executed.
+	 */
+	amdgpu_ring_write(ring, shadow_va ? 1 : 0);
+	amdgpu_ring_write(ring, 0);
+
+	if (ring->set_q_mode_offs) {
+		uint64_t addr;
+
+		addr = amdgpu_bo_gpu_offset(ring->ring_obj);
+		addr += ring->set_q_mode_offs << 2;
+		end = gfx_v11_0_ring_emit_init_cond_exec(ring, addr);
+	}
+
+	/*
+	 * When the postfix SET_Q_MODE packet executes we need to make sure that the
+	 * next prefix SET_Q_MODE packet executes as well.
+	 */
+	if (!shadow_va) {
+		uint64_t addr;
+
+		addr = amdgpu_bo_gpu_offset(ring->ring_obj);
+		addr += offs << 2;
+		amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+		amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM);
+		amdgpu_ring_write(ring, lower_32_bits(addr));
+		amdgpu_ring_write(ring, upper_32_bits(addr));
+		amdgpu_ring_write(ring, 0x1);
+	}
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_SET_Q_PREEMPTION_MODE, 7));
+	amdgpu_ring_write(ring, lower_32_bits(shadow_va));
+	amdgpu_ring_write(ring, upper_32_bits(shadow_va));
+	amdgpu_ring_write(ring, lower_32_bits(gds_va));
+	amdgpu_ring_write(ring, upper_32_bits(gds_va));
+	amdgpu_ring_write(ring, lower_32_bits(csa_va));
+	amdgpu_ring_write(ring, upper_32_bits(csa_va));
+	amdgpu_ring_write(ring, shadow_va ?
+			  PACKET3_SET_Q_PREEMPTION_MODE_IB_VMID(vmid) : 0);
+	amdgpu_ring_write(ring, init_shadow ?
+			  PACKET3_SET_Q_PREEMPTION_MODE_INIT_SHADOW_MEM : 0);
+
+	if (ring->set_q_mode_offs)
+		amdgpu_ring_patch_cond_exec(ring, end);
+
+	if (shadow_va) {
+		uint64_t token = shadow_va ^ csa_va ^ gds_va ^ vmid;
+
+		/*
+		 * If the tokens match try to skip the last postfix SET_Q_MODE
+		 * packet to avoid saving/restoring the state all the time.
+		 */
+		if (ring->set_q_mode_ptr && ring->set_q_mode_token == token)
+			*ring->set_q_mode_ptr = 0;
+
+		ring->set_q_mode_token = token;
+	} else {
+		ring->set_q_mode_ptr = &ring->ring[ring->set_q_mode_offs];
+	}
+
+	ring->set_q_mode_offs = offs;
 }
 
 static int gfx_v11_0_ring_preempt_ib(struct amdgpu_ring *ring)
 {
 	int i, r = 0;
 	struct amdgpu_device *adev = ring->adev;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
 	struct amdgpu_ring *kiq_ring = &kiq->ring;
 	unsigned long flags;
 
+	if (adev->enable_mes)
+		return -EINVAL;
+
 	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
 		return -EINVAL;
 
@@ -5658,28 +6212,13 @@ static void gfx_v11_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume)
 	void *de_payload_cpu_addr;
 	int cnt;
 
-	if (ring->is_mes_queue) {
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  gfx[0].gfx_meta_data) +
-			offsetof(struct v10_gfx_meta_data, de_payload);
-		de_payload_gpu_addr =
-			amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		de_payload_cpu_addr =
-			amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  gfx[0].gds_backup) +
-			offsetof(struct v10_gfx_meta_data, de_payload);
-		gds_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-	} else {
-		offset = offsetof(struct v10_gfx_meta_data, de_payload);
-		de_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
-		de_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
+	offset = offsetof(struct v10_gfx_meta_data, de_payload);
+	de_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
+	de_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
 
-		gds_addr = ALIGN(amdgpu_csa_vaddr(ring->adev) +
-				 AMDGPU_CSA_SIZE - adev->gds.gds_size,
-				 PAGE_SIZE);
-	}
+	gds_addr = ALIGN(amdgpu_csa_vaddr(ring->adev) +
+			 AMDGPU_CSA_SIZE - adev->gds.gds_size,
+			 PAGE_SIZE);
 
 	de_payload.gds_backup_addrlo = lower_32_bits(gds_addr);
 	de_payload.gds_backup_addrhi = upper_32_bits(gds_addr);
@@ -5766,19 +6305,6 @@ static void gfx_v11_0_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring,
 			       ref, mask, 0x20);
 }
 
-static void gfx_v11_0_ring_soft_recovery(struct amdgpu_ring *ring,
-					 unsigned vmid)
-{
-	struct amdgpu_device *adev = ring->adev;
-	uint32_t value = 0;
-
-	value = REG_SET_FIELD(value, SQ_CMD, CMD, 0x03);
-	value = REG_SET_FIELD(value, SQ_CMD, MODE, 0x01);
-	value = REG_SET_FIELD(value, SQ_CMD, CHECK_VMID, 1);
-	value = REG_SET_FIELD(value, SQ_CMD, VM_ID, vmid);
-	WREG32_SOC15(GC, 0, regSQ_CMD, value);
-}
-
 static void
 gfx_v11_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev,
 				      uint32_t me, uint32_t pipe,
@@ -5882,36 +6408,6 @@ static void gfx_v11_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev
 	}
 }
 
-#define CP_ME1_PIPE_INST_ADDR_INTERVAL  0x1
-#define SET_ECC_ME_PIPE_STATE(reg_addr, state) \
-	do { \
-		uint32_t tmp = RREG32_SOC15_IP(GC, reg_addr); \
-		tmp = REG_SET_FIELD(tmp, CP_ME1_PIPE0_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, state); \
-		WREG32_SOC15_IP(GC, reg_addr, tmp); \
-	} while (0)
-
-static int gfx_v11_0_set_cp_ecc_error_state(struct amdgpu_device *adev,
-							struct amdgpu_irq_src *source,
-							unsigned type,
-							enum amdgpu_interrupt_state state)
-{
-	uint32_t ecc_irq_state = 0;
-	uint32_t pipe0_int_cntl_addr = 0;
-	int i = 0;
-
-	ecc_irq_state = (state == AMDGPU_IRQ_STATE_ENABLE) ? 1 : 0;
-
-	pipe0_int_cntl_addr = SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE0_INT_CNTL);
-
-	WREG32_FIELD15_PREREG(GC, 0, CP_INT_CNTL_RING0, CP_ECC_ERROR_INT_ENABLE, ecc_irq_state);
-
-	for (i = 0; i < adev->gfx.mec.num_pipe_per_mec; i++)
-		SET_ECC_ME_PIPE_STATE(pipe0_int_cntl_addr + i * CP_ME1_PIPE_INST_ADDR_INTERVAL,
-					ecc_irq_state);
-
-	return 0;
-}
-
 static int gfx_v11_0_set_eop_interrupt_state(struct amdgpu_device *adev,
 					    struct amdgpu_irq_src *src,
 					    unsigned type,
@@ -5946,25 +6442,23 @@ static int gfx_v11_0_eop_irq(struct amdgpu_device *adev,
 			     struct amdgpu_irq_src *source,
 			     struct amdgpu_iv_entry *entry)
 {
-	int i;
+	u32 doorbell_offset = entry->src_data[0];
 	u8 me_id, pipe_id, queue_id;
 	struct amdgpu_ring *ring;
-	uint32_t mes_queue_id = entry->src_data[0];
+	int i;
 
 	DRM_DEBUG("IH: CP EOP\n");
 
-	if (adev->enable_mes && (mes_queue_id & AMDGPU_FENCE_MES_QUEUE_FLAG)) {
-		struct amdgpu_mes_queue *queue;
-
-		mes_queue_id &= AMDGPU_FENCE_MES_QUEUE_ID_MASK;
+	if (adev->enable_mes && doorbell_offset) {
+		struct amdgpu_userq_fence_driver *fence_drv = NULL;
+		struct xarray *xa = &adev->userq_xa;
+		unsigned long flags;
 
-		spin_lock(&adev->mes.queue_id_lock);
-		queue = idr_find(&adev->mes.queue_id_idr, mes_queue_id);
-		if (queue) {
-			DRM_DEBUG("process mes queue id = %d\n", mes_queue_id);
-			amdgpu_fence_process(queue->ring);
-		}
-		spin_unlock(&adev->mes.queue_id_lock);
+		xa_lock_irqsave(xa, flags);
+		fence_drv = xa_load(xa, doorbell_offset);
+		if (fence_drv)
+			amdgpu_userq_fence_driver_process(fence_drv);
+		xa_unlock_irqrestore(xa, flags);
 	} else {
 		me_id = (entry->ring_id & 0x0c) >> 2;
 		pipe_id = (entry->ring_id & 0x03) >> 0;
@@ -5999,15 +6493,42 @@ static int gfx_v11_0_eop_irq(struct amdgpu_device *adev,
 
 static int gfx_v11_0_set_priv_reg_fault_state(struct amdgpu_device *adev,
 					      struct amdgpu_irq_src *source,
-					      unsigned type,
+					      unsigned int type,
 					      enum amdgpu_interrupt_state state)
 {
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
 	case AMDGPU_IRQ_STATE_ENABLE:
-		WREG32_FIELD15_PREREG(GC, 0, CP_INT_CNTL_RING0,
-			       PRIV_REG_INT_ENABLE,
-			       state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v11_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    PRIV_REG_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				/* MECs start at 1 */
+				cp_int_cntl_reg = gfx_v11_0_get_cpc_int_cntl(adev, i + 1, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+								    PRIV_REG_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
 		break;
 	default:
 		break;
@@ -6016,17 +6537,75 @@ static int gfx_v11_0_set_priv_reg_fault_state(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int gfx_v11_0_set_bad_op_fault_state(struct amdgpu_device *adev,
+					    struct amdgpu_irq_src *source,
+					    unsigned type,
+					    enum amdgpu_interrupt_state state)
+{
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v11_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    OPCODE_ERROR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				/* MECs start at 1 */
+				cp_int_cntl_reg = gfx_v11_0_get_cpc_int_cntl(adev, i + 1, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+								    OPCODE_ERROR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
 static int gfx_v11_0_set_priv_inst_fault_state(struct amdgpu_device *adev,
 					       struct amdgpu_irq_src *source,
-					       unsigned type,
+					       unsigned int type,
 					       enum amdgpu_interrupt_state state)
 {
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
 	case AMDGPU_IRQ_STATE_ENABLE:
-		WREG32_FIELD15_PREREG(GC, 0, CP_INT_CNTL_RING0,
-			       PRIV_INSTR_INT_ENABLE,
-			       state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v11_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    PRIV_INSTR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
 		break;
 	default:
 		break;
@@ -6046,27 +6625,29 @@ static void gfx_v11_0_handle_priv_fault(struct amdgpu_device *adev,
 	pipe_id = (entry->ring_id & 0x03) >> 0;
 	queue_id = (entry->ring_id & 0x70) >> 4;
 
-	switch (me_id) {
-	case 0:
-		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
-			ring = &adev->gfx.gfx_ring[i];
-			/* we only enabled 1 gfx queue per pipe for now */
-			if (ring->me == me_id && ring->pipe == pipe_id)
-				drm_sched_fault(&ring->sched);
-		}
-		break;
-	case 1:
-	case 2:
-		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-			ring = &adev->gfx.compute_ring[i];
-			if (ring->me == me_id && ring->pipe == pipe_id &&
-			    ring->queue == queue_id)
-				drm_sched_fault(&ring->sched);
+	if (!adev->gfx.disable_kq) {
+		switch (me_id) {
+		case 0:
+			for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+				ring = &adev->gfx.gfx_ring[i];
+				if (ring->me == me_id && ring->pipe == pipe_id &&
+				    ring->queue == queue_id)
+					drm_sched_fault(&ring->sched);
+			}
+			break;
+		case 1:
+		case 2:
+			for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+				ring = &adev->gfx.compute_ring[i];
+				if (ring->me == me_id && ring->pipe == pipe_id &&
+				    ring->queue == queue_id)
+					drm_sched_fault(&ring->sched);
+			}
+			break;
+		default:
+			BUG();
+			break;
 		}
-		break;
-	default:
-		BUG();
-		break;
 	}
 }
 
@@ -6079,6 +6660,15 @@ static int gfx_v11_0_priv_reg_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int gfx_v11_0_bad_op_irq(struct amdgpu_device *adev,
+				struct amdgpu_irq_src *source,
+				struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal opcode in command stream \n");
+	gfx_v11_0_handle_priv_fault(adev, entry);
+	return 0;
+}
+
 static int gfx_v11_0_priv_inst_irq(struct amdgpu_device *adev,
 				   struct amdgpu_irq_src *source,
 				   struct amdgpu_iv_entry *entry)
@@ -6105,7 +6695,7 @@ static int gfx_v11_0_kiq_set_interrupt_state(struct amdgpu_device *adev,
 					     enum amdgpu_interrupt_state state)
 {
 	uint32_t tmp, target;
-	struct amdgpu_ring *ring = &(adev->gfx.kiq.ring);
+	struct amdgpu_ring *ring = &(adev->gfx.kiq[0].ring);
 
 	target = SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE0_INT_CNTL);
 	target += ring->pipe;
@@ -6165,6 +6755,426 @@ static void gfx_v11_0_emit_mem_sync(struct amdgpu_ring *ring)
 	amdgpu_ring_write(ring, gcr_cntl); /* GCR_CNTL */
 }
 
+static bool gfx_v11_pipe_reset_support(struct amdgpu_device *adev)
+{
+	/* Disable the pipe reset until the CPFW fully support it.*/
+	dev_warn_once(adev->dev, "The CPFW hasn't support pipe reset yet.\n");
+	return false;
+}
+
+
+static int gfx_v11_reset_gfx_pipe(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t reset_pipe = 0, clean_pipe = 0;
+	int r;
+
+	if (!gfx_v11_pipe_reset_support(adev))
+		return -EOPNOTSUPP;
+
+	gfx_v11_0_set_safe_mode(adev, 0);
+	mutex_lock(&adev->srbm_mutex);
+	soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+
+	switch (ring->pipe) {
+	case 0:
+		reset_pipe = REG_SET_FIELD(reset_pipe, CP_ME_CNTL,
+					   PFP_PIPE0_RESET, 1);
+		reset_pipe = REG_SET_FIELD(reset_pipe, CP_ME_CNTL,
+					   ME_PIPE0_RESET, 1);
+		clean_pipe = REG_SET_FIELD(clean_pipe, CP_ME_CNTL,
+					   PFP_PIPE0_RESET, 0);
+		clean_pipe = REG_SET_FIELD(clean_pipe, CP_ME_CNTL,
+					   ME_PIPE0_RESET, 0);
+		break;
+	case 1:
+		reset_pipe = REG_SET_FIELD(reset_pipe, CP_ME_CNTL,
+					   PFP_PIPE1_RESET, 1);
+		reset_pipe = REG_SET_FIELD(reset_pipe, CP_ME_CNTL,
+					   ME_PIPE1_RESET, 1);
+		clean_pipe = REG_SET_FIELD(clean_pipe, CP_ME_CNTL,
+					   PFP_PIPE1_RESET, 0);
+		clean_pipe = REG_SET_FIELD(clean_pipe, CP_ME_CNTL,
+					   ME_PIPE1_RESET, 0);
+		break;
+	default:
+		break;
+	}
+
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, reset_pipe);
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, clean_pipe);
+
+	r = (RREG32(SOC15_REG_OFFSET(GC, 0, regCP_GFX_RS64_INSTR_PNTR1)) << 2) -
+						RS64_FW_UC_START_ADDR_LO;
+	soc21_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	gfx_v11_0_unset_safe_mode(adev, 0);
+
+	dev_info(adev->dev, "The ring %s pipe reset to the ME firmware start PC: %s\n", ring->name,
+			r == 0 ? "successfully" : "failed");
+	/* FIXME: Sometimes driver can't cache the ME firmware start PC correctly,
+	 * so the pipe reset status relies on the later gfx ring test result.
+	 */
+	return 0;
+}
+
+static int gfx_v11_0_reset_kgq(struct amdgpu_ring *ring,
+			       unsigned int vmid,
+			       struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int r;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	r = amdgpu_mes_reset_legacy_queue(ring->adev, ring, vmid, false);
+	if (r) {
+
+		dev_warn(adev->dev, "reset via MES failed and try pipe reset %d\n", r);
+		r = gfx_v11_reset_gfx_pipe(ring);
+		if (r)
+			return r;
+	}
+
+	r = gfx_v11_0_kgq_init_queue(ring, true);
+	if (r) {
+		dev_err(adev->dev, "failed to init kgq\n");
+		return r;
+	}
+
+	r = amdgpu_mes_map_legacy_queue(adev, ring);
+	if (r) {
+		dev_err(adev->dev, "failed to remap kgq\n");
+		return r;
+	}
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static int gfx_v11_0_reset_compute_pipe(struct amdgpu_ring *ring)
+{
+
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t reset_pipe = 0, clean_pipe = 0;
+	int r;
+
+	if (!gfx_v11_pipe_reset_support(adev))
+		return -EOPNOTSUPP;
+
+	gfx_v11_0_set_safe_mode(adev, 0);
+	mutex_lock(&adev->srbm_mutex);
+	soc21_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+
+	reset_pipe = RREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL);
+	clean_pipe = reset_pipe;
+
+	if (adev->gfx.rs64_enable) {
+
+		switch (ring->pipe) {
+		case 0:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE0_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE0_RESET, 0);
+			break;
+		case 1:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE1_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE1_RESET, 0);
+			break;
+		case 2:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE2_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE2_RESET, 0);
+			break;
+		case 3:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE3_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE3_RESET, 0);
+			break;
+		default:
+			break;
+		}
+		WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, reset_pipe);
+		WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, clean_pipe);
+		r = (RREG32_SOC15(GC, 0, regCP_MEC_RS64_INSTR_PNTR) << 2) -
+					RS64_FW_UC_START_ADDR_LO;
+	} else {
+		if (ring->me == 1) {
+			switch (ring->pipe) {
+			case 0:
+				reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE0_RESET, 1);
+				clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE0_RESET, 0);
+				break;
+			case 1:
+				reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE1_RESET, 1);
+				clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE1_RESET, 0);
+				break;
+			case 2:
+				reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE2_RESET, 1);
+				clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE2_RESET, 0);
+				break;
+			case 3:
+				reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE3_RESET, 1);
+				clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE3_RESET, 0);
+				break;
+			default:
+				break;
+			}
+			/* mec1 fw pc: CP_MEC1_INSTR_PNTR */
+		} else {
+			switch (ring->pipe) {
+			case 0:
+				reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME2_PIPE0_RESET, 1);
+				clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME2_PIPE0_RESET, 0);
+				break;
+			case 1:
+				reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME2_PIPE1_RESET, 1);
+				clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME2_PIPE1_RESET, 0);
+				break;
+			case 2:
+				reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME2_PIPE2_RESET, 1);
+				clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME2_PIPE2_RESET, 0);
+				break;
+			case 3:
+				reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME2_PIPE3_RESET, 1);
+				clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME2_PIPE3_RESET, 0);
+				break;
+			default:
+				break;
+			}
+			/* mec2 fw pc: CP:CP_MEC2_INSTR_PNTR */
+		}
+		WREG32_SOC15(GC, 0, regCP_MEC_CNTL, reset_pipe);
+		WREG32_SOC15(GC, 0, regCP_MEC_CNTL, clean_pipe);
+		r = RREG32(SOC15_REG_OFFSET(GC, 0, regCP_MEC1_INSTR_PNTR));
+	}
+
+	soc21_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	gfx_v11_0_unset_safe_mode(adev, 0);
+
+	dev_info(adev->dev, "The ring %s pipe resets to MEC FW start PC: %s\n", ring->name,
+			r == 0 ? "successfully" : "failed");
+	/*FIXME:Sometimes driver can't cache the MEC firmware start PC correctly, so the pipe
+	 * reset status relies on the compute ring test result.
+	 */
+	return 0;
+}
+
+static int gfx_v11_0_reset_kcq(struct amdgpu_ring *ring,
+			       unsigned int vmid,
+			       struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int r = 0;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	r = amdgpu_mes_reset_legacy_queue(ring->adev, ring, vmid, true);
+	if (r) {
+		dev_warn(adev->dev, "fail(%d) to reset kcq and try pipe reset\n", r);
+		r = gfx_v11_0_reset_compute_pipe(ring);
+		if (r)
+			return r;
+	}
+
+	r = gfx_v11_0_kcq_init_queue(ring, true);
+	if (r) {
+		dev_err(adev->dev, "fail to init kcq\n");
+		return r;
+	}
+	r = amdgpu_mes_map_legacy_queue(adev, ring);
+	if (r) {
+		dev_err(adev->dev, "failed to remap kcq\n");
+		return r;
+	}
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static void gfx_v11_ip_print(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k, reg, index = 0;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_11_0);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	for (i = 0; i < reg_count; i++)
+		drm_printf(p, "%-50s \t 0x%08x\n",
+			   gc_reg_list_11_0[i].reg_name,
+			   adev->gfx.ip_dump_core[i]);
+
+	/* print compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_11);
+	drm_printf(p, "\nnum_mec: %d num_pipe: %d num_queue: %d\n",
+		   adev->gfx.mec.num_mec,
+		   adev->gfx.mec.num_pipe_per_mec,
+		   adev->gfx.mec.num_queue_per_pipe);
+
+	for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+		for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+			for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+				drm_printf(p, "\nmec %d, pipe %d, queue %d\n", i, j, k);
+				for (reg = 0; reg < reg_count; reg++) {
+					if (i && gc_cp_reg_list_11[reg].reg_offset == regCP_MEC_ME1_HEADER_DUMP)
+						drm_printf(p, "%-50s \t 0x%08x\n",
+							   "regCP_MEC_ME2_HEADER_DUMP",
+							   adev->gfx.ip_dump_compute_queues[index + reg]);
+					else
+						drm_printf(p, "%-50s \t 0x%08x\n",
+							   gc_cp_reg_list_11[reg].reg_name,
+							   adev->gfx.ip_dump_compute_queues[index + reg]);
+				}
+				index += reg_count;
+			}
+		}
+	}
+
+	/* print gfx queue registers for all instances */
+	if (!adev->gfx.ip_dump_gfx_queues)
+		return;
+
+	index = 0;
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_11);
+	drm_printf(p, "\nnum_me: %d num_pipe: %d num_queue: %d\n",
+		   adev->gfx.me.num_me,
+		   adev->gfx.me.num_pipe_per_me,
+		   adev->gfx.me.num_queue_per_pipe);
+
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			for (k = 0; k < adev->gfx.me.num_queue_per_pipe; k++) {
+				drm_printf(p, "\nme %d, pipe %d, queue %d\n", i, j, k);
+				for (reg = 0; reg < reg_count; reg++) {
+					drm_printf(p, "%-50s \t 0x%08x\n",
+						   gc_gfx_queue_reg_list_11[reg].reg_name,
+						   adev->gfx.ip_dump_gfx_queues[index + reg]);
+				}
+				index += reg_count;
+			}
+		}
+	}
+}
+
+static void gfx_v11_ip_dump(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k, reg, index = 0;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_11_0);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	amdgpu_gfx_off_ctrl(adev, false);
+	for (i = 0; i < reg_count; i++)
+		adev->gfx.ip_dump_core[i] = RREG32(SOC15_REG_ENTRY_OFFSET(gc_reg_list_11_0[i]));
+	amdgpu_gfx_off_ctrl(adev, true);
+
+	/* dump compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_11);
+	amdgpu_gfx_off_ctrl(adev, false);
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+		for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+			for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+				/* ME0 is for GFX so start from 1 for CP */
+				soc21_grbm_select(adev, adev->gfx.me.num_me + i, j, k, 0);
+				for (reg = 0; reg < reg_count; reg++) {
+					if (i &&
+					    gc_cp_reg_list_11[reg].reg_offset ==
+						    regCP_MEC_ME1_HEADER_DUMP)
+						adev->gfx.ip_dump_compute_queues[index + reg] =
+							RREG32(SOC15_REG_OFFSET(GC, 0,
+							       regCP_MEC_ME2_HEADER_DUMP));
+					else
+						adev->gfx.ip_dump_compute_queues[index + reg] =
+							RREG32(SOC15_REG_ENTRY_OFFSET(
+								       gc_cp_reg_list_11[reg]));
+				}
+				index += reg_count;
+			}
+		}
+	}
+	soc21_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_off_ctrl(adev, true);
+
+	/* dump gfx queue registers for all instances */
+	if (!adev->gfx.ip_dump_gfx_queues)
+		return;
+
+	index = 0;
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_11);
+	amdgpu_gfx_off_ctrl(adev, false);
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			for (k = 0; k < adev->gfx.me.num_queue_per_pipe; k++) {
+				soc21_grbm_select(adev, i, j, k, 0);
+
+				for (reg = 0; reg < reg_count; reg++) {
+					adev->gfx.ip_dump_gfx_queues[index + reg] =
+						RREG32(SOC15_REG_ENTRY_OFFSET(
+							gc_gfx_queue_reg_list_11[reg]));
+				}
+				index += reg_count;
+			}
+		}
+	}
+	soc21_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_off_ctrl(adev, true);
+}
+
+static void gfx_v11_0_ring_emit_cleaner_shader(struct amdgpu_ring *ring)
+{
+	/* Emit the cleaner shader */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_RUN_CLEANER_SHADER, 0));
+	amdgpu_ring_write(ring, 0);  /* RESERVED field, programmed to zero */
+}
+
+static void gfx_v11_0_ring_begin_use(struct amdgpu_ring *ring)
+{
+	amdgpu_gfx_profile_ring_begin_use(ring);
+
+	amdgpu_gfx_enforce_isolation_ring_begin_use(ring);
+}
+
+static void gfx_v11_0_ring_end_use(struct amdgpu_ring *ring)
+{
+	amdgpu_gfx_profile_ring_end_use(ring);
+
+	amdgpu_gfx_enforce_isolation_ring_end_use(ring);
+}
+
 static const struct amd_ip_funcs gfx_v11_0_ip_funcs = {
 	.name = "gfx_v11_0",
 	.early_init = gfx_v11_0_early_init,
@@ -6183,6 +7193,8 @@ static const struct amd_ip_funcs gfx_v11_0_ip_funcs = {
 	.set_clockgating_state = gfx_v11_0_set_clockgating_state,
 	.set_powergating_state = gfx_v11_0_set_powergating_state,
 	.get_clockgating_state = gfx_v11_0_get_clockgating_state,
+	.dump_ip_state = gfx_v11_ip_dump,
+	.print_ip_state = gfx_v11_ip_print,
 };
 
 static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_gfx = {
@@ -6191,16 +7203,17 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_gfx = {
 	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
 	.support_64bit_ptrs = true,
 	.secure_submission_supported = true,
-	.vmhub = AMDGPU_GFXHUB_0,
 	.get_rptr = gfx_v11_0_ring_get_rptr_gfx,
 	.get_wptr = gfx_v11_0_ring_get_wptr_gfx,
 	.set_wptr = gfx_v11_0_ring_set_wptr_gfx,
-	.emit_frame_size = /* totally 242 maximum if 16 IBs */
+	.emit_frame_size = /* totally 247 maximum if 16 IBs */
+		5 + /* update_spm_vmid */
 		5 + /* COND_EXEC */
+		22 + /* SET_Q_PREEMPTION_MODE */
 		7 + /* PIPELINE_SYNC */
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
-		2 + /* VM_FLUSH */
+		4 + /* VM_FLUSH */
 		8 + /* FENCE for VM_FLUSH */
 		20 + /* GDS switch */
 		5 + /* COND_EXEC */
@@ -6209,8 +7222,10 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_gfx = {
 		31 + /*	DE_META */
 		3 + /* CNTX_CTRL */
 		5 + /* HDP_INVL */
+		22 + /* SET_Q_PREEMPTION_MODE */
 		8 + 8 + /* FENCE x2 */
-		8, /* gfx_v11_0_emit_mem_sync */
+		8 + /* gfx_v11_0_emit_mem_sync */
+		2, /* gfx_v11_0_ring_emit_cleaner_shader */
 	.emit_ib_size =	4, /* gfx_v11_0_ring_emit_ib_gfx */
 	.emit_ib = gfx_v11_0_ring_emit_ib_gfx,
 	.emit_fence = gfx_v11_0_ring_emit_fence,
@@ -6220,18 +7235,21 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_gfx = {
 	.emit_hdp_flush = gfx_v11_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v11_0_ring_test_ring,
 	.test_ib = gfx_v11_0_ring_test_ib,
-	.insert_nop = amdgpu_ring_insert_nop,
+	.insert_nop = gfx_v11_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_cntxcntl = gfx_v11_0_ring_emit_cntxcntl,
+	.emit_gfx_shadow = gfx_v11_0_ring_emit_gfx_shadow,
 	.init_cond_exec = gfx_v11_0_ring_emit_init_cond_exec,
-	.patch_cond_exec = gfx_v11_0_ring_emit_patch_cond_exec,
 	.preempt_ib = gfx_v11_0_ring_preempt_ib,
 	.emit_frame_cntl = gfx_v11_0_ring_emit_frame_cntl,
 	.emit_wreg = gfx_v11_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v11_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v11_0_ring_emit_reg_write_reg_wait,
-	.soft_recovery = gfx_v11_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v11_0_emit_mem_sync,
+	.reset = gfx_v11_0_reset_kgq,
+	.emit_cleaner_shader = gfx_v11_0_ring_emit_cleaner_shader,
+	.begin_use = gfx_v11_0_ring_begin_use,
+	.end_use = gfx_v11_0_ring_end_use,
 };
 
 static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_compute = {
@@ -6239,11 +7257,11 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_compute = {
 	.align_mask = 0xff,
 	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
 	.support_64bit_ptrs = true,
-	.vmhub = AMDGPU_GFXHUB_0,
 	.get_rptr = gfx_v11_0_ring_get_rptr_compute,
 	.get_wptr = gfx_v11_0_ring_get_wptr_compute,
 	.set_wptr = gfx_v11_0_ring_set_wptr_compute,
 	.emit_frame_size =
+		5 + /* update_spm_vmid */
 		20 + /* gfx_v11_0_ring_emit_gds_switch */
 		7 + /* gfx_v11_0_ring_emit_hdp_flush */
 		5 + /* hdp invalidate */
@@ -6252,7 +7270,8 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_compute = {
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
 		2 + /* gfx_v11_0_ring_emit_vm_flush */
 		8 + 8 + 8 + /* gfx_v11_0_ring_emit_fence x3 for user fence, vm fence */
-		8, /* gfx_v11_0_emit_mem_sync */
+		8 + /* gfx_v11_0_emit_mem_sync */
+		2, /* gfx_v11_0_ring_emit_cleaner_shader */
 	.emit_ib_size =	7, /* gfx_v11_0_ring_emit_ib_compute */
 	.emit_ib = gfx_v11_0_ring_emit_ib_compute,
 	.emit_fence = gfx_v11_0_ring_emit_fence,
@@ -6262,12 +7281,16 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_compute = {
 	.emit_hdp_flush = gfx_v11_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v11_0_ring_test_ring,
 	.test_ib = gfx_v11_0_ring_test_ib,
-	.insert_nop = amdgpu_ring_insert_nop,
+	.insert_nop = gfx_v11_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_wreg = gfx_v11_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v11_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v11_0_ring_emit_reg_write_reg_wait,
 	.emit_mem_sync = gfx_v11_0_emit_mem_sync,
+	.reset = gfx_v11_0_reset_kcq,
+	.emit_cleaner_shader = gfx_v11_0_ring_emit_cleaner_shader,
+	.begin_use = gfx_v11_0_ring_begin_use,
+	.end_use = gfx_v11_0_ring_end_use,
 };
 
 static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_kiq = {
@@ -6275,7 +7298,6 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_kiq = {
 	.align_mask = 0xff,
 	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
 	.support_64bit_ptrs = true,
-	.vmhub = AMDGPU_GFXHUB_0,
 	.get_rptr = gfx_v11_0_ring_get_rptr_compute,
 	.get_wptr = gfx_v11_0_ring_get_wptr_compute,
 	.set_wptr = gfx_v11_0_ring_set_wptr_compute,
@@ -6286,7 +7308,6 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_kiq = {
 		7 + /* gfx_v11_0_ring_emit_pipeline_sync */
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
-		2 + /* gfx_v11_0_ring_emit_vm_flush */
 		8 + 8 + 8, /* gfx_v11_0_ring_emit_fence_kiq x3 for user fence, vm fence */
 	.emit_ib_size =	7, /* gfx_v11_0_ring_emit_ib_compute */
 	.emit_ib = gfx_v11_0_ring_emit_ib_compute,
@@ -6299,13 +7320,14 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_kiq = {
 	.emit_wreg = gfx_v11_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v11_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v11_0_ring_emit_reg_write_reg_wait,
+	.emit_hdp_flush = gfx_v11_0_ring_emit_hdp_flush,
 };
 
 static void gfx_v11_0_set_ring_funcs(struct amdgpu_device *adev)
 {
 	int i;
 
-	adev->gfx.kiq.ring.funcs = &gfx_v11_0_ring_funcs_kiq;
+	adev->gfx.kiq[0].ring.funcs = &gfx_v11_0_ring_funcs_kiq;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 		adev->gfx.gfx_ring[i].funcs = &gfx_v11_0_ring_funcs_gfx;
@@ -6324,16 +7346,16 @@ static const struct amdgpu_irq_src_funcs gfx_v11_0_priv_reg_irq_funcs = {
 	.process = gfx_v11_0_priv_reg_irq,
 };
 
+static const struct amdgpu_irq_src_funcs gfx_v11_0_bad_op_irq_funcs = {
+	.set = gfx_v11_0_set_bad_op_fault_state,
+	.process = gfx_v11_0_bad_op_irq,
+};
+
 static const struct amdgpu_irq_src_funcs gfx_v11_0_priv_inst_irq_funcs = {
 	.set = gfx_v11_0_set_priv_inst_fault_state,
 	.process = gfx_v11_0_priv_inst_irq,
 };
 
-static const struct amdgpu_irq_src_funcs gfx_v11_0_cp_ecc_error_irq_funcs = {
-	.set = gfx_v11_0_set_cp_ecc_error_state,
-	.process = amdgpu_gfx_cp_ecc_error_irq,
-};
-
 static const struct amdgpu_irq_src_funcs gfx_v11_0_rlc_gc_fed_irq_funcs = {
 	.process = gfx_v11_0_rlc_gc_fed_irq,
 };
@@ -6346,12 +7368,12 @@ static void gfx_v11_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->gfx.priv_reg_irq.num_types = 1;
 	adev->gfx.priv_reg_irq.funcs = &gfx_v11_0_priv_reg_irq_funcs;
 
+	adev->gfx.bad_op_irq.num_types = 1;
+	adev->gfx.bad_op_irq.funcs = &gfx_v11_0_bad_op_irq_funcs;
+
 	adev->gfx.priv_inst_irq.num_types = 1;
 	adev->gfx.priv_inst_irq.funcs = &gfx_v11_0_priv_inst_irq_funcs;
 
-	adev->gfx.cp_ecc_error_irq.num_types = 1; /* CP ECC error */
-	adev->gfx.cp_ecc_error_irq.funcs = &gfx_v11_0_cp_ecc_error_irq_funcs;
-
 	adev->gfx.rlc_gc_fed_irq.num_types = 1; /* 0x80 FED error */
 	adev->gfx.rlc_gc_fed_irq.funcs = &gfx_v11_0_rlc_gc_fed_irq_funcs;
 
@@ -6460,9 +7482,12 @@ static int gfx_v11_0_get_cu_info(struct amdgpu_device *adev,
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
+			bitmap = i * adev->gfx.config.max_sh_per_se + j;
+			if (!((gfx_v11_0_get_sa_active_bitmap(adev) >> bitmap) & 1))
+				continue;
 			mask = 1;
 			counter = 0;
-			gfx_v11_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v11_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			if (i < 8 && j < 2)
 				gfx_v11_0_set_user_wgp_inactive_bitmap_per_sh(
 					adev, disable_masks[i * 2 + j]);
@@ -6483,7 +7508,7 @@ static int gfx_v11_0_get_cu_info(struct amdgpu_device *adev,
 			 *    SE6: {SH0,SH1} --> {bitmap[2][2], bitmap[2][3]}
 			 *    SE7: {SH0,SH1} --> {bitmap[3][2], bitmap[3][3]}
 			 */
-			cu_info->bitmap[i % 4][j + (i / 4) * 2] = bitmap;
+			cu_info->bitmap[0][i % 4][j + (i / 4) * 2] = bitmap;
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
 				if (bitmap & mask)
@@ -6494,7 +7519,7 @@ static int gfx_v11_0_get_cu_info(struct amdgpu_device *adev,
 			active_cu_number += counter;
 		}
 	}
-	gfx_v11_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v11_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	cu_info->number = active_cu_number;
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.h b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.h
index 10cfc29c27c9..157a5c812259 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.h
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.h
@@ -26,4 +26,7 @@
 
 extern const struct amdgpu_ip_block_version gfx_v11_0_ip_block;
 
+int gfx_v11_0_request_gfx_index_mutex(struct amdgpu_device *adev,
+				      bool req);
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_3.c
index b07a72ca25d9..999bb3cc88b7 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_3.c
@@ -62,10 +62,18 @@ static int gfx_v11_0_3_rlc_gc_fed_irq(struct amdgpu_device *adev,
 		return -EINVAL;
 	}
 
-	ih_data.head = *ras_if;
-
 	dev_warn(adev->dev, "RLC %s FED IRQ\n", ras_if->name);
-	amdgpu_ras_interrupt_dispatch(adev, &ih_data);
+
+	if (!amdgpu_sriov_vf(adev)) {
+		ih_data.head = *ras_if;
+		amdgpu_ras_interrupt_dispatch(adev, &ih_data);
+	} else {
+		if (adev->virt.ops && adev->virt.ops->ras_poison_handler)
+			adev->virt.ops->ras_poison_handler(adev, ras_if->block);
+		else
+			dev_warn(adev->dev,
+				"No ras_poison_handler interface in SRIOV for %s!\n", ras_if->name);
+	}
 
 	return 0;
 }
@@ -76,8 +84,22 @@ static int gfx_v11_0_3_poison_consumption_handler(struct amdgpu_device *adev,
 	/* Workaround: when vmid and pasid are both zero, trigger gpu reset in KGD. */
 	if (entry && (entry->client_id == SOC21_IH_CLIENTID_GFX) &&
 	    (entry->src_id == GFX_11_0_0__SRCID__RLC_GC_FED_INTERRUPT) &&
-	     !entry->vmid && !entry->pasid)
-		amdgpu_ras_reset_gpu(adev);
+	     !entry->vmid && !entry->pasid) {
+		struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+		uint32_t rlc_status0 = 0;
+
+		rlc_status0 = RREG32_SOC15(GC, 0, regRLC_RLCS_FED_STATUS_0);
+
+		if (REG_GET_FIELD(rlc_status0, RLC_RLCS_FED_STATUS_0, SDMA0_FED_ERR) ||
+		    REG_GET_FIELD(rlc_status0, RLC_RLCS_FED_STATUS_0, SDMA1_FED_ERR)) {
+			struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
+
+			ras->gpu_reset_flags |= AMDGPU_RAS_GPU_RESET_MODE2_RESET;
+		}
+
+		if (con && !amdgpu_ras_is_rma(adev))
+			amdgpu_ras_reset_gpu(adev);
+	}
 
 	return 0;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_3_cleaner_shader.asm b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_3_cleaner_shader.asm
new file mode 100644
index 000000000000..9b90b66368c7
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_3_cleaner_shader.asm
@@ -0,0 +1,118 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+// This shader is to clean LDS, SGPRs and VGPRs. It is  first 64 Dwords or 256 bytes of 192 Dwords cleaner shader.
+//To turn this shader program on for complitaion change this to main and lower shader main to main_1
+
+// Navi3 : Clear SGPRs, VGPRs and LDS
+//   Launch 32 waves per CU (16 per SIMD) as a workgroup (threadgroup) to fill every wave slot
+//   Waves are "wave32" and have 64 VGPRs each, which uses all 1024 VGPRs per SIMD
+//   Waves are launched in "CU" mode, and the workgroup shares 64KB of LDS (half of the WGP's LDS)
+//      It takes 2 workgroups to use all of LDS: one on each CU of the WGP
+//   Each wave clears SGPRs 0 - 107
+//   Each wave clears VGPRs 0 - 63
+//   The first wave of the workgroup clears its 64KB of LDS
+//   The shader starts with "S_BARRIER" to ensure SPI has launched all waves of the workgroup
+//       before any wave in the workgroup could end.  Without this, it is possible not all SGPRs get cleared.
+
+shader main
+  asic(GFX11)
+  type(CS)
+  wave_size(32)
+// Note: original source code from SQ team
+
+// Takes about 2500 clocks to run.
+//   (theorhetical fastest = 1024clks vgpr + 640lds = 1660 clks)
+//
+  S_BARRIER
+
+  //
+  // CLEAR VGPRs
+  //
+  s_mov_b32     m0, 0x00000058  // Loop 96/8=12 times  (loop unrolled for performance)
+
+label_0005:
+  v_movreld_b32     v0, 0
+  v_movreld_b32     v1, 0
+  v_movreld_b32     v2, 0
+  v_movreld_b32     v3, 0
+  v_movreld_b32     v4, 0
+  v_movreld_b32     v5, 0
+  v_movreld_b32     v6, 0
+  v_movreld_b32     v7, 0
+  s_sub_u32     m0, m0, 8
+  s_cbranch_scc0  label_0005
+  //
+  //
+
+  s_mov_b32     s2, 0x80000000                      // Bit31 is first_wave
+  s_and_b32     s2, s2, s0                          // sgpr0 has tg_size (first_wave) term as in ucode only COMPUTE_PGM_RSRC2.tg_size_en is set
+  s_cbranch_scc0  label_0023                        // Clean LDS if its first wave of ThreadGroup/WorkGroup
+  // CLEAR LDS
+  //
+  s_mov_b32 exec_lo, 0xffffffff
+  s_mov_b32 exec_hi, 0xffffffff
+  v_mbcnt_lo_u32_b32  v1, exec_hi, 0          // Set V1 to thread-ID (0..63)
+  v_mbcnt_hi_u32_b32  v1, exec_lo, v1        // Set V1 to thread-ID (0..63)
+  v_mul_u32_u24  v1, 0x00000008, v1          // * 8, so each thread is a double-dword address (8byte)
+  s_mov_b32     s2, 0x00000003f                    // 64 loop iterations
+  s_mov_b32     m0, 0xffffffff
+  // Clear all of LDS space
+  // Each FirstWave of WorkGroup clears 64kbyte block
+
+label_001F:
+  ds_write2_b64  v1, v[2:3], v[2:3] offset1:32
+  ds_write2_b64  v1, v[4:5], v[4:5] offset0:64 offset1:96
+  v_add_co_u32     v1, vcc, 0x00000400, v1
+  s_sub_u32     s2, s2, 1
+  s_cbranch_scc0  label_001F
+  //
+  // CLEAR SGPRs
+  //
+label_0023:
+  s_mov_b32     m0, 0x00000068  // Loop 108/4=27 times  (loop unrolled for performance)
+label_sgpr_loop:
+  s_movreld_b32     s0, 0
+  s_movreld_b32     s1, 0
+  s_movreld_b32     s2, 0
+  s_movreld_b32     s3, 0
+  s_sub_u32         m0, m0, 4
+  s_cbranch_scc0  label_sgpr_loop
+
+  //clear vcc
+  s_mov_b64 vcc, 0          //clear vcc
+  s_mov_b32 flat_scratch_lo, 0   //clear  flat scratch lo SGPR
+  s_mov_b32 flat_scratch_hi, 0   //clear  flat scratch hi SGPR
+  s_mov_b64 ttmp0, 0        //Clear ttmp0 and ttmp1
+  s_mov_b64 ttmp2, 0        //Clear ttmp2 and ttmp3
+  s_mov_b64 ttmp4, 0        //Clear ttmp4 and ttmp5
+  s_mov_b64 ttmp6, 0        //Clear ttmp6 and ttmp7
+  s_mov_b64 ttmp8, 0        //Clear ttmp8 and ttmp9
+  s_mov_b64 ttmp10, 0       //Clear ttmp10 and ttmp11
+  s_mov_b64 ttmp12, 0       //Clear ttmp12 and ttmp13
+  s_mov_b64 ttmp14, 0       //Clear ttmp14 and ttmp15
+
+ s_endpgm
+
+end
+
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_cleaner_shader.h b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_cleaner_shader.h
new file mode 100644
index 000000000000..3218cc04f543
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0_cleaner_shader.h
@@ -0,0 +1,56 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/* Define the cleaner shader gfx_11_0_3 */
+static const u32 gfx_11_0_3_cleaner_shader_hex[] = {
+	0xb0804006, 0xbe8200ff,
+	0x00000058, 0xbefd0080,
+	0x7e008480, 0x7e028480,
+	0x7e048480, 0x7e068480,
+	0x7e088480, 0x7e0a8480,
+	0x7e0c8480, 0x7e0e8480,
+	0xbefd0002, 0x80828802,
+	0xbfa1fff5, 0xbe8200ff,
+	0x80000000, 0x8b020002,
+	0xbfa10012, 0xbefe00c1,
+	0xbeff00c1, 0xd71f0001,
+	0x0001007f, 0xd7200001,
+	0x0002027e, 0x16020288,
+	0xbe8200bf, 0xbefd00c1,
+	0xd9382000, 0x00020201,
+	0xd9386040, 0x00040401,
+	0xd7006a01, 0x000202ff,
+	0x00000400, 0x80828102,
+	0xbfa1fff7, 0xbefd00ff,
+	0x00000068, 0xbe804280,
+	0xbe814280, 0xbe824280,
+	0xbe834280, 0x80fd847d,
+	0xbfa1fffa, 0xbeea0180,
+	0xbeec0180, 0xbeee0180,
+	0xbef00180, 0xbef20180,
+	0xbef40180, 0xbef60180,
+	0xbef80180, 0xbefa0180,
+	0xbfb00000, 0xbf9f0000,
+	0xbf9f0000, 0xbf9f0000,
+	0xbf9f0000, 0xbf9f0000,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c
new file mode 100644
index 000000000000..d01d2712cf57
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c
@@ -0,0 +1,5793 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include <linux/delay.h>
+#include <linux/kernel.h>
+#include <linux/firmware.h>
+#include <linux/module.h>
+#include <linux/pci.h>
+#include "amdgpu.h"
+#include "amdgpu_gfx.h"
+#include "amdgpu_psp.h"
+#include "amdgpu_smu.h"
+#include "imu_v12_0.h"
+#include "soc24.h"
+#include "nvd.h"
+
+#include "gc/gc_12_0_0_offset.h"
+#include "gc/gc_12_0_0_sh_mask.h"
+#include "soc24_enum.h"
+#include "ivsrcid/gfx/irqsrcs_gfx_12_0_0.h"
+
+#include "soc15.h"
+#include "clearstate_gfx12.h"
+#include "v12_structs.h"
+#include "gfx_v12_0.h"
+#include "nbif_v6_3_1.h"
+#include "mes_v12_0.h"
+#include "mes_userqueue.h"
+#include "amdgpu_userq_fence.h"
+
+#define GFX12_NUM_GFX_RINGS	1
+#define GFX12_MEC_HPD_SIZE	2048
+
+#define RLCG_UCODE_LOADING_START_ADDRESS	0x00002000L
+
+#define regCP_GFX_MQD_CONTROL_DEFAULT                                             0x00000100
+#define regCP_GFX_HQD_VMID_DEFAULT                                                0x00000000
+#define regCP_GFX_HQD_QUEUE_PRIORITY_DEFAULT                                      0x00000000
+#define regCP_GFX_HQD_QUANTUM_DEFAULT                                             0x00000a01
+#define regCP_GFX_HQD_CNTL_DEFAULT                                                0x00f00000
+#define regCP_RB_DOORBELL_CONTROL_DEFAULT                                         0x00000000
+#define regCP_GFX_HQD_RPTR_DEFAULT                                                0x00000000
+
+#define regCP_HQD_EOP_CONTROL_DEFAULT                                             0x00000006
+#define regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT                                     0x00000000
+#define regCP_MQD_CONTROL_DEFAULT                                                 0x00000100
+#define regCP_HQD_PQ_CONTROL_DEFAULT                                              0x00308509
+#define regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT                                     0x00000000
+#define regCP_HQD_PQ_RPTR_DEFAULT                                                 0x00000000
+#define regCP_HQD_PERSISTENT_STATE_DEFAULT                                        0x0be05501
+#define regCP_HQD_IB_CONTROL_DEFAULT                                              0x00300000
+
+
+MODULE_FIRMWARE("amdgpu/gc_12_0_0_pfp.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_0_me.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_0_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_0_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_0_toc.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_pfp.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_me.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_rlc_kicker.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_toc.bin");
+
+static const struct amdgpu_hwip_reg_entry gc_reg_list_12_0[] = {
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS2),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS3),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT3),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_BUSY_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_BUSY_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HPD_STATUS0),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB0_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB0_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB0_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB2_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB2_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB2_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB2_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPF_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPG_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regIA_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regIA_UTCL1_STATUS_2),
+	SOC15_REG_ENTRY_STR(GC, 0, regPA_CL_CNTL_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regRMI_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSQC_CACHES),
+	SOC15_REG_ENTRY_STR(GC, 0, regSQG_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regWD_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regGCVM_L2_PROTECTION_FAULT_STATUS_LO32),
+	SOC15_REG_ENTRY_STR(GC, 0, regGCVM_L2_PROTECTION_FAULT_STATUS_HI32),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_DEBUG),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_RS64_INSTR_PNTR0),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_RS64_INSTR_PNTR1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_RS64_INSTR_PNTR),
+	/* cp header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MES_HEADER_DUMP),
+	/* SE status registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE0),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE1),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE2),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE3)
+};
+
+static const struct amdgpu_hwip_reg_entry gc_cp_reg_list_12[] = {
+	/* compute registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_VMID),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PERSISTENT_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PIPE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_QUEUE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_QUANTUM),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_DEQUEUE_REQUEST),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_EVENTS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_BASE_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CNTL_STACK_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CNTL_STACK_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_WG_STATE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_GDS_RESOURCE_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_WPTR_MEM),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_SUSPEND_CNTL_STACK_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_SUSPEND_CNTL_STACK_DW_CNT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_SUSPEND_WG_STATE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_DEQUEUE_STATUS),
+	/* cp header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+};
+
+static const struct amdgpu_hwip_reg_entry gc_gfx_queue_reg_list_12[] = {
+	/* gfx queue registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_ACTIVE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_VMID),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_QUEUE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_QUANTUM),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_CSMD_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_DEQUEUE_REQUEST),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_MAPPED),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_QUE_MGR_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_HQ_CONTROL0),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_HQD_HQ_STATUS0),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_MQD_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_MQD_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_WPTR_POLL_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_WPTR_POLL_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BUFSZ),
+	/* cp header registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_ME_HEADER_DUMP),
+};
+
+static const struct soc15_reg_golden golden_settings_gc_12_0_rev0[] = {
+	SOC15_REG_GOLDEN_VALUE(GC, 0, regDB_MEM_CONFIG, 0x0000000f, 0x0000000f),
+	SOC15_REG_GOLDEN_VALUE(GC, 0, regCB_HW_CONTROL_1, 0x03000000, 0x03000000),
+	SOC15_REG_GOLDEN_VALUE(GC, 0, regGL2C_CTRL5, 0x00000070, 0x00000020)
+};
+
+static const struct soc15_reg_golden golden_settings_gc_12_0[] = {
+	SOC15_REG_GOLDEN_VALUE(GC, 0, regDB_MEM_CONFIG, 0x00008000, 0x00008000),
+};
+
+#define DEFAULT_SH_MEM_CONFIG \
+	((SH_MEM_ADDRESS_MODE_64 << SH_MEM_CONFIG__ADDRESS_MODE__SHIFT) | \
+	 (SH_MEM_ALIGNMENT_MODE_UNALIGNED << SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT) | \
+	 (3 << SH_MEM_CONFIG__INITIAL_INST_PREFETCH__SHIFT))
+
+static void gfx_v12_0_disable_gpa_mode(struct amdgpu_device *adev);
+static void gfx_v12_0_set_ring_funcs(struct amdgpu_device *adev);
+static void gfx_v12_0_set_irq_funcs(struct amdgpu_device *adev);
+static void gfx_v12_0_set_rlc_funcs(struct amdgpu_device *adev);
+static void gfx_v12_0_set_mqd_funcs(struct amdgpu_device *adev);
+static void gfx_v12_0_set_imu_funcs(struct amdgpu_device *adev);
+static int gfx_v12_0_get_cu_info(struct amdgpu_device *adev,
+				 struct amdgpu_cu_info *cu_info);
+static uint64_t gfx_v12_0_get_gpu_clock_counter(struct amdgpu_device *adev);
+static void gfx_v12_0_select_se_sh(struct amdgpu_device *adev, u32 se_num,
+				   u32 sh_num, u32 instance, int xcc_id);
+static u32 gfx_v12_0_get_wgp_active_bitmap_per_sh(struct amdgpu_device *adev);
+
+static void gfx_v12_0_ring_emit_frame_cntl(struct amdgpu_ring *ring, bool start, bool secure);
+static void gfx_v12_0_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg,
+				     uint32_t val);
+static int gfx_v12_0_wait_for_rlc_autoload_complete(struct amdgpu_device *adev);
+static void gfx_v12_0_ring_invalidate_tlbs(struct amdgpu_ring *ring,
+					   uint16_t pasid, uint32_t flush_type,
+					   bool all_hub, uint8_t dst_sel);
+static void gfx_v12_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id);
+static void gfx_v12_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id);
+static void gfx_v12_0_update_perf_clk(struct amdgpu_device *adev,
+				      bool enable);
+
+static void gfx_v12_0_kiq_set_resources(struct amdgpu_ring *kiq_ring,
+					uint64_t queue_mask)
+{
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_RESOURCES, 6));
+	amdgpu_ring_write(kiq_ring, PACKET3_SET_RESOURCES_VMID_MASK(0) |
+			  PACKET3_SET_RESOURCES_QUEUE_TYPE(0));	/* vmid_mask:0 queue_type:0 (KIQ) */
+	amdgpu_ring_write(kiq_ring, lower_32_bits(queue_mask));	/* queue mask lo */
+	amdgpu_ring_write(kiq_ring, upper_32_bits(queue_mask));	/* queue mask hi */
+	amdgpu_ring_write(kiq_ring, 0);	/* gws mask lo */
+	amdgpu_ring_write(kiq_ring, 0);	/* gws mask hi */
+	amdgpu_ring_write(kiq_ring, 0);	/* oac mask */
+	amdgpu_ring_write(kiq_ring, 0);
+}
+
+static void gfx_v12_0_kiq_map_queues(struct amdgpu_ring *kiq_ring,
+				     struct amdgpu_ring *ring)
+{
+	uint64_t mqd_addr = amdgpu_bo_gpu_offset(ring->mqd_obj);
+	uint64_t wptr_addr = ring->wptr_gpu_addr;
+	uint32_t me = 0, eng_sel = 0;
+
+	switch (ring->funcs->type) {
+	case AMDGPU_RING_TYPE_COMPUTE:
+		me = 1;
+		eng_sel = 0;
+		break;
+	case AMDGPU_RING_TYPE_GFX:
+		me = 0;
+		eng_sel = 4;
+		break;
+	case AMDGPU_RING_TYPE_MES:
+		me = 2;
+		eng_sel = 5;
+		break;
+	default:
+		WARN_ON(1);
+	}
+
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_MAP_QUEUES, 5));
+	/* Q_sel:0, vmid:0, vidmem: 1, engine:0, num_Q:1*/
+	amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
+			  PACKET3_MAP_QUEUES_QUEUE_SEL(0) | /* Queue_Sel */
+			  PACKET3_MAP_QUEUES_VMID(0) | /* VMID */
+			  PACKET3_MAP_QUEUES_QUEUE(ring->queue) |
+			  PACKET3_MAP_QUEUES_PIPE(ring->pipe) |
+			  PACKET3_MAP_QUEUES_ME((me)) |
+			  PACKET3_MAP_QUEUES_QUEUE_TYPE(0) | /*queue_type: normal compute queue */
+			  PACKET3_MAP_QUEUES_ALLOC_FORMAT(0) | /* alloc format: all_on_one_pipe */
+			  PACKET3_MAP_QUEUES_ENGINE_SEL(eng_sel) |
+			  PACKET3_MAP_QUEUES_NUM_QUEUES(1)); /* num_queues: must be 1 */
+	amdgpu_ring_write(kiq_ring, PACKET3_MAP_QUEUES_DOORBELL_OFFSET(ring->doorbell_index));
+	amdgpu_ring_write(kiq_ring, lower_32_bits(mqd_addr));
+	amdgpu_ring_write(kiq_ring, upper_32_bits(mqd_addr));
+	amdgpu_ring_write(kiq_ring, lower_32_bits(wptr_addr));
+	amdgpu_ring_write(kiq_ring, upper_32_bits(wptr_addr));
+}
+
+static void gfx_v12_0_kiq_unmap_queues(struct amdgpu_ring *kiq_ring,
+				       struct amdgpu_ring *ring,
+				       enum amdgpu_unmap_queues_action action,
+				       u64 gpu_addr, u64 seq)
+{
+	struct amdgpu_device *adev = kiq_ring->adev;
+	uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0;
+
+	if (adev->enable_mes && !adev->gfx.kiq[0].ring.sched.ready) {
+		amdgpu_mes_unmap_legacy_queue(adev, ring, action, gpu_addr, seq);
+		return;
+	}
+
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4));
+	amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
+			  PACKET3_UNMAP_QUEUES_ACTION(action) |
+			  PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) |
+			  PACKET3_UNMAP_QUEUES_ENGINE_SEL(eng_sel) |
+			  PACKET3_UNMAP_QUEUES_NUM_QUEUES(1));
+	amdgpu_ring_write(kiq_ring,
+		  PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index));
+
+	if (action == PREEMPT_QUEUES_NO_UNMAP) {
+		amdgpu_ring_write(kiq_ring, lower_32_bits(gpu_addr));
+		amdgpu_ring_write(kiq_ring, upper_32_bits(gpu_addr));
+		amdgpu_ring_write(kiq_ring, seq);
+	} else {
+		amdgpu_ring_write(kiq_ring, 0);
+		amdgpu_ring_write(kiq_ring, 0);
+		amdgpu_ring_write(kiq_ring, 0);
+	}
+}
+
+static void gfx_v12_0_kiq_query_status(struct amdgpu_ring *kiq_ring,
+				       struct amdgpu_ring *ring,
+				       u64 addr, u64 seq)
+{
+	uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0;
+
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_QUERY_STATUS, 5));
+	amdgpu_ring_write(kiq_ring,
+			  PACKET3_QUERY_STATUS_CONTEXT_ID(0) |
+			  PACKET3_QUERY_STATUS_INTERRUPT_SEL(0) |
+			  PACKET3_QUERY_STATUS_COMMAND(2));
+	amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
+			  PACKET3_QUERY_STATUS_DOORBELL_OFFSET(ring->doorbell_index) |
+			  PACKET3_QUERY_STATUS_ENG_SEL(eng_sel));
+	amdgpu_ring_write(kiq_ring, lower_32_bits(addr));
+	amdgpu_ring_write(kiq_ring, upper_32_bits(addr));
+	amdgpu_ring_write(kiq_ring, lower_32_bits(seq));
+	amdgpu_ring_write(kiq_ring, upper_32_bits(seq));
+}
+
+static void gfx_v12_0_kiq_invalidate_tlbs(struct amdgpu_ring *kiq_ring,
+					  uint16_t pasid,
+					  uint32_t flush_type,
+					  bool all_hub)
+{
+	gfx_v12_0_ring_invalidate_tlbs(kiq_ring, pasid, flush_type, all_hub, 1);
+}
+
+static const struct kiq_pm4_funcs gfx_v12_0_kiq_pm4_funcs = {
+	.kiq_set_resources = gfx_v12_0_kiq_set_resources,
+	.kiq_map_queues = gfx_v12_0_kiq_map_queues,
+	.kiq_unmap_queues = gfx_v12_0_kiq_unmap_queues,
+	.kiq_query_status = gfx_v12_0_kiq_query_status,
+	.kiq_invalidate_tlbs = gfx_v12_0_kiq_invalidate_tlbs,
+	.set_resources_size = 8,
+	.map_queues_size = 7,
+	.unmap_queues_size = 6,
+	.query_status_size = 7,
+	.invalidate_tlbs_size = 2,
+};
+
+static void gfx_v12_0_set_kiq_pm4_funcs(struct amdgpu_device *adev)
+{
+	adev->gfx.kiq[0].pmf = &gfx_v12_0_kiq_pm4_funcs;
+}
+
+static void gfx_v12_0_wait_reg_mem(struct amdgpu_ring *ring, int eng_sel,
+				   int mem_space, int opt, uint32_t addr0,
+				   uint32_t addr1, uint32_t ref,
+				   uint32_t mask, uint32_t inv)
+{
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5));
+	amdgpu_ring_write(ring,
+			  /* memory (1) or register (0) */
+			  (WAIT_REG_MEM_MEM_SPACE(mem_space) |
+			   WAIT_REG_MEM_OPERATION(opt) | /* wait */
+			   WAIT_REG_MEM_FUNCTION(3) |  /* equal */
+			   WAIT_REG_MEM_ENGINE(eng_sel)));
+
+	if (mem_space)
+		BUG_ON(addr0 & 0x3); /* Dword align */
+	amdgpu_ring_write(ring, addr0);
+	amdgpu_ring_write(ring, addr1);
+	amdgpu_ring_write(ring, ref);
+	amdgpu_ring_write(ring, mask);
+	amdgpu_ring_write(ring, inv); /* poll interval */
+}
+
+static int gfx_v12_0_ring_test_ring(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t scratch = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG0);
+	uint32_t tmp = 0;
+	unsigned i;
+	int r;
+
+	WREG32(scratch, 0xCAFEDEAD);
+	r = amdgpu_ring_alloc(ring, 5);
+	if (r) {
+		dev_err(adev->dev,
+			"amdgpu: cp failed to lock ring %d (%d).\n",
+			ring->idx, r);
+		return r;
+	}
+
+	if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ) {
+		gfx_v12_0_ring_emit_wreg(ring, scratch, 0xDEADBEEF);
+	} else {
+		amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
+		amdgpu_ring_write(ring, scratch -
+				  PACKET3_SET_UCONFIG_REG_START);
+		amdgpu_ring_write(ring, 0xDEADBEEF);
+	}
+	amdgpu_ring_commit(ring);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32(scratch);
+		if (tmp == 0xDEADBEEF)
+			break;
+		if (amdgpu_emu_mode == 1)
+			msleep(1);
+		else
+			udelay(1);
+	}
+
+	if (i >= adev->usec_timeout)
+		r = -ETIMEDOUT;
+	return r;
+}
+
+static int gfx_v12_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_ib ib;
+	struct dma_fence *f = NULL;
+	unsigned index;
+	uint64_t gpu_addr;
+	uint32_t *cpu_ptr;
+	long r;
+
+	/* MES KIQ fw hasn't indirect buffer support for now */
+	if (adev->enable_mes_kiq &&
+	    ring->funcs->type == AMDGPU_RING_TYPE_KIQ)
+		return 0;
+
+	memset(&ib, 0, sizeof(ib));
+
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r)
+		return r;
+
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
+	cpu_ptr = &adev->wb.wb[index];
+
+	r = amdgpu_ib_get(adev, NULL, 16, AMDGPU_IB_POOL_DIRECT, &ib);
+	if (r) {
+		dev_err(adev->dev, "amdgpu: failed to get ib (%ld).\n", r);
+		goto err1;
+	}
+
+	ib.ptr[0] = PACKET3(PACKET3_WRITE_DATA, 3);
+	ib.ptr[1] = WRITE_DATA_DST_SEL(5) | WR_CONFIRM;
+	ib.ptr[2] = lower_32_bits(gpu_addr);
+	ib.ptr[3] = upper_32_bits(gpu_addr);
+	ib.ptr[4] = 0xDEADBEEF;
+	ib.length_dw = 5;
+
+	r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f);
+	if (r)
+		goto err2;
+
+	r = dma_fence_wait_timeout(f, false, timeout);
+	if (r == 0) {
+		r = -ETIMEDOUT;
+		goto err2;
+	} else if (r < 0) {
+		goto err2;
+	}
+
+	if (le32_to_cpu(*cpu_ptr) == 0xDEADBEEF)
+		r = 0;
+	else
+		r = -EINVAL;
+err2:
+	amdgpu_ib_free(&ib, NULL);
+	dma_fence_put(f);
+err1:
+	amdgpu_device_wb_free(adev, index);
+	return r;
+}
+
+static void gfx_v12_0_free_microcode(struct amdgpu_device *adev)
+{
+	amdgpu_ucode_release(&adev->gfx.pfp_fw);
+	amdgpu_ucode_release(&adev->gfx.me_fw);
+	amdgpu_ucode_release(&adev->gfx.rlc_fw);
+	amdgpu_ucode_release(&adev->gfx.mec_fw);
+
+	kfree(adev->gfx.rlc.register_list_format);
+}
+
+static int gfx_v12_0_init_toc_microcode(struct amdgpu_device *adev, const char *ucode_prefix)
+{
+	const struct psp_firmware_header_v1_0 *toc_hdr;
+	int err = 0;
+
+	err = amdgpu_ucode_request(adev, &adev->psp.toc_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_toc.bin", ucode_prefix);
+	if (err)
+		goto out;
+
+	toc_hdr = (const struct psp_firmware_header_v1_0 *)adev->psp.toc_fw->data;
+	adev->psp.toc.fw_version = le32_to_cpu(toc_hdr->header.ucode_version);
+	adev->psp.toc.feature_version = le32_to_cpu(toc_hdr->sos.fw_version);
+	adev->psp.toc.size_bytes = le32_to_cpu(toc_hdr->header.ucode_size_bytes);
+	adev->psp.toc.start_addr = (uint8_t *)toc_hdr +
+			le32_to_cpu(toc_hdr->header.ucode_array_offset_bytes);
+	return 0;
+out:
+	amdgpu_ucode_release(&adev->psp.toc_fw);
+	return err;
+}
+
+static int gfx_v12_0_init_microcode(struct amdgpu_device *adev)
+{
+	char ucode_prefix[30];
+	int err;
+	const struct rlc_firmware_header_v2_0 *rlc_hdr;
+	uint16_t version_major;
+	uint16_t version_minor;
+
+	DRM_DEBUG("\n");
+
+	amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix));
+
+	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_pfp.bin", ucode_prefix);
+	if (err)
+		goto out;
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_PFP);
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_PFP_P0_STACK);
+
+	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_me.bin", ucode_prefix);
+	if (err)
+		goto out;
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_ME);
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_ME_P0_STACK);
+
+	if (!amdgpu_sriov_vf(adev)) {
+		if (amdgpu_is_kicker_fw(adev))
+			err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_rlc_kicker.bin", ucode_prefix);
+		else
+			err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_rlc.bin", ucode_prefix);
+		if (err)
+			goto out;
+		rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
+		version_major = le16_to_cpu(rlc_hdr->header.header_version_major);
+		version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor);
+		err = amdgpu_gfx_rlc_init_microcode(adev, version_major, version_minor);
+		if (err)
+			goto out;
+	}
+
+	err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_mec.bin", ucode_prefix);
+	if (err)
+		goto out;
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_MEC);
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_MEC_P0_STACK);
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_RS64_MEC_P1_STACK);
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO)
+		err = gfx_v12_0_init_toc_microcode(adev, ucode_prefix);
+
+	/* only one MEC for gfx 12 */
+	adev->gfx.mec2_fw = NULL;
+
+	if (adev->gfx.imu.funcs) {
+		if (adev->gfx.imu.funcs->init_microcode) {
+			err = adev->gfx.imu.funcs->init_microcode(adev);
+			if (err)
+				dev_err(adev->dev, "Failed to load imu firmware!\n");
+		}
+	}
+
+out:
+	if (err) {
+		amdgpu_ucode_release(&adev->gfx.pfp_fw);
+		amdgpu_ucode_release(&adev->gfx.me_fw);
+		amdgpu_ucode_release(&adev->gfx.rlc_fw);
+		amdgpu_ucode_release(&adev->gfx.mec_fw);
+	}
+
+	return err;
+}
+
+static u32 gfx_v12_0_get_csb_size(struct amdgpu_device *adev)
+{
+	u32 count = 0;
+	const struct cs_section_def *sect = NULL;
+	const struct cs_extent_def *ext = NULL;
+
+	count += 1;
+
+	for (sect = gfx12_cs_data; sect->section != NULL; ++sect) {
+		if (sect->id == SECT_CONTEXT) {
+			for (ext = sect->section; ext->extent != NULL; ++ext)
+				count += 2 + ext->reg_count;
+		} else
+			return 0;
+	}
+
+	return count;
+}
+
+static void gfx_v12_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer)
+{
+	u32 count = 0, clustercount = 0, i;
+	const struct cs_section_def *sect = NULL;
+	const struct cs_extent_def *ext = NULL;
+
+	if (adev->gfx.rlc.cs_data == NULL)
+		return;
+	if (buffer == NULL)
+		return;
+
+	count += 1;
+
+	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
+		if (sect->id == SECT_CONTEXT) {
+			for (ext = sect->section; ext->extent != NULL; ++ext) {
+				clustercount++;
+				buffer[count++] = ext->reg_count;
+				buffer[count++] = ext->reg_index;
+
+				for (i = 0; i < ext->reg_count; i++)
+					buffer[count++] = cpu_to_le32(ext->extent[i]);
+			}
+		} else
+			return;
+	}
+
+	buffer[0] = clustercount;
+}
+
+static void gfx_v12_0_rlc_fini(struct amdgpu_device *adev)
+{
+	/* clear state block */
+	amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj,
+			&adev->gfx.rlc.clear_state_gpu_addr,
+			(void **)&adev->gfx.rlc.cs_ptr);
+
+	/* jump table block */
+	amdgpu_bo_free_kernel(&adev->gfx.rlc.cp_table_obj,
+			&adev->gfx.rlc.cp_table_gpu_addr,
+			(void **)&adev->gfx.rlc.cp_table_ptr);
+}
+
+static void gfx_v12_0_init_rlcg_reg_access_ctrl(struct amdgpu_device *adev)
+{
+	struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl;
+
+	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl[0];
+	reg_access_ctrl->scratch_reg0 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG0);
+	reg_access_ctrl->scratch_reg1 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG1);
+	reg_access_ctrl->scratch_reg2 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG2);
+	reg_access_ctrl->scratch_reg3 = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG3);
+	reg_access_ctrl->grbm_cntl = SOC15_REG_OFFSET(GC, 0, regGRBM_GFX_CNTL);
+	reg_access_ctrl->grbm_idx = SOC15_REG_OFFSET(GC, 0, regGRBM_GFX_INDEX);
+	reg_access_ctrl->spare_int = SOC15_REG_OFFSET(GC, 0, regRLC_SPARE_INT_0);
+	adev->gfx.rlc.rlcg_reg_access_supported = true;
+}
+
+static int gfx_v12_0_rlc_init(struct amdgpu_device *adev)
+{
+	const struct cs_section_def *cs_data;
+	int r;
+
+	adev->gfx.rlc.cs_data = gfx12_cs_data;
+
+	cs_data = adev->gfx.rlc.cs_data;
+
+	if (cs_data) {
+		/* init clear state block */
+		r = amdgpu_gfx_rlc_init_csb(adev);
+		if (r)
+			return r;
+	}
+
+	/* init spm vmid with 0xf */
+	if (adev->gfx.rlc.funcs->update_spm_vmid)
+		adev->gfx.rlc.funcs->update_spm_vmid(adev, NULL, 0xf);
+
+	return 0;
+}
+
+static void gfx_v12_0_mec_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_free_kernel(&adev->gfx.mec.hpd_eop_obj, NULL, NULL);
+	amdgpu_bo_free_kernel(&adev->gfx.mec.mec_fw_obj, NULL, NULL);
+	amdgpu_bo_free_kernel(&adev->gfx.mec.mec_fw_data_obj, NULL, NULL);
+}
+
+static void gfx_v12_0_me_init(struct amdgpu_device *adev)
+{
+	bitmap_zero(adev->gfx.me.queue_bitmap, AMDGPU_MAX_GFX_QUEUES);
+
+	amdgpu_gfx_graphics_queue_acquire(adev);
+}
+
+static int gfx_v12_0_mec_init(struct amdgpu_device *adev)
+{
+	int r;
+	u32 *hpd;
+	size_t mec_hpd_size;
+
+	bitmap_zero(adev->gfx.mec_bitmap[0].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
+
+	/* take ownership of the relevant compute queues */
+	amdgpu_gfx_compute_queue_acquire(adev);
+	mec_hpd_size = adev->gfx.num_compute_rings * GFX12_MEC_HPD_SIZE;
+
+	if (mec_hpd_size) {
+		r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE,
+					      AMDGPU_GEM_DOMAIN_GTT,
+					      &adev->gfx.mec.hpd_eop_obj,
+					      &adev->gfx.mec.hpd_eop_gpu_addr,
+					      (void **)&hpd);
+		if (r) {
+			dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r);
+			gfx_v12_0_mec_fini(adev);
+			return r;
+		}
+
+		memset(hpd, 0, mec_hpd_size);
+
+		amdgpu_bo_kunmap(adev->gfx.mec.hpd_eop_obj);
+		amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj);
+	}
+
+	return 0;
+}
+
+static uint32_t wave_read_ind(struct amdgpu_device *adev, uint32_t wave, uint32_t address)
+{
+	WREG32_SOC15(GC, 0, regSQ_IND_INDEX,
+		(wave << SQ_IND_INDEX__WAVE_ID__SHIFT) |
+		(address << SQ_IND_INDEX__INDEX__SHIFT));
+	return RREG32_SOC15(GC, 0, regSQ_IND_DATA);
+}
+
+static void wave_read_regs(struct amdgpu_device *adev, uint32_t wave,
+			   uint32_t thread, uint32_t regno,
+			   uint32_t num, uint32_t *out)
+{
+	WREG32_SOC15(GC, 0, regSQ_IND_INDEX,
+		(wave << SQ_IND_INDEX__WAVE_ID__SHIFT) |
+		(regno << SQ_IND_INDEX__INDEX__SHIFT) |
+		(thread << SQ_IND_INDEX__WORKITEM_ID__SHIFT) |
+		(SQ_IND_INDEX__AUTO_INCR_MASK));
+	while (num--)
+		*(out++) = RREG32_SOC15(GC, 0, regSQ_IND_DATA);
+}
+
+static void gfx_v12_0_read_wave_data(struct amdgpu_device *adev,
+				     uint32_t xcc_id,
+				     uint32_t simd, uint32_t wave,
+				     uint32_t *dst, int *no_fields)
+{
+	/* in gfx12 the SIMD_ID is specified as part of the INSTANCE
+	 * field when performing a select_se_sh so it should be
+	 * zero here */
+	WARN_ON(simd != 0);
+
+	/* type 4 wave data */
+	dst[(*no_fields)++] = 4;
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_STATUS);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_PC_LO);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_PC_HI);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_EXEC_LO);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_EXEC_HI);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_HW_ID1);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_HW_ID2);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_GPR_ALLOC);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_LDS_ALLOC);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_IB_STS);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_IB_STS2);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_IB_DBG1);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_M0);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_MODE);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_STATE_PRIV);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_EXCP_FLAG_PRIV);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_EXCP_FLAG_USER);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_TRAP_CTRL);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_ACTIVE);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_VALID_AND_IDLE);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_DVGPR_ALLOC_LO);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_DVGPR_ALLOC_HI);
+	dst[(*no_fields)++] = wave_read_ind(adev, wave, ixSQ_WAVE_SCHED_MODE);
+}
+
+static void gfx_v12_0_read_wave_sgprs(struct amdgpu_device *adev,
+				      uint32_t xcc_id, uint32_t simd,
+				      uint32_t wave, uint32_t start,
+				      uint32_t size, uint32_t *dst)
+{
+	WARN_ON(simd != 0);
+
+	wave_read_regs(
+		adev, wave, 0, start + SQIND_WAVE_SGPRS_OFFSET, size,
+		dst);
+}
+
+static void gfx_v12_0_read_wave_vgprs(struct amdgpu_device *adev,
+				      uint32_t xcc_id, uint32_t simd,
+				      uint32_t wave, uint32_t thread,
+				      uint32_t start, uint32_t size,
+				      uint32_t *dst)
+{
+	wave_read_regs(
+		adev, wave, thread,
+		start + SQIND_WAVE_VGPRS_OFFSET, size, dst);
+}
+
+static void gfx_v12_0_select_me_pipe_q(struct amdgpu_device *adev,
+				       u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
+{
+	soc24_grbm_select(adev, me, pipe, q, vm);
+}
+
+/* all sizes are in bytes */
+#define MQD_SHADOW_BASE_SIZE      73728
+#define MQD_SHADOW_BASE_ALIGNMENT 256
+#define MQD_FWWORKAREA_SIZE       484
+#define MQD_FWWORKAREA_ALIGNMENT  256
+
+static void gfx_v12_0_get_gfx_shadow_info_nocheck(struct amdgpu_device *adev,
+						  struct amdgpu_gfx_shadow_info *shadow_info)
+{
+	shadow_info->shadow_size = MQD_SHADOW_BASE_SIZE;
+	shadow_info->shadow_alignment = MQD_SHADOW_BASE_ALIGNMENT;
+	shadow_info->csa_size = MQD_FWWORKAREA_SIZE;
+	shadow_info->csa_alignment = MQD_FWWORKAREA_ALIGNMENT;
+}
+
+static int gfx_v12_0_get_gfx_shadow_info(struct amdgpu_device *adev,
+					 struct amdgpu_gfx_shadow_info *shadow_info,
+					 bool skip_check)
+{
+	if (adev->gfx.cp_gfx_shadow || skip_check) {
+		gfx_v12_0_get_gfx_shadow_info_nocheck(adev, shadow_info);
+		return 0;
+	}
+
+	memset(shadow_info, 0, sizeof(struct amdgpu_gfx_shadow_info));
+	return -EINVAL;
+}
+
+static const struct amdgpu_gfx_funcs gfx_v12_0_gfx_funcs = {
+	.get_gpu_clock_counter = &gfx_v12_0_get_gpu_clock_counter,
+	.select_se_sh = &gfx_v12_0_select_se_sh,
+	.read_wave_data = &gfx_v12_0_read_wave_data,
+	.read_wave_sgprs = &gfx_v12_0_read_wave_sgprs,
+	.read_wave_vgprs = &gfx_v12_0_read_wave_vgprs,
+	.select_me_pipe_q = &gfx_v12_0_select_me_pipe_q,
+	.update_perfmon_mgcg = &gfx_v12_0_update_perf_clk,
+	.get_gfx_shadow_info = &gfx_v12_0_get_gfx_shadow_info,
+};
+
+static int gfx_v12_0_gpu_early_init(struct amdgpu_device *adev)
+{
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		adev->gfx.config.max_hw_contexts = 8;
+		adev->gfx.config.sc_prim_fifo_size_frontend = 0x20;
+		adev->gfx.config.sc_prim_fifo_size_backend = 0x100;
+		adev->gfx.config.sc_hiz_tile_fifo_size = 0;
+		adev->gfx.config.sc_earlyz_tile_fifo_size = 0x4C0;
+		break;
+	default:
+		BUG();
+		break;
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_gfx_ring_init(struct amdgpu_device *adev, int ring_id,
+				   int me, int pipe, int queue)
+{
+	int r;
+	struct amdgpu_ring *ring;
+	unsigned int irq_type;
+
+	ring = &adev->gfx.gfx_ring[ring_id];
+
+	ring->me = me;
+	ring->pipe = pipe;
+	ring->queue = queue;
+
+	ring->ring_obj = NULL;
+	ring->use_doorbell = true;
+
+	if (!ring_id)
+		ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
+	else
+		ring->doorbell_index = adev->doorbell_index.gfx_ring1 << 1;
+	ring->vm_hub = AMDGPU_GFXHUB(0);
+	sprintf(ring->name, "gfx_%d.%d.%d", ring->me, ring->pipe, ring->queue);
+
+	irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + ring->pipe;
+	r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type,
+			     AMDGPU_RING_PRIO_DEFAULT, NULL);
+	if (r)
+		return r;
+	return 0;
+}
+
+static int gfx_v12_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
+				       int mec, int pipe, int queue)
+{
+	int r;
+	unsigned irq_type;
+	struct amdgpu_ring *ring;
+	unsigned int hw_prio;
+
+	ring = &adev->gfx.compute_ring[ring_id];
+
+	/* mec0 is me1 */
+	ring->me = mec + 1;
+	ring->pipe = pipe;
+	ring->queue = queue;
+
+	ring->ring_obj = NULL;
+	ring->use_doorbell = true;
+	ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
+	ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
+				+ (ring_id * GFX12_MEC_HPD_SIZE);
+	ring->vm_hub = AMDGPU_GFXHUB(0);
+	sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
+
+	irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
+		+ ((ring->me - 1) * adev->gfx.mec.num_pipe_per_mec)
+		+ ring->pipe;
+	hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev, ring) ?
+			AMDGPU_GFX_PIPE_PRIO_HIGH : AMDGPU_GFX_PIPE_PRIO_NORMAL;
+	/* type-2 packets are deprecated on MEC, use type-3 instead */
+	r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type,
+			     hw_prio, NULL);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static struct {
+	SOC24_FIRMWARE_ID	id;
+	unsigned int		offset;
+	unsigned int		size;
+	unsigned int		size_x16;
+} rlc_autoload_info[SOC24_FIRMWARE_ID_MAX];
+
+#define RLC_TOC_OFFSET_DWUNIT   8
+#define RLC_SIZE_MULTIPLE       1024
+#define RLC_TOC_UMF_SIZE_inM	23ULL
+#define RLC_TOC_FORMAT_API	165ULL
+
+static void gfx_v12_0_parse_rlc_toc(struct amdgpu_device *adev, void *rlc_toc)
+{
+	RLC_TABLE_OF_CONTENT_V2 *ucode = rlc_toc;
+
+	while (ucode && (ucode->id > SOC24_FIRMWARE_ID_INVALID)) {
+		rlc_autoload_info[ucode->id].id = ucode->id;
+		rlc_autoload_info[ucode->id].offset =
+			ucode->offset * RLC_TOC_OFFSET_DWUNIT * 4;
+		rlc_autoload_info[ucode->id].size =
+			ucode->size_x16 ? ucode->size * RLC_SIZE_MULTIPLE * 4 :
+					  ucode->size * 4;
+		ucode++;
+	}
+}
+
+static uint32_t gfx_v12_0_calc_toc_total_size(struct amdgpu_device *adev)
+{
+	uint32_t total_size = 0;
+	SOC24_FIRMWARE_ID id;
+
+	gfx_v12_0_parse_rlc_toc(adev, adev->psp.toc.start_addr);
+
+	for (id = SOC24_FIRMWARE_ID_RLC_G_UCODE; id < SOC24_FIRMWARE_ID_MAX; id++)
+		total_size += rlc_autoload_info[id].size;
+
+	/* In case the offset in rlc toc ucode is aligned */
+	if (total_size < rlc_autoload_info[SOC24_FIRMWARE_ID_MAX-1].offset)
+		total_size = rlc_autoload_info[SOC24_FIRMWARE_ID_MAX-1].offset +
+			rlc_autoload_info[SOC24_FIRMWARE_ID_MAX-1].size;
+	if (total_size < (RLC_TOC_UMF_SIZE_inM << 20))
+		total_size = RLC_TOC_UMF_SIZE_inM << 20;
+
+	return total_size;
+}
+
+static int gfx_v12_0_rlc_autoload_buffer_init(struct amdgpu_device *adev)
+{
+	int r;
+	uint32_t total_size;
+
+	total_size = gfx_v12_0_calc_toc_total_size(adev);
+
+	r = amdgpu_bo_create_reserved(adev, total_size, 64 * 1024,
+				      AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->gfx.rlc.rlc_autoload_bo,
+				      &adev->gfx.rlc.rlc_autoload_gpu_addr,
+				      (void **)&adev->gfx.rlc.rlc_autoload_ptr);
+
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create fw autoload bo\n", r);
+		return r;
+	}
+
+	return 0;
+}
+
+static void gfx_v12_0_rlc_backdoor_autoload_copy_ucode(struct amdgpu_device *adev,
+						       SOC24_FIRMWARE_ID id,
+						       const void *fw_data,
+						       uint32_t fw_size)
+{
+	uint32_t toc_offset;
+	uint32_t toc_fw_size;
+	char *ptr = adev->gfx.rlc.rlc_autoload_ptr;
+
+	if (id <= SOC24_FIRMWARE_ID_INVALID || id >= SOC24_FIRMWARE_ID_MAX)
+		return;
+
+	toc_offset = rlc_autoload_info[id].offset;
+	toc_fw_size = rlc_autoload_info[id].size;
+
+	if (fw_size == 0)
+		fw_size = toc_fw_size;
+
+	if (fw_size > toc_fw_size)
+		fw_size = toc_fw_size;
+
+	memcpy(ptr + toc_offset, fw_data, fw_size);
+
+	if (fw_size < toc_fw_size)
+		memset(ptr + toc_offset + fw_size, 0, toc_fw_size - fw_size);
+}
+
+static void
+gfx_v12_0_rlc_backdoor_autoload_copy_toc_ucode(struct amdgpu_device *adev)
+{
+	void *data;
+	uint32_t size;
+	uint32_t *toc_ptr;
+
+	data = adev->psp.toc.start_addr;
+	size = rlc_autoload_info[SOC24_FIRMWARE_ID_RLC_TOC].size;
+
+	toc_ptr = (uint32_t *)data + size / 4 - 2;
+	*toc_ptr = (RLC_TOC_FORMAT_API << 24) | 0x1;
+
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RLC_TOC,
+						   data, size);
+}
+
+static void
+gfx_v12_0_rlc_backdoor_autoload_copy_gfx_ucode(struct amdgpu_device *adev)
+{
+	const __le32 *fw_data;
+	uint32_t fw_size;
+	const struct gfx_firmware_header_v2_0 *cpv2_hdr;
+	const struct rlc_firmware_header_v2_0 *rlc_hdr;
+	const struct rlc_firmware_header_v2_1 *rlcv21_hdr;
+	const struct rlc_firmware_header_v2_2 *rlcv22_hdr;
+	uint16_t version_major, version_minor;
+
+	/* pfp ucode */
+	cpv2_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.pfp_fw->data;
+	/* instruction */
+	fw_data = (const __le32 *)(adev->gfx.pfp_fw->data +
+		le32_to_cpu(cpv2_hdr->ucode_offset_bytes));
+	fw_size = le32_to_cpu(cpv2_hdr->ucode_size_bytes);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_PFP,
+						   fw_data, fw_size);
+	/* data */
+	fw_data = (const __le32 *)(adev->gfx.pfp_fw->data +
+		le32_to_cpu(cpv2_hdr->data_offset_bytes));
+	fw_size = le32_to_cpu(cpv2_hdr->data_size_bytes);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_PFP_P0_STACK,
+						   fw_data, fw_size);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_PFP_P1_STACK,
+						   fw_data, fw_size);
+	/* me ucode */
+	cpv2_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.me_fw->data;
+	/* instruction */
+	fw_data = (const __le32 *)(adev->gfx.me_fw->data +
+		le32_to_cpu(cpv2_hdr->ucode_offset_bytes));
+	fw_size = le32_to_cpu(cpv2_hdr->ucode_size_bytes);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_ME,
+						   fw_data, fw_size);
+	/* data */
+	fw_data = (const __le32 *)(adev->gfx.me_fw->data +
+		le32_to_cpu(cpv2_hdr->data_offset_bytes));
+	fw_size = le32_to_cpu(cpv2_hdr->data_size_bytes);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_ME_P0_STACK,
+						   fw_data, fw_size);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_ME_P1_STACK,
+						   fw_data, fw_size);
+	/* mec ucode */
+	cpv2_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.mec_fw->data;
+	/* instruction */
+	fw_data = (const __le32 *) (adev->gfx.mec_fw->data +
+		le32_to_cpu(cpv2_hdr->ucode_offset_bytes));
+	fw_size = le32_to_cpu(cpv2_hdr->ucode_size_bytes);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_MEC,
+						   fw_data, fw_size);
+	/* data */
+	fw_data = (const __le32 *) (adev->gfx.mec_fw->data +
+		le32_to_cpu(cpv2_hdr->data_offset_bytes));
+	fw_size = le32_to_cpu(cpv2_hdr->data_size_bytes);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_MEC_P0_STACK,
+						   fw_data, fw_size);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_MEC_P1_STACK,
+						   fw_data, fw_size);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_MEC_P2_STACK,
+						   fw_data, fw_size);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RS64_MEC_P3_STACK,
+						   fw_data, fw_size);
+
+	/* rlc ucode */
+	rlc_hdr = (const struct rlc_firmware_header_v2_0 *)
+		adev->gfx.rlc_fw->data;
+	fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+			le32_to_cpu(rlc_hdr->header.ucode_array_offset_bytes));
+	fw_size = le32_to_cpu(rlc_hdr->header.ucode_size_bytes);
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RLC_G_UCODE,
+						   fw_data, fw_size);
+
+	version_major = le16_to_cpu(rlc_hdr->header.header_version_major);
+	version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor);
+	if (version_major == 2) {
+		if (version_minor >= 1) {
+			rlcv21_hdr = (const struct rlc_firmware_header_v2_1 *)adev->gfx.rlc_fw->data;
+
+			fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+					le32_to_cpu(rlcv21_hdr->save_restore_list_gpm_offset_bytes));
+			fw_size = le32_to_cpu(rlcv21_hdr->save_restore_list_gpm_size_bytes);
+			gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RLCG_SCRATCH,
+						   fw_data, fw_size);
+
+			fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+					le32_to_cpu(rlcv21_hdr->save_restore_list_srm_offset_bytes));
+			fw_size = le32_to_cpu(rlcv21_hdr->save_restore_list_srm_size_bytes);
+			gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RLC_SRM_ARAM,
+						   fw_data, fw_size);
+		}
+		if (version_minor >= 2) {
+			rlcv22_hdr = (const struct rlc_firmware_header_v2_2 *)adev->gfx.rlc_fw->data;
+
+			fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+					le32_to_cpu(rlcv22_hdr->rlc_iram_ucode_offset_bytes));
+			fw_size = le32_to_cpu(rlcv22_hdr->rlc_iram_ucode_size_bytes);
+			gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RLX6_UCODE,
+						   fw_data, fw_size);
+
+			fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+					le32_to_cpu(rlcv22_hdr->rlc_dram_ucode_offset_bytes));
+			fw_size = le32_to_cpu(rlcv22_hdr->rlc_dram_ucode_size_bytes);
+			gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_RLX6_DRAM_BOOT,
+						   fw_data, fw_size);
+		}
+	}
+}
+
+static void
+gfx_v12_0_rlc_backdoor_autoload_copy_sdma_ucode(struct amdgpu_device *adev)
+{
+	const __le32 *fw_data;
+	uint32_t fw_size;
+	const struct sdma_firmware_header_v3_0 *sdma_hdr;
+
+	sdma_hdr = (const struct sdma_firmware_header_v3_0 *)
+		adev->sdma.instance[0].fw->data;
+	fw_data = (const __le32 *) (adev->sdma.instance[0].fw->data +
+			le32_to_cpu(sdma_hdr->ucode_offset_bytes));
+	fw_size = le32_to_cpu(sdma_hdr->ucode_size_bytes);
+
+	gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, SOC24_FIRMWARE_ID_SDMA_UCODE_TH0,
+						   fw_data, fw_size);
+}
+
+static void
+gfx_v12_0_rlc_backdoor_autoload_copy_mes_ucode(struct amdgpu_device *adev)
+{
+	const __le32 *fw_data;
+	unsigned fw_size;
+	const struct mes_firmware_header_v1_0 *mes_hdr;
+	int pipe, ucode_id, data_id;
+
+	for (pipe = 0; pipe < 2; pipe++) {
+		if (pipe == 0) {
+			ucode_id = SOC24_FIRMWARE_ID_RS64_MES_P0;
+			data_id  = SOC24_FIRMWARE_ID_RS64_MES_P0_STACK;
+		} else {
+			ucode_id = SOC24_FIRMWARE_ID_RS64_MES_P1;
+			data_id  = SOC24_FIRMWARE_ID_RS64_MES_P1_STACK;
+		}
+
+		mes_hdr = (const struct mes_firmware_header_v1_0 *)
+			adev->mes.fw[pipe]->data;
+
+		fw_data = (const __le32 *)(adev->mes.fw[pipe]->data +
+				le32_to_cpu(mes_hdr->mes_ucode_offset_bytes));
+		fw_size = le32_to_cpu(mes_hdr->mes_ucode_size_bytes);
+
+		gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, ucode_id, fw_data, fw_size);
+
+		fw_data = (const __le32 *)(adev->mes.fw[pipe]->data +
+				le32_to_cpu(mes_hdr->mes_ucode_data_offset_bytes));
+		fw_size = le32_to_cpu(mes_hdr->mes_ucode_data_size_bytes);
+
+		gfx_v12_0_rlc_backdoor_autoload_copy_ucode(adev, data_id, fw_data, fw_size);
+	}
+}
+
+static int gfx_v12_0_rlc_backdoor_autoload_enable(struct amdgpu_device *adev)
+{
+	uint32_t rlc_g_offset, rlc_g_size;
+	uint64_t gpu_addr;
+	uint32_t data;
+
+	/* RLC autoload sequence 2: copy ucode */
+	gfx_v12_0_rlc_backdoor_autoload_copy_sdma_ucode(adev);
+	gfx_v12_0_rlc_backdoor_autoload_copy_gfx_ucode(adev);
+	gfx_v12_0_rlc_backdoor_autoload_copy_mes_ucode(adev);
+	gfx_v12_0_rlc_backdoor_autoload_copy_toc_ucode(adev);
+
+	rlc_g_offset = rlc_autoload_info[SOC24_FIRMWARE_ID_RLC_G_UCODE].offset;
+	rlc_g_size = rlc_autoload_info[SOC24_FIRMWARE_ID_RLC_G_UCODE].size;
+	gpu_addr = adev->gfx.rlc.rlc_autoload_gpu_addr + rlc_g_offset - adev->gmc.vram_start;
+
+	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_BOOTLOADER_ADDR_HI, upper_32_bits(gpu_addr));
+	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_BOOTLOADER_ADDR_LO, lower_32_bits(gpu_addr));
+
+	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_BOOTLOADER_SIZE, rlc_g_size);
+
+	if (adev->gfx.imu.funcs && (amdgpu_dpm > 0)) {
+		/* RLC autoload sequence 3: load IMU fw */
+		if (adev->gfx.imu.funcs->load_microcode)
+			adev->gfx.imu.funcs->load_microcode(adev);
+		/* RLC autoload sequence 4 init IMU fw */
+		if (adev->gfx.imu.funcs->setup_imu)
+			adev->gfx.imu.funcs->setup_imu(adev);
+		if (adev->gfx.imu.funcs->start_imu)
+			adev->gfx.imu.funcs->start_imu(adev);
+
+		/* RLC autoload sequence 5 disable gpa mode */
+		gfx_v12_0_disable_gpa_mode(adev);
+	} else {
+		/* unhalt rlc to start autoload without imu */
+		data = RREG32_SOC15(GC, 0, regRLC_GPM_THREAD_ENABLE);
+		data = REG_SET_FIELD(data, RLC_GPM_THREAD_ENABLE, THREAD0_ENABLE, 1);
+		data = REG_SET_FIELD(data, RLC_GPM_THREAD_ENABLE, THREAD1_ENABLE, 1);
+		WREG32_SOC15(GC, 0, regRLC_GPM_THREAD_ENABLE, data);
+		WREG32_SOC15(GC, 0, regRLC_CNTL, RLC_CNTL__RLC_ENABLE_F32_MASK);
+	}
+
+	return 0;
+}
+
+static void gfx_v12_0_alloc_ip_dump(struct amdgpu_device *adev)
+{
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_12_0);
+	uint32_t *ptr;
+	uint32_t inst;
+
+	ptr = kcalloc(reg_count, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for GFX IP Dump\n");
+		adev->gfx.ip_dump_core = NULL;
+	} else {
+		adev->gfx.ip_dump_core = ptr;
+	}
+
+	/* Allocate memory for compute queue registers for all the instances */
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_12);
+	inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec *
+		adev->gfx.mec.num_queue_per_pipe;
+
+	ptr = kcalloc(reg_count * inst, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for Compute Queues IP Dump\n");
+		adev->gfx.ip_dump_compute_queues = NULL;
+	} else {
+		adev->gfx.ip_dump_compute_queues = ptr;
+	}
+
+	/* Allocate memory for gfx queue registers for all the instances */
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_12);
+	inst = adev->gfx.me.num_me * adev->gfx.me.num_pipe_per_me *
+		adev->gfx.me.num_queue_per_pipe;
+
+	ptr = kcalloc(reg_count * inst, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for GFX Queues IP Dump\n");
+		adev->gfx.ip_dump_gfx_queues = NULL;
+	} else {
+		adev->gfx.ip_dump_gfx_queues = ptr;
+	}
+}
+
+static int gfx_v12_0_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	int i, j, k, r, ring_id = 0;
+	unsigned num_compute_rings;
+	int xcc_id = 0;
+	struct amdgpu_device *adev = ip_block->adev;
+	int num_queue_per_pipe = 1; /* we only enable 1 KGQ per pipe */
+
+	INIT_DELAYED_WORK(&adev->gfx.idle_work, amdgpu_gfx_profile_idle_work_handler);
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		adev->gfx.me.num_me = 1;
+		adev->gfx.me.num_pipe_per_me = 1;
+		adev->gfx.me.num_queue_per_pipe = 8;
+		adev->gfx.mec.num_mec = 1;
+		adev->gfx.mec.num_pipe_per_mec = 2;
+		adev->gfx.mec.num_queue_per_pipe = 4;
+		break;
+	default:
+		adev->gfx.me.num_me = 1;
+		adev->gfx.me.num_pipe_per_me = 1;
+		adev->gfx.me.num_queue_per_pipe = 1;
+		adev->gfx.mec.num_mec = 1;
+		adev->gfx.mec.num_pipe_per_mec = 4;
+		adev->gfx.mec.num_queue_per_pipe = 8;
+		break;
+	}
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		if (!adev->gfx.disable_uq &&
+		    adev->gfx.me_fw_version  >= 2780 &&
+		    adev->gfx.pfp_fw_version >= 2840 &&
+		    adev->gfx.mec_fw_version >= 3050 &&
+		    adev->mes.fw_version[0] >= 123) {
+			adev->userq_funcs[AMDGPU_HW_IP_GFX] = &userq_mes_funcs;
+			adev->userq_funcs[AMDGPU_HW_IP_COMPUTE] = &userq_mes_funcs;
+		}
+		break;
+	default:
+		break;
+	}
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		if (adev->gfx.me_fw_version  >= 2480 &&
+		    adev->gfx.pfp_fw_version >= 2530 &&
+		    adev->gfx.mec_fw_version >= 2680 &&
+		    adev->mes.fw_version[0] >= 100)
+			adev->gfx.enable_cleaner_shader = true;
+		break;
+	default:
+		adev->gfx.enable_cleaner_shader = false;
+		break;
+	}
+
+	if (adev->gfx.num_compute_rings) {
+		/* recalculate compute rings to use based on hardware configuration */
+		num_compute_rings = (adev->gfx.mec.num_pipe_per_mec *
+				     adev->gfx.mec.num_queue_per_pipe) / 2;
+		adev->gfx.num_compute_rings = min(adev->gfx.num_compute_rings,
+						  num_compute_rings);
+	}
+
+	/* EOP Event */
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP,
+			      GFX_12_0_0__SRCID__CP_EOP_INTERRUPT,
+			      &adev->gfx.eop_irq);
+	if (r)
+		return r;
+
+	/* Bad opcode Event */
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP,
+			      GFX_12_0_0__SRCID__CP_BAD_OPCODE_ERROR,
+			      &adev->gfx.bad_op_irq);
+	if (r)
+		return r;
+
+	/* Privileged reg */
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP,
+			      GFX_12_0_0__SRCID__CP_PRIV_REG_FAULT,
+			      &adev->gfx.priv_reg_irq);
+	if (r)
+		return r;
+
+	/* Privileged inst */
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GRBM_CP,
+			      GFX_12_0_0__SRCID__CP_PRIV_INSTR_FAULT,
+			      &adev->gfx.priv_inst_irq);
+	if (r)
+		return r;
+
+	adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE;
+
+	gfx_v12_0_me_init(adev);
+
+	r = gfx_v12_0_rlc_init(adev);
+	if (r) {
+		dev_err(adev->dev, "Failed to init rlc BOs!\n");
+		return r;
+	}
+
+	r = gfx_v12_0_mec_init(adev);
+	if (r) {
+		dev_err(adev->dev, "Failed to init MEC BOs!\n");
+		return r;
+	}
+
+	if (adev->gfx.num_gfx_rings) {
+		/* set up the gfx ring */
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < num_queue_per_pipe; j++) {
+				for (k = 0; k < adev->gfx.me.num_pipe_per_me; k++) {
+					if (!amdgpu_gfx_is_me_queue_enabled(adev, i, k, j))
+						continue;
+
+					r = gfx_v12_0_gfx_ring_init(adev, ring_id,
+								    i, k, j);
+					if (r)
+						return r;
+					ring_id++;
+				}
+			}
+		}
+	}
+
+	if (adev->gfx.num_compute_rings) {
+		ring_id = 0;
+		/* set up the compute queues - allocate horizontally across pipes */
+		for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
+			for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
+				for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
+					if (!amdgpu_gfx_is_mec_queue_enabled(adev,
+									     0, i, k, j))
+						continue;
+
+					r = gfx_v12_0_compute_ring_init(adev, ring_id,
+									i, k, j);
+					if (r)
+						return r;
+
+					ring_id++;
+				}
+			}
+		}
+	}
+
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		if ((adev->gfx.me_fw_version >= 2660) &&
+		    (adev->gfx.mec_fw_version >= 2920) &&
+		    !amdgpu_sriov_vf(adev) &&
+		    !adev->debug_disable_gpu_ring_reset) {
+			adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+			adev->gfx.gfx_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+		}
+		break;
+	default:
+		break;
+	}
+
+	if (!adev->enable_mes_kiq) {
+		r = amdgpu_gfx_kiq_init(adev, GFX12_MEC_HPD_SIZE, 0);
+		if (r) {
+			dev_err(adev->dev, "Failed to init KIQ BOs!\n");
+			return r;
+		}
+
+		r = amdgpu_gfx_kiq_init_ring(adev, xcc_id);
+		if (r)
+			return r;
+	}
+
+	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v12_compute_mqd), 0);
+	if (r)
+		return r;
+
+	/* allocate visible FB for rlc auto-loading fw */
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) {
+		r = gfx_v12_0_rlc_autoload_buffer_init(adev);
+		if (r)
+			return r;
+	}
+
+	r = gfx_v12_0_gpu_early_init(adev);
+	if (r)
+		return r;
+
+	gfx_v12_0_alloc_ip_dump(adev);
+
+	r = amdgpu_gfx_sysfs_init(adev);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static void gfx_v12_0_pfp_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_free_kernel(&adev->gfx.pfp.pfp_fw_obj,
+			      &adev->gfx.pfp.pfp_fw_gpu_addr,
+			      (void **)&adev->gfx.pfp.pfp_fw_ptr);
+
+	amdgpu_bo_free_kernel(&adev->gfx.pfp.pfp_fw_data_obj,
+			      &adev->gfx.pfp.pfp_fw_data_gpu_addr,
+			      (void **)&adev->gfx.pfp.pfp_fw_data_ptr);
+}
+
+static void gfx_v12_0_me_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_free_kernel(&adev->gfx.me.me_fw_obj,
+			      &adev->gfx.me.me_fw_gpu_addr,
+			      (void **)&adev->gfx.me.me_fw_ptr);
+
+	amdgpu_bo_free_kernel(&adev->gfx.me.me_fw_data_obj,
+			       &adev->gfx.me.me_fw_data_gpu_addr,
+			       (void **)&adev->gfx.me.me_fw_data_ptr);
+}
+
+static void gfx_v12_0_rlc_autoload_buffer_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_free_kernel(&adev->gfx.rlc.rlc_autoload_bo,
+			&adev->gfx.rlc.rlc_autoload_gpu_addr,
+			(void **)&adev->gfx.rlc.rlc_autoload_ptr);
+}
+
+static int gfx_v12_0_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	int i;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
+		amdgpu_ring_fini(&adev->gfx.gfx_ring[i]);
+	for (i = 0; i < adev->gfx.num_compute_rings; i++)
+		amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
+
+	amdgpu_gfx_mqd_sw_fini(adev, 0);
+
+	if (!adev->enable_mes_kiq) {
+		amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq[0].ring);
+		amdgpu_gfx_kiq_fini(adev, 0);
+	}
+
+	gfx_v12_0_pfp_fini(adev);
+	gfx_v12_0_me_fini(adev);
+	gfx_v12_0_rlc_fini(adev);
+	gfx_v12_0_mec_fini(adev);
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO)
+		gfx_v12_0_rlc_autoload_buffer_fini(adev);
+
+	gfx_v12_0_free_microcode(adev);
+
+	amdgpu_gfx_sysfs_fini(adev);
+
+	kfree(adev->gfx.ip_dump_core);
+	kfree(adev->gfx.ip_dump_compute_queues);
+	kfree(adev->gfx.ip_dump_gfx_queues);
+
+	return 0;
+}
+
+static void gfx_v12_0_select_se_sh(struct amdgpu_device *adev, u32 se_num,
+				   u32 sh_num, u32 instance, int xcc_id)
+{
+	u32 data;
+
+	if (instance == 0xffffffff)
+		data = REG_SET_FIELD(0, GRBM_GFX_INDEX,
+				     INSTANCE_BROADCAST_WRITES, 1);
+	else
+		data = REG_SET_FIELD(0, GRBM_GFX_INDEX, INSTANCE_INDEX,
+				     instance);
+
+	if (se_num == 0xffffffff)
+		data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_BROADCAST_WRITES,
+				     1);
+	else
+		data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_INDEX, se_num);
+
+	if (sh_num == 0xffffffff)
+		data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SA_BROADCAST_WRITES,
+				     1);
+	else
+		data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SA_INDEX, sh_num);
+
+	WREG32_SOC15(GC, 0, regGRBM_GFX_INDEX, data);
+}
+
+static u32 gfx_v12_0_get_sa_active_bitmap(struct amdgpu_device *adev)
+{
+	u32 gc_disabled_sa_mask, gc_user_disabled_sa_mask, sa_mask;
+
+	gc_disabled_sa_mask = RREG32_SOC15(GC, 0, regGRBM_CC_GC_SA_UNIT_DISABLE);
+	gc_disabled_sa_mask = REG_GET_FIELD(gc_disabled_sa_mask,
+					    GRBM_CC_GC_SA_UNIT_DISABLE,
+					    SA_DISABLE);
+	gc_user_disabled_sa_mask = RREG32_SOC15(GC, 0, regGRBM_GC_USER_SA_UNIT_DISABLE);
+	gc_user_disabled_sa_mask = REG_GET_FIELD(gc_user_disabled_sa_mask,
+						 GRBM_GC_USER_SA_UNIT_DISABLE,
+						 SA_DISABLE);
+	sa_mask = amdgpu_gfx_create_bitmask(adev->gfx.config.max_sh_per_se *
+					    adev->gfx.config.max_shader_engines);
+
+	return sa_mask & (~(gc_disabled_sa_mask | gc_user_disabled_sa_mask));
+}
+
+static u32 gfx_v12_0_get_rb_active_bitmap(struct amdgpu_device *adev)
+{
+	u32 gc_disabled_rb_mask, gc_user_disabled_rb_mask;
+	u32 rb_mask;
+
+	gc_disabled_rb_mask = RREG32_SOC15(GC, 0, regCC_RB_BACKEND_DISABLE);
+	gc_disabled_rb_mask = REG_GET_FIELD(gc_disabled_rb_mask,
+					    CC_RB_BACKEND_DISABLE,
+					    BACKEND_DISABLE);
+	gc_user_disabled_rb_mask = RREG32_SOC15(GC, 0, regGC_USER_RB_BACKEND_DISABLE);
+	gc_user_disabled_rb_mask = REG_GET_FIELD(gc_user_disabled_rb_mask,
+						 GC_USER_RB_BACKEND_DISABLE,
+						 BACKEND_DISABLE);
+	rb_mask = amdgpu_gfx_create_bitmask(adev->gfx.config.max_backends_per_se *
+					    adev->gfx.config.max_shader_engines);
+
+	return rb_mask & (~(gc_disabled_rb_mask | gc_user_disabled_rb_mask));
+}
+
+static void gfx_v12_0_setup_rb(struct amdgpu_device *adev)
+{
+	u32 rb_bitmap_per_sa;
+	u32 rb_bitmap_width_per_sa;
+	u32 max_sa;
+	u32 active_sa_bitmap;
+	u32 global_active_rb_bitmap;
+	u32 active_rb_bitmap = 0;
+	u32 i;
+
+	/* query sa bitmap from SA_UNIT_DISABLE registers */
+	active_sa_bitmap = gfx_v12_0_get_sa_active_bitmap(adev);
+	/* query rb bitmap from RB_BACKEND_DISABLE registers */
+	global_active_rb_bitmap = gfx_v12_0_get_rb_active_bitmap(adev);
+
+	/* generate active rb bitmap according to active sa bitmap */
+	max_sa = adev->gfx.config.max_shader_engines *
+		 adev->gfx.config.max_sh_per_se;
+	rb_bitmap_width_per_sa = adev->gfx.config.max_backends_per_se /
+				 adev->gfx.config.max_sh_per_se;
+	rb_bitmap_per_sa = amdgpu_gfx_create_bitmask(rb_bitmap_width_per_sa);
+
+	for (i = 0; i < max_sa; i++) {
+		if (active_sa_bitmap & (1 << i))
+			active_rb_bitmap |= (rb_bitmap_per_sa << (i * rb_bitmap_width_per_sa));
+	}
+
+	active_rb_bitmap &= global_active_rb_bitmap;
+	adev->gfx.config.backend_enable_mask = active_rb_bitmap;
+	adev->gfx.config.num_rbs = hweight32(active_rb_bitmap);
+}
+
+#define LDS_APP_BASE           0x1
+#define SCRATCH_APP_BASE       0x2
+
+static void gfx_v12_0_init_compute_vmid(struct amdgpu_device *adev)
+{
+	int i;
+	uint32_t sh_mem_bases;
+	uint32_t data;
+
+	/*
+	 * Configure apertures:
+	 * LDS:         0x60000000'00000000 - 0x60000001'00000000 (4GB)
+	 * Scratch:     0x60000001'00000000 - 0x60000002'00000000 (4GB)
+	 * GPUVM:       0x60010000'00000000 - 0x60020000'00000000 (1TB)
+	 */
+	sh_mem_bases = (LDS_APP_BASE << SH_MEM_BASES__SHARED_BASE__SHIFT) |
+			SCRATCH_APP_BASE;
+
+	mutex_lock(&adev->srbm_mutex);
+	for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) {
+		soc24_grbm_select(adev, 0, 0, 0, i);
+		/* CP and shaders */
+		WREG32_SOC15(GC, 0, regSH_MEM_CONFIG, DEFAULT_SH_MEM_CONFIG);
+		WREG32_SOC15(GC, 0, regSH_MEM_BASES, sh_mem_bases);
+
+		/* Enable trap for each kfd vmid. */
+		data = RREG32_SOC15(GC, 0, regSPI_GDBG_PER_VMID_CNTL);
+		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+		WREG32_SOC15(GC, 0, regSPI_GDBG_PER_VMID_CNTL, data);
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+
+static void gfx_v12_0_tcp_harvest(struct amdgpu_device *adev)
+{
+	/* TODO: harvest feature to be added later. */
+}
+
+static void gfx_v12_0_get_tcc_info(struct amdgpu_device *adev)
+{
+}
+
+static void gfx_v12_0_constants_init(struct amdgpu_device *adev)
+{
+	u32 tmp;
+	int i;
+
+	if (!amdgpu_sriov_vf(adev))
+		WREG32_FIELD15_PREREG(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff);
+
+	gfx_v12_0_setup_rb(adev);
+	gfx_v12_0_get_cu_info(adev, &adev->gfx.cu_info);
+	gfx_v12_0_get_tcc_info(adev);
+	adev->gfx.config.pa_sc_tile_steering_override = 0;
+
+	/* XXX SH_MEM regs */
+	/* where to put LDS, scratch, GPUVM in FSA64 space */
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
+		soc24_grbm_select(adev, 0, 0, 0, i);
+		/* CP and shaders */
+		WREG32_SOC15(GC, 0, regSH_MEM_CONFIG, DEFAULT_SH_MEM_CONFIG);
+		if (i != 0) {
+			tmp = REG_SET_FIELD(0, SH_MEM_BASES, PRIVATE_BASE,
+				(adev->gmc.private_aperture_start >> 48));
+			tmp = REG_SET_FIELD(tmp, SH_MEM_BASES, SHARED_BASE,
+				(adev->gmc.shared_aperture_start >> 48));
+			WREG32_SOC15(GC, 0, regSH_MEM_BASES, tmp);
+		}
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+
+	mutex_unlock(&adev->srbm_mutex);
+
+	gfx_v12_0_init_compute_vmid(adev);
+}
+
+static u32 gfx_v12_0_get_cpg_int_cntl(struct amdgpu_device *adev,
+				      int me, int pipe)
+{
+	if (me != 0)
+		return 0;
+
+	switch (pipe) {
+	case 0:
+		return SOC15_REG_OFFSET(GC, 0, regCP_INT_CNTL_RING0);
+	default:
+		return 0;
+	}
+}
+
+static u32 gfx_v12_0_get_cpc_int_cntl(struct amdgpu_device *adev,
+				      int me, int pipe)
+{
+	/*
+	 * amdgpu controls only the first MEC. That's why this function only
+	 * handles the setting of interrupts for this specific MEC. All other
+	 * pipes' interrupts are set by amdkfd.
+	 */
+	if (me != 1)
+		return 0;
+
+	switch (pipe) {
+	case 0:
+		return SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE0_INT_CNTL);
+	case 1:
+		return SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE1_INT_CNTL);
+	default:
+		return 0;
+	}
+}
+
+static void gfx_v12_0_enable_gui_idle_interrupt(struct amdgpu_device *adev,
+					       bool enable)
+{
+	u32 tmp, cp_int_cntl_reg;
+	int i, j;
+
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			cp_int_cntl_reg = gfx_v12_0_get_cpg_int_cntl(adev, i, j);
+
+			if (cp_int_cntl_reg) {
+				tmp = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE,
+						    enable ? 1 : 0);
+				tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE,
+						    enable ? 1 : 0);
+				WREG32_SOC15_IP(GC, cp_int_cntl_reg, tmp);
+			}
+		}
+	}
+}
+
+static int gfx_v12_0_init_csb(struct amdgpu_device *adev)
+{
+	adev->gfx.rlc.funcs->get_csb_buffer(adev, adev->gfx.rlc.cs_ptr);
+
+	WREG32_SOC15(GC, 0, regRLC_CSIB_ADDR_HI,
+			adev->gfx.rlc.clear_state_gpu_addr >> 32);
+	WREG32_SOC15(GC, 0, regRLC_CSIB_ADDR_LO,
+			adev->gfx.rlc.clear_state_gpu_addr & 0xfffffffc);
+	WREG32_SOC15(GC, 0, regRLC_CSIB_LENGTH, adev->gfx.rlc.clear_state_size);
+
+	return 0;
+}
+
+static void gfx_v12_0_rlc_stop(struct amdgpu_device *adev)
+{
+	u32 tmp = RREG32_SOC15(GC, 0, regRLC_CNTL);
+
+	tmp = REG_SET_FIELD(tmp, RLC_CNTL, RLC_ENABLE_F32, 0);
+	WREG32_SOC15(GC, 0, regRLC_CNTL, tmp);
+}
+
+static void gfx_v12_0_rlc_reset(struct amdgpu_device *adev)
+{
+	WREG32_FIELD15_PREREG(GC, 0, GRBM_SOFT_RESET, SOFT_RESET_RLC, 1);
+	udelay(50);
+	WREG32_FIELD15_PREREG(GC, 0, GRBM_SOFT_RESET, SOFT_RESET_RLC, 0);
+	udelay(50);
+}
+
+static void gfx_v12_0_rlc_smu_handshake_cntl(struct amdgpu_device *adev,
+					     bool enable)
+{
+	uint32_t rlc_pg_cntl;
+
+	rlc_pg_cntl = RREG32_SOC15(GC, 0, regRLC_PG_CNTL);
+
+	if (!enable) {
+		/* RLC_PG_CNTL[23] = 0 (default)
+		 * RLC will wait for handshake acks with SMU
+		 * GFXOFF will be enabled
+		 * RLC_PG_CNTL[23] = 1
+		 * RLC will not issue any message to SMU
+		 * hence no handshake between SMU & RLC
+		 * GFXOFF will be disabled
+		 */
+		rlc_pg_cntl |= RLC_PG_CNTL__SMU_HANDSHAKE_DISABLE_MASK;
+	} else
+		rlc_pg_cntl &= ~RLC_PG_CNTL__SMU_HANDSHAKE_DISABLE_MASK;
+	WREG32_SOC15(GC, 0, regRLC_PG_CNTL, rlc_pg_cntl);
+}
+
+static void gfx_v12_0_rlc_start(struct amdgpu_device *adev)
+{
+	/* TODO: enable rlc & smu handshake until smu
+	 * and gfxoff feature works as expected */
+	if (!(amdgpu_pp_feature_mask & PP_GFXOFF_MASK))
+		gfx_v12_0_rlc_smu_handshake_cntl(adev, false);
+
+	WREG32_FIELD15_PREREG(GC, 0, RLC_CNTL, RLC_ENABLE_F32, 1);
+	udelay(50);
+}
+
+static void gfx_v12_0_rlc_enable_srm(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	/* enable Save Restore Machine */
+	tmp = RREG32(SOC15_REG_OFFSET(GC, 0, regRLC_SRM_CNTL));
+	tmp |= RLC_SRM_CNTL__AUTO_INCR_ADDR_MASK;
+	tmp |= RLC_SRM_CNTL__SRM_ENABLE_MASK;
+	WREG32(SOC15_REG_OFFSET(GC, 0, regRLC_SRM_CNTL), tmp);
+}
+
+static void gfx_v12_0_load_rlcg_microcode(struct amdgpu_device *adev)
+{
+	const struct rlc_firmware_header_v2_0 *hdr;
+	const __le32 *fw_data;
+	unsigned i, fw_size;
+
+	hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
+	fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+			   le32_to_cpu(hdr->header.ucode_array_offset_bytes));
+	fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4;
+
+	WREG32_SOC15(GC, 0, regRLC_GPM_UCODE_ADDR,
+		     RLCG_UCODE_LOADING_START_ADDRESS);
+
+	for (i = 0; i < fw_size; i++)
+		WREG32_SOC15(GC, 0, regRLC_GPM_UCODE_DATA,
+			     le32_to_cpup(fw_data++));
+
+	WREG32_SOC15(GC, 0, regRLC_GPM_UCODE_ADDR, adev->gfx.rlc_fw_version);
+}
+
+static void gfx_v12_0_load_rlc_iram_dram_microcode(struct amdgpu_device *adev)
+{
+	const struct rlc_firmware_header_v2_2 *hdr;
+	const __le32 *fw_data;
+	unsigned i, fw_size;
+	u32 tmp;
+
+	hdr = (const struct rlc_firmware_header_v2_2 *)adev->gfx.rlc_fw->data;
+
+	fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+			le32_to_cpu(hdr->rlc_iram_ucode_offset_bytes));
+	fw_size = le32_to_cpu(hdr->rlc_iram_ucode_size_bytes) / 4;
+
+	WREG32_SOC15(GC, 0, regRLC_LX6_IRAM_ADDR, 0);
+
+	for (i = 0; i < fw_size; i++) {
+		if ((amdgpu_emu_mode == 1) && (i % 100 == 99))
+			msleep(1);
+		WREG32_SOC15(GC, 0, regRLC_LX6_IRAM_DATA,
+				le32_to_cpup(fw_data++));
+	}
+
+	WREG32_SOC15(GC, 0, regRLC_LX6_IRAM_ADDR, adev->gfx.rlc_fw_version);
+
+	fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+			le32_to_cpu(hdr->rlc_dram_ucode_offset_bytes));
+	fw_size = le32_to_cpu(hdr->rlc_dram_ucode_size_bytes) / 4;
+
+	WREG32_SOC15(GC, 0, regRLC_LX6_DRAM_ADDR, 0);
+	for (i = 0; i < fw_size; i++) {
+		if ((amdgpu_emu_mode == 1) && (i % 100 == 99))
+			msleep(1);
+		WREG32_SOC15(GC, 0, regRLC_LX6_DRAM_DATA,
+				le32_to_cpup(fw_data++));
+	}
+
+	WREG32_SOC15(GC, 0, regRLC_LX6_IRAM_ADDR, adev->gfx.rlc_fw_version);
+
+	tmp = RREG32_SOC15(GC, 0, regRLC_LX6_CNTL);
+	tmp = REG_SET_FIELD(tmp, RLC_LX6_CNTL, PDEBUG_ENABLE, 1);
+	tmp = REG_SET_FIELD(tmp, RLC_LX6_CNTL, BRESET, 0);
+	WREG32_SOC15(GC, 0, regRLC_LX6_CNTL, tmp);
+}
+
+static int gfx_v12_0_rlc_load_microcode(struct amdgpu_device *adev)
+{
+	const struct rlc_firmware_header_v2_0 *hdr;
+	uint16_t version_major;
+	uint16_t version_minor;
+
+	if (!adev->gfx.rlc_fw)
+		return -EINVAL;
+
+	hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
+	amdgpu_ucode_print_rlc_hdr(&hdr->header);
+
+	version_major = le16_to_cpu(hdr->header.header_version_major);
+	version_minor = le16_to_cpu(hdr->header.header_version_minor);
+
+	if (version_major == 2) {
+		gfx_v12_0_load_rlcg_microcode(adev);
+		if (amdgpu_dpm == 1) {
+			if (version_minor >= 2)
+				gfx_v12_0_load_rlc_iram_dram_microcode(adev);
+		}
+
+		return 0;
+	}
+
+	return -EINVAL;
+}
+
+static int gfx_v12_0_rlc_resume(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		gfx_v12_0_init_csb(adev);
+
+		if (!amdgpu_sriov_vf(adev)) /* enable RLC SRM */
+			gfx_v12_0_rlc_enable_srm(adev);
+	} else {
+		if (amdgpu_sriov_vf(adev)) {
+			gfx_v12_0_init_csb(adev);
+			return 0;
+		}
+
+		adev->gfx.rlc.funcs->stop(adev);
+
+		/* disable CG */
+		WREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL, 0);
+
+		/* disable PG */
+		WREG32_SOC15(GC, 0, regRLC_PG_CNTL, 0);
+
+		if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
+			/* legacy rlc firmware loading */
+			r = gfx_v12_0_rlc_load_microcode(adev);
+			if (r)
+				return r;
+		}
+
+		gfx_v12_0_init_csb(adev);
+
+		adev->gfx.rlc.funcs->start(adev);
+	}
+
+	return 0;
+}
+
+static void gfx_v12_0_config_gfx_rs64(struct amdgpu_device *adev)
+{
+	const struct gfx_firmware_header_v2_0 *pfp_hdr;
+	const struct gfx_firmware_header_v2_0 *me_hdr;
+	const struct gfx_firmware_header_v2_0 *mec_hdr;
+	uint32_t pipe_id, tmp;
+
+	mec_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.mec_fw->data;
+	me_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.me_fw->data;
+	pfp_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.pfp_fw->data;
+
+	/* config pfp program start addr */
+	for (pipe_id = 0; pipe_id < 2; pipe_id++) {
+		soc24_grbm_select(adev, 0, pipe_id, 0, 0);
+		WREG32_SOC15(GC, 0, regCP_PFP_PRGRM_CNTR_START,
+			(pfp_hdr->ucode_start_addr_hi << 30) |
+			(pfp_hdr->ucode_start_addr_lo >> 2));
+		WREG32_SOC15(GC, 0, regCP_PFP_PRGRM_CNTR_START_HI,
+			pfp_hdr->ucode_start_addr_hi >> 2);
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+
+	/* reset pfp pipe */
+	tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE0_RESET, 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE1_RESET, 1);
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+
+	/* clear pfp pipe reset */
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE0_RESET, 0);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE1_RESET, 0);
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+
+	/* config me program start addr */
+	for (pipe_id = 0; pipe_id < 2; pipe_id++) {
+		soc24_grbm_select(adev, 0, pipe_id, 0, 0);
+		WREG32_SOC15(GC, 0, regCP_ME_PRGRM_CNTR_START,
+			(me_hdr->ucode_start_addr_hi << 30) |
+			(me_hdr->ucode_start_addr_lo >> 2));
+		WREG32_SOC15(GC, 0, regCP_ME_PRGRM_CNTR_START_HI,
+			me_hdr->ucode_start_addr_hi>>2);
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+
+	/* reset me pipe */
+	tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE0_RESET, 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE1_RESET, 1);
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+
+	/* clear me pipe reset */
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE0_RESET, 0);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE1_RESET, 0);
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+
+	/* config mec program start addr */
+	for (pipe_id = 0; pipe_id < 4; pipe_id++) {
+		soc24_grbm_select(adev, 1, pipe_id, 0, 0);
+		WREG32_SOC15(GC, 0, regCP_MEC_RS64_PRGRM_CNTR_START,
+					mec_hdr->ucode_start_addr_lo >> 2 |
+					mec_hdr->ucode_start_addr_hi << 30);
+		WREG32_SOC15(GC, 0, regCP_MEC_RS64_PRGRM_CNTR_START_HI,
+					mec_hdr->ucode_start_addr_hi >> 2);
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+
+	/* reset mec pipe */
+	tmp = RREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE0_RESET, 1);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE1_RESET, 1);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE2_RESET, 1);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE3_RESET, 1);
+	WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, tmp);
+
+	/* clear mec pipe reset */
+	tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE0_RESET, 0);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE1_RESET, 0);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE2_RESET, 0);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_RS64_CNTL, MEC_PIPE3_RESET, 0);
+	WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, tmp);
+}
+
+static void gfx_v12_0_set_pfp_ucode_start_addr(struct amdgpu_device *adev)
+{
+	const struct gfx_firmware_header_v2_0 *cp_hdr;
+	unsigned pipe_id, tmp;
+
+	cp_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.pfp_fw->data;
+	mutex_lock(&adev->srbm_mutex);
+	for (pipe_id = 0; pipe_id < adev->gfx.me.num_pipe_per_me; pipe_id++) {
+		soc24_grbm_select(adev, 0, pipe_id, 0, 0);
+		WREG32_SOC15(GC, 0, regCP_PFP_PRGRM_CNTR_START,
+			     (cp_hdr->ucode_start_addr_hi << 30) |
+			     (cp_hdr->ucode_start_addr_lo >> 2));
+		WREG32_SOC15(GC, 0, regCP_PFP_PRGRM_CNTR_START_HI,
+			     cp_hdr->ucode_start_addr_hi>>2);
+
+		/*
+		 * Program CP_ME_CNTL to reset given PIPE to take
+		 * effect of CP_PFP_PRGRM_CNTR_START.
+		 */
+		tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL);
+		if (pipe_id == 0)
+			tmp = REG_SET_FIELD(tmp, CP_ME_CNTL,
+					PFP_PIPE0_RESET, 1);
+		else
+			tmp = REG_SET_FIELD(tmp, CP_ME_CNTL,
+					PFP_PIPE1_RESET, 1);
+		WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+
+		/* Clear pfp pipe0 reset bit. */
+		if (pipe_id == 0)
+			tmp = REG_SET_FIELD(tmp, CP_ME_CNTL,
+					PFP_PIPE0_RESET, 0);
+		else
+			tmp = REG_SET_FIELD(tmp, CP_ME_CNTL,
+					PFP_PIPE1_RESET, 0);
+		WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+
+static void gfx_v12_0_set_me_ucode_start_addr(struct amdgpu_device *adev)
+{
+	const struct gfx_firmware_header_v2_0 *cp_hdr;
+	unsigned pipe_id, tmp;
+
+	cp_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.me_fw->data;
+	mutex_lock(&adev->srbm_mutex);
+	for (pipe_id = 0; pipe_id < adev->gfx.me.num_pipe_per_me; pipe_id++) {
+		soc24_grbm_select(adev, 0, pipe_id, 0, 0);
+		WREG32_SOC15(GC, 0, regCP_ME_PRGRM_CNTR_START,
+			     (cp_hdr->ucode_start_addr_hi << 30) |
+			     (cp_hdr->ucode_start_addr_lo >> 2) );
+		WREG32_SOC15(GC, 0, regCP_ME_PRGRM_CNTR_START_HI,
+			     cp_hdr->ucode_start_addr_hi>>2);
+
+		/*
+		 * Program CP_ME_CNTL to reset given PIPE to take
+		 * effect of CP_ME_PRGRM_CNTR_START.
+		 */
+		tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL);
+		if (pipe_id == 0)
+			tmp = REG_SET_FIELD(tmp, CP_ME_CNTL,
+					ME_PIPE0_RESET, 1);
+		else
+			tmp = REG_SET_FIELD(tmp, CP_ME_CNTL,
+					ME_PIPE1_RESET, 1);
+		WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+
+		/* Clear pfp pipe0 reset bit. */
+		if (pipe_id == 0)
+			tmp = REG_SET_FIELD(tmp, CP_ME_CNTL,
+					ME_PIPE0_RESET, 0);
+		else
+			tmp = REG_SET_FIELD(tmp, CP_ME_CNTL,
+					ME_PIPE1_RESET, 0);
+		WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+
+static void gfx_v12_0_set_mec_ucode_start_addr(struct amdgpu_device *adev)
+{
+	const struct gfx_firmware_header_v2_0 *cp_hdr;
+	unsigned pipe_id;
+
+	cp_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.mec_fw->data;
+	mutex_lock(&adev->srbm_mutex);
+	for (pipe_id = 0; pipe_id < adev->gfx.mec.num_pipe_per_mec; pipe_id++) {
+		soc24_grbm_select(adev, 1, pipe_id, 0, 0);
+		WREG32_SOC15(GC, 0, regCP_MEC_RS64_PRGRM_CNTR_START,
+			     cp_hdr->ucode_start_addr_lo >> 2 |
+			     cp_hdr->ucode_start_addr_hi << 30);
+		WREG32_SOC15(GC, 0, regCP_MEC_RS64_PRGRM_CNTR_START_HI,
+			     cp_hdr->ucode_start_addr_hi >> 2);
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+
+static int gfx_v12_0_wait_for_rlc_autoload_complete(struct amdgpu_device *adev)
+{
+	uint32_t cp_status;
+	uint32_t bootload_status;
+	int i;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		cp_status = RREG32_SOC15(GC, 0, regCP_STAT);
+		bootload_status = RREG32_SOC15(GC, 0, regRLC_RLCS_BOOTLOAD_STATUS);
+
+		if ((cp_status == 0) &&
+		    (REG_GET_FIELD(bootload_status,
+			RLC_RLCS_BOOTLOAD_STATUS, BOOTLOAD_COMPLETE) == 1)) {
+			break;
+		}
+		udelay(1);
+		if (amdgpu_emu_mode)
+			msleep(10);
+	}
+
+	if (i >= adev->usec_timeout) {
+		dev_err(adev->dev, "rlc autoload: gc ucode autoload timeout\n");
+		return -ETIMEDOUT;
+	}
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) {
+		gfx_v12_0_set_pfp_ucode_start_addr(adev);
+		gfx_v12_0_set_me_ucode_start_addr(adev);
+		gfx_v12_0_set_mec_ucode_start_addr(adev);
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable)
+{
+	int i;
+	u32 tmp = RREG32_SOC15(GC, 0, regCP_ME_CNTL);
+
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_HALT, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_HALT, enable ? 0 : 1);
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, tmp);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (RREG32_SOC15(GC, 0, regCP_STAT) == 0)
+			break;
+		udelay(1);
+	}
+
+	if (i >= adev->usec_timeout)
+		DRM_ERROR("failed to %s cp gfx\n", enable ? "unhalt" : "halt");
+
+	return 0;
+}
+
+static int gfx_v12_0_cp_gfx_load_pfp_microcode_rs64(struct amdgpu_device *adev)
+{
+	int r;
+	const struct gfx_firmware_header_v2_0 *pfp_hdr;
+	const __le32 *fw_ucode, *fw_data;
+	unsigned i, pipe_id, fw_ucode_size, fw_data_size;
+	uint32_t tmp;
+	uint32_t usec_timeout = 50000;  /* wait for 50ms */
+
+	pfp_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.pfp_fw->data;
+
+	amdgpu_ucode_print_gfx_hdr(&pfp_hdr->header);
+
+	/* instruction */
+	fw_ucode = (const __le32 *)(adev->gfx.pfp_fw->data +
+		le32_to_cpu(pfp_hdr->ucode_offset_bytes));
+	fw_ucode_size = le32_to_cpu(pfp_hdr->ucode_size_bytes);
+	/* data */
+	fw_data = (const __le32 *)(adev->gfx.pfp_fw->data +
+		le32_to_cpu(pfp_hdr->data_offset_bytes));
+	fw_data_size = le32_to_cpu(pfp_hdr->data_size_bytes);
+
+	/* 64kb align */
+	r = amdgpu_bo_create_reserved(adev, fw_ucode_size,
+				      64 * 1024, AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->gfx.pfp.pfp_fw_obj,
+				      &adev->gfx.pfp.pfp_fw_gpu_addr,
+				      (void **)&adev->gfx.pfp.pfp_fw_ptr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create pfp ucode fw bo\n", r);
+		gfx_v12_0_pfp_fini(adev);
+		return r;
+	}
+
+	r = amdgpu_bo_create_reserved(adev, fw_data_size,
+				      64 * 1024, AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->gfx.pfp.pfp_fw_data_obj,
+				      &adev->gfx.pfp.pfp_fw_data_gpu_addr,
+				      (void **)&adev->gfx.pfp.pfp_fw_data_ptr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create pfp data fw bo\n", r);
+		gfx_v12_0_pfp_fini(adev);
+		return r;
+	}
+
+	memcpy(adev->gfx.pfp.pfp_fw_ptr, fw_ucode, fw_ucode_size);
+	memcpy(adev->gfx.pfp.pfp_fw_data_ptr, fw_data, fw_data_size);
+
+	amdgpu_bo_kunmap(adev->gfx.pfp.pfp_fw_obj);
+	amdgpu_bo_kunmap(adev->gfx.pfp.pfp_fw_data_obj);
+	amdgpu_bo_unreserve(adev->gfx.pfp.pfp_fw_obj);
+	amdgpu_bo_unreserve(adev->gfx.pfp.pfp_fw_data_obj);
+
+	if (amdgpu_emu_mode == 1)
+		amdgpu_device_flush_hdp(adev, NULL);
+
+	WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_LO,
+		lower_32_bits(adev->gfx.pfp.pfp_fw_gpu_addr));
+	WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_HI,
+		upper_32_bits(adev->gfx.pfp.pfp_fw_gpu_addr));
+
+	tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, VMID, 0);
+	tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, CACHE_POLICY, 0);
+	tmp = REG_SET_FIELD(tmp, CP_PFP_IC_BASE_CNTL, EXE_DISABLE, 0);
+	WREG32_SOC15(GC, 0, regCP_PFP_IC_BASE_CNTL, tmp);
+
+	/*
+	 * Programming any of the CP_PFP_IC_BASE registers
+	 * forces invalidation of the ME L1 I$. Wait for the
+	 * invalidation complete
+	 */
+	for (i = 0; i < usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL);
+		if (1 == REG_GET_FIELD(tmp, CP_PFP_IC_OP_CNTL,
+			INVALIDATE_CACHE_COMPLETE))
+			break;
+		udelay(1);
+	}
+
+	if (i >= usec_timeout) {
+		dev_err(adev->dev, "failed to invalidate instruction cache\n");
+		return -EINVAL;
+	}
+
+	/* Prime the L1 instruction caches */
+	tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_PFP_IC_OP_CNTL, PRIME_ICACHE, 1);
+	WREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL, tmp);
+	/* Waiting for cache primed*/
+	for (i = 0; i < usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regCP_PFP_IC_OP_CNTL);
+		if (1 == REG_GET_FIELD(tmp, CP_PFP_IC_OP_CNTL,
+			ICACHE_PRIMED))
+			break;
+		udelay(1);
+	}
+
+	if (i >= usec_timeout) {
+		dev_err(adev->dev, "failed to prime instruction cache\n");
+		return -EINVAL;
+	}
+
+	mutex_lock(&adev->srbm_mutex);
+	for (pipe_id = 0; pipe_id < adev->gfx.me.num_pipe_per_me; pipe_id++) {
+		soc24_grbm_select(adev, 0, pipe_id, 0, 0);
+
+		WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE0_LO,
+			lower_32_bits(adev->gfx.pfp.pfp_fw_data_gpu_addr));
+		WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE0_HI,
+			upper_32_bits(adev->gfx.pfp.pfp_fw_data_gpu_addr));
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+
+	tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_BASE_CNTL, VMID, 0);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_BASE_CNTL, CACHE_POLICY, 0);
+	WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE_CNTL, tmp);
+
+	/* Invalidate the data caches */
+	tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_OP_CNTL, INVALIDATE_DCACHE, 1);
+	WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL, tmp);
+
+	for (i = 0; i < usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL);
+		if (1 == REG_GET_FIELD(tmp, CP_GFX_RS64_DC_OP_CNTL,
+			INVALIDATE_DCACHE_COMPLETE))
+			break;
+		udelay(1);
+	}
+
+	if (i >= usec_timeout) {
+		dev_err(adev->dev, "failed to invalidate RS64 data cache\n");
+		return -EINVAL;
+	}
+
+	gfx_v12_0_set_pfp_ucode_start_addr(adev);
+
+	return 0;
+}
+
+static int gfx_v12_0_cp_gfx_load_me_microcode_rs64(struct amdgpu_device *adev)
+{
+	int r;
+	const struct gfx_firmware_header_v2_0 *me_hdr;
+	const __le32 *fw_ucode, *fw_data;
+	unsigned i, pipe_id, fw_ucode_size, fw_data_size;
+	uint32_t tmp;
+	uint32_t usec_timeout = 50000;  /* wait for 50ms */
+
+	me_hdr = (const struct gfx_firmware_header_v2_0 *)
+		adev->gfx.me_fw->data;
+
+	amdgpu_ucode_print_gfx_hdr(&me_hdr->header);
+
+	/* instruction */
+	fw_ucode = (const __le32 *)(adev->gfx.me_fw->data +
+		le32_to_cpu(me_hdr->ucode_offset_bytes));
+	fw_ucode_size = le32_to_cpu(me_hdr->ucode_size_bytes);
+	/* data */
+	fw_data = (const __le32 *)(adev->gfx.me_fw->data +
+		le32_to_cpu(me_hdr->data_offset_bytes));
+	fw_data_size = le32_to_cpu(me_hdr->data_size_bytes);
+
+	/* 64kb align*/
+	r = amdgpu_bo_create_reserved(adev, fw_ucode_size,
+				      64 * 1024, AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->gfx.me.me_fw_obj,
+				      &adev->gfx.me.me_fw_gpu_addr,
+				      (void **)&adev->gfx.me.me_fw_ptr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create me ucode bo\n", r);
+		gfx_v12_0_me_fini(adev);
+		return r;
+	}
+
+	r = amdgpu_bo_create_reserved(adev, fw_data_size,
+				      64 * 1024, AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->gfx.me.me_fw_data_obj,
+				      &adev->gfx.me.me_fw_data_gpu_addr,
+				      (void **)&adev->gfx.me.me_fw_data_ptr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create me data bo\n", r);
+		gfx_v12_0_me_fini(adev);
+		return r;
+	}
+
+	memcpy(adev->gfx.me.me_fw_ptr, fw_ucode, fw_ucode_size);
+	memcpy(adev->gfx.me.me_fw_data_ptr, fw_data, fw_data_size);
+
+	amdgpu_bo_kunmap(adev->gfx.me.me_fw_obj);
+	amdgpu_bo_kunmap(adev->gfx.me.me_fw_data_obj);
+	amdgpu_bo_unreserve(adev->gfx.me.me_fw_obj);
+	amdgpu_bo_unreserve(adev->gfx.me.me_fw_data_obj);
+
+	if (amdgpu_emu_mode == 1)
+		amdgpu_device_flush_hdp(adev, NULL);
+
+	WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_LO,
+		lower_32_bits(adev->gfx.me.me_fw_gpu_addr));
+	WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_HI,
+		upper_32_bits(adev->gfx.me.me_fw_gpu_addr));
+
+	tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_BASE_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, VMID, 0);
+	tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, CACHE_POLICY, 0);
+	tmp = REG_SET_FIELD(tmp, CP_ME_IC_BASE_CNTL, EXE_DISABLE, 0);
+	WREG32_SOC15(GC, 0, regCP_ME_IC_BASE_CNTL, tmp);
+
+	/*
+	 * Programming any of the CP_ME_IC_BASE registers
+	 * forces invalidation of the ME L1 I$. Wait for the
+	 * invalidation complete
+	 */
+	for (i = 0; i < usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL);
+		if (1 == REG_GET_FIELD(tmp, CP_ME_IC_OP_CNTL,
+			INVALIDATE_CACHE_COMPLETE))
+			break;
+		udelay(1);
+	}
+
+	if (i >= usec_timeout) {
+		dev_err(adev->dev, "failed to invalidate instruction cache\n");
+		return -EINVAL;
+	}
+
+	/* Prime the instruction caches */
+	tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_ME_IC_OP_CNTL, PRIME_ICACHE, 1);
+	WREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL, tmp);
+
+	/* Waiting for instruction cache primed*/
+	for (i = 0; i < usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regCP_ME_IC_OP_CNTL);
+		if (1 == REG_GET_FIELD(tmp, CP_ME_IC_OP_CNTL,
+			ICACHE_PRIMED))
+			break;
+		udelay(1);
+	}
+
+	if (i >= usec_timeout) {
+		dev_err(adev->dev, "failed to prime instruction cache\n");
+		return -EINVAL;
+	}
+
+	mutex_lock(&adev->srbm_mutex);
+	for (pipe_id = 0; pipe_id < adev->gfx.me.num_pipe_per_me; pipe_id++) {
+		soc24_grbm_select(adev, 0, pipe_id, 0, 0);
+
+		WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE1_LO,
+			lower_32_bits(adev->gfx.me.me_fw_data_gpu_addr));
+		WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE1_HI,
+			upper_32_bits(adev->gfx.me.me_fw_data_gpu_addr));
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+
+	tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_BASE_CNTL, VMID, 0);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_BASE_CNTL, CACHE_POLICY, 0);
+	WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_BASE_CNTL, tmp);
+
+	/* Invalidate the data caches */
+	tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_RS64_DC_OP_CNTL, INVALIDATE_DCACHE, 1);
+	WREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL, tmp);
+
+	for (i = 0; i < usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regCP_GFX_RS64_DC_OP_CNTL);
+		if (1 == REG_GET_FIELD(tmp, CP_GFX_RS64_DC_OP_CNTL,
+			INVALIDATE_DCACHE_COMPLETE))
+			break;
+		udelay(1);
+	}
+
+	if (i >= usec_timeout) {
+		dev_err(adev->dev, "failed to invalidate RS64 data cache\n");
+		return -EINVAL;
+	}
+
+	gfx_v12_0_set_me_ucode_start_addr(adev);
+
+	return 0;
+}
+
+static int gfx_v12_0_cp_gfx_load_microcode(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (!adev->gfx.me_fw || !adev->gfx.pfp_fw)
+		return -EINVAL;
+
+	gfx_v12_0_cp_gfx_enable(adev, false);
+
+	r = gfx_v12_0_cp_gfx_load_pfp_microcode_rs64(adev);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to load pfp fw\n", r);
+		return r;
+	}
+
+	r = gfx_v12_0_cp_gfx_load_me_microcode_rs64(adev);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to load me fw\n", r);
+		return r;
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_cp_gfx_start(struct amdgpu_device *adev)
+{
+	/* init the CP */
+	WREG32_SOC15(GC, 0, regCP_MAX_CONTEXT,
+		     adev->gfx.config.max_hw_contexts - 1);
+	WREG32_SOC15(GC, 0, regCP_DEVICE_ID, 1);
+
+	if (!amdgpu_async_gfx_ring)
+		gfx_v12_0_cp_gfx_enable(adev, true);
+
+	return 0;
+}
+
+static void gfx_v12_0_cp_gfx_switch_pipe(struct amdgpu_device *adev,
+					 CP_PIPE_ID pipe)
+{
+	u32 tmp;
+
+	tmp = RREG32_SOC15(GC, 0, regGRBM_GFX_CNTL);
+	tmp = REG_SET_FIELD(tmp, GRBM_GFX_CNTL, PIPEID, pipe);
+
+	WREG32_SOC15(GC, 0, regGRBM_GFX_CNTL, tmp);
+}
+
+static void gfx_v12_0_cp_gfx_set_doorbell(struct amdgpu_device *adev,
+					  struct amdgpu_ring *ring)
+{
+	u32 tmp;
+
+	tmp = RREG32_SOC15(GC, 0, regCP_RB_DOORBELL_CONTROL);
+	if (ring->use_doorbell) {
+		tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL,
+				    DOORBELL_OFFSET, ring->doorbell_index);
+		tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL,
+				    DOORBELL_EN, 1);
+	} else {
+		tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL,
+				    DOORBELL_EN, 0);
+	}
+	WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_CONTROL, tmp);
+
+	tmp = REG_SET_FIELD(0, CP_RB_DOORBELL_RANGE_LOWER,
+			    DOORBELL_RANGE_LOWER, ring->doorbell_index);
+	WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_RANGE_LOWER, tmp);
+
+	WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_RANGE_UPPER,
+		     CP_RB_DOORBELL_RANGE_UPPER__DOORBELL_RANGE_UPPER_MASK);
+}
+
+static int gfx_v12_0_cp_gfx_resume(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	u32 tmp;
+	u32 rb_bufsz;
+	u64 rb_addr, rptr_addr, wptr_gpu_addr;
+
+	/* Set the write pointer delay */
+	WREG32_SOC15(GC, 0, regCP_RB_WPTR_DELAY, 0);
+
+	/* set the RB to use vmid 0 */
+	WREG32_SOC15(GC, 0, regCP_RB_VMID, 0);
+
+	/* Init gfx ring 0 for pipe 0 */
+	mutex_lock(&adev->srbm_mutex);
+	gfx_v12_0_cp_gfx_switch_pipe(adev, PIPE_ID0);
+
+	/* Set ring buffer size */
+	ring = &adev->gfx.gfx_ring[0];
+	rb_bufsz = order_base_2(ring->ring_size / 8);
+	tmp = REG_SET_FIELD(0, CP_RB0_CNTL, RB_BUFSZ, rb_bufsz);
+	tmp = REG_SET_FIELD(tmp, CP_RB0_CNTL, RB_BLKSZ, rb_bufsz - 2);
+	WREG32_SOC15(GC, 0, regCP_RB0_CNTL, tmp);
+
+	/* Initialize the ring buffer's write pointers */
+	ring->wptr = 0;
+	WREG32_SOC15(GC, 0, regCP_RB0_WPTR, lower_32_bits(ring->wptr));
+	WREG32_SOC15(GC, 0, regCP_RB0_WPTR_HI, upper_32_bits(ring->wptr));
+
+	/* set the wb address whether it's enabled or not */
+	rptr_addr = ring->rptr_gpu_addr;
+	WREG32_SOC15(GC, 0, regCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr));
+	WREG32_SOC15(GC, 0, regCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) &
+		     CP_RB_RPTR_ADDR_HI__RB_RPTR_ADDR_HI_MASK);
+
+	wptr_gpu_addr = ring->wptr_gpu_addr;
+	WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_ADDR_LO,
+		     lower_32_bits(wptr_gpu_addr));
+	WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_ADDR_HI,
+		     upper_32_bits(wptr_gpu_addr));
+
+	mdelay(1);
+	WREG32_SOC15(GC, 0, regCP_RB0_CNTL, tmp);
+
+	rb_addr = ring->gpu_addr >> 8;
+	WREG32_SOC15(GC, 0, regCP_RB0_BASE, rb_addr);
+	WREG32_SOC15(GC, 0, regCP_RB0_BASE_HI, upper_32_bits(rb_addr));
+
+	WREG32_SOC15(GC, 0, regCP_RB_ACTIVE, 1);
+
+	gfx_v12_0_cp_gfx_set_doorbell(adev, ring);
+	mutex_unlock(&adev->srbm_mutex);
+
+	/* Switch to pipe 0 */
+	mutex_lock(&adev->srbm_mutex);
+	gfx_v12_0_cp_gfx_switch_pipe(adev, PIPE_ID0);
+	mutex_unlock(&adev->srbm_mutex);
+
+	/* start the ring */
+	gfx_v12_0_cp_gfx_start(adev);
+	return 0;
+}
+
+static void gfx_v12_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
+{
+	u32 data;
+
+	data = RREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_INVALIDATE_ICACHE,
+						 enable ? 0 : 1);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE0_RESET,
+						 enable ? 0 : 1);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE1_RESET,
+						 enable ? 0 : 1);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE2_RESET,
+						 enable ? 0 : 1);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE3_RESET,
+						 enable ? 0 : 1);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE0_ACTIVE,
+						 enable ? 1 : 0);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE1_ACTIVE,
+			                         enable ? 1 : 0);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE2_ACTIVE,
+						 enable ? 1 : 0);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_PIPE3_ACTIVE,
+						 enable ? 1 : 0);
+	data = REG_SET_FIELD(data, CP_MEC_RS64_CNTL, MEC_HALT,
+						 enable ? 0 : 1);
+	WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, data);
+
+	adev->gfx.kiq[0].ring.sched.ready = enable;
+
+	udelay(50);
+}
+
+static int gfx_v12_0_cp_compute_load_microcode_rs64(struct amdgpu_device *adev)
+{
+	const struct gfx_firmware_header_v2_0 *mec_hdr;
+	const __le32 *fw_ucode, *fw_data;
+	u32 tmp, fw_ucode_size, fw_data_size;
+	u32 i, usec_timeout = 50000; /* Wait for 50 ms */
+	u32 *fw_ucode_ptr, *fw_data_ptr;
+	int r;
+
+	if (!adev->gfx.mec_fw)
+		return -EINVAL;
+
+	gfx_v12_0_cp_compute_enable(adev, false);
+
+	mec_hdr = (const struct gfx_firmware_header_v2_0 *)adev->gfx.mec_fw->data;
+	amdgpu_ucode_print_gfx_hdr(&mec_hdr->header);
+
+	fw_ucode = (const __le32 *) (adev->gfx.mec_fw->data +
+				le32_to_cpu(mec_hdr->ucode_offset_bytes));
+	fw_ucode_size = le32_to_cpu(mec_hdr->ucode_size_bytes);
+
+	fw_data = (const __le32 *) (adev->gfx.mec_fw->data +
+				le32_to_cpu(mec_hdr->data_offset_bytes));
+	fw_data_size = le32_to_cpu(mec_hdr->data_size_bytes);
+
+	r = amdgpu_bo_create_reserved(adev, fw_ucode_size,
+				      64 * 1024, AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->gfx.mec.mec_fw_obj,
+				      &adev->gfx.mec.mec_fw_gpu_addr,
+				      (void **)&fw_ucode_ptr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create mec fw ucode bo\n", r);
+		gfx_v12_0_mec_fini(adev);
+		return r;
+	}
+
+	r = amdgpu_bo_create_reserved(adev,
+				      ALIGN(fw_data_size, 64 * 1024) *
+				      adev->gfx.mec.num_pipe_per_mec,
+				      64 * 1024, AMDGPU_GEM_DOMAIN_VRAM,
+				      &adev->gfx.mec.mec_fw_data_obj,
+				      &adev->gfx.mec.mec_fw_data_gpu_addr,
+				      (void **)&fw_data_ptr);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to create mec fw ucode bo\n", r);
+		gfx_v12_0_mec_fini(adev);
+		return r;
+	}
+
+	memcpy(fw_ucode_ptr, fw_ucode, fw_ucode_size);
+	for (i = 0; i < adev->gfx.mec.num_pipe_per_mec; i++) {
+		memcpy(fw_data_ptr + i * ALIGN(fw_data_size, 64 * 1024) / 4, fw_data, fw_data_size);
+	}
+
+	amdgpu_bo_kunmap(adev->gfx.mec.mec_fw_obj);
+	amdgpu_bo_kunmap(adev->gfx.mec.mec_fw_data_obj);
+	amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_obj);
+	amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_data_obj);
+
+	tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, VMID, 0);
+	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, EXE_DISABLE, 0);
+	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, CACHE_POLICY, 0);
+	WREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_CNTL, tmp);
+
+	tmp = RREG32_SOC15(GC, 0, regCP_MEC_DC_BASE_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_DC_BASE_CNTL, VMID, 0);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_DC_BASE_CNTL, CACHE_POLICY, 0);
+	WREG32_SOC15(GC, 0, regCP_MEC_DC_BASE_CNTL, tmp);
+
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->gfx.mec.num_pipe_per_mec; i++) {
+		soc24_grbm_select(adev, 1, i, 0, 0);
+
+		WREG32_SOC15(GC, 0, regCP_MEC_MDBASE_LO,
+			     lower_32_bits(adev->gfx.mec.mec_fw_data_gpu_addr +
+					   i * ALIGN(fw_data_size, 64 * 1024)));
+		WREG32_SOC15(GC, 0, regCP_MEC_MDBASE_HI,
+			     upper_32_bits(adev->gfx.mec.mec_fw_data_gpu_addr +
+					   i * ALIGN(fw_data_size, 64 * 1024)));
+
+		WREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_LO,
+			     lower_32_bits(adev->gfx.mec.mec_fw_gpu_addr));
+		WREG32_SOC15(GC, 0, regCP_CPC_IC_BASE_HI,
+			     upper_32_bits(adev->gfx.mec.mec_fw_gpu_addr));
+	}
+	mutex_unlock(&adev->srbm_mutex);
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+
+	/* Trigger an invalidation of the L1 instruction caches */
+	tmp = RREG32_SOC15(GC, 0, regCP_MEC_DC_OP_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_MEC_DC_OP_CNTL, INVALIDATE_DCACHE, 1);
+	WREG32_SOC15(GC, 0, regCP_MEC_DC_OP_CNTL, tmp);
+
+	/* Wait for invalidation complete */
+	for (i = 0; i < usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regCP_MEC_DC_OP_CNTL);
+		if (1 == REG_GET_FIELD(tmp, CP_MEC_DC_OP_CNTL,
+				       INVALIDATE_DCACHE_COMPLETE))
+			break;
+		udelay(1);
+	}
+
+	if (i >= usec_timeout) {
+		dev_err(adev->dev, "failed to invalidate instruction cache\n");
+		return -EINVAL;
+	}
+
+	/* Trigger an invalidation of the L1 instruction caches */
+	tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_OP_CNTL);
+	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_OP_CNTL, INVALIDATE_CACHE, 1);
+	WREG32_SOC15(GC, 0, regCP_CPC_IC_OP_CNTL, tmp);
+
+	/* Wait for invalidation complete */
+	for (i = 0; i < usec_timeout; i++) {
+		tmp = RREG32_SOC15(GC, 0, regCP_CPC_IC_OP_CNTL);
+		if (1 == REG_GET_FIELD(tmp, CP_CPC_IC_OP_CNTL,
+				       INVALIDATE_CACHE_COMPLETE))
+			break;
+		udelay(1);
+	}
+
+	if (i >= usec_timeout) {
+		dev_err(adev->dev, "failed to invalidate instruction cache\n");
+		return -EINVAL;
+	}
+
+	gfx_v12_0_set_mec_ucode_start_addr(adev);
+
+	return 0;
+}
+
+static void gfx_v12_0_kiq_setting(struct amdgpu_ring *ring)
+{
+	uint32_t tmp;
+	struct amdgpu_device *adev = ring->adev;
+
+	/* tell RLC which is KIQ queue */
+	tmp = RREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS);
+	tmp &= 0xffffff00;
+	tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
+	WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp | 0x80);
+}
+
+static void gfx_v12_0_cp_set_doorbell_range(struct amdgpu_device *adev)
+{
+	/* set graphics engine doorbell range */
+	WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_RANGE_LOWER,
+		     (adev->doorbell_index.gfx_ring0 * 2) << 2);
+	WREG32_SOC15(GC, 0, regCP_RB_DOORBELL_RANGE_UPPER,
+		     (adev->doorbell_index.gfx_userqueue_end * 2) << 2);
+
+	/* set compute engine doorbell range */
+	WREG32_SOC15(GC, 0, regCP_MEC_DOORBELL_RANGE_LOWER,
+		     (adev->doorbell_index.kiq * 2) << 2);
+	WREG32_SOC15(GC, 0, regCP_MEC_DOORBELL_RANGE_UPPER,
+		     (adev->doorbell_index.userqueue_end * 2) << 2);
+}
+
+static int gfx_v12_0_gfx_mqd_init(struct amdgpu_device *adev, void *m,
+				  struct amdgpu_mqd_prop *prop)
+{
+	struct v12_gfx_mqd *mqd = m;
+	uint64_t hqd_gpu_addr, wb_gpu_addr;
+	uint32_t tmp;
+	uint32_t rb_bufsz;
+
+	/* set up gfx hqd wptr */
+	mqd->cp_gfx_hqd_wptr = 0;
+	mqd->cp_gfx_hqd_wptr_hi = 0;
+
+	/* set the pointer to the MQD */
+	mqd->cp_mqd_base_addr = prop->mqd_gpu_addr & 0xfffffffc;
+	mqd->cp_mqd_base_addr_hi = upper_32_bits(prop->mqd_gpu_addr);
+
+	/* set up mqd control */
+	tmp = regCP_GFX_MQD_CONTROL_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, VMID, 0);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, PRIV_STATE, 1);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_MQD_CONTROL, CACHE_POLICY, 0);
+	mqd->cp_gfx_mqd_control = tmp;
+
+	/* set up gfx_hqd_vimd with 0x0 to indicate the ring buffer's vmid */
+	tmp = regCP_GFX_HQD_VMID_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_VMID, VMID, 0);
+	mqd->cp_gfx_hqd_vmid = 0;
+
+	/* set up default queue priority level
+	 * 0x0 = low priority, 0x1 = high priority */
+	tmp = regCP_GFX_HQD_QUEUE_PRIORITY_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUEUE_PRIORITY, PRIORITY_LEVEL, 0);
+	mqd->cp_gfx_hqd_queue_priority = tmp;
+
+	/* set up time quantum */
+	tmp = regCP_GFX_HQD_QUANTUM_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_QUANTUM, QUANTUM_EN, 1);
+	mqd->cp_gfx_hqd_quantum = tmp;
+
+	/* set up gfx hqd base. this is similar as CP_RB_BASE */
+	hqd_gpu_addr = prop->hqd_base_gpu_addr >> 8;
+	mqd->cp_gfx_hqd_base = hqd_gpu_addr;
+	mqd->cp_gfx_hqd_base_hi = upper_32_bits(hqd_gpu_addr);
+
+	/* set up hqd_rptr_addr/_hi, similar as CP_RB_RPTR */
+	wb_gpu_addr = prop->rptr_gpu_addr;
+	mqd->cp_gfx_hqd_rptr_addr = wb_gpu_addr & 0xfffffffc;
+	mqd->cp_gfx_hqd_rptr_addr_hi =
+		upper_32_bits(wb_gpu_addr) & 0xffff;
+
+	/* set up rb_wptr_poll addr */
+	wb_gpu_addr = prop->wptr_gpu_addr;
+	mqd->cp_rb_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc;
+	mqd->cp_rb_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff;
+
+	/* set up the gfx_hqd_control, similar as CP_RB0_CNTL */
+	rb_bufsz = order_base_2(prop->queue_size / 4) - 1;
+	tmp = regCP_GFX_HQD_CNTL_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_BUFSZ, rb_bufsz);
+	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_BLKSZ, rb_bufsz - 2);
+#ifdef __BIG_ENDIAN
+	tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, BUF_SWAP, 1);
+#endif
+	if (prop->tmz_queue)
+		tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, TMZ_MATCH, 1);
+	if (!prop->kernel_queue)
+		tmp = REG_SET_FIELD(tmp, CP_GFX_HQD_CNTL, RB_NON_PRIV, 1);
+	mqd->cp_gfx_hqd_cntl = tmp;
+
+	/* set up cp_doorbell_control */
+	tmp = regCP_RB_DOORBELL_CONTROL_DEFAULT;
+	if (prop->use_doorbell) {
+		tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL,
+				    DOORBELL_OFFSET, prop->doorbell_index);
+		tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL,
+				    DOORBELL_EN, 1);
+	} else
+		tmp = REG_SET_FIELD(tmp, CP_RB_DOORBELL_CONTROL,
+				    DOORBELL_EN, 0);
+	mqd->cp_rb_doorbell_control = tmp;
+
+	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
+	mqd->cp_gfx_hqd_rptr = regCP_GFX_HQD_RPTR_DEFAULT;
+
+	/* active the queue */
+	mqd->cp_gfx_hqd_active = 1;
+
+	/* set gfx UQ items */
+	mqd->shadow_base_lo = lower_32_bits(prop->shadow_addr);
+	mqd->shadow_base_hi = upper_32_bits(prop->shadow_addr);
+	mqd->fw_work_area_base_lo = lower_32_bits(prop->csa_addr);
+	mqd->fw_work_area_base_hi = upper_32_bits(prop->csa_addr);
+	mqd->fence_address_lo = lower_32_bits(prop->fence_address);
+	mqd->fence_address_hi = upper_32_bits(prop->fence_address);
+
+	return 0;
+}
+
+static int gfx_v12_0_kgq_init_queue(struct amdgpu_ring *ring, bool reset)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct v12_gfx_mqd *mqd = ring->mqd_ptr;
+	int mqd_idx = ring - &adev->gfx.gfx_ring[0];
+
+	if (!reset && !amdgpu_in_reset(adev) && !adev->in_suspend) {
+		memset((void *)mqd, 0, sizeof(*mqd));
+		mutex_lock(&adev->srbm_mutex);
+		soc24_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		amdgpu_ring_init_mqd(ring);
+		soc24_grbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+		if (adev->gfx.me.mqd_backup[mqd_idx])
+			memcpy_fromio(adev->gfx.me.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+	} else {
+		/* restore mqd with the backup copy */
+		if (adev->gfx.me.mqd_backup[mqd_idx])
+			memcpy_toio(mqd, adev->gfx.me.mqd_backup[mqd_idx], sizeof(*mqd));
+		/* reset the ring */
+		ring->wptr = 0;
+		*ring->wptr_cpu_addr = 0;
+		amdgpu_ring_clear_ring(ring);
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_cp_async_gfx_ring_resume(struct amdgpu_device *adev)
+{
+	int i, r;
+
+	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+		r = gfx_v12_0_kgq_init_queue(&adev->gfx.gfx_ring[i], false);
+		if (r)
+			return r;
+	}
+
+	r = amdgpu_gfx_enable_kgq(adev, 0);
+	if (r)
+		return r;
+
+	return gfx_v12_0_cp_gfx_start(adev);
+}
+
+static int gfx_v12_0_compute_mqd_init(struct amdgpu_device *adev, void *m,
+				      struct amdgpu_mqd_prop *prop)
+{
+	struct v12_compute_mqd *mqd = m;
+	uint64_t hqd_gpu_addr, wb_gpu_addr, eop_base_addr;
+	uint32_t tmp;
+
+	mqd->header = 0xC0310800;
+	mqd->compute_pipelinestat_enable = 0x00000001;
+	mqd->compute_static_thread_mgmt_se0 = 0xffffffff;
+	mqd->compute_static_thread_mgmt_se1 = 0xffffffff;
+	mqd->compute_static_thread_mgmt_se2 = 0xffffffff;
+	mqd->compute_static_thread_mgmt_se3 = 0xffffffff;
+	mqd->compute_misc_reserved = 0x00000007;
+
+	eop_base_addr = prop->eop_gpu_addr >> 8;
+	mqd->cp_hqd_eop_base_addr_lo = eop_base_addr;
+	mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr);
+
+	/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
+	tmp = regCP_HQD_EOP_CONTROL_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_HQD_EOP_CONTROL, EOP_SIZE,
+			(order_base_2(GFX12_MEC_HPD_SIZE / 4) - 1));
+
+	mqd->cp_hqd_eop_control = tmp;
+
+	/* enable doorbell? */
+	tmp = regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT;
+
+	if (prop->use_doorbell) {
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_OFFSET, prop->doorbell_index);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_EN, 1);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_SOURCE, 0);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_HIT, 0);
+	} else {
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_EN, 0);
+	}
+
+	mqd->cp_hqd_pq_doorbell_control = tmp;
+
+	/* disable the queue if it's active */
+	mqd->cp_hqd_dequeue_request = 0;
+	mqd->cp_hqd_pq_rptr = 0;
+	mqd->cp_hqd_pq_wptr_lo = 0;
+	mqd->cp_hqd_pq_wptr_hi = 0;
+
+	/* set the pointer to the MQD */
+	mqd->cp_mqd_base_addr_lo = prop->mqd_gpu_addr & 0xfffffffc;
+	mqd->cp_mqd_base_addr_hi = upper_32_bits(prop->mqd_gpu_addr);
+
+	/* set MQD vmid to 0 */
+	tmp = regCP_MQD_CONTROL_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_MQD_CONTROL, VMID, 0);
+	mqd->cp_mqd_control = tmp;
+
+	/* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */
+	hqd_gpu_addr = prop->hqd_base_gpu_addr >> 8;
+	mqd->cp_hqd_pq_base_lo = hqd_gpu_addr;
+	mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr);
+
+	/* set up the HQD, this is similar to CP_RB0_CNTL */
+	tmp = regCP_HQD_PQ_CONTROL_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE,
+			    (order_base_2(prop->queue_size / 4) - 1));
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE,
+			    (order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1));
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 1);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TUNNEL_DISPATCH, 0);
+	if (prop->kernel_queue) {
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1);
+	}
+	if (prop->tmz_queue)
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TMZ, 1);
+	mqd->cp_hqd_pq_control = tmp;
+
+	/* set the wb address whether it's enabled or not */
+	wb_gpu_addr = prop->rptr_gpu_addr;
+	mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc;
+	mqd->cp_hqd_pq_rptr_report_addr_hi =
+		upper_32_bits(wb_gpu_addr) & 0xffff;
+
+	/* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */
+	wb_gpu_addr = prop->wptr_gpu_addr;
+	mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc;
+	mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff;
+
+	tmp = 0;
+	/* enable the doorbell if requested */
+	if (prop->use_doorbell) {
+		tmp = regCP_HQD_PQ_DOORBELL_CONTROL_DEFAULT;
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				DOORBELL_OFFSET, prop->doorbell_index);
+
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_EN, 1);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_SOURCE, 0);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_HIT, 0);
+	}
+
+	mqd->cp_hqd_pq_doorbell_control = tmp;
+
+	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
+	mqd->cp_hqd_pq_rptr = regCP_HQD_PQ_RPTR_DEFAULT;
+
+	/* set the vmid for the queue */
+	mqd->cp_hqd_vmid = 0;
+
+	tmp = regCP_HQD_PERSISTENT_STATE_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PERSISTENT_STATE, PRELOAD_SIZE, 0x55);
+	mqd->cp_hqd_persistent_state = tmp;
+
+	/* set MIN_IB_AVAIL_SIZE */
+	tmp = regCP_HQD_IB_CONTROL_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, CP_HQD_IB_CONTROL, MIN_IB_AVAIL_SIZE, 3);
+	mqd->cp_hqd_ib_control = tmp;
+
+	/* set static priority for a compute queue/ring */
+	mqd->cp_hqd_pipe_priority = prop->hqd_pipe_priority;
+	mqd->cp_hqd_queue_priority = prop->hqd_queue_priority;
+
+	mqd->cp_hqd_active = prop->hqd_active;
+
+	/* set UQ fenceaddress */
+	mqd->fence_address_lo = lower_32_bits(prop->fence_address);
+	mqd->fence_address_hi = upper_32_bits(prop->fence_address);
+
+	return 0;
+}
+
+static int gfx_v12_0_kiq_init_register(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct v12_compute_mqd *mqd = ring->mqd_ptr;
+	int j;
+
+	/* inactivate the queue */
+	if (amdgpu_sriov_vf(adev))
+		WREG32_SOC15(GC, 0, regCP_HQD_ACTIVE, 0);
+
+	/* disable wptr polling */
+	WREG32_FIELD15_PREREG(GC, 0, CP_PQ_WPTR_POLL_CNTL, EN, 0);
+
+	/* write the EOP addr */
+	WREG32_SOC15(GC, 0, regCP_HQD_EOP_BASE_ADDR,
+	       mqd->cp_hqd_eop_base_addr_lo);
+	WREG32_SOC15(GC, 0, regCP_HQD_EOP_BASE_ADDR_HI,
+	       mqd->cp_hqd_eop_base_addr_hi);
+
+	/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
+	WREG32_SOC15(GC, 0, regCP_HQD_EOP_CONTROL,
+	       mqd->cp_hqd_eop_control);
+
+	/* enable doorbell? */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL,
+	       mqd->cp_hqd_pq_doorbell_control);
+
+	/* disable the queue if it's active */
+	if (RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) & 1) {
+		WREG32_SOC15(GC, 0, regCP_HQD_DEQUEUE_REQUEST, 1);
+		for (j = 0; j < adev->usec_timeout; j++) {
+			if (!(RREG32_SOC15(GC, 0, regCP_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+		WREG32_SOC15(GC, 0, regCP_HQD_DEQUEUE_REQUEST,
+		       mqd->cp_hqd_dequeue_request);
+		WREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR,
+		       mqd->cp_hqd_pq_rptr);
+		WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_LO,
+		       mqd->cp_hqd_pq_wptr_lo);
+		WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_HI,
+		       mqd->cp_hqd_pq_wptr_hi);
+	}
+
+	/* set the pointer to the MQD */
+	WREG32_SOC15(GC, 0, regCP_MQD_BASE_ADDR,
+	       mqd->cp_mqd_base_addr_lo);
+	WREG32_SOC15(GC, 0, regCP_MQD_BASE_ADDR_HI,
+	       mqd->cp_mqd_base_addr_hi);
+
+	/* set MQD vmid to 0 */
+	WREG32_SOC15(GC, 0, regCP_MQD_CONTROL,
+	       mqd->cp_mqd_control);
+
+	/* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_BASE,
+	       mqd->cp_hqd_pq_base_lo);
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_BASE_HI,
+	       mqd->cp_hqd_pq_base_hi);
+
+	/* set up the HQD, this is similar to CP_RB0_CNTL */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_CONTROL,
+	       mqd->cp_hqd_pq_control);
+
+	/* set the wb address whether it's enabled or not */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR_REPORT_ADDR,
+		mqd->cp_hqd_pq_rptr_report_addr_lo);
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_RPTR_REPORT_ADDR_HI,
+		mqd->cp_hqd_pq_rptr_report_addr_hi);
+
+	/* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR,
+	       mqd->cp_hqd_pq_wptr_poll_addr_lo);
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR_HI,
+	       mqd->cp_hqd_pq_wptr_poll_addr_hi);
+
+	/* enable the doorbell if requested */
+	if (ring->use_doorbell) {
+		WREG32_SOC15(GC, 0, regCP_MEC_DOORBELL_RANGE_LOWER,
+			(adev->doorbell_index.kiq * 2) << 2);
+		WREG32_SOC15(GC, 0, regCP_MEC_DOORBELL_RANGE_UPPER,
+			(adev->doorbell_index.userqueue_end * 2) << 2);
+	}
+
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL,
+	       mqd->cp_hqd_pq_doorbell_control);
+
+	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_LO,
+	       mqd->cp_hqd_pq_wptr_lo);
+	WREG32_SOC15(GC, 0, regCP_HQD_PQ_WPTR_HI,
+	       mqd->cp_hqd_pq_wptr_hi);
+
+	/* set the vmid for the queue */
+	WREG32_SOC15(GC, 0, regCP_HQD_VMID, mqd->cp_hqd_vmid);
+
+	WREG32_SOC15(GC, 0, regCP_HQD_PERSISTENT_STATE,
+	       mqd->cp_hqd_persistent_state);
+
+	/* activate the queue */
+	WREG32_SOC15(GC, 0, regCP_HQD_ACTIVE,
+	       mqd->cp_hqd_active);
+
+	if (ring->use_doorbell)
+		WREG32_FIELD15_PREREG(GC, 0, CP_PQ_STATUS, DOORBELL_ENABLE, 1);
+
+	return 0;
+}
+
+static int gfx_v12_0_kiq_init_queue(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct v12_compute_mqd *mqd = ring->mqd_ptr;
+	int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS;
+
+	gfx_v12_0_kiq_setting(ring);
+
+	if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
+		/* reset MQD to a clean status */
+		if (adev->gfx.mec.mqd_backup[mqd_idx])
+			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
+
+		/* reset ring buffer */
+		ring->wptr = 0;
+		amdgpu_ring_clear_ring(ring);
+
+		mutex_lock(&adev->srbm_mutex);
+		soc24_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		gfx_v12_0_kiq_init_register(ring);
+		soc24_grbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+	} else {
+		memset((void *)mqd, 0, sizeof(*mqd));
+		if (amdgpu_sriov_vf(adev) && adev->in_suspend)
+			amdgpu_ring_clear_ring(ring);
+		mutex_lock(&adev->srbm_mutex);
+		soc24_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		amdgpu_ring_init_mqd(ring);
+		gfx_v12_0_kiq_init_register(ring);
+		soc24_grbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+
+		if (adev->gfx.mec.mqd_backup[mqd_idx])
+			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_kcq_init_queue(struct amdgpu_ring *ring, bool reset)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct v12_compute_mqd *mqd = ring->mqd_ptr;
+	int mqd_idx = ring - &adev->gfx.compute_ring[0];
+
+	if (!reset && !amdgpu_in_reset(adev) && !adev->in_suspend) {
+		memset((void *)mqd, 0, sizeof(*mqd));
+		mutex_lock(&adev->srbm_mutex);
+		soc24_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		amdgpu_ring_init_mqd(ring);
+		soc24_grbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+
+		if (adev->gfx.mec.mqd_backup[mqd_idx])
+			memcpy_fromio(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd));
+	} else {
+		/* restore MQD to a clean status */
+		if (adev->gfx.mec.mqd_backup[mqd_idx])
+			memcpy_toio(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd));
+		/* reset ring buffer */
+		ring->wptr = 0;
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, 0);
+		amdgpu_ring_clear_ring(ring);
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_kiq_resume(struct amdgpu_device *adev)
+{
+	gfx_v12_0_kiq_init_queue(&adev->gfx.kiq[0].ring);
+	adev->gfx.kiq[0].ring.sched.ready = true;
+	return 0;
+}
+
+static int gfx_v12_0_kcq_resume(struct amdgpu_device *adev)
+{
+	int i, r;
+
+	if (!amdgpu_async_gfx_ring)
+		gfx_v12_0_cp_compute_enable(adev, true);
+
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		r = gfx_v12_0_kcq_init_queue(&adev->gfx.compute_ring[i], false);
+		if (r)
+			return r;
+	}
+
+	return amdgpu_gfx_enable_kcq(adev, 0);
+}
+
+static int gfx_v12_0_cp_resume(struct amdgpu_device *adev)
+{
+	int r, i;
+	struct amdgpu_ring *ring;
+
+	if (!(adev->flags & AMD_IS_APU))
+		gfx_v12_0_enable_gui_idle_interrupt(adev, false);
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
+		/* legacy firmware loading */
+		r = gfx_v12_0_cp_gfx_load_microcode(adev);
+		if (r)
+			return r;
+
+		r = gfx_v12_0_cp_compute_load_microcode_rs64(adev);
+		if (r)
+			return r;
+	}
+
+	gfx_v12_0_cp_set_doorbell_range(adev);
+
+	if (amdgpu_async_gfx_ring) {
+		gfx_v12_0_cp_compute_enable(adev, true);
+		gfx_v12_0_cp_gfx_enable(adev, true);
+	}
+
+	if (adev->enable_mes_kiq && adev->mes.kiq_hw_init)
+		r = amdgpu_mes_kiq_hw_init(adev);
+	else
+		r = gfx_v12_0_kiq_resume(adev);
+	if (r)
+		return r;
+
+	r = gfx_v12_0_kcq_resume(adev);
+	if (r)
+		return r;
+
+	if (!amdgpu_async_gfx_ring) {
+		r = gfx_v12_0_cp_gfx_resume(adev);
+		if (r)
+			return r;
+	} else {
+		r = gfx_v12_0_cp_async_gfx_ring_resume(adev);
+		if (r)
+			return r;
+	}
+
+	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+		ring = &adev->gfx.gfx_ring[i];
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
+	}
+
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		ring = &adev->gfx.compute_ring[i];
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static void gfx_v12_0_cp_enable(struct amdgpu_device *adev, bool enable)
+{
+	gfx_v12_0_cp_gfx_enable(adev, enable);
+	gfx_v12_0_cp_compute_enable(adev, enable);
+}
+
+static int gfx_v12_0_gfxhub_enable(struct amdgpu_device *adev)
+{
+	int r;
+	bool value;
+
+	r = adev->gfxhub.funcs->gart_enable(adev);
+	if (r)
+		return r;
+
+	amdgpu_device_flush_hdp(adev, NULL);
+
+	value = amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS;
+
+	adev->gfxhub.funcs->set_fault_enable_default(adev, value);
+	/* TODO investigate why this and the hdp flush above is needed,
+	 * are we missing a flush somewhere else? */
+	adev->gmc.gmc_funcs->flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB(0), 0);
+
+	return 0;
+}
+
+static int get_gb_addr_config(struct amdgpu_device *adev)
+{
+	u32 gb_addr_config;
+
+	gb_addr_config = RREG32_SOC15(GC, 0, regGB_ADDR_CONFIG);
+	if (gb_addr_config == 0)
+		return -EINVAL;
+
+	adev->gfx.config.gb_addr_config_fields.num_pkrs =
+		1 << REG_GET_FIELD(gb_addr_config, GB_ADDR_CONFIG, NUM_PKRS);
+
+	adev->gfx.config.gb_addr_config = gb_addr_config;
+
+	adev->gfx.config.gb_addr_config_fields.num_pipes = 1 <<
+			REG_GET_FIELD(adev->gfx.config.gb_addr_config,
+				      GB_ADDR_CONFIG, NUM_PIPES);
+
+	adev->gfx.config.max_tile_pipes =
+		adev->gfx.config.gb_addr_config_fields.num_pipes;
+
+	adev->gfx.config.gb_addr_config_fields.max_compress_frags = 1 <<
+			REG_GET_FIELD(adev->gfx.config.gb_addr_config,
+				      GB_ADDR_CONFIG, MAX_COMPRESSED_FRAGS);
+	adev->gfx.config.gb_addr_config_fields.num_rb_per_se = 1 <<
+			REG_GET_FIELD(adev->gfx.config.gb_addr_config,
+				      GB_ADDR_CONFIG, NUM_RB_PER_SE);
+	adev->gfx.config.gb_addr_config_fields.num_se = 1 <<
+			REG_GET_FIELD(adev->gfx.config.gb_addr_config,
+				      GB_ADDR_CONFIG, NUM_SHADER_ENGINES);
+	adev->gfx.config.gb_addr_config_fields.pipe_interleave_size = 1 << (8 +
+			REG_GET_FIELD(adev->gfx.config.gb_addr_config,
+				      GB_ADDR_CONFIG, PIPE_INTERLEAVE_SIZE));
+
+	return 0;
+}
+
+static void gfx_v12_0_disable_gpa_mode(struct amdgpu_device *adev)
+{
+	uint32_t data;
+
+	data = RREG32_SOC15(GC, 0, regCPC_PSP_DEBUG);
+	data |= CPC_PSP_DEBUG__GPA_OVERRIDE_MASK;
+	WREG32_SOC15(GC, 0, regCPC_PSP_DEBUG, data);
+
+	data = RREG32_SOC15(GC, 0, regCPG_PSP_DEBUG);
+	data |= CPG_PSP_DEBUG__GPA_OVERRIDE_MASK;
+	WREG32_SOC15(GC, 0, regCPG_PSP_DEBUG, data);
+}
+
+static void gfx_v12_0_init_golden_registers(struct amdgpu_device *adev)
+{
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		soc15_program_register_sequence(adev,
+						golden_settings_gc_12_0,
+						(const u32)ARRAY_SIZE(golden_settings_gc_12_0));
+
+		if (adev->rev_id == 0)
+			soc15_program_register_sequence(adev,
+					golden_settings_gc_12_0_rev0,
+					(const u32)ARRAY_SIZE(golden_settings_gc_12_0_rev0));
+		break;
+	default:
+		break;
+	}
+}
+
+static int gfx_v12_0_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) {
+		if (adev->gfx.imu.funcs && (amdgpu_dpm > 0)) {
+			/* RLC autoload sequence 1: Program rlc ram */
+			if (adev->gfx.imu.funcs->program_rlc_ram)
+				adev->gfx.imu.funcs->program_rlc_ram(adev);
+		}
+		/* rlc autoload firmware */
+		r = gfx_v12_0_rlc_backdoor_autoload_enable(adev);
+		if (r)
+			return r;
+	} else {
+		if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) {
+			if (adev->gfx.imu.funcs && (amdgpu_dpm > 0)) {
+				if (adev->gfx.imu.funcs->load_microcode)
+					adev->gfx.imu.funcs->load_microcode(adev);
+				if (adev->gfx.imu.funcs->setup_imu)
+					adev->gfx.imu.funcs->setup_imu(adev);
+				if (adev->gfx.imu.funcs->start_imu)
+					adev->gfx.imu.funcs->start_imu(adev);
+			}
+
+			/* disable gpa mode in backdoor loading */
+			gfx_v12_0_disable_gpa_mode(adev);
+		}
+	}
+
+	if ((adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) ||
+	    (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP)) {
+		r = gfx_v12_0_wait_for_rlc_autoload_complete(adev);
+		if (r) {
+			dev_err(adev->dev, "(%d) failed to wait rlc autoload complete\n", r);
+			return r;
+		}
+	}
+
+	if (!amdgpu_emu_mode)
+		gfx_v12_0_init_golden_registers(adev);
+
+	adev->gfx.is_poweron = true;
+
+	if (get_gb_addr_config(adev))
+		DRM_WARN("Invalid gb_addr_config !\n");
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP)
+		gfx_v12_0_config_gfx_rs64(adev);
+
+	r = gfx_v12_0_gfxhub_enable(adev);
+	if (r)
+		return r;
+
+	if ((adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT ||
+	     adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO) &&
+	     (amdgpu_dpm == 1)) {
+		/**
+		 * For gfx 12, rlc firmware loading relies on smu firmware is
+		 * loaded firstly, so in direct type, it has to load smc ucode
+		 * here before rlc.
+		 */
+		r = amdgpu_pm_load_smu_firmware(adev, NULL);
+		if (r)
+			return r;
+	}
+
+	gfx_v12_0_constants_init(adev);
+
+	if (adev->nbio.funcs->gc_doorbell_init)
+		adev->nbio.funcs->gc_doorbell_init(adev);
+
+	r = gfx_v12_0_rlc_resume(adev);
+	if (r)
+		return r;
+
+	/*
+	 * init golden registers and rlc resume may override some registers,
+	 * reconfig them here
+	 */
+	gfx_v12_0_tcp_harvest(adev);
+
+	r = gfx_v12_0_cp_resume(adev);
+	if (r)
+		return r;
+
+	return r;
+}
+
+static int gfx_v12_0_set_userq_eop_interrupts(struct amdgpu_device *adev,
+					      bool enable)
+{
+	unsigned int irq_type;
+	int m, p, r;
+
+	if (adev->userq_funcs[AMDGPU_HW_IP_GFX]) {
+		for (m = 0; m < adev->gfx.me.num_me; m++) {
+			for (p = 0; p < adev->gfx.me.num_pipe_per_me; p++) {
+				irq_type = AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP + p;
+				if (enable)
+					r = amdgpu_irq_get(adev, &adev->gfx.eop_irq,
+							   irq_type);
+				else
+					r = amdgpu_irq_put(adev, &adev->gfx.eop_irq,
+							   irq_type);
+				if (r)
+					return r;
+			}
+		}
+	}
+
+	if (adev->userq_funcs[AMDGPU_HW_IP_COMPUTE]) {
+		for (m = 0; m < adev->gfx.mec.num_mec; ++m) {
+			for (p = 0; p < adev->gfx.mec.num_pipe_per_mec; p++) {
+				irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
+					+ (m * adev->gfx.mec.num_pipe_per_mec)
+					+ p;
+				if (enable)
+					r = amdgpu_irq_get(adev, &adev->gfx.eop_irq,
+							   irq_type);
+				else
+					r = amdgpu_irq_put(adev, &adev->gfx.eop_irq,
+							   irq_type);
+				if (r)
+					return r;
+			}
+		}
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t tmp;
+
+	cancel_delayed_work_sync(&adev->gfx.idle_work);
+
+	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
+	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
+	amdgpu_irq_put(adev, &adev->gfx.bad_op_irq, 0);
+	gfx_v12_0_set_userq_eop_interrupts(adev, false);
+
+	if (!adev->no_hw_access) {
+		if (amdgpu_async_gfx_ring) {
+			if (amdgpu_gfx_disable_kgq(adev, 0))
+				DRM_ERROR("KGQ disable failed\n");
+		}
+
+		if (amdgpu_gfx_disable_kcq(adev, 0))
+			DRM_ERROR("KCQ disable failed\n");
+
+		amdgpu_mes_kiq_hw_fini(adev);
+	}
+
+	if (amdgpu_sriov_vf(adev)) {
+		gfx_v12_0_cp_gfx_enable(adev, false);
+		/* Program KIQ position of RLC_CP_SCHEDULERS during destroy */
+		tmp = RREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS);
+		tmp &= 0xffffff00;
+		WREG32_SOC15(GC, 0, regRLC_CP_SCHEDULERS, tmp);
+
+		return 0;
+	}
+	gfx_v12_0_cp_enable(adev, false);
+	gfx_v12_0_enable_gui_idle_interrupt(adev, false);
+
+	adev->gfxhub.funcs->gart_disable(adev);
+
+	adev->gfx.is_poweron = false;
+
+	return 0;
+}
+
+static int gfx_v12_0_suspend(struct amdgpu_ip_block *ip_block)
+{
+	return gfx_v12_0_hw_fini(ip_block);
+}
+
+static int gfx_v12_0_resume(struct amdgpu_ip_block *ip_block)
+{
+	return gfx_v12_0_hw_init(ip_block);
+}
+
+static bool gfx_v12_0_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (REG_GET_FIELD(RREG32_SOC15(GC, 0, regGRBM_STATUS),
+				GRBM_STATUS, GUI_ACTIVE))
+		return false;
+	else
+		return true;
+}
+
+static int gfx_v12_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	unsigned i;
+	u32 tmp;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		/* read MC_STATUS */
+		tmp = RREG32_SOC15(GC, 0, regGRBM_STATUS) &
+			GRBM_STATUS__GUI_ACTIVE_MASK;
+
+		if (!REG_GET_FIELD(tmp, GRBM_STATUS, GUI_ACTIVE))
+			return 0;
+		udelay(1);
+	}
+	return -ETIMEDOUT;
+}
+
+static uint64_t gfx_v12_0_get_gpu_clock_counter(struct amdgpu_device *adev)
+{
+	uint64_t clock = 0;
+
+	if (adev->smuio.funcs &&
+	    adev->smuio.funcs->get_gpu_clock_counter)
+		clock = adev->smuio.funcs->get_gpu_clock_counter(adev);
+	else
+		dev_warn(adev->dev, "query gpu clock counter is not supported\n");
+
+	return clock;
+}
+
+static int gfx_v12_0_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	switch (amdgpu_user_queue) {
+	case -1:
+	case 0:
+	default:
+		adev->gfx.disable_kq = false;
+		adev->gfx.disable_uq = true;
+		break;
+	case 1:
+		adev->gfx.disable_kq = false;
+		adev->gfx.disable_uq = false;
+		break;
+	case 2:
+		adev->gfx.disable_kq = true;
+		adev->gfx.disable_uq = false;
+		break;
+	}
+
+	adev->gfx.funcs = &gfx_v12_0_gfx_funcs;
+
+	if (adev->gfx.disable_kq) {
+		adev->gfx.num_gfx_rings = 0;
+		adev->gfx.num_compute_rings = 0;
+	} else {
+		adev->gfx.num_gfx_rings = GFX12_NUM_GFX_RINGS;
+		adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
+						  AMDGPU_MAX_COMPUTE_RINGS);
+	}
+
+	gfx_v12_0_set_kiq_pm4_funcs(adev);
+	gfx_v12_0_set_ring_funcs(adev);
+	gfx_v12_0_set_irq_funcs(adev);
+	gfx_v12_0_set_rlc_funcs(adev);
+	gfx_v12_0_set_mqd_funcs(adev);
+	gfx_v12_0_set_imu_funcs(adev);
+
+	gfx_v12_0_init_rlcg_reg_access_ctrl(adev);
+
+	return gfx_v12_0_init_microcode(adev);
+}
+
+static int gfx_v12_0_late_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0);
+	if (r)
+		return r;
+
+	r = amdgpu_irq_get(adev, &adev->gfx.priv_inst_irq, 0);
+	if (r)
+		return r;
+
+	r = amdgpu_irq_get(adev, &adev->gfx.bad_op_irq, 0);
+	if (r)
+		return r;
+
+	r = gfx_v12_0_set_userq_eop_interrupts(adev, true);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static bool gfx_v12_0_is_rlc_enabled(struct amdgpu_device *adev)
+{
+	uint32_t rlc_cntl;
+
+	/* if RLC is not enabled, do nothing */
+	rlc_cntl = RREG32_SOC15(GC, 0, regRLC_CNTL);
+	return (REG_GET_FIELD(rlc_cntl, RLC_CNTL, RLC_ENABLE_F32)) ? true : false;
+}
+
+static void gfx_v12_0_set_safe_mode(struct amdgpu_device *adev,
+				    int xcc_id)
+{
+	uint32_t data;
+	unsigned i;
+
+	data = RLC_SAFE_MODE__CMD_MASK;
+	data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT);
+
+	WREG32_SOC15(GC, 0, regRLC_SAFE_MODE, data);
+
+	/* wait for RLC_SAFE_MODE */
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (!REG_GET_FIELD(RREG32_SOC15(GC, 0, regRLC_SAFE_MODE),
+				   RLC_SAFE_MODE, CMD))
+			break;
+		udelay(1);
+	}
+}
+
+static void gfx_v12_0_unset_safe_mode(struct amdgpu_device *adev,
+				      int xcc_id)
+{
+	WREG32_SOC15(GC, 0, regRLC_SAFE_MODE, RLC_SAFE_MODE__CMD_MASK);
+}
+
+static void gfx_v12_0_update_perf_clk(struct amdgpu_device *adev,
+				      bool enable)
+{
+	uint32_t def, data;
+
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_GFX_PERF_CLK))
+		return;
+
+	def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE);
+
+	if (enable)
+		data &= ~RLC_CGTT_MGCG_OVERRIDE__PERFMON_CLOCK_STATE_MASK;
+	else
+		data |= RLC_CGTT_MGCG_OVERRIDE__PERFMON_CLOCK_STATE_MASK;
+
+	if (def != data)
+		WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data);
+}
+
+static void gfx_v12_0_update_spm_vmid(struct amdgpu_device *adev,
+				      struct amdgpu_ring *ring,
+				      unsigned vmid)
+{
+	u32 reg, data;
+
+	reg = SOC15_REG_OFFSET(GC, 0, regRLC_SPM_MC_CNTL);
+	if (amdgpu_sriov_is_pp_one_vf(adev))
+		data = RREG32_NO_KIQ(reg);
+	else
+		data = RREG32(reg);
+
+	data &= ~RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK;
+	data |= (vmid & RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK) << RLC_SPM_MC_CNTL__RLC_SPM_VMID__SHIFT;
+
+	if (amdgpu_sriov_is_pp_one_vf(adev))
+		WREG32_SOC15_NO_KIQ(GC, 0, regRLC_SPM_MC_CNTL, data);
+	else
+		WREG32_SOC15(GC, 0, regRLC_SPM_MC_CNTL, data);
+
+	if (ring
+	    && amdgpu_sriov_is_pp_one_vf(adev)
+	    && ((ring->funcs->type == AMDGPU_RING_TYPE_GFX)
+		|| (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE))) {
+		uint32_t reg = SOC15_REG_OFFSET(GC, 0, regRLC_SPM_MC_CNTL);
+		amdgpu_ring_emit_wreg(ring, reg, data);
+	}
+}
+
+static const struct amdgpu_rlc_funcs gfx_v12_0_rlc_funcs = {
+	.is_rlc_enabled = gfx_v12_0_is_rlc_enabled,
+	.set_safe_mode = gfx_v12_0_set_safe_mode,
+	.unset_safe_mode = gfx_v12_0_unset_safe_mode,
+	.init = gfx_v12_0_rlc_init,
+	.get_csb_size = gfx_v12_0_get_csb_size,
+	.get_csb_buffer = gfx_v12_0_get_csb_buffer,
+	.resume = gfx_v12_0_rlc_resume,
+	.stop = gfx_v12_0_rlc_stop,
+	.reset = gfx_v12_0_rlc_reset,
+	.start = gfx_v12_0_rlc_start,
+	.update_spm_vmid = gfx_v12_0_update_spm_vmid,
+};
+
+#if 0
+static void gfx_v12_cntl_power_gating(struct amdgpu_device *adev, bool enable)
+{
+	/* TODO */
+}
+
+static void gfx_v12_cntl_pg(struct amdgpu_device *adev, bool enable)
+{
+	/* TODO */
+}
+#endif
+
+static int gfx_v12_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					   enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = (state == AMD_PG_STATE_GATE);
+
+	if (amdgpu_sriov_vf(adev))
+		return 0;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		amdgpu_gfx_off_ctrl(adev, enable);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static void gfx_v12_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev,
+						       bool enable)
+{
+	uint32_t def, data;
+
+	if (!(adev->cg_flags &
+	      (AMD_CG_SUPPORT_GFX_CGCG |
+	      AMD_CG_SUPPORT_GFX_CGLS |
+	      AMD_CG_SUPPORT_GFX_3D_CGCG |
+	      AMD_CG_SUPPORT_GFX_3D_CGLS)))
+		return;
+
+	if (enable) {
+		def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE);
+
+		/* unset CGCG override */
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)
+			data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_CGCG_OVERRIDE_MASK;
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS)
+			data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_CGLS_OVERRIDE_MASK;
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGCG ||
+		    adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGLS)
+			data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_GFX3D_CG_OVERRIDE_MASK;
+
+		/* update CGCG override bits */
+		if (def != data)
+			WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data);
+
+		/* enable cgcg FSM(0x0000363F) */
+		def = data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL);
+
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG) {
+			data &= ~RLC_CGCG_CGLS_CTRL__CGCG_GFX_IDLE_THRESHOLD_MASK;
+			data |= (0x36 << RLC_CGCG_CGLS_CTRL__CGCG_GFX_IDLE_THRESHOLD__SHIFT) |
+				 RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK;
+		}
+
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS) {
+			data &= ~RLC_CGCG_CGLS_CTRL__CGLS_REP_COMPANSAT_DELAY_MASK;
+			data |= (0x000F << RLC_CGCG_CGLS_CTRL__CGLS_REP_COMPANSAT_DELAY__SHIFT) |
+				 RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK;
+		}
+
+		if (def != data)
+			WREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL, data);
+
+		/* Program RLC_CGCG_CGLS_CTRL_3D */
+		def = data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL_3D);
+
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGCG) {
+			data &= ~RLC_CGCG_CGLS_CTRL_3D__CGCG_GFX_IDLE_THRESHOLD_MASK;
+			data |= (0x36 << RLC_CGCG_CGLS_CTRL_3D__CGCG_GFX_IDLE_THRESHOLD__SHIFT) |
+				 RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK;
+		}
+
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGLS) {
+			data &= ~RLC_CGCG_CGLS_CTRL_3D__CGLS_REP_COMPANSAT_DELAY_MASK;
+			data |= (0xf << RLC_CGCG_CGLS_CTRL_3D__CGLS_REP_COMPANSAT_DELAY__SHIFT) |
+				 RLC_CGCG_CGLS_CTRL_3D__CGLS_EN_MASK;
+		}
+
+		if (def != data)
+			WREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL_3D, data);
+
+		/* set IDLE_POLL_COUNT(0x00900100) */
+		def = data = RREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_CNTL);
+
+		data &= ~(CP_RB_WPTR_POLL_CNTL__POLL_FREQUENCY_MASK | CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT_MASK);
+		data |= (0x0100 << CP_RB_WPTR_POLL_CNTL__POLL_FREQUENCY__SHIFT) |
+			(0x0090 << CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT__SHIFT);
+
+		if (def != data)
+			WREG32_SOC15(GC, 0, regCP_RB_WPTR_POLL_CNTL, data);
+
+		data = RREG32_SOC15(GC, 0, regCP_INT_CNTL);
+		data = REG_SET_FIELD(data, CP_INT_CNTL, CNTX_BUSY_INT_ENABLE, 1);
+		data = REG_SET_FIELD(data, CP_INT_CNTL, CNTX_EMPTY_INT_ENABLE, 1);
+		data = REG_SET_FIELD(data, CP_INT_CNTL, CMP_BUSY_INT_ENABLE, 1);
+		data = REG_SET_FIELD(data, CP_INT_CNTL, GFX_IDLE_INT_ENABLE, 1);
+		WREG32_SOC15(GC, 0, regCP_INT_CNTL, data);
+
+		data = RREG32_SOC15(GC, 0, regSDMA0_RLC_CGCG_CTRL);
+		data = REG_SET_FIELD(data, SDMA0_RLC_CGCG_CTRL, CGCG_INT_ENABLE, 1);
+		WREG32_SOC15(GC, 0, regSDMA0_RLC_CGCG_CTRL, data);
+
+		/* Some ASICs only have one SDMA instance, not need to configure SDMA1 */
+		if (adev->sdma.num_instances > 1) {
+			data = RREG32_SOC15(GC, 0, regSDMA1_RLC_CGCG_CTRL);
+			data = REG_SET_FIELD(data, SDMA1_RLC_CGCG_CTRL, CGCG_INT_ENABLE, 1);
+			WREG32_SOC15(GC, 0, regSDMA1_RLC_CGCG_CTRL, data);
+		}
+	} else {
+		/* Program RLC_CGCG_CGLS_CTRL */
+		def = data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL);
+
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)
+			data &= ~RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK;
+
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS)
+			data &= ~RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK;
+
+		if (def != data)
+			WREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL, data);
+
+		/* Program RLC_CGCG_CGLS_CTRL_3D */
+		def = data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL_3D);
+
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGCG)
+			data &= ~RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK;
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGLS)
+			data &= ~RLC_CGCG_CGLS_CTRL_3D__CGLS_EN_MASK;
+
+		if (def != data)
+			WREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL_3D, data);
+	}
+}
+
+static void gfx_v12_0_update_medium_grain_clock_gating(struct amdgpu_device *adev,
+						       bool enable)
+{
+	uint32_t data, def;
+	if (!(adev->cg_flags & (AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS)))
+		return;
+
+	/* It is disabled by HW by default */
+	if (enable) {
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG) {
+			/* 1 - RLC_CGTT_MGCG_OVERRIDE */
+			def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE);
+
+			data &= ~(RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK |
+				  RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK |
+				  RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK);
+
+			if (def != data)
+				WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data);
+		}
+	} else {
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG) {
+			def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE);
+
+			data |= (RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK |
+				 RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK |
+				 RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK);
+
+			if (def != data)
+				WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data);
+		}
+	}
+}
+
+static void gfx_v12_0_update_repeater_fgcg(struct amdgpu_device *adev,
+					   bool enable)
+{
+	uint32_t def, data;
+
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_REPEATER_FGCG))
+		return;
+
+	def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE);
+
+	if (enable)
+		data &= ~(RLC_CGTT_MGCG_OVERRIDE__GFXIP_REPEATER_FGCG_OVERRIDE_MASK |
+				  RLC_CGTT_MGCG_OVERRIDE__RLC_REPEATER_FGCG_OVERRIDE_MASK);
+	else
+		data |= RLC_CGTT_MGCG_OVERRIDE__GFXIP_REPEATER_FGCG_OVERRIDE_MASK |
+				RLC_CGTT_MGCG_OVERRIDE__RLC_REPEATER_FGCG_OVERRIDE_MASK;
+
+	if (def != data)
+		WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data);
+}
+
+static void gfx_v12_0_update_sram_fgcg(struct amdgpu_device *adev,
+				       bool enable)
+{
+	uint32_t def, data;
+
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_GFX_FGCG))
+		return;
+
+	def = data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE);
+
+	if (enable)
+		data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_FGCG_OVERRIDE_MASK;
+	else
+		data |= RLC_CGTT_MGCG_OVERRIDE__GFXIP_FGCG_OVERRIDE_MASK;
+
+	if (def != data)
+		WREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE, data);
+}
+
+static int gfx_v12_0_update_gfx_clock_gating(struct amdgpu_device *adev,
+					    bool enable)
+{
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+
+	gfx_v12_0_update_coarse_grain_clock_gating(adev, enable);
+
+	gfx_v12_0_update_medium_grain_clock_gating(adev, enable);
+
+	gfx_v12_0_update_repeater_fgcg(adev, enable);
+
+	gfx_v12_0_update_sram_fgcg(adev, enable);
+
+	gfx_v12_0_update_perf_clk(adev, enable);
+
+	if (adev->cg_flags &
+	    (AMD_CG_SUPPORT_GFX_MGCG |
+	     AMD_CG_SUPPORT_GFX_CGLS |
+	     AMD_CG_SUPPORT_GFX_CGCG |
+	     AMD_CG_SUPPORT_GFX_3D_CGCG |
+	     AMD_CG_SUPPORT_GFX_3D_CGLS))
+		gfx_v12_0_enable_gui_idle_interrupt(adev, enable);
+
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
+
+	return 0;
+}
+
+static int gfx_v12_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					   enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (amdgpu_sriov_vf(adev))
+		return 0;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		gfx_v12_0_update_gfx_clock_gating(adev,
+						  state == AMD_CG_STATE_GATE);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static void gfx_v12_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int data;
+
+	/* AMD_CG_SUPPORT_GFX_MGCG */
+	data = RREG32_SOC15(GC, 0, regRLC_CGTT_MGCG_OVERRIDE);
+	if (!(data & RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK))
+		*flags |= AMD_CG_SUPPORT_GFX_MGCG;
+
+	/* AMD_CG_SUPPORT_REPEATER_FGCG */
+	if (!(data & RLC_CGTT_MGCG_OVERRIDE__GFXIP_REPEATER_FGCG_OVERRIDE_MASK))
+		*flags |= AMD_CG_SUPPORT_REPEATER_FGCG;
+
+	/* AMD_CG_SUPPORT_GFX_FGCG */
+	if (!(data & RLC_CGTT_MGCG_OVERRIDE__GFXIP_FGCG_OVERRIDE_MASK))
+		*flags |= AMD_CG_SUPPORT_GFX_FGCG;
+
+	/* AMD_CG_SUPPORT_GFX_PERF_CLK */
+	if (!(data & RLC_CGTT_MGCG_OVERRIDE__PERFMON_CLOCK_STATE_MASK))
+		*flags |= AMD_CG_SUPPORT_GFX_PERF_CLK;
+
+	/* AMD_CG_SUPPORT_GFX_CGCG */
+	data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL);
+	if (data & RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_GFX_CGCG;
+
+	/* AMD_CG_SUPPORT_GFX_CGLS */
+	if (data & RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_GFX_CGLS;
+
+	/* AMD_CG_SUPPORT_GFX_3D_CGCG */
+	data = RREG32_SOC15(GC, 0, regRLC_CGCG_CGLS_CTRL_3D);
+	if (data & RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_GFX_3D_CGCG;
+
+	/* AMD_CG_SUPPORT_GFX_3D_CGLS */
+	if (data & RLC_CGCG_CGLS_CTRL_3D__CGLS_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_GFX_3D_CGLS;
+}
+
+static u64 gfx_v12_0_ring_get_rptr_gfx(struct amdgpu_ring *ring)
+{
+	/* gfx12 is 32bit rptr*/
+	return *(uint32_t *)ring->rptr_cpu_addr;
+}
+
+static u64 gfx_v12_0_ring_get_wptr_gfx(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u64 wptr;
+
+	/* XXX check if swapping is necessary on BE */
+	if (ring->use_doorbell) {
+		wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr);
+	} else {
+		wptr = RREG32_SOC15(GC, 0, regCP_RB0_WPTR);
+		wptr += (u64)RREG32_SOC15(GC, 0, regCP_RB0_WPTR_HI) << 32;
+	}
+
+	return wptr;
+}
+
+static void gfx_v12_0_ring_set_wptr_gfx(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->use_doorbell) {
+		/* XXX check if swapping is necessary on BE */
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
+			     ring->wptr);
+		WDOORBELL64(ring->doorbell_index, ring->wptr);
+	} else {
+		WREG32_SOC15(GC, 0, regCP_RB0_WPTR,
+			     lower_32_bits(ring->wptr));
+		WREG32_SOC15(GC, 0, regCP_RB0_WPTR_HI,
+			     upper_32_bits(ring->wptr));
+	}
+}
+
+static u64 gfx_v12_0_ring_get_rptr_compute(struct amdgpu_ring *ring)
+{
+	/* gfx12 hardware is 32bit rptr */
+	return *(uint32_t *)ring->rptr_cpu_addr;
+}
+
+static u64 gfx_v12_0_ring_get_wptr_compute(struct amdgpu_ring *ring)
+{
+	u64 wptr;
+
+	/* XXX check if swapping is necessary on BE */
+	if (ring->use_doorbell)
+		wptr = atomic64_read((atomic64_t *)ring->wptr_cpu_addr);
+	else
+		BUG();
+	return wptr;
+}
+
+static void gfx_v12_0_ring_set_wptr_compute(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	/* XXX check if swapping is necessary on BE */
+	if (ring->use_doorbell) {
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr,
+			     ring->wptr);
+		WDOORBELL64(ring->doorbell_index, ring->wptr);
+	} else {
+		BUG(); /* only DOORBELL method supported on gfx12 now */
+	}
+}
+
+static void gfx_v12_0_ring_emit_hdp_flush(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 ref_and_mask, reg_mem_engine;
+	const struct nbio_hdp_flush_reg *nbio_hf_reg = adev->nbio.hdp_flush_reg;
+
+	if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) {
+		switch (ring->me) {
+		case 1:
+			ref_and_mask = nbio_hf_reg->ref_and_mask_cp2 << ring->pipe;
+			break;
+		case 2:
+			ref_and_mask = nbio_hf_reg->ref_and_mask_cp6 << ring->pipe;
+			break;
+		default:
+			return;
+		}
+		reg_mem_engine = 0;
+	} else {
+		ref_and_mask = nbio_hf_reg->ref_and_mask_cp0;
+		reg_mem_engine = 1; /* pfp */
+	}
+
+	gfx_v12_0_wait_reg_mem(ring, reg_mem_engine, 0, 1,
+			       adev->nbio.funcs->get_hdp_flush_req_offset(adev),
+			       adev->nbio.funcs->get_hdp_flush_done_offset(adev),
+			       ref_and_mask, ref_and_mask, 0x20);
+}
+
+static void gfx_v12_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
+				       struct amdgpu_job *job,
+				       struct amdgpu_ib *ib,
+				       uint32_t flags)
+{
+	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
+	u32 header, control = 0;
+
+	header = PACKET3(PACKET3_INDIRECT_BUFFER, 2);
+
+	control |= ib->length_dw | (vmid << 24);
+
+	amdgpu_ring_write(ring, header);
+	BUG_ON(ib->gpu_addr & 0x3); /* Dword align */
+	amdgpu_ring_write(ring,
+#ifdef __BIG_ENDIAN
+		(2 << 0) |
+#endif
+		lower_32_bits(ib->gpu_addr));
+	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
+	amdgpu_ring_write(ring, control);
+}
+
+static void gfx_v12_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
+					   struct amdgpu_job *job,
+					   struct amdgpu_ib *ib,
+					   uint32_t flags)
+{
+	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
+	u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24);
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
+	BUG_ON(ib->gpu_addr & 0x3); /* Dword align */
+	amdgpu_ring_write(ring,
+#ifdef __BIG_ENDIAN
+				(2 << 0) |
+#endif
+				lower_32_bits(ib->gpu_addr));
+	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
+	amdgpu_ring_write(ring, control);
+}
+
+static void gfx_v12_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
+				     u64 seq, unsigned flags)
+{
+	bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT;
+	bool int_sel = flags & AMDGPU_FENCE_FLAG_INT;
+
+	/* RELEASE_MEM - flush caches, send int */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_RELEASE_MEM, 6));
+	amdgpu_ring_write(ring, (PACKET3_RELEASE_MEM_GCR_SEQ |
+				 PACKET3_RELEASE_MEM_GCR_GL2_WB |
+				 PACKET3_RELEASE_MEM_CACHE_POLICY(3) |
+				 PACKET3_RELEASE_MEM_EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) |
+				 PACKET3_RELEASE_MEM_EVENT_INDEX(5)));
+	amdgpu_ring_write(ring, (PACKET3_RELEASE_MEM_DATA_SEL(write64bit ? 2 : 1) |
+				 PACKET3_RELEASE_MEM_INT_SEL(int_sel ? 2 : 0)));
+
+	/*
+	 * the address should be Qword aligned if 64bit write, Dword
+	 * aligned if only send 32bit data low (discard data high)
+	 */
+	if (write64bit)
+		BUG_ON(addr & 0x7);
+	else
+		BUG_ON(addr & 0x3);
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	amdgpu_ring_write(ring, lower_32_bits(seq));
+	amdgpu_ring_write(ring, upper_32_bits(seq));
+	amdgpu_ring_write(ring, 0);
+}
+
+static void gfx_v12_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
+{
+	int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX);
+	uint32_t seq = ring->fence_drv.sync_seq;
+	uint64_t addr = ring->fence_drv.gpu_addr;
+
+	gfx_v12_0_wait_reg_mem(ring, usepfp, 1, 0, lower_32_bits(addr),
+			       upper_32_bits(addr), seq, 0xffffffff, 4);
+}
+
+static void gfx_v12_0_ring_invalidate_tlbs(struct amdgpu_ring *ring,
+				   uint16_t pasid, uint32_t flush_type,
+				   bool all_hub, uint8_t dst_sel)
+{
+	amdgpu_ring_write(ring, PACKET3(PACKET3_INVALIDATE_TLBS, 0));
+	amdgpu_ring_write(ring,
+			  PACKET3_INVALIDATE_TLBS_DST_SEL(dst_sel) |
+			  PACKET3_INVALIDATE_TLBS_ALL_HUB(all_hub) |
+			  PACKET3_INVALIDATE_TLBS_PASID(pasid) |
+			  PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(flush_type));
+}
+
+static void gfx_v12_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
+					 unsigned vmid, uint64_t pd_addr)
+{
+	amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
+
+	/* compute doesn't have PFP */
+	if (ring->funcs->type == AMDGPU_RING_TYPE_GFX) {
+		/* sync PFP to ME, otherwise we might get invalid PFP reads */
+		amdgpu_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0));
+		amdgpu_ring_write(ring, 0x0);
+	}
+}
+
+static void gfx_v12_0_ring_emit_fence_kiq(struct amdgpu_ring *ring, u64 addr,
+					  u64 seq, unsigned int flags)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	/* we only allocate 32bit for each seq wb address */
+	BUG_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
+
+	/* write fence seq to the "addr" */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+				 WRITE_DATA_DST_SEL(5) | WR_CONFIRM));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	amdgpu_ring_write(ring, lower_32_bits(seq));
+
+	if (flags & AMDGPU_FENCE_FLAG_INT) {
+		/* set register to trigger INT */
+		amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+		amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+					 WRITE_DATA_DST_SEL(0) | WR_CONFIRM));
+		amdgpu_ring_write(ring, SOC15_REG_OFFSET(GC, 0, regCPC_INT_STATUS));
+		amdgpu_ring_write(ring, 0);
+		amdgpu_ring_write(ring, 0x20000000); /* src_id is 178 */
+	}
+}
+
+static void gfx_v12_0_ring_emit_cntxcntl(struct amdgpu_ring *ring,
+					 uint32_t flags)
+{
+	uint32_t dw2 = 0;
+
+	dw2 |= 0x80000000; /* set load_enable otherwise this package is just NOPs */
+	if (flags & AMDGPU_HAVE_CTX_SWITCH) {
+		/* set load_global_config & load_global_uconfig */
+		dw2 |= 0x8001;
+		/* set load_cs_sh_regs */
+		dw2 |= 0x01000000;
+		/* set load_per_context_state & load_gfx_sh_regs for GFX */
+		dw2 |= 0x10002;
+	}
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_CONTEXT_CONTROL, 1));
+	amdgpu_ring_write(ring, dw2);
+	amdgpu_ring_write(ring, 0);
+}
+
+static unsigned gfx_v12_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring,
+						   uint64_t addr)
+{
+	unsigned ret;
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	/* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, 0);
+	ret = ring->wptr & ring->buf_mask;
+	/* patch dummy value later */
+	amdgpu_ring_write(ring, 0);
+
+	return ret;
+}
+
+static int gfx_v12_0_ring_preempt_ib(struct amdgpu_ring *ring)
+{
+	int i, r = 0;
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	unsigned long flags;
+
+	if (adev->enable_mes)
+		return -EINVAL;
+
+	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
+		return -EINVAL;
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+
+	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size)) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+
+	/* assert preemption condition */
+	amdgpu_ring_set_preempt_cond_exec(ring, false);
+
+	/* assert IB preemption, emit the trailing fence */
+	kiq->pmf->kiq_unmap_queues(kiq_ring, ring, PREEMPT_QUEUES_NO_UNMAP,
+				   ring->trail_fence_gpu_addr,
+				   ++ring->trail_seq);
+	amdgpu_ring_commit(kiq_ring);
+
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+
+	/* poll the trailing fence */
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (ring->trail_seq ==
+		    le32_to_cpu(*(ring->trail_fence_cpu_addr)))
+			break;
+		udelay(1);
+	}
+
+	if (i >= adev->usec_timeout) {
+		r = -EINVAL;
+		DRM_ERROR("ring %d failed to preempt ib\n", ring->idx);
+	}
+
+	/* deassert preemption condition */
+	amdgpu_ring_set_preempt_cond_exec(ring, true);
+	return r;
+}
+
+static void gfx_v12_0_ring_emit_frame_cntl(struct amdgpu_ring *ring,
+					   bool start,
+					   bool secure)
+{
+	uint32_t v = secure ? FRAME_TMZ : 0;
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_FRAME_CONTROL, 0));
+	amdgpu_ring_write(ring, v | FRAME_CMD(start ? 0 : 1));
+}
+
+static void gfx_v12_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg,
+				     uint32_t reg_val_offs)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_COPY_DATA, 4));
+	amdgpu_ring_write(ring, 0 |	/* src: register*/
+				(5 << 8) |	/* dst: memory */
+				(1 << 20));	/* write confirm */
+	amdgpu_ring_write(ring, reg);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +
+				reg_val_offs * 4));
+	amdgpu_ring_write(ring, upper_32_bits(adev->wb.gpu_addr +
+				reg_val_offs * 4));
+}
+
+static void gfx_v12_0_ring_emit_wreg(struct amdgpu_ring *ring,
+				     uint32_t reg,
+				     uint32_t val)
+{
+	uint32_t cmd = 0;
+
+	switch (ring->funcs->type) {
+	case AMDGPU_RING_TYPE_GFX:
+		cmd = WRITE_DATA_ENGINE_SEL(1) | WR_CONFIRM;
+		break;
+	case AMDGPU_RING_TYPE_KIQ:
+		cmd = (1 << 16); /* no inc addr */
+		break;
+	default:
+		cmd = WR_CONFIRM;
+		break;
+	}
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, cmd);
+	amdgpu_ring_write(ring, reg);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, val);
+}
+
+static void gfx_v12_0_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
+					uint32_t val, uint32_t mask)
+{
+	gfx_v12_0_wait_reg_mem(ring, 0, 0, 0, reg, 0, val, mask, 0x20);
+}
+
+static void gfx_v12_0_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring,
+						   uint32_t reg0, uint32_t reg1,
+						   uint32_t ref, uint32_t mask)
+{
+	int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX);
+
+	gfx_v12_0_wait_reg_mem(ring, usepfp, 0, 1, reg0, reg1,
+			       ref, mask, 0x20);
+}
+
+static void
+gfx_v12_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev,
+				      uint32_t me, uint32_t pipe,
+				      enum amdgpu_interrupt_state state)
+{
+	uint32_t cp_int_cntl, cp_int_cntl_reg;
+
+	if (!me) {
+		switch (pipe) {
+		case 0:
+			cp_int_cntl_reg = SOC15_REG_OFFSET(GC, 0, regCP_INT_CNTL_RING0);
+			break;
+		default:
+			DRM_DEBUG("invalid pipe %d\n", pipe);
+			return;
+		}
+	} else {
+		DRM_DEBUG("invalid me %d\n", me);
+		return;
+	}
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+		cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+					    TIME_STAMP_INT_ENABLE, 0);
+		cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+					    GENERIC0_INT_ENABLE, 0);
+		WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+		cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+					    TIME_STAMP_INT_ENABLE, 1);
+		cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+					    GENERIC0_INT_ENABLE, 1);
+		WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+		break;
+	default:
+		break;
+	}
+}
+
+static void gfx_v12_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev,
+						     int me, int pipe,
+						     enum amdgpu_interrupt_state state)
+{
+	u32 mec_int_cntl, mec_int_cntl_reg;
+
+	/*
+	 * amdgpu controls only the first MEC. That's why this function only
+	 * handles the setting of interrupts for this specific MEC. All other
+	 * pipes' interrupts are set by amdkfd.
+	 */
+
+	if (me == 1) {
+		switch (pipe) {
+		case 0:
+			mec_int_cntl_reg = SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE0_INT_CNTL);
+			break;
+		case 1:
+			mec_int_cntl_reg = SOC15_REG_OFFSET(GC, 0, regCP_ME1_PIPE1_INT_CNTL);
+			break;
+		default:
+			DRM_DEBUG("invalid pipe %d\n", pipe);
+			return;
+		}
+	} else {
+		DRM_DEBUG("invalid me %d\n", me);
+		return;
+	}
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		mec_int_cntl = RREG32_SOC15_IP(GC, mec_int_cntl_reg);
+		mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+					     TIME_STAMP_INT_ENABLE, 0);
+		mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+					     GENERIC0_INT_ENABLE, 0);
+		WREG32_SOC15_IP(GC, mec_int_cntl_reg, mec_int_cntl);
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		mec_int_cntl = RREG32_SOC15_IP(GC, mec_int_cntl_reg);
+		mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+					     TIME_STAMP_INT_ENABLE, 1);
+		mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+					     GENERIC0_INT_ENABLE, 1);
+		WREG32_SOC15_IP(GC, mec_int_cntl_reg, mec_int_cntl);
+		break;
+	default:
+		break;
+	}
+}
+
+static int gfx_v12_0_set_eop_interrupt_state(struct amdgpu_device *adev,
+					    struct amdgpu_irq_src *src,
+					    unsigned type,
+					    enum amdgpu_interrupt_state state)
+{
+	switch (type) {
+	case AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP:
+		gfx_v12_0_set_gfx_eop_interrupt_state(adev, 0, 0, state);
+		break;
+	case AMDGPU_CP_IRQ_GFX_ME0_PIPE1_EOP:
+		gfx_v12_0_set_gfx_eop_interrupt_state(adev, 0, 1, state);
+		break;
+	case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP:
+		gfx_v12_0_set_compute_eop_interrupt_state(adev, 1, 0, state);
+		break;
+	case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE1_EOP:
+		gfx_v12_0_set_compute_eop_interrupt_state(adev, 1, 1, state);
+		break;
+	case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE2_EOP:
+		gfx_v12_0_set_compute_eop_interrupt_state(adev, 1, 2, state);
+		break;
+	case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE3_EOP:
+		gfx_v12_0_set_compute_eop_interrupt_state(adev, 1, 3, state);
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
+static int gfx_v12_0_eop_irq(struct amdgpu_device *adev,
+			     struct amdgpu_irq_src *source,
+			     struct amdgpu_iv_entry *entry)
+{
+	u32 doorbell_offset = entry->src_data[0];
+	u8 me_id, pipe_id, queue_id;
+	struct amdgpu_ring *ring;
+	int i;
+
+	DRM_DEBUG("IH: CP EOP\n");
+
+	if (adev->enable_mes && doorbell_offset) {
+		struct amdgpu_userq_fence_driver *fence_drv = NULL;
+		struct xarray *xa = &adev->userq_xa;
+		unsigned long flags;
+
+		xa_lock_irqsave(xa, flags);
+		fence_drv = xa_load(xa, doorbell_offset);
+		if (fence_drv)
+			amdgpu_userq_fence_driver_process(fence_drv);
+		xa_unlock_irqrestore(xa, flags);
+	} else {
+		me_id = (entry->ring_id & 0x0c) >> 2;
+		pipe_id = (entry->ring_id & 0x03) >> 0;
+		queue_id = (entry->ring_id & 0x70) >> 4;
+
+		switch (me_id) {
+		case 0:
+			if (pipe_id == 0)
+				amdgpu_fence_process(&adev->gfx.gfx_ring[0]);
+			else
+				amdgpu_fence_process(&adev->gfx.gfx_ring[1]);
+			break;
+		case 1:
+		case 2:
+			for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+				ring = &adev->gfx.compute_ring[i];
+				/* Per-queue interrupt is supported for MEC starting from VI.
+				 * The interrupt can only be enabled/disabled per pipe instead
+				 * of per queue.
+				 */
+				if ((ring->me == me_id) &&
+				    (ring->pipe == pipe_id) &&
+				    (ring->queue == queue_id))
+					amdgpu_fence_process(ring);
+			}
+			break;
+		}
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_set_priv_reg_fault_state(struct amdgpu_device *adev,
+					      struct amdgpu_irq_src *source,
+					      unsigned int type,
+					      enum amdgpu_interrupt_state state)
+{
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v12_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    PRIV_REG_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				/* MECs start at 1 */
+				cp_int_cntl_reg = gfx_v12_0_get_cpc_int_cntl(adev, i + 1, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+								    PRIV_REG_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int gfx_v12_0_set_bad_op_fault_state(struct amdgpu_device *adev,
+					    struct amdgpu_irq_src *source,
+					    unsigned type,
+					    enum amdgpu_interrupt_state state)
+{
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v12_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    OPCODE_ERROR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				/* MECs start at 1 */
+				cp_int_cntl_reg = gfx_v12_0_get_cpc_int_cntl(adev, i + 1, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+								    OPCODE_ERROR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
+static int gfx_v12_0_set_priv_inst_fault_state(struct amdgpu_device *adev,
+					       struct amdgpu_irq_src *source,
+					       unsigned int type,
+					       enum amdgpu_interrupt_state state)
+{
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < adev->gfx.me.num_me; i++) {
+			for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+				cp_int_cntl_reg = gfx_v12_0_get_cpg_int_cntl(adev, i, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_INT_CNTL_RING0,
+								    PRIV_INSTR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static void gfx_v12_0_handle_priv_fault(struct amdgpu_device *adev,
+					struct amdgpu_iv_entry *entry)
+{
+	u8 me_id, pipe_id, queue_id;
+	struct amdgpu_ring *ring;
+	int i;
+
+	me_id = (entry->ring_id & 0x0c) >> 2;
+	pipe_id = (entry->ring_id & 0x03) >> 0;
+	queue_id = (entry->ring_id & 0x70) >> 4;
+
+	if (!adev->gfx.disable_kq) {
+		switch (me_id) {
+		case 0:
+			for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+				ring = &adev->gfx.gfx_ring[i];
+				if (ring->me == me_id && ring->pipe == pipe_id &&
+				    ring->queue == queue_id)
+					drm_sched_fault(&ring->sched);
+			}
+			break;
+		case 1:
+		case 2:
+			for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+				ring = &adev->gfx.compute_ring[i];
+				if (ring->me == me_id && ring->pipe == pipe_id &&
+				    ring->queue == queue_id)
+					drm_sched_fault(&ring->sched);
+			}
+			break;
+		default:
+			BUG();
+			break;
+		}
+	}
+}
+
+static int gfx_v12_0_priv_reg_irq(struct amdgpu_device *adev,
+				  struct amdgpu_irq_src *source,
+				  struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal register access in command stream\n");
+	gfx_v12_0_handle_priv_fault(adev, entry);
+	return 0;
+}
+
+static int gfx_v12_0_bad_op_irq(struct amdgpu_device *adev,
+				struct amdgpu_irq_src *source,
+				struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal opcode in command stream \n");
+	gfx_v12_0_handle_priv_fault(adev, entry);
+	return 0;
+}
+
+static int gfx_v12_0_priv_inst_irq(struct amdgpu_device *adev,
+				   struct amdgpu_irq_src *source,
+				   struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal instruction in command stream\n");
+	gfx_v12_0_handle_priv_fault(adev, entry);
+	return 0;
+}
+
+static void gfx_v12_0_emit_mem_sync(struct amdgpu_ring *ring)
+{
+	const unsigned int gcr_cntl =
+			PACKET3_ACQUIRE_MEM_GCR_CNTL_GL2_INV(1) |
+			PACKET3_ACQUIRE_MEM_GCR_CNTL_GL2_WB(1) |
+			PACKET3_ACQUIRE_MEM_GCR_CNTL_GLM_INV(1) |
+			PACKET3_ACQUIRE_MEM_GCR_CNTL_GLM_WB(1) |
+			PACKET3_ACQUIRE_MEM_GCR_CNTL_GL1_INV(1) |
+			PACKET3_ACQUIRE_MEM_GCR_CNTL_GLV_INV(1) |
+			PACKET3_ACQUIRE_MEM_GCR_CNTL_GLK_INV(1) |
+			PACKET3_ACQUIRE_MEM_GCR_CNTL_GLI_INV(1);
+
+	/* ACQUIRE_MEM - make one or more surfaces valid for use by the subsequent operations */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_ACQUIRE_MEM, 6));
+	amdgpu_ring_write(ring, 0); /* CP_COHER_CNTL */
+	amdgpu_ring_write(ring, 0xffffffff);  /* CP_COHER_SIZE */
+	amdgpu_ring_write(ring, 0xffffff);  /* CP_COHER_SIZE_HI */
+	amdgpu_ring_write(ring, 0); /* CP_COHER_BASE */
+	amdgpu_ring_write(ring, 0);  /* CP_COHER_BASE_HI */
+	amdgpu_ring_write(ring, 0x0000000A); /* POLL_INTERVAL */
+	amdgpu_ring_write(ring, gcr_cntl); /* GCR_CNTL */
+}
+
+static void gfx_v12_ring_insert_nop(struct amdgpu_ring *ring, uint32_t num_nop)
+{
+	/* Header itself is a NOP packet */
+	if (num_nop == 1) {
+		amdgpu_ring_write(ring, ring->funcs->nop);
+		return;
+	}
+
+	/* Max HW optimization till 0x3ffe, followed by remaining one NOP at a time*/
+	amdgpu_ring_write(ring, PACKET3(PACKET3_NOP, min(num_nop - 2, 0x3ffe)));
+
+	/* Header is at index 0, followed by num_nops - 1 NOP packet's */
+	amdgpu_ring_insert_nop(ring, num_nop - 1);
+}
+
+static void gfx_v12_0_ring_emit_cleaner_shader(struct amdgpu_ring *ring)
+{
+	/* Emit the cleaner shader */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_RUN_CLEANER_SHADER, 0));
+	amdgpu_ring_write(ring, 0);  /* RESERVED field, programmed to zero */
+}
+
+static void gfx_v12_ip_print(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k, reg, index = 0;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_12_0);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	for (i = 0; i < reg_count; i++)
+		drm_printf(p, "%-50s \t 0x%08x\n",
+			   gc_reg_list_12_0[i].reg_name,
+			   adev->gfx.ip_dump_core[i]);
+
+	/* print compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_12);
+	drm_printf(p, "\nnum_mec: %d num_pipe: %d num_queue: %d\n",
+		   adev->gfx.mec.num_mec,
+		   adev->gfx.mec.num_pipe_per_mec,
+		   adev->gfx.mec.num_queue_per_pipe);
+
+	for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+		for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+			for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+				drm_printf(p, "\nmec %d, pipe %d, queue %d\n", i, j, k);
+				for (reg = 0; reg < reg_count; reg++) {
+					drm_printf(p, "%-50s \t 0x%08x\n",
+						   gc_cp_reg_list_12[reg].reg_name,
+						   adev->gfx.ip_dump_compute_queues[index + reg]);
+				}
+				index += reg_count;
+			}
+		}
+	}
+
+	/* print gfx queue registers for all instances */
+	if (!adev->gfx.ip_dump_gfx_queues)
+		return;
+
+	index = 0;
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_12);
+	drm_printf(p, "\nnum_me: %d num_pipe: %d num_queue: %d\n",
+		   adev->gfx.me.num_me,
+		   adev->gfx.me.num_pipe_per_me,
+		   adev->gfx.me.num_queue_per_pipe);
+
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			for (k = 0; k < adev->gfx.me.num_queue_per_pipe; k++) {
+				drm_printf(p, "\nme %d, pipe %d, queue %d\n", i, j, k);
+				for (reg = 0; reg < reg_count; reg++) {
+					drm_printf(p, "%-50s \t 0x%08x\n",
+						   gc_gfx_queue_reg_list_12[reg].reg_name,
+						   adev->gfx.ip_dump_gfx_queues[index + reg]);
+				}
+				index += reg_count;
+			}
+		}
+	}
+}
+
+static void gfx_v12_ip_dump(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k, reg, index = 0;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_12_0);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	amdgpu_gfx_off_ctrl(adev, false);
+	for (i = 0; i < reg_count; i++)
+		adev->gfx.ip_dump_core[i] = RREG32(SOC15_REG_ENTRY_OFFSET(gc_reg_list_12_0[i]));
+	amdgpu_gfx_off_ctrl(adev, true);
+
+	/* dump compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_12);
+	amdgpu_gfx_off_ctrl(adev, false);
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+		for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+			for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+				/* ME0 is for GFX so start from 1 for CP */
+				soc24_grbm_select(adev, adev->gfx.me.num_me + i, j, k, 0);
+				for (reg = 0; reg < reg_count; reg++) {
+					adev->gfx.ip_dump_compute_queues[index + reg] =
+						RREG32(SOC15_REG_ENTRY_OFFSET(
+							gc_cp_reg_list_12[reg]));
+				}
+				index += reg_count;
+			}
+		}
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_off_ctrl(adev, true);
+
+	/* dump gfx queue registers for all instances */
+	if (!adev->gfx.ip_dump_gfx_queues)
+		return;
+
+	index = 0;
+	reg_count = ARRAY_SIZE(gc_gfx_queue_reg_list_12);
+	amdgpu_gfx_off_ctrl(adev, false);
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->gfx.me.num_me; i++) {
+		for (j = 0; j < adev->gfx.me.num_pipe_per_me; j++) {
+			for (k = 0; k < adev->gfx.me.num_queue_per_pipe; k++) {
+				soc24_grbm_select(adev, i, j, k, 0);
+
+				for (reg = 0; reg < reg_count; reg++) {
+					adev->gfx.ip_dump_gfx_queues[index + reg] =
+						RREG32(SOC15_REG_ENTRY_OFFSET(
+							gc_gfx_queue_reg_list_12[reg]));
+				}
+				index += reg_count;
+			}
+		}
+	}
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_off_ctrl(adev, true);
+}
+
+static bool gfx_v12_pipe_reset_support(struct amdgpu_device *adev)
+{
+	/* Disable the pipe reset until the CPFW fully support it.*/
+	dev_warn_once(adev->dev, "The CPFW hasn't support pipe reset yet.\n");
+	return false;
+}
+
+static int gfx_v12_reset_gfx_pipe(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t reset_pipe = 0, clean_pipe = 0;
+	int r;
+
+	if (!gfx_v12_pipe_reset_support(adev))
+		return -EOPNOTSUPP;
+
+	gfx_v12_0_set_safe_mode(adev, 0);
+	mutex_lock(&adev->srbm_mutex);
+	soc24_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+
+	switch (ring->pipe) {
+	case 0:
+		reset_pipe = REG_SET_FIELD(reset_pipe, CP_ME_CNTL,
+					   PFP_PIPE0_RESET, 1);
+		reset_pipe = REG_SET_FIELD(reset_pipe, CP_ME_CNTL,
+					   ME_PIPE0_RESET, 1);
+		clean_pipe = REG_SET_FIELD(clean_pipe, CP_ME_CNTL,
+					   PFP_PIPE0_RESET, 0);
+		clean_pipe = REG_SET_FIELD(clean_pipe, CP_ME_CNTL,
+					   ME_PIPE0_RESET, 0);
+		break;
+	case 1:
+		reset_pipe = REG_SET_FIELD(reset_pipe, CP_ME_CNTL,
+					   PFP_PIPE1_RESET, 1);
+		reset_pipe = REG_SET_FIELD(reset_pipe, CP_ME_CNTL,
+					   ME_PIPE1_RESET, 1);
+		clean_pipe = REG_SET_FIELD(clean_pipe, CP_ME_CNTL,
+					   PFP_PIPE1_RESET, 0);
+		clean_pipe = REG_SET_FIELD(clean_pipe, CP_ME_CNTL,
+					   ME_PIPE1_RESET, 0);
+		break;
+	default:
+		break;
+	}
+
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, reset_pipe);
+	WREG32_SOC15(GC, 0, regCP_ME_CNTL, clean_pipe);
+
+	r = (RREG32(SOC15_REG_OFFSET(GC, 0, regCP_GFX_RS64_INSTR_PNTR1)) << 2) -
+					RS64_FW_UC_START_ADDR_LO;
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	gfx_v12_0_unset_safe_mode(adev, 0);
+
+	dev_info(adev->dev, "The ring %s pipe reset: %s\n", ring->name,
+			r == 0 ? "successfully" : "failed");
+	/* Sometimes the ME start pc counter can't cache correctly, so the
+	 * PC check only as a reference and pipe reset result rely on the
+	 * later ring test.
+	 */
+	return 0;
+}
+
+static int gfx_v12_0_reset_kgq(struct amdgpu_ring *ring,
+			       unsigned int vmid,
+			       struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int r;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	r = amdgpu_mes_reset_legacy_queue(ring->adev, ring, vmid, false);
+	if (r) {
+		dev_warn(adev->dev, "reset via MES failed and try pipe reset %d\n", r);
+		r = gfx_v12_reset_gfx_pipe(ring);
+		if (r)
+			return r;
+	}
+
+	r = gfx_v12_0_kgq_init_queue(ring, true);
+	if (r) {
+		dev_err(adev->dev, "failed to init kgq\n");
+		return r;
+	}
+
+	r = amdgpu_mes_map_legacy_queue(adev, ring);
+	if (r) {
+		dev_err(adev->dev, "failed to remap kgq\n");
+		return r;
+	}
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static int gfx_v12_0_reset_compute_pipe(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t reset_pipe = 0, clean_pipe = 0;
+	int r = 0;
+
+	if (!gfx_v12_pipe_reset_support(adev))
+		return -EOPNOTSUPP;
+
+	gfx_v12_0_set_safe_mode(adev, 0);
+	mutex_lock(&adev->srbm_mutex);
+	soc24_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+
+	reset_pipe = RREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL);
+	clean_pipe = reset_pipe;
+
+	if (adev->gfx.rs64_enable) {
+		switch (ring->pipe) {
+		case 0:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE0_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE0_RESET, 0);
+			break;
+		case 1:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE1_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE1_RESET, 0);
+			break;
+		case 2:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE2_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE2_RESET, 0);
+			break;
+		case 3:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE3_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_RS64_CNTL,
+						   MEC_PIPE3_RESET, 0);
+			break;
+		default:
+			break;
+		}
+		WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, reset_pipe);
+		WREG32_SOC15(GC, 0, regCP_MEC_RS64_CNTL, clean_pipe);
+		r = (RREG32_SOC15(GC, 0, regCP_MEC_RS64_INSTR_PNTR) << 2) -
+				RS64_FW_UC_START_ADDR_LO;
+	} else {
+		switch (ring->pipe) {
+		case 0:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE0_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE0_RESET, 0);
+			break;
+		case 1:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE1_RESET, 1);
+			clean_pipe = REG_SET_FIELD(clean_pipe, CP_MEC_CNTL,
+							   MEC_ME1_PIPE1_RESET, 0);
+			break;
+		default:
+		break;
+		}
+		WREG32_SOC15(GC, 0, regCP_MEC_CNTL, reset_pipe);
+		WREG32_SOC15(GC, 0, regCP_MEC_CNTL, clean_pipe);
+		/* Doesn't find the F32 MEC instruction pointer register, and suppose
+		 * the driver won't run into the F32 mode.
+		 */
+	}
+
+	soc24_grbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	gfx_v12_0_unset_safe_mode(adev, 0);
+
+	dev_info(adev->dev, "The ring %s pipe resets: %s\n", ring->name,
+			r == 0 ? "successfully" : "failed");
+	/* Need the ring test to verify the pipe reset result.*/
+	return 0;
+}
+
+static int gfx_v12_0_reset_kcq(struct amdgpu_ring *ring,
+			       unsigned int vmid,
+			       struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int r;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	r = amdgpu_mes_reset_legacy_queue(ring->adev, ring, vmid, true);
+	if (r) {
+		dev_warn(adev->dev, "fail(%d) to reset kcq  and try pipe reset\n", r);
+		r = gfx_v12_0_reset_compute_pipe(ring);
+		if (r)
+			return r;
+	}
+
+	r = gfx_v12_0_kcq_init_queue(ring, true);
+	if (r) {
+		dev_err(adev->dev, "failed to init kcq\n");
+		return r;
+	}
+	r = amdgpu_mes_map_legacy_queue(adev, ring);
+	if (r) {
+		dev_err(adev->dev, "failed to remap kcq\n");
+		return r;
+	}
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static void gfx_v12_0_ring_begin_use(struct amdgpu_ring *ring)
+{
+	amdgpu_gfx_profile_ring_begin_use(ring);
+
+	amdgpu_gfx_enforce_isolation_ring_begin_use(ring);
+}
+
+static void gfx_v12_0_ring_end_use(struct amdgpu_ring *ring)
+{
+	amdgpu_gfx_profile_ring_end_use(ring);
+
+	amdgpu_gfx_enforce_isolation_ring_end_use(ring);
+}
+
+static const struct amd_ip_funcs gfx_v12_0_ip_funcs = {
+	.name = "gfx_v12_0",
+	.early_init = gfx_v12_0_early_init,
+	.late_init = gfx_v12_0_late_init,
+	.sw_init = gfx_v12_0_sw_init,
+	.sw_fini = gfx_v12_0_sw_fini,
+	.hw_init = gfx_v12_0_hw_init,
+	.hw_fini = gfx_v12_0_hw_fini,
+	.suspend = gfx_v12_0_suspend,
+	.resume = gfx_v12_0_resume,
+	.is_idle = gfx_v12_0_is_idle,
+	.wait_for_idle = gfx_v12_0_wait_for_idle,
+	.set_clockgating_state = gfx_v12_0_set_clockgating_state,
+	.set_powergating_state = gfx_v12_0_set_powergating_state,
+	.get_clockgating_state = gfx_v12_0_get_clockgating_state,
+	.dump_ip_state = gfx_v12_ip_dump,
+	.print_ip_state = gfx_v12_ip_print,
+};
+
+static const struct amdgpu_ring_funcs gfx_v12_0_ring_funcs_gfx = {
+	.type = AMDGPU_RING_TYPE_GFX,
+	.align_mask = 0xff,
+	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
+	.support_64bit_ptrs = true,
+	.secure_submission_supported = true,
+	.get_rptr = gfx_v12_0_ring_get_rptr_gfx,
+	.get_wptr = gfx_v12_0_ring_get_wptr_gfx,
+	.set_wptr = gfx_v12_0_ring_set_wptr_gfx,
+	.emit_frame_size = /* totally 242 maximum if 16 IBs */
+		5 + /* COND_EXEC */
+		7 + /* PIPELINE_SYNC */
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
+		2 + /* VM_FLUSH */
+		8 + /* FENCE for VM_FLUSH */
+		5 + /* COND_EXEC */
+		7 + /* HDP_flush */
+		4 + /* VGT_flush */
+		31 + /*	DE_META */
+		3 + /* CNTX_CTRL */
+		5 + /* HDP_INVL */
+		8 + 8 + /* FENCE x2 */
+		8 + /* gfx_v12_0_emit_mem_sync */
+		2, /* gfx_v12_0_ring_emit_cleaner_shader */
+	.emit_ib_size =	4, /* gfx_v12_0_ring_emit_ib_gfx */
+	.emit_ib = gfx_v12_0_ring_emit_ib_gfx,
+	.emit_fence = gfx_v12_0_ring_emit_fence,
+	.emit_pipeline_sync = gfx_v12_0_ring_emit_pipeline_sync,
+	.emit_vm_flush = gfx_v12_0_ring_emit_vm_flush,
+	.emit_hdp_flush = gfx_v12_0_ring_emit_hdp_flush,
+	.test_ring = gfx_v12_0_ring_test_ring,
+	.test_ib = gfx_v12_0_ring_test_ib,
+	.insert_nop = gfx_v12_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.emit_cntxcntl = gfx_v12_0_ring_emit_cntxcntl,
+	.init_cond_exec = gfx_v12_0_ring_emit_init_cond_exec,
+	.preempt_ib = gfx_v12_0_ring_preempt_ib,
+	.emit_frame_cntl = gfx_v12_0_ring_emit_frame_cntl,
+	.emit_wreg = gfx_v12_0_ring_emit_wreg,
+	.emit_reg_wait = gfx_v12_0_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = gfx_v12_0_ring_emit_reg_write_reg_wait,
+	.emit_mem_sync = gfx_v12_0_emit_mem_sync,
+	.reset = gfx_v12_0_reset_kgq,
+	.emit_cleaner_shader = gfx_v12_0_ring_emit_cleaner_shader,
+	.begin_use = gfx_v12_0_ring_begin_use,
+	.end_use = gfx_v12_0_ring_end_use,
+};
+
+static const struct amdgpu_ring_funcs gfx_v12_0_ring_funcs_compute = {
+	.type = AMDGPU_RING_TYPE_COMPUTE,
+	.align_mask = 0xff,
+	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
+	.support_64bit_ptrs = true,
+	.get_rptr = gfx_v12_0_ring_get_rptr_compute,
+	.get_wptr = gfx_v12_0_ring_get_wptr_compute,
+	.set_wptr = gfx_v12_0_ring_set_wptr_compute,
+	.emit_frame_size =
+		7 + /* gfx_v12_0_ring_emit_hdp_flush */
+		5 + /* hdp invalidate */
+		7 + /* gfx_v12_0_ring_emit_pipeline_sync */
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
+		2 + /* gfx_v12_0_ring_emit_vm_flush */
+		8 + 8 + 8 + /* gfx_v12_0_ring_emit_fence x3 for user fence, vm fence */
+		8 + /* gfx_v12_0_emit_mem_sync */
+		2, /* gfx_v12_0_ring_emit_cleaner_shader */
+	.emit_ib_size =	7, /* gfx_v12_0_ring_emit_ib_compute */
+	.emit_ib = gfx_v12_0_ring_emit_ib_compute,
+	.emit_fence = gfx_v12_0_ring_emit_fence,
+	.emit_pipeline_sync = gfx_v12_0_ring_emit_pipeline_sync,
+	.emit_vm_flush = gfx_v12_0_ring_emit_vm_flush,
+	.emit_hdp_flush = gfx_v12_0_ring_emit_hdp_flush,
+	.test_ring = gfx_v12_0_ring_test_ring,
+	.test_ib = gfx_v12_0_ring_test_ib,
+	.insert_nop = gfx_v12_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.emit_wreg = gfx_v12_0_ring_emit_wreg,
+	.emit_reg_wait = gfx_v12_0_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = gfx_v12_0_ring_emit_reg_write_reg_wait,
+	.emit_mem_sync = gfx_v12_0_emit_mem_sync,
+	.reset = gfx_v12_0_reset_kcq,
+	.emit_cleaner_shader = gfx_v12_0_ring_emit_cleaner_shader,
+	.begin_use = gfx_v12_0_ring_begin_use,
+	.end_use = gfx_v12_0_ring_end_use,
+};
+
+static const struct amdgpu_ring_funcs gfx_v12_0_ring_funcs_kiq = {
+	.type = AMDGPU_RING_TYPE_KIQ,
+	.align_mask = 0xff,
+	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
+	.support_64bit_ptrs = true,
+	.get_rptr = gfx_v12_0_ring_get_rptr_compute,
+	.get_wptr = gfx_v12_0_ring_get_wptr_compute,
+	.set_wptr = gfx_v12_0_ring_set_wptr_compute,
+	.emit_frame_size =
+		7 + /* gfx_v12_0_ring_emit_hdp_flush */
+		5 + /*hdp invalidate */
+		7 + /* gfx_v12_0_ring_emit_pipeline_sync */
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
+		2 + /* gfx_v12_0_ring_emit_vm_flush */
+		8 + 8 + 8, /* gfx_v12_0_ring_emit_fence_kiq x3 for user fence, vm fence */
+	.emit_ib_size =	7, /* gfx_v12_0_ring_emit_ib_compute */
+	.emit_ib = gfx_v12_0_ring_emit_ib_compute,
+	.emit_fence = gfx_v12_0_ring_emit_fence_kiq,
+	.test_ring = gfx_v12_0_ring_test_ring,
+	.test_ib = gfx_v12_0_ring_test_ib,
+	.insert_nop = amdgpu_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.emit_rreg = gfx_v12_0_ring_emit_rreg,
+	.emit_wreg = gfx_v12_0_ring_emit_wreg,
+	.emit_reg_wait = gfx_v12_0_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = gfx_v12_0_ring_emit_reg_write_reg_wait,
+	.emit_hdp_flush = gfx_v12_0_ring_emit_hdp_flush,
+};
+
+static void gfx_v12_0_set_ring_funcs(struct amdgpu_device *adev)
+{
+	int i;
+
+	adev->gfx.kiq[0].ring.funcs = &gfx_v12_0_ring_funcs_kiq;
+
+	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
+		adev->gfx.gfx_ring[i].funcs = &gfx_v12_0_ring_funcs_gfx;
+
+	for (i = 0; i < adev->gfx.num_compute_rings; i++)
+		adev->gfx.compute_ring[i].funcs = &gfx_v12_0_ring_funcs_compute;
+}
+
+static const struct amdgpu_irq_src_funcs gfx_v12_0_eop_irq_funcs = {
+	.set = gfx_v12_0_set_eop_interrupt_state,
+	.process = gfx_v12_0_eop_irq,
+};
+
+static const struct amdgpu_irq_src_funcs gfx_v12_0_priv_reg_irq_funcs = {
+	.set = gfx_v12_0_set_priv_reg_fault_state,
+	.process = gfx_v12_0_priv_reg_irq,
+};
+
+static const struct amdgpu_irq_src_funcs gfx_v12_0_bad_op_irq_funcs = {
+	.set = gfx_v12_0_set_bad_op_fault_state,
+	.process = gfx_v12_0_bad_op_irq,
+};
+
+static const struct amdgpu_irq_src_funcs gfx_v12_0_priv_inst_irq_funcs = {
+	.set = gfx_v12_0_set_priv_inst_fault_state,
+	.process = gfx_v12_0_priv_inst_irq,
+};
+
+static void gfx_v12_0_set_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->gfx.eop_irq.num_types = AMDGPU_CP_IRQ_LAST;
+	adev->gfx.eop_irq.funcs = &gfx_v12_0_eop_irq_funcs;
+
+	adev->gfx.priv_reg_irq.num_types = 1;
+	adev->gfx.priv_reg_irq.funcs = &gfx_v12_0_priv_reg_irq_funcs;
+
+	adev->gfx.bad_op_irq.num_types = 1;
+	adev->gfx.bad_op_irq.funcs = &gfx_v12_0_bad_op_irq_funcs;
+
+	adev->gfx.priv_inst_irq.num_types = 1;
+	adev->gfx.priv_inst_irq.funcs = &gfx_v12_0_priv_inst_irq_funcs;
+}
+
+static void gfx_v12_0_set_imu_funcs(struct amdgpu_device *adev)
+{
+	if (adev->flags & AMD_IS_APU)
+		adev->gfx.imu.mode = MISSION_MODE;
+	else
+		adev->gfx.imu.mode = DEBUG_MODE;
+
+	adev->gfx.imu.funcs = &gfx_v12_0_imu_funcs;
+}
+
+static void gfx_v12_0_set_rlc_funcs(struct amdgpu_device *adev)
+{
+	adev->gfx.rlc.funcs = &gfx_v12_0_rlc_funcs;
+}
+
+static void gfx_v12_0_set_mqd_funcs(struct amdgpu_device *adev)
+{
+	/* set gfx eng mqd */
+	adev->mqds[AMDGPU_HW_IP_GFX].mqd_size =
+		sizeof(struct v12_gfx_mqd);
+	adev->mqds[AMDGPU_HW_IP_GFX].init_mqd =
+		gfx_v12_0_gfx_mqd_init;
+	/* set compute eng mqd */
+	adev->mqds[AMDGPU_HW_IP_COMPUTE].mqd_size =
+		sizeof(struct v12_compute_mqd);
+	adev->mqds[AMDGPU_HW_IP_COMPUTE].init_mqd =
+		gfx_v12_0_compute_mqd_init;
+}
+
+static void gfx_v12_0_set_user_wgp_inactive_bitmap_per_sh(struct amdgpu_device *adev,
+							  u32 bitmap)
+{
+	u32 data;
+
+	if (!bitmap)
+		return;
+
+	data = bitmap << GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_WGPS__SHIFT;
+	data &= GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_WGPS_MASK;
+
+	WREG32_SOC15(GC, 0, regGC_USER_SHADER_ARRAY_CONFIG, data);
+}
+
+static u32 gfx_v12_0_get_wgp_active_bitmap_per_sh(struct amdgpu_device *adev)
+{
+	u32 data, wgp_bitmask;
+	data = RREG32_SOC15(GC, 0, regCC_GC_SHADER_ARRAY_CONFIG);
+	data |= RREG32_SOC15(GC, 0, regGC_USER_SHADER_ARRAY_CONFIG);
+
+	data &= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_WGPS_MASK;
+	data >>= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_WGPS__SHIFT;
+
+	wgp_bitmask =
+		amdgpu_gfx_create_bitmask(adev->gfx.config.max_cu_per_sh >> 1);
+
+	return (~data) & wgp_bitmask;
+}
+
+static u32 gfx_v12_0_get_cu_active_bitmap_per_sh(struct amdgpu_device *adev)
+{
+	u32 wgp_idx, wgp_active_bitmap;
+	u32 cu_bitmap_per_wgp, cu_active_bitmap;
+
+	wgp_active_bitmap = gfx_v12_0_get_wgp_active_bitmap_per_sh(adev);
+	cu_active_bitmap = 0;
+
+	for (wgp_idx = 0; wgp_idx < 16; wgp_idx++) {
+		/* if there is one WGP enabled, it means 2 CUs will be enabled */
+		cu_bitmap_per_wgp = 3 << (2 * wgp_idx);
+		if (wgp_active_bitmap & (1 << wgp_idx))
+			cu_active_bitmap |= cu_bitmap_per_wgp;
+	}
+
+	return cu_active_bitmap;
+}
+
+static int gfx_v12_0_get_cu_info(struct amdgpu_device *adev,
+				 struct amdgpu_cu_info *cu_info)
+{
+	int i, j, k, counter, active_cu_number = 0;
+	u32 mask, bitmap;
+	unsigned disable_masks[8 * 2];
+
+	if (!adev || !cu_info)
+		return -EINVAL;
+
+	amdgpu_gfx_parse_disable_cu(disable_masks, 8, 2);
+
+	mutex_lock(&adev->grbm_idx_mutex);
+	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
+		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
+			bitmap = i * adev->gfx.config.max_sh_per_se + j;
+			if (!((gfx_v12_0_get_sa_active_bitmap(adev) >> bitmap) & 1))
+				continue;
+			mask = 1;
+			counter = 0;
+			gfx_v12_0_select_se_sh(adev, i, j, 0xffffffff, 0);
+			if (i < 8 && j < 2)
+				gfx_v12_0_set_user_wgp_inactive_bitmap_per_sh(
+					adev, disable_masks[i * 2 + j]);
+			bitmap = gfx_v12_0_get_cu_active_bitmap_per_sh(adev);
+
+			/**
+			 * GFX12 could support more than 4 SEs, while the bitmap
+			 * in cu_info struct is 4x4 and ioctl interface struct
+			 * drm_amdgpu_info_device should keep stable.
+			 * So we use last two columns of bitmap to store cu mask for
+			 * SEs 4 to 7, the layout of the bitmap is as below:
+			 *    SE0: {SH0,SH1} --> {bitmap[0][0], bitmap[0][1]}
+			 *    SE1: {SH0,SH1} --> {bitmap[1][0], bitmap[1][1]}
+			 *    SE2: {SH0,SH1} --> {bitmap[2][0], bitmap[2][1]}
+			 *    SE3: {SH0,SH1} --> {bitmap[3][0], bitmap[3][1]}
+			 *    SE4: {SH0,SH1} --> {bitmap[0][2], bitmap[0][3]}
+			 *    SE5: {SH0,SH1} --> {bitmap[1][2], bitmap[1][3]}
+			 *    SE6: {SH0,SH1} --> {bitmap[2][2], bitmap[2][3]}
+			 *    SE7: {SH0,SH1} --> {bitmap[3][2], bitmap[3][3]}
+			 */
+			cu_info->bitmap[0][i % 4][j + (i / 4) * 2] = bitmap;
+
+			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
+				if (bitmap & mask)
+					counter++;
+
+				mask <<= 1;
+			}
+			active_cu_number += counter;
+		}
+	}
+	gfx_v12_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	cu_info->number = active_cu_number;
+	cu_info->simd_per_cu = NUM_SIMD_PER_CU;
+
+	return 0;
+}
+
+const struct amdgpu_ip_block_version gfx_v12_0_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_GFX,
+	.major = 12,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &gfx_v12_0_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.h b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.h
new file mode 100644
index 000000000000..f7184b2dc4e8
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.h
@@ -0,0 +1,31 @@
+/*
+ * Copyright 2023 dvanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __GFX_V12_0_H__
+#define __GFX_V12_0_H__
+
+extern const struct amdgpu_ip_block_version gfx_v12_0_ip_block;
+
+int gfx_v12_0_request_gfx_index_mutex(struct amdgpu_device *adev,
+				      bool req);
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c
index c41219e23151..80565392313f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c
@@ -28,19 +28,33 @@
 #include "amdgpu_gfx.h"
 #include "amdgpu_ucode.h"
 #include "clearstate_si.h"
+#include "si.h"
+#include "sid.h"
+
 #include "bif/bif_3_0_d.h"
 #include "bif/bif_3_0_sh_mask.h"
+
 #include "oss/oss_1_0_d.h"
 #include "oss/oss_1_0_sh_mask.h"
+
 #include "gca/gfx_6_0_d.h"
 #include "gca/gfx_6_0_sh_mask.h"
+#include "gca/gfx_7_2_enum.h"
+
 #include "gmc/gmc_6_0_d.h"
 #include "gmc/gmc_6_0_sh_mask.h"
+
 #include "dce/dce_6_0_d.h"
 #include "dce/dce_6_0_sh_mask.h"
-#include "gca/gfx_7_2_enum.h"
+
 #include "si_enums.h"
-#include "si.h"
+
+#define TAHITI_GB_ADDR_CONFIG_GOLDEN        0x12011003
+#define VERDE_GB_ADDR_CONFIG_GOLDEN         0x12010002
+#define HAINAN_GB_ADDR_CONFIG_GOLDEN        0x02010001
+
+#define GFX6_NUM_GFX_RINGS     1
+#define GFX6_NUM_COMPUTE_RINGS 2
 
 static void gfx_v6_0_set_ring_funcs(struct amdgpu_device *adev);
 static void gfx_v6_0_set_irq_funcs(struct amdgpu_device *adev);
@@ -72,7 +86,7 @@ MODULE_FIRMWARE("amdgpu/hainan_ce.bin");
 MODULE_FIRMWARE("amdgpu/hainan_rlc.bin");
 
 static u32 gfx_v6_0_get_csb_size(struct amdgpu_device *adev);
-static void gfx_v6_0_get_csb_buffer(struct amdgpu_device *adev, volatile u32 *buffer);
+static void gfx_v6_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer);
 //static void gfx_v6_0_init_cp_pg_table(struct amdgpu_device *adev);
 static void gfx_v6_0_init_pg(struct amdgpu_device *adev);
 
@@ -311,7 +325,6 @@ static const u32 verde_rlc_save_restore_register_list[] =
 static int gfx_v6_0_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err;
 	const struct gfx_firmware_header_v1_0 *cp_hdr;
 	const struct rlc_firmware_header_v1_0 *rlc_hdr;
@@ -337,32 +350,36 @@ static int gfx_v6_0_init_microcode(struct amdgpu_device *adev)
 	default: BUG();
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_pfp.bin", chip_name);
 	if (err)
 		goto out;
 	cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.pfp_fw->data;
 	adev->gfx.pfp_fw_version = le32_to_cpu(cp_hdr->header.ucode_version);
 	adev->gfx.pfp_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_me.bin", chip_name);
 	if (err)
 		goto out;
 	cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.me_fw->data;
 	adev->gfx.me_fw_version = le32_to_cpu(cp_hdr->header.ucode_version);
 	adev->gfx.me_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_ce.bin", chip_name);
 	if (err)
 		goto out;
 	cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data;
 	adev->gfx.ce_fw_version = le32_to_cpu(cp_hdr->header.ucode_version);
 	adev->gfx.ce_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_rlc.bin", chip_name);
 	if (err)
 		goto out;
 	rlc_hdr = (const struct rlc_firmware_header_v1_0 *)adev->gfx.rlc_fw->data;
@@ -371,7 +388,7 @@ static int gfx_v6_0_init_microcode(struct amdgpu_device *adev)
 
 out:
 	if (err) {
-		pr_err("gfx6: Failed to load firmware \"%s\"\n", fw_name);
+		pr_err("gfx6: Failed to load firmware %s gfx firmware\n", chip_name);
 		amdgpu_ucode_release(&adev->gfx.pfp_fw);
 		amdgpu_ucode_release(&adev->gfx.me_fw);
 		amdgpu_ucode_release(&adev->gfx.ce_fw);
@@ -1285,7 +1302,7 @@ static void gfx_v6_0_tiling_mode_table_init(struct amdgpu_device *adev)
 }
 
 static void gfx_v6_0_select_se_sh(struct amdgpu_device *adev, u32 se_num,
-				  u32 sh_num, u32 instance)
+				  u32 sh_num, u32 instance, int xcc_id)
 {
 	u32 data;
 
@@ -1438,12 +1455,12 @@ static void gfx_v6_0_write_harvested_raster_configs(struct amdgpu_device *adev,
 		}
 
 		/* GRBM_GFX_INDEX has a different offset on SI */
-		gfx_v6_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff);
+		gfx_v6_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff, 0);
 		WREG32(mmPA_SC_RASTER_CONFIG, raster_config_se);
 	}
 
 	/* GRBM_GFX_INDEX has a different offset on SI */
-	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 }
 
 static void gfx_v6_0_setup_rb(struct amdgpu_device *adev)
@@ -1459,14 +1476,14 @@ static void gfx_v6_0_setup_rb(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v6_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v6_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			data = gfx_v6_0_get_rb_active_bitmap(adev);
 			active_rbs |= data <<
 				((i * adev->gfx.config.max_sh_per_se + j) *
 				 rb_bitmap_width_per_sh);
 		}
 	}
-	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 
 	adev->gfx.config.backend_enable_mask = active_rbs;
 	adev->gfx.config.num_rbs = hweight32(active_rbs);
@@ -1487,7 +1504,7 @@ static void gfx_v6_0_setup_rb(struct amdgpu_device *adev)
 	/* cache the values for userspace */
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v6_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v6_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			adev->gfx.config.rb_config[i][j].rb_backend_disable =
 				RREG32(mmCC_RB_BACKEND_DISABLE);
 			adev->gfx.config.rb_config[i][j].user_rb_backend_disable =
@@ -1496,7 +1513,7 @@ static void gfx_v6_0_setup_rb(struct amdgpu_device *adev)
 				RREG32(mmPA_SC_RASTER_CONFIG);
 		}
 	}
-	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 }
 
@@ -1535,7 +1552,7 @@ static void gfx_v6_0_setup_spi(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v6_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v6_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			data = RREG32(mmSPI_STATIC_THREAD_MGMT_3);
 			active_cu = gfx_v6_0_get_cu_enabled(adev);
 
@@ -1550,7 +1567,7 @@ static void gfx_v6_0_setup_spi(struct amdgpu_device *adev)
 			}
 		}
 	}
-	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 }
 
@@ -1718,10 +1735,14 @@ static void gfx_v6_0_constants_init(struct amdgpu_device *adev)
 	gfx_v6_0_get_cu_info(adev);
 	gfx_v6_0_config_init(adev);
 
-	WREG32(mmCP_QUEUE_THRESHOLDS, ((0x16 << CP_QUEUE_THRESHOLDS__ROQ_IB1_START__SHIFT) |
-				       (0x2b << CP_QUEUE_THRESHOLDS__ROQ_IB2_START__SHIFT)));
-	WREG32(mmCP_MEQ_THRESHOLDS, (0x30 << CP_MEQ_THRESHOLDS__MEQ1_START__SHIFT) |
-				    (0x60 << CP_MEQ_THRESHOLDS__MEQ2_START__SHIFT));
+	WREG32(mmCP_QUEUE_THRESHOLDS,
+		((0x16 << CP_QUEUE_THRESHOLDS__ROQ_IB1_START__SHIFT) |
+		(0x2b << CP_QUEUE_THRESHOLDS__ROQ_IB2_START__SHIFT)));
+
+	/* set HW defaults for 3D engine */
+	WREG32(mmCP_MEQ_THRESHOLDS,
+		(0x30 << CP_MEQ_THRESHOLDS__MEQ1_START__SHIFT) |
+		(0x60 << CP_MEQ_THRESHOLDS__MEQ2_START__SHIFT));
 
 	sx_debug_1 = RREG32(mmSX_DEBUG_1);
 	WREG32(mmSX_DEBUG_1, sx_debug_1);
@@ -1907,7 +1928,7 @@ static int gfx_v6_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 error:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 	return r;
 }
@@ -2333,7 +2354,7 @@ static void gfx_v6_0_ring_emit_wreg(struct amdgpu_ring *ring,
 static int gfx_v6_0_rlc_init(struct amdgpu_device *adev)
 {
 	const u32 *src_ptr;
-	volatile u32 *dst_ptr;
+	u32 *dst_ptr;
 	u32 dws;
 	u64 reg_list_mc_addr;
 	const struct cs_section_def *cs_data;
@@ -2391,7 +2412,7 @@ static void gfx_v6_0_enable_lbpw(struct amdgpu_device *adev, bool enable)
 	WREG32_FIELD(RLC_LB_CNTL, LOAD_BALANCE_ENABLE, enable ? 1 : 0);
 
 	if (!enable) {
-		gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+		gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 		WREG32(mmSPI_LB_CU_MASK, 0x00ff);
 	}
 }
@@ -2834,47 +2855,23 @@ static u32 gfx_v6_0_get_csb_size(struct amdgpu_device *adev)
 	return count;
 }
 
-static void gfx_v6_0_get_csb_buffer(struct amdgpu_device *adev,
-				    volatile u32 *buffer)
+static void gfx_v6_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer)
 {
-	u32 count = 0, i;
-	const struct cs_section_def *sect = NULL;
-	const struct cs_extent_def *ext = NULL;
+	u32 count = 0;
 
 	if (adev->gfx.rlc.cs_data == NULL)
 		return;
 	if (buffer == NULL)
 		return;
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1));
-	buffer[count++] = cpu_to_le32(0x80000000);
-	buffer[count++] = cpu_to_le32(0x80000000);
-
-	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
-		for (ext = sect->section; ext->extent != NULL; ++ext) {
-			if (sect->id == SECT_CONTEXT) {
-				buffer[count++] =
-					cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
-				buffer[count++] = cpu_to_le32(ext->reg_index - 0xa000);
-				for (i = 0; i < ext->reg_count; i++)
-					buffer[count++] = cpu_to_le32(ext->extent[i]);
-			} else {
-				return;
-			}
-		}
-	}
+	count = amdgpu_gfx_csb_preamble_start(buffer);
+	count = amdgpu_gfx_csb_data_parser(adev, buffer, count);
 
 	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, 1));
 	buffer[count++] = cpu_to_le32(mmPA_SC_RASTER_CONFIG - PACKET3_SET_CONTEXT_REG_START);
 	buffer[count++] = cpu_to_le32(adev->gfx.config.rb_config[0][0].raster_config);
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0));
-	buffer[count++] = cpu_to_le32(0);
+	amdgpu_gfx_csb_preamble_end(buffer, count);
 }
 
 static void gfx_v6_0_init_pg(struct amdgpu_device *adev)
@@ -2968,7 +2965,7 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd,
 		*(out++) = RREG32(mmSQ_IND_DATA);
 }
 
-static void gfx_v6_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
+static void gfx_v6_0_read_wave_data(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
 {
 	/* type 0 wave data */
 	dst[(*no_fields)++] = 0;
@@ -2993,7 +2990,7 @@ static void gfx_v6_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, u
 	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_MODE);
 }
 
-static void gfx_v6_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v6_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				     uint32_t wave, uint32_t start,
 				     uint32_t size, uint32_t *dst)
 {
@@ -3003,7 +3000,7 @@ static void gfx_v6_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
 }
 
 static void gfx_v6_0_select_me_pipe_q(struct amdgpu_device *adev,
-				  u32 me, u32 pipe, u32 q, u32 vm)
+				  u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
 {
 	DRM_INFO("Not implemented\n");
 }
@@ -3024,10 +3021,11 @@ static const struct amdgpu_rlc_funcs gfx_v6_0_rlc_funcs = {
 	.start = gfx_v6_0_rlc_start
 };
 
-static int gfx_v6_0_early_init(void *handle)
+static int gfx_v6_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
+	adev->gfx.xcc_mask = 1;
 	adev->gfx.num_gfx_rings = GFX6_NUM_GFX_RINGS;
 	adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
 					  GFX6_NUM_COMPUTE_RINGS);
@@ -3039,10 +3037,10 @@ static int gfx_v6_0_early_init(void *handle)
 	return 0;
 }
 
-static int gfx_v6_0_sw_init(void *handle)
+static int gfx_v6_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i, r;
 
 	r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 181, &adev->gfx.eop_irq);
@@ -3073,7 +3071,7 @@ static int gfx_v6_0_sw_init(void *handle)
 		ring = &adev->gfx.gfx_ring[i];
 		ring->ring_obj = NULL;
 		sprintf(ring->name, "gfx");
-		r = amdgpu_ring_init(adev, ring, 1024,
+		r = amdgpu_ring_init(adev, ring, 2048,
 				     &adev->gfx.eop_irq,
 				     AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP,
 				     AMDGPU_RING_PRIO_DEFAULT, NULL);
@@ -3104,13 +3102,18 @@ static int gfx_v6_0_sw_init(void *handle)
 			return r;
 	}
 
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+
 	return r;
 }
 
-static int gfx_v6_0_sw_fini(void *handle)
+static int gfx_v6_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.gfx_ring[i]);
@@ -3122,10 +3125,10 @@ static int gfx_v6_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int gfx_v6_0_hw_init(void *handle)
+static int gfx_v6_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gfx_v6_0_constants_init(adev);
 
@@ -3142,9 +3145,9 @@ static int gfx_v6_0_hw_init(void *handle)
 	return r;
 }
 
-static int gfx_v6_0_hw_fini(void *handle)
+static int gfx_v6_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gfx_v6_0_cp_enable(adev, false);
 	adev->gfx.rlc.funcs->stop(adev);
@@ -3153,23 +3156,19 @@ static int gfx_v6_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int gfx_v6_0_suspend(void *handle)
+static int gfx_v6_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return gfx_v6_0_hw_fini(adev);
+	return gfx_v6_0_hw_fini(ip_block);
 }
 
-static int gfx_v6_0_resume(void *handle)
+static int gfx_v6_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return gfx_v6_0_hw_init(adev);
+	return gfx_v6_0_hw_init(ip_block);
 }
 
-static bool gfx_v6_0_is_idle(void *handle)
+static bool gfx_v6_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (RREG32(mmGRBM_STATUS) & GRBM_STATUS__GUI_ACTIVE_MASK)
 		return false;
@@ -3177,24 +3176,19 @@ static bool gfx_v6_0_is_idle(void *handle)
 		return true;
 }
 
-static int gfx_v6_0_wait_for_idle(void *handle)
+static int gfx_v6_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (gfx_v6_0_is_idle(handle))
+		if (gfx_v6_0_is_idle(ip_block))
 			return 0;
 		udelay(1);
 	}
 	return -ETIMEDOUT;
 }
 
-static int gfx_v6_0_soft_reset(void *handle)
-{
-	return 0;
-}
-
 static void gfx_v6_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev,
 						 enum amdgpu_interrupt_state state)
 {
@@ -3382,11 +3376,11 @@ static int gfx_v6_0_priv_inst_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int gfx_v6_0_set_clockgating_state(void *handle,
+static int gfx_v6_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	bool gate = false;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_CG_STATE_GATE)
 		gate = true;
@@ -3404,11 +3398,11 @@ static int gfx_v6_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int gfx_v6_0_set_powergating_state(void *handle,
+static int gfx_v6_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	bool gate = false;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_PG_STATE_GATE)
 		gate = true;
@@ -3444,7 +3438,6 @@ static void gfx_v6_0_emit_mem_sync(struct amdgpu_ring *ring)
 static const struct amd_ip_funcs gfx_v6_0_ip_funcs = {
 	.name = "gfx_v6_0",
 	.early_init = gfx_v6_0_early_init,
-	.late_init = NULL,
 	.sw_init = gfx_v6_0_sw_init,
 	.sw_fini = gfx_v6_0_sw_fini,
 	.hw_init = gfx_v6_0_hw_init,
@@ -3453,7 +3446,6 @@ static const struct amd_ip_funcs gfx_v6_0_ip_funcs = {
 	.resume = gfx_v6_0_resume,
 	.is_idle = gfx_v6_0_is_idle,
 	.wait_for_idle = gfx_v6_0_wait_for_idle,
-	.soft_reset = gfx_v6_0_soft_reset,
 	.set_clockgating_state = gfx_v6_0_set_clockgating_state,
 	.set_powergating_state = gfx_v6_0_set_powergating_state,
 };
@@ -3571,12 +3563,12 @@ static void gfx_v6_0_get_cu_info(struct amdgpu_device *adev)
 			mask = 1;
 			ao_bitmap = 0;
 			counter = 0;
-			gfx_v6_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v6_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			if (i < 4 && j < 2)
 				gfx_v6_0_set_user_cu_inactive_bitmap(
 					adev, disable_masks[i * 2 + j]);
 			bitmap = gfx_v6_0_get_cu_enabled(adev);
-			cu_info->bitmap[i][j] = bitmap;
+			cu_info->bitmap[0][i][j] = bitmap;
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
 				if (bitmap & mask) {
@@ -3593,7 +3585,7 @@ static void gfx_v6_0_get_cu_info(struct amdgpu_device *adev)
 		}
 	}
 
-	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v6_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	cu_info->number = active_cu_number;
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
index 9d5c1e29b4a3..2b7aba22ecc1 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
@@ -55,6 +55,9 @@
 #define GFX7_NUM_GFX_RINGS     1
 #define GFX7_MEC_HPD_SIZE      2048
 
+#define BONAIRE_GB_ADDR_CONFIG_GOLDEN        0x12010001
+#define HAWAII_GB_ADDR_CONFIG_GOLDEN         0x12011003
+
 static void gfx_v7_0_set_ring_funcs(struct amdgpu_device *adev);
 static void gfx_v7_0_set_irq_funcs(struct amdgpu_device *adev);
 static void gfx_v7_0_set_gds_init(struct amdgpu_device *adev);
@@ -90,8 +93,7 @@ MODULE_FIRMWARE("amdgpu/mullins_ce.bin");
 MODULE_FIRMWARE("amdgpu/mullins_rlc.bin");
 MODULE_FIRMWARE("amdgpu/mullins_mec.bin");
 
-static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] =
-{
+static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] = {
 	{mmGDS_VMID0_BASE, mmGDS_VMID0_SIZE, mmGDS_GWS_VMID0, mmGDS_OA_VMID0},
 	{mmGDS_VMID1_BASE, mmGDS_VMID1_SIZE, mmGDS_GWS_VMID1, mmGDS_OA_VMID1},
 	{mmGDS_VMID2_BASE, mmGDS_VMID2_SIZE, mmGDS_GWS_VMID2, mmGDS_OA_VMID2},
@@ -110,8 +112,7 @@ static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] =
 	{mmGDS_VMID15_BASE, mmGDS_VMID15_SIZE, mmGDS_GWS_VMID15, mmGDS_OA_VMID15}
 };
 
-static const u32 spectre_rlc_save_restore_register_list[] =
-{
+static const u32 spectre_rlc_save_restore_register_list[] = {
 	(0x0e00 << 16) | (0xc12c >> 2),
 	0x00000000,
 	(0x0e00 << 16) | (0xc140 >> 2),
@@ -557,8 +558,7 @@ static const u32 spectre_rlc_save_restore_register_list[] =
 	(0x0e00 << 16) | (0x9600 >> 2),
 };
 
-static const u32 kalindi_rlc_save_restore_register_list[] =
-{
+static const u32 kalindi_rlc_save_restore_register_list[] = {
 	(0x0e00 << 16) | (0xc12c >> 2),
 	0x00000000,
 	(0x0e00 << 16) | (0xc140 >> 2),
@@ -883,7 +883,7 @@ static const u32 kalindi_rlc_save_restore_register_list[] =
 };
 
 static u32 gfx_v7_0_get_csb_size(struct amdgpu_device *adev);
-static void gfx_v7_0_get_csb_buffer(struct amdgpu_device *adev, volatile u32 *buffer);
+static void gfx_v7_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer);
 static void gfx_v7_0_init_pg(struct amdgpu_device *adev);
 static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev);
 
@@ -912,7 +912,6 @@ static void gfx_v7_0_free_microcode(struct amdgpu_device *adev)
 static int gfx_v7_0_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err;
 
 	DRM_DEBUG("\n");
@@ -933,43 +932,48 @@ static int gfx_v7_0_init_microcode(struct amdgpu_device *adev)
 	case CHIP_MULLINS:
 		chip_name = "mullins";
 		break;
-	default: BUG();
+	default:
+		BUG();
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_pfp.bin", chip_name);
 	if (err)
 		goto out;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_me.bin", chip_name);
 	if (err)
 		goto out;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_ce.bin", chip_name);
 	if (err)
 		goto out;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_mec.bin", chip_name);
 	if (err)
 		goto out;
 
 	if (adev->asic_type == CHIP_KAVERI) {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_mec2.bin", chip_name);
 		if (err)
 			goto out;
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, fw_name);
-	if (err)
-		goto out;
+	err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_rlc.bin", chip_name);
 out:
 	if (err) {
-		pr_err("gfx7: Failed to load firmware \"%s\"\n", fw_name);
+		pr_err("gfx7: Failed to load firmware %s gfx firmware\n", chip_name);
 		gfx_v7_0_free_microcode(adev);
 	}
 	return err;
@@ -1548,11 +1552,12 @@ static void gfx_v7_0_tiling_mode_table_init(struct amdgpu_device *adev)
  * @sh_num: sh block to address
  * @instance: Certain registers are instanced per SE or SH.
  *            0xffffffff means broadcast to all SEs or SHs (CIK).
- *
+ * @xcc_id: xcc accelerated compute core id
  * Select which SE, SH combinations to address.
  */
 static void gfx_v7_0_select_se_sh(struct amdgpu_device *adev,
-				  u32 se_num, u32 sh_num, u32 instance)
+				  u32 se_num, u32 sh_num, u32 instance,
+				  int xcc_id)
 {
 	u32 data;
 
@@ -1732,13 +1737,13 @@ gfx_v7_0_write_harvested_raster_configs(struct amdgpu_device *adev,
 		}
 
 		/* GRBM_GFX_INDEX has a different offset on CI+ */
-		gfx_v7_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff);
+		gfx_v7_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff, 0);
 		WREG32(mmPA_SC_RASTER_CONFIG, raster_config_se);
 		WREG32(mmPA_SC_RASTER_CONFIG_1, raster_config_1);
 	}
 
 	/* GRBM_GFX_INDEX has a different offset on CI+ */
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 }
 
 /**
@@ -1761,13 +1766,13 @@ static void gfx_v7_0_setup_rb(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			data = gfx_v7_0_get_rb_active_bitmap(adev);
 			active_rbs |= data << ((i * adev->gfx.config.max_sh_per_se + j) *
 					       rb_bitmap_width_per_sh);
 		}
 	}
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 
 	adev->gfx.config.backend_enable_mask = active_rbs;
 	adev->gfx.config.num_rbs = hweight32(active_rbs);
@@ -1790,7 +1795,7 @@ static void gfx_v7_0_setup_rb(struct amdgpu_device *adev)
 	/* cache the values for userspace */
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			adev->gfx.config.rb_config[i][j].rb_backend_disable =
 				RREG32(mmCC_RB_BACKEND_DISABLE);
 			adev->gfx.config.rb_config[i][j].user_rb_backend_disable =
@@ -1801,7 +1806,7 @@ static void gfx_v7_0_setup_rb(struct amdgpu_device *adev)
 				RREG32(mmPA_SC_RASTER_CONFIG_1);
 		}
 	}
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 }
 
@@ -1911,7 +1916,7 @@ static void gfx_v7_0_constants_init(struct amdgpu_device *adev)
 	 * making sure that the following register writes will be broadcasted
 	 * to all the shaders
 	 */
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 
 	/* XXX SH_MEM regs */
 	/* where to put LDS, scratch, GPUVM in FSA64 space */
@@ -2118,6 +2123,8 @@ static void gfx_v7_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr,
 {
 	bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT;
 	bool int_sel = flags & AMDGPU_FENCE_FLAG_INT;
+	bool exec = flags & AMDGPU_FENCE_FLAG_EXEC;
+
 	/* Workaround for cache flush problems. First send a dummy EOP
 	 * event down the pipe with seq one below.
 	 */
@@ -2137,7 +2144,8 @@ static void gfx_v7_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr,
 	amdgpu_ring_write(ring, (EOP_TCL1_ACTION_EN |
 				 EOP_TC_ACTION_EN |
 				 EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) |
-				 EVENT_INDEX(5)));
+				 EVENT_INDEX(5) |
+				 (exec ? EOP_EXEC : 0)));
 	amdgpu_ring_write(ring, addr & 0xfffffffc);
 	amdgpu_ring_write(ring, (upper_32_bits(addr) & 0xffff) |
 				DATA_SEL(write64bit ? 2 : 1) | INT_SEL(int_sel ? 2 : 0));
@@ -2325,7 +2333,7 @@ static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 error:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 	return r;
 }
@@ -2560,7 +2568,7 @@ static int gfx_v7_0_cp_gfx_resume(struct amdgpu_device *adev)
 	ring->wptr = 0;
 	WREG32(mmCP_RB0_WPTR, lower_32_bits(ring->wptr));
 
-	/* set the wb address wether it's enabled or not */
+	/* set the wb address whether it's enabled or not */
 	rptr_addr = ring->rptr_gpu_addr;
 	WREG32(mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr));
 	WREG32(mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & 0xFF);
@@ -2728,7 +2736,7 @@ static int gfx_v7_0_mec_init(struct amdgpu_device *adev)
 	u32 *hpd;
 	size_t mec_hpd_size;
 
-	bitmap_zero(adev->gfx.mec.queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
+	bitmap_zero(adev->gfx.mec_bitmap[0].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
 
 	/* take ownership of the relevant compute queues */
 	amdgpu_gfx_compute_queue_acquire(adev);
@@ -2758,45 +2766,6 @@ static int gfx_v7_0_mec_init(struct amdgpu_device *adev)
 	return 0;
 }
 
-struct hqd_registers
-{
-	u32 cp_mqd_base_addr;
-	u32 cp_mqd_base_addr_hi;
-	u32 cp_hqd_active;
-	u32 cp_hqd_vmid;
-	u32 cp_hqd_persistent_state;
-	u32 cp_hqd_pipe_priority;
-	u32 cp_hqd_queue_priority;
-	u32 cp_hqd_quantum;
-	u32 cp_hqd_pq_base;
-	u32 cp_hqd_pq_base_hi;
-	u32 cp_hqd_pq_rptr;
-	u32 cp_hqd_pq_rptr_report_addr;
-	u32 cp_hqd_pq_rptr_report_addr_hi;
-	u32 cp_hqd_pq_wptr_poll_addr;
-	u32 cp_hqd_pq_wptr_poll_addr_hi;
-	u32 cp_hqd_pq_doorbell_control;
-	u32 cp_hqd_pq_wptr;
-	u32 cp_hqd_pq_control;
-	u32 cp_hqd_ib_base_addr;
-	u32 cp_hqd_ib_base_addr_hi;
-	u32 cp_hqd_ib_rptr;
-	u32 cp_hqd_ib_control;
-	u32 cp_hqd_iq_timer;
-	u32 cp_hqd_iq_rptr;
-	u32 cp_hqd_dequeue_request;
-	u32 cp_hqd_dma_offload;
-	u32 cp_hqd_sema_cmd;
-	u32 cp_hqd_msg_type;
-	u32 cp_hqd_atomic0_preop_lo;
-	u32 cp_hqd_atomic0_preop_hi;
-	u32 cp_hqd_atomic1_preop_lo;
-	u32 cp_hqd_atomic1_preop_hi;
-	u32 cp_hqd_hq_scheduler0;
-	u32 cp_hqd_hq_scheduler1;
-	u32 cp_mqd_control;
-};
-
 static void gfx_v7_0_compute_pipe_init(struct amdgpu_device *adev,
 				       int mec, int pipe)
 {
@@ -2916,7 +2885,7 @@ static void gfx_v7_0_mqd_init(struct amdgpu_device *adev,
 	mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc;
 	mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff;
 
-	/* set the wb address wether it's enabled or not */
+	/* set the wb address whether it's enabled or not */
 	wb_gpu_addr = ring->rptr_gpu_addr;
 	mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc;
 	mqd->cp_hqd_pq_rptr_report_addr_hi =
@@ -3276,7 +3245,7 @@ static int gfx_v7_0_rlc_init(struct amdgpu_device *adev)
 
 	/* init spm vmid with 0xf */
 	if (adev->gfx.rlc.funcs->update_spm_vmid)
-		adev->gfx.rlc.funcs->update_spm_vmid(adev, 0xf);
+		adev->gfx.rlc.funcs->update_spm_vmid(adev, NULL, 0xf);
 
 	return 0;
 }
@@ -3301,7 +3270,7 @@ static void gfx_v7_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			for (k = 0; k < adev->usec_timeout; k++) {
 				if (RREG32(mmRLC_SERDES_CU_MASTER_BUSY) == 0)
 					break;
@@ -3309,7 +3278,7 @@ static void gfx_v7_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 			}
 		}
 	}
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	mask = RLC_SERDES_NONCU_MASTER_BUSY__SE_MASTER_BUSY_MASK |
@@ -3361,7 +3330,7 @@ static bool gfx_v7_0_is_rlc_enabled(struct amdgpu_device *adev)
 	return true;
 }
 
-static void gfx_v7_0_set_safe_mode(struct amdgpu_device *adev)
+static void gfx_v7_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	u32 tmp, i, mask;
 
@@ -3383,7 +3352,7 @@ static void gfx_v7_0_set_safe_mode(struct amdgpu_device *adev)
 	}
 }
 
-static void gfx_v7_0_unset_safe_mode(struct amdgpu_device *adev)
+static void gfx_v7_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	u32 tmp;
 
@@ -3474,7 +3443,7 @@ static int gfx_v7_0_rlc_resume(struct amdgpu_device *adev)
 	WREG32(mmRLC_LB_CNTR_MAX, 0x00008000);
 
 	mutex_lock(&adev->grbm_idx_mutex);
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	WREG32(mmRLC_LB_INIT_CU_MASK, 0xffffffff);
 	WREG32(mmRLC_LB_PARAMS, 0x00600408);
 	WREG32(mmRLC_LB_CNTL, 0x80000004);
@@ -3502,7 +3471,7 @@ static int gfx_v7_0_rlc_resume(struct amdgpu_device *adev)
 	return 0;
 }
 
-static void gfx_v7_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid)
+static void gfx_v7_0_update_spm_vmid(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned vmid)
 {
 	u32 data;
 
@@ -3530,7 +3499,7 @@ static void gfx_v7_0_enable_cgcg(struct amdgpu_device *adev, bool enable)
 		tmp = gfx_v7_0_halt_rlc(adev);
 
 		mutex_lock(&adev->grbm_idx_mutex);
-		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 		WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff);
 		WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff);
 		tmp2 = RLC_SERDES_WR_CTRL__BPM_ADDR_MASK |
@@ -3584,7 +3553,7 @@ static void gfx_v7_0_enable_mgcg(struct amdgpu_device *adev, bool enable)
 		tmp = gfx_v7_0_halt_rlc(adev);
 
 		mutex_lock(&adev->grbm_idx_mutex);
-		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 		WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff);
 		WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff);
 		data = RLC_SERDES_WR_CTRL__BPM_ADDR_MASK |
@@ -3635,7 +3604,7 @@ static void gfx_v7_0_enable_mgcg(struct amdgpu_device *adev, bool enable)
 		tmp = gfx_v7_0_halt_rlc(adev);
 
 		mutex_lock(&adev->grbm_idx_mutex);
-		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 		WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff);
 		WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff);
 		data = RLC_SERDES_WR_CTRL__BPM_ADDR_MASK | RLC_SERDES_WR_CTRL__MGCG_OVERRIDE_1_MASK;
@@ -3913,70 +3882,24 @@ static u32 gfx_v7_0_get_csb_size(struct amdgpu_device *adev)
 	return count;
 }
 
-static void gfx_v7_0_get_csb_buffer(struct amdgpu_device *adev,
-				    volatile u32 *buffer)
+static void gfx_v7_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer)
 {
-	u32 count = 0, i;
-	const struct cs_section_def *sect = NULL;
-	const struct cs_extent_def *ext = NULL;
+	u32 count = 0;
 
 	if (adev->gfx.rlc.cs_data == NULL)
 		return;
 	if (buffer == NULL)
 		return;
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1));
-	buffer[count++] = cpu_to_le32(0x80000000);
-	buffer[count++] = cpu_to_le32(0x80000000);
-
-	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
-		for (ext = sect->section; ext->extent != NULL; ++ext) {
-			if (sect->id == SECT_CONTEXT) {
-				buffer[count++] =
-					cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
-				buffer[count++] = cpu_to_le32(ext->reg_index - PACKET3_SET_CONTEXT_REG_START);
-				for (i = 0; i < ext->reg_count; i++)
-					buffer[count++] = cpu_to_le32(ext->extent[i]);
-			} else {
-				return;
-			}
-		}
-	}
+	count = amdgpu_gfx_csb_preamble_start(buffer);
+	count = amdgpu_gfx_csb_data_parser(adev, buffer, count);
 
 	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, 2));
 	buffer[count++] = cpu_to_le32(mmPA_SC_RASTER_CONFIG - PACKET3_SET_CONTEXT_REG_START);
-	switch (adev->asic_type) {
-	case CHIP_BONAIRE:
-		buffer[count++] = cpu_to_le32(0x16000012);
-		buffer[count++] = cpu_to_le32(0x00000000);
-		break;
-	case CHIP_KAVERI:
-		buffer[count++] = cpu_to_le32(0x00000000); /* XXX */
-		buffer[count++] = cpu_to_le32(0x00000000);
-		break;
-	case CHIP_KABINI:
-	case CHIP_MULLINS:
-		buffer[count++] = cpu_to_le32(0x00000000); /* XXX */
-		buffer[count++] = cpu_to_le32(0x00000000);
-		break;
-	case CHIP_HAWAII:
-		buffer[count++] = cpu_to_le32(0x3a00161a);
-		buffer[count++] = cpu_to_le32(0x0000002e);
-		break;
-	default:
-		buffer[count++] = cpu_to_le32(0x00000000);
-		buffer[count++] = cpu_to_le32(0x00000000);
-		break;
-	}
+	buffer[count++] = cpu_to_le32(adev->gfx.config.rb_config[0][0].raster_config);
+	buffer[count++] = cpu_to_le32(adev->gfx.config.rb_config[0][0].raster_config_1);
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0));
-	buffer[count++] = cpu_to_le32(0);
+	amdgpu_gfx_csb_preamble_end(buffer, count);
 }
 
 static void gfx_v7_0_init_pg(struct amdgpu_device *adev)
@@ -4111,7 +4034,7 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd,
 		*(out++) = RREG32(mmSQ_IND_DATA);
 }
 
-static void gfx_v7_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
+static void gfx_v7_0_read_wave_data(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
 {
 	/* type 0 wave data */
 	dst[(*no_fields)++] = 0;
@@ -4136,7 +4059,7 @@ static void gfx_v7_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, u
 	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_MODE);
 }
 
-static void gfx_v7_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v7_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				     uint32_t wave, uint32_t start,
 				     uint32_t size, uint32_t *dst)
 {
@@ -4146,7 +4069,7 @@ static void gfx_v7_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
 }
 
 static void gfx_v7_0_select_me_pipe_q(struct amdgpu_device *adev,
-				  u32 me, u32 pipe, u32 q, u32 vm)
+				  u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
 {
 	cik_srbm_select(adev, me, pipe, q, vm);
 }
@@ -4174,10 +4097,11 @@ static const struct amdgpu_rlc_funcs gfx_v7_0_rlc_funcs = {
 	.update_spm_vmid = gfx_v7_0_update_spm_vmid
 };
 
-static int gfx_v7_0_early_init(void *handle)
+static int gfx_v7_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
+	adev->gfx.xcc_mask = 1;
 	adev->gfx.num_gfx_rings = GFX7_NUM_GFX_RINGS;
 	adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
 					  AMDGPU_MAX_COMPUTE_RINGS);
@@ -4190,9 +4114,9 @@ static int gfx_v7_0_early_init(void *handle)
 	return 0;
 }
 
-static int gfx_v7_0_late_init(void *handle)
+static int gfx_v7_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0);
@@ -4382,10 +4306,10 @@ static int gfx_v7_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 	return 0;
 }
 
-static int gfx_v7_0_sw_init(void *handle)
+static int gfx_v7_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	struct amdgpu_ring *ring;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i, j, k, r, ring_id;
 
 	switch (adev->asic_type) {
@@ -4456,7 +4380,8 @@ static int gfx_v7_0_sw_init(void *handle)
 	for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
 		for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
 			for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
-				if (!amdgpu_gfx_is_mec_queue_enabled(adev, i, k, j))
+				if (!amdgpu_gfx_is_mec_queue_enabled(adev, 0, i,
+								     k, j))
 					continue;
 
 				r = gfx_v7_0_compute_ring_init(adev,
@@ -4474,12 +4399,17 @@ static int gfx_v7_0_sw_init(void *handle)
 
 	gfx_v7_0_gpu_early_init(adev);
 
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+
 	return r;
 }
 
-static int gfx_v7_0_sw_fini(void *handle)
+static int gfx_v7_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
@@ -4503,10 +4433,10 @@ static int gfx_v7_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int gfx_v7_0_hw_init(void *handle)
+static int gfx_v7_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gfx_v7_0_constants_init(adev);
 
@@ -4524,9 +4454,9 @@ static int gfx_v7_0_hw_init(void *handle)
 	return r;
 }
 
-static int gfx_v7_0_hw_fini(void *handle)
+static int gfx_v7_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
@@ -4537,23 +4467,19 @@ static int gfx_v7_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int gfx_v7_0_suspend(void *handle)
+static int gfx_v7_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return gfx_v7_0_hw_fini(adev);
+	return gfx_v7_0_hw_fini(ip_block);
 }
 
-static int gfx_v7_0_resume(void *handle)
+static int gfx_v7_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return gfx_v7_0_hw_init(adev);
+	return gfx_v7_0_hw_init(ip_block);
 }
 
-static bool gfx_v7_0_is_idle(void *handle)
+static bool gfx_v7_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (RREG32(mmGRBM_STATUS) & GRBM_STATUS__GUI_ACTIVE_MASK)
 		return false;
@@ -4561,11 +4487,11 @@ static bool gfx_v7_0_is_idle(void *handle)
 		return true;
 }
 
-static int gfx_v7_0_wait_for_idle(void *handle)
+static int gfx_v7_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		/* read MC_STATUS */
@@ -4578,11 +4504,11 @@ static int gfx_v7_0_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int gfx_v7_0_soft_reset(void *handle)
+static int gfx_v7_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 grbm_soft_reset = 0, srbm_soft_reset = 0;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* GRBM_STATUS */
 	tmp = RREG32(mmGRBM_STATUS);
@@ -4888,11 +4814,11 @@ static int gfx_v7_0_priv_inst_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int gfx_v7_0_set_clockgating_state(void *handle,
+static int gfx_v7_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	bool gate = false;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_CG_STATE_GATE)
 		gate = true;
@@ -4911,11 +4837,11 @@ static int gfx_v7_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int gfx_v7_0_set_powergating_state(void *handle,
+static int gfx_v7_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	bool gate = false;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_PG_STATE_GATE)
 		gate = true;
@@ -5041,6 +4967,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = {
 	.insert_nop = amdgpu_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_wreg = gfx_v7_0_ring_emit_wreg,
+	.soft_recovery = gfx_v7_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v7_0_emit_mem_sync_compute,
 };
 
@@ -5114,18 +5041,18 @@ static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev)
 			mask = 1;
 			ao_bitmap = 0;
 			counter = 0;
-			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v7_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			if (i < 4 && j < 2)
 				gfx_v7_0_set_user_cu_inactive_bitmap(
 					adev, disable_masks[i * 2 + j]);
 			bitmap = gfx_v7_0_get_cu_active_bitmap(adev);
-			cu_info->bitmap[i][j] = bitmap;
+			cu_info->bitmap[0][i][j] = bitmap;
 
-			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) {
+			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
 				if (bitmap & mask) {
 					if (counter < ao_cu_num)
 						ao_bitmap |= mask;
-					counter ++;
+					counter++;
 				}
 				mask <<= 1;
 			}
@@ -5135,7 +5062,7 @@ static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev)
 			cu_info->ao_cu_bitmap[i][j] = ao_bitmap;
 		}
 	}
-	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	cu_info->number = active_cu_number;
@@ -5147,8 +5074,7 @@ static void gfx_v7_0_get_cu_info(struct amdgpu_device *adev)
 	cu_info->lds_size = 64;
 }
 
-const struct amdgpu_ip_block_version gfx_v7_1_ip_block =
-{
+const struct amdgpu_ip_block_version gfx_v7_1_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GFX,
 	.major = 7,
 	.minor = 1,
@@ -5156,8 +5082,7 @@ const struct amdgpu_ip_block_version gfx_v7_1_ip_block =
 	.funcs = &gfx_v7_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version gfx_v7_2_ip_block =
-{
+const struct amdgpu_ip_block_version gfx_v7_2_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GFX,
 	.major = 7,
 	.minor = 2,
@@ -5165,8 +5090,7 @@ const struct amdgpu_ip_block_version gfx_v7_2_ip_block =
 	.funcs = &gfx_v7_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version gfx_v7_3_ip_block =
-{
+const struct amdgpu_ip_block_version gfx_v7_3_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GFX,
 	.major = 7,
 	.minor = 3,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
index b1f2684d854a..1c87375e1dd5 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
@@ -883,8 +883,8 @@ static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	gpu_addr = adev->wb.gpu_addr + (index * 4);
 	adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
 	memset(&ib, 0, sizeof(ib));
-	r = amdgpu_ib_get(adev, NULL, 16,
-					AMDGPU_IB_POOL_DIRECT, &ib);
+
+	r = amdgpu_ib_get(adev, NULL, 20, AMDGPU_IB_POOL_DIRECT, &ib);
 	if (r)
 		goto err1;
 
@@ -914,7 +914,7 @@ static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 err2:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err1:
 	amdgpu_device_wb_free(adev, index);
@@ -939,7 +939,6 @@ static void gfx_v8_0_free_microcode(struct amdgpu_device *adev)
 static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err;
 	struct amdgpu_firmware_info *info = NULL;
 	const struct common_firmware_header *header = NULL;
@@ -982,15 +981,18 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 	}
 
 	if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp_2.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+					   AMDGPU_UCODE_OPTIONAL,
+					   "amdgpu/%s_pfp_2.bin", chip_name);
 		if (err == -ENODEV) {
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name);
-			err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_pfp.bin", chip_name);
 		}
 	} else {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_pfp.bin", chip_name);
 	}
 	if (err)
 		goto out;
@@ -999,15 +1001,18 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 	adev->gfx.pfp_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
 	if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me_2.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+					   AMDGPU_UCODE_OPTIONAL,
+					   "amdgpu/%s_me_2.bin", chip_name);
 		if (err == -ENODEV) {
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
-			err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_me.bin", chip_name);
 		}
 	} else {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_me.bin", chip_name);
 	}
 	if (err)
 		goto out;
@@ -1017,15 +1022,18 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 	adev->gfx.me_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version);
 
 	if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce_2.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw,
+					   AMDGPU_UCODE_OPTIONAL,
+					   "amdgpu/%s_ce_2.bin", chip_name);
 		if (err == -ENODEV) {
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
-			err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_ce.bin", chip_name);
 		}
 	} else {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_ce.bin", chip_name);
 	}
 	if (err)
 		goto out;
@@ -1044,8 +1052,9 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 	} else
 		adev->virt.chained_ib_support = false;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_rlc.bin", chip_name);
 	if (err)
 		goto out;
 	rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
@@ -1093,15 +1102,18 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 		adev->gfx.rlc.register_restore[i] = le32_to_cpu(tmp[i]);
 
 	if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec_2.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+					   AMDGPU_UCODE_OPTIONAL,
+					   "amdgpu/%s_mec_2.bin", chip_name);
 		if (err == -ENODEV) {
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
-			err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_mec.bin", chip_name);
 		}
 	} else {
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
-		err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_mec.bin", chip_name);
 	}
 	if (err)
 		goto out;
@@ -1112,15 +1124,18 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 	if ((adev->asic_type != CHIP_STONEY) &&
 	    (adev->asic_type != CHIP_TOPAZ)) {
 		if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) {
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2_2.bin", chip_name);
-			err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw,
+						   AMDGPU_UCODE_OPTIONAL,
+						   "amdgpu/%s_mec2_2.bin", chip_name);
 			if (err == -ENODEV) {
-				snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name);
-				err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, fw_name);
+				err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw,
+							   AMDGPU_UCODE_REQUIRED,
+							   "amdgpu/%s_mec2.bin", chip_name);
 			}
 		} else {
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name);
-			err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_mec2.bin", chip_name);
 		}
 		if (!err) {
 			cp_hdr = (const struct gfx_firmware_header_v1_0 *)
@@ -1194,9 +1209,7 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev)
 
 out:
 	if (err) {
-		dev_err(adev->dev,
-			"gfx8: Failed to load firmware \"%s\"\n",
-			fw_name);
+		dev_err(adev->dev, "gfx8: Failed to load firmware %s gfx firmware\n", chip_name);
 		amdgpu_ucode_release(&adev->gfx.pfp_fw);
 		amdgpu_ucode_release(&adev->gfx.me_fw);
 		amdgpu_ucode_release(&adev->gfx.ce_fw);
@@ -1207,51 +1220,24 @@ out:
 	return err;
 }
 
-static void gfx_v8_0_get_csb_buffer(struct amdgpu_device *adev,
-				    volatile u32 *buffer)
+static void gfx_v8_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer)
 {
-	u32 count = 0, i;
-	const struct cs_section_def *sect = NULL;
-	const struct cs_extent_def *ext = NULL;
+	u32 count = 0;
 
 	if (adev->gfx.rlc.cs_data == NULL)
 		return;
 	if (buffer == NULL)
 		return;
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1));
-	buffer[count++] = cpu_to_le32(0x80000000);
-	buffer[count++] = cpu_to_le32(0x80000000);
-
-	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
-		for (ext = sect->section; ext->extent != NULL; ++ext) {
-			if (sect->id == SECT_CONTEXT) {
-				buffer[count++] =
-					cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
-				buffer[count++] = cpu_to_le32(ext->reg_index -
-						PACKET3_SET_CONTEXT_REG_START);
-				for (i = 0; i < ext->reg_count; i++)
-					buffer[count++] = cpu_to_le32(ext->extent[i]);
-			} else {
-				return;
-			}
-		}
-	}
+	count = amdgpu_gfx_csb_preamble_start(buffer);
+	count = amdgpu_gfx_csb_data_parser(adev, buffer, count);
 
 	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, 2));
-	buffer[count++] = cpu_to_le32(mmPA_SC_RASTER_CONFIG -
-			PACKET3_SET_CONTEXT_REG_START);
+	buffer[count++] = cpu_to_le32(mmPA_SC_RASTER_CONFIG - PACKET3_SET_CONTEXT_REG_START);
 	buffer[count++] = cpu_to_le32(adev->gfx.config.rb_config[0][0].raster_config);
 	buffer[count++] = cpu_to_le32(adev->gfx.config.rb_config[0][0].raster_config_1);
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0));
-	buffer[count++] = cpu_to_le32(0);
+	amdgpu_gfx_csb_preamble_end(buffer, count);
 }
 
 static int gfx_v8_0_cp_jump_table_num(struct amdgpu_device *adev)
@@ -1288,7 +1274,7 @@ static int gfx_v8_0_rlc_init(struct amdgpu_device *adev)
 
 	/* init spm vmid with 0xf */
 	if (adev->gfx.rlc.funcs->update_spm_vmid)
-		adev->gfx.rlc.funcs->update_spm_vmid(adev, 0xf);
+		adev->gfx.rlc.funcs->update_spm_vmid(adev, NULL, 0xf);
 
 	return 0;
 }
@@ -1304,7 +1290,7 @@ static int gfx_v8_0_mec_init(struct amdgpu_device *adev)
 	u32 *hpd;
 	size_t mec_hpd_size;
 
-	bitmap_zero(adev->gfx.mec.queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
+	bitmap_zero(adev->gfx.mec_bitmap[0].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
 
 	/* take ownership of the relevant compute queues */
 	amdgpu_gfx_compute_queue_acquire(adev);
@@ -1643,7 +1629,7 @@ static int gfx_v8_0_do_edc_gpr_workarounds(struct amdgpu_device *adev)
 		RREG32(sec_ded_counter_registers[i]);
 
 fail:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 
 	return r;
@@ -1897,12 +1883,12 @@ static int gfx_v8_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 
 static void gfx_v8_0_sq_irq_work_func(struct work_struct *work);
 
-static int gfx_v8_0_sw_init(void *handle)
+static int gfx_v8_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int i, j, k, r, ring_id;
+	int xcc_id = 0;
 	struct amdgpu_ring *ring;
-	struct amdgpu_kiq *kiq;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	switch (adev->asic_type) {
 	case CHIP_TONGA:
@@ -2001,7 +1987,8 @@ static int gfx_v8_0_sw_init(void *handle)
 	for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
 		for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
 			for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
-				if (!amdgpu_gfx_is_mec_queue_enabled(adev, i, k, j))
+				if (!amdgpu_gfx_is_mec_queue_enabled(adev, 0, i,
+								     k, j))
 					continue;
 
 				r = gfx_v8_0_compute_ring_init(adev,
@@ -2015,19 +2002,18 @@ static int gfx_v8_0_sw_init(void *handle)
 		}
 	}
 
-	r = amdgpu_gfx_kiq_init(adev, GFX8_MEC_HPD_SIZE);
+	r = amdgpu_gfx_kiq_init(adev, GFX8_MEC_HPD_SIZE, 0);
 	if (r) {
 		DRM_ERROR("Failed to init KIQ BOs!\n");
 		return r;
 	}
 
-	kiq = &adev->gfx.kiq;
-	r = amdgpu_gfx_kiq_init_ring(adev, &kiq->ring, &kiq->irq);
+	r = amdgpu_gfx_kiq_init_ring(adev, xcc_id);
 	if (r)
 		return r;
 
 	/* create MQD for all compute queues as well as KIQ for SRIOV case */
-	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct vi_mqd_allocation));
+	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct vi_mqd_allocation), 0);
 	if (r)
 		return r;
 
@@ -2037,12 +2023,17 @@ static int gfx_v8_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+
 	return 0;
 }
 
-static int gfx_v8_0_sw_fini(void *handle)
+static int gfx_v8_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int i;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
@@ -2050,9 +2041,9 @@ static int gfx_v8_0_sw_fini(void *handle)
 	for (i = 0; i < adev->gfx.num_compute_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
 
-	amdgpu_gfx_mqd_sw_fini(adev);
-	amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring);
-	amdgpu_gfx_kiq_fini(adev);
+	amdgpu_gfx_mqd_sw_fini(adev, 0);
+	amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq[0].ring);
+	amdgpu_gfx_kiq_fini(adev, 0);
 
 	gfx_v8_0_mec_fini(adev);
 	amdgpu_gfx_rlc_fini(adev);
@@ -3394,7 +3385,8 @@ static void gfx_v8_0_tiling_mode_table_init(struct amdgpu_device *adev)
 }
 
 static void gfx_v8_0_select_se_sh(struct amdgpu_device *adev,
-				  u32 se_num, u32 sh_num, u32 instance)
+				  u32 se_num, u32 sh_num, u32 instance,
+				  int xcc_id)
 {
 	u32 data;
 
@@ -3417,7 +3409,7 @@ static void gfx_v8_0_select_se_sh(struct amdgpu_device *adev,
 }
 
 static void gfx_v8_0_select_me_pipe_q(struct amdgpu_device *adev,
-				  u32 me, u32 pipe, u32 q, u32 vm)
+				  u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
 {
 	vi_srbm_select(adev, me, pipe, q, vm);
 }
@@ -3578,13 +3570,13 @@ gfx_v8_0_write_harvested_raster_configs(struct amdgpu_device *adev,
 		}
 
 		/* GRBM_GFX_INDEX has a different offset on VI */
-		gfx_v8_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff);
+		gfx_v8_0_select_se_sh(adev, se, 0xffffffff, 0xffffffff, 0);
 		WREG32(mmPA_SC_RASTER_CONFIG, raster_config_se);
 		WREG32(mmPA_SC_RASTER_CONFIG_1, raster_config_1);
 	}
 
 	/* GRBM_GFX_INDEX has a different offset on VI */
-	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 }
 
 static void gfx_v8_0_setup_rb(struct amdgpu_device *adev)
@@ -3600,13 +3592,13 @@ static void gfx_v8_0_setup_rb(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			data = gfx_v8_0_get_rb_active_bitmap(adev);
 			active_rbs |= data << ((i * adev->gfx.config.max_sh_per_se + j) *
 					       rb_bitmap_width_per_sh);
 		}
 	}
-	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 
 	adev->gfx.config.backend_enable_mask = active_rbs;
 	adev->gfx.config.num_rbs = hweight32(active_rbs);
@@ -3629,7 +3621,7 @@ static void gfx_v8_0_setup_rb(struct amdgpu_device *adev)
 	/* cache the values for userspace */
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			adev->gfx.config.rb_config[i][j].rb_backend_disable =
 				RREG32(mmCC_RB_BACKEND_DISABLE);
 			adev->gfx.config.rb_config[i][j].user_rb_backend_disable =
@@ -3640,7 +3632,7 @@ static void gfx_v8_0_setup_rb(struct amdgpu_device *adev)
 				RREG32(mmPA_SC_RASTER_CONFIG_1);
 		}
 	}
-	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 }
 
@@ -3787,7 +3779,7 @@ static void gfx_v8_0_constants_init(struct amdgpu_device *adev)
 	 * making sure that the following register writes will be broadcasted
 	 * to all the shaders
 	 */
-	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 
 	WREG32(mmPA_SC_FIFO_SIZE,
 		   (adev->gfx.config.sc_prim_fifo_size_frontend <<
@@ -3818,7 +3810,7 @@ static void gfx_v8_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			for (k = 0; k < adev->usec_timeout; k++) {
 				if (RREG32(mmRLC_SERDES_CU_MASTER_BUSY) == 0)
 					break;
@@ -3826,7 +3818,7 @@ static void gfx_v8_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 			}
 			if (k == adev->usec_timeout) {
 				gfx_v8_0_select_se_sh(adev, 0xffffffff,
-						      0xffffffff, 0xffffffff);
+						      0xffffffff, 0xffffffff, 0);
 				mutex_unlock(&adev->grbm_idx_mutex);
 				DRM_INFO("Timeout wait for RLC serdes %u,%u\n",
 					 i, j);
@@ -3834,7 +3826,7 @@ static void gfx_v8_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 			}
 		}
 	}
-	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	mask = RLC_SERDES_NONCU_MASTER_BUSY__SE_MASTER_BUSY_MASK |
@@ -4262,7 +4254,7 @@ static int gfx_v8_0_cp_gfx_resume(struct amdgpu_device *adev)
 	ring->wptr = 0;
 	WREG32(mmCP_RB0_WPTR, lower_32_bits(ring->wptr));
 
-	/* set the wb address wether it's enabled or not */
+	/* set the wb address whether it's enabled or not */
 	rptr_addr = ring->rptr_gpu_addr;
 	WREG32(mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr));
 	WREG32(mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & 0xFF);
@@ -4281,7 +4273,6 @@ static int gfx_v8_0_cp_gfx_resume(struct amdgpu_device *adev)
 	/* start the ring */
 	amdgpu_ring_clear_ring(ring);
 	gfx_v8_0_cp_gfx_start(adev);
-	ring->sched.ready = true;
 
 	return 0;
 }
@@ -4292,7 +4283,7 @@ static void gfx_v8_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
 		WREG32(mmCP_MEC_CNTL, 0);
 	} else {
 		WREG32(mmCP_MEC_CNTL, (CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK));
-		adev->gfx.kiq.ring.sched.ready = false;
+		adev->gfx.kiq[0].ring.sched.ready = false;
 	}
 	udelay(50);
 }
@@ -4307,19 +4298,17 @@ static void gfx_v8_0_kiq_setting(struct amdgpu_ring *ring)
 	tmp = RREG32(mmRLC_CP_SCHEDULERS);
 	tmp &= 0xffffff00;
 	tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
-	WREG32(mmRLC_CP_SCHEDULERS, tmp);
-	tmp |= 0x80;
-	WREG32(mmRLC_CP_SCHEDULERS, tmp);
+	WREG32(mmRLC_CP_SCHEDULERS, tmp | 0x80);
 }
 
 static int gfx_v8_0_kiq_kcq_enable(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[0].ring;
 	uint64_t queue_mask = 0;
 	int r, i;
 
 	for (i = 0; i < AMDGPU_MAX_COMPUTE_QUEUES; ++i) {
-		if (!test_bit(i, adev->gfx.mec.queue_bitmap))
+		if (!test_bit(i, adev->gfx.mec_bitmap[0].queue_bitmap))
 			continue;
 
 		/* This situation may be hit in the future if a new HW
@@ -4595,14 +4584,13 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct vi_mqd *mqd = ring->mqd_ptr;
-	int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS;
 
 	gfx_v8_0_kiq_setting(ring);
 
 	if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
 		/* reset MQD to a clean status */
-		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct vi_mqd_allocation));
+		if (adev->gfx.kiq[0].mqd_backup)
+			memcpy(mqd, adev->gfx.kiq[0].mqd_backup, sizeof(struct vi_mqd_allocation));
 
 		/* reset ring buffer */
 		ring->wptr = 0;
@@ -4625,8 +4613,8 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring)
 		vi_srbm_select(adev, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 
-		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct vi_mqd_allocation));
+		if (adev->gfx.kiq[0].mqd_backup)
+			memcpy(adev->gfx.kiq[0].mqd_backup, mqd, sizeof(struct vi_mqd_allocation));
 	}
 
 	return 0;
@@ -4650,14 +4638,13 @@ static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring)
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
 			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct vi_mqd_allocation));
-	} else if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
-		/* reset MQD to a clean status */
+	} else {
+		/* restore MQD to a clean status */
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
 			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct vi_mqd_allocation));
 		/* reset ring buffer */
 		ring->wptr = 0;
-		amdgpu_ring_clear_ring(ring);
-	} else {
+		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, 0);
 		amdgpu_ring_clear_ring(ring);
 	}
 	return 0;
@@ -4675,59 +4662,25 @@ static void gfx_v8_0_set_mec_doorbell_range(struct amdgpu_device *adev)
 
 static int gfx_v8_0_kiq_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring;
-	int r;
-
-	ring = &adev->gfx.kiq.ring;
-
-	r = amdgpu_bo_reserve(ring->mqd_obj, false);
-	if (unlikely(r != 0))
-		return r;
-
-	r = amdgpu_bo_kmap(ring->mqd_obj, &ring->mqd_ptr);
-	if (unlikely(r != 0))
-		return r;
-
-	gfx_v8_0_kiq_init_queue(ring);
-	amdgpu_bo_kunmap(ring->mqd_obj);
-	ring->mqd_ptr = NULL;
-	amdgpu_bo_unreserve(ring->mqd_obj);
-	ring->sched.ready = true;
+	gfx_v8_0_kiq_init_queue(&adev->gfx.kiq[0].ring);
 	return 0;
 }
 
 static int gfx_v8_0_kcq_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring = NULL;
-	int r = 0, i;
+	int i, r;
 
 	gfx_v8_0_cp_compute_enable(adev, true);
 
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-		ring = &adev->gfx.compute_ring[i];
-
-		r = amdgpu_bo_reserve(ring->mqd_obj, false);
-		if (unlikely(r != 0))
-			goto done;
-		r = amdgpu_bo_kmap(ring->mqd_obj, &ring->mqd_ptr);
-		if (!r) {
-			r = gfx_v8_0_kcq_init_queue(ring);
-			amdgpu_bo_kunmap(ring->mqd_obj);
-			ring->mqd_ptr = NULL;
-		}
-		amdgpu_bo_unreserve(ring->mqd_obj);
+		r = gfx_v8_0_kcq_init_queue(&adev->gfx.compute_ring[i]);
 		if (r)
-			goto done;
+			return r;
 	}
 
 	gfx_v8_0_set_mec_doorbell_range(adev);
 
-	r = gfx_v8_0_kiq_kcq_enable(adev);
-	if (r)
-		goto done;
-
-done:
-	return r;
+	return gfx_v8_0_kiq_kcq_enable(adev);
 }
 
 static int gfx_v8_0_cp_test_all_rings(struct amdgpu_device *adev)
@@ -4741,7 +4694,7 @@ static int gfx_v8_0_cp_test_all_rings(struct amdgpu_device *adev)
 	if (r)
 		return r;
 
-	ring = &adev->gfx.kiq.ring;
+	ring = &adev->gfx.kiq[0].ring;
 	r = amdgpu_ring_test_helper(ring);
 	if (r)
 		return r;
@@ -4788,10 +4741,10 @@ static void gfx_v8_0_cp_enable(struct amdgpu_device *adev, bool enable)
 	gfx_v8_0_cp_compute_enable(adev, enable);
 }
 
-static int gfx_v8_0_hw_init(void *handle)
+static int gfx_v8_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gfx_v8_0_init_golden_registers(adev);
 	gfx_v8_0_constants_init(adev);
@@ -4808,7 +4761,7 @@ static int gfx_v8_0_hw_init(void *handle)
 static int gfx_v8_0_kcq_disable(struct amdgpu_device *adev)
 {
 	int r, i;
-	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring;
+	struct amdgpu_ring *kiq_ring = &adev->gfx.kiq[0].ring;
 
 	r = amdgpu_ring_alloc(kiq_ring, 6 * adev->gfx.num_compute_rings);
 	if (r)
@@ -4828,6 +4781,13 @@ static int gfx_v8_0_kcq_disable(struct amdgpu_device *adev)
 		amdgpu_ring_write(kiq_ring, 0);
 		amdgpu_ring_write(kiq_ring, 0);
 	}
+	/* Submit unmap queue packet */
+	amdgpu_ring_commit(kiq_ring);
+	/*
+	 * Ring test will do a basic scratch register change check. Just run
+	 * this to ensure that unmap queues that is submitted before got
+	 * processed successfully before returning.
+	 */
 	r = amdgpu_ring_test_helper(kiq_ring);
 	if (r)
 		DRM_ERROR("KCQ disable failed\n");
@@ -4835,9 +4795,9 @@ static int gfx_v8_0_kcq_disable(struct amdgpu_device *adev)
 	return r;
 }
 
-static bool gfx_v8_0_is_idle(void *handle)
+static bool gfx_v8_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (REG_GET_FIELD(RREG32(mmGRBM_STATUS), GRBM_STATUS, GUI_ACTIVE)
 		|| RREG32(mmGRBM_STATUS2) != 0x8)
@@ -4870,13 +4830,13 @@ static int gfx_v8_0_wait_for_rlc_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int gfx_v8_0_wait_for_idle(void *handle)
+static int gfx_v8_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (gfx_v8_0_is_idle(handle))
+		if (gfx_v8_0_is_idle(ip_block))
 			return 0;
 
 		udelay(1);
@@ -4884,9 +4844,9 @@ static int gfx_v8_0_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int gfx_v8_0_hw_fini(void *handle)
+static int gfx_v8_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
@@ -4902,8 +4862,9 @@ static int gfx_v8_0_hw_fini(void *handle)
 		pr_debug("For SRIOV client, shouldn't do anything.\n");
 		return 0;
 	}
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
-	if (!gfx_v8_0_wait_for_idle(adev))
+
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+	if (!gfx_v8_0_wait_for_idle(ip_block))
 		gfx_v8_0_cp_enable(adev, false);
 	else
 		pr_err("cp is busy, skip halt cp\n");
@@ -4911,24 +4872,24 @@ static int gfx_v8_0_hw_fini(void *handle)
 		adev->gfx.rlc.funcs->stop(adev);
 	else
 		pr_err("rlc is busy, skip halt rlc\n");
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 
 	return 0;
 }
 
-static int gfx_v8_0_suspend(void *handle)
+static int gfx_v8_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	return gfx_v8_0_hw_fini(handle);
+	return gfx_v8_0_hw_fini(ip_block);
 }
 
-static int gfx_v8_0_resume(void *handle)
+static int gfx_v8_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	return gfx_v8_0_hw_init(handle);
+	return gfx_v8_0_hw_init(ip_block);
 }
 
-static bool gfx_v8_0_check_soft_reset(void *handle)
+static bool gfx_v8_0_check_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 grbm_soft_reset = 0, srbm_soft_reset = 0;
 	u32 tmp;
 
@@ -4988,9 +4949,9 @@ static bool gfx_v8_0_check_soft_reset(void *handle)
 	}
 }
 
-static int gfx_v8_0_pre_soft_reset(void *handle)
+static int gfx_v8_0_pre_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 grbm_soft_reset = 0;
 
 	if ((!adev->gfx.grbm_soft_reset) &&
@@ -5029,9 +4990,9 @@ static int gfx_v8_0_pre_soft_reset(void *handle)
 	return 0;
 }
 
-static int gfx_v8_0_soft_reset(void *handle)
+static int gfx_v8_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 grbm_soft_reset = 0, srbm_soft_reset = 0;
 	u32 tmp;
 
@@ -5091,9 +5052,9 @@ static int gfx_v8_0_soft_reset(void *handle)
 	return 0;
 }
 
-static int gfx_v8_0_post_soft_reset(void *handle)
+static int gfx_v8_0_post_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 grbm_soft_reset = 0;
 
 	if ((!adev->gfx.grbm_soft_reset) &&
@@ -5216,7 +5177,7 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd,
 		*(out++) = RREG32(mmSQ_IND_DATA);
 }
 
-static void gfx_v8_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
+static void gfx_v8_0_read_wave_data(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
 {
 	/* type 0 wave data */
 	dst[(*no_fields)++] = 0;
@@ -5241,7 +5202,7 @@ static void gfx_v8_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, u
 	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_MODE);
 }
 
-static void gfx_v8_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v8_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				     uint32_t wave, uint32_t start,
 				     uint32_t size, uint32_t *dst)
 {
@@ -5259,10 +5220,11 @@ static const struct amdgpu_gfx_funcs gfx_v8_0_gfx_funcs = {
 	.select_me_pipe_q = &gfx_v8_0_select_me_pipe_q
 };
 
-static int gfx_v8_0_early_init(void *handle)
+static int gfx_v8_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
+	adev->gfx.xcc_mask = 1;
 	adev->gfx.num_gfx_rings = GFX8_NUM_GFX_RINGS;
 	adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
 					  AMDGPU_MAX_COMPUTE_RINGS);
@@ -5275,9 +5237,9 @@ static int gfx_v8_0_early_init(void *handle)
 	return 0;
 }
 
-static int gfx_v8_0_late_init(void *handle)
+static int gfx_v8_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0);
@@ -5317,7 +5279,7 @@ static void gfx_v8_0_enable_gfx_static_mg_power_gating(struct amdgpu_device *ade
 	    (adev->asic_type == CHIP_POLARIS12) ||
 	    (adev->asic_type == CHIP_VEGAM))
 		/* Send msg to SMU via Powerplay */
-		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, enable);
+		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, enable, 0);
 
 	WREG32_FIELD(RLC_PG_CNTL, STATIC_PER_CU_PG_ENABLE, enable ? 1 : 0);
 }
@@ -5363,10 +5325,10 @@ static void cz_update_gfx_cg_power_gating(struct amdgpu_device *adev,
 	}
 }
 
-static int gfx_v8_0_set_powergating_state(void *handle,
+static int gfx_v8_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_PG_STATE_GATE);
 
 	if (amdgpu_sriov_vf(adev))
@@ -5376,7 +5338,7 @@ static int gfx_v8_0_set_powergating_state(void *handle,
 				AMD_PG_SUPPORT_RLC_SMU_HS |
 				AMD_PG_SUPPORT_CP |
 				AMD_PG_SUPPORT_GFX_DMG))
-		amdgpu_gfx_rlc_enter_safe_mode(adev);
+		amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 	switch (adev->asic_type) {
 	case CHIP_CARRIZO:
 	case CHIP_STONEY:
@@ -5430,13 +5392,13 @@ static int gfx_v8_0_set_powergating_state(void *handle,
 				AMD_PG_SUPPORT_RLC_SMU_HS |
 				AMD_PG_SUPPORT_CP |
 				AMD_PG_SUPPORT_GFX_DMG))
-		amdgpu_gfx_rlc_exit_safe_mode(adev);
+		amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 	return 0;
 }
 
-static void gfx_v8_0_get_clockgating_state(void *handle, u64 *flags)
+static void gfx_v8_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	if (amdgpu_sriov_vf(adev))
@@ -5481,7 +5443,7 @@ static void gfx_v8_0_send_serdes_cmd(struct amdgpu_device *adev,
 {
 	uint32_t data;
 
-	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 
 	WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff);
 	WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff);
@@ -5535,7 +5497,7 @@ static bool gfx_v8_0_is_rlc_enabled(struct amdgpu_device *adev)
 	return true;
 }
 
-static void gfx_v8_0_set_safe_mode(struct amdgpu_device *adev)
+static void gfx_v8_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	uint32_t data;
 	unsigned i;
@@ -5562,7 +5524,7 @@ static void gfx_v8_0_set_safe_mode(struct amdgpu_device *adev)
 	}
 }
 
-static void gfx_v8_0_unset_safe_mode(struct amdgpu_device *adev)
+static void gfx_v8_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	uint32_t data;
 	unsigned i;
@@ -5579,7 +5541,7 @@ static void gfx_v8_0_unset_safe_mode(struct amdgpu_device *adev)
 	}
 }
 
-static void gfx_v8_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid)
+static void gfx_v8_0_update_spm_vmid(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned vmid)
 {
 	u32 data;
 
@@ -5621,8 +5583,6 @@ static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev
 {
 	uint32_t temp, data;
 
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
-
 	/* It is disabled by HW by default */
 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) {
 		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGLS) {
@@ -5716,8 +5676,6 @@ static void gfx_v8_0_update_medium_grain_clock_gating(struct amdgpu_device *adev
 		/* 7- wait for RLC_SERDES_CU_MASTER & RLC_SERDES_NONCU_MASTER idle */
 		gfx_v8_0_wait_for_rlc_serdes(adev);
 	}
-
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
 }
 
 static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev,
@@ -5727,8 +5685,6 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev
 
 	temp = data = RREG32(mmRLC_CGCG_CGLS_CTRL);
 
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
-
 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) {
 		temp1 = data1 =	RREG32(mmRLC_CGTT_MGCG_OVERRIDE);
 		data1 &= ~RLC_CGTT_MGCG_OVERRIDE__CGCG_MASK;
@@ -5809,12 +5765,12 @@ static void gfx_v8_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev
 	}
 
 	gfx_v8_0_wait_for_rlc_serdes(adev);
-
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
 }
 static int gfx_v8_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 					    bool enable)
 {
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+
 	if (enable) {
 		/* CGCG/CGLS should be enabled after MGCG/MGLS/TS(CG/LS)
 		 * ===  MGCG + MGLS + TS(CG/LS) ===
@@ -5828,6 +5784,8 @@ static int gfx_v8_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 		gfx_v8_0_update_coarse_grain_clock_gating(adev, enable);
 		gfx_v8_0_update_medium_grain_clock_gating(adev, enable);
 	}
+
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 	return 0;
 }
 
@@ -5978,10 +5936,10 @@ static int gfx_v8_0_polaris_update_gfx_clock_gating(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int gfx_v8_0_set_clockgating_state(void *handle,
+static int gfx_v8_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
@@ -6153,6 +6111,7 @@ static void gfx_v8_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr,
 {
 	bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT;
 	bool int_sel = flags & AMDGPU_FENCE_FLAG_INT;
+	bool exec = flags & AMDGPU_FENCE_FLAG_EXEC;
 
 	/* Workaround for cache flush problems. First send a dummy EOP
 	 * event down the pipe with seq one below.
@@ -6176,7 +6135,8 @@ static void gfx_v8_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr,
 				 EOP_TC_ACTION_EN |
 				 EOP_TC_WB_ACTION_EN |
 				 EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) |
-				 EVENT_INDEX(5)));
+				 EVENT_INDEX(5) |
+				 (exec ? EOP_EXEC : 0)));
 	amdgpu_ring_write(ring, addr & 0xfffffffc);
 	amdgpu_ring_write(ring, (upper_32_bits(addr) & 0xffff) |
 			  DATA_SEL(write64bit ? 2 : 1) | INT_SEL(int_sel ? 2 : 0));
@@ -6327,33 +6287,22 @@ static void gfx_v8_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags)
 	amdgpu_ring_write(ring, 0);
 }
 
-static unsigned gfx_v8_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned gfx_v8_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring,
+						  uint64_t addr)
 {
 	unsigned ret;
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, 0); /* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	/* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, 0);
 	ret = ring->wptr & ring->buf_mask;
-	amdgpu_ring_write(ring, 0x55aa55aa); /* patch dummy value later */
+	/* patch dummy value later */
+	amdgpu_ring_write(ring, 0);
 	return ret;
 }
 
-static void gfx_v8_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigned offset)
-{
-	unsigned cur;
-
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
-
-	cur = (ring->wptr & ring->buf_mask) - 1;
-	if (likely(cur > offset))
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->ring_size >> 2) - offset + cur;
-}
-
 static void gfx_v8_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg,
 				    uint32_t reg_val_offs)
 {
@@ -6723,11 +6672,11 @@ static void gfx_v8_0_parse_sq_irq(struct amdgpu_device *adev, unsigned ih_data,
 			 */
 			if (from_wq) {
 				mutex_lock(&adev->grbm_idx_mutex);
-				gfx_v8_0_select_se_sh(adev, se_id, sh_id, cu_id);
+				gfx_v8_0_select_se_sh(adev, se_id, sh_id, cu_id, 0);
 
 				sq_edc_source = REG_GET_FIELD(RREG32(mmSQ_EDC_INFO), SQ_EDC_INFO, SOURCE);
 
-				gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+				gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 				mutex_unlock(&adev->grbm_idx_mutex);
 			}
 
@@ -6933,7 +6882,6 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = {
 	.emit_switch_buffer = gfx_v8_ring_emit_sb,
 	.emit_cntxcntl = gfx_v8_ring_emit_cntxcntl,
 	.init_cond_exec = gfx_v8_0_ring_emit_init_cond_exec,
-	.patch_cond_exec = gfx_v8_0_ring_emit_patch_cond_exec,
 	.emit_wreg = gfx_v8_0_ring_emit_wreg,
 	.soft_recovery = gfx_v8_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v8_0_emit_mem_sync,
@@ -6969,6 +6917,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = {
 	.insert_nop = amdgpu_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_wreg = gfx_v8_0_ring_emit_wreg,
+	.soft_recovery = gfx_v8_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v8_0_emit_mem_sync_compute,
 	.emit_wave_limit = gfx_v8_0_emit_wave_limit,
 };
@@ -6995,13 +6944,14 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_kiq = {
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_rreg = gfx_v8_0_ring_emit_rreg,
 	.emit_wreg = gfx_v8_0_ring_emit_wreg,
+	.emit_hdp_flush = gfx_v8_0_ring_emit_hdp_flush,
 };
 
 static void gfx_v8_0_set_ring_funcs(struct amdgpu_device *adev)
 {
 	int i;
 
-	adev->gfx.kiq.ring.funcs = &gfx_v8_0_ring_funcs_kiq;
+	adev->gfx.kiq[0].ring.funcs = &gfx_v8_0_ring_funcs_kiq;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 		adev->gfx.gfx_ring[i].funcs = &gfx_v8_0_ring_funcs_gfx;
@@ -7116,12 +7066,12 @@ static void gfx_v8_0_get_cu_info(struct amdgpu_device *adev)
 			mask = 1;
 			ao_bitmap = 0;
 			counter = 0;
-			gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff);
+			gfx_v8_0_select_se_sh(adev, i, j, 0xffffffff, 0);
 			if (i < 4 && j < 2)
 				gfx_v8_0_set_user_cu_inactive_bitmap(
 					adev, disable_masks[i * 2 + j]);
 			bitmap = gfx_v8_0_get_cu_active_bitmap(adev);
-			cu_info->bitmap[i][j] = bitmap;
+			cu_info->bitmap[0][i][j] = bitmap;
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) {
 				if (bitmap & mask) {
@@ -7137,7 +7087,7 @@ static void gfx_v8_0_get_cu_info(struct amdgpu_device *adev)
 			cu_info->ao_cu_bitmap[i][j] = ao_bitmap;
 		}
 	}
-	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	cu_info->number = active_cu_number;
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
index ae09fc1cfe6b..0148d7ff34d9 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c
@@ -50,6 +50,7 @@
 #include "amdgpu_ring_mux.h"
 #include "gfx_v9_4.h"
 #include "gfx_v9_0.h"
+#include "gfx_v9_0_cleaner_shader.h"
 #include "gfx_v9_4_2.h"
 
 #include "asic_reg/pwr/pwr_10_0_offset.h"
@@ -149,6 +150,162 @@ MODULE_FIRMWARE("amdgpu/aldebaran_sjt_mec2.bin");
 #define mmGOLDEN_TSC_COUNT_LOWER_Renoir                0x0026
 #define mmGOLDEN_TSC_COUNT_LOWER_Renoir_BASE_IDX       1
 
+static const struct amdgpu_hwip_reg_entry gc_reg_list_9[] = {
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_STALLED_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPF_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_GFX_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB0_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB0_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB0_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB1_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB1_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB1_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB2_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB2_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_RB2_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_CMD_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB1_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_IB2_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB1_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_BASE_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_IB2_BUFSZ),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCPF_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCPC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCPG_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGDS_PROTECTION_FAULT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGDS_VM_PROTECTION_FAULT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmIA_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmIA_UTCL1_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmPA_CL_CNTL_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRMI_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSQC_DCACHE_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSQC_ICACHE_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSQ_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmTCP_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmWD_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmVM_L2_PROTECTION_FAULT_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmVM_L2_PROTECTION_FAULT_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_DEBUG),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC1_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC2_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CPC_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_COMMAND),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_MESSAGE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_ARGUMENT_1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_ARGUMENT_2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmSMU_RLC_RESPONSE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SAFE_MODE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_SMU_SAFE_MODE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_INT_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, mmRLC_GPM_GENERAL_6),
+	/* SE status registers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE0),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE1),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE2),
+	SOC15_REG_ENTRY_STR(GC, 0, mmGRBM_STATUS_SE3),
+	/* packet headers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_CE_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_PFP_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_ME_HEADER_DUMP)
+};
+
+static const struct amdgpu_hwip_reg_entry gc_cp_reg_list_9[] = {
+	/* compute queue registers */
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_VMID),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_ACTIVE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PERSISTENT_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PIPE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_QUEUE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_QUANTUM),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_IB_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_DEQUEUE_REQUEST),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_EVENTS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_BASE_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CNTL_STACK_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CNTL_STACK_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_WG_STATE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_CTX_SAVE_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_GDS_RESOURCE_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_EOP_WPTR_MEM),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_PQ_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_HQD_GFX_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, mmCP_MEC_ME1_HEADER_DUMP)
+};
+
 enum ta_ras_gfx_subblock {
 	/*CPC*/
 	TA_RAS_BLOCK__GFX_CPC_INDEX_START = 0,
@@ -755,17 +912,27 @@ static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev);
 static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev,
 				struct amdgpu_cu_info *cu_info);
 static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev);
-static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume);
+static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume, bool usegds);
 static u64 gfx_v9_0_ring_get_rptr_compute(struct amdgpu_ring *ring);
 static void gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev,
 					  void *ras_error_status);
 static int gfx_v9_0_ras_error_inject(struct amdgpu_device *adev,
-				     void *inject_if);
+				     void *inject_if, uint32_t instance_mask);
 static void gfx_v9_0_reset_ras_error_count(struct amdgpu_device *adev);
+static void gfx_v9_0_update_spm_vmid_internal(struct amdgpu_device *adev,
+					      unsigned int vmid);
+static void gfx_v9_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id);
+static void gfx_v9_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id);
 
 static void gfx_v9_0_kiq_set_resources(struct amdgpu_ring *kiq_ring,
 				uint64_t queue_mask)
 {
+	struct amdgpu_device *adev = kiq_ring->adev;
+	u64 shader_mc_addr;
+
+	/* Cleaner shader MC address */
+	shader_mc_addr = adev->gfx.cleaner_shader_gpu_addr >> 8;
+
 	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_RESOURCES, 6));
 	amdgpu_ring_write(kiq_ring,
 		PACKET3_SET_RESOURCES_VMID_MASK(0) |
@@ -775,8 +942,8 @@ static void gfx_v9_0_kiq_set_resources(struct amdgpu_ring *kiq_ring,
 			lower_32_bits(queue_mask));	/* queue mask lo */
 	amdgpu_ring_write(kiq_ring,
 			upper_32_bits(queue_mask));	/* queue mask hi */
-	amdgpu_ring_write(kiq_ring, 0);	/* gws mask lo */
-	amdgpu_ring_write(kiq_ring, 0);	/* gws mask hi */
+	amdgpu_ring_write(kiq_ring, lower_32_bits(shader_mc_addr)); /* cleaner shader addr lo */
+	amdgpu_ring_write(kiq_ring, upper_32_bits(shader_mc_addr)); /* cleaner shader addr hi */
 	amdgpu_ring_write(kiq_ring, 0);	/* oac mask */
 	amdgpu_ring_write(kiq_ring, 0);	/* gds heap base:0, gds heap size:0 */
 }
@@ -873,12 +1040,47 @@ static void gfx_v9_0_kiq_invalidate_tlbs(struct amdgpu_ring *kiq_ring,
 			PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(flush_type));
 }
 
+
+static void gfx_v9_0_kiq_reset_hw_queue(struct amdgpu_ring *kiq_ring, uint32_t queue_type,
+					uint32_t me_id, uint32_t pipe_id, uint32_t queue_id,
+					uint32_t xcc_id, uint32_t vmid)
+{
+	struct amdgpu_device *adev = kiq_ring->adev;
+	unsigned i;
+
+	/* enter save mode */
+	amdgpu_gfx_rlc_enter_safe_mode(adev, xcc_id);
+	mutex_lock(&adev->srbm_mutex);
+	soc15_grbm_select(adev, me_id, pipe_id, queue_id, 0, 0);
+
+	if (queue_type == AMDGPU_RING_TYPE_COMPUTE) {
+		WREG32_SOC15(GC, 0, mmCP_HQD_DEQUEUE_REQUEST, 0x2);
+		WREG32_SOC15(GC, 0, mmSPI_COMPUTE_QUEUE_RESET, 0x1);
+		/* wait till dequeue take effects */
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (!(RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+		if (i >= adev->usec_timeout)
+			dev_err(adev->dev, "fail to wait on hqd deactive\n");
+	} else {
+		dev_err(adev->dev, "reset queue_type(%d) not supported\n", queue_type);
+	}
+
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	/* exit safe mode */
+	amdgpu_gfx_rlc_exit_safe_mode(adev, xcc_id);
+}
+
 static const struct kiq_pm4_funcs gfx_v9_0_kiq_pm4_funcs = {
 	.kiq_set_resources = gfx_v9_0_kiq_set_resources,
 	.kiq_map_queues = gfx_v9_0_kiq_map_queues,
 	.kiq_unmap_queues = gfx_v9_0_kiq_unmap_queues,
 	.kiq_query_status = gfx_v9_0_kiq_query_status,
 	.kiq_invalidate_tlbs = gfx_v9_0_kiq_invalidate_tlbs,
+	.kiq_reset_hw_queue = gfx_v9_0_kiq_reset_hw_queue,
 	.set_resources_size = 8,
 	.map_queues_size = 7,
 	.unmap_queues_size = 6,
@@ -888,12 +1090,12 @@ static const struct kiq_pm4_funcs gfx_v9_0_kiq_pm4_funcs = {
 
 static void gfx_v9_0_set_kiq_pm4_funcs(struct amdgpu_device *adev)
 {
-	adev->gfx.kiq.pmf = &gfx_v9_0_kiq_pm4_funcs;
+	adev->gfx.kiq[0].pmf = &gfx_v9_0_kiq_pm4_funcs;
 }
 
 static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 		soc15_program_register_sequence(adev,
 						golden_settings_gc_9_0,
@@ -949,8 +1151,8 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev)
 		break;
 	}
 
-	if ((adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 1)) &&
-	    (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 2)))
+	if ((amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 1)) &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 2)))
 		soc15_program_register_sequence(adev, golden_settings_gc_9_x_common,
 						(const u32)ARRAY_SIZE(golden_settings_gc_9_x_common));
 }
@@ -1037,8 +1239,8 @@ static int gfx_v9_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 	gpu_addr = adev->wb.gpu_addr + (index * 4);
 	adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
 	memset(&ib, 0, sizeof(ib));
-	r = amdgpu_ib_get(adev, NULL, 16,
-					AMDGPU_IB_POOL_DIRECT, &ib);
+
+	r = amdgpu_ib_get(adev, NULL, 20, AMDGPU_IB_POOL_DIRECT, &ib);
 	if (r)
 		goto err1;
 
@@ -1068,7 +1270,7 @@ static int gfx_v9_0_ring_test_ib(struct amdgpu_ring *ring, long timeout)
 		r = -EINVAL;
 
 err2:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 err1:
 	amdgpu_device_wb_free(adev, index);
@@ -1093,14 +1295,15 @@ static void gfx_v9_0_check_fw_write_wait(struct amdgpu_device *adev)
 	adev->gfx.me_fw_write_wait = false;
 	adev->gfx.mec_fw_write_wait = false;
 
-	if ((adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 1)) &&
+	if ((amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 1)) &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 2)) &&
 	    ((adev->gfx.mec_fw_version < 0x000001a5) ||
-	    (adev->gfx.mec_feature_version < 46) ||
-	    (adev->gfx.pfp_fw_version < 0x000000b7) ||
-	    (adev->gfx.pfp_feature_version < 46)))
+	     (adev->gfx.mec_feature_version < 46) ||
+	     (adev->gfx.pfp_fw_version < 0x000000b7) ||
+	     (adev->gfx.pfp_feature_version < 46)))
 		DRM_WARN_ONCE("CP firmware version too old, please update!");
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 		if ((adev->gfx.me_fw_version >= 0x0000009c) &&
 		    (adev->gfx.me_feature_version >= 42) &&
@@ -1170,6 +1373,10 @@ static const struct amdgpu_gfxoff_quirk amdgpu_gfxoff_quirk_list[] = {
 	{ 0x1002, 0x15dd, 0x1002, 0x15dd, 0xc6 },
 	/* Apple MacBook Pro (15-inch, 2019) Radeon Pro Vega 20 4 GB */
 	{ 0x1002, 0x69af, 0x106b, 0x019a, 0xc0 },
+	/* https://bbs.openkylin.top/t/topic/171497 */
+	{ 0x1002, 0x15d8, 0x19e5, 0x3e14, 0xc2 },
+	/* HP 705G4 DM with R5 2400G */
+	{ 0x1002, 0x15dd, 0x103c, 0x8464, 0xd6 },
 	{ 0, 0, 0, 0, 0 },
 };
 
@@ -1200,7 +1407,7 @@ static bool is_raven_kicker(struct amdgpu_device *adev)
 
 static bool check_if_enlarge_doorbell_range(struct amdgpu_device *adev)
 {
-	if ((adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 3, 0)) &&
+	if ((amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 3, 0)) &&
 	    (adev->gfx.me_fw_version >= 0x000000a5) &&
 	    (adev->gfx.me_feature_version >= 52))
 		return true;
@@ -1213,7 +1420,7 @@ static void gfx_v9_0_check_if_need_gfxoff(struct amdgpu_device *adev)
 	if (gfx_v9_0_should_disable_gfxoff(adev->pdev))
 		adev->pm.pp_feature &= ~PP_GFXOFF_MASK;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 2, 1):
 	case IP_VERSION(9, 4, 0):
@@ -1247,23 +1454,25 @@ static void gfx_v9_0_check_if_need_gfxoff(struct amdgpu_device *adev)
 static int gfx_v9_0_init_cp_gfx_microcode(struct amdgpu_device *adev,
 					  char *chip_name)
 {
-	char fw_name[30];
 	int err;
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.pfp_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_pfp.bin", chip_name);
 	if (err)
 		goto out;
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_PFP);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.me_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_me.bin", chip_name);
 	if (err)
 		goto out;
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_ME);
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gfx.ce_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_ce.bin", chip_name);
 	if (err)
 		goto out;
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_CE);
@@ -1280,7 +1489,6 @@ out:
 static int gfx_v9_0_init_rlc_microcode(struct amdgpu_device *adev,
 				       char *chip_name)
 {
-	char fw_name[30];
 	int err;
 	const struct rlc_firmware_header_v2_0 *rlc_hdr;
 	uint16_t version_major;
@@ -1298,20 +1506,25 @@ static int gfx_v9_0_init_rlc_microcode(struct amdgpu_device *adev,
 	if (!strcmp(chip_name, "picasso") &&
 		(((adev->pdev->revision >= 0xC8) && (adev->pdev->revision <= 0xCF)) ||
 		((adev->pdev->revision >= 0xD8) && (adev->pdev->revision <= 0xDF))))
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc_am4.bin", chip_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_rlc_am4.bin", chip_name);
 	else if (!strcmp(chip_name, "raven") && (amdgpu_pm_load_smu_firmware(adev, &smu_version) == 0) &&
 		(smu_version >= 0x41e2b))
 		/**
 		*SMC is loaded by SBIOS on APU and it's able to get the SMU version directly.
 		*/
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_kicker_rlc.bin", chip_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_kicker_rlc.bin", chip_name);
 	else
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_rlc.bin", chip_name);
 	if (err)
 		goto out;
-	rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
 
+	rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
 	version_major = le16_to_cpu(rlc_hdr->header.header_version_major);
 	version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor);
 	err = amdgpu_gfx_rlc_init_microcode(adev, version_major, version_minor);
@@ -1324,9 +1537,9 @@ out:
 
 static bool gfx_v9_0_load_mec2_fw_bin_support(struct amdgpu_device *adev)
 {
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) ||
-	    adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1) ||
-	    adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 3, 0))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 3, 0))
 		return false;
 
 	return true;
@@ -1335,28 +1548,31 @@ static bool gfx_v9_0_load_mec2_fw_bin_support(struct amdgpu_device *adev)
 static int gfx_v9_0_init_cp_compute_microcode(struct amdgpu_device *adev,
 					      char *chip_name)
 {
-	char fw_name[30];
 	int err;
 
 	if (amdgpu_sriov_vf(adev) && (adev->asic_type == CHIP_ALDEBARAN))
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sjt_mec.bin", chip_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_sjt_mec.bin", chip_name);
 	else
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name);
-
-	err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, fw_name);
+		err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_mec.bin", chip_name);
 	if (err)
 		goto out;
+
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1);
 	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1_JT);
 
 	if (gfx_v9_0_load_mec2_fw_bin_support(adev)) {
 		if (amdgpu_sriov_vf(adev) && (adev->asic_type == CHIP_ALDEBARAN))
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sjt_mec2.bin", chip_name);
+			err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_sjt_mec2.bin", chip_name);
 		else
-			snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name);
-
-		/* ignore failures to load */
-		err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw, fw_name);
+			err = amdgpu_ucode_request(adev, &adev->gfx.mec2_fw,
+						   AMDGPU_UCODE_REQUIRED,
+						   "amdgpu/%s_mec2.bin", chip_name);
 		if (!err) {
 			amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2);
 			amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC2_JT);
@@ -1432,45 +1648,18 @@ static u32 gfx_v9_0_get_csb_size(struct amdgpu_device *adev)
 	return count;
 }
 
-static void gfx_v9_0_get_csb_buffer(struct amdgpu_device *adev,
-				    volatile u32 *buffer)
+static void gfx_v9_0_get_csb_buffer(struct amdgpu_device *adev, u32 *buffer)
 {
-	u32 count = 0, i;
-	const struct cs_section_def *sect = NULL;
-	const struct cs_extent_def *ext = NULL;
+	u32 count = 0;
 
 	if (adev->gfx.rlc.cs_data == NULL)
 		return;
 	if (buffer == NULL)
 		return;
 
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1));
-	buffer[count++] = cpu_to_le32(0x80000000);
-	buffer[count++] = cpu_to_le32(0x80000000);
-
-	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
-		for (ext = sect->section; ext->extent != NULL; ++ext) {
-			if (sect->id == SECT_CONTEXT) {
-				buffer[count++] =
-					cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
-				buffer[count++] = cpu_to_le32(ext->reg_index -
-						PACKET3_SET_CONTEXT_REG_START);
-				for (i = 0; i < ext->reg_count; i++)
-					buffer[count++] = cpu_to_le32(ext->extent[i]);
-			} else {
-				return;
-			}
-		}
-	}
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
-	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE);
-
-	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0));
-	buffer[count++] = cpu_to_le32(0);
+	count = amdgpu_gfx_csb_preamble_start(buffer);
+	count = amdgpu_gfx_csb_data_parser(adev, buffer, count);
+	amdgpu_gfx_csb_preamble_end(buffer, count);
 }
 
 static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev)
@@ -1483,7 +1672,7 @@ static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev)
 
 	if (adev->flags & AMD_IS_APU)
 		always_on_cu_num = 4;
-	else if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 2, 1))
+	else if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 2, 1))
 		always_on_cu_num = 8;
 	else
 		always_on_cu_num = 12;
@@ -1494,10 +1683,10 @@ static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev)
 			mask = 1;
 			cu_bitmap = 0;
 			counter = 0;
-			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff);
+			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff, 0);
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) {
-				if (cu_info->bitmap[i][j] & mask) {
+				if (cu_info->bitmap[0][i][j] & mask) {
 					if (counter == pg_always_on_cu_num)
 						WREG32_SOC15(GC, 0, mmRLC_PG_ALWAYS_ON_CU_MASK, cu_bitmap);
 					if (counter < always_on_cu_num)
@@ -1513,7 +1702,7 @@ static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev)
 			cu_info->ao_cu_bitmap[i][j] = cu_bitmap;
 		}
 	}
-	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 }
 
@@ -1535,7 +1724,7 @@ static void gfx_v9_0_init_lbpw(struct amdgpu_device *adev)
 
 	mutex_lock(&adev->grbm_idx_mutex);
 	/* set mmRLC_LB_INIT_CU_MASK thru broadcast mode to enable all SE/SH*/
-	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	WREG32_SOC15(GC, 0, mmRLC_LB_INIT_CU_MASK, 0xffffffff);
 
 	/* set mmRLC_LB_PARAMS = 0x003F_1006 */
@@ -1584,7 +1773,7 @@ static void gfx_v9_4_init_lbpw(struct amdgpu_device *adev)
 
 	mutex_lock(&adev->grbm_idx_mutex);
 	/* set mmRLC_LB_INIT_CU_MASK thru broadcast mode to enable all SE/SH*/
-	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	WREG32_SOC15(GC, 0, mmRLC_LB_INIT_CU_MASK, 0xffffffff);
 
 	/* set mmRLC_LB_PARAMS = 0x003F_1006 */
@@ -1632,7 +1821,7 @@ static void gfx_v9_0_init_rlcg_reg_access_ctrl(struct amdgpu_device *adev)
 {
 	struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl;
 
-	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl;
+	reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl[0];
 	reg_access_ctrl->scratch_reg0 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG0);
 	reg_access_ctrl->scratch_reg1 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG1);
 	reg_access_ctrl->scratch_reg2 = SOC15_REG_OFFSET(GC, 0, mmSCRATCH_REG2);
@@ -1667,22 +1856,6 @@ static int gfx_v9_0_rlc_init(struct amdgpu_device *adev)
 			return r;
 	}
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
-	case IP_VERSION(9, 2, 2):
-	case IP_VERSION(9, 1, 0):
-		gfx_v9_0_init_lbpw(adev);
-		break;
-	case IP_VERSION(9, 4, 0):
-		gfx_v9_4_init_lbpw(adev);
-		break;
-	default:
-		break;
-	}
-
-	/* init spm vmid with 0xf */
-	if (adev->gfx.rlc.funcs->update_spm_vmid)
-		adev->gfx.rlc.funcs->update_spm_vmid(adev, 0xf);
-
 	return 0;
 }
 
@@ -1703,7 +1876,7 @@ static int gfx_v9_0_mec_init(struct amdgpu_device *adev)
 
 	const struct gfx_firmware_header_v1_0 *mec_hdr;
 
-	bitmap_zero(adev->gfx.mec.queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
+	bitmap_zero(adev->gfx.mec_bitmap[0].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
 
 	/* take ownership of the relevant compute queues */
 	amdgpu_gfx_compute_queue_acquire(adev);
@@ -1778,7 +1951,7 @@ static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd,
 		*(out++) = RREG32_SOC15(GC, 0, mmSQ_IND_DATA);
 }
 
-static void gfx_v9_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
+static void gfx_v9_0_read_wave_data(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields)
 {
 	/* type 1 wave data */
 	dst[(*no_fields)++] = 1;
@@ -1799,7 +1972,7 @@ static void gfx_v9_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, u
 	dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_MODE);
 }
 
-static void gfx_v9_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v9_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				     uint32_t wave, uint32_t start,
 				     uint32_t size, uint32_t *dst)
 {
@@ -1808,7 +1981,7 @@ static void gfx_v9_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd,
 		start + SQIND_WAVE_SGPRS_OFFSET, size, dst);
 }
 
-static void gfx_v9_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd,
+static void gfx_v9_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
 				     uint32_t wave, uint32_t thread,
 				     uint32_t start, uint32_t size,
 				     uint32_t *dst)
@@ -1819,9 +1992,9 @@ static void gfx_v9_0_read_wave_vgprs(struct amdgpu_device *adev, uint32_t simd,
 }
 
 static void gfx_v9_0_select_me_pipe_q(struct amdgpu_device *adev,
-				  u32 me, u32 pipe, u32 q, u32 vm)
+				  u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
 {
-	soc15_grbm_select(adev, me, pipe, q, vm);
+	soc15_grbm_select(adev, me, pipe, q, vm, 0);
 }
 
 static const struct amdgpu_gfx_funcs gfx_v9_0_gfx_funcs = {
@@ -1850,7 +2023,7 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev)
 	u32 gb_addr_config;
 	int err;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 		adev->gfx.config.max_hw_contexts = 8;
 		adev->gfx.config.sc_prim_fifo_size_frontend = 0x20;
@@ -1995,6 +2168,7 @@ static int gfx_v9_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 	ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
 	ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
 				+ (ring_id * GFX9_MEC_HPD_SIZE);
+	ring->vm_hub = AMDGPU_GFXHUB(0);
 	sprintf(ring->name, "comp_%d.%d.%d", ring->me, ring->pipe, ring->queue);
 
 	irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
@@ -2007,15 +2181,43 @@ static int gfx_v9_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 				hw_prio, NULL);
 }
 
-static int gfx_v9_0_sw_init(void *handle)
+static void gfx_v9_0_alloc_ip_dump(struct amdgpu_device *adev)
+{
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9);
+	uint32_t *ptr;
+	uint32_t inst;
+
+	ptr = kcalloc(reg_count, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for GFX IP Dump\n");
+		adev->gfx.ip_dump_core = NULL;
+	} else {
+		adev->gfx.ip_dump_core = ptr;
+	}
+
+	/* Allocate memory for compute queue registers for all the instances */
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_9);
+	inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec *
+		adev->gfx.mec.num_queue_per_pipe;
+
+	ptr = kcalloc(reg_count * inst, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for Compute Queues IP Dump\n");
+		adev->gfx.ip_dump_compute_queues = NULL;
+	} else {
+		adev->gfx.ip_dump_compute_queues = ptr;
+	}
+}
+
+static int gfx_v9_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int i, j, k, r, ring_id;
+	int xcc_id = 0;
 	struct amdgpu_ring *ring;
-	struct amdgpu_kiq *kiq;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	unsigned int hw_prio;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 2, 1):
 	case IP_VERSION(9, 4, 0):
@@ -2031,6 +2233,43 @@ static int gfx_v9_0_sw_init(void *handle)
 		break;
 	}
 
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(9, 0, 1):
+	case IP_VERSION(9, 2, 1):
+	case IP_VERSION(9, 4, 0):
+	case IP_VERSION(9, 2, 2):
+	case IP_VERSION(9, 1, 0):
+	case IP_VERSION(9, 3, 0):
+		adev->gfx.cleaner_shader_ptr = gfx_9_4_2_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_9_4_2_cleaner_shader_hex);
+		if (adev->gfx.me_fw_version  >= 167 &&
+		    adev->gfx.pfp_fw_version >= 196 &&
+		    adev->gfx.mec_fw_version >= 474) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	case IP_VERSION(9, 4, 2):
+		adev->gfx.cleaner_shader_ptr = gfx_9_4_2_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_9_4_2_cleaner_shader_hex);
+		if (adev->gfx.mec_fw_version >= 88) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	default:
+		adev->gfx.enable_cleaner_shader = false;
+		break;
+	}
+
 	adev->gfx.mec.num_pipe_per_mec = 4;
 	adev->gfx.mec.num_queue_per_pipe = 8;
 
@@ -2039,6 +2278,13 @@ static int gfx_v9_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	/* Bad opcode Event */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP,
+			      GFX_9_0__SRCID__CP_BAD_OPCODE_ERROR,
+			      &adev->gfx.bad_op_irq);
+	if (r)
+		return r;
+
 	/* Privileged reg */
 	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_9_0__SRCID__CP_PRIV_REG_FAULT,
 			      &adev->gfx.priv_reg_irq);
@@ -2093,7 +2339,8 @@ static int gfx_v9_0_sw_init(void *handle)
 		ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
 
 		/* disable scheduler on the real ring */
-		ring->no_scheduler = true;
+		ring->no_scheduler = adev->gfx.mcbp;
+		ring->vm_hub = AMDGPU_GFXHUB(0);
 		r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq,
 				     AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP,
 				     AMDGPU_RING_PRIO_DEFAULT, NULL);
@@ -2102,7 +2349,7 @@ static int gfx_v9_0_sw_init(void *handle)
 	}
 
 	/* set up the software rings */
-	if (adev->gfx.num_gfx_rings) {
+	if (adev->gfx.mcbp && adev->gfx.num_gfx_rings) {
 		for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++) {
 			ring = &adev->gfx.sw_gfx_ring[i];
 			ring->ring_obj = NULL;
@@ -2111,6 +2358,7 @@ static int gfx_v9_0_sw_init(void *handle)
 			ring->doorbell_index = adev->doorbell_index.gfx_ring0 << 1;
 			ring->is_sw_ring = true;
 			hw_prio = amdgpu_sw_ring_priority(i);
+			ring->vm_hub = AMDGPU_GFXHUB(0);
 			r = amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq,
 					     AMDGPU_CP_IRQ_GFX_ME0_PIPE0_EOP, hw_prio,
 					     NULL);
@@ -2141,7 +2389,8 @@ static int gfx_v9_0_sw_init(void *handle)
 	for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
 		for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
 			for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
-				if (!amdgpu_gfx_is_mec_queue_enabled(adev, i, k, j))
+				if (!amdgpu_gfx_is_mec_queue_enabled(adev, 0, i,
+								     k, j))
 					continue;
 
 				r = gfx_v9_0_compute_ring_init(adev,
@@ -2155,19 +2404,26 @@ static int gfx_v9_0_sw_init(void *handle)
 		}
 	}
 
-	r = amdgpu_gfx_kiq_init(adev, GFX9_MEC_HPD_SIZE);
+	/* TODO: Add queue reset mask when FW fully supports it */
+	adev->gfx.gfx_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.gfx_ring[0]);
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+	if (!amdgpu_sriov_vf(adev) && !adev->debug_disable_gpu_ring_reset)
+		adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+
+	r = amdgpu_gfx_kiq_init(adev, GFX9_MEC_HPD_SIZE, 0);
 	if (r) {
 		DRM_ERROR("Failed to init KIQ BOs!\n");
 		return r;
 	}
 
-	kiq = &adev->gfx.kiq;
-	r = amdgpu_gfx_kiq_init_ring(adev, &kiq->ring, &kiq->irq);
+	r = amdgpu_gfx_kiq_init_ring(adev, xcc_id);
 	if (r)
 		return r;
 
 	/* create MQD for all compute queues as wel as KIQ for SRIOV case */
-	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v9_mqd_allocation));
+	r = amdgpu_gfx_mqd_sw_init(adev, sizeof(struct v9_mqd_allocation), 0);
 	if (r)
 		return r;
 
@@ -2182,16 +2438,22 @@ static int gfx_v9_0_sw_init(void *handle)
 		return -EINVAL;
 	}
 
+	gfx_v9_0_alloc_ip_dump(adev);
+
+	r = amdgpu_gfx_sysfs_init(adev);
+	if (r)
+		return r;
+
 	return 0;
 }
 
 
-static int gfx_v9_0_sw_fini(void *handle)
+static int gfx_v9_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
 	int i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	if (adev->gfx.num_gfx_rings) {
+	if (adev->gfx.mcbp && adev->gfx.num_gfx_rings) {
 		for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++)
 			amdgpu_ring_fini(&adev->gfx.sw_gfx_ring[i]);
 		amdgpu_ring_mux_fini(&adev->gfx.muxer);
@@ -2202,9 +2464,11 @@ static int gfx_v9_0_sw_fini(void *handle)
 	for (i = 0; i < adev->gfx.num_compute_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
 
-	amdgpu_gfx_mqd_sw_fini(adev);
-	amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring);
-	amdgpu_gfx_kiq_fini(adev);
+	amdgpu_gfx_mqd_sw_fini(adev, 0);
+	amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq[0].ring);
+	amdgpu_gfx_kiq_fini(adev, 0);
+
+	amdgpu_gfx_cleaner_shader_sw_fini(adev);
 
 	gfx_v9_0_mec_fini(adev);
 	amdgpu_bo_free_kernel(&adev->gfx.rlc.clear_state_obj,
@@ -2217,6 +2481,11 @@ static int gfx_v9_0_sw_fini(void *handle)
 	}
 	gfx_v9_0_free_microcode(adev);
 
+	amdgpu_gfx_sysfs_fini(adev);
+
+	kfree(adev->gfx.ip_dump_core);
+	kfree(adev->gfx.ip_dump_compute_queues);
+
 	return 0;
 }
 
@@ -2227,7 +2496,7 @@ static void gfx_v9_0_tiling_mode_table_init(struct amdgpu_device *adev)
 }
 
 void gfx_v9_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num,
-			   u32 instance)
+			   u32 instance, int xcc_id)
 {
 	u32 data;
 
@@ -2276,19 +2545,42 @@ static void gfx_v9_0_setup_rb(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff);
+			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff, 0);
 			data = gfx_v9_0_get_rb_active_bitmap(adev);
 			active_rbs |= data << ((i * adev->gfx.config.max_sh_per_se + j) *
 					       rb_bitmap_width_per_sh);
 		}
 	}
-	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	adev->gfx.config.backend_enable_mask = active_rbs;
 	adev->gfx.config.num_rbs = hweight32(active_rbs);
 }
 
+static void gfx_v9_0_debug_trap_config_init(struct amdgpu_device *adev,
+				uint32_t first_vmid,
+				uint32_t last_vmid)
+{
+	uint32_t data;
+	uint32_t trap_config_vmid_mask = 0;
+	int i;
+
+	/* Calculate trap config vmid mask */
+	for (i = first_vmid; i < last_vmid; i++)
+		trap_config_vmid_mask |= (1 << i);
+
+	data = REG_SET_FIELD(0, SPI_GDBG_TRAP_CONFIG,
+			VMID_SEL, trap_config_vmid_mask);
+	data = REG_SET_FIELD(data, SPI_GDBG_TRAP_CONFIG,
+			TRAP_EN, 1);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_CONFIG), data);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_MASK), 0);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA0), 0);
+	WREG32(SOC15_REG_OFFSET(GC, 0, mmSPI_GDBG_TRAP_DATA1), 0);
+}
+
 #define DEFAULT_SH_MEM_BASES	(0x6000)
 static void gfx_v9_0_init_compute_vmid(struct amdgpu_device *adev)
 {
@@ -2310,12 +2602,12 @@ static void gfx_v9_0_init_compute_vmid(struct amdgpu_device *adev)
 
 	mutex_lock(&adev->srbm_mutex);
 	for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) {
-		soc15_grbm_select(adev, 0, 0, 0, i);
+		soc15_grbm_select(adev, 0, 0, 0, i, 0);
 		/* CP and shaders */
 		WREG32_SOC15_RLC(GC, 0, mmSH_MEM_CONFIG, sh_mem_config);
 		WREG32_SOC15_RLC(GC, 0, mmSH_MEM_BASES, sh_mem_bases);
 	}
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
 
 	/* Initialize all compute VMIDs to have no GDS, GWS, or OA
@@ -2350,13 +2642,16 @@ static void gfx_v9_0_init_sq_config(struct amdgpu_device *adev)
 {
 	uint32_t tmp;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 4, 1):
 		tmp = RREG32_SOC15(GC, 0, mmSQ_CONFIG);
-		tmp = REG_SET_FIELD(tmp, SQ_CONFIG,
-					DISABLE_BARRIER_WAITCNT, 1);
+		tmp = REG_SET_FIELD(tmp, SQ_CONFIG, DISABLE_BARRIER_WAITCNT,
+				!READ_ONCE(adev->barrier_has_auto_waitcnt));
 		WREG32_SOC15(GC, 0, mmSQ_CONFIG, tmp);
 		break;
+	case IP_VERSION(9, 4, 2):
+		gfx_v9_4_2_init_sq(adev);
+		break;
 	default:
 		break;
 	}
@@ -2367,7 +2662,10 @@ static void gfx_v9_0_constants_init(struct amdgpu_device *adev)
 	u32 tmp;
 	int i;
 
-	WREG32_FIELD15_RLC(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff);
+	if (!amdgpu_sriov_vf(adev) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 2)) {
+		WREG32_FIELD15_RLC(GC, 0, GRBM_CNTL, READ_TIMEOUT, 0xff);
+	}
 
 	gfx_v9_0_tiling_mode_table_init(adev);
 
@@ -2379,8 +2677,8 @@ static void gfx_v9_0_constants_init(struct amdgpu_device *adev)
 	/* XXX SH_MEM regs */
 	/* where to put LDS, scratch, GPUVM in FSA64 space */
 	mutex_lock(&adev->srbm_mutex);
-	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB_0].num_ids; i++) {
-		soc15_grbm_select(adev, 0, 0, 0, i);
+	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
+		soc15_grbm_select(adev, 0, 0, 0, i, 0);
 		/* CP and shaders */
 		if (i == 0) {
 			tmp = REG_SET_FIELD(0, SH_MEM_CONFIG, ALIGNMENT_MODE,
@@ -2402,7 +2700,7 @@ static void gfx_v9_0_constants_init(struct amdgpu_device *adev)
 			WREG32_SOC15_RLC(GC, 0, mmSH_MEM_BASES, tmp);
 		}
 	}
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 
 	mutex_unlock(&adev->srbm_mutex);
 
@@ -2419,7 +2717,7 @@ static void gfx_v9_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 	mutex_lock(&adev->grbm_idx_mutex);
 	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
 		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
-			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff);
+			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff, 0);
 			for (k = 0; k < adev->usec_timeout; k++) {
 				if (RREG32_SOC15(GC, 0, mmRLC_SERDES_CU_MASTER_BUSY) == 0)
 					break;
@@ -2427,7 +2725,7 @@ static void gfx_v9_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 			}
 			if (k == adev->usec_timeout) {
 				amdgpu_gfx_select_se_sh(adev, 0xffffffff,
-						      0xffffffff, 0xffffffff);
+						      0xffffffff, 0xffffffff, 0);
 				mutex_unlock(&adev->grbm_idx_mutex);
 				DRM_INFO("Timeout wait for RLC serdes %u,%u\n",
 					 i, j);
@@ -2435,7 +2733,7 @@ static void gfx_v9_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
 			}
 		}
 	}
-	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	mask = RLC_SERDES_NONCU_MASTER_BUSY__SE_MASTER_BUSY_MASK |
@@ -2461,7 +2759,7 @@ static void gfx_v9_0_enable_gui_idle_interrupt(struct amdgpu_device *adev,
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE, enable ? 1 : 0);
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE, enable ? 1 : 0);
 	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE, enable ? 1 : 0);
-	if(adev->gfx.num_gfx_rings)
+	if (adev->gfx.num_gfx_rings)
 		tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, GFX_IDLE_INT_ENABLE, enable ? 1 : 0);
 
 	WREG32_SOC15(GC, 0, mmCP_INT_CNTL_RING0, tmp);
@@ -2687,7 +2985,7 @@ static void gfx_v9_0_init_gfx_power_gating(struct amdgpu_device *adev)
 		/* program GRBM_REG_SAVE_GFX_IDLE_THRESHOLD to 0x55f0 */
 		data |= (0x55f0 << RLC_AUTO_PG_CTRL__GRBM_REG_SAVE_GFX_IDLE_THRESHOLD__SHIFT);
 		WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_AUTO_PG_CTRL), data);
-		if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 3, 0))
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 3, 0))
 			pwr_10_0_gfxip_control_over_cgpg(adev, true);
 	}
 }
@@ -2799,7 +3097,8 @@ static void gfx_v9_0_init_pg(struct amdgpu_device *adev)
 	 * And it's needed by gfxoff feature.
 	 */
 	if (adev->gfx.rlc.is_rlc_v2_1) {
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 2, 1) ||
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			    IP_VERSION(9, 2, 1) ||
 		    (adev->apu_flags & AMD_APU_IS_RAVEN2))
 			gfx_v9_1_init_rlc_save_restore_list(adev);
 		gfx_v9_0_enable_save_restore_machine(adev);
@@ -2912,15 +3211,17 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev)
 			return r;
 	}
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 2, 2):
 	case IP_VERSION(9, 1, 0):
+		gfx_v9_0_init_lbpw(adev);
 		if (amdgpu_lbpw == 0)
 			gfx_v9_0_enable_lbpw(adev, false);
 		else
 			gfx_v9_0_enable_lbpw(adev, true);
 		break;
 	case IP_VERSION(9, 4, 0):
+		gfx_v9_4_init_lbpw(adev);
 		if (amdgpu_lbpw > 0)
 			gfx_v9_0_enable_lbpw(adev, true);
 		else
@@ -2930,6 +3231,8 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev)
 		break;
 	}
 
+	gfx_v9_0_update_spm_vmid_internal(adev, 0xf);
+
 	adev->gfx.rlc.funcs->start(adev);
 
 	return 0;
@@ -2939,6 +3242,15 @@ static void gfx_v9_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable)
 {
 	u32 tmp = RREG32_SOC15(GC, 0, mmCP_ME_CNTL);
 
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_INVALIDATE_ICACHE, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_INVALIDATE_ICACHE, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_INVALIDATE_ICACHE, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_PIPE0_RESET, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_PIPE1_RESET, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE0_RESET, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_PIPE1_RESET, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE0_RESET, enable ? 0 : 1);
+	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_PIPE1_RESET, enable ? 0 : 1);
 	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, ME_HALT, enable ? 0 : 1);
 	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, PFP_HALT, enable ? 0 : 1);
 	tmp = REG_SET_FIELD(tmp, CP_ME_CNTL, CE_HALT, enable ? 0 : 1);
@@ -3016,6 +3328,14 @@ static int gfx_v9_0_cp_gfx_start(struct amdgpu_device *adev)
 
 	gfx_v9_0_cp_gfx_enable(adev, true);
 
+	/* Now only limit the quirk on the APU gfx9 series and already
+	 * confirmed that the APU gfx10/gfx11 needn't such update.
+	 */
+	if (adev->flags & AMD_IS_APU &&
+			adev->in_s3 && !pm_resume_via_firmware()) {
+		DRM_INFO("Will skip the CSB packet resubmit\n");
+		return 0;
+	}
 	r = amdgpu_ring_alloc(ring, gfx_v9_0_get_csb_size(adev) + 4 + 3);
 	if (r) {
 		DRM_ERROR("amdgpu: cp failed to lock ring (%d).\n", r);
@@ -3093,7 +3413,7 @@ static int gfx_v9_0_cp_gfx_resume(struct amdgpu_device *adev)
 	WREG32_SOC15(GC, 0, mmCP_RB0_WPTR, lower_32_bits(ring->wptr));
 	WREG32_SOC15(GC, 0, mmCP_RB0_WPTR_HI, upper_32_bits(ring->wptr));
 
-	/* set the wb address wether it's enabled or not */
+	/* set the wb address whether it's enabled or not */
 	rptr_addr = ring->rptr_gpu_addr;
 	WREG32_SOC15(GC, 0, mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr));
 	WREG32_SOC15(GC, 0, mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & CP_RB_RPTR_ADDR_HI__RB_RPTR_ADDR_HI_MASK);
@@ -3130,7 +3450,6 @@ static int gfx_v9_0_cp_gfx_resume(struct amdgpu_device *adev)
 
 	/* start the ring */
 	gfx_v9_0_cp_gfx_start(adev);
-	ring->sched.ready = true;
 
 	return 0;
 }
@@ -3141,8 +3460,16 @@ static void gfx_v9_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
 		WREG32_SOC15_RLC(GC, 0, mmCP_MEC_CNTL, 0);
 	} else {
 		WREG32_SOC15_RLC(GC, 0, mmCP_MEC_CNTL,
-			(CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK));
-		adev->gfx.kiq.ring.sched.ready = false;
+				 (CP_MEC_CNTL__MEC_INVALIDATE_ICACHE_MASK |
+				  CP_MEC_CNTL__MEC_ME1_PIPE0_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME1_PIPE1_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME1_PIPE2_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME1_PIPE3_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME2_PIPE0_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME2_PIPE1_RESET_MASK |
+				  CP_MEC_CNTL__MEC_ME1_HALT_MASK |
+				  CP_MEC_CNTL__MEC_ME2_HALT_MASK));
+		adev->gfx.kiq[0].ring.sched.ready = false;
 	}
 	udelay(50);
 }
@@ -3199,9 +3526,7 @@ static void gfx_v9_0_kiq_setting(struct amdgpu_ring *ring)
 	tmp = RREG32_SOC15(GC, 0, mmRLC_CP_SCHEDULERS);
 	tmp &= 0xffffff00;
 	tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
-	WREG32_SOC15_RLC(GC, 0, mmRLC_CP_SCHEDULERS, tmp);
-	tmp |= 0x80;
-	WREG32_SOC15_RLC(GC, 0, mmRLC_CP_SCHEDULERS, tmp);
+	WREG32_SOC15_RLC(GC, 0, mmRLC_CP_SCHEDULERS, tmp | 0x80);
 }
 
 static void gfx_v9_0_mqd_set_priority(struct amdgpu_ring *ring, struct v9_mqd *mqd)
@@ -3506,7 +3831,6 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v9_mqd *mqd = ring->mqd_ptr;
-	int mqd_idx = AMDGPU_MAX_COMPUTE_RINGS;
 	struct v9_mqd *tmp_mqd;
 
 	gfx_v9_0_kiq_setting(ring);
@@ -3516,20 +3840,20 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring)
 	 * driver need to re-init the mqd.
 	 * check mqd->cp_hqd_pq_control since this value should not be 0
 	 */
-	tmp_mqd = (struct v9_mqd *)adev->gfx.mec.mqd_backup[mqd_idx];
+	tmp_mqd = (struct v9_mqd *)adev->gfx.kiq[0].mqd_backup;
 	if (amdgpu_in_reset(adev) && tmp_mqd->cp_hqd_pq_control){
 		/* for GPU_RESET case , reset MQD to a clean status */
-		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation));
+		if (adev->gfx.kiq[0].mqd_backup)
+			memcpy(mqd, adev->gfx.kiq[0].mqd_backup, sizeof(struct v9_mqd_allocation));
 
 		/* reset ring buffer */
 		ring->wptr = 0;
 		amdgpu_ring_clear_ring(ring);
 
 		mutex_lock(&adev->srbm_mutex);
-		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0);
 		gfx_v9_0_kiq_init_register(ring);
-		soc15_grbm_select(adev, 0, 0, 0, 0);
+		soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 	} else {
 		memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation));
@@ -3538,20 +3862,20 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring)
 		if (amdgpu_sriov_vf(adev) && adev->in_suspend)
 			amdgpu_ring_clear_ring(ring);
 		mutex_lock(&adev->srbm_mutex);
-		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0);
 		gfx_v9_0_mqd_init(ring);
 		gfx_v9_0_kiq_init_register(ring);
-		soc15_grbm_select(adev, 0, 0, 0, 0);
+		soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 
-		if (adev->gfx.mec.mqd_backup[mqd_idx])
-			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation));
+		if (adev->gfx.kiq[0].mqd_backup)
+			memcpy(adev->gfx.kiq[0].mqd_backup, mqd, sizeof(struct v9_mqd_allocation));
 	}
 
 	return 0;
 }
 
-static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring)
+static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring, bool restore)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v9_mqd *mqd = ring->mqd_ptr;
@@ -3563,30 +3887,27 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring)
 	 */
 	tmp_mqd = (struct v9_mqd *)adev->gfx.mec.mqd_backup[mqd_idx];
 
-	if (!tmp_mqd->cp_hqd_pq_control ||
-	    (!amdgpu_in_reset(adev) && !adev->in_suspend)) {
+	if (!restore && (!tmp_mqd->cp_hqd_pq_control ||
+	    (!amdgpu_in_reset(adev) && !adev->in_suspend))) {
 		memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation));
 		((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
 		((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
 		mutex_lock(&adev->srbm_mutex);
-		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0);
 		gfx_v9_0_mqd_init(ring);
-		soc15_grbm_select(adev, 0, 0, 0, 0);
+		soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
 			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation));
-	} else if (amdgpu_in_reset(adev)) { /* for GPU_RESET case */
-		/* reset MQD to a clean status */
+	} else {
+		/* restore MQD to a clean status */
 		if (adev->gfx.mec.mqd_backup[mqd_idx])
 			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation));
-
 		/* reset ring buffer */
 		ring->wptr = 0;
 		atomic64_set((atomic64_t *)ring->wptr_cpu_addr, 0);
 		amdgpu_ring_clear_ring(ring);
-	} else {
-		amdgpu_ring_clear_ring(ring);
 	}
 
 	return 0;
@@ -3594,54 +3915,23 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring)
 
 static int gfx_v9_0_kiq_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring;
-	int r;
-
-	ring = &adev->gfx.kiq.ring;
-
-	r = amdgpu_bo_reserve(ring->mqd_obj, false);
-	if (unlikely(r != 0))
-		return r;
-
-	r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
-	if (unlikely(r != 0))
-		return r;
-
-	gfx_v9_0_kiq_init_queue(ring);
-	amdgpu_bo_kunmap(ring->mqd_obj);
-	ring->mqd_ptr = NULL;
-	amdgpu_bo_unreserve(ring->mqd_obj);
-	ring->sched.ready = true;
+	gfx_v9_0_kiq_init_queue(&adev->gfx.kiq[0].ring);
 	return 0;
 }
 
 static int gfx_v9_0_kcq_resume(struct amdgpu_device *adev)
 {
-	struct amdgpu_ring *ring = NULL;
-	int r = 0, i;
+	int i, r;
 
 	gfx_v9_0_cp_compute_enable(adev, true);
 
 	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-		ring = &adev->gfx.compute_ring[i];
-
-		r = amdgpu_bo_reserve(ring->mqd_obj, false);
-		if (unlikely(r != 0))
-			goto done;
-		r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr);
-		if (!r) {
-			r = gfx_v9_0_kcq_init_queue(ring);
-			amdgpu_bo_kunmap(ring->mqd_obj);
-			ring->mqd_ptr = NULL;
-		}
-		amdgpu_bo_unreserve(ring->mqd_obj);
+		r = gfx_v9_0_kcq_init_queue(&adev->gfx.compute_ring[i], false);
 		if (r)
-			goto done;
+			return r;
 	}
 
-	r = amdgpu_gfx_enable_kcq(adev);
-done:
-	return r;
+	return amdgpu_gfx_enable_kcq(adev, 0);
 }
 
 static int gfx_v9_0_cp_resume(struct amdgpu_device *adev)
@@ -3665,6 +3955,10 @@ static int gfx_v9_0_cp_resume(struct amdgpu_device *adev)
 			return r;
 	}
 
+	if (adev->gfx.num_gfx_rings)
+		gfx_v9_0_cp_gfx_enable(adev, false);
+	gfx_v9_0_cp_compute_enable(adev, false);
+
 	r = gfx_v9_0_kiq_resume(adev);
 	if (r)
 		return r;
@@ -3700,8 +3994,8 @@ static void gfx_v9_0_init_tcp_config(struct amdgpu_device *adev)
 {
 	u32 tmp;
 
-	if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 1) &&
-	    adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 2))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 1) &&
+	    amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 2))
 		return;
 
 	tmp = RREG32_SOC15(GC, 0, mmTCP_ADDR_CONFIG);
@@ -3721,10 +4015,13 @@ static void gfx_v9_0_cp_enable(struct amdgpu_device *adev, bool enable)
 	gfx_v9_0_cp_compute_enable(adev, enable);
 }
 
-static int gfx_v9_0_hw_init(void *handle)
+static int gfx_v9_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	amdgpu_gfx_cleaner_shader_init(adev, adev->gfx.cleaner_shader_size,
+				       adev->gfx.cleaner_shader_ptr);
 
 	if (!amdgpu_sriov_vf(adev))
 		gfx_v9_0_init_golden_registers(adev);
@@ -3741,24 +4038,27 @@ static int gfx_v9_0_hw_init(void *handle)
 	if (r)
 		return r;
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2) &&
+	    !amdgpu_sriov_vf(adev))
 		gfx_v9_4_2_set_power_brake_sequence(adev);
 
 	return r;
 }
 
-static int gfx_v9_0_hw_fini(void *handle)
+static int gfx_v9_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	amdgpu_irq_put(adev, &adev->gfx.cp_ecc_error_irq, 0);
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX))
+		amdgpu_irq_put(adev, &adev->gfx.cp_ecc_error_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
 	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
+	amdgpu_irq_put(adev, &adev->gfx.bad_op_irq, 0);
 
 	/* DF freeze and kcq disable will fail */
 	if (!amdgpu_ras_intr_triggered())
 		/* disable KCQ to avoid CPC touch memory not valid anymore */
-		amdgpu_gfx_disable_kcq(adev);
+		amdgpu_gfx_disable_kcq(adev, 0);
 
 	if (amdgpu_sriov_vf(adev)) {
 		gfx_v9_0_cp_gfx_enable(adev, false);
@@ -3776,11 +4076,11 @@ static int gfx_v9_0_hw_fini(void *handle)
 	 */
 	if (!amdgpu_in_reset(adev) && !adev->in_suspend) {
 		mutex_lock(&adev->srbm_mutex);
-		soc15_grbm_select(adev, adev->gfx.kiq.ring.me,
-				adev->gfx.kiq.ring.pipe,
-				adev->gfx.kiq.ring.queue, 0);
-		gfx_v9_0_kiq_fini_register(&adev->gfx.kiq.ring);
-		soc15_grbm_select(adev, 0, 0, 0, 0);
+		soc15_grbm_select(adev, adev->gfx.kiq[0].ring.me,
+				adev->gfx.kiq[0].ring.pipe,
+				adev->gfx.kiq[0].ring.queue, 0, 0);
+		gfx_v9_0_kiq_fini_register(&adev->gfx.kiq[0].ring);
+		soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 		mutex_unlock(&adev->srbm_mutex);
 	}
 
@@ -3788,7 +4088,7 @@ static int gfx_v9_0_hw_fini(void *handle)
 
 	/* Skip stopping RLC with A+A reset or when RLC controls GFX clock */
 	if ((adev->gmc.xgmi.connected_to_cpu && amdgpu_in_reset(adev)) ||
-	    (adev->ip_versions[GC_HWIP][0] >= IP_VERSION(9, 4, 2))) {
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(9, 4, 2))) {
 		dev_dbg(adev->dev, "Skipping RLC halt\n");
 		return 0;
 	}
@@ -3797,19 +4097,19 @@ static int gfx_v9_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int gfx_v9_0_suspend(void *handle)
+static int gfx_v9_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	return gfx_v9_0_hw_fini(handle);
+	return gfx_v9_0_hw_fini(ip_block);
 }
 
-static int gfx_v9_0_resume(void *handle)
+static int gfx_v9_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	return gfx_v9_0_hw_init(handle);
+	return gfx_v9_0_hw_init(ip_block);
 }
 
-static bool gfx_v9_0_is_idle(void *handle)
+static bool gfx_v9_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (REG_GET_FIELD(RREG32_SOC15(GC, 0, mmGRBM_STATUS),
 				GRBM_STATUS, GUI_ACTIVE))
@@ -3818,24 +4118,24 @@ static bool gfx_v9_0_is_idle(void *handle)
 		return true;
 }
 
-static int gfx_v9_0_wait_for_idle(void *handle)
+static int gfx_v9_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (gfx_v9_0_is_idle(handle))
+		if (gfx_v9_0_is_idle(ip_block))
 			return 0;
 		udelay(1);
 	}
 	return -ETIMEDOUT;
 }
 
-static int gfx_v9_0_soft_reset(void *handle)
+static int gfx_v9_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 grbm_soft_reset = 0;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/* GRBM_STATUS */
 	tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS);
@@ -3874,19 +4174,17 @@ static int gfx_v9_0_soft_reset(void *handle)
 		/* Disable MEC parsing/prefetching */
 		gfx_v9_0_cp_compute_enable(adev, false);
 
-		if (grbm_soft_reset) {
-			tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
-			tmp |= grbm_soft_reset;
-			dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp);
-			WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
-			tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
+		tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
+		tmp |= grbm_soft_reset;
+		dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
+		tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
 
-			udelay(50);
+		udelay(50);
 
-			tmp &= ~grbm_soft_reset;
-			WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
-			tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
-		}
+		tmp &= ~grbm_soft_reset;
+		WREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET, tmp);
+		tmp = RREG32_SOC15(GC, 0, mmGRBM_SOFT_RESET);
 
 		/* Wait a little for things to settle down */
 		udelay(50);
@@ -3900,7 +4198,7 @@ static uint64_t gfx_v9_0_kiq_read_clock(struct amdgpu_device *adev)
 	unsigned long flags;
 	uint32_t seq, reg_val_offs = 0;
 	uint64_t value = 0;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
 	struct amdgpu_ring *ring = &kiq->ring;
 
 	BUG_ON(!ring->funcs->emit_rreg);
@@ -3972,7 +4270,7 @@ static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev)
 {
 	uint64_t clock, clock_lo, clock_hi, hi_check;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 3, 0):
 		preempt_disable();
 		clock_hi = RREG32_SOC15_NO_KIQ(SMUIO, 0, mmGOLDEN_TSC_COUNT_UPPER_Renoir);
@@ -3991,7 +4289,9 @@ static uint64_t gfx_v9_0_get_gpu_clock_counter(struct amdgpu_device *adev)
 	default:
 		amdgpu_gfx_off_ctrl(adev, false);
 		mutex_lock(&adev->gfx.gpu_clock_mutex);
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 0, 1) && amdgpu_sriov_runtime(adev)) {
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) ==
+			    IP_VERSION(9, 0, 1) &&
+		    amdgpu_sriov_runtime(adev)) {
 			clock = gfx_v9_0_kiq_read_clock(adev);
 		} else {
 			WREG32_SOC15(GC, 0, mmRLC_CAPTURE_GPU_CLOCK_COUNT, 1);
@@ -4343,7 +4643,7 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev)
 	if (!ring->sched.ready)
 		return 0;
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1)) {
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1)) {
 		vgpr_init_shader_ptr = vgpr_init_compute_shader_arcturus;
 		vgpr_init_shader_size = sizeof(vgpr_init_compute_shader_arcturus);
 		vgpr_init_regs_ptr = vgpr_init_regs_arcturus;
@@ -4483,23 +4783,24 @@ static int gfx_v9_0_do_edc_gpr_workarounds(struct amdgpu_device *adev)
 	}
 
 fail:
-	amdgpu_ib_free(adev, &ib, NULL);
+	amdgpu_ib_free(&ib, NULL);
 	dma_fence_put(f);
 
 	return r;
 }
 
-static int gfx_v9_0_early_init(void *handle)
+static int gfx_v9_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->gfx.funcs = &gfx_v9_0_gfx_funcs;
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1) ||
-	    adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2))
 		adev->gfx.num_gfx_rings = 0;
 	else
 		adev->gfx.num_gfx_rings = GFX9_NUM_GFX_RINGS;
+	adev->gfx.xcc_mask = 1;
 	adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
 					  AMDGPU_MAX_COMPUTE_RINGS);
 	gfx_v9_0_set_kiq_pm4_funcs(adev);
@@ -4514,9 +4815,9 @@ static int gfx_v9_0_early_init(void *handle)
 	return gfx_v9_0_init_microcode(adev);
 }
 
-static int gfx_v9_0_ecc_late_init(void *handle)
+static int gfx_v9_0_ecc_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	/*
@@ -4533,7 +4834,7 @@ static int gfx_v9_0_ecc_late_init(void *handle)
 	}
 
 	/* requires IBs so do in late init after IB pool is initialized */
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2))
 		r = gfx_v9_4_2_do_edc_gpr_workarounds(adev);
 	else
 		r = gfx_v9_0_do_edc_gpr_workarounds(adev);
@@ -4548,9 +4849,9 @@ static int gfx_v9_0_ecc_late_init(void *handle)
 	return 0;
 }
 
-static int gfx_v9_0_late_init(void *handle)
+static int gfx_v9_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0);
@@ -4561,10 +4862,21 @@ static int gfx_v9_0_late_init(void *handle)
 	if (r)
 		return r;
 
-	r = gfx_v9_0_ecc_late_init(handle);
+	r = amdgpu_irq_get(adev, &adev->gfx.bad_op_irq, 0);
+	if (r)
+		return r;
+
+	r = gfx_v9_0_ecc_late_init(ip_block);
 	if (r)
 		return r;
 
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2))
+		gfx_v9_4_2_debug_trap_config_init(adev,
+			adev->vm_manager.first_kfd_vmid, AMDGPU_NUM_VMID);
+	else
+		gfx_v9_0_debug_trap_config_init(adev,
+			adev->vm_manager.first_kfd_vmid, AMDGPU_NUM_VMID);
+
 	return 0;
 }
 
@@ -4580,7 +4892,7 @@ static bool gfx_v9_0_is_rlc_enabled(struct amdgpu_device *adev)
 	return true;
 }
 
-static void gfx_v9_0_set_safe_mode(struct amdgpu_device *adev)
+static void gfx_v9_0_set_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	uint32_t data;
 	unsigned i;
@@ -4597,7 +4909,7 @@ static void gfx_v9_0_set_safe_mode(struct amdgpu_device *adev)
 	}
 }
 
-static void gfx_v9_0_unset_safe_mode(struct amdgpu_device *adev)
+static void gfx_v9_0_unset_safe_mode(struct amdgpu_device *adev, int xcc_id)
 {
 	uint32_t data;
 
@@ -4608,7 +4920,7 @@ static void gfx_v9_0_unset_safe_mode(struct amdgpu_device *adev)
 static void gfx_v9_0_update_gfx_cg_power_gating(struct amdgpu_device *adev,
 						bool enable)
 {
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 
 	if ((adev->pg_flags & AMD_PG_SUPPORT_GFX_PG) && enable) {
 		gfx_v9_0_enable_gfx_cg_power_gating(adev, true);
@@ -4620,7 +4932,7 @@ static void gfx_v9_0_update_gfx_cg_power_gating(struct amdgpu_device *adev,
 			gfx_v9_0_enable_gfx_pipeline_powergating(adev, false);
 	}
 
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 }
 
 static void gfx_v9_0_update_gfx_mg_power_gating(struct amdgpu_device *adev,
@@ -4647,14 +4959,12 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev
 {
 	uint32_t data, def;
 
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
-
 	/* It is disabled by HW by default */
 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) {
 		/* 1 - RLC_CGTT_MGCG_OVERRIDE */
 		def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE);
 
-		if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 2, 1))
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 2, 1))
 			data &= ~RLC_CGTT_MGCG_OVERRIDE__CPF_CGTT_SCLK_OVERRIDE_MASK;
 
 		data &= ~(RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK |
@@ -4688,7 +4998,7 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev
 		/* 1 - MGCG_OVERRIDE */
 		def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE);
 
-		if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 2, 1))
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 2, 1))
 			data |= RLC_CGTT_MGCG_OVERRIDE__CPF_CGTT_SCLK_OVERRIDE_MASK;
 
 		data |= (RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK |
@@ -4713,8 +5023,6 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev
 			WREG32_SOC15(GC, 0, mmCP_MEM_SLP_CNTL, data);
 		}
 	}
-
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
 }
 
 static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev,
@@ -4725,8 +5033,6 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev,
 	if (!adev->gfx.num_gfx_rings)
 		return;
 
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
-
 	/* Enable 3D CGCG/CGLS */
 	if (enable) {
 		/* write cmd to clear cgcg/cgls ov */
@@ -4768,8 +5074,6 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev,
 		if (def != data)
 			WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D, data);
 	}
-
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
 }
 
 static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev,
@@ -4777,8 +5081,6 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev
 {
 	uint32_t def, data;
 
-	amdgpu_gfx_rlc_enter_safe_mode(adev);
-
 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) {
 		def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE);
 		/* unset CGCG override */
@@ -4794,7 +5096,7 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev
 		/* enable cgcg FSM(0x0000363F) */
 		def = RREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL);
 
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1))
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1))
 			data = (0x2000 << RLC_CGCG_CGLS_CTRL__CGCG_GFX_IDLE_THRESHOLD__SHIFT) |
 				RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK;
 		else
@@ -4820,13 +5122,12 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev
 		if (def != data)
 			WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL, data);
 	}
-
-	amdgpu_gfx_rlc_exit_safe_mode(adev);
 }
 
 static int gfx_v9_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 					    bool enable)
 {
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 	if (enable) {
 		/* CGCG/CGLS should be enabled after MGCG/MGLS
 		 * ===  MGCG + MGLS ===
@@ -4846,15 +5147,15 @@ static int gfx_v9_0_update_gfx_clock_gating(struct amdgpu_device *adev,
 		/* ===  MGCG + MGLS === */
 		gfx_v9_0_update_medium_grain_clock_gating(adev, enable);
 	}
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 	return 0;
 }
 
-static void gfx_v9_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid)
+static void gfx_v9_0_update_spm_vmid_internal(struct amdgpu_device *adev,
+					      unsigned int vmid)
 {
 	u32 reg, data;
 
-	amdgpu_gfx_off_ctrl(adev, false);
-
 	reg = SOC15_REG_OFFSET(GC, 0, mmRLC_SPM_MC_CNTL);
 	if (amdgpu_sriov_is_pp_one_vf(adev))
 		data = RREG32_NO_KIQ(reg);
@@ -4868,6 +5169,13 @@ static void gfx_v9_0_update_spm_vmid(struct amdgpu_device *adev, unsigned vmid)
 		WREG32_SOC15_NO_KIQ(GC, 0, mmRLC_SPM_MC_CNTL, data);
 	else
 		WREG32_SOC15(GC, 0, mmRLC_SPM_MC_CNTL, data);
+}
+
+static void gfx_v9_0_update_spm_vmid(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned int vmid)
+{
+	amdgpu_gfx_off_ctrl(adev, false);
+
+	gfx_v9_0_update_spm_vmid_internal(adev, vmid);
 
 	amdgpu_gfx_off_ctrl(adev, true);
 }
@@ -4917,18 +5225,18 @@ static const struct amdgpu_rlc_funcs gfx_v9_0_rlc_funcs = {
 	.is_rlcg_access_range = gfx_v9_0_is_rlcg_access_range,
 };
 
-static int gfx_v9_0_set_powergating_state(void *handle,
+static int gfx_v9_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_PG_STATE_GATE);
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 2, 2):
 	case IP_VERSION(9, 1, 0):
 	case IP_VERSION(9, 3, 0):
 		if (!enable)
-			amdgpu_gfx_off_ctrl(adev, false);
+			amdgpu_gfx_off_ctrl_immediate(adev, false);
 
 		if (adev->pg_flags & AMD_PG_SUPPORT_RLC_SMU_HS) {
 			gfx_v9_0_enable_sck_slow_down_on_power_up(adev, true);
@@ -4950,10 +5258,10 @@ static int gfx_v9_0_set_powergating_state(void *handle,
 		gfx_v9_0_update_gfx_mg_power_gating(adev, enable);
 
 		if (enable)
-			amdgpu_gfx_off_ctrl(adev, true);
+			amdgpu_gfx_off_ctrl_immediate(adev, true);
 		break;
 	case IP_VERSION(9, 2, 1):
-		amdgpu_gfx_off_ctrl(adev, enable);
+		amdgpu_gfx_off_ctrl_immediate(adev, enable);
 		break;
 	default:
 		break;
@@ -4962,15 +5270,15 @@ static int gfx_v9_0_set_powergating_state(void *handle,
 	return 0;
 }
 
-static int gfx_v9_0_set_clockgating_state(void *handle,
+static int gfx_v9_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 2, 1):
 	case IP_VERSION(9, 4, 0):
@@ -4988,9 +5296,9 @@ static int gfx_v9_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static void gfx_v9_0_get_clockgating_state(void *handle, u64 *flags)
+static void gfx_v9_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	if (amdgpu_sriov_vf(adev))
@@ -5020,7 +5328,7 @@ static void gfx_v9_0_get_clockgating_state(void *handle, u64 *flags)
 	if (data & CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK)
 		*flags |= AMD_CG_SUPPORT_GFX_CP_LS | AMD_CG_SUPPORT_GFX_MGLS;
 
-	if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 1)) {
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 1)) {
 		/* AMD_CG_SUPPORT_GFX_3D_CGCG */
 		data = RREG32_KIQ(SOC15_REG_OFFSET(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D));
 		if (data & RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK)
@@ -5121,7 +5429,8 @@ static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
 			gfx_v9_0_ring_emit_de_meta(ring,
 						   (!amdgpu_sriov_vf(ring->adev) &&
 						   flags & AMDGPU_IB_PREEMPTED) ?
-						   true : false);
+						   true : false,
+						   job->gds_size > 0 && job->gds_base != 0);
 	}
 
 	amdgpu_ring_write(ring, header);
@@ -5132,9 +5441,70 @@ static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring,
 #endif
 		lower_32_bits(ib->gpu_addr));
 	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
+	amdgpu_ring_ib_on_emit_cntl(ring);
 	amdgpu_ring_write(ring, control);
 }
 
+static void gfx_v9_0_ring_patch_cntl(struct amdgpu_ring *ring,
+				     unsigned offset)
+{
+	u32 control = ring->ring[offset];
+
+	control |= INDIRECT_BUFFER_PRE_RESUME(1);
+	ring->ring[offset] = control;
+}
+
+static void gfx_v9_0_ring_patch_ce_meta(struct amdgpu_ring *ring,
+					unsigned offset)
+{
+	struct amdgpu_device *adev = ring->adev;
+	void *ce_payload_cpu_addr;
+	uint64_t payload_offset, payload_size;
+
+	payload_size = sizeof(struct v9_ce_ib_state);
+
+	payload_offset = offsetof(struct v9_gfx_meta_data, ce_payload);
+	ce_payload_cpu_addr = adev->virt.csa_cpu_addr + payload_offset;
+
+	if (offset + (payload_size >> 2) <= ring->buf_mask + 1) {
+		memcpy((void *)&ring->ring[offset], ce_payload_cpu_addr, payload_size);
+	} else {
+		memcpy((void *)&ring->ring[offset], ce_payload_cpu_addr,
+		       (ring->buf_mask + 1 - offset) << 2);
+		payload_size -= (ring->buf_mask + 1 - offset) << 2;
+		memcpy((void *)&ring->ring[0],
+		       ce_payload_cpu_addr + ((ring->buf_mask + 1 - offset) << 2),
+		       payload_size);
+	}
+}
+
+static void gfx_v9_0_ring_patch_de_meta(struct amdgpu_ring *ring,
+					unsigned offset)
+{
+	struct amdgpu_device *adev = ring->adev;
+	void *de_payload_cpu_addr;
+	uint64_t payload_offset, payload_size;
+
+	payload_size = sizeof(struct v9_de_ib_state);
+
+	payload_offset = offsetof(struct v9_gfx_meta_data, de_payload);
+	de_payload_cpu_addr = adev->virt.csa_cpu_addr + payload_offset;
+
+	((struct v9_de_ib_state *)de_payload_cpu_addr)->ib_completion_status =
+		IB_COMPLETION_STATUS_PREEMPTED;
+
+	if (offset + (payload_size >> 2) <= ring->buf_mask + 1) {
+		memcpy((void *)&ring->ring[offset], de_payload_cpu_addr, payload_size);
+	} else {
+		memcpy((void *)&ring->ring[offset], de_payload_cpu_addr,
+		       (ring->buf_mask + 1 - offset) << 2);
+		payload_size -= (ring->buf_mask + 1 - offset) << 2;
+		memcpy((void *)&ring->ring[0],
+		       de_payload_cpu_addr + ((ring->buf_mask + 1 - offset) << 2),
+		       payload_size);
+	}
+}
+
 static void gfx_v9_0_ring_emit_ib_compute(struct amdgpu_ring *ring,
 					  struct amdgpu_job *job,
 					  struct amdgpu_ib *ib,
@@ -5308,19 +5678,9 @@ static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring, bool resume)
 
 	cnt = (sizeof(ce_payload) >> 2) + 4 - 2;
 
-	if (ring->is_mes_queue) {
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  gfx[0].gfx_meta_data) +
-			offsetof(struct v9_gfx_meta_data, ce_payload);
-		ce_payload_gpu_addr =
-			amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		ce_payload_cpu_addr =
-			amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-	} else {
-		offset = offsetof(struct v9_gfx_meta_data, ce_payload);
-		ce_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
-		ce_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
-	}
+	offset = offsetof(struct v9_gfx_meta_data, ce_payload);
+	ce_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
+	ce_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, cnt));
 	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(2) |
@@ -5330,6 +5690,8 @@ static void gfx_v9_0_ring_emit_ce_meta(struct amdgpu_ring *ring, bool resume)
 	amdgpu_ring_write(ring, lower_32_bits(ce_payload_gpu_addr));
 	amdgpu_ring_write(ring, upper_32_bits(ce_payload_gpu_addr));
 
+	amdgpu_ring_ib_on_emit_ce(ring);
+
 	if (resume)
 		amdgpu_ring_write_multiple(ring, ce_payload_cpu_addr,
 					   sizeof(ce_payload) >> 2);
@@ -5342,7 +5704,7 @@ static int gfx_v9_0_ring_preempt_ib(struct amdgpu_ring *ring)
 {
 	int i, r = 0;
 	struct amdgpu_device *adev = ring->adev;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
 	struct amdgpu_ring *kiq_ring = &kiq->ring;
 	unsigned long flags;
 
@@ -5363,10 +5725,6 @@ static int gfx_v9_0_ring_preempt_ib(struct amdgpu_ring *ring)
 	amdgpu_ring_alloc(ring, 13);
 	gfx_v9_0_ring_emit_fence(ring, ring->trail_fence_gpu_addr,
 				 ring->trail_seq, AMDGPU_FENCE_FLAG_EXEC | AMDGPU_FENCE_FLAG_INT);
-	/*reset the CP_VMID_PREEMPT after trailing fence*/
-	amdgpu_ring_emit_wreg(ring,
-			      SOC15_REG_OFFSET(GC, 0, mmCP_VMID_PREEMPT),
-			      0x0);
 
 	/* assert IB preemption, emit the trailing fence */
 	kiq->pmf->kiq_unmap_queues(kiq_ring, ring, PREEMPT_QUEUES_NO_UNMAP,
@@ -5389,6 +5747,10 @@ static int gfx_v9_0_ring_preempt_ib(struct amdgpu_ring *ring)
 		DRM_WARN("ring %d timeout to preempt ib\n", ring->idx);
 	}
 
+	/*reset the CP_VMID_PREEMPT after trailing fence*/
+	amdgpu_ring_emit_wreg(ring,
+			      SOC15_REG_OFFSET(GC, 0, mmCP_VMID_PREEMPT),
+			      0x0);
 	amdgpu_ring_commit(ring);
 
 	/* deassert preemption condition */
@@ -5396,7 +5758,7 @@ static int gfx_v9_0_ring_preempt_ib(struct amdgpu_ring *ring)
 	return r;
 }
 
-static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume)
+static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume, bool usegds)
 {
 	struct amdgpu_device *adev = ring->adev;
 	struct v9_de_ib_state de_payload = {0};
@@ -5404,31 +5766,18 @@ static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume)
 	void *de_payload_cpu_addr;
 	int cnt;
 
-	if (ring->is_mes_queue) {
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  gfx[0].gfx_meta_data) +
-			offsetof(struct v9_gfx_meta_data, de_payload);
-		de_payload_gpu_addr =
-			amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-		de_payload_cpu_addr =
-			amdgpu_mes_ctx_get_offs_cpu_addr(ring, offset);
-
-		offset = offsetof(struct amdgpu_mes_ctx_meta_data,
-				  gfx[0].gds_backup) +
-			offsetof(struct v9_gfx_meta_data, de_payload);
-		gds_addr = amdgpu_mes_ctx_get_offs_gpu_addr(ring, offset);
-	} else {
-		offset = offsetof(struct v9_gfx_meta_data, de_payload);
-		de_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
-		de_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
+	offset = offsetof(struct v9_gfx_meta_data, de_payload);
+	de_payload_gpu_addr = amdgpu_csa_vaddr(ring->adev) + offset;
+	de_payload_cpu_addr = adev->virt.csa_cpu_addr + offset;
 
-		gds_addr = ALIGN(amdgpu_csa_vaddr(ring->adev) +
-				 AMDGPU_CSA_SIZE - adev->gds.gds_size,
-				 PAGE_SIZE);
-	}
+	gds_addr = ALIGN(amdgpu_csa_vaddr(ring->adev) +
+			 AMDGPU_CSA_SIZE - adev->gds.gds_size,
+			 PAGE_SIZE);
 
-	de_payload.gds_backup_addrlo = lower_32_bits(gds_addr);
-	de_payload.gds_backup_addrhi = upper_32_bits(gds_addr);
+	if (usegds) {
+		de_payload.gds_backup_addrlo = lower_32_bits(gds_addr);
+		de_payload.gds_backup_addrhi = upper_32_bits(gds_addr);
+	}
 
 	cnt = (sizeof(de_payload) >> 2) + 4 - 2;
 	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, cnt));
@@ -5439,6 +5788,7 @@ static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring, bool resume)
 	amdgpu_ring_write(ring, lower_32_bits(de_payload_gpu_addr));
 	amdgpu_ring_write(ring, upper_32_bits(de_payload_gpu_addr));
 
+	amdgpu_ring_ib_on_emit_de(ring);
 	if (resume)
 		amdgpu_ring_write_multiple(ring, de_payload_cpu_addr,
 					   sizeof(de_payload) >> 2);
@@ -5489,31 +5839,21 @@ static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags)
 	amdgpu_ring_write(ring, 0);
 }
 
-static unsigned gfx_v9_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring)
+static unsigned gfx_v9_0_ring_emit_init_cond_exec(struct amdgpu_ring *ring,
+						  uint64_t addr)
 {
 	unsigned ret;
 	amdgpu_ring_write(ring, PACKET3(PACKET3_COND_EXEC, 3));
-	amdgpu_ring_write(ring, lower_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, upper_32_bits(ring->cond_exe_gpu_addr));
-	amdgpu_ring_write(ring, 0); /* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	/* discard following DWs if *cond_exec_gpu_addr==0 */
+	amdgpu_ring_write(ring, 0);
 	ret = ring->wptr & ring->buf_mask;
-	amdgpu_ring_write(ring, 0x55aa55aa); /* patch dummy value later */
+	/* patch dummy value later */
+	amdgpu_ring_write(ring, 0);
 	return ret;
 }
 
-static void gfx_v9_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigned offset)
-{
-	unsigned cur;
-	BUG_ON(offset > ring->buf_mask);
-	BUG_ON(ring->ring[offset] != 0x55aa55aa);
-
-	cur = (ring->wptr - 1) & ring->buf_mask;
-	if (likely(cur > offset))
-		ring->ring[offset] = cur - offset;
-	else
-		ring->ring[offset] = (ring->ring_size>>2) - offset + cur;
-}
-
 static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg,
 				    uint32_t reg_val_offs)
 {
@@ -5586,7 +5926,9 @@ static void gfx_v9_0_ring_soft_recovery(struct amdgpu_ring *ring, unsigned vmid)
 	value = REG_SET_FIELD(value, SQ_CMD, MODE, 0x01);
 	value = REG_SET_FIELD(value, SQ_CMD, CHECK_VMID, 1);
 	value = REG_SET_FIELD(value, SQ_CMD, VM_ID, vmid);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
 	WREG32_SOC15(GC, 0, mmSQ_CMD, value);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
 }
 
 static void gfx_v9_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev,
@@ -5657,17 +5999,95 @@ static void gfx_v9_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev,
 	}
 }
 
+static u32 gfx_v9_0_get_cpc_int_cntl(struct amdgpu_device *adev,
+				     int me, int pipe)
+{
+	/*
+	 * amdgpu controls only the first MEC. That's why this function only
+	 * handles the setting of interrupts for this specific MEC. All other
+	 * pipes' interrupts are set by amdkfd.
+	 */
+	if (me != 1)
+		return 0;
+
+	switch (pipe) {
+	case 0:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE0_INT_CNTL);
+	case 1:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE1_INT_CNTL);
+	case 2:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE2_INT_CNTL);
+	case 3:
+		return SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE3_INT_CNTL);
+	default:
+		return 0;
+	}
+}
+
 static int gfx_v9_0_set_priv_reg_fault_state(struct amdgpu_device *adev,
 					     struct amdgpu_irq_src *source,
 					     unsigned type,
 					     enum amdgpu_interrupt_state state)
 {
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
 	case AMDGPU_IRQ_STATE_ENABLE:
 		WREG32_FIELD15(GC, 0, CP_INT_CNTL_RING0,
 			       PRIV_REG_INT_ENABLE,
 			       state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				/* MECs start at 1 */
+				cp_int_cntl_reg = gfx_v9_0_get_cpc_int_cntl(adev, i + 1, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+								    PRIV_REG_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int gfx_v9_0_set_bad_op_fault_state(struct amdgpu_device *adev,
+					   struct amdgpu_irq_src *source,
+					   unsigned type,
+					   enum amdgpu_interrupt_state state)
+{
+	u32 cp_int_cntl_reg, cp_int_cntl;
+	int i, j;
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		WREG32_FIELD15(GC, 0, CP_INT_CNTL_RING0,
+			       OPCODE_ERROR_INT_ENABLE,
+			       state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				/* MECs start at 1 */
+				cp_int_cntl_reg = gfx_v9_0_get_cpc_int_cntl(adev, i + 1, j);
+
+				if (cp_int_cntl_reg) {
+					cp_int_cntl = RREG32_SOC15_IP(GC, cp_int_cntl_reg);
+					cp_int_cntl = REG_SET_FIELD(cp_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+								    OPCODE_ERROR_INT_ENABLE,
+								    state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+					WREG32_SOC15_IP(GC, cp_int_cntl_reg, cp_int_cntl);
+				}
+			}
+		}
 		break;
 	default:
 		break;
@@ -5788,11 +6208,14 @@ static int gfx_v9_0_eop_irq(struct amdgpu_device *adev,
 
 	switch (me_id) {
 	case 0:
-		if (adev->gfx.num_gfx_rings &&
-		    !amdgpu_mcbp_handle_trailing_fence_irq(&adev->gfx.muxer)) {
-			/* Fence signals are handled on the software rings*/
-			for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++)
-				amdgpu_fence_process(&adev->gfx.sw_gfx_ring[i]);
+		if (adev->gfx.num_gfx_rings) {
+			if (!adev->gfx.mcbp) {
+				amdgpu_fence_process(&adev->gfx.gfx_ring[0]);
+			} else if (!amdgpu_mcbp_handle_trailing_fence_irq(&adev->gfx.muxer)) {
+				/* Fence signals are handled on the software rings*/
+				for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++)
+					amdgpu_fence_process(&adev->gfx.sw_gfx_ring[i]);
+			}
 		}
 		break;
 	case 1:
@@ -5846,6 +6269,15 @@ static int gfx_v9_0_priv_reg_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int gfx_v9_0_bad_op_irq(struct amdgpu_device *adev,
+			       struct amdgpu_irq_src *source,
+			       struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal opcode in command stream\n");
+	gfx_v9_0_fault(adev, entry);
+	return 0;
+}
+
 static int gfx_v9_0_priv_inst_irq(struct amdgpu_device *adev,
 				  struct amdgpu_irq_src *source,
 				  struct amdgpu_iv_entry *entry)
@@ -6298,7 +6730,7 @@ static const struct soc15_ras_field_entry gfx_v9_0_ras_fields[] = {
 };
 
 static int gfx_v9_0_ras_error_inject(struct amdgpu_device *adev,
-				     void *inject_if)
+				     void *inject_if, uint32_t instance_mask)
 {
 	struct ras_inject_if *info = (struct ras_inject_if *)inject_if;
 	int ret;
@@ -6337,13 +6769,13 @@ static int gfx_v9_0_ras_error_inject(struct amdgpu_device *adev,
 	block_info.value = info->value;
 
 	mutex_lock(&adev->grbm_idx_mutex);
-	ret = psp_ras_trigger_error(&adev->psp, &block_info);
+	ret = psp_ras_trigger_error(&adev->psp, &block_info, instance_mask);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	return ret;
 }
 
-static const char *vml2_mems[] = {
+static const char * const vml2_mems[] = {
 	"UTC_VML2_BANK_CACHE_0_BIGK_MEM0",
 	"UTC_VML2_BANK_CACHE_0_BIGK_MEM1",
 	"UTC_VML2_BANK_CACHE_0_4K_MEM0",
@@ -6362,7 +6794,7 @@ static const char *vml2_mems[] = {
 	"UTC_VML2_BANK_CACHE_3_4K_MEM1",
 };
 
-static const char *vml2_walker_mems[] = {
+static const char * const vml2_walker_mems[] = {
 	"UTC_VML2_CACHE_PDE0_MEM0",
 	"UTC_VML2_CACHE_PDE0_MEM1",
 	"UTC_VML2_CACHE_PDE1_MEM0",
@@ -6372,7 +6804,7 @@ static const char *vml2_walker_mems[] = {
 	"UTC_VML2_RDIF_LOG_FIFO",
 };
 
-static const char *atc_l2_cache_2m_mems[] = {
+static const char * const atc_l2_cache_2m_mems[] = {
 	"UTC_ATCL2_CACHE_2M_BANK0_WAY0_MEM",
 	"UTC_ATCL2_CACHE_2M_BANK0_WAY1_MEM",
 	"UTC_ATCL2_CACHE_2M_BANK1_WAY0_MEM",
@@ -6565,7 +6997,7 @@ static void gfx_v9_0_reset_ras_error_count(struct amdgpu_device *adev)
 	for (i = 0; i < ARRAY_SIZE(gfx_v9_0_edc_counter_regs); i++) {
 		for (j = 0; j < gfx_v9_0_edc_counter_regs[i].se_num; j++) {
 			for (k = 0; k < gfx_v9_0_edc_counter_regs[i].instance; k++) {
-				amdgpu_gfx_select_se_sh(adev, j, 0x0, k);
+				amdgpu_gfx_select_se_sh(adev, j, 0x0, k, 0);
 				RREG32(SOC15_REG_ENTRY_OFFSET(gfx_v9_0_edc_counter_regs[i]));
 			}
 		}
@@ -6627,7 +7059,7 @@ static void gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev,
 	for (i = 0; i < ARRAY_SIZE(gfx_v9_0_edc_counter_regs); i++) {
 		for (j = 0; j < gfx_v9_0_edc_counter_regs[i].se_num; j++) {
 			for (k = 0; k < gfx_v9_0_edc_counter_regs[i].instance; k++) {
-				amdgpu_gfx_select_se_sh(adev, j, 0, k);
+				amdgpu_gfx_select_se_sh(adev, j, 0, k, 0);
 				reg_value =
 					RREG32(SOC15_REG_ENTRY_OFFSET(gfx_v9_0_edc_counter_regs[i]));
 				if (reg_value)
@@ -6642,7 +7074,7 @@ static void gfx_v9_0_query_ras_error_count(struct amdgpu_device *adev,
 	err_data->ce_count += sec_count;
 	err_data->ue_count += ded_count;
 
-	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	gfx_v9_0_query_utc_edc_status(adev, err_data);
@@ -6726,6 +7158,230 @@ static void gfx_v9_0_emit_wave_limit(struct amdgpu_ring *ring, bool enable)
 	}
 }
 
+static void gfx_v9_ring_insert_nop(struct amdgpu_ring *ring, uint32_t num_nop)
+{
+	/* Header itself is a NOP packet */
+	if (num_nop == 1) {
+		amdgpu_ring_write(ring, ring->funcs->nop);
+		return;
+	}
+
+	/* Max HW optimization till 0x3ffe, followed by remaining one NOP at a time*/
+	amdgpu_ring_write(ring, PACKET3(PACKET3_NOP, min(num_nop - 2, 0x3ffe)));
+
+	/* Header is at index 0, followed by num_nops - 1 NOP packet's */
+	amdgpu_ring_insert_nop(ring, num_nop - 1);
+}
+
+static int gfx_v9_0_reset_kcq(struct amdgpu_ring *ring,
+			      unsigned int vmid,
+			      struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[0];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	unsigned long flags;
+	int i, r;
+
+	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
+		return -EINVAL;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+
+	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size)) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+
+	kiq->pmf->kiq_unmap_queues(kiq_ring, ring, RESET_QUEUES,
+				   0, 0);
+	amdgpu_ring_commit(kiq_ring);
+
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+
+	r = amdgpu_ring_test_ring(kiq_ring);
+	if (r)
+		return r;
+
+	/* make sure dequeue is complete*/
+	amdgpu_gfx_rlc_enter_safe_mode(adev, 0);
+	mutex_lock(&adev->srbm_mutex);
+	soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, 0);
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (!(RREG32_SOC15(GC, 0, mmCP_HQD_ACTIVE) & 1))
+			break;
+		udelay(1);
+	}
+	if (i >= adev->usec_timeout)
+		r = -ETIMEDOUT;
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, 0);
+	if (r) {
+		dev_err(adev->dev, "fail to wait on hqd deactive\n");
+		return r;
+	}
+
+	r = gfx_v9_0_kcq_init_queue(ring, true);
+	if (r) {
+		dev_err(adev->dev, "fail to init kcq\n");
+		return r;
+	}
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+	r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size);
+	if (r) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+	kiq->pmf->kiq_map_queues(kiq_ring, ring);
+	amdgpu_ring_commit(kiq_ring);
+	r = amdgpu_ring_test_ring(kiq_ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+	if (r) {
+		DRM_ERROR("fail to remap queue\n");
+		return r;
+	}
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+static void gfx_v9_ip_print(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k, reg, index = 0;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	for (i = 0; i < reg_count; i++)
+		drm_printf(p, "%-50s \t 0x%08x\n",
+			   gc_reg_list_9[i].reg_name,
+			   adev->gfx.ip_dump_core[i]);
+
+	/* print compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_9);
+	drm_printf(p, "\nnum_mec: %d num_pipe: %d num_queue: %d\n",
+		   adev->gfx.mec.num_mec,
+		   adev->gfx.mec.num_pipe_per_mec,
+		   adev->gfx.mec.num_queue_per_pipe);
+
+	for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+		for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+			for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+				drm_printf(p, "\nmec %d, pipe %d, queue %d\n", i, j, k);
+				for (reg = 0; reg < reg_count; reg++) {
+					if (i && gc_cp_reg_list_9[reg].reg_offset == mmCP_MEC_ME1_HEADER_DUMP)
+						drm_printf(p, "%-50s \t 0x%08x\n",
+							   "mmCP_MEC_ME2_HEADER_DUMP",
+							   adev->gfx.ip_dump_compute_queues[index + reg]);
+					else
+						drm_printf(p, "%-50s \t 0x%08x\n",
+							   gc_cp_reg_list_9[reg].reg_name,
+							   adev->gfx.ip_dump_compute_queues[index + reg]);
+				}
+				index += reg_count;
+			}
+		}
+	}
+
+}
+
+static void gfx_v9_ip_dump(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k, reg, index = 0;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9);
+
+	if (!adev->gfx.ip_dump_core || !adev->gfx.num_gfx_rings)
+		return;
+
+	amdgpu_gfx_off_ctrl(adev, false);
+	for (i = 0; i < reg_count; i++)
+		adev->gfx.ip_dump_core[i] = RREG32(SOC15_REG_ENTRY_OFFSET(gc_reg_list_9[i]));
+	amdgpu_gfx_off_ctrl(adev, true);
+
+	/* dump compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_9);
+	amdgpu_gfx_off_ctrl(adev, false);
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+		for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+			for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+				/* ME0 is for GFX so start from 1 for CP */
+				soc15_grbm_select(adev, 1 + i, j, k, 0, 0);
+
+				for (reg = 0; reg < reg_count; reg++) {
+					if (i && gc_cp_reg_list_9[reg].reg_offset == mmCP_MEC_ME1_HEADER_DUMP)
+						adev->gfx.ip_dump_compute_queues[index + reg] =
+							RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_MEC_ME2_HEADER_DUMP));
+					else
+						adev->gfx.ip_dump_compute_queues[index + reg] =
+							RREG32(SOC15_REG_ENTRY_OFFSET(
+								       gc_cp_reg_list_9[reg]));
+				}
+				index += reg_count;
+			}
+		}
+	}
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	amdgpu_gfx_off_ctrl(adev, true);
+
+}
+
+static void gfx_v9_0_ring_emit_cleaner_shader(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	/* Emit the cleaner shader */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2))
+		amdgpu_ring_write(ring, PACKET3(PACKET3_RUN_CLEANER_SHADER, 0));
+	else
+		amdgpu_ring_write(ring, PACKET3(PACKET3_RUN_CLEANER_SHADER_9_0, 0));
+
+	amdgpu_ring_write(ring, 0);  /* RESERVED field, programmed to zero */
+}
+
+static void gfx_v9_0_ring_begin_use_compute(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_ip_block *gfx_block =
+		amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GFX);
+
+	amdgpu_gfx_enforce_isolation_ring_begin_use(ring);
+
+	/* Raven and PCO APUs seem to have stability issues
+	 * with compute and gfxoff and gfx pg.  Disable gfx pg during
+	 * submission and allow again afterwards.
+	 */
+	if (gfx_block && amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 1, 0))
+		gfx_v9_0_set_powergating_state(gfx_block, AMD_PG_STATE_UNGATE);
+}
+
+static void gfx_v9_0_ring_end_use_compute(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_ip_block *gfx_block =
+		amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GFX);
+
+	/* Raven and PCO APUs seem to have stability issues
+	 * with compute and gfxoff and gfx pg.  Disable gfx pg during
+	 * submission and allow again afterwards.
+	 */
+	if (gfx_block && amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 1, 0))
+		gfx_v9_0_set_powergating_state(gfx_block, AMD_PG_STATE_GATE);
+
+	amdgpu_gfx_enforce_isolation_ring_end_use(ring);
+}
+
 static const struct amd_ip_funcs gfx_v9_0_ip_funcs = {
 	.name = "gfx_v9_0",
 	.early_init = gfx_v9_0_early_init,
@@ -6742,6 +7398,8 @@ static const struct amd_ip_funcs gfx_v9_0_ip_funcs = {
 	.set_clockgating_state = gfx_v9_0_set_clockgating_state,
 	.set_powergating_state = gfx_v9_0_set_powergating_state,
 	.get_clockgating_state = gfx_v9_0_get_clockgating_state,
+	.dump_ip_state = gfx_v9_ip_dump,
+	.print_ip_state = gfx_v9_ip_print,
 };
 
 static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = {
@@ -6750,7 +7408,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = {
 	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
 	.support_64bit_ptrs = true,
 	.secure_submission_supported = true,
-	.vmhub = AMDGPU_GFXHUB_0,
 	.get_rptr = gfx_v9_0_ring_get_rptr_gfx,
 	.get_wptr = gfx_v9_0_ring_get_wptr_gfx,
 	.set_wptr = gfx_v9_0_ring_set_wptr_gfx,
@@ -6774,7 +7431,8 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = {
 		5 + /* HDP_INVL */
 		8 + 8 + /* FENCE x2 */
 		2 + /* SWITCH_BUFFER */
-		7, /* gfx_v9_0_emit_mem_sync */
+		7 + /* gfx_v9_0_emit_mem_sync */
+		2, /* gfx_v9_0_ring_emit_cleaner_shader */
 	.emit_ib_size =	4, /* gfx_v9_0_ring_emit_ib_gfx */
 	.emit_ib = gfx_v9_0_ring_emit_ib_gfx,
 	.emit_fence = gfx_v9_0_ring_emit_fence,
@@ -6783,12 +7441,11 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = {
 	.emit_gds_switch = gfx_v9_0_ring_emit_gds_switch,
 	.emit_hdp_flush = gfx_v9_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v9_0_ring_test_ring,
-	.insert_nop = amdgpu_ring_insert_nop,
+	.insert_nop = gfx_v9_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_switch_buffer = gfx_v9_ring_emit_sb,
 	.emit_cntxcntl = gfx_v9_ring_emit_cntxcntl,
 	.init_cond_exec = gfx_v9_0_ring_emit_init_cond_exec,
-	.patch_cond_exec = gfx_v9_0_ring_emit_patch_cond_exec,
 	.preempt_ib = gfx_v9_0_ring_preempt_ib,
 	.emit_frame_cntl = gfx_v9_0_ring_emit_frame_cntl,
 	.emit_wreg = gfx_v9_0_ring_emit_wreg,
@@ -6796,6 +7453,9 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = {
 	.emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait,
 	.soft_recovery = gfx_v9_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v9_0_emit_mem_sync,
+	.emit_cleaner_shader = gfx_v9_0_ring_emit_cleaner_shader,
+	.begin_use = amdgpu_gfx_enforce_isolation_ring_begin_use,
+	.end_use = amdgpu_gfx_enforce_isolation_ring_end_use,
 };
 
 static const struct amdgpu_ring_funcs gfx_v9_0_sw_ring_funcs_gfx = {
@@ -6804,7 +7464,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_sw_ring_funcs_gfx = {
 	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
 	.support_64bit_ptrs = true,
 	.secure_submission_supported = true,
-	.vmhub = AMDGPU_GFXHUB_0,
 	.get_rptr = amdgpu_sw_ring_get_rptr_gfx,
 	.get_wptr = amdgpu_sw_ring_get_wptr_gfx,
 	.set_wptr = amdgpu_sw_ring_set_wptr_gfx,
@@ -6829,7 +7488,8 @@ static const struct amdgpu_ring_funcs gfx_v9_0_sw_ring_funcs_gfx = {
 		5 + /* HDP_INVL */
 		8 + 8 + /* FENCE x2 */
 		2 + /* SWITCH_BUFFER */
-		7, /* gfx_v9_0_emit_mem_sync */
+		7 + /* gfx_v9_0_emit_mem_sync */
+		2, /* gfx_v9_0_ring_emit_cleaner_shader */
 	.emit_ib_size =	4, /* gfx_v9_0_ring_emit_ib_gfx */
 	.emit_ib = gfx_v9_0_ring_emit_ib_gfx,
 	.emit_fence = gfx_v9_0_ring_emit_fence,
@@ -6839,18 +7499,23 @@ static const struct amdgpu_ring_funcs gfx_v9_0_sw_ring_funcs_gfx = {
 	.emit_hdp_flush = gfx_v9_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v9_0_ring_test_ring,
 	.test_ib = gfx_v9_0_ring_test_ib,
-	.insert_nop = amdgpu_sw_ring_insert_nop,
+	.insert_nop = gfx_v9_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_switch_buffer = gfx_v9_ring_emit_sb,
 	.emit_cntxcntl = gfx_v9_ring_emit_cntxcntl,
 	.init_cond_exec = gfx_v9_0_ring_emit_init_cond_exec,
-	.patch_cond_exec = gfx_v9_0_ring_emit_patch_cond_exec,
 	.emit_frame_cntl = gfx_v9_0_ring_emit_frame_cntl,
 	.emit_wreg = gfx_v9_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v9_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait,
 	.soft_recovery = gfx_v9_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v9_0_emit_mem_sync,
+	.patch_cntl = gfx_v9_0_ring_patch_cntl,
+	.patch_de = gfx_v9_0_ring_patch_de_meta,
+	.patch_ce = gfx_v9_0_ring_patch_ce_meta,
+	.emit_cleaner_shader = gfx_v9_0_ring_emit_cleaner_shader,
+	.begin_use = amdgpu_gfx_enforce_isolation_ring_begin_use,
+	.end_use = amdgpu_gfx_enforce_isolation_ring_end_use,
 };
 
 static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = {
@@ -6858,7 +7523,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = {
 	.align_mask = 0xff,
 	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
 	.support_64bit_ptrs = true,
-	.vmhub = AMDGPU_GFXHUB_0,
 	.get_rptr = gfx_v9_0_ring_get_rptr_compute,
 	.get_wptr = gfx_v9_0_ring_get_wptr_compute,
 	.set_wptr = gfx_v9_0_ring_set_wptr_compute,
@@ -6869,11 +7533,11 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = {
 		7 + /* gfx_v9_0_ring_emit_pipeline_sync */
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
-		2 + /* gfx_v9_0_ring_emit_vm_flush */
 		8 + 8 + 8 + /* gfx_v9_0_ring_emit_fence x3 for user fence, vm fence */
 		7 + /* gfx_v9_0_emit_mem_sync */
 		5 + /* gfx_v9_0_emit_wave_limit for updating mmSPI_WCL_PIPE_PERCENT_GFX register */
-		15, /* for updating 3 mmSPI_WCL_PIPE_PERCENT_CS registers */
+		15 + /* for updating 3 mmSPI_WCL_PIPE_PERCENT_CS registers */
+		2, /* gfx_v9_0_ring_emit_cleaner_shader */
 	.emit_ib_size =	7, /* gfx_v9_0_ring_emit_ib_compute */
 	.emit_ib = gfx_v9_0_ring_emit_ib_compute,
 	.emit_fence = gfx_v9_0_ring_emit_fence,
@@ -6883,13 +7547,18 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = {
 	.emit_hdp_flush = gfx_v9_0_ring_emit_hdp_flush,
 	.test_ring = gfx_v9_0_ring_test_ring,
 	.test_ib = gfx_v9_0_ring_test_ib,
-	.insert_nop = amdgpu_ring_insert_nop,
+	.insert_nop = gfx_v9_ring_insert_nop,
 	.pad_ib = amdgpu_ring_generic_pad_ib,
 	.emit_wreg = gfx_v9_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v9_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait,
+	.soft_recovery = gfx_v9_0_ring_soft_recovery,
 	.emit_mem_sync = gfx_v9_0_emit_mem_sync,
 	.emit_wave_limit = gfx_v9_0_emit_wave_limit,
+	.reset = gfx_v9_0_reset_kcq,
+	.emit_cleaner_shader = gfx_v9_0_ring_emit_cleaner_shader,
+	.begin_use = gfx_v9_0_ring_begin_use_compute,
+	.end_use = gfx_v9_0_ring_end_use_compute,
 };
 
 static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = {
@@ -6897,7 +7566,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = {
 	.align_mask = 0xff,
 	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
 	.support_64bit_ptrs = true,
-	.vmhub = AMDGPU_GFXHUB_0,
 	.get_rptr = gfx_v9_0_ring_get_rptr_compute,
 	.get_wptr = gfx_v9_0_ring_get_wptr_compute,
 	.set_wptr = gfx_v9_0_ring_set_wptr_compute,
@@ -6908,7 +7576,6 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = {
 		7 + /* gfx_v9_0_ring_emit_pipeline_sync */
 		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
 		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
-		2 + /* gfx_v9_0_ring_emit_vm_flush */
 		8 + 8 + 8, /* gfx_v9_0_ring_emit_fence_kiq x3 for user fence, vm fence */
 	.emit_ib_size =	7, /* gfx_v9_0_ring_emit_ib_compute */
 	.emit_fence = gfx_v9_0_ring_emit_fence_kiq,
@@ -6919,18 +7586,19 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = {
 	.emit_wreg = gfx_v9_0_ring_emit_wreg,
 	.emit_reg_wait = gfx_v9_0_ring_emit_reg_wait,
 	.emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait,
+	.emit_hdp_flush = gfx_v9_0_ring_emit_hdp_flush,
 };
 
 static void gfx_v9_0_set_ring_funcs(struct amdgpu_device *adev)
 {
 	int i;
 
-	adev->gfx.kiq.ring.funcs = &gfx_v9_0_ring_funcs_kiq;
+	adev->gfx.kiq[0].ring.funcs = &gfx_v9_0_ring_funcs_kiq;
 
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
 		adev->gfx.gfx_ring[i].funcs = &gfx_v9_0_ring_funcs_gfx;
 
-	if (adev->gfx.num_gfx_rings) {
+	if (adev->gfx.mcbp && adev->gfx.num_gfx_rings) {
 		for (i = 0; i < GFX9_NUM_SW_GFX_RINGS; i++)
 			adev->gfx.sw_gfx_ring[i].funcs = &gfx_v9_0_sw_ring_funcs_gfx;
 	}
@@ -6949,6 +7617,11 @@ static const struct amdgpu_irq_src_funcs gfx_v9_0_priv_reg_irq_funcs = {
 	.process = gfx_v9_0_priv_reg_irq,
 };
 
+static const struct amdgpu_irq_src_funcs gfx_v9_0_bad_op_irq_funcs = {
+	.set = gfx_v9_0_set_bad_op_fault_state,
+	.process = gfx_v9_0_bad_op_irq,
+};
+
 static const struct amdgpu_irq_src_funcs gfx_v9_0_priv_inst_irq_funcs = {
 	.set = gfx_v9_0_set_priv_inst_fault_state,
 	.process = gfx_v9_0_priv_inst_irq,
@@ -6968,6 +7641,9 @@ static void gfx_v9_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->gfx.priv_reg_irq.num_types = 1;
 	adev->gfx.priv_reg_irq.funcs = &gfx_v9_0_priv_reg_irq_funcs;
 
+	adev->gfx.bad_op_irq.num_types = 1;
+	adev->gfx.bad_op_irq.funcs = &gfx_v9_0_bad_op_irq_funcs;
+
 	adev->gfx.priv_inst_irq.num_types = 1;
 	adev->gfx.priv_inst_irq.funcs = &gfx_v9_0_priv_inst_irq_funcs;
 
@@ -6977,7 +7653,7 @@ static void gfx_v9_0_set_irq_funcs(struct amdgpu_device *adev)
 
 static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 2, 1):
 	case IP_VERSION(9, 4, 0):
@@ -6996,7 +7672,7 @@ static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev)
 static void gfx_v9_0_set_gds_init(struct amdgpu_device *adev)
 {
 	/* init asci gds info */
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 2, 1):
 	case IP_VERSION(9, 4, 0):
@@ -7018,7 +7694,7 @@ static void gfx_v9_0_set_gds_init(struct amdgpu_device *adev)
 		break;
 	}
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 0, 1):
 	case IP_VERSION(9, 4, 0):
 		adev->gds.gds_compute_max_wave_id = 0x7ff;
@@ -7106,7 +7782,7 @@ static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev,
 			mask = 1;
 			ao_bitmap = 0;
 			counter = 0;
-			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff);
+			amdgpu_gfx_select_se_sh(adev, i, j, 0xffffffff, 0);
 			gfx_v9_0_set_user_cu_inactive_bitmap(
 				adev, disable_masks[i * adev->gfx.config.max_sh_per_se + j]);
 			bitmap = gfx_v9_0_get_cu_active_bitmap(adev);
@@ -7123,7 +7799,7 @@ static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev,
 			 *    SE6,SH0 --> bitmap[2][1]
 			 *    SE7,SH0 --> bitmap[3][1]
 			 */
-			cu_info->bitmap[i % 4][j + i / 4] = bitmap;
+			cu_info->bitmap[0][i % 4][j + i / 4] = bitmap;
 
 			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) {
 				if (bitmap & mask) {
@@ -7139,7 +7815,7 @@ static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev,
 			cu_info->ao_cu_bitmap[i % 4][j + i / 4] = ao_bitmap;
 		}
 	}
-	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 	mutex_unlock(&adev->grbm_idx_mutex);
 
 	cu_info->number = active_cu_number;
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.h
index dfe8d4841f58..f9f6edc5e558 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.h
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.h
@@ -27,6 +27,6 @@
 extern const struct amdgpu_ip_block_version gfx_v9_0_ip_block;
 
 void gfx_v9_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num,
-			   u32 instance);
+			   u32 instance, int xcc_id);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0_cleaner_shader.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0_cleaner_shader.h
new file mode 100644
index 000000000000..0b6bd09b7529
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0_cleaner_shader.h
@@ -0,0 +1,68 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+/* Define the cleaner shader gfx_9_0 */
+static const u32 __maybe_unused gfx_9_0_cleaner_shader_hex[] = {
+	/* Add the cleaner shader code here */
+};
+
+/* Define the cleaner shader gfx_9_4_2 */
+static const u32 gfx_9_4_2_cleaner_shader_hex[] = {
+	0xbf068100, 0xbf84003b,
+	0xbf8a0000, 0xb07c0000,
+	0xbe8200ff, 0x00000078,
+	0xbf110802, 0x7e000280,
+	0x7e020280, 0x7e040280,
+	0x7e060280, 0x7e080280,
+	0x7e0a0280, 0x7e0c0280,
+	0x7e0e0280, 0x80828802,
+	0xbe803202, 0xbf84fff5,
+	0xbf9c0000, 0xbe8200ff,
+	0x80000000, 0x86020102,
+	0xbf840011, 0xbefe00c1,
+	0xbeff00c1, 0xd28c0001,
+	0x0001007f, 0xd28d0001,
+	0x0002027e, 0x10020288,
+	0xbe8200bf, 0xbefc00c1,
+	0xd89c2000, 0x00020201,
+	0xd89c6040, 0x00040401,
+	0x320202ff, 0x00000400,
+	0x80828102, 0xbf84fff8,
+	0xbefc00ff, 0x0000005c,
+	0xbf800000, 0xbe802c80,
+	0xbe812c80, 0xbe822c80,
+	0xbe832c80, 0x80fc847c,
+	0xbf84fffa, 0xbee60080,
+	0xbee70080, 0xbeea0180,
+	0xbeec0180, 0xbeee0180,
+	0xbef00180, 0xbef20180,
+	0xbef40180, 0xbef60180,
+	0xbef80180, 0xbefa0180,
+	0xbf810000, 0xbf8d0001,
+	0xbefc00ff, 0x0000005c,
+	0xbf800000, 0xbe802c80,
+	0xbe812c80, 0xbe822c80,
+	0xbe832c80, 0x80fc847c,
+	0xbf84fffa, 0xbee60080,
+	0xbee70080, 0xbeea01ff,
+	0x000000ee, 0xbf810000,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c
index c67e387a97f5..6028afd81690 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4.c
@@ -27,7 +27,6 @@
 #include "amdgpu_gfx.h"
 #include "soc15.h"
 #include "soc15d.h"
-#include "amdgpu_atomfirmware.h"
 #include "amdgpu_pm.h"
 
 #include "gc/gc_9_4_1_offset.h"
@@ -970,31 +969,9 @@ static void gfx_v9_4_reset_ras_error_count(struct amdgpu_device *adev)
 	WREG32_SOC15(GC, 0, mmATC_L2_CACHE_4K_DSM_INDEX, 255);
 }
 
-static int gfx_v9_4_ras_error_inject(struct amdgpu_device *adev,
-				     void *inject_if)
-{
-	struct ras_inject_if *info = (struct ras_inject_if *)inject_if;
-	int ret;
-	struct ta_ras_trigger_error_input block_info = { 0 };
-
-	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX))
-		return -EINVAL;
-
-	block_info.block_id = amdgpu_ras_block_to_ta(info->head.block);
-	block_info.sub_block_index = info->head.sub_block_index;
-	block_info.inject_error_type = amdgpu_ras_error_to_ta(info->head.type);
-	block_info.address = info->address;
-	block_info.value = info->value;
-
-	mutex_lock(&adev->grbm_idx_mutex);
-	ret = psp_ras_trigger_error(&adev->psp, &block_info);
-	mutex_unlock(&adev->grbm_idx_mutex);
-
-	return ret;
-}
-
-static const struct soc15_reg_entry gfx_v9_4_ea_err_status_regs =
-	{ SOC15_REG_ENTRY(GC, 0, mmGCEA_ERR_STATUS), 0, 1, 32 };
+static const struct soc15_reg_entry gfx_v9_4_ea_err_status_regs = {
+	SOC15_REG_ENTRY(GC, 0, mmGCEA_ERR_STATUS), 0, 1, 32
+};
 
 static void gfx_v9_4_query_ras_error_status(struct amdgpu_device *adev)
 {
@@ -1030,7 +1007,6 @@ static void gfx_v9_4_query_ras_error_status(struct amdgpu_device *adev)
 
 
 const struct amdgpu_ras_block_hw_ops  gfx_v9_4_ras_ops = {
-	.ras_error_inject = &gfx_v9_4_ras_error_inject,
 	.query_ras_error_count = &gfx_v9_4_query_ras_error_count,
 	.reset_ras_error_count = &gfx_v9_4_reset_ras_error_count,
 	.query_ras_error_status = &gfx_v9_4_query_ras_error_status,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c
index 3a797424579c..8058ea91ecaf 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.c
@@ -412,7 +412,7 @@ static int gfx_v9_4_2_run_shader(struct amdgpu_device *adev,
 	r = amdgpu_ib_schedule(ring, 1, ib, NULL, fence_ptr);
 	if (r) {
 		dev_err(adev->dev, "ib submit failed (%d).\n", r);
-		amdgpu_ib_free(adev, ib, NULL);
+		amdgpu_ib_free(ib, NULL);
 	}
 	return r;
 }
@@ -611,16 +611,16 @@ static int gfx_v9_4_2_do_sgprs_init(struct amdgpu_device *adev)
 	}
 
 disp2_failed:
-	amdgpu_ib_free(adev, &disp_ibs[2], NULL);
+	amdgpu_ib_free(&disp_ibs[2], NULL);
 	dma_fence_put(fences[2]);
 disp1_failed:
-	amdgpu_ib_free(adev, &disp_ibs[1], NULL);
+	amdgpu_ib_free(&disp_ibs[1], NULL);
 	dma_fence_put(fences[1]);
 disp0_failed:
-	amdgpu_ib_free(adev, &disp_ibs[0], NULL);
+	amdgpu_ib_free(&disp_ibs[0], NULL);
 	dma_fence_put(fences[0]);
 pro_end:
-	amdgpu_ib_free(adev, &wb_ib, NULL);
+	amdgpu_ib_free(&wb_ib, NULL);
 
 	if (r)
 		dev_info(adev->dev, "Init SGPRS Failed\n");
@@ -687,10 +687,10 @@ static int gfx_v9_4_2_do_vgprs_init(struct amdgpu_device *adev)
 	}
 
 disp_failed:
-	amdgpu_ib_free(adev, &disp_ib, NULL);
+	amdgpu_ib_free(&disp_ib, NULL);
 	dma_fence_put(fence);
 pro_end:
-	amdgpu_ib_free(adev, &wb_ib, NULL);
+	amdgpu_ib_free(&wb_ib, NULL);
 
 	if (r)
 		dev_info(adev->dev, "Init VGPRS Failed\n");
@@ -746,8 +746,18 @@ void gfx_v9_4_2_init_golden_registers(struct amdgpu_device *adev,
 			 die_id);
 		break;
 	}
+}
 
-	return;
+void gfx_v9_4_2_init_sq(struct amdgpu_device *adev)
+{
+	uint32_t data;
+
+	if (adev->gfx.mec_fw_version >= 98) {
+		adev->gmc.xnack_flags |= AMDGPU_GMC_XNACK_FLAG_CHAIN;
+		data = RREG32_SOC15(GC, 0, regSQ_CONFIG1);
+		data = REG_SET_FIELD(data, SQ_CONFIG1, DISABLE_XNACK_CHECK_IN_RETRY_DISABLE, 1);
+		WREG32_SOC15(GC, 0, regSQ_CONFIG1, data);
+	}
 }
 
 void gfx_v9_4_2_debug_trap_config_init(struct amdgpu_device *adev,
@@ -761,7 +771,7 @@ void gfx_v9_4_2_debug_trap_config_init(struct amdgpu_device *adev,
 
 	for (i = first_vmid; i < last_vmid; i++) {
 		data = 0;
-		soc15_grbm_select(adev, 0, 0, 0, i);
+		soc15_grbm_select(adev, 0, 0, 0, i, 0);
 		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
 		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_EN, 0);
 		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, EXCP_REPLACE,
@@ -769,15 +779,18 @@ void gfx_v9_4_2_debug_trap_config_init(struct amdgpu_device *adev,
 		WREG32(SOC15_REG_OFFSET(GC, 0, regSPI_GDBG_PER_VMID_CNTL), data);
 	}
 
-	soc15_grbm_select(adev, 0, 0, 0, 0);
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
 	mutex_unlock(&adev->srbm_mutex);
+
+	WREG32(SOC15_REG_OFFSET(GC, 0, regSPI_GDBG_TRAP_DATA0), 0);
+	WREG32(SOC15_REG_OFFSET(GC, 0, regSPI_GDBG_TRAP_DATA1), 0);
 }
 
 void gfx_v9_4_2_set_power_brake_sequence(struct amdgpu_device *adev)
 {
 	u32 tmp;
 
-	gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+	gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff, 0);
 
 	tmp = 0;
 	tmp = REG_SET_FIELD(tmp, GC_THROTTLE_CTRL, PATTERN_MODE, 1);
@@ -1545,8 +1558,8 @@ static void gfx_v9_4_2_log_utc_edc_count(struct amdgpu_device *adev,
 					 uint32_t ded_cnt)
 {
 	uint32_t bank, way, mem;
-	static const char *vml2_way_str[] = { "BIGK", "4K" };
-	static const char *utcl2_rounter_str[] = { "VMC", "APT" };
+	static const char * const vml2_way_str[] = { "BIGK", "4K" };
+	static const char * const utcl2_router_str[] = { "VMC", "APT" };
 
 	mem = instance % blk->num_mem_blocks;
 	way = (instance / blk->num_mem_blocks) % blk->num_ways;
@@ -1567,7 +1580,7 @@ static void gfx_v9_4_2_log_utc_edc_count(struct amdgpu_device *adev,
 		dev_info(
 			adev->dev,
 			"GFX SubBlock UTCL2_ROUTER_IFIF%d_GROUP0_%s, SED %d, DED %d\n",
-			bank, utcl2_rounter_str[mem], sec_cnt, ded_cnt);
+			bank, utcl2_router_str[mem], sec_cnt, ded_cnt);
 		break;
 	case ATC_L2_CACHE_2M:
 		dev_info(
@@ -1699,28 +1712,6 @@ static void gfx_v9_4_2_reset_ras_error_count(struct amdgpu_device *adev)
 	gfx_v9_4_2_query_utc_edc_count(adev, NULL, NULL);
 }
 
-static int gfx_v9_4_2_ras_error_inject(struct amdgpu_device *adev, void *inject_if)
-{
-	struct ras_inject_if *info = (struct ras_inject_if *)inject_if;
-	int ret;
-	struct ta_ras_trigger_error_input block_info = { 0 };
-
-	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX))
-		return -EINVAL;
-
-	block_info.block_id = amdgpu_ras_block_to_ta(info->head.block);
-	block_info.sub_block_index = info->head.sub_block_index;
-	block_info.inject_error_type = amdgpu_ras_error_to_ta(info->head.type);
-	block_info.address = info->address;
-	block_info.value = info->value;
-
-	mutex_lock(&adev->grbm_idx_mutex);
-	ret = psp_ras_trigger_error(&adev->psp, &block_info);
-	mutex_unlock(&adev->grbm_idx_mutex);
-
-	return ret;
-}
-
 static void gfx_v9_4_2_query_ea_err_status(struct amdgpu_device *adev)
 {
 	uint32_t i, j;
@@ -1930,21 +1921,9 @@ static void gfx_v9_4_2_reset_sq_timeout_status(struct amdgpu_device *adev)
 	mutex_unlock(&adev->grbm_idx_mutex);
 }
 
-static bool gfx_v9_4_2_query_uctl2_poison_status(struct amdgpu_device *adev)
-{
-	u32 status = 0;
-	struct amdgpu_vmhub *hub;
-
-	hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	status = RREG32(hub->vm_l2_pro_fault_status);
-	/* reset page fault status */
-	WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1);
 
-	return REG_GET_FIELD(status, VM_L2_PROTECTION_FAULT_STATUS, FED);
-}
 
 struct amdgpu_ras_block_hw_ops  gfx_v9_4_2_ras_ops = {
-		.ras_error_inject = &gfx_v9_4_2_ras_error_inject,
 		.query_ras_error_count = &gfx_v9_4_2_query_ras_error_count,
 		.reset_ras_error_count = &gfx_v9_4_2_reset_ras_error_count,
 		.query_ras_error_status = &gfx_v9_4_2_query_ras_error_status,
@@ -1956,5 +1935,4 @@ struct amdgpu_gfx_ras gfx_v9_4_2_ras = {
 		.hw_ops = &gfx_v9_4_2_ras_ops,
 	},
 	.enable_watchdog_timer = &gfx_v9_4_2_enable_watchdog_timer,
-	.query_utcl2_poison_status = gfx_v9_4_2_query_uctl2_poison_status,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h
index 7584624b641c..a603724c1dfc 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2.h
@@ -28,6 +28,7 @@ void gfx_v9_4_2_debug_trap_config_init(struct amdgpu_device *adev,
 				uint32_t first_vmid, uint32_t last_vmid);
 void gfx_v9_4_2_init_golden_registers(struct amdgpu_device *adev,
 				      uint32_t die_id);
+void gfx_v9_4_2_init_sq(struct amdgpu_device *adev);
 void gfx_v9_4_2_set_power_brake_sequence(struct amdgpu_device *adev);
 int gfx_v9_4_2_do_edc_gpr_workarounds(struct amdgpu_device *adev);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2_cleaner_shader.asm b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2_cleaner_shader.asm
new file mode 100644
index 000000000000..35b8cf9070bd
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_2_cleaner_shader.asm
@@ -0,0 +1,153 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+// This shader is to clean LDS, SGPRs and VGPRs. It is  first 64 Dwords or 256 bytes of 192 Dwords cleaner shader.
+//To turn this shader program on for complitaion change this to main and lower shader main to main_1
+
+// MI200 : Clear SGPRs, VGPRs and LDS
+//   Uses two kernels launched separately:
+//   1. Clean VGPRs, LDS, and lower SGPRs
+//        Launches one workgroup per CU, each workgroup with 4x wave64 per SIMD in the CU
+//        Waves are "wave64" and have 128 VGPRs each, which uses all 512 VGPRs per SIMD
+//        Waves in the workgroup share the 64KB of LDS
+//        Each wave clears SGPRs 0 - 95. Because there are 4 waves/SIMD, this is physical SGPRs 0-383
+//        Each wave clears 128 VGPRs, so all 512 in the SIMD
+//        The first wave of the workgroup clears its 64KB of LDS
+//        The shader starts with "S_BARRIER" to ensure SPI has launched all waves of the workgroup
+//          before any wave in the workgroup could end.  Without this, it is possible not all SGPRs get cleared.
+//    2. Clean remaining SGPRs
+//        Launches a workgroup with 24 waves per workgroup, yielding 6 waves per SIMD in each CU
+//        Waves are allocating 96 SGPRs
+//          CP sets up SPI_RESOURCE_RESERVE_* registers to prevent these waves from allocating SGPRs 0-223.
+//          As such, these 6 waves per SIMD are allocated physical SGPRs 224-799
+//        Barriers do not work for >16 waves per workgroup, so we cannot start with S_BARRIER
+//          Instead, the shader starts with an S_SETHALT 1. Once all waves are launched CP will send unhalt command
+//        The shader then clears all SGPRs allocated to it, cleaning out physical SGPRs 224-799
+
+shader main
+  asic(MI200)
+  type(CS)
+  wave_size(64)
+// Note: original source code from SQ team
+
+//   (theorhetical fastest = ~512clks vgpr + 1536 lds + ~128 sgpr  = 2176 clks)
+
+  s_cmp_eq_u32 s0, 1                                // Bit0 is set, sgpr0 is set then clear VGPRS and LDS as FW set COMPUTE_USER_DATA_3
+  s_cbranch_scc0  label_0023                        // Clean VGPRs and LDS if sgpr0 of wave is set, scc = (s3 == 1)
+  S_BARRIER
+
+  s_movk_i32    m0, 0x0000
+  s_mov_b32     s2, 0x00000078  // Loop 128/8=16 times  (loop unrolled for performance)
+  //
+  // CLEAR VGPRs
+  //
+  s_set_gpr_idx_on  s2, 0x8    // enable Dest VGPR indexing
+label_0005:
+  v_mov_b32     v0, 0
+  v_mov_b32     v1, 0
+  v_mov_b32     v2, 0
+  v_mov_b32     v3, 0
+  v_mov_b32     v4, 0
+  v_mov_b32     v5, 0
+  v_mov_b32     v6, 0
+  v_mov_b32     v7, 0
+  s_sub_u32     s2, s2, 8
+  s_set_gpr_idx_idx  s2
+  s_cbranch_scc0  label_0005
+  s_set_gpr_idx_off
+
+  //
+  //
+
+  s_mov_b32     s2, 0x80000000                      // Bit31 is first_wave
+  s_and_b32     s2, s2, s1                          // sgpr0 has tg_size (first_wave) term as in ucode only COMPUTE_PGM_RSRC2.tg_size_en is set
+  s_cbranch_scc0  label_clean_sgpr_1                // Clean LDS if its first wave of ThreadGroup/WorkGroup
+  // CLEAR LDS
+  //
+  s_mov_b32 exec_lo, 0xffffffff
+  s_mov_b32 exec_hi, 0xffffffff
+  v_mbcnt_lo_u32_b32  v1, exec_hi, 0          // Set V1 to thread-ID (0..63)
+  v_mbcnt_hi_u32_b32  v1, exec_lo, v1         // Set V1 to thread-ID (0..63)
+  v_mul_u32_u24  v1, 0x00000008, v1           // * 8, so each thread is a double-dword address (8byte)
+  s_mov_b32     s2, 0x00000003f               // 64 loop iterations
+  s_mov_b32     m0, 0xffffffff
+  // Clear all of LDS space
+  // Each FirstWave of WorkGroup clears 64kbyte block
+
+label_001F:
+  ds_write2_b64  v1, v[2:3], v[2:3] offset1:32
+  ds_write2_b64  v1, v[4:5], v[4:5] offset0:64 offset1:96
+  v_add_co_u32     v1, vcc, 0x00000400, v1
+  s_sub_u32     s2, s2, 1
+  s_cbranch_scc0  label_001F
+  //
+  // CLEAR SGPRs
+  //
+label_clean_sgpr_1:
+  s_mov_b32     m0, 0x0000005c   // Loop 96/4=24 times  (loop unrolled for performance)
+  s_nop 0
+label_sgpr_loop:
+  s_movreld_b32     s0, 0
+  s_movreld_b32     s1, 0
+  s_movreld_b32     s2, 0
+  s_movreld_b32     s3, 0
+  s_sub_u32         m0, m0, 4
+  s_cbranch_scc0  label_sgpr_loop
+
+  //clear vcc, flat scratch
+  s_mov_b32 flat_scratch_lo, 0   //clear  flat scratch lo SGPR
+  s_mov_b32 flat_scratch_hi, 0   //clear  flat scratch hi SGPR
+  s_mov_b64 vcc, 0               //clear vcc
+  s_mov_b64 ttmp0, 0             //Clear ttmp0 and ttmp1
+  s_mov_b64 ttmp2, 0             //Clear ttmp2 and ttmp3
+  s_mov_b64 ttmp4, 0             //Clear ttmp4 and ttmp5
+  s_mov_b64 ttmp6, 0             //Clear ttmp6 and ttmp7
+  s_mov_b64 ttmp8, 0             //Clear ttmp8 and ttmp9
+  s_mov_b64 ttmp10, 0            //Clear ttmp10 and ttmp11
+  s_mov_b64 ttmp12, 0            //Clear ttmp12 and ttmp13
+  s_mov_b64 ttmp14, 0            //Clear ttmp14 and ttmp15
+s_endpgm
+
+label_0023:
+
+  s_sethalt 1
+
+  s_mov_b32     m0, 0x0000005c   // Loop 96/4=24 times  (loop unrolled for performance)
+  s_nop 0
+label_sgpr_loop1:
+
+  s_movreld_b32     s0, 0
+  s_movreld_b32     s1, 0
+  s_movreld_b32     s2, 0
+  s_movreld_b32     s3, 0
+  s_sub_u32         m0, m0, 4
+  s_cbranch_scc0  label_sgpr_loop1
+
+  //clear vcc, flat scratch
+  s_mov_b32 flat_scratch_lo, 0   //clear  flat scratch lo SGPR
+  s_mov_b32 flat_scratch_hi, 0   //clear  flat scratch hi SGPR
+  s_mov_b64 vcc, 0xee            //clear vcc
+
+s_endpgm
+end
+
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
new file mode 100644
index 000000000000..cbb74ffc4792
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
@@ -0,0 +1,5062 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include <linux/firmware.h>
+
+#include "amdgpu.h"
+#include "amdgpu_gfx.h"
+#include "soc15.h"
+#include "soc15d.h"
+#include "soc15_common.h"
+#include "vega10_enum.h"
+
+#include "v9_structs.h"
+
+#include "ivsrcid/gfx/irqsrcs_gfx_9_0.h"
+
+#include "gc/gc_9_4_3_offset.h"
+#include "gc/gc_9_4_3_sh_mask.h"
+
+#include "gfx_v9_4_3.h"
+#include "gfx_v9_4_3_cleaner_shader.h"
+#include "amdgpu_xcp.h"
+#include "amdgpu_aca.h"
+
+MODULE_FIRMWARE("amdgpu/gc_9_4_3_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_9_4_4_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_9_5_0_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_9_4_3_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_9_4_4_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_9_5_0_rlc.bin");
+MODULE_FIRMWARE("amdgpu/gc_9_4_3_sjt_mec.bin");
+MODULE_FIRMWARE("amdgpu/gc_9_4_4_sjt_mec.bin");
+
+#define GFX9_MEC_HPD_SIZE 4096
+#define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L
+
+#define GOLDEN_GB_ADDR_CONFIG 0x2a114042
+#define CP_HQD_PERSISTENT_STATE_DEFAULT 0xbe05301
+
+#define XCC_REG_RANGE_0_LOW  0x2000     /* XCC gfxdec0 lower Bound */
+#define XCC_REG_RANGE_0_HIGH 0x3400     /* XCC gfxdec0 upper Bound */
+#define XCC_REG_RANGE_1_LOW  0xA000     /* XCC gfxdec1 lower Bound */
+#define XCC_REG_RANGE_1_HIGH 0x10000    /* XCC gfxdec1 upper Bound */
+
+#define NORMALIZE_XCC_REG_OFFSET(offset) \
+	(offset & 0xFFFF)
+
+static const struct amdgpu_hwip_reg_entry gc_reg_list_9_4_3[] = {
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS2),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_STALLED_STAT2),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_STALLED_STAT1),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_BUSY_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPF_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_GFX_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPF_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCPG_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regGDS_PROTECTION_FAULT),
+	SOC15_REG_ENTRY_STR(GC, 0, regGDS_VM_PROTECTION_FAULT),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regRMI_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSQC_DCACHE_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSQC_ICACHE_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regSQ_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regTCP_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regWD_UTCL1_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regVM_L2_PROTECTION_FAULT_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_DEBUG),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_CNTL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC1_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC2_INSTR_PNTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_CPC_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_SMU_COMMAND),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_SMU_MESSAGE),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_SMU_ARGUMENT_1),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_SMU_ARGUMENT_2),
+	SOC15_REG_ENTRY_STR(GC, 0, regSMU_RLC_RESPONSE),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_SAFE_MODE),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_SMU_SAFE_MODE),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_INT_STAT),
+	SOC15_REG_ENTRY_STR(GC, 0, regRLC_GPM_GENERAL_6),
+	/* SE status registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE0),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE1),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE2),
+	SOC15_REG_ENTRY_STR(GC, 0, regGRBM_STATUS_SE3)
+};
+
+static const struct amdgpu_hwip_reg_entry gc_cp_reg_list_9_4_3[] = {
+	/* compute queue registers */
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_VMID),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_ACTIVE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PERSISTENT_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PIPE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_QUEUE_PRIORITY),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_QUANTUM),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_BASE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_BASE_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_POLL_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_DOORBELL_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_IB_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_DEQUEUE_REQUEST),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_BASE_ADDR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_RPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_WPTR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_EVENTS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_BASE_ADDR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_BASE_ADDR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_CONTROL),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CNTL_STACK_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CNTL_STACK_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_WG_STATE_OFFSET),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_CTX_SAVE_SIZE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_GDS_RESOURCE_STATE),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_ERROR),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_EOP_WPTR_MEM),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_LO),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_PQ_WPTR_HI),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_HQD_GFX_STATUS),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+	SOC15_REG_ENTRY_STR(GC, 0, regCP_MEC_ME1_HEADER_DUMP),
+};
+
+struct amdgpu_gfx_ras gfx_v9_4_3_ras;
+
+static void gfx_v9_4_3_set_ring_funcs(struct amdgpu_device *adev);
+static void gfx_v9_4_3_set_irq_funcs(struct amdgpu_device *adev);
+static void gfx_v9_4_3_set_gds_init(struct amdgpu_device *adev);
+static void gfx_v9_4_3_set_rlc_funcs(struct amdgpu_device *adev);
+static int gfx_v9_4_3_get_cu_info(struct amdgpu_device *adev,
+				struct amdgpu_cu_info *cu_info);
+static void gfx_v9_4_3_xcc_set_safe_mode(struct amdgpu_device *adev, int xcc_id);
+static void gfx_v9_4_3_xcc_unset_safe_mode(struct amdgpu_device *adev, int xcc_id);
+
+static void gfx_v9_4_3_kiq_set_resources(struct amdgpu_ring *kiq_ring,
+				uint64_t queue_mask)
+{
+	struct amdgpu_device *adev = kiq_ring->adev;
+	u64 shader_mc_addr;
+
+	/* Cleaner shader MC address */
+	shader_mc_addr = adev->gfx.cleaner_shader_gpu_addr >> 8;
+
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_RESOURCES, 6));
+	amdgpu_ring_write(kiq_ring,
+		PACKET3_SET_RESOURCES_VMID_MASK(0) |
+		/* vmid_mask:0* queue_type:0 (KIQ) */
+		PACKET3_SET_RESOURCES_QUEUE_TYPE(0));
+	amdgpu_ring_write(kiq_ring,
+			lower_32_bits(queue_mask));	/* queue mask lo */
+	amdgpu_ring_write(kiq_ring,
+			upper_32_bits(queue_mask));	/* queue mask hi */
+	amdgpu_ring_write(kiq_ring, lower_32_bits(shader_mc_addr)); /* cleaner shader addr lo */
+	amdgpu_ring_write(kiq_ring, upper_32_bits(shader_mc_addr)); /* cleaner shader addr hi */
+	amdgpu_ring_write(kiq_ring, 0);	/* oac mask */
+	amdgpu_ring_write(kiq_ring, 0);	/* gds heap base:0, gds heap size:0 */
+}
+
+static void gfx_v9_4_3_kiq_map_queues(struct amdgpu_ring *kiq_ring,
+				 struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = kiq_ring->adev;
+	uint64_t mqd_addr = amdgpu_bo_gpu_offset(ring->mqd_obj);
+	uint64_t wptr_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4);
+	uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0;
+
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_MAP_QUEUES, 5));
+	/* Q_sel:0, vmid:0, vidmem: 1, engine:0, num_Q:1*/
+	amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
+			 PACKET3_MAP_QUEUES_QUEUE_SEL(0) | /* Queue_Sel */
+			 PACKET3_MAP_QUEUES_VMID(0) | /* VMID */
+			 PACKET3_MAP_QUEUES_QUEUE(ring->queue) |
+			 PACKET3_MAP_QUEUES_PIPE(ring->pipe) |
+			 PACKET3_MAP_QUEUES_ME((ring->me == 1 ? 0 : 1)) |
+			 /*queue_type: normal compute queue */
+			 PACKET3_MAP_QUEUES_QUEUE_TYPE(0) |
+			 /* alloc format: all_on_one_pipe */
+			 PACKET3_MAP_QUEUES_ALLOC_FORMAT(0) |
+			 PACKET3_MAP_QUEUES_ENGINE_SEL(eng_sel) |
+			 /* num_queues: must be 1 */
+			 PACKET3_MAP_QUEUES_NUM_QUEUES(1));
+	amdgpu_ring_write(kiq_ring,
+			PACKET3_MAP_QUEUES_DOORBELL_OFFSET(ring->doorbell_index));
+	amdgpu_ring_write(kiq_ring, lower_32_bits(mqd_addr));
+	amdgpu_ring_write(kiq_ring, upper_32_bits(mqd_addr));
+	amdgpu_ring_write(kiq_ring, lower_32_bits(wptr_addr));
+	amdgpu_ring_write(kiq_ring, upper_32_bits(wptr_addr));
+}
+
+static void gfx_v9_4_3_kiq_unmap_queues(struct amdgpu_ring *kiq_ring,
+				   struct amdgpu_ring *ring,
+				   enum amdgpu_unmap_queues_action action,
+				   u64 gpu_addr, u64 seq)
+{
+	uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0;
+
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4));
+	amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
+			  PACKET3_UNMAP_QUEUES_ACTION(action) |
+			  PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) |
+			  PACKET3_UNMAP_QUEUES_ENGINE_SEL(eng_sel) |
+			  PACKET3_UNMAP_QUEUES_NUM_QUEUES(1));
+	amdgpu_ring_write(kiq_ring,
+			PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index));
+
+	if (action == PREEMPT_QUEUES_NO_UNMAP) {
+		amdgpu_ring_write(kiq_ring, lower_32_bits(gpu_addr));
+		amdgpu_ring_write(kiq_ring, upper_32_bits(gpu_addr));
+		amdgpu_ring_write(kiq_ring, seq);
+	} else {
+		amdgpu_ring_write(kiq_ring, 0);
+		amdgpu_ring_write(kiq_ring, 0);
+		amdgpu_ring_write(kiq_ring, 0);
+	}
+}
+
+static void gfx_v9_4_3_kiq_query_status(struct amdgpu_ring *kiq_ring,
+				   struct amdgpu_ring *ring,
+				   u64 addr,
+				   u64 seq)
+{
+	uint32_t eng_sel = ring->funcs->type == AMDGPU_RING_TYPE_GFX ? 4 : 0;
+
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_QUERY_STATUS, 5));
+	amdgpu_ring_write(kiq_ring,
+			  PACKET3_QUERY_STATUS_CONTEXT_ID(0) |
+			  PACKET3_QUERY_STATUS_INTERRUPT_SEL(0) |
+			  PACKET3_QUERY_STATUS_COMMAND(2));
+	/* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */
+	amdgpu_ring_write(kiq_ring,
+			PACKET3_QUERY_STATUS_DOORBELL_OFFSET(ring->doorbell_index) |
+			PACKET3_QUERY_STATUS_ENG_SEL(eng_sel));
+	amdgpu_ring_write(kiq_ring, lower_32_bits(addr));
+	amdgpu_ring_write(kiq_ring, upper_32_bits(addr));
+	amdgpu_ring_write(kiq_ring, lower_32_bits(seq));
+	amdgpu_ring_write(kiq_ring, upper_32_bits(seq));
+}
+
+static void gfx_v9_4_3_kiq_invalidate_tlbs(struct amdgpu_ring *kiq_ring,
+				uint16_t pasid, uint32_t flush_type,
+				bool all_hub)
+{
+	amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_INVALIDATE_TLBS, 0));
+	amdgpu_ring_write(kiq_ring,
+			PACKET3_INVALIDATE_TLBS_DST_SEL(1) |
+			PACKET3_INVALIDATE_TLBS_ALL_HUB(all_hub) |
+			PACKET3_INVALIDATE_TLBS_PASID(pasid) |
+			PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(flush_type));
+}
+
+static void gfx_v9_4_3_kiq_reset_hw_queue(struct amdgpu_ring *kiq_ring, uint32_t queue_type,
+					  uint32_t me_id, uint32_t pipe_id, uint32_t queue_id,
+					  uint32_t xcc_id, uint32_t vmid)
+{
+	struct amdgpu_device *adev = kiq_ring->adev;
+	unsigned i;
+
+	/* enter save mode */
+	amdgpu_gfx_rlc_enter_safe_mode(adev, xcc_id);
+	mutex_lock(&adev->srbm_mutex);
+	soc15_grbm_select(adev, me_id, pipe_id, queue_id, 0, xcc_id);
+
+	if (queue_type == AMDGPU_RING_TYPE_COMPUTE) {
+		WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_DEQUEUE_REQUEST, 0x2);
+		WREG32_SOC15(GC, GET_INST(GC, xcc_id), regSPI_COMPUTE_QUEUE_RESET, 0x1);
+		/* wait till dequeue take effects */
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (!(RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+		if (i >= adev->usec_timeout)
+			dev_err(adev->dev, "fail to wait on hqd deactive\n");
+	} else {
+		dev_err(adev->dev, "reset queue_type(%d) not supported\n\n", queue_type);
+	}
+
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+	/* exit safe mode */
+	amdgpu_gfx_rlc_exit_safe_mode(adev, xcc_id);
+}
+
+static const struct kiq_pm4_funcs gfx_v9_4_3_kiq_pm4_funcs = {
+	.kiq_set_resources = gfx_v9_4_3_kiq_set_resources,
+	.kiq_map_queues = gfx_v9_4_3_kiq_map_queues,
+	.kiq_unmap_queues = gfx_v9_4_3_kiq_unmap_queues,
+	.kiq_query_status = gfx_v9_4_3_kiq_query_status,
+	.kiq_invalidate_tlbs = gfx_v9_4_3_kiq_invalidate_tlbs,
+	.kiq_reset_hw_queue = gfx_v9_4_3_kiq_reset_hw_queue,
+	.set_resources_size = 8,
+	.map_queues_size = 7,
+	.unmap_queues_size = 6,
+	.query_status_size = 7,
+	.invalidate_tlbs_size = 2,
+};
+
+static void gfx_v9_4_3_set_kiq_pm4_funcs(struct amdgpu_device *adev)
+{
+	int i, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++)
+		adev->gfx.kiq[i].pmf = &gfx_v9_4_3_kiq_pm4_funcs;
+}
+
+static void gfx_v9_4_3_init_golden_registers(struct amdgpu_device *adev)
+{
+	int i, num_xcc, dev_inst;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++) {
+		dev_inst = GET_INST(GC, i);
+
+		WREG32_SOC15(GC, dev_inst, regGB_ADDR_CONFIG,
+			     GOLDEN_GB_ADDR_CONFIG);
+		WREG32_FIELD15_PREREG(GC, dev_inst, TCP_UTCL1_CNTL2, SPARE, 0x1);
+	}
+}
+
+static uint32_t gfx_v9_4_3_normalize_xcc_reg_offset(uint32_t reg)
+{
+	uint32_t normalized_reg = NORMALIZE_XCC_REG_OFFSET(reg);
+
+	/* If it is an XCC reg, normalize the reg to keep
+	   lower 16 bits in local xcc */
+
+	if (((normalized_reg >= XCC_REG_RANGE_0_LOW) && (normalized_reg < XCC_REG_RANGE_0_HIGH)) ||
+		((normalized_reg >= XCC_REG_RANGE_1_LOW) && (normalized_reg < XCC_REG_RANGE_1_HIGH)))
+		return normalized_reg;
+	else
+		return reg;
+}
+
+static void gfx_v9_4_3_write_data_to_reg(struct amdgpu_ring *ring, int eng_sel,
+				       bool wc, uint32_t reg, uint32_t val)
+{
+	reg = gfx_v9_4_3_normalize_xcc_reg_offset(reg);
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, WRITE_DATA_ENGINE_SEL(eng_sel) |
+				WRITE_DATA_DST_SEL(0) |
+				(wc ? WR_CONFIRM : 0));
+	amdgpu_ring_write(ring, reg);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, val);
+}
+
+static void gfx_v9_4_3_wait_reg_mem(struct amdgpu_ring *ring, int eng_sel,
+				  int mem_space, int opt, uint32_t addr0,
+				  uint32_t addr1, uint32_t ref, uint32_t mask,
+				  uint32_t inv)
+{
+	/* Only do the normalization on regspace */
+	if (mem_space == 0) {
+		addr0 = gfx_v9_4_3_normalize_xcc_reg_offset(addr0);
+		addr1 = gfx_v9_4_3_normalize_xcc_reg_offset(addr1);
+	}
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5));
+	amdgpu_ring_write(ring,
+				 /* memory (1) or register (0) */
+				 (WAIT_REG_MEM_MEM_SPACE(mem_space) |
+				 WAIT_REG_MEM_OPERATION(opt) | /* wait */
+				 WAIT_REG_MEM_FUNCTION(3) |  /* equal */
+				 WAIT_REG_MEM_ENGINE(eng_sel)));
+
+	if (mem_space)
+		BUG_ON(addr0 & 0x3); /* Dword align */
+	amdgpu_ring_write(ring, addr0);
+	amdgpu_ring_write(ring, addr1);
+	amdgpu_ring_write(ring, ref);
+	amdgpu_ring_write(ring, mask);
+	amdgpu_ring_write(ring, inv); /* poll interval */
+}
+
+static int gfx_v9_4_3_ring_test_ring(struct amdgpu_ring *ring)
+{
+	uint32_t scratch_reg0_offset, xcc_offset;
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t tmp = 0;
+	unsigned i;
+	int r;
+
+	/* Use register offset which is local to XCC in the packet */
+	xcc_offset = SOC15_REG_OFFSET(GC, 0, regSCRATCH_REG0);
+	scratch_reg0_offset = SOC15_REG_OFFSET(GC, GET_INST(GC, ring->xcc_id), regSCRATCH_REG0);
+	WREG32(scratch_reg0_offset, 0xCAFEDEAD);
+	tmp = RREG32(scratch_reg0_offset);
+
+	r = amdgpu_ring_alloc(ring, 3);
+	if (r)
+		return r;
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
+	amdgpu_ring_write(ring, xcc_offset - PACKET3_SET_UCONFIG_REG_START);
+	amdgpu_ring_write(ring, 0xDEADBEEF);
+	amdgpu_ring_commit(ring);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32(scratch_reg0_offset);
+		if (tmp == 0xDEADBEEF)
+			break;
+		udelay(1);
+	}
+
+	if (i >= adev->usec_timeout)
+		r = -ETIMEDOUT;
+	return r;
+}
+
+static int gfx_v9_4_3_ring_test_ib(struct amdgpu_ring *ring, long timeout)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_ib ib;
+	struct dma_fence *f = NULL;
+
+	unsigned index;
+	uint64_t gpu_addr;
+	uint32_t tmp;
+	long r;
+
+	r = amdgpu_device_wb_get(adev, &index);
+	if (r)
+		return r;
+
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD);
+	memset(&ib, 0, sizeof(ib));
+
+	r = amdgpu_ib_get(adev, NULL, 20, AMDGPU_IB_POOL_DIRECT, &ib);
+	if (r)
+		goto err1;
+
+	ib.ptr[0] = PACKET3(PACKET3_WRITE_DATA, 3);
+	ib.ptr[1] = WRITE_DATA_DST_SEL(5) | WR_CONFIRM;
+	ib.ptr[2] = lower_32_bits(gpu_addr);
+	ib.ptr[3] = upper_32_bits(gpu_addr);
+	ib.ptr[4] = 0xDEADBEEF;
+	ib.length_dw = 5;
+
+	r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f);
+	if (r)
+		goto err2;
+
+	r = dma_fence_wait_timeout(f, false, timeout);
+	if (r == 0) {
+		r = -ETIMEDOUT;
+		goto err2;
+	} else if (r < 0) {
+		goto err2;
+	}
+
+	tmp = adev->wb.wb[index];
+	if (tmp == 0xDEADBEEF)
+		r = 0;
+	else
+		r = -EINVAL;
+
+err2:
+	amdgpu_ib_free(&ib, NULL);
+	dma_fence_put(f);
+err1:
+	amdgpu_device_wb_free(adev, index);
+	return r;
+}
+
+
+/* This value might differs per partition */
+static uint64_t gfx_v9_4_3_get_gpu_clock_counter(struct amdgpu_device *adev)
+{
+	uint64_t clock;
+
+	mutex_lock(&adev->gfx.gpu_clock_mutex);
+	WREG32_SOC15(GC, GET_INST(GC, 0), regRLC_CAPTURE_GPU_CLOCK_COUNT, 1);
+	clock = (uint64_t)RREG32_SOC15(GC, GET_INST(GC, 0), regRLC_GPU_CLOCK_COUNT_LSB) |
+		((uint64_t)RREG32_SOC15(GC, GET_INST(GC, 0), regRLC_GPU_CLOCK_COUNT_MSB) << 32ULL);
+	mutex_unlock(&adev->gfx.gpu_clock_mutex);
+
+	return clock;
+}
+
+static void gfx_v9_4_3_free_microcode(struct amdgpu_device *adev)
+{
+	amdgpu_ucode_release(&adev->gfx.pfp_fw);
+	amdgpu_ucode_release(&adev->gfx.me_fw);
+	amdgpu_ucode_release(&adev->gfx.ce_fw);
+	amdgpu_ucode_release(&adev->gfx.rlc_fw);
+	amdgpu_ucode_release(&adev->gfx.mec_fw);
+	amdgpu_ucode_release(&adev->gfx.mec2_fw);
+
+	kfree(adev->gfx.rlc.register_list_format);
+}
+
+static int gfx_v9_4_3_init_rlc_microcode(struct amdgpu_device *adev,
+					  const char *chip_name)
+{
+	int err;
+	const struct rlc_firmware_header_v2_0 *rlc_hdr;
+	uint16_t version_major;
+	uint16_t version_minor;
+
+
+	err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw,
+				   AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_rlc.bin", chip_name);
+	if (err)
+		goto out;
+	rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
+
+	version_major = le16_to_cpu(rlc_hdr->header.header_version_major);
+	version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor);
+	err = amdgpu_gfx_rlc_init_microcode(adev, version_major, version_minor);
+out:
+	if (err)
+		amdgpu_ucode_release(&adev->gfx.rlc_fw);
+
+	return err;
+}
+
+static int gfx_v9_4_3_init_cp_compute_microcode(struct amdgpu_device *adev,
+					  const char *chip_name)
+{
+	int err;
+
+	if (amdgpu_sriov_vf(adev)) {
+		err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_sjt_mec.bin", chip_name);
+
+		if (err)
+			err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+							AMDGPU_UCODE_REQUIRED,
+							"amdgpu/%s_mec.bin", chip_name);
+	} else
+		err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw,
+					   AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_mec.bin", chip_name);
+	if (err)
+		goto out;
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1);
+	amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1_JT);
+
+	adev->gfx.mec2_fw_version = adev->gfx.mec_fw_version;
+	adev->gfx.mec2_feature_version = adev->gfx.mec_feature_version;
+
+out:
+	if (err)
+		amdgpu_ucode_release(&adev->gfx.mec_fw);
+	return err;
+}
+
+static int gfx_v9_4_3_init_microcode(struct amdgpu_device *adev)
+{
+	char ucode_prefix[15];
+	int r;
+
+	amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix));
+
+	r = gfx_v9_4_3_init_rlc_microcode(adev, ucode_prefix);
+	if (r)
+		return r;
+
+	r = gfx_v9_4_3_init_cp_compute_microcode(adev, ucode_prefix);
+	if (r)
+		return r;
+
+	return r;
+}
+
+static void gfx_v9_4_3_mec_fini(struct amdgpu_device *adev)
+{
+	amdgpu_bo_free_kernel(&adev->gfx.mec.hpd_eop_obj, NULL, NULL);
+	amdgpu_bo_free_kernel(&adev->gfx.mec.mec_fw_obj, NULL, NULL);
+}
+
+static int gfx_v9_4_3_mec_init(struct amdgpu_device *adev)
+{
+	int r, i, num_xcc;
+	u32 *hpd;
+	const __le32 *fw_data;
+	unsigned fw_size;
+	u32 *fw;
+	size_t mec_hpd_size;
+
+	const struct gfx_firmware_header_v1_0 *mec_hdr;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++)
+		bitmap_zero(adev->gfx.mec_bitmap[i].queue_bitmap,
+			AMDGPU_MAX_COMPUTE_QUEUES);
+
+	/* take ownership of the relevant compute queues */
+	amdgpu_gfx_compute_queue_acquire(adev);
+	mec_hpd_size =
+		adev->gfx.num_compute_rings * num_xcc * GFX9_MEC_HPD_SIZE;
+	if (mec_hpd_size) {
+		r = amdgpu_bo_create_reserved(adev, mec_hpd_size, PAGE_SIZE,
+					      AMDGPU_GEM_DOMAIN_VRAM |
+					      AMDGPU_GEM_DOMAIN_GTT,
+					      &adev->gfx.mec.hpd_eop_obj,
+					      &adev->gfx.mec.hpd_eop_gpu_addr,
+					      (void **)&hpd);
+		if (r) {
+			dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r);
+			gfx_v9_4_3_mec_fini(adev);
+			return r;
+		}
+
+		if (amdgpu_emu_mode == 1) {
+			for (i = 0; i < mec_hpd_size / 4; i++) {
+				memset((void *)(hpd + i), 0, 4);
+				if (i % 50 == 0)
+					msleep(1);
+			}
+		} else {
+			memset(hpd, 0, mec_hpd_size);
+		}
+
+		amdgpu_bo_kunmap(adev->gfx.mec.hpd_eop_obj);
+		amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj);
+	}
+
+	mec_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data;
+
+	fw_data = (const __le32 *)
+		(adev->gfx.mec_fw->data +
+		 le32_to_cpu(mec_hdr->header.ucode_array_offset_bytes));
+	fw_size = le32_to_cpu(mec_hdr->header.ucode_size_bytes);
+
+	r = amdgpu_bo_create_reserved(adev, mec_hdr->header.ucode_size_bytes,
+				      PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT,
+				      &adev->gfx.mec.mec_fw_obj,
+				      &adev->gfx.mec.mec_fw_gpu_addr,
+				      (void **)&fw);
+	if (r) {
+		dev_warn(adev->dev, "(%d) create mec firmware bo failed\n", r);
+		gfx_v9_4_3_mec_fini(adev);
+		return r;
+	}
+
+	memcpy(fw, fw_data, fw_size);
+
+	amdgpu_bo_kunmap(adev->gfx.mec.mec_fw_obj);
+	amdgpu_bo_unreserve(adev->gfx.mec.mec_fw_obj);
+
+	return 0;
+}
+
+static void gfx_v9_4_3_xcc_select_se_sh(struct amdgpu_device *adev, u32 se_num,
+					u32 sh_num, u32 instance, int xcc_id)
+{
+	u32 data;
+
+	if (instance == 0xffffffff)
+		data = REG_SET_FIELD(0, GRBM_GFX_INDEX,
+				     INSTANCE_BROADCAST_WRITES, 1);
+	else
+		data = REG_SET_FIELD(0, GRBM_GFX_INDEX,
+				     INSTANCE_INDEX, instance);
+
+	if (se_num == 0xffffffff)
+		data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
+				     SE_BROADCAST_WRITES, 1);
+	else
+		data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SE_INDEX, se_num);
+
+	if (sh_num == 0xffffffff)
+		data = REG_SET_FIELD(data, GRBM_GFX_INDEX,
+				     SH_BROADCAST_WRITES, 1);
+	else
+		data = REG_SET_FIELD(data, GRBM_GFX_INDEX, SH_INDEX, sh_num);
+
+	WREG32_SOC15_RLC_SHADOW_EX(reg, GC, GET_INST(GC, xcc_id), regGRBM_GFX_INDEX, data);
+}
+
+static uint32_t wave_read_ind(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd, uint32_t wave, uint32_t address)
+{
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regSQ_IND_INDEX,
+		(wave << SQ_IND_INDEX__WAVE_ID__SHIFT) |
+		(simd << SQ_IND_INDEX__SIMD_ID__SHIFT) |
+		(address << SQ_IND_INDEX__INDEX__SHIFT) |
+		(SQ_IND_INDEX__FORCE_READ_MASK));
+	return RREG32_SOC15(GC, GET_INST(GC, xcc_id), regSQ_IND_DATA);
+}
+
+static void wave_read_regs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
+			   uint32_t wave, uint32_t thread,
+			   uint32_t regno, uint32_t num, uint32_t *out)
+{
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regSQ_IND_INDEX,
+		(wave << SQ_IND_INDEX__WAVE_ID__SHIFT) |
+		(simd << SQ_IND_INDEX__SIMD_ID__SHIFT) |
+		(regno << SQ_IND_INDEX__INDEX__SHIFT) |
+		(thread << SQ_IND_INDEX__THREAD_ID__SHIFT) |
+		(SQ_IND_INDEX__FORCE_READ_MASK) |
+		(SQ_IND_INDEX__AUTO_INCR_MASK));
+	while (num--)
+		*(out++) = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regSQ_IND_DATA);
+}
+
+static void gfx_v9_4_3_read_wave_data(struct amdgpu_device *adev,
+				      uint32_t xcc_id, uint32_t simd, uint32_t wave,
+				      uint32_t *dst, int *no_fields)
+{
+	/* type 1 wave data */
+	dst[(*no_fields)++] = 1;
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_STATUS);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_PC_LO);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_PC_HI);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_EXEC_LO);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_EXEC_HI);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_HW_ID);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_INST_DW0);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_INST_DW1);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_GPR_ALLOC);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_LDS_ALLOC);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_TRAPSTS);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_IB_STS);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_IB_DBG0);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_M0);
+	dst[(*no_fields)++] = wave_read_ind(adev, xcc_id, simd, wave, ixSQ_WAVE_MODE);
+}
+
+static void gfx_v9_4_3_read_wave_sgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
+				       uint32_t wave, uint32_t start,
+				       uint32_t size, uint32_t *dst)
+{
+	wave_read_regs(adev, xcc_id, simd, wave, 0,
+		       start + SQIND_WAVE_SGPRS_OFFSET, size, dst);
+}
+
+static void gfx_v9_4_3_read_wave_vgprs(struct amdgpu_device *adev, uint32_t xcc_id, uint32_t simd,
+				       uint32_t wave, uint32_t thread,
+				       uint32_t start, uint32_t size,
+				       uint32_t *dst)
+{
+	wave_read_regs(adev, xcc_id, simd, wave, thread,
+		       start + SQIND_WAVE_VGPRS_OFFSET, size, dst);
+}
+
+static void gfx_v9_4_3_select_me_pipe_q(struct amdgpu_device *adev,
+					u32 me, u32 pipe, u32 q, u32 vm, u32 xcc_id)
+{
+	soc15_grbm_select(adev, me, pipe, q, vm, GET_INST(GC, xcc_id));
+}
+
+static int gfx_v9_4_3_get_xccs_per_xcp(struct amdgpu_device *adev)
+{
+	u32 xcp_ctl;
+
+	/* Value is expected to be the same on all, fetch from first instance */
+	xcp_ctl = RREG32_SOC15(GC, GET_INST(GC, 0), regCP_HYP_XCP_CTL);
+
+	return REG_GET_FIELD(xcp_ctl, CP_HYP_XCP_CTL, NUM_XCC_IN_XCP);
+}
+
+static int gfx_v9_4_3_switch_compute_partition(struct amdgpu_device *adev,
+						int num_xccs_per_xcp)
+{
+	int ret, i, num_xcc;
+	u32 tmp = 0;
+
+	if (adev->psp.funcs) {
+		ret = psp_spatial_partition(&adev->psp,
+					    NUM_XCC(adev->gfx.xcc_mask) /
+						    num_xccs_per_xcp);
+		if (ret)
+			return ret;
+	} else {
+		num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+
+		for (i = 0; i < num_xcc; i++) {
+			tmp = REG_SET_FIELD(tmp, CP_HYP_XCP_CTL, NUM_XCC_IN_XCP,
+					    num_xccs_per_xcp);
+			tmp = REG_SET_FIELD(tmp, CP_HYP_XCP_CTL, VIRTUAL_XCC_ID,
+					    i % num_xccs_per_xcp);
+			WREG32_SOC15(GC, GET_INST(GC, i), regCP_HYP_XCP_CTL,
+				     tmp);
+		}
+		ret = 0;
+	}
+
+	adev->gfx.num_xcc_per_xcp = num_xccs_per_xcp;
+
+	return ret;
+}
+
+static int gfx_v9_4_3_ih_to_xcc_inst(struct amdgpu_device *adev, int ih_node)
+{
+	int xcc;
+
+	xcc = hweight8(adev->gfx.xcc_mask & GENMASK(ih_node / 2, 0));
+	if (!xcc) {
+		dev_err(adev->dev, "Couldn't find xcc mapping from IH node");
+		return -EINVAL;
+	}
+
+	return xcc - 1;
+}
+
+static const struct amdgpu_gfx_funcs gfx_v9_4_3_gfx_funcs = {
+	.get_gpu_clock_counter = &gfx_v9_4_3_get_gpu_clock_counter,
+	.select_se_sh = &gfx_v9_4_3_xcc_select_se_sh,
+	.read_wave_data = &gfx_v9_4_3_read_wave_data,
+	.read_wave_sgprs = &gfx_v9_4_3_read_wave_sgprs,
+	.read_wave_vgprs = &gfx_v9_4_3_read_wave_vgprs,
+	.select_me_pipe_q = &gfx_v9_4_3_select_me_pipe_q,
+	.switch_partition_mode = &gfx_v9_4_3_switch_compute_partition,
+	.ih_node_to_logical_xcc = &gfx_v9_4_3_ih_to_xcc_inst,
+	.get_xccs_per_xcp = &gfx_v9_4_3_get_xccs_per_xcp,
+};
+
+static int gfx_v9_4_3_aca_bank_parser(struct aca_handle *handle,
+				      struct aca_bank *bank, enum aca_smu_type type,
+				      void *data)
+{
+	struct aca_bank_info info;
+	u64 misc0;
+	u32 instlo;
+	int ret;
+
+	ret = aca_bank_info_decode(bank, &info);
+	if (ret)
+		return ret;
+
+	/* NOTE: overwrite info.die_id with xcd id for gfx */
+	instlo = ACA_REG__IPID__INSTANCEIDLO(bank->regs[ACA_REG_IDX_IPID]);
+	instlo &= GENMASK(31, 1);
+	info.die_id = instlo == mmSMNAID_XCD0_MCA_SMU ? 0 : 1;
+
+	misc0 = bank->regs[ACA_REG_IDX_MISC0];
+
+	switch (type) {
+	case ACA_SMU_TYPE_UE:
+		bank->aca_err_type = ACA_ERROR_TYPE_UE;
+		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type, 1ULL);
+		break;
+	case ACA_SMU_TYPE_CE:
+		bank->aca_err_type = ACA_ERROR_TYPE_CE;
+		ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type,
+						     ACA_REG__MISC0__ERRCNT(misc0));
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return ret;
+}
+
+static bool gfx_v9_4_3_aca_bank_is_valid(struct aca_handle *handle, struct aca_bank *bank,
+					 enum aca_smu_type type, void *data)
+{
+	u32 instlo;
+
+	instlo = ACA_REG__IPID__INSTANCEIDLO(bank->regs[ACA_REG_IDX_IPID]);
+	instlo &= GENMASK(31, 1);
+	switch (instlo) {
+	case mmSMNAID_XCD0_MCA_SMU:
+	case mmSMNAID_XCD1_MCA_SMU:
+	case mmSMNXCD_XCD0_MCA_SMU:
+		return true;
+	default:
+		break;
+	}
+
+	return false;
+}
+
+static const struct aca_bank_ops gfx_v9_4_3_aca_bank_ops = {
+	.aca_bank_parser = gfx_v9_4_3_aca_bank_parser,
+	.aca_bank_is_valid = gfx_v9_4_3_aca_bank_is_valid,
+};
+
+static const struct aca_info gfx_v9_4_3_aca_info = {
+	.hwip = ACA_HWIP_TYPE_SMU,
+	.mask = ACA_ERROR_UE_MASK | ACA_ERROR_CE_MASK,
+	.bank_ops = &gfx_v9_4_3_aca_bank_ops,
+};
+
+static int gfx_v9_4_3_gpu_early_init(struct amdgpu_device *adev)
+{
+	adev->gfx.funcs = &gfx_v9_4_3_gfx_funcs;
+	adev->gfx.ras = &gfx_v9_4_3_ras;
+
+	adev->gfx.config.max_hw_contexts = 8;
+	adev->gfx.config.sc_prim_fifo_size_frontend = 0x20;
+	adev->gfx.config.sc_prim_fifo_size_backend = 0x100;
+	adev->gfx.config.sc_hiz_tile_fifo_size = 0x30;
+	adev->gfx.config.sc_earlyz_tile_fifo_size = 0x4C0;
+	adev->gfx.config.gb_addr_config = GOLDEN_GB_ADDR_CONFIG;
+
+	adev->gfx.config.gb_addr_config_fields.num_pipes = 1 <<
+			REG_GET_FIELD(
+					adev->gfx.config.gb_addr_config,
+					GB_ADDR_CONFIG,
+					NUM_PIPES);
+
+	adev->gfx.config.max_tile_pipes =
+		adev->gfx.config.gb_addr_config_fields.num_pipes;
+
+	adev->gfx.config.gb_addr_config_fields.num_banks = 1 <<
+			REG_GET_FIELD(
+					adev->gfx.config.gb_addr_config,
+					GB_ADDR_CONFIG,
+					NUM_BANKS);
+	adev->gfx.config.gb_addr_config_fields.max_compress_frags = 1 <<
+			REG_GET_FIELD(
+					adev->gfx.config.gb_addr_config,
+					GB_ADDR_CONFIG,
+					MAX_COMPRESSED_FRAGS);
+	adev->gfx.config.gb_addr_config_fields.num_rb_per_se = 1 <<
+			REG_GET_FIELD(
+					adev->gfx.config.gb_addr_config,
+					GB_ADDR_CONFIG,
+					NUM_RB_PER_SE);
+	adev->gfx.config.gb_addr_config_fields.num_se = 1 <<
+			REG_GET_FIELD(
+					adev->gfx.config.gb_addr_config,
+					GB_ADDR_CONFIG,
+					NUM_SHADER_ENGINES);
+	adev->gfx.config.gb_addr_config_fields.pipe_interleave_size = 1 << (8 +
+			REG_GET_FIELD(
+					adev->gfx.config.gb_addr_config,
+					GB_ADDR_CONFIG,
+					PIPE_INTERLEAVE_SIZE));
+
+	return 0;
+}
+
+static int gfx_v9_4_3_compute_ring_init(struct amdgpu_device *adev, int ring_id,
+				        int xcc_id, int mec, int pipe, int queue)
+{
+	unsigned irq_type;
+	struct amdgpu_ring *ring = &adev->gfx.compute_ring[ring_id];
+	unsigned int hw_prio;
+	uint32_t xcc_doorbell_start;
+
+	ring = &adev->gfx.compute_ring[xcc_id * adev->gfx.num_compute_rings +
+				       ring_id];
+
+	/* mec0 is me1 */
+	ring->xcc_id = xcc_id;
+	ring->me = mec + 1;
+	ring->pipe = pipe;
+	ring->queue = queue;
+
+	ring->ring_obj = NULL;
+	ring->use_doorbell = true;
+	xcc_doorbell_start = adev->doorbell_index.mec_ring0 +
+			     xcc_id * adev->doorbell_index.xcc_doorbell_range;
+	ring->doorbell_index = (xcc_doorbell_start + ring_id) << 1;
+	ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr +
+			     (ring_id + xcc_id * adev->gfx.num_compute_rings) *
+				     GFX9_MEC_HPD_SIZE;
+	ring->vm_hub = AMDGPU_GFXHUB(xcc_id);
+	sprintf(ring->name, "comp_%d.%d.%d.%d",
+			ring->xcc_id, ring->me, ring->pipe, ring->queue);
+
+	irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP
+		+ ((ring->me - 1) * adev->gfx.mec.num_pipe_per_mec)
+		+ ring->pipe;
+	hw_prio = amdgpu_gfx_is_high_priority_compute_queue(adev, ring) ?
+			AMDGPU_GFX_PIPE_PRIO_HIGH : AMDGPU_GFX_PIPE_PRIO_NORMAL;
+	/* type-2 packets are deprecated on MEC, use type-3 instead */
+	return amdgpu_ring_init(adev, ring, 1024, &adev->gfx.eop_irq, irq_type,
+				hw_prio, NULL);
+}
+
+static void gfx_v9_4_3_alloc_ip_dump(struct amdgpu_device *adev)
+{
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9_4_3);
+	uint32_t *ptr, num_xcc, inst;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+
+	ptr = kcalloc(reg_count * num_xcc, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for GFX IP Dump\n");
+		adev->gfx.ip_dump_core = NULL;
+	} else {
+		adev->gfx.ip_dump_core = ptr;
+	}
+
+	/* Allocate memory for compute queue registers for all the instances */
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_9_4_3);
+	inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec *
+		adev->gfx.mec.num_queue_per_pipe;
+
+	ptr = kcalloc(reg_count * inst * num_xcc, sizeof(uint32_t), GFP_KERNEL);
+	if (!ptr) {
+		DRM_ERROR("Failed to allocate memory for Compute Queues IP Dump\n");
+		adev->gfx.ip_dump_compute_queues = NULL;
+	} else {
+		adev->gfx.ip_dump_compute_queues = ptr;
+	}
+}
+
+static int gfx_v9_4_3_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	int i, j, k, r, ring_id, xcc_id, num_xcc;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+		adev->gfx.cleaner_shader_ptr = gfx_9_4_3_cleaner_shader_hex;
+		adev->gfx.cleaner_shader_size = sizeof(gfx_9_4_3_cleaner_shader_hex);
+		if (adev->gfx.mec_fw_version >= 153) {
+			adev->gfx.enable_cleaner_shader = true;
+			r = amdgpu_gfx_cleaner_shader_sw_init(adev, adev->gfx.cleaner_shader_size);
+			if (r) {
+				adev->gfx.enable_cleaner_shader = false;
+				dev_err(adev->dev, "Failed to initialize cleaner shader\n");
+			}
+		}
+		break;
+	default:
+		adev->gfx.enable_cleaner_shader = false;
+		break;
+	}
+
+	adev->gfx.mec.num_mec = 2;
+	adev->gfx.mec.num_pipe_per_mec = 4;
+	adev->gfx.mec.num_queue_per_pipe = 8;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+
+	/* EOP Event */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_9_0__SRCID__CP_EOP_INTERRUPT, &adev->gfx.eop_irq);
+	if (r)
+		return r;
+
+	/* Bad opcode Event */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP,
+			      GFX_9_0__SRCID__CP_BAD_OPCODE_ERROR,
+			      &adev->gfx.bad_op_irq);
+	if (r)
+		return r;
+
+	/* Privileged reg */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_9_0__SRCID__CP_PRIV_REG_FAULT,
+			      &adev->gfx.priv_reg_irq);
+	if (r)
+		return r;
+
+	/* Privileged inst */
+	r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_9_0__SRCID__CP_PRIV_INSTR_FAULT,
+			      &adev->gfx.priv_inst_irq);
+	if (r)
+		return r;
+
+	adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE;
+
+	r = adev->gfx.rlc.funcs->init(adev);
+	if (r) {
+		DRM_ERROR("Failed to init rlc BOs!\n");
+		return r;
+	}
+
+	r = gfx_v9_4_3_mec_init(adev);
+	if (r) {
+		DRM_ERROR("Failed to init MEC BOs!\n");
+		return r;
+	}
+
+	/* set up the compute queues - allocate horizontally across pipes */
+	for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) {
+		ring_id = 0;
+		for (i = 0; i < adev->gfx.mec.num_mec; ++i) {
+			for (j = 0; j < adev->gfx.mec.num_queue_per_pipe; j++) {
+				for (k = 0; k < adev->gfx.mec.num_pipe_per_mec;
+				     k++) {
+					if (!amdgpu_gfx_is_mec_queue_enabled(
+							adev, xcc_id, i, k, j))
+						continue;
+
+					r = gfx_v9_4_3_compute_ring_init(adev,
+								       ring_id,
+								       xcc_id,
+								       i, k, j);
+					if (r)
+						return r;
+
+					ring_id++;
+				}
+			}
+		}
+
+		r = amdgpu_gfx_kiq_init(adev, GFX9_MEC_HPD_SIZE, xcc_id);
+		if (r) {
+			DRM_ERROR("Failed to init KIQ BOs!\n");
+			return r;
+		}
+
+		r = amdgpu_gfx_kiq_init_ring(adev, xcc_id);
+		if (r)
+			return r;
+
+		/* create MQD for all compute queues as wel as KIQ for SRIOV case */
+		r = amdgpu_gfx_mqd_sw_init(adev,
+				sizeof(struct v9_mqd_allocation), xcc_id);
+		if (r)
+			return r;
+	}
+
+	adev->gfx.compute_supported_reset =
+		amdgpu_get_soft_full_reset_mask(&adev->gfx.compute_ring[0]);
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+		if ((adev->gfx.mec_fw_version >= 155) &&
+		    !amdgpu_sriov_vf(adev) &&
+		    !adev->debug_disable_gpu_ring_reset) {
+			adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+			adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_PIPE;
+		}
+		break;
+	case IP_VERSION(9, 5, 0):
+		if ((adev->gfx.mec_fw_version >= 21) &&
+		    !amdgpu_sriov_vf(adev) &&
+		    !adev->debug_disable_gpu_ring_reset) {
+			adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE;
+			adev->gfx.compute_supported_reset |= AMDGPU_RESET_TYPE_PER_PIPE;
+		}
+		break;
+	default:
+		break;
+	}
+	r = gfx_v9_4_3_gpu_early_init(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_gfx_ras_sw_init(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_gfx_sysfs_init(adev);
+	if (r)
+		return r;
+
+	gfx_v9_4_3_alloc_ip_dump(adev);
+
+	return 0;
+}
+
+static int gfx_v9_4_3_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	int i, num_xcc;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < adev->gfx.num_compute_rings * num_xcc; i++)
+		amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
+
+	for (i = 0; i < num_xcc; i++) {
+		amdgpu_gfx_mqd_sw_fini(adev, i);
+		amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq[i].ring);
+		amdgpu_gfx_kiq_fini(adev, i);
+	}
+
+	amdgpu_gfx_cleaner_shader_sw_fini(adev);
+
+	gfx_v9_4_3_mec_fini(adev);
+	amdgpu_bo_unref(&adev->gfx.rlc.clear_state_obj);
+	gfx_v9_4_3_free_microcode(adev);
+	amdgpu_gfx_sysfs_fini(adev);
+
+	kfree(adev->gfx.ip_dump_core);
+	kfree(adev->gfx.ip_dump_compute_queues);
+
+	return 0;
+}
+
+#define DEFAULT_SH_MEM_BASES	(0x6000)
+static void gfx_v9_4_3_xcc_init_compute_vmid(struct amdgpu_device *adev,
+					     int xcc_id)
+{
+	int i;
+	uint32_t sh_mem_config;
+	uint32_t sh_mem_bases;
+	uint32_t data;
+
+	/*
+	 * Configure apertures:
+	 * LDS:         0x60000000'00000000 - 0x60000001'00000000 (4GB)
+	 * Scratch:     0x60000001'00000000 - 0x60000002'00000000 (4GB)
+	 * GPUVM:       0x60010000'00000000 - 0x60020000'00000000 (1TB)
+	 */
+	sh_mem_bases = DEFAULT_SH_MEM_BASES | (DEFAULT_SH_MEM_BASES << 16);
+
+	sh_mem_config = SH_MEM_ADDRESS_MODE_64 |
+			SH_MEM_ALIGNMENT_MODE_UNALIGNED <<
+			SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT;
+
+	mutex_lock(&adev->srbm_mutex);
+	for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) {
+		soc15_grbm_select(adev, 0, 0, 0, i, GET_INST(GC, xcc_id));
+		/* CP and shaders */
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regSH_MEM_CONFIG, sh_mem_config);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regSH_MEM_BASES, sh_mem_bases);
+
+		/* Enable trap for each kfd vmid. */
+		data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regSPI_GDBG_PER_VMID_CNTL);
+		data = REG_SET_FIELD(data, SPI_GDBG_PER_VMID_CNTL, TRAP_EN, 1);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regSPI_GDBG_PER_VMID_CNTL, data);
+	}
+	soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
+	mutex_unlock(&adev->srbm_mutex);
+
+	/*
+	 * Initialize all compute VMIDs to have no GDS, GWS, or OA
+	 * access. These should be enabled by FW for target VMIDs.
+	 */
+	for (i = adev->vm_manager.first_kfd_vmid; i < AMDGPU_NUM_VMID; i++) {
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_VMID0_BASE, 2 * i, 0);
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_VMID0_SIZE, 2 * i, 0);
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_GWS_VMID0, i, 0);
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_OA_VMID0, i, 0);
+	}
+}
+
+static void gfx_v9_4_3_xcc_init_gds_vmid(struct amdgpu_device *adev, int xcc_id)
+{
+	int vmid;
+
+	/*
+	 * Initialize all compute and user-gfx VMIDs to have no GDS, GWS, or OA
+	 * access. Compute VMIDs should be enabled by FW for target VMIDs,
+	 * the driver can enable them for graphics. VMID0 should maintain
+	 * access so that HWS firmware can save/restore entries.
+	 */
+	for (vmid = 1; vmid < AMDGPU_NUM_VMID; vmid++) {
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_VMID0_BASE, 2 * vmid, 0);
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_VMID0_SIZE, 2 * vmid, 0);
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_GWS_VMID0, vmid, 0);
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, xcc_id), regGDS_OA_VMID0, vmid, 0);
+	}
+}
+
+/* For ASICs that needs xnack chain and MEC version supports, set SG_CONFIG1
+ * DISABLE_XNACK_CHECK_IN_RETRY_DISABLE bit and inform KFD to set xnack_chain
+ * bit in SET_RESOURCES
+ */
+static void gfx_v9_4_3_xcc_init_sq(struct amdgpu_device *adev, int xcc_id)
+{
+	uint32_t data;
+
+	if (!(adev->gmc.xnack_flags & AMDGPU_GMC_XNACK_FLAG_CHAIN))
+		return;
+
+	data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regSQ_CONFIG1);
+	data = REG_SET_FIELD(data, SQ_CONFIG1, DISABLE_XNACK_CHECK_IN_RETRY_DISABLE, 1);
+	WREG32_SOC15(GC, xcc_id, regSQ_CONFIG1, data);
+}
+
+static void gfx_v9_4_3_xcc_constants_init(struct amdgpu_device *adev,
+					  int xcc_id)
+{
+	u32 tmp;
+	int i;
+
+	/* XXX SH_MEM regs */
+	/* where to put LDS, scratch, GPUVM in FSA64 space */
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < adev->vm_manager.id_mgr[AMDGPU_GFXHUB(0)].num_ids; i++) {
+		soc15_grbm_select(adev, 0, 0, 0, i, GET_INST(GC, xcc_id));
+		/* CP and shaders */
+		if (i == 0) {
+			tmp = REG_SET_FIELD(0, SH_MEM_CONFIG, ALIGNMENT_MODE,
+					    SH_MEM_ALIGNMENT_MODE_UNALIGNED);
+			tmp = REG_SET_FIELD(tmp, SH_MEM_CONFIG, RETRY_DISABLE,
+					    !!adev->gmc.noretry);
+			WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id),
+					 regSH_MEM_CONFIG, tmp);
+			WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id),
+					 regSH_MEM_BASES, 0);
+		} else {
+			tmp = REG_SET_FIELD(0, SH_MEM_CONFIG, ALIGNMENT_MODE,
+					    SH_MEM_ALIGNMENT_MODE_UNALIGNED);
+			tmp = REG_SET_FIELD(tmp, SH_MEM_CONFIG, RETRY_DISABLE,
+					    !!adev->gmc.noretry);
+			WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id),
+					 regSH_MEM_CONFIG, tmp);
+			tmp = REG_SET_FIELD(0, SH_MEM_BASES, PRIVATE_BASE,
+					    (adev->gmc.private_aperture_start >>
+					     48));
+			tmp = REG_SET_FIELD(tmp, SH_MEM_BASES, SHARED_BASE,
+					    (adev->gmc.shared_aperture_start >>
+					     48));
+			WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id),
+					 regSH_MEM_BASES, tmp);
+		}
+	}
+	soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, 0));
+
+	mutex_unlock(&adev->srbm_mutex);
+
+	gfx_v9_4_3_xcc_init_compute_vmid(adev, xcc_id);
+	gfx_v9_4_3_xcc_init_gds_vmid(adev, xcc_id);
+	gfx_v9_4_3_xcc_init_sq(adev, xcc_id);
+}
+
+static void gfx_v9_4_3_constants_init(struct amdgpu_device *adev)
+{
+	int i, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+
+	gfx_v9_4_3_get_cu_info(adev, &adev->gfx.cu_info);
+	adev->gfx.config.db_debug2 =
+		RREG32_SOC15(GC, GET_INST(GC, 0), regDB_DEBUG2);
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	/* ToDo: GC 9.4.4 */
+	case IP_VERSION(9, 4, 3):
+		if (adev->gfx.mec_fw_version >= 184 &&
+		    (amdgpu_sriov_reg_access_sq_config(adev) ||
+		     !amdgpu_sriov_vf(adev)))
+			adev->gmc.xnack_flags |= AMDGPU_GMC_XNACK_FLAG_CHAIN;
+		break;
+	case IP_VERSION(9, 5, 0):
+		if (adev->gfx.mec_fw_version >= 23)
+			adev->gmc.xnack_flags |= AMDGPU_GMC_XNACK_FLAG_CHAIN;
+		break;
+	default:
+		break;
+	}
+
+	for (i = 0; i < num_xcc; i++)
+		gfx_v9_4_3_xcc_constants_init(adev, i);
+}
+
+static void
+gfx_v9_4_3_xcc_enable_save_restore_machine(struct amdgpu_device *adev,
+					   int xcc_id)
+{
+	WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), RLC_SRM_CNTL, SRM_ENABLE, 1);
+}
+
+static void gfx_v9_4_3_xcc_init_pg(struct amdgpu_device *adev, int xcc_id)
+{
+	/*
+	 * Rlc save restore list is workable since v2_1.
+	 */
+	gfx_v9_4_3_xcc_enable_save_restore_machine(adev, xcc_id);
+}
+
+static void gfx_v9_4_3_xcc_disable_gpa_mode(struct amdgpu_device *adev, int xcc_id)
+{
+	uint32_t data;
+
+	data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCPC_PSP_DEBUG);
+	data |= CPC_PSP_DEBUG__UTCL2IUGPAOVERRIDE_MASK;
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCPC_PSP_DEBUG, data);
+}
+
+static bool gfx_v9_4_3_is_rlc_enabled(struct amdgpu_device *adev)
+{
+	uint32_t rlc_setting;
+
+	/* if RLC is not enabled, do nothing */
+	rlc_setting = RREG32_SOC15(GC, GET_INST(GC, 0), regRLC_CNTL);
+	if (!(rlc_setting & RLC_CNTL__RLC_ENABLE_F32_MASK))
+		return false;
+
+	return true;
+}
+
+static void gfx_v9_4_3_xcc_set_safe_mode(struct amdgpu_device *adev, int xcc_id)
+{
+	uint32_t data;
+	unsigned i;
+
+	data = RLC_SAFE_MODE__CMD_MASK;
+	data |= (1 << RLC_SAFE_MODE__MESSAGE__SHIFT);
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_SAFE_MODE, data);
+
+	/* wait for RLC_SAFE_MODE */
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (!REG_GET_FIELD(RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_SAFE_MODE), RLC_SAFE_MODE, CMD))
+			break;
+		udelay(1);
+	}
+}
+
+static void gfx_v9_4_3_xcc_unset_safe_mode(struct amdgpu_device *adev,
+					   int xcc_id)
+{
+	uint32_t data;
+
+	data = RLC_SAFE_MODE__CMD_MASK;
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_SAFE_MODE, data);
+}
+
+static void gfx_v9_4_3_init_rlcg_reg_access_ctrl(struct amdgpu_device *adev)
+{
+	int xcc_id, num_xcc;
+	struct amdgpu_rlcg_reg_access_ctrl *reg_access_ctrl;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) {
+		reg_access_ctrl = &adev->gfx.rlc.reg_access_ctrl[GET_INST(GC, xcc_id)];
+		reg_access_ctrl->scratch_reg0 = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regSCRATCH_REG0);
+		reg_access_ctrl->scratch_reg1 = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regSCRATCH_REG1);
+		reg_access_ctrl->scratch_reg2 = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regSCRATCH_REG2);
+		reg_access_ctrl->scratch_reg3 = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regSCRATCH_REG3);
+		reg_access_ctrl->grbm_cntl = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regGRBM_GFX_CNTL);
+		reg_access_ctrl->grbm_idx = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regGRBM_GFX_INDEX);
+		reg_access_ctrl->spare_int = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regRLC_SPARE_INT);
+	}
+	adev->gfx.rlc.rlcg_reg_access_supported = true;
+}
+
+static int gfx_v9_4_3_rlc_init(struct amdgpu_device *adev)
+{
+	/* init spm vmid with 0xf */
+	if (adev->gfx.rlc.funcs->update_spm_vmid)
+		adev->gfx.rlc.funcs->update_spm_vmid(adev, NULL, 0xf);
+
+	return 0;
+}
+
+static void gfx_v9_4_3_xcc_wait_for_rlc_serdes(struct amdgpu_device *adev,
+					       int xcc_id)
+{
+	u32 i, j, k;
+	u32 mask;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
+		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
+			gfx_v9_4_3_xcc_select_se_sh(adev, i, j, 0xffffffff,
+						    xcc_id);
+			for (k = 0; k < adev->usec_timeout; k++) {
+				if (RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_SERDES_CU_MASTER_BUSY) == 0)
+					break;
+				udelay(1);
+			}
+			if (k == adev->usec_timeout) {
+				gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff,
+							    0xffffffff,
+							    0xffffffff, xcc_id);
+				mutex_unlock(&adev->grbm_idx_mutex);
+				DRM_INFO("Timeout wait for RLC serdes %u,%u\n",
+					 i, j);
+				return;
+			}
+		}
+	}
+	gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff,
+				    xcc_id);
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	mask = RLC_SERDES_NONCU_MASTER_BUSY__SE_MASTER_BUSY_MASK |
+		RLC_SERDES_NONCU_MASTER_BUSY__GC_MASTER_BUSY_MASK |
+		RLC_SERDES_NONCU_MASTER_BUSY__TC0_MASTER_BUSY_MASK |
+		RLC_SERDES_NONCU_MASTER_BUSY__TC1_MASTER_BUSY_MASK;
+	for (k = 0; k < adev->usec_timeout; k++) {
+		if ((RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_SERDES_NONCU_MASTER_BUSY) & mask) == 0)
+			break;
+		udelay(1);
+	}
+}
+
+static void gfx_v9_4_3_xcc_enable_gui_idle_interrupt(struct amdgpu_device *adev,
+						     bool enable, int xcc_id)
+{
+	u32 tmp;
+
+	/* These interrupts should be enabled to drive DS clock */
+
+	tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_INT_CNTL_RING0);
+
+	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE, enable ? 1 : 0);
+	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CNTX_EMPTY_INT_ENABLE, enable ? 1 : 0);
+	tmp = REG_SET_FIELD(tmp, CP_INT_CNTL_RING0, CMP_BUSY_INT_ENABLE, enable ? 1 : 0);
+
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_INT_CNTL_RING0, tmp);
+}
+
+static void gfx_v9_4_3_xcc_rlc_stop(struct amdgpu_device *adev, int xcc_id)
+{
+	WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), RLC_CNTL,
+			      RLC_ENABLE_F32, 0);
+	gfx_v9_4_3_xcc_enable_gui_idle_interrupt(adev, false, xcc_id);
+	gfx_v9_4_3_xcc_wait_for_rlc_serdes(adev, xcc_id);
+}
+
+static void gfx_v9_4_3_rlc_stop(struct amdgpu_device *adev)
+{
+	int i, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++)
+		gfx_v9_4_3_xcc_rlc_stop(adev, i);
+}
+
+static void gfx_v9_4_3_xcc_rlc_reset(struct amdgpu_device *adev, int xcc_id)
+{
+	WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), GRBM_SOFT_RESET,
+			      SOFT_RESET_RLC, 1);
+	udelay(50);
+	WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), GRBM_SOFT_RESET,
+			      SOFT_RESET_RLC, 0);
+	udelay(50);
+}
+
+static void gfx_v9_4_3_rlc_reset(struct amdgpu_device *adev)
+{
+	int i, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++)
+		gfx_v9_4_3_xcc_rlc_reset(adev, i);
+}
+
+static void gfx_v9_4_3_xcc_rlc_start(struct amdgpu_device *adev, int xcc_id)
+{
+	WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), RLC_CNTL,
+			      RLC_ENABLE_F32, 1);
+	udelay(50);
+
+	/* carrizo do enable cp interrupt after cp inited */
+	if (!(adev->flags & AMD_IS_APU)) {
+		gfx_v9_4_3_xcc_enable_gui_idle_interrupt(adev, true, xcc_id);
+		udelay(50);
+	}
+}
+
+static void gfx_v9_4_3_rlc_start(struct amdgpu_device *adev)
+{
+#ifdef AMDGPU_RLC_DEBUG_RETRY
+	u32 rlc_ucode_ver;
+#endif
+	int i, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++) {
+		gfx_v9_4_3_xcc_rlc_start(adev, i);
+#ifdef AMDGPU_RLC_DEBUG_RETRY
+		/* RLC_GPM_GENERAL_6 : RLC Ucode version */
+		rlc_ucode_ver = RREG32_SOC15(GC, GET_INST(GC, i), regRLC_GPM_GENERAL_6);
+		if (rlc_ucode_ver == 0x108) {
+			dev_info(adev->dev,
+				 "Using rlc debug ucode. regRLC_GPM_GENERAL_6 ==0x08%x / fw_ver == %i \n",
+				 rlc_ucode_ver, adev->gfx.rlc_fw_version);
+			/* RLC_GPM_TIMER_INT_3 : Timer interval in RefCLK cycles,
+			 * default is 0x9C4 to create a 100us interval */
+			WREG32_SOC15(GC, GET_INST(GC, i), regRLC_GPM_TIMER_INT_3, 0x9C4);
+			/* RLC_GPM_GENERAL_12 : Minimum gap between wptr and rptr
+			 * to disable the page fault retry interrupts, default is
+			 * 0x100 (256) */
+			WREG32_SOC15(GC, GET_INST(GC, i), regRLC_GPM_GENERAL_12, 0x100);
+		}
+#endif
+	}
+}
+
+static int gfx_v9_4_3_xcc_rlc_load_microcode(struct amdgpu_device *adev,
+					     int xcc_id)
+{
+	const struct rlc_firmware_header_v2_0 *hdr;
+	const __le32 *fw_data;
+	unsigned i, fw_size;
+
+	if (!adev->gfx.rlc_fw)
+		return -EINVAL;
+
+	hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data;
+	amdgpu_ucode_print_rlc_hdr(&hdr->header);
+
+	fw_data = (const __le32 *)(adev->gfx.rlc_fw->data +
+			   le32_to_cpu(hdr->header.ucode_array_offset_bytes));
+	fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4;
+
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_GPM_UCODE_ADDR,
+			RLCG_UCODE_LOADING_START_ADDRESS);
+	for (i = 0; i < fw_size; i++) {
+		if (amdgpu_emu_mode == 1 && i % 100 == 0) {
+			dev_info(adev->dev, "Write RLC ucode data %u DWs\n", i);
+			msleep(1);
+		}
+		WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_GPM_UCODE_DATA, le32_to_cpup(fw_data++));
+	}
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_GPM_UCODE_ADDR, adev->gfx.rlc_fw_version);
+
+	return 0;
+}
+
+static int gfx_v9_4_3_xcc_rlc_resume(struct amdgpu_device *adev, int xcc_id)
+{
+	int r;
+
+	if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
+		gfx_v9_4_3_xcc_rlc_stop(adev, xcc_id);
+		/* legacy rlc firmware loading */
+		r = gfx_v9_4_3_xcc_rlc_load_microcode(adev, xcc_id);
+		if (r)
+			return r;
+		gfx_v9_4_3_xcc_rlc_start(adev, xcc_id);
+	}
+
+	amdgpu_gfx_rlc_enter_safe_mode(adev, xcc_id);
+	/* disable CG */
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGCG_CGLS_CTRL, 0);
+	gfx_v9_4_3_xcc_init_pg(adev, xcc_id);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, xcc_id);
+
+	return 0;
+}
+
+static int gfx_v9_4_3_rlc_resume(struct amdgpu_device *adev)
+{
+	int r, i, num_xcc;
+
+	if (amdgpu_sriov_vf(adev))
+		return 0;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++) {
+		r = gfx_v9_4_3_xcc_rlc_resume(adev, i);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static void gfx_v9_4_3_update_spm_vmid(struct amdgpu_device *adev, struct amdgpu_ring *ring,
+				       unsigned vmid)
+{
+	u32 reg, pre_data, data;
+
+	reg = SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regRLC_SPM_MC_CNTL);
+	if (amdgpu_sriov_is_pp_one_vf(adev) && !amdgpu_sriov_runtime(adev))
+		pre_data = RREG32_NO_KIQ(reg);
+	else
+		pre_data = RREG32(reg);
+
+	data =	pre_data & (~RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK);
+	data |= (vmid & RLC_SPM_MC_CNTL__RLC_SPM_VMID_MASK) << RLC_SPM_MC_CNTL__RLC_SPM_VMID__SHIFT;
+
+	if (pre_data != data) {
+		if (amdgpu_sriov_is_pp_one_vf(adev) && !amdgpu_sriov_runtime(adev)) {
+			WREG32_SOC15_NO_KIQ(GC, GET_INST(GC, 0), regRLC_SPM_MC_CNTL, data);
+		} else
+			WREG32_SOC15(GC, GET_INST(GC, 0), regRLC_SPM_MC_CNTL, data);
+	}
+}
+
+static const struct soc15_reg_rlcg rlcg_access_gc_9_4_3[] = {
+	{SOC15_REG_ENTRY(GC, 0, regGRBM_GFX_INDEX)},
+	{SOC15_REG_ENTRY(GC, 0, regSQ_IND_INDEX)},
+};
+
+static bool gfx_v9_4_3_check_rlcg_range(struct amdgpu_device *adev,
+					uint32_t offset,
+					struct soc15_reg_rlcg *entries, int arr_size)
+{
+	int i, inst;
+	uint32_t reg;
+
+	if (!entries)
+		return false;
+
+	for (i = 0; i < arr_size; i++) {
+		const struct soc15_reg_rlcg *entry;
+
+		entry = &entries[i];
+		inst = adev->ip_map.logical_to_dev_inst ?
+			       adev->ip_map.logical_to_dev_inst(
+				       adev, entry->hwip, entry->instance) :
+			       entry->instance;
+		reg = adev->reg_offset[entry->hwip][inst][entry->segment] +
+		      entry->reg;
+		if (offset == reg)
+			return true;
+	}
+
+	return false;
+}
+
+static bool gfx_v9_4_3_is_rlcg_access_range(struct amdgpu_device *adev, u32 offset)
+{
+	return gfx_v9_4_3_check_rlcg_range(adev, offset,
+					(void *)rlcg_access_gc_9_4_3,
+					ARRAY_SIZE(rlcg_access_gc_9_4_3));
+}
+
+static void gfx_v9_4_3_xcc_cp_compute_enable(struct amdgpu_device *adev,
+					     bool enable, int xcc_id)
+{
+	if (enable) {
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_MEC_CNTL, 0);
+	} else {
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_MEC_CNTL,
+			(CP_MEC_CNTL__MEC_INVALIDATE_ICACHE_MASK |
+			 CP_MEC_CNTL__MEC_ME1_PIPE0_RESET_MASK |
+			 CP_MEC_CNTL__MEC_ME1_PIPE1_RESET_MASK |
+			 CP_MEC_CNTL__MEC_ME1_PIPE2_RESET_MASK |
+			 CP_MEC_CNTL__MEC_ME1_PIPE3_RESET_MASK |
+			 CP_MEC_CNTL__MEC_ME2_PIPE0_RESET_MASK |
+			 CP_MEC_CNTL__MEC_ME2_PIPE1_RESET_MASK |
+			 CP_MEC_CNTL__MEC_ME1_HALT_MASK |
+			 CP_MEC_CNTL__MEC_ME2_HALT_MASK));
+		adev->gfx.kiq[xcc_id].ring.sched.ready = false;
+	}
+	udelay(50);
+}
+
+static int gfx_v9_4_3_xcc_cp_compute_load_microcode(struct amdgpu_device *adev,
+						    int xcc_id)
+{
+	const struct gfx_firmware_header_v1_0 *mec_hdr;
+	const __le32 *fw_data;
+	unsigned i;
+	u32 tmp;
+	u32 mec_ucode_addr_offset;
+	u32 mec_ucode_data_offset;
+
+	if (!adev->gfx.mec_fw)
+		return -EINVAL;
+
+	gfx_v9_4_3_xcc_cp_compute_enable(adev, false, xcc_id);
+
+	mec_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data;
+	amdgpu_ucode_print_gfx_hdr(&mec_hdr->header);
+
+	fw_data = (const __le32 *)
+		(adev->gfx.mec_fw->data +
+		 le32_to_cpu(mec_hdr->header.ucode_array_offset_bytes));
+	tmp = 0;
+	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, VMID, 0);
+	tmp = REG_SET_FIELD(tmp, CP_CPC_IC_BASE_CNTL, CACHE_POLICY, 0);
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_CPC_IC_BASE_CNTL, tmp);
+
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_CPC_IC_BASE_LO,
+		adev->gfx.mec.mec_fw_gpu_addr & 0xFFFFF000);
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_CPC_IC_BASE_HI,
+		upper_32_bits(adev->gfx.mec.mec_fw_gpu_addr));
+
+	mec_ucode_addr_offset =
+		SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_MEC_ME1_UCODE_ADDR);
+	mec_ucode_data_offset =
+		SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_MEC_ME1_UCODE_DATA);
+
+	/* MEC1 */
+	WREG32(mec_ucode_addr_offset, mec_hdr->jt_offset);
+	for (i = 0; i < mec_hdr->jt_size; i++)
+		WREG32(mec_ucode_data_offset,
+		       le32_to_cpup(fw_data + mec_hdr->jt_offset + i));
+
+	WREG32(mec_ucode_addr_offset, adev->gfx.mec_fw_version);
+	/* Todo : Loading MEC2 firmware is only necessary if MEC2 should run different microcode than MEC1. */
+
+	return 0;
+}
+
+/* KIQ functions */
+static void gfx_v9_4_3_xcc_kiq_setting(struct amdgpu_ring *ring, int xcc_id)
+{
+	uint32_t tmp;
+	struct amdgpu_device *adev = ring->adev;
+
+	/* tell RLC which is KIQ queue */
+	tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CP_SCHEDULERS);
+	tmp &= 0xffffff00;
+	tmp |= (ring->me << 5) | (ring->pipe << 3) | (ring->queue);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regRLC_CP_SCHEDULERS, tmp | 0x80);
+}
+
+static void gfx_v9_4_3_mqd_set_priority(struct amdgpu_ring *ring, struct v9_mqd *mqd)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) {
+		if (amdgpu_gfx_is_high_priority_compute_queue(adev, ring)) {
+			mqd->cp_hqd_pipe_priority = AMDGPU_GFX_PIPE_PRIO_HIGH;
+			mqd->cp_hqd_queue_priority =
+				AMDGPU_GFX_QUEUE_PRIORITY_MAXIMUM;
+		}
+	}
+}
+
+static int gfx_v9_4_3_xcc_mqd_init(struct amdgpu_ring *ring, int xcc_id)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct v9_mqd *mqd = ring->mqd_ptr;
+	uint64_t hqd_gpu_addr, wb_gpu_addr, eop_base_addr;
+	uint32_t tmp;
+
+	mqd->header = 0xC0310800;
+	mqd->compute_pipelinestat_enable = 0x00000001;
+	mqd->compute_static_thread_mgmt_se0 = 0xffffffff;
+	mqd->compute_static_thread_mgmt_se1 = 0xffffffff;
+	mqd->compute_static_thread_mgmt_se2 = 0xffffffff;
+	mqd->compute_static_thread_mgmt_se3 = 0xffffffff;
+	mqd->compute_misc_reserved = 0x00000003;
+
+	mqd->dynamic_cu_mask_addr_lo =
+		lower_32_bits(ring->mqd_gpu_addr
+			      + offsetof(struct v9_mqd_allocation, dynamic_cu_mask));
+	mqd->dynamic_cu_mask_addr_hi =
+		upper_32_bits(ring->mqd_gpu_addr
+			      + offsetof(struct v9_mqd_allocation, dynamic_cu_mask));
+
+	eop_base_addr = ring->eop_gpu_addr >> 8;
+	mqd->cp_hqd_eop_base_addr_lo = eop_base_addr;
+	mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr);
+
+	/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
+	tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_EOP_CONTROL);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_EOP_CONTROL, EOP_SIZE,
+			(order_base_2(GFX9_MEC_HPD_SIZE / 4) - 1));
+
+	mqd->cp_hqd_eop_control = tmp;
+
+	/* enable doorbell? */
+	tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_DOORBELL_CONTROL);
+
+	if (ring->use_doorbell) {
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_OFFSET, ring->doorbell_index);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_EN, 1);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_SOURCE, 0);
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+				    DOORBELL_HIT, 0);
+		if (amdgpu_sriov_multi_vf_mode(adev))
+			tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+					    DOORBELL_MODE, 1);
+	} else {
+		tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL,
+					 DOORBELL_EN, 0);
+	}
+
+	mqd->cp_hqd_pq_doorbell_control = tmp;
+
+	/* disable the queue if it's active */
+	ring->wptr = 0;
+	mqd->cp_hqd_dequeue_request = 0;
+	mqd->cp_hqd_pq_rptr = 0;
+	mqd->cp_hqd_pq_wptr_lo = 0;
+	mqd->cp_hqd_pq_wptr_hi = 0;
+
+	/* set the pointer to the MQD */
+	mqd->cp_mqd_base_addr_lo = ring->mqd_gpu_addr & 0xfffffffc;
+	mqd->cp_mqd_base_addr_hi = upper_32_bits(ring->mqd_gpu_addr);
+
+	/* set MQD vmid to 0 */
+	tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_MQD_CONTROL);
+	tmp = REG_SET_FIELD(tmp, CP_MQD_CONTROL, VMID, 0);
+	mqd->cp_mqd_control = tmp;
+
+	/* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */
+	hqd_gpu_addr = ring->gpu_addr >> 8;
+	mqd->cp_hqd_pq_base_lo = hqd_gpu_addr;
+	mqd->cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr);
+
+	/* set up the HQD, this is similar to CP_RB0_CNTL */
+	tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_CONTROL);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, QUEUE_SIZE,
+			    (order_base_2(ring->ring_size / 4) - 1));
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE,
+			((order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1) << 8));
+#ifdef __BIG_ENDIAN
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, ENDIAN_SWAP, 1);
+#endif
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 0);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, ROQ_PQ_IB_FLIP, 0);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1);
+	mqd->cp_hqd_pq_control = tmp;
+
+	/* set the wb address whether it's enabled or not */
+	wb_gpu_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4);
+	mqd->cp_hqd_pq_rptr_report_addr_lo = wb_gpu_addr & 0xfffffffc;
+	mqd->cp_hqd_pq_rptr_report_addr_hi =
+		upper_32_bits(wb_gpu_addr) & 0xffff;
+
+	/* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */
+	wb_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4);
+	mqd->cp_hqd_pq_wptr_poll_addr_lo = wb_gpu_addr & 0xfffffffc;
+	mqd->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff;
+
+	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
+	ring->wptr = 0;
+	mqd->cp_hqd_pq_rptr = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_RPTR);
+
+	/* set the vmid for the queue */
+	mqd->cp_hqd_vmid = 0;
+
+	tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_PERSISTENT_STATE);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_PERSISTENT_STATE, PRELOAD_SIZE, 0x53);
+	mqd->cp_hqd_persistent_state = tmp;
+
+	/* set MIN_IB_AVAIL_SIZE */
+	tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_IB_CONTROL);
+	tmp = REG_SET_FIELD(tmp, CP_HQD_IB_CONTROL, MIN_IB_AVAIL_SIZE, 3);
+	mqd->cp_hqd_ib_control = tmp;
+
+	/* set static priority for a queue/ring */
+	gfx_v9_4_3_mqd_set_priority(ring, mqd);
+	mqd->cp_hqd_quantum = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_QUANTUM);
+
+	/* map_queues packet doesn't need activate the queue,
+	 * so only kiq need set this field.
+	 */
+	if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ)
+		mqd->cp_hqd_active = 1;
+
+	return 0;
+}
+
+static int gfx_v9_4_3_xcc_kiq_init_register(struct amdgpu_ring *ring,
+					    int xcc_id)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct v9_mqd *mqd = ring->mqd_ptr;
+	int j;
+
+	/* disable wptr polling */
+	WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), CP_PQ_WPTR_POLL_CNTL, EN, 0);
+
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_EOP_BASE_ADDR,
+	       mqd->cp_hqd_eop_base_addr_lo);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_EOP_BASE_ADDR_HI,
+	       mqd->cp_hqd_eop_base_addr_hi);
+
+	/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_EOP_CONTROL,
+	       mqd->cp_hqd_eop_control);
+
+	/* enable doorbell? */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_DOORBELL_CONTROL,
+	       mqd->cp_hqd_pq_doorbell_control);
+
+	/* disable the queue if it's active */
+	if (RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1) {
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_DEQUEUE_REQUEST, 1);
+		for (j = 0; j < adev->usec_timeout; j++) {
+			if (!(RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_DEQUEUE_REQUEST,
+		       mqd->cp_hqd_dequeue_request);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_RPTR,
+		       mqd->cp_hqd_pq_rptr);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_LO,
+		       mqd->cp_hqd_pq_wptr_lo);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_HI,
+		       mqd->cp_hqd_pq_wptr_hi);
+	}
+
+	/* set the pointer to the MQD */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_MQD_BASE_ADDR,
+	       mqd->cp_mqd_base_addr_lo);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_MQD_BASE_ADDR_HI,
+	       mqd->cp_mqd_base_addr_hi);
+
+	/* set MQD vmid to 0 */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_MQD_CONTROL,
+	       mqd->cp_mqd_control);
+
+	/* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_BASE,
+	       mqd->cp_hqd_pq_base_lo);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_BASE_HI,
+	       mqd->cp_hqd_pq_base_hi);
+
+	/* set up the HQD, this is similar to CP_RB0_CNTL */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_CONTROL,
+	       mqd->cp_hqd_pq_control);
+
+	/* set the wb address whether it's enabled or not */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_RPTR_REPORT_ADDR,
+				mqd->cp_hqd_pq_rptr_report_addr_lo);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_RPTR_REPORT_ADDR_HI,
+				mqd->cp_hqd_pq_rptr_report_addr_hi);
+
+	/* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_POLL_ADDR,
+	       mqd->cp_hqd_pq_wptr_poll_addr_lo);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_POLL_ADDR_HI,
+	       mqd->cp_hqd_pq_wptr_poll_addr_hi);
+
+	/* enable the doorbell if requested */
+	if (ring->use_doorbell) {
+		WREG32_SOC15(
+			GC, GET_INST(GC, xcc_id),
+			regCP_MEC_DOORBELL_RANGE_LOWER,
+			((adev->doorbell_index.kiq +
+			  xcc_id * adev->doorbell_index.xcc_doorbell_range) *
+			 2) << 2);
+		WREG32_SOC15(
+			GC, GET_INST(GC, xcc_id),
+			regCP_MEC_DOORBELL_RANGE_UPPER,
+			((adev->doorbell_index.userqueue_end +
+			  xcc_id * adev->doorbell_index.xcc_doorbell_range) *
+			 2) << 2);
+	}
+
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_DOORBELL_CONTROL,
+	       mqd->cp_hqd_pq_doorbell_control);
+
+	/* reset read and write pointers, similar to CP_RB0_WPTR/_RPTR */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_LO,
+	       mqd->cp_hqd_pq_wptr_lo);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_HI,
+	       mqd->cp_hqd_pq_wptr_hi);
+
+	/* set the vmid for the queue */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_VMID, mqd->cp_hqd_vmid);
+
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PERSISTENT_STATE,
+	       mqd->cp_hqd_persistent_state);
+
+	/* activate the queue */
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE,
+	       mqd->cp_hqd_active);
+
+	if (ring->use_doorbell)
+		WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), CP_PQ_STATUS, DOORBELL_ENABLE, 1);
+
+	return 0;
+}
+
+static int gfx_v9_4_3_xcc_q_fini_register(struct amdgpu_ring *ring,
+					    int xcc_id)
+{
+	struct amdgpu_device *adev = ring->adev;
+	int j;
+
+	/* disable the queue if it's active */
+	if (RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1) {
+
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_DEQUEUE_REQUEST, 1);
+
+		for (j = 0; j < adev->usec_timeout; j++) {
+			if (!(RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1))
+				break;
+			udelay(1);
+		}
+
+		if (j == AMDGPU_MAX_USEC_TIMEOUT) {
+			DRM_DEBUG("%s dequeue request failed.\n", ring->name);
+
+			/* Manual disable if dequeue request times out */
+			WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE, 0);
+		}
+
+		WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_DEQUEUE_REQUEST,
+		      0);
+	}
+
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_IQ_TIMER, 0);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_IB_CONTROL, 0);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PERSISTENT_STATE, CP_HQD_PERSISTENT_STATE_DEFAULT);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_DOORBELL_CONTROL, 0x40000000);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_DOORBELL_CONTROL, 0);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_RPTR, 0);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_HI, 0);
+	WREG32_SOC15_RLC(GC, GET_INST(GC, xcc_id), regCP_HQD_PQ_WPTR_LO, 0);
+
+	return 0;
+}
+
+static int gfx_v9_4_3_xcc_kiq_init_queue(struct amdgpu_ring *ring, int xcc_id)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct v9_mqd *mqd = ring->mqd_ptr;
+	struct v9_mqd *tmp_mqd;
+
+	gfx_v9_4_3_xcc_kiq_setting(ring, xcc_id);
+
+	/* GPU could be in bad state during probe, driver trigger the reset
+	 * after load the SMU, in this case , the mqd is not be initialized.
+	 * driver need to re-init the mqd.
+	 * check mqd->cp_hqd_pq_control since this value should not be 0
+	 */
+	tmp_mqd = (struct v9_mqd *)adev->gfx.kiq[xcc_id].mqd_backup;
+	if (amdgpu_in_reset(adev) && tmp_mqd->cp_hqd_pq_control) {
+		/* for GPU_RESET case , reset MQD to a clean status */
+		if (adev->gfx.kiq[xcc_id].mqd_backup)
+			memcpy(mqd, adev->gfx.kiq[xcc_id].mqd_backup, sizeof(struct v9_mqd_allocation));
+
+		/* reset ring buffer */
+		ring->wptr = 0;
+		amdgpu_ring_clear_ring(ring);
+		mutex_lock(&adev->srbm_mutex);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, GET_INST(GC, xcc_id));
+		gfx_v9_4_3_xcc_kiq_init_register(ring, xcc_id);
+		soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
+		mutex_unlock(&adev->srbm_mutex);
+	} else {
+		memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation));
+		((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
+		((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
+		mutex_lock(&adev->srbm_mutex);
+		if (amdgpu_sriov_vf(adev) && adev->in_suspend)
+			amdgpu_ring_clear_ring(ring);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, GET_INST(GC, xcc_id));
+		gfx_v9_4_3_xcc_mqd_init(ring, xcc_id);
+		gfx_v9_4_3_xcc_kiq_init_register(ring, xcc_id);
+		soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
+		mutex_unlock(&adev->srbm_mutex);
+
+		if (adev->gfx.kiq[xcc_id].mqd_backup)
+			memcpy(adev->gfx.kiq[xcc_id].mqd_backup, mqd, sizeof(struct v9_mqd_allocation));
+	}
+
+	return 0;
+}
+
+static void gfx_v9_4_3_xcc_kcq_init_queue(struct amdgpu_ring *ring, int xcc_id,
+					  bool restore)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct v9_mqd *mqd = ring->mqd_ptr;
+	int mqd_idx = ring - &adev->gfx.compute_ring[0];
+	struct v9_mqd *tmp_mqd;
+
+	/* Same as above kiq init, driver need to re-init the mqd if mqd->cp_hqd_pq_control
+	 * is not be initialized before
+	 */
+	tmp_mqd = (struct v9_mqd *)adev->gfx.mec.mqd_backup[mqd_idx];
+
+	if (!restore && (!tmp_mqd->cp_hqd_pq_control ||
+	    (!amdgpu_in_reset(adev) && !adev->in_suspend))) {
+		memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation));
+		((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF;
+		((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF;
+		mutex_lock(&adev->srbm_mutex);
+		soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0, GET_INST(GC, xcc_id));
+		gfx_v9_4_3_xcc_mqd_init(ring, xcc_id);
+		soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
+		mutex_unlock(&adev->srbm_mutex);
+
+		if (adev->gfx.mec.mqd_backup[mqd_idx])
+			memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation));
+	} else {
+		/* restore MQD to a clean status */
+		if (adev->gfx.mec.mqd_backup[mqd_idx])
+			memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation));
+		/* reset ring buffer */
+		ring->wptr = 0;
+		atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], 0);
+		amdgpu_ring_clear_ring(ring);
+	}
+}
+
+static int gfx_v9_4_3_xcc_kcq_fini_register(struct amdgpu_device *adev, int xcc_id)
+{
+	struct amdgpu_ring *ring;
+	int j;
+
+	for (j = 0; j < adev->gfx.num_compute_rings; j++) {
+		ring = &adev->gfx.compute_ring[j +  xcc_id * adev->gfx.num_compute_rings];
+		if (!amdgpu_in_reset(adev) && !adev->in_suspend) {
+			mutex_lock(&adev->srbm_mutex);
+			soc15_grbm_select(adev, ring->me,
+					ring->pipe,
+					ring->queue, 0, GET_INST(GC, xcc_id));
+			gfx_v9_4_3_xcc_q_fini_register(ring, xcc_id);
+			soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
+			mutex_unlock(&adev->srbm_mutex);
+		}
+	}
+
+	return 0;
+}
+
+static int gfx_v9_4_3_xcc_kiq_resume(struct amdgpu_device *adev, int xcc_id)
+{
+	gfx_v9_4_3_xcc_kiq_init_queue(&adev->gfx.kiq[xcc_id].ring, xcc_id);
+	return 0;
+}
+
+static int gfx_v9_4_3_xcc_kcq_resume(struct amdgpu_device *adev, int xcc_id)
+{
+	struct amdgpu_ring *ring;
+	int i;
+
+	gfx_v9_4_3_xcc_cp_compute_enable(adev, true, xcc_id);
+
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		ring = &adev->gfx.compute_ring[i + xcc_id *
+			adev->gfx.num_compute_rings];
+
+		gfx_v9_4_3_xcc_kcq_init_queue(ring, xcc_id, false);
+	}
+
+	return amdgpu_gfx_enable_kcq(adev, xcc_id);
+}
+
+static int gfx_v9_4_3_xcc_cp_resume(struct amdgpu_device *adev, int xcc_id)
+{
+	struct amdgpu_ring *ring;
+	int r, j;
+
+	gfx_v9_4_3_xcc_enable_gui_idle_interrupt(adev, false, xcc_id);
+
+	if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) {
+		gfx_v9_4_3_xcc_disable_gpa_mode(adev, xcc_id);
+
+		r = gfx_v9_4_3_xcc_cp_compute_load_microcode(adev, xcc_id);
+		if (r)
+			return r;
+	} else {
+		gfx_v9_4_3_xcc_cp_compute_enable(adev, false, xcc_id);
+	}
+
+	r = gfx_v9_4_3_xcc_kiq_resume(adev, xcc_id);
+	if (r)
+		return r;
+
+	r = gfx_v9_4_3_xcc_kcq_resume(adev, xcc_id);
+	if (r)
+		return r;
+
+	for (j = 0; j < adev->gfx.num_compute_rings; j++) {
+		ring = &adev->gfx.compute_ring
+				[j + xcc_id * adev->gfx.num_compute_rings];
+		r = amdgpu_ring_test_helper(ring);
+		if (r)
+			return r;
+	}
+
+	gfx_v9_4_3_xcc_enable_gui_idle_interrupt(adev, true, xcc_id);
+
+	return 0;
+}
+
+static int gfx_v9_4_3_cp_resume(struct amdgpu_device *adev)
+{
+	int r = 0, i, num_xcc, num_xcp, num_xcc_per_xcp;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	if (amdgpu_sriov_vf(adev)) {
+		enum amdgpu_gfx_partition mode;
+
+		mode = amdgpu_xcp_query_partition_mode(adev->xcp_mgr,
+						       AMDGPU_XCP_FL_NONE);
+		if (mode == AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE)
+			return -EINVAL;
+		num_xcc_per_xcp = gfx_v9_4_3_get_xccs_per_xcp(adev);
+		adev->gfx.num_xcc_per_xcp = num_xcc_per_xcp;
+		num_xcp = num_xcc / num_xcc_per_xcp;
+		r = amdgpu_xcp_init(adev->xcp_mgr, num_xcp, mode);
+
+	} else {
+		if (adev->in_suspend)
+			amdgpu_xcp_restore_partition_mode(adev->xcp_mgr);
+		else if (amdgpu_xcp_query_partition_mode(adev->xcp_mgr,
+						    AMDGPU_XCP_FL_NONE) ==
+		    AMDGPU_UNKNOWN_COMPUTE_PARTITION_MODE)
+			r = amdgpu_xcp_switch_partition_mode(
+				adev->xcp_mgr, amdgpu_user_partt_mode);
+	}
+	if (r)
+		return r;
+
+	for (i = 0; i < num_xcc; i++) {
+		r = gfx_v9_4_3_xcc_cp_resume(adev, i);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static void gfx_v9_4_3_xcc_fini(struct amdgpu_device *adev, int xcc_id)
+{
+	if (amdgpu_gfx_disable_kcq(adev, xcc_id))
+		DRM_ERROR("XCD %d KCQ disable failed\n", xcc_id);
+
+	if (amdgpu_sriov_vf(adev)) {
+		/* must disable polling for SRIOV when hw finished, otherwise
+		 * CPC engine may still keep fetching WB address which is already
+		 * invalid after sw finished and trigger DMAR reading error in
+		 * hypervisor side.
+		 */
+		WREG32_FIELD15_PREREG(GC, GET_INST(GC, xcc_id), CP_PQ_WPTR_POLL_CNTL, EN, 0);
+		return;
+	}
+
+	/* Use deinitialize sequence from CAIL when unbinding device
+	 * from driver, otherwise KIQ is hanging when binding back
+	 */
+	if (!amdgpu_in_reset(adev) && !adev->in_suspend) {
+		mutex_lock(&adev->srbm_mutex);
+		soc15_grbm_select(adev, adev->gfx.kiq[xcc_id].ring.me,
+				  adev->gfx.kiq[xcc_id].ring.pipe,
+				  adev->gfx.kiq[xcc_id].ring.queue, 0,
+				  GET_INST(GC, xcc_id));
+		gfx_v9_4_3_xcc_q_fini_register(&adev->gfx.kiq[xcc_id].ring,
+						 xcc_id);
+		soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
+		mutex_unlock(&adev->srbm_mutex);
+	}
+
+	gfx_v9_4_3_xcc_kcq_fini_register(adev, xcc_id);
+	gfx_v9_4_3_xcc_cp_compute_enable(adev, false, xcc_id);
+}
+
+static int gfx_v9_4_3_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	amdgpu_gfx_cleaner_shader_init(adev, adev->gfx.cleaner_shader_size,
+				       adev->gfx.cleaner_shader_ptr);
+
+	if (!amdgpu_sriov_vf(adev))
+		gfx_v9_4_3_init_golden_registers(adev);
+
+	gfx_v9_4_3_constants_init(adev);
+
+	r = adev->gfx.rlc.funcs->resume(adev);
+	if (r)
+		return r;
+
+	r = gfx_v9_4_3_cp_resume(adev);
+	if (r)
+		return r;
+
+	return r;
+}
+
+static int gfx_v9_4_3_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, num_xcc;
+
+	amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0);
+	amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0);
+	amdgpu_irq_put(adev, &adev->gfx.bad_op_irq, 0);
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++) {
+		gfx_v9_4_3_xcc_fini(adev, i);
+	}
+
+	return 0;
+}
+
+static int gfx_v9_4_3_suspend(struct amdgpu_ip_block *ip_block)
+{
+	return gfx_v9_4_3_hw_fini(ip_block);
+}
+
+static int gfx_v9_4_3_resume(struct amdgpu_ip_block *ip_block)
+{
+	return gfx_v9_4_3_hw_init(ip_block);
+}
+
+static bool gfx_v9_4_3_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++) {
+		if (REG_GET_FIELD(RREG32_SOC15(GC, GET_INST(GC, i), regGRBM_STATUS),
+					GRBM_STATUS, GUI_ACTIVE))
+			return false;
+	}
+	return true;
+}
+
+static int gfx_v9_4_3_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	unsigned i;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (gfx_v9_4_3_is_idle(ip_block))
+			return 0;
+		udelay(1);
+	}
+	return -ETIMEDOUT;
+}
+
+static int gfx_v9_4_3_soft_reset(struct amdgpu_ip_block *ip_block)
+{
+	u32 grbm_soft_reset = 0;
+	u32 tmp;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	/* GRBM_STATUS */
+	tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_STATUS);
+	if (tmp & (GRBM_STATUS__PA_BUSY_MASK | GRBM_STATUS__SC_BUSY_MASK |
+		   GRBM_STATUS__BCI_BUSY_MASK | GRBM_STATUS__SX_BUSY_MASK |
+		   GRBM_STATUS__TA_BUSY_MASK | GRBM_STATUS__VGT_BUSY_MASK |
+		   GRBM_STATUS__DB_BUSY_MASK | GRBM_STATUS__CB_BUSY_MASK |
+		   GRBM_STATUS__GDS_BUSY_MASK | GRBM_STATUS__SPI_BUSY_MASK |
+		   GRBM_STATUS__IA_BUSY_MASK | GRBM_STATUS__IA_BUSY_NO_DMA_MASK)) {
+		grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset,
+						GRBM_SOFT_RESET, SOFT_RESET_CP, 1);
+		grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset,
+						GRBM_SOFT_RESET, SOFT_RESET_GFX, 1);
+	}
+
+	if (tmp & (GRBM_STATUS__CP_BUSY_MASK | GRBM_STATUS__CP_COHERENCY_BUSY_MASK)) {
+		grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset,
+						GRBM_SOFT_RESET, SOFT_RESET_CP, 1);
+	}
+
+	/* GRBM_STATUS2 */
+	tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_STATUS2);
+	if (REG_GET_FIELD(tmp, GRBM_STATUS2, RLC_BUSY))
+		grbm_soft_reset = REG_SET_FIELD(grbm_soft_reset,
+						GRBM_SOFT_RESET, SOFT_RESET_RLC, 1);
+
+
+	if (grbm_soft_reset) {
+		/* stop the rlc */
+		adev->gfx.rlc.funcs->stop(adev);
+
+		/* Disable MEC parsing/prefetching */
+		gfx_v9_4_3_xcc_cp_compute_enable(adev, false, 0);
+
+		tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_SOFT_RESET);
+		tmp |= grbm_soft_reset;
+		dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_SOFT_RESET, tmp);
+		tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_SOFT_RESET);
+
+		udelay(50);
+
+		tmp &= ~grbm_soft_reset;
+		WREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_SOFT_RESET, tmp);
+		tmp = RREG32_SOC15(GC, GET_INST(GC, 0), regGRBM_SOFT_RESET);
+
+		/* Wait a little for things to settle down */
+		udelay(50);
+	}
+	return 0;
+}
+
+static void gfx_v9_4_3_ring_emit_gds_switch(struct amdgpu_ring *ring,
+					  uint32_t vmid,
+					  uint32_t gds_base, uint32_t gds_size,
+					  uint32_t gws_base, uint32_t gws_size,
+					  uint32_t oa_base, uint32_t oa_size)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	/* GDS Base */
+	gfx_v9_4_3_write_data_to_reg(ring, 0, false,
+				   SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regGDS_VMID0_BASE) + 2 * vmid,
+				   gds_base);
+
+	/* GDS Size */
+	gfx_v9_4_3_write_data_to_reg(ring, 0, false,
+				   SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regGDS_VMID0_SIZE) + 2 * vmid,
+				   gds_size);
+
+	/* GWS */
+	gfx_v9_4_3_write_data_to_reg(ring, 0, false,
+				   SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regGDS_GWS_VMID0) + vmid,
+				   gws_size << GDS_GWS_VMID0__SIZE__SHIFT | gws_base);
+
+	/* OA */
+	gfx_v9_4_3_write_data_to_reg(ring, 0, false,
+				   SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regGDS_OA_VMID0) + vmid,
+				   (1 << (oa_size + oa_base)) - (1 << oa_base));
+}
+
+static int gfx_v9_4_3_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
+					  AMDGPU_MAX_COMPUTE_RINGS);
+	gfx_v9_4_3_set_kiq_pm4_funcs(adev);
+	gfx_v9_4_3_set_ring_funcs(adev);
+	gfx_v9_4_3_set_irq_funcs(adev);
+	gfx_v9_4_3_set_gds_init(adev);
+	gfx_v9_4_3_set_rlc_funcs(adev);
+
+	/* init rlcg reg access ctrl */
+	gfx_v9_4_3_init_rlcg_reg_access_ctrl(adev);
+
+	return gfx_v9_4_3_init_microcode(adev);
+}
+
+static int gfx_v9_4_3_late_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_irq_get(adev, &adev->gfx.priv_reg_irq, 0);
+	if (r)
+		return r;
+
+	r = amdgpu_irq_get(adev, &adev->gfx.priv_inst_irq, 0);
+	if (r)
+		return r;
+
+	r = amdgpu_irq_get(adev, &adev->gfx.bad_op_irq, 0);
+	if (r)
+		return r;
+
+	if (adev->gfx.ras &&
+	    adev->gfx.ras->enable_watchdog_timer)
+		adev->gfx.ras->enable_watchdog_timer(adev);
+
+	return 0;
+}
+
+static void gfx_v9_4_3_xcc_update_sram_fgcg(struct amdgpu_device *adev,
+					    bool enable, int xcc_id)
+{
+	uint32_t def, data;
+
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_GFX_FGCG))
+		return;
+
+	def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id),
+				  regRLC_CGTT_MGCG_OVERRIDE);
+
+	if (enable)
+		data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_FGCG_OVERRIDE_MASK;
+	else
+		data |= RLC_CGTT_MGCG_OVERRIDE__GFXIP_FGCG_OVERRIDE_MASK;
+
+	if (def != data)
+		WREG32_SOC15(GC, GET_INST(GC, xcc_id),
+			     regRLC_CGTT_MGCG_OVERRIDE, data);
+
+}
+
+static void gfx_v9_4_3_xcc_update_repeater_fgcg(struct amdgpu_device *adev,
+						bool enable, int xcc_id)
+{
+	uint32_t def, data;
+
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_REPEATER_FGCG))
+		return;
+
+	def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id),
+				  regRLC_CGTT_MGCG_OVERRIDE);
+
+	if (enable)
+		data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_REP_FGCG_OVERRIDE_MASK;
+	else
+		data |= RLC_CGTT_MGCG_OVERRIDE__GFXIP_REP_FGCG_OVERRIDE_MASK;
+
+	if (def != data)
+		WREG32_SOC15(GC, GET_INST(GC, xcc_id),
+			     regRLC_CGTT_MGCG_OVERRIDE, data);
+}
+
+static void
+gfx_v9_4_3_xcc_update_medium_grain_clock_gating(struct amdgpu_device *adev,
+						bool enable, int xcc_id)
+{
+	uint32_t data, def;
+
+	/* It is disabled by HW by default */
+	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) {
+		/* 1 - RLC_CGTT_MGCG_OVERRIDE */
+		def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE);
+
+		data &= ~(RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK |
+			  RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK |
+			  RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK |
+			  RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGLS_OVERRIDE_MASK);
+
+		if (def != data)
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE, data);
+
+		/* MGLS is a global flag to control all MGLS in GFX */
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGLS) {
+			/* 2 - RLC memory Light sleep */
+			if (adev->cg_flags & AMD_CG_SUPPORT_GFX_RLC_LS) {
+				def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_MEM_SLP_CNTL);
+				data |= RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK;
+				if (def != data)
+					WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_MEM_SLP_CNTL, data);
+			}
+			/* 3 - CP memory Light sleep */
+			if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CP_LS) {
+				def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_MEM_SLP_CNTL);
+				data |= CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK;
+				if (def != data)
+					WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_MEM_SLP_CNTL, data);
+			}
+		}
+	} else {
+		/* 1 - MGCG_OVERRIDE */
+		def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE);
+
+		data |= (RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK |
+			 RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK |
+			 RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK |
+			 RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGLS_OVERRIDE_MASK);
+
+		if (def != data)
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE, data);
+
+		/* 2 - disable MGLS in RLC */
+		data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_MEM_SLP_CNTL);
+		if (data & RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK) {
+			data &= ~RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK;
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_MEM_SLP_CNTL, data);
+		}
+
+		/* 3 - disable MGLS in CP */
+		data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_MEM_SLP_CNTL);
+		if (data & CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK) {
+			data &= ~CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK;
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_MEM_SLP_CNTL, data);
+		}
+	}
+
+}
+
+static void
+gfx_v9_4_3_xcc_update_coarse_grain_clock_gating(struct amdgpu_device *adev,
+						bool enable, int xcc_id)
+{
+	uint32_t def, data;
+
+	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGCG)) {
+
+		def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE);
+		/* unset CGCG override */
+		data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_CGCG_OVERRIDE_MASK;
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS)
+			data &= ~RLC_CGTT_MGCG_OVERRIDE__GFXIP_CGLS_OVERRIDE_MASK;
+		else
+			data |= RLC_CGTT_MGCG_OVERRIDE__GFXIP_CGLS_OVERRIDE_MASK;
+		/* update CGCG and CGLS override bits */
+		if (def != data)
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGTT_MGCG_OVERRIDE, data);
+
+		/* CGCG Hysteresis: 400us */
+		def = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGCG_CGLS_CTRL);
+
+		data = (0x2710
+			<< RLC_CGCG_CGLS_CTRL__CGCG_GFX_IDLE_THRESHOLD__SHIFT) |
+		       RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK;
+		if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS)
+			data |= (0x000F << RLC_CGCG_CGLS_CTRL__CGLS_REP_COMPANSAT_DELAY__SHIFT) |
+				RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK;
+		if (def != data)
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGCG_CGLS_CTRL, data);
+
+		/* set IDLE_POLL_COUNT(0x33450100)*/
+		def = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_RB_WPTR_POLL_CNTL);
+		data = (0x0100 << CP_RB_WPTR_POLL_CNTL__POLL_FREQUENCY__SHIFT) |
+			(0x3345 << CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT__SHIFT);
+		if (def != data)
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_RB_WPTR_POLL_CNTL, data);
+	} else {
+		def = data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGCG_CGLS_CTRL);
+		/* reset CGCG/CGLS bits */
+		data &= ~(RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK | RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK);
+		/* disable cgcg and cgls in FSM */
+		if (def != data)
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regRLC_CGCG_CGLS_CTRL, data);
+	}
+
+}
+
+static int gfx_v9_4_3_xcc_update_gfx_clock_gating(struct amdgpu_device *adev,
+						  bool enable, int xcc_id)
+{
+	amdgpu_gfx_rlc_enter_safe_mode(adev, xcc_id);
+
+	if (enable) {
+		/* FGCG */
+		gfx_v9_4_3_xcc_update_sram_fgcg(adev, enable, xcc_id);
+		gfx_v9_4_3_xcc_update_repeater_fgcg(adev, enable, xcc_id);
+
+		/* CGCG/CGLS should be enabled after MGCG/MGLS
+		 * ===  MGCG + MGLS ===
+		 */
+		gfx_v9_4_3_xcc_update_medium_grain_clock_gating(adev, enable,
+								xcc_id);
+		/* ===  CGCG + CGLS === */
+		gfx_v9_4_3_xcc_update_coarse_grain_clock_gating(adev, enable,
+								xcc_id);
+	} else {
+		/* CGCG/CGLS should be disabled before MGCG/MGLS
+		 * ===  CGCG + CGLS ===
+		 */
+		gfx_v9_4_3_xcc_update_coarse_grain_clock_gating(adev, enable,
+								xcc_id);
+		/* ===  MGCG + MGLS === */
+		gfx_v9_4_3_xcc_update_medium_grain_clock_gating(adev, enable,
+								xcc_id);
+
+		/* FGCG */
+		gfx_v9_4_3_xcc_update_sram_fgcg(adev, enable, xcc_id);
+		gfx_v9_4_3_xcc_update_repeater_fgcg(adev, enable, xcc_id);
+	}
+
+	amdgpu_gfx_rlc_exit_safe_mode(adev, xcc_id);
+
+	return 0;
+}
+
+static const struct amdgpu_rlc_funcs gfx_v9_4_3_rlc_funcs = {
+	.is_rlc_enabled = gfx_v9_4_3_is_rlc_enabled,
+	.set_safe_mode = gfx_v9_4_3_xcc_set_safe_mode,
+	.unset_safe_mode = gfx_v9_4_3_xcc_unset_safe_mode,
+	.init = gfx_v9_4_3_rlc_init,
+	.resume = gfx_v9_4_3_rlc_resume,
+	.stop = gfx_v9_4_3_rlc_stop,
+	.reset = gfx_v9_4_3_rlc_reset,
+	.start = gfx_v9_4_3_rlc_start,
+	.update_spm_vmid = gfx_v9_4_3_update_spm_vmid,
+	.is_rlcg_access_range = gfx_v9_4_3_is_rlcg_access_range,
+};
+
+static int gfx_v9_4_3_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_powergating_state state)
+{
+	return 0;
+}
+
+static int gfx_v9_4_3_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					  enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int i, num_xcc;
+
+	if (amdgpu_sriov_vf(adev))
+		return 0;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++)
+		gfx_v9_4_3_xcc_update_gfx_clock_gating(
+			adev, state == AMD_CG_STATE_GATE, i);
+
+	return 0;
+}
+
+static void gfx_v9_4_3_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int data;
+
+	if (amdgpu_sriov_vf(adev))
+		*flags = 0;
+
+	/* AMD_CG_SUPPORT_GFX_MGCG */
+	data = RREG32_KIQ(SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regRLC_CGTT_MGCG_OVERRIDE));
+	if (!(data & RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK))
+		*flags |= AMD_CG_SUPPORT_GFX_MGCG;
+
+	/* AMD_CG_SUPPORT_GFX_CGCG */
+	data = RREG32_KIQ(SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regRLC_CGCG_CGLS_CTRL));
+	if (data & RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_GFX_CGCG;
+
+	/* AMD_CG_SUPPORT_GFX_CGLS */
+	if (data & RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_GFX_CGLS;
+
+	/* AMD_CG_SUPPORT_GFX_RLC_LS */
+	data = RREG32_KIQ(SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regRLC_MEM_SLP_CNTL));
+	if (data & RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_GFX_RLC_LS | AMD_CG_SUPPORT_GFX_MGLS;
+
+	/* AMD_CG_SUPPORT_GFX_CP_LS */
+	data = RREG32_KIQ(SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regCP_MEM_SLP_CNTL));
+	if (data & CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_GFX_CP_LS | AMD_CG_SUPPORT_GFX_MGLS;
+}
+
+static void gfx_v9_4_3_ring_emit_hdp_flush(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 ref_and_mask, reg_mem_engine;
+	const struct nbio_hdp_flush_reg *nbio_hf_reg = adev->nbio.hdp_flush_reg;
+
+	if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) {
+		switch (ring->me) {
+		case 1:
+			ref_and_mask = nbio_hf_reg->ref_and_mask_cp2 << ring->pipe;
+			break;
+		case 2:
+			ref_and_mask = nbio_hf_reg->ref_and_mask_cp6 << ring->pipe;
+			break;
+		default:
+			return;
+		}
+		reg_mem_engine = 0;
+	} else {
+		ref_and_mask = nbio_hf_reg->ref_and_mask_cp0;
+		reg_mem_engine = 1; /* pfp */
+	}
+
+	gfx_v9_4_3_wait_reg_mem(ring, reg_mem_engine, 0, 1,
+			      adev->nbio.funcs->get_hdp_flush_req_offset(adev),
+			      adev->nbio.funcs->get_hdp_flush_done_offset(adev),
+			      ref_and_mask, ref_and_mask, 0x20);
+}
+
+static void gfx_v9_4_3_ring_emit_ib_compute(struct amdgpu_ring *ring,
+					  struct amdgpu_job *job,
+					  struct amdgpu_ib *ib,
+					  uint32_t flags)
+{
+	unsigned vmid = AMDGPU_JOB_GET_VMID(job);
+	u32 control = INDIRECT_BUFFER_VALID | ib->length_dw | (vmid << 24);
+
+	/* Currently, there is a high possibility to get wave ID mismatch
+	 * between ME and GDS, leading to a hw deadlock, because ME generates
+	 * different wave IDs than the GDS expects. This situation happens
+	 * randomly when at least 5 compute pipes use GDS ordered append.
+	 * The wave IDs generated by ME are also wrong after suspend/resume.
+	 * Those are probably bugs somewhere else in the kernel driver.
+	 *
+	 * Writing GDS_COMPUTE_MAX_WAVE_ID resets wave ID counters in ME and
+	 * GDS to 0 for this ring (me/pipe).
+	 */
+	if (ib->flags & AMDGPU_IB_FLAG_RESET_GDS_MAX_WAVE_ID) {
+		amdgpu_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1));
+		amdgpu_ring_write(ring, regGDS_COMPUTE_MAX_WAVE_ID);
+		amdgpu_ring_write(ring, ring->adev->gds.gds_compute_max_wave_id);
+	}
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_INDIRECT_BUFFER, 2));
+	BUG_ON(ib->gpu_addr & 0x3); /* Dword align */
+	amdgpu_ring_write(ring,
+#ifdef __BIG_ENDIAN
+				(2 << 0) |
+#endif
+				lower_32_bits(ib->gpu_addr));
+	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr));
+	amdgpu_ring_write(ring, control);
+}
+
+static void gfx_v9_4_3_ring_emit_fence(struct amdgpu_ring *ring, u64 addr,
+				     u64 seq, unsigned flags)
+{
+	bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT;
+	bool int_sel = flags & AMDGPU_FENCE_FLAG_INT;
+	bool writeback = flags & AMDGPU_FENCE_FLAG_TC_WB_ONLY;
+
+	/* RELEASE_MEM - flush caches, send int */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_RELEASE_MEM, 6));
+	amdgpu_ring_write(ring, ((writeback ? (EOP_TC_WB_ACTION_EN |
+					       EOP_TC_NC_ACTION_EN) :
+					      (EOP_TCL1_ACTION_EN |
+					       EOP_TC_ACTION_EN |
+					       EOP_TC_WB_ACTION_EN |
+					       EOP_TC_MD_ACTION_EN)) |
+				 EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) |
+				 EVENT_INDEX(5)));
+	amdgpu_ring_write(ring, DATA_SEL(write64bit ? 2 : 1) | INT_SEL(int_sel ? 2 : 0));
+
+	/*
+	 * the address should be Qword aligned if 64bit write, Dword
+	 * aligned if only send 32bit data low (discard data high)
+	 */
+	if (write64bit)
+		BUG_ON(addr & 0x7);
+	else
+		BUG_ON(addr & 0x3);
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	amdgpu_ring_write(ring, lower_32_bits(seq));
+	amdgpu_ring_write(ring, upper_32_bits(seq));
+	amdgpu_ring_write(ring, 0);
+}
+
+static void gfx_v9_4_3_ring_emit_pipeline_sync(struct amdgpu_ring *ring)
+{
+	int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX);
+	uint32_t seq = ring->fence_drv.sync_seq;
+	uint64_t addr = ring->fence_drv.gpu_addr;
+
+	gfx_v9_4_3_wait_reg_mem(ring, usepfp, 1, 0,
+			      lower_32_bits(addr), upper_32_bits(addr),
+			      seq, 0xffffffff, 4);
+}
+
+static void gfx_v9_4_3_ring_emit_vm_flush(struct amdgpu_ring *ring,
+					unsigned vmid, uint64_t pd_addr)
+{
+	amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr);
+}
+
+static u64 gfx_v9_4_3_ring_get_rptr_compute(struct amdgpu_ring *ring)
+{
+	return ring->adev->wb.wb[ring->rptr_offs]; /* gfx9 hardware is 32bit rptr */
+}
+
+static u64 gfx_v9_4_3_ring_get_wptr_compute(struct amdgpu_ring *ring)
+{
+	u64 wptr;
+
+	/* XXX check if swapping is necessary on BE */
+	if (ring->use_doorbell)
+		wptr = atomic64_read((atomic64_t *)&ring->adev->wb.wb[ring->wptr_offs]);
+	else
+		BUG();
+	return wptr;
+}
+
+static void gfx_v9_4_3_ring_set_wptr_compute(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	/* XXX check if swapping is necessary on BE */
+	if (ring->use_doorbell) {
+		atomic64_set((atomic64_t *)&adev->wb.wb[ring->wptr_offs], ring->wptr);
+		WDOORBELL64(ring->doorbell_index, ring->wptr);
+	} else {
+		BUG(); /* only DOORBELL method supported on gfx9 now */
+	}
+}
+
+static void gfx_v9_4_3_ring_emit_fence_kiq(struct amdgpu_ring *ring, u64 addr,
+					 u64 seq, unsigned int flags)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	/* we only allocate 32bit for each seq wb address */
+	BUG_ON(flags & AMDGPU_FENCE_FLAG_64BIT);
+
+	/* write fence seq to the "addr" */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+				 WRITE_DATA_DST_SEL(5) | WR_CONFIRM));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	amdgpu_ring_write(ring, lower_32_bits(seq));
+
+	if (flags & AMDGPU_FENCE_FLAG_INT) {
+		/* set register to trigger INT */
+		amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+		amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+					 WRITE_DATA_DST_SEL(0) | WR_CONFIRM));
+		amdgpu_ring_write(ring, SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regCPC_INT_STATUS));
+		amdgpu_ring_write(ring, 0);
+		amdgpu_ring_write(ring, 0x20000000); /* src_id is 178 */
+	}
+}
+
+static void gfx_v9_4_3_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg,
+				    uint32_t reg_val_offs)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	reg = gfx_v9_4_3_normalize_xcc_reg_offset(reg);
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_COPY_DATA, 4));
+	amdgpu_ring_write(ring, 0 |	/* src: register*/
+				(5 << 8) |	/* dst: memory */
+				(1 << 20));	/* write confirm */
+	amdgpu_ring_write(ring, reg);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, lower_32_bits(adev->wb.gpu_addr +
+				reg_val_offs * 4));
+	amdgpu_ring_write(ring, upper_32_bits(adev->wb.gpu_addr +
+				reg_val_offs * 4));
+}
+
+static void gfx_v9_4_3_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg,
+				    uint32_t val)
+{
+	uint32_t cmd = 0;
+
+	reg = gfx_v9_4_3_normalize_xcc_reg_offset(reg);
+
+	switch (ring->funcs->type) {
+	case AMDGPU_RING_TYPE_GFX:
+		cmd = WRITE_DATA_ENGINE_SEL(1) | WR_CONFIRM;
+		break;
+	case AMDGPU_RING_TYPE_KIQ:
+		cmd = (1 << 16); /* no inc addr */
+		break;
+	default:
+		cmd = WR_CONFIRM;
+		break;
+	}
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, cmd);
+	amdgpu_ring_write(ring, reg);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, val);
+}
+
+static void gfx_v9_4_3_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg,
+					uint32_t val, uint32_t mask)
+{
+	gfx_v9_4_3_wait_reg_mem(ring, 0, 0, 0, reg, 0, val, mask, 0x20);
+}
+
+static void gfx_v9_4_3_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring,
+						  uint32_t reg0, uint32_t reg1,
+						  uint32_t ref, uint32_t mask)
+{
+	amdgpu_ring_emit_reg_write_reg_wait_helper(ring, reg0, reg1,
+						   ref, mask);
+}
+
+static void gfx_v9_4_3_ring_soft_recovery(struct amdgpu_ring *ring,
+					  unsigned vmid)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t value = 0;
+
+	value = REG_SET_FIELD(value, SQ_CMD, CMD, 0x03);
+	value = REG_SET_FIELD(value, SQ_CMD, MODE, 0x01);
+	value = REG_SET_FIELD(value, SQ_CMD, CHECK_VMID, 1);
+	value = REG_SET_FIELD(value, SQ_CMD, VM_ID, vmid);
+	amdgpu_gfx_rlc_enter_safe_mode(adev, ring->xcc_id);
+	WREG32_SOC15(GC, GET_INST(GC, ring->xcc_id), regSQ_CMD, value);
+	amdgpu_gfx_rlc_exit_safe_mode(adev, ring->xcc_id);
+}
+
+static void gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+	struct amdgpu_device *adev, int me, int pipe,
+	enum amdgpu_interrupt_state state, int xcc_id)
+{
+	u32 mec_int_cntl, mec_int_cntl_reg;
+
+	/*
+	 * amdgpu controls only the first MEC. That's why this function only
+	 * handles the setting of interrupts for this specific MEC. All other
+	 * pipes' interrupts are set by amdkfd.
+	 */
+
+	if (me == 1) {
+		switch (pipe) {
+		case 0:
+			mec_int_cntl_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE0_INT_CNTL);
+			break;
+		case 1:
+			mec_int_cntl_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE1_INT_CNTL);
+			break;
+		case 2:
+			mec_int_cntl_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE2_INT_CNTL);
+			break;
+		case 3:
+			mec_int_cntl_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE3_INT_CNTL);
+			break;
+		default:
+			DRM_DEBUG("invalid pipe %d\n", pipe);
+			return;
+		}
+	} else {
+		DRM_DEBUG("invalid me %d\n", me);
+		return;
+	}
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		mec_int_cntl = RREG32_XCC(mec_int_cntl_reg, xcc_id);
+		mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+					     TIME_STAMP_INT_ENABLE, 0);
+		WREG32_XCC(mec_int_cntl_reg, mec_int_cntl, xcc_id);
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		mec_int_cntl = RREG32_XCC(mec_int_cntl_reg, xcc_id);
+		mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+					     TIME_STAMP_INT_ENABLE, 1);
+		WREG32_XCC(mec_int_cntl_reg, mec_int_cntl, xcc_id);
+		break;
+	default:
+		break;
+	}
+}
+
+static u32 gfx_v9_4_3_get_cpc_int_cntl(struct amdgpu_device *adev,
+				     int xcc_id, int me, int pipe)
+{
+	/*
+	 * amdgpu controls only the first MEC. That's why this function only
+	 * handles the setting of interrupts for this specific MEC. All other
+	 * pipes' interrupts are set by amdkfd.
+	 */
+	if (me != 1)
+		return 0;
+
+	switch (pipe) {
+	case 0:
+		return SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE0_INT_CNTL);
+	case 1:
+		return SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE1_INT_CNTL);
+	case 2:
+		return SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE2_INT_CNTL);
+	case 3:
+		return SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id), regCP_ME1_PIPE3_INT_CNTL);
+	default:
+		return 0;
+	}
+}
+
+static int gfx_v9_4_3_set_priv_reg_fault_state(struct amdgpu_device *adev,
+					     struct amdgpu_irq_src *source,
+					     unsigned type,
+					     enum amdgpu_interrupt_state state)
+{
+	u32 mec_int_cntl_reg, mec_int_cntl;
+	int i, j, k, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < num_xcc; i++) {
+			WREG32_FIELD15_PREREG(GC, GET_INST(GC, i), CP_INT_CNTL_RING0,
+					      PRIV_REG_INT_ENABLE,
+					      state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+			for (j = 0; j < adev->gfx.mec.num_mec; j++) {
+				for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
+					/* MECs start at 1 */
+					mec_int_cntl_reg = gfx_v9_4_3_get_cpc_int_cntl(adev, i, j + 1, k);
+
+					if (mec_int_cntl_reg) {
+						mec_int_cntl = RREG32_XCC(mec_int_cntl_reg, i);
+						mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+									     PRIV_REG_INT_ENABLE,
+									     state == AMDGPU_IRQ_STATE_ENABLE ?
+									     1 : 0);
+						WREG32_XCC(mec_int_cntl_reg, mec_int_cntl, i);
+					}
+				}
+			}
+		}
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int gfx_v9_4_3_set_bad_op_fault_state(struct amdgpu_device *adev,
+					     struct amdgpu_irq_src *source,
+					     unsigned type,
+					     enum amdgpu_interrupt_state state)
+{
+	u32 mec_int_cntl_reg, mec_int_cntl;
+	int i, j, k, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < num_xcc; i++) {
+			WREG32_FIELD15_PREREG(GC, GET_INST(GC, i), CP_INT_CNTL_RING0,
+					      OPCODE_ERROR_INT_ENABLE,
+					      state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+			for (j = 0; j < adev->gfx.mec.num_mec; j++) {
+				for (k = 0; k < adev->gfx.mec.num_pipe_per_mec; k++) {
+					/* MECs start at 1 */
+					mec_int_cntl_reg = gfx_v9_4_3_get_cpc_int_cntl(adev, i, j + 1, k);
+
+					if (mec_int_cntl_reg) {
+						mec_int_cntl = RREG32_XCC(mec_int_cntl_reg, i);
+						mec_int_cntl = REG_SET_FIELD(mec_int_cntl, CP_ME1_PIPE0_INT_CNTL,
+									     OPCODE_ERROR_INT_ENABLE,
+									     state == AMDGPU_IRQ_STATE_ENABLE ?
+									     1 : 0);
+						WREG32_XCC(mec_int_cntl_reg, mec_int_cntl, i);
+					}
+				}
+			}
+		}
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int gfx_v9_4_3_set_priv_inst_fault_state(struct amdgpu_device *adev,
+					      struct amdgpu_irq_src *source,
+					      unsigned type,
+					      enum amdgpu_interrupt_state state)
+{
+	int i, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+	case AMDGPU_IRQ_STATE_ENABLE:
+		for (i = 0; i < num_xcc; i++)
+			WREG32_FIELD15_PREREG(GC, GET_INST(GC, i), CP_INT_CNTL_RING0,
+				PRIV_INSTR_INT_ENABLE,
+				state == AMDGPU_IRQ_STATE_ENABLE ? 1 : 0);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int gfx_v9_4_3_set_eop_interrupt_state(struct amdgpu_device *adev,
+					    struct amdgpu_irq_src *src,
+					    unsigned type,
+					    enum amdgpu_interrupt_state state)
+{
+	int i, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++) {
+		switch (type) {
+		case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP:
+			gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+				adev, 1, 0, state, i);
+			break;
+		case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE1_EOP:
+			gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+				adev, 1, 1, state, i);
+			break;
+		case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE2_EOP:
+			gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+				adev, 1, 2, state, i);
+			break;
+		case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE3_EOP:
+			gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+				adev, 1, 3, state, i);
+			break;
+		case AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE0_EOP:
+			gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+				adev, 2, 0, state, i);
+			break;
+		case AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE1_EOP:
+			gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+				adev, 2, 1, state, i);
+			break;
+		case AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE2_EOP:
+			gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+				adev, 2, 2, state, i);
+			break;
+		case AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE3_EOP:
+			gfx_v9_4_3_xcc_set_compute_eop_interrupt_state(
+				adev, 2, 3, state, i);
+			break;
+		default:
+			break;
+		}
+	}
+
+	return 0;
+}
+
+static int gfx_v9_4_3_eop_irq(struct amdgpu_device *adev,
+			    struct amdgpu_irq_src *source,
+			    struct amdgpu_iv_entry *entry)
+{
+	int i, xcc_id;
+	u8 me_id, pipe_id, queue_id;
+	struct amdgpu_ring *ring;
+
+	DRM_DEBUG("IH: CP EOP\n");
+	me_id = (entry->ring_id & 0x0c) >> 2;
+	pipe_id = (entry->ring_id & 0x03) >> 0;
+	queue_id = (entry->ring_id & 0x70) >> 4;
+
+	xcc_id = gfx_v9_4_3_ih_to_xcc_inst(adev, entry->node_id);
+
+	if (xcc_id == -EINVAL)
+		return -EINVAL;
+
+	switch (me_id) {
+	case 0:
+	case 1:
+	case 2:
+		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+			ring = &adev->gfx.compute_ring
+					[i +
+					 xcc_id * adev->gfx.num_compute_rings];
+			/* Per-queue interrupt is supported for MEC starting from VI.
+			  * The interrupt can only be enabled/disabled per pipe instead of per queue.
+			  */
+
+			if ((ring->me == me_id) && (ring->pipe == pipe_id) && (ring->queue == queue_id))
+				amdgpu_fence_process(ring);
+		}
+		break;
+	}
+	return 0;
+}
+
+static void gfx_v9_4_3_fault(struct amdgpu_device *adev,
+			   struct amdgpu_iv_entry *entry)
+{
+	u8 me_id, pipe_id, queue_id;
+	struct amdgpu_ring *ring;
+	int i, xcc_id;
+
+	me_id = (entry->ring_id & 0x0c) >> 2;
+	pipe_id = (entry->ring_id & 0x03) >> 0;
+	queue_id = (entry->ring_id & 0x70) >> 4;
+
+	xcc_id = gfx_v9_4_3_ih_to_xcc_inst(adev, entry->node_id);
+
+	if (xcc_id == -EINVAL)
+		return;
+
+	switch (me_id) {
+	case 0:
+	case 1:
+	case 2:
+		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+			ring = &adev->gfx.compute_ring
+					[i +
+					 xcc_id * adev->gfx.num_compute_rings];
+			if (ring->me == me_id && ring->pipe == pipe_id &&
+			    ring->queue == queue_id)
+				drm_sched_fault(&ring->sched);
+		}
+		break;
+	}
+}
+
+static int gfx_v9_4_3_priv_reg_irq(struct amdgpu_device *adev,
+				 struct amdgpu_irq_src *source,
+				 struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal register access in command stream\n");
+	gfx_v9_4_3_fault(adev, entry);
+	return 0;
+}
+
+static int gfx_v9_4_3_bad_op_irq(struct amdgpu_device *adev,
+				 struct amdgpu_irq_src *source,
+				 struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal opcode in command stream\n");
+	gfx_v9_4_3_fault(adev, entry);
+	return 0;
+}
+
+static int gfx_v9_4_3_priv_inst_irq(struct amdgpu_device *adev,
+				  struct amdgpu_irq_src *source,
+				  struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal instruction in command stream\n");
+	gfx_v9_4_3_fault(adev, entry);
+	return 0;
+}
+
+static void gfx_v9_4_3_emit_mem_sync(struct amdgpu_ring *ring)
+{
+	const unsigned int cp_coher_cntl =
+			PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_SH_ICACHE_ACTION_ENA(1) |
+			PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_SH_KCACHE_ACTION_ENA(1) |
+			PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_TC_ACTION_ENA(1) |
+			PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_TCL1_ACTION_ENA(1) |
+			PACKET3_ACQUIRE_MEM_CP_COHER_CNTL_TC_WB_ACTION_ENA(1);
+
+	/* ACQUIRE_MEM -make one or more surfaces valid for use by the subsequent operations */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_ACQUIRE_MEM, 5));
+	amdgpu_ring_write(ring, cp_coher_cntl); /* CP_COHER_CNTL */
+	amdgpu_ring_write(ring, 0xffffffff);  /* CP_COHER_SIZE */
+	amdgpu_ring_write(ring, 0xffffff);  /* CP_COHER_SIZE_HI */
+	amdgpu_ring_write(ring, 0); /* CP_COHER_BASE */
+	amdgpu_ring_write(ring, 0);  /* CP_COHER_BASE_HI */
+	amdgpu_ring_write(ring, 0x0000000A); /* POLL_INTERVAL */
+}
+
+static void gfx_v9_4_3_emit_wave_limit_cs(struct amdgpu_ring *ring,
+					uint32_t pipe, bool enable)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t val;
+	uint32_t wcl_cs_reg;
+
+	/* regSPI_WCL_PIPE_PERCENT_CS[0-7]_DEFAULT values are same */
+	val = enable ? 0x1 : 0x7f;
+
+	switch (pipe) {
+	case 0:
+		wcl_cs_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regSPI_WCL_PIPE_PERCENT_CS0);
+		break;
+	case 1:
+		wcl_cs_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regSPI_WCL_PIPE_PERCENT_CS1);
+		break;
+	case 2:
+		wcl_cs_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regSPI_WCL_PIPE_PERCENT_CS2);
+		break;
+	case 3:
+		wcl_cs_reg = SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regSPI_WCL_PIPE_PERCENT_CS3);
+		break;
+	default:
+		DRM_DEBUG("invalid pipe %d\n", pipe);
+		return;
+	}
+
+	amdgpu_ring_emit_wreg(ring, wcl_cs_reg, val);
+
+}
+static void gfx_v9_4_3_emit_wave_limit(struct amdgpu_ring *ring, bool enable)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t val;
+	int i;
+
+	/* regSPI_WCL_PIPE_PERCENT_GFX is 7 bit multiplier register to limit
+	 * number of gfx waves. Setting 5 bit will make sure gfx only gets
+	 * around 25% of gpu resources.
+	 */
+	val = enable ? 0x1f : 0x07ffffff;
+	amdgpu_ring_emit_wreg(ring,
+			      SOC15_REG_OFFSET(GC, GET_INST(GC, 0), regSPI_WCL_PIPE_PERCENT_GFX),
+			      val);
+
+	/* Restrict waves for normal/low priority compute queues as well
+	 * to get best QoS for high priority compute jobs.
+	 *
+	 * amdgpu controls only 1st ME(0-3 CS pipes).
+	 */
+	for (i = 0; i < adev->gfx.mec.num_pipe_per_mec; i++) {
+		if (i != ring->pipe)
+			gfx_v9_4_3_emit_wave_limit_cs(ring, i, enable);
+
+	}
+}
+
+static int gfx_v9_4_3_unmap_done(struct amdgpu_device *adev, uint32_t me,
+				uint32_t pipe, uint32_t queue,
+				uint32_t xcc_id)
+{
+	int i, r;
+	/* make sure dequeue is complete*/
+	gfx_v9_4_3_xcc_set_safe_mode(adev, xcc_id);
+	mutex_lock(&adev->srbm_mutex);
+	soc15_grbm_select(adev, me, pipe, queue, 0, GET_INST(GC, xcc_id));
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (!(RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_HQD_ACTIVE) & 1))
+			break;
+		udelay(1);
+	}
+	if (i >= adev->usec_timeout)
+		r = -ETIMEDOUT;
+	else
+		r = 0;
+	soc15_grbm_select(adev, 0, 0, 0, 0, GET_INST(GC, xcc_id));
+	mutex_unlock(&adev->srbm_mutex);
+	gfx_v9_4_3_xcc_unset_safe_mode(adev, xcc_id);
+
+	return r;
+
+}
+
+static bool gfx_v9_4_3_pipe_reset_support(struct amdgpu_device *adev)
+{
+	if (!!(adev->gfx.compute_supported_reset & AMDGPU_RESET_TYPE_PER_PIPE))
+		return true;
+	else
+		dev_warn_once(adev->dev, "Please use the latest MEC version to see whether support pipe reset\n");
+
+	return false;
+}
+
+static int gfx_v9_4_3_reset_hw_pipe(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t reset_pipe, clean_pipe;
+	int r;
+
+	if (!gfx_v9_4_3_pipe_reset_support(adev))
+		return -EINVAL;
+
+	gfx_v9_4_3_xcc_set_safe_mode(adev, ring->xcc_id);
+	mutex_lock(&adev->srbm_mutex);
+
+	reset_pipe = RREG32_SOC15(GC, GET_INST(GC, ring->xcc_id), regCP_MEC_CNTL);
+	clean_pipe = reset_pipe;
+
+	if (ring->me == 1) {
+		switch (ring->pipe) {
+		case 0:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+						   MEC_ME1_PIPE0_RESET, 1);
+			break;
+		case 1:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+						   MEC_ME1_PIPE1_RESET, 1);
+			break;
+		case 2:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+						   MEC_ME1_PIPE2_RESET, 1);
+			break;
+		case 3:
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+						   MEC_ME1_PIPE3_RESET, 1);
+			break;
+		default:
+			break;
+		}
+	} else {
+		if (ring->pipe)
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+						   MEC_ME2_PIPE1_RESET, 1);
+		else
+			reset_pipe = REG_SET_FIELD(reset_pipe, CP_MEC_CNTL,
+						   MEC_ME2_PIPE0_RESET, 1);
+	}
+
+	WREG32_SOC15(GC, GET_INST(GC, ring->xcc_id), regCP_MEC_CNTL, reset_pipe);
+	WREG32_SOC15(GC, GET_INST(GC, ring->xcc_id), regCP_MEC_CNTL, clean_pipe);
+	mutex_unlock(&adev->srbm_mutex);
+	gfx_v9_4_3_xcc_unset_safe_mode(adev, ring->xcc_id);
+
+	r = gfx_v9_4_3_unmap_done(adev, ring->me, ring->pipe, ring->queue, ring->xcc_id);
+	return r;
+}
+
+static int gfx_v9_4_3_reset_kcq(struct amdgpu_ring *ring,
+				unsigned int vmid,
+				struct amdgpu_fence *timedout_fence)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_kiq *kiq = &adev->gfx.kiq[ring->xcc_id];
+	struct amdgpu_ring *kiq_ring = &kiq->ring;
+	int reset_mode = AMDGPU_RESET_TYPE_PER_QUEUE;
+	unsigned long flags;
+	int r;
+
+	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
+		return -EINVAL;
+
+	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+
+	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size)) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+
+	kiq->pmf->kiq_unmap_queues(kiq_ring, ring, RESET_QUEUES,
+				   0, 0);
+	amdgpu_ring_commit(kiq_ring);
+
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+
+	r = amdgpu_ring_test_ring(kiq_ring);
+	if (r) {
+		dev_err(adev->dev, "kiq ring test failed after ring: %s queue reset\n",
+				ring->name);
+		goto pipe_reset;
+	}
+
+	r = gfx_v9_4_3_unmap_done(adev, ring->me, ring->pipe, ring->queue, ring->xcc_id);
+	if (r)
+		dev_err(adev->dev, "fail to wait on hqd deactive and will try pipe reset\n");
+
+pipe_reset:
+	if (r) {
+		if (!(adev->gfx.compute_supported_reset & AMDGPU_RESET_TYPE_PER_PIPE))
+			return -EOPNOTSUPP;
+		r = gfx_v9_4_3_reset_hw_pipe(ring);
+		reset_mode = AMDGPU_RESET_TYPE_PER_PIPE;
+		dev_info(adev->dev, "ring: %s pipe reset :%s\n", ring->name,
+				r ? "failed" : "successfully");
+		if (r)
+			return r;
+	}
+
+	gfx_v9_4_3_xcc_kcq_init_queue(ring, ring->xcc_id, true);
+
+	spin_lock_irqsave(&kiq->ring_lock, flags);
+	r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size);
+	if (r) {
+		spin_unlock_irqrestore(&kiq->ring_lock, flags);
+		return -ENOMEM;
+	}
+	kiq->pmf->kiq_map_queues(kiq_ring, ring);
+	amdgpu_ring_commit(kiq_ring);
+	r = amdgpu_ring_test_ring(kiq_ring);
+	spin_unlock_irqrestore(&kiq->ring_lock, flags);
+	if (r) {
+		if (reset_mode == AMDGPU_RESET_TYPE_PER_QUEUE)
+			goto pipe_reset;
+
+		dev_err(adev->dev, "fail to remap queue\n");
+		return r;
+	}
+
+	if (reset_mode == AMDGPU_RESET_TYPE_PER_QUEUE) {
+		r = amdgpu_ring_test_ring(ring);
+		if (r)
+			goto pipe_reset;
+	}
+
+
+	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
+}
+
+enum amdgpu_gfx_cp_ras_mem_id {
+	AMDGPU_GFX_CP_MEM1 = 1,
+	AMDGPU_GFX_CP_MEM2,
+	AMDGPU_GFX_CP_MEM3,
+	AMDGPU_GFX_CP_MEM4,
+	AMDGPU_GFX_CP_MEM5,
+};
+
+enum amdgpu_gfx_gcea_ras_mem_id {
+	AMDGPU_GFX_GCEA_IOWR_CMDMEM = 4,
+	AMDGPU_GFX_GCEA_IORD_CMDMEM,
+	AMDGPU_GFX_GCEA_GMIWR_CMDMEM,
+	AMDGPU_GFX_GCEA_GMIRD_CMDMEM,
+	AMDGPU_GFX_GCEA_DRAMWR_CMDMEM,
+	AMDGPU_GFX_GCEA_DRAMRD_CMDMEM,
+	AMDGPU_GFX_GCEA_MAM_DMEM0,
+	AMDGPU_GFX_GCEA_MAM_DMEM1,
+	AMDGPU_GFX_GCEA_MAM_DMEM2,
+	AMDGPU_GFX_GCEA_MAM_DMEM3,
+	AMDGPU_GFX_GCEA_MAM_AMEM0,
+	AMDGPU_GFX_GCEA_MAM_AMEM1,
+	AMDGPU_GFX_GCEA_MAM_AMEM2,
+	AMDGPU_GFX_GCEA_MAM_AMEM3,
+	AMDGPU_GFX_GCEA_MAM_AFLUSH_BUFFER,
+	AMDGPU_GFX_GCEA_WRET_TAGMEM,
+	AMDGPU_GFX_GCEA_RRET_TAGMEM,
+	AMDGPU_GFX_GCEA_IOWR_DATAMEM,
+	AMDGPU_GFX_GCEA_GMIWR_DATAMEM,
+	AMDGPU_GFX_GCEA_DRAM_DATAMEM,
+};
+
+enum amdgpu_gfx_gc_cane_ras_mem_id {
+	AMDGPU_GFX_GC_CANE_MEM0 = 0,
+};
+
+enum amdgpu_gfx_gcutcl2_ras_mem_id {
+	AMDGPU_GFX_GCUTCL2_MEM2P512X95 = 160,
+};
+
+enum amdgpu_gfx_gds_ras_mem_id {
+	AMDGPU_GFX_GDS_MEM0 = 0,
+};
+
+enum amdgpu_gfx_lds_ras_mem_id {
+	AMDGPU_GFX_LDS_BANK0 = 0,
+	AMDGPU_GFX_LDS_BANK1,
+	AMDGPU_GFX_LDS_BANK2,
+	AMDGPU_GFX_LDS_BANK3,
+	AMDGPU_GFX_LDS_BANK4,
+	AMDGPU_GFX_LDS_BANK5,
+	AMDGPU_GFX_LDS_BANK6,
+	AMDGPU_GFX_LDS_BANK7,
+	AMDGPU_GFX_LDS_BANK8,
+	AMDGPU_GFX_LDS_BANK9,
+	AMDGPU_GFX_LDS_BANK10,
+	AMDGPU_GFX_LDS_BANK11,
+	AMDGPU_GFX_LDS_BANK12,
+	AMDGPU_GFX_LDS_BANK13,
+	AMDGPU_GFX_LDS_BANK14,
+	AMDGPU_GFX_LDS_BANK15,
+	AMDGPU_GFX_LDS_BANK16,
+	AMDGPU_GFX_LDS_BANK17,
+	AMDGPU_GFX_LDS_BANK18,
+	AMDGPU_GFX_LDS_BANK19,
+	AMDGPU_GFX_LDS_BANK20,
+	AMDGPU_GFX_LDS_BANK21,
+	AMDGPU_GFX_LDS_BANK22,
+	AMDGPU_GFX_LDS_BANK23,
+	AMDGPU_GFX_LDS_BANK24,
+	AMDGPU_GFX_LDS_BANK25,
+	AMDGPU_GFX_LDS_BANK26,
+	AMDGPU_GFX_LDS_BANK27,
+	AMDGPU_GFX_LDS_BANK28,
+	AMDGPU_GFX_LDS_BANK29,
+	AMDGPU_GFX_LDS_BANK30,
+	AMDGPU_GFX_LDS_BANK31,
+	AMDGPU_GFX_LDS_SP_BUFFER_A,
+	AMDGPU_GFX_LDS_SP_BUFFER_B,
+};
+
+enum amdgpu_gfx_rlc_ras_mem_id {
+	AMDGPU_GFX_RLC_GPMF32 = 1,
+	AMDGPU_GFX_RLC_RLCVF32,
+	AMDGPU_GFX_RLC_SCRATCH,
+	AMDGPU_GFX_RLC_SRM_ARAM,
+	AMDGPU_GFX_RLC_SRM_DRAM,
+	AMDGPU_GFX_RLC_TCTAG,
+	AMDGPU_GFX_RLC_SPM_SE,
+	AMDGPU_GFX_RLC_SPM_GRBMT,
+};
+
+enum amdgpu_gfx_sp_ras_mem_id {
+	AMDGPU_GFX_SP_SIMDID0 = 0,
+};
+
+enum amdgpu_gfx_spi_ras_mem_id {
+	AMDGPU_GFX_SPI_MEM0 = 0,
+	AMDGPU_GFX_SPI_MEM1,
+	AMDGPU_GFX_SPI_MEM2,
+	AMDGPU_GFX_SPI_MEM3,
+};
+
+enum amdgpu_gfx_sqc_ras_mem_id {
+	AMDGPU_GFX_SQC_INST_CACHE_A = 100,
+	AMDGPU_GFX_SQC_INST_CACHE_B = 101,
+	AMDGPU_GFX_SQC_INST_CACHE_TAG_A = 102,
+	AMDGPU_GFX_SQC_INST_CACHE_TAG_B = 103,
+	AMDGPU_GFX_SQC_INST_CACHE_MISS_FIFO_A = 104,
+	AMDGPU_GFX_SQC_INST_CACHE_MISS_FIFO_B = 105,
+	AMDGPU_GFX_SQC_INST_CACHE_GATCL1_MISS_FIFO_A = 106,
+	AMDGPU_GFX_SQC_INST_CACHE_GATCL1_MISS_FIFO_B = 107,
+	AMDGPU_GFX_SQC_DATA_CACHE_A = 200,
+	AMDGPU_GFX_SQC_DATA_CACHE_B = 201,
+	AMDGPU_GFX_SQC_DATA_CACHE_TAG_A = 202,
+	AMDGPU_GFX_SQC_DATA_CACHE_TAG_B = 203,
+	AMDGPU_GFX_SQC_DATA_CACHE_MISS_FIFO_A = 204,
+	AMDGPU_GFX_SQC_DATA_CACHE_MISS_FIFO_B = 205,
+	AMDGPU_GFX_SQC_DATA_CACHE_HIT_FIFO_A = 206,
+	AMDGPU_GFX_SQC_DATA_CACHE_HIT_FIFO_B = 207,
+	AMDGPU_GFX_SQC_DIRTY_BIT_A = 208,
+	AMDGPU_GFX_SQC_DIRTY_BIT_B = 209,
+	AMDGPU_GFX_SQC_WRITE_DATA_BUFFER_CU0 = 210,
+	AMDGPU_GFX_SQC_WRITE_DATA_BUFFER_CU1 = 211,
+	AMDGPU_GFX_SQC_UTCL1_MISS_LFIFO_DATA_CACHE_A = 212,
+	AMDGPU_GFX_SQC_UTCL1_MISS_LFIFO_DATA_CACHE_B = 213,
+	AMDGPU_GFX_SQC_UTCL1_MISS_LFIFO_INST_CACHE = 108,
+};
+
+enum amdgpu_gfx_sq_ras_mem_id {
+	AMDGPU_GFX_SQ_SGPR_MEM0 = 0,
+	AMDGPU_GFX_SQ_SGPR_MEM1,
+	AMDGPU_GFX_SQ_SGPR_MEM2,
+	AMDGPU_GFX_SQ_SGPR_MEM3,
+};
+
+enum amdgpu_gfx_ta_ras_mem_id {
+	AMDGPU_GFX_TA_FS_AFIFO_RAM_LO = 1,
+	AMDGPU_GFX_TA_FS_AFIFO_RAM_HI,
+	AMDGPU_GFX_TA_FS_CFIFO_RAM,
+	AMDGPU_GFX_TA_FSX_LFIFO,
+	AMDGPU_GFX_TA_FS_DFIFO_RAM,
+};
+
+enum amdgpu_gfx_tcc_ras_mem_id {
+	AMDGPU_GFX_TCC_MEM1 = 1,
+};
+
+enum amdgpu_gfx_tca_ras_mem_id {
+	AMDGPU_GFX_TCA_MEM1 = 1,
+};
+
+enum amdgpu_gfx_tci_ras_mem_id {
+	AMDGPU_GFX_TCIW_MEM = 1,
+};
+
+enum amdgpu_gfx_tcp_ras_mem_id {
+	AMDGPU_GFX_TCP_LFIFO0 = 1,
+	AMDGPU_GFX_TCP_SET0BANK0_RAM,
+	AMDGPU_GFX_TCP_SET0BANK1_RAM,
+	AMDGPU_GFX_TCP_SET0BANK2_RAM,
+	AMDGPU_GFX_TCP_SET0BANK3_RAM,
+	AMDGPU_GFX_TCP_SET1BANK0_RAM,
+	AMDGPU_GFX_TCP_SET1BANK1_RAM,
+	AMDGPU_GFX_TCP_SET1BANK2_RAM,
+	AMDGPU_GFX_TCP_SET1BANK3_RAM,
+	AMDGPU_GFX_TCP_SET2BANK0_RAM,
+	AMDGPU_GFX_TCP_SET2BANK1_RAM,
+	AMDGPU_GFX_TCP_SET2BANK2_RAM,
+	AMDGPU_GFX_TCP_SET2BANK3_RAM,
+	AMDGPU_GFX_TCP_SET3BANK0_RAM,
+	AMDGPU_GFX_TCP_SET3BANK1_RAM,
+	AMDGPU_GFX_TCP_SET3BANK2_RAM,
+	AMDGPU_GFX_TCP_SET3BANK3_RAM,
+	AMDGPU_GFX_TCP_VM_FIFO,
+	AMDGPU_GFX_TCP_DB_TAGRAM0,
+	AMDGPU_GFX_TCP_DB_TAGRAM1,
+	AMDGPU_GFX_TCP_DB_TAGRAM2,
+	AMDGPU_GFX_TCP_DB_TAGRAM3,
+	AMDGPU_GFX_TCP_UTCL1_LFIFO_PROBE0,
+	AMDGPU_GFX_TCP_UTCL1_LFIFO_PROBE1,
+	AMDGPU_GFX_TCP_CMD_FIFO,
+};
+
+enum amdgpu_gfx_td_ras_mem_id {
+	AMDGPU_GFX_TD_UTD_CS_FIFO_MEM = 1,
+	AMDGPU_GFX_TD_UTD_SS_FIFO_LO_MEM,
+	AMDGPU_GFX_TD_UTD_SS_FIFO_HI_MEM,
+};
+
+enum amdgpu_gfx_tcx_ras_mem_id {
+	AMDGPU_GFX_TCX_FIFOD0 = 0,
+	AMDGPU_GFX_TCX_FIFOD1,
+	AMDGPU_GFX_TCX_FIFOD2,
+	AMDGPU_GFX_TCX_FIFOD3,
+	AMDGPU_GFX_TCX_FIFOD4,
+	AMDGPU_GFX_TCX_FIFOD5,
+	AMDGPU_GFX_TCX_FIFOD6,
+	AMDGPU_GFX_TCX_FIFOD7,
+	AMDGPU_GFX_TCX_FIFOB0,
+	AMDGPU_GFX_TCX_FIFOB1,
+	AMDGPU_GFX_TCX_FIFOB2,
+	AMDGPU_GFX_TCX_FIFOB3,
+	AMDGPU_GFX_TCX_FIFOB4,
+	AMDGPU_GFX_TCX_FIFOB5,
+	AMDGPU_GFX_TCX_FIFOB6,
+	AMDGPU_GFX_TCX_FIFOB7,
+	AMDGPU_GFX_TCX_FIFOA0,
+	AMDGPU_GFX_TCX_FIFOA1,
+	AMDGPU_GFX_TCX_FIFOA2,
+	AMDGPU_GFX_TCX_FIFOA3,
+	AMDGPU_GFX_TCX_FIFOA4,
+	AMDGPU_GFX_TCX_FIFOA5,
+	AMDGPU_GFX_TCX_FIFOA6,
+	AMDGPU_GFX_TCX_FIFOA7,
+	AMDGPU_GFX_TCX_CFIFO0,
+	AMDGPU_GFX_TCX_CFIFO1,
+	AMDGPU_GFX_TCX_CFIFO2,
+	AMDGPU_GFX_TCX_CFIFO3,
+	AMDGPU_GFX_TCX_CFIFO4,
+	AMDGPU_GFX_TCX_CFIFO5,
+	AMDGPU_GFX_TCX_CFIFO6,
+	AMDGPU_GFX_TCX_CFIFO7,
+	AMDGPU_GFX_TCX_FIFO_ACKB0,
+	AMDGPU_GFX_TCX_FIFO_ACKB1,
+	AMDGPU_GFX_TCX_FIFO_ACKB2,
+	AMDGPU_GFX_TCX_FIFO_ACKB3,
+	AMDGPU_GFX_TCX_FIFO_ACKB4,
+	AMDGPU_GFX_TCX_FIFO_ACKB5,
+	AMDGPU_GFX_TCX_FIFO_ACKB6,
+	AMDGPU_GFX_TCX_FIFO_ACKB7,
+	AMDGPU_GFX_TCX_FIFO_ACKD0,
+	AMDGPU_GFX_TCX_FIFO_ACKD1,
+	AMDGPU_GFX_TCX_FIFO_ACKD2,
+	AMDGPU_GFX_TCX_FIFO_ACKD3,
+	AMDGPU_GFX_TCX_FIFO_ACKD4,
+	AMDGPU_GFX_TCX_FIFO_ACKD5,
+	AMDGPU_GFX_TCX_FIFO_ACKD6,
+	AMDGPU_GFX_TCX_FIFO_ACKD7,
+	AMDGPU_GFX_TCX_DST_FIFOA0,
+	AMDGPU_GFX_TCX_DST_FIFOA1,
+	AMDGPU_GFX_TCX_DST_FIFOA2,
+	AMDGPU_GFX_TCX_DST_FIFOA3,
+	AMDGPU_GFX_TCX_DST_FIFOA4,
+	AMDGPU_GFX_TCX_DST_FIFOA5,
+	AMDGPU_GFX_TCX_DST_FIFOA6,
+	AMDGPU_GFX_TCX_DST_FIFOA7,
+	AMDGPU_GFX_TCX_DST_FIFOB0,
+	AMDGPU_GFX_TCX_DST_FIFOB1,
+	AMDGPU_GFX_TCX_DST_FIFOB2,
+	AMDGPU_GFX_TCX_DST_FIFOB3,
+	AMDGPU_GFX_TCX_DST_FIFOB4,
+	AMDGPU_GFX_TCX_DST_FIFOB5,
+	AMDGPU_GFX_TCX_DST_FIFOB6,
+	AMDGPU_GFX_TCX_DST_FIFOB7,
+	AMDGPU_GFX_TCX_DST_FIFOD0,
+	AMDGPU_GFX_TCX_DST_FIFOD1,
+	AMDGPU_GFX_TCX_DST_FIFOD2,
+	AMDGPU_GFX_TCX_DST_FIFOD3,
+	AMDGPU_GFX_TCX_DST_FIFOD4,
+	AMDGPU_GFX_TCX_DST_FIFOD5,
+	AMDGPU_GFX_TCX_DST_FIFOD6,
+	AMDGPU_GFX_TCX_DST_FIFOD7,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKB0,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKB1,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKB2,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKB3,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKB4,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKB5,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKB6,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKB7,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKD0,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKD1,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKD2,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKD3,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKD4,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKD5,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKD6,
+	AMDGPU_GFX_TCX_DST_FIFO_ACKD7,
+};
+
+enum amdgpu_gfx_atc_l2_ras_mem_id {
+	AMDGPU_GFX_ATC_L2_MEM0 = 0,
+};
+
+enum amdgpu_gfx_utcl2_ras_mem_id {
+	AMDGPU_GFX_UTCL2_MEM0 = 0,
+};
+
+enum amdgpu_gfx_vml2_ras_mem_id {
+	AMDGPU_GFX_VML2_MEM0 = 0,
+};
+
+enum amdgpu_gfx_vml2_walker_ras_mem_id {
+	AMDGPU_GFX_VML2_WALKER_MEM0 = 0,
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_cp_mem_list[] = {
+	{AMDGPU_GFX_CP_MEM1, "CP_MEM1"},
+	{AMDGPU_GFX_CP_MEM2, "CP_MEM2"},
+	{AMDGPU_GFX_CP_MEM3, "CP_MEM3"},
+	{AMDGPU_GFX_CP_MEM4, "CP_MEM4"},
+	{AMDGPU_GFX_CP_MEM5, "CP_MEM5"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_gcea_mem_list[] = {
+	{AMDGPU_GFX_GCEA_IOWR_CMDMEM, "GCEA_IOWR_CMDMEM"},
+	{AMDGPU_GFX_GCEA_IORD_CMDMEM, "GCEA_IORD_CMDMEM"},
+	{AMDGPU_GFX_GCEA_GMIWR_CMDMEM, "GCEA_GMIWR_CMDMEM"},
+	{AMDGPU_GFX_GCEA_GMIRD_CMDMEM, "GCEA_GMIRD_CMDMEM"},
+	{AMDGPU_GFX_GCEA_DRAMWR_CMDMEM, "GCEA_DRAMWR_CMDMEM"},
+	{AMDGPU_GFX_GCEA_DRAMRD_CMDMEM, "GCEA_DRAMRD_CMDMEM"},
+	{AMDGPU_GFX_GCEA_MAM_DMEM0, "GCEA_MAM_DMEM0"},
+	{AMDGPU_GFX_GCEA_MAM_DMEM1, "GCEA_MAM_DMEM1"},
+	{AMDGPU_GFX_GCEA_MAM_DMEM2, "GCEA_MAM_DMEM2"},
+	{AMDGPU_GFX_GCEA_MAM_DMEM3, "GCEA_MAM_DMEM3"},
+	{AMDGPU_GFX_GCEA_MAM_AMEM0, "GCEA_MAM_AMEM0"},
+	{AMDGPU_GFX_GCEA_MAM_AMEM1, "GCEA_MAM_AMEM1"},
+	{AMDGPU_GFX_GCEA_MAM_AMEM2, "GCEA_MAM_AMEM2"},
+	{AMDGPU_GFX_GCEA_MAM_AMEM3, "GCEA_MAM_AMEM3"},
+	{AMDGPU_GFX_GCEA_MAM_AFLUSH_BUFFER, "GCEA_MAM_AFLUSH_BUFFER"},
+	{AMDGPU_GFX_GCEA_WRET_TAGMEM, "GCEA_WRET_TAGMEM"},
+	{AMDGPU_GFX_GCEA_RRET_TAGMEM, "GCEA_RRET_TAGMEM"},
+	{AMDGPU_GFX_GCEA_IOWR_DATAMEM, "GCEA_IOWR_DATAMEM"},
+	{AMDGPU_GFX_GCEA_GMIWR_DATAMEM, "GCEA_GMIWR_DATAMEM"},
+	{AMDGPU_GFX_GCEA_DRAM_DATAMEM, "GCEA_DRAM_DATAMEM"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_gc_cane_mem_list[] = {
+	{AMDGPU_GFX_GC_CANE_MEM0, "GC_CANE_MEM0"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_gcutcl2_mem_list[] = {
+	{AMDGPU_GFX_GCUTCL2_MEM2P512X95, "GCUTCL2_MEM2P512X95"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_gds_mem_list[] = {
+	{AMDGPU_GFX_GDS_MEM0, "GDS_MEM"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_lds_mem_list[] = {
+	{AMDGPU_GFX_LDS_BANK0, "LDS_BANK0"},
+	{AMDGPU_GFX_LDS_BANK1, "LDS_BANK1"},
+	{AMDGPU_GFX_LDS_BANK2, "LDS_BANK2"},
+	{AMDGPU_GFX_LDS_BANK3, "LDS_BANK3"},
+	{AMDGPU_GFX_LDS_BANK4, "LDS_BANK4"},
+	{AMDGPU_GFX_LDS_BANK5, "LDS_BANK5"},
+	{AMDGPU_GFX_LDS_BANK6, "LDS_BANK6"},
+	{AMDGPU_GFX_LDS_BANK7, "LDS_BANK7"},
+	{AMDGPU_GFX_LDS_BANK8, "LDS_BANK8"},
+	{AMDGPU_GFX_LDS_BANK9, "LDS_BANK9"},
+	{AMDGPU_GFX_LDS_BANK10, "LDS_BANK10"},
+	{AMDGPU_GFX_LDS_BANK11, "LDS_BANK11"},
+	{AMDGPU_GFX_LDS_BANK12, "LDS_BANK12"},
+	{AMDGPU_GFX_LDS_BANK13, "LDS_BANK13"},
+	{AMDGPU_GFX_LDS_BANK14, "LDS_BANK14"},
+	{AMDGPU_GFX_LDS_BANK15, "LDS_BANK15"},
+	{AMDGPU_GFX_LDS_BANK16, "LDS_BANK16"},
+	{AMDGPU_GFX_LDS_BANK17, "LDS_BANK17"},
+	{AMDGPU_GFX_LDS_BANK18, "LDS_BANK18"},
+	{AMDGPU_GFX_LDS_BANK19, "LDS_BANK19"},
+	{AMDGPU_GFX_LDS_BANK20, "LDS_BANK20"},
+	{AMDGPU_GFX_LDS_BANK21, "LDS_BANK21"},
+	{AMDGPU_GFX_LDS_BANK22, "LDS_BANK22"},
+	{AMDGPU_GFX_LDS_BANK23, "LDS_BANK23"},
+	{AMDGPU_GFX_LDS_BANK24, "LDS_BANK24"},
+	{AMDGPU_GFX_LDS_BANK25, "LDS_BANK25"},
+	{AMDGPU_GFX_LDS_BANK26, "LDS_BANK26"},
+	{AMDGPU_GFX_LDS_BANK27, "LDS_BANK27"},
+	{AMDGPU_GFX_LDS_BANK28, "LDS_BANK28"},
+	{AMDGPU_GFX_LDS_BANK29, "LDS_BANK29"},
+	{AMDGPU_GFX_LDS_BANK30, "LDS_BANK30"},
+	{AMDGPU_GFX_LDS_BANK31, "LDS_BANK31"},
+	{AMDGPU_GFX_LDS_SP_BUFFER_A, "LDS_SP_BUFFER_A"},
+	{AMDGPU_GFX_LDS_SP_BUFFER_B, "LDS_SP_BUFFER_B"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_rlc_mem_list[] = {
+	{AMDGPU_GFX_RLC_GPMF32, "RLC_GPMF32"},
+	{AMDGPU_GFX_RLC_RLCVF32, "RLC_RLCVF32"},
+	{AMDGPU_GFX_RLC_SCRATCH, "RLC_SCRATCH"},
+	{AMDGPU_GFX_RLC_SRM_ARAM, "RLC_SRM_ARAM"},
+	{AMDGPU_GFX_RLC_SRM_DRAM, "RLC_SRM_DRAM"},
+	{AMDGPU_GFX_RLC_TCTAG, "RLC_TCTAG"},
+	{AMDGPU_GFX_RLC_SPM_SE, "RLC_SPM_SE"},
+	{AMDGPU_GFX_RLC_SPM_GRBMT, "RLC_SPM_GRBMT"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_sp_mem_list[] = {
+	{AMDGPU_GFX_SP_SIMDID0, "SP_SIMDID0"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_spi_mem_list[] = {
+	{AMDGPU_GFX_SPI_MEM0, "SPI_MEM0"},
+	{AMDGPU_GFX_SPI_MEM1, "SPI_MEM1"},
+	{AMDGPU_GFX_SPI_MEM2, "SPI_MEM2"},
+	{AMDGPU_GFX_SPI_MEM3, "SPI_MEM3"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_sqc_mem_list[] = {
+	{AMDGPU_GFX_SQC_INST_CACHE_A, "SQC_INST_CACHE_A"},
+	{AMDGPU_GFX_SQC_INST_CACHE_B, "SQC_INST_CACHE_B"},
+	{AMDGPU_GFX_SQC_INST_CACHE_TAG_A, "SQC_INST_CACHE_TAG_A"},
+	{AMDGPU_GFX_SQC_INST_CACHE_TAG_B, "SQC_INST_CACHE_TAG_B"},
+	{AMDGPU_GFX_SQC_INST_CACHE_MISS_FIFO_A, "SQC_INST_CACHE_MISS_FIFO_A"},
+	{AMDGPU_GFX_SQC_INST_CACHE_MISS_FIFO_B, "SQC_INST_CACHE_MISS_FIFO_B"},
+	{AMDGPU_GFX_SQC_INST_CACHE_GATCL1_MISS_FIFO_A, "SQC_INST_CACHE_GATCL1_MISS_FIFO_A"},
+	{AMDGPU_GFX_SQC_INST_CACHE_GATCL1_MISS_FIFO_B, "SQC_INST_CACHE_GATCL1_MISS_FIFO_B"},
+	{AMDGPU_GFX_SQC_DATA_CACHE_A, "SQC_DATA_CACHE_A"},
+	{AMDGPU_GFX_SQC_DATA_CACHE_B, "SQC_DATA_CACHE_B"},
+	{AMDGPU_GFX_SQC_DATA_CACHE_TAG_A, "SQC_DATA_CACHE_TAG_A"},
+	{AMDGPU_GFX_SQC_DATA_CACHE_TAG_B, "SQC_DATA_CACHE_TAG_B"},
+	{AMDGPU_GFX_SQC_DATA_CACHE_MISS_FIFO_A, "SQC_DATA_CACHE_MISS_FIFO_A"},
+	{AMDGPU_GFX_SQC_DATA_CACHE_MISS_FIFO_B, "SQC_DATA_CACHE_MISS_FIFO_B"},
+	{AMDGPU_GFX_SQC_DATA_CACHE_HIT_FIFO_A, "SQC_DATA_CACHE_HIT_FIFO_A"},
+	{AMDGPU_GFX_SQC_DATA_CACHE_HIT_FIFO_B, "SQC_DATA_CACHE_HIT_FIFO_B"},
+	{AMDGPU_GFX_SQC_DIRTY_BIT_A, "SQC_DIRTY_BIT_A"},
+	{AMDGPU_GFX_SQC_DIRTY_BIT_B, "SQC_DIRTY_BIT_B"},
+	{AMDGPU_GFX_SQC_WRITE_DATA_BUFFER_CU0, "SQC_WRITE_DATA_BUFFER_CU0"},
+	{AMDGPU_GFX_SQC_WRITE_DATA_BUFFER_CU1, "SQC_WRITE_DATA_BUFFER_CU1"},
+	{AMDGPU_GFX_SQC_UTCL1_MISS_LFIFO_DATA_CACHE_A, "SQC_UTCL1_MISS_LFIFO_DATA_CACHE_A"},
+	{AMDGPU_GFX_SQC_UTCL1_MISS_LFIFO_DATA_CACHE_B, "SQC_UTCL1_MISS_LFIFO_DATA_CACHE_B"},
+	{AMDGPU_GFX_SQC_UTCL1_MISS_LFIFO_INST_CACHE, "SQC_UTCL1_MISS_LFIFO_INST_CACHE"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_sq_mem_list[] = {
+	{AMDGPU_GFX_SQ_SGPR_MEM0, "SQ_SGPR_MEM0"},
+	{AMDGPU_GFX_SQ_SGPR_MEM1, "SQ_SGPR_MEM1"},
+	{AMDGPU_GFX_SQ_SGPR_MEM2, "SQ_SGPR_MEM2"},
+	{AMDGPU_GFX_SQ_SGPR_MEM3, "SQ_SGPR_MEM3"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_ta_mem_list[] = {
+	{AMDGPU_GFX_TA_FS_AFIFO_RAM_LO, "TA_FS_AFIFO_RAM_LO"},
+	{AMDGPU_GFX_TA_FS_AFIFO_RAM_HI, "TA_FS_AFIFO_RAM_HI"},
+	{AMDGPU_GFX_TA_FS_CFIFO_RAM, "TA_FS_CFIFO_RAM"},
+	{AMDGPU_GFX_TA_FSX_LFIFO, "TA_FSX_LFIFO"},
+	{AMDGPU_GFX_TA_FS_DFIFO_RAM, "TA_FS_DFIFO_RAM"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_tcc_mem_list[] = {
+	{AMDGPU_GFX_TCC_MEM1, "TCC_MEM1"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_tca_mem_list[] = {
+	{AMDGPU_GFX_TCA_MEM1, "TCA_MEM1"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_tci_mem_list[] = {
+	{AMDGPU_GFX_TCIW_MEM, "TCIW_MEM"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_tcp_mem_list[] = {
+	{AMDGPU_GFX_TCP_LFIFO0, "TCP_LFIFO0"},
+	{AMDGPU_GFX_TCP_SET0BANK0_RAM, "TCP_SET0BANK0_RAM"},
+	{AMDGPU_GFX_TCP_SET0BANK1_RAM, "TCP_SET0BANK1_RAM"},
+	{AMDGPU_GFX_TCP_SET0BANK2_RAM, "TCP_SET0BANK2_RAM"},
+	{AMDGPU_GFX_TCP_SET0BANK3_RAM, "TCP_SET0BANK3_RAM"},
+	{AMDGPU_GFX_TCP_SET1BANK0_RAM, "TCP_SET1BANK0_RAM"},
+	{AMDGPU_GFX_TCP_SET1BANK1_RAM, "TCP_SET1BANK1_RAM"},
+	{AMDGPU_GFX_TCP_SET1BANK2_RAM, "TCP_SET1BANK2_RAM"},
+	{AMDGPU_GFX_TCP_SET1BANK3_RAM, "TCP_SET1BANK3_RAM"},
+	{AMDGPU_GFX_TCP_SET2BANK0_RAM, "TCP_SET2BANK0_RAM"},
+	{AMDGPU_GFX_TCP_SET2BANK1_RAM, "TCP_SET2BANK1_RAM"},
+	{AMDGPU_GFX_TCP_SET2BANK2_RAM, "TCP_SET2BANK2_RAM"},
+	{AMDGPU_GFX_TCP_SET2BANK3_RAM, "TCP_SET2BANK3_RAM"},
+	{AMDGPU_GFX_TCP_SET3BANK0_RAM, "TCP_SET3BANK0_RAM"},
+	{AMDGPU_GFX_TCP_SET3BANK1_RAM, "TCP_SET3BANK1_RAM"},
+	{AMDGPU_GFX_TCP_SET3BANK2_RAM, "TCP_SET3BANK2_RAM"},
+	{AMDGPU_GFX_TCP_SET3BANK3_RAM, "TCP_SET3BANK3_RAM"},
+	{AMDGPU_GFX_TCP_VM_FIFO, "TCP_VM_FIFO"},
+	{AMDGPU_GFX_TCP_DB_TAGRAM0, "TCP_DB_TAGRAM0"},
+	{AMDGPU_GFX_TCP_DB_TAGRAM1, "TCP_DB_TAGRAM1"},
+	{AMDGPU_GFX_TCP_DB_TAGRAM2, "TCP_DB_TAGRAM2"},
+	{AMDGPU_GFX_TCP_DB_TAGRAM3, "TCP_DB_TAGRAM3"},
+	{AMDGPU_GFX_TCP_UTCL1_LFIFO_PROBE0, "TCP_UTCL1_LFIFO_PROBE0"},
+	{AMDGPU_GFX_TCP_UTCL1_LFIFO_PROBE1, "TCP_UTCL1_LFIFO_PROBE1"},
+	{AMDGPU_GFX_TCP_CMD_FIFO, "TCP_CMD_FIFO"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_td_mem_list[] = {
+	{AMDGPU_GFX_TD_UTD_CS_FIFO_MEM, "TD_UTD_CS_FIFO_MEM"},
+	{AMDGPU_GFX_TD_UTD_SS_FIFO_LO_MEM, "TD_UTD_SS_FIFO_LO_MEM"},
+	{AMDGPU_GFX_TD_UTD_SS_FIFO_HI_MEM, "TD_UTD_SS_FIFO_HI_MEM"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_tcx_mem_list[] = {
+	{AMDGPU_GFX_TCX_FIFOD0, "TCX_FIFOD0"},
+	{AMDGPU_GFX_TCX_FIFOD1, "TCX_FIFOD1"},
+	{AMDGPU_GFX_TCX_FIFOD2, "TCX_FIFOD2"},
+	{AMDGPU_GFX_TCX_FIFOD3, "TCX_FIFOD3"},
+	{AMDGPU_GFX_TCX_FIFOD4, "TCX_FIFOD4"},
+	{AMDGPU_GFX_TCX_FIFOD5, "TCX_FIFOD5"},
+	{AMDGPU_GFX_TCX_FIFOD6, "TCX_FIFOD6"},
+	{AMDGPU_GFX_TCX_FIFOD7, "TCX_FIFOD7"},
+	{AMDGPU_GFX_TCX_FIFOB0, "TCX_FIFOB0"},
+	{AMDGPU_GFX_TCX_FIFOB1, "TCX_FIFOB1"},
+	{AMDGPU_GFX_TCX_FIFOB2, "TCX_FIFOB2"},
+	{AMDGPU_GFX_TCX_FIFOB3, "TCX_FIFOB3"},
+	{AMDGPU_GFX_TCX_FIFOB4, "TCX_FIFOB4"},
+	{AMDGPU_GFX_TCX_FIFOB5, "TCX_FIFOB5"},
+	{AMDGPU_GFX_TCX_FIFOB6, "TCX_FIFOB6"},
+	{AMDGPU_GFX_TCX_FIFOB7, "TCX_FIFOB7"},
+	{AMDGPU_GFX_TCX_FIFOA0, "TCX_FIFOA0"},
+	{AMDGPU_GFX_TCX_FIFOA1, "TCX_FIFOA1"},
+	{AMDGPU_GFX_TCX_FIFOA2, "TCX_FIFOA2"},
+	{AMDGPU_GFX_TCX_FIFOA3, "TCX_FIFOA3"},
+	{AMDGPU_GFX_TCX_FIFOA4, "TCX_FIFOA4"},
+	{AMDGPU_GFX_TCX_FIFOA5, "TCX_FIFOA5"},
+	{AMDGPU_GFX_TCX_FIFOA6, "TCX_FIFOA6"},
+	{AMDGPU_GFX_TCX_FIFOA7, "TCX_FIFOA7"},
+	{AMDGPU_GFX_TCX_CFIFO0, "TCX_CFIFO0"},
+	{AMDGPU_GFX_TCX_CFIFO1, "TCX_CFIFO1"},
+	{AMDGPU_GFX_TCX_CFIFO2, "TCX_CFIFO2"},
+	{AMDGPU_GFX_TCX_CFIFO3, "TCX_CFIFO3"},
+	{AMDGPU_GFX_TCX_CFIFO4, "TCX_CFIFO4"},
+	{AMDGPU_GFX_TCX_CFIFO5, "TCX_CFIFO5"},
+	{AMDGPU_GFX_TCX_CFIFO6, "TCX_CFIFO6"},
+	{AMDGPU_GFX_TCX_CFIFO7, "TCX_CFIFO7"},
+	{AMDGPU_GFX_TCX_FIFO_ACKB0, "TCX_FIFO_ACKB0"},
+	{AMDGPU_GFX_TCX_FIFO_ACKB1, "TCX_FIFO_ACKB1"},
+	{AMDGPU_GFX_TCX_FIFO_ACKB2, "TCX_FIFO_ACKB2"},
+	{AMDGPU_GFX_TCX_FIFO_ACKB3, "TCX_FIFO_ACKB3"},
+	{AMDGPU_GFX_TCX_FIFO_ACKB4, "TCX_FIFO_ACKB4"},
+	{AMDGPU_GFX_TCX_FIFO_ACKB5, "TCX_FIFO_ACKB5"},
+	{AMDGPU_GFX_TCX_FIFO_ACKB6, "TCX_FIFO_ACKB6"},
+	{AMDGPU_GFX_TCX_FIFO_ACKB7, "TCX_FIFO_ACKB7"},
+	{AMDGPU_GFX_TCX_FIFO_ACKD0, "TCX_FIFO_ACKD0"},
+	{AMDGPU_GFX_TCX_FIFO_ACKD1, "TCX_FIFO_ACKD1"},
+	{AMDGPU_GFX_TCX_FIFO_ACKD2, "TCX_FIFO_ACKD2"},
+	{AMDGPU_GFX_TCX_FIFO_ACKD3, "TCX_FIFO_ACKD3"},
+	{AMDGPU_GFX_TCX_FIFO_ACKD4, "TCX_FIFO_ACKD4"},
+	{AMDGPU_GFX_TCX_FIFO_ACKD5, "TCX_FIFO_ACKD5"},
+	{AMDGPU_GFX_TCX_FIFO_ACKD6, "TCX_FIFO_ACKD6"},
+	{AMDGPU_GFX_TCX_FIFO_ACKD7, "TCX_FIFO_ACKD7"},
+	{AMDGPU_GFX_TCX_DST_FIFOA0, "TCX_DST_FIFOA0"},
+	{AMDGPU_GFX_TCX_DST_FIFOA1, "TCX_DST_FIFOA1"},
+	{AMDGPU_GFX_TCX_DST_FIFOA2, "TCX_DST_FIFOA2"},
+	{AMDGPU_GFX_TCX_DST_FIFOA3, "TCX_DST_FIFOA3"},
+	{AMDGPU_GFX_TCX_DST_FIFOA4, "TCX_DST_FIFOA4"},
+	{AMDGPU_GFX_TCX_DST_FIFOA5, "TCX_DST_FIFOA5"},
+	{AMDGPU_GFX_TCX_DST_FIFOA6, "TCX_DST_FIFOA6"},
+	{AMDGPU_GFX_TCX_DST_FIFOA7, "TCX_DST_FIFOA7"},
+	{AMDGPU_GFX_TCX_DST_FIFOB0, "TCX_DST_FIFOB0"},
+	{AMDGPU_GFX_TCX_DST_FIFOB1, "TCX_DST_FIFOB1"},
+	{AMDGPU_GFX_TCX_DST_FIFOB2, "TCX_DST_FIFOB2"},
+	{AMDGPU_GFX_TCX_DST_FIFOB3, "TCX_DST_FIFOB3"},
+	{AMDGPU_GFX_TCX_DST_FIFOB4, "TCX_DST_FIFOB4"},
+	{AMDGPU_GFX_TCX_DST_FIFOB5, "TCX_DST_FIFOB5"},
+	{AMDGPU_GFX_TCX_DST_FIFOB6, "TCX_DST_FIFOB6"},
+	{AMDGPU_GFX_TCX_DST_FIFOB7, "TCX_DST_FIFOB7"},
+	{AMDGPU_GFX_TCX_DST_FIFOD0, "TCX_DST_FIFOD0"},
+	{AMDGPU_GFX_TCX_DST_FIFOD1, "TCX_DST_FIFOD1"},
+	{AMDGPU_GFX_TCX_DST_FIFOD2, "TCX_DST_FIFOD2"},
+	{AMDGPU_GFX_TCX_DST_FIFOD3, "TCX_DST_FIFOD3"},
+	{AMDGPU_GFX_TCX_DST_FIFOD4, "TCX_DST_FIFOD4"},
+	{AMDGPU_GFX_TCX_DST_FIFOD5, "TCX_DST_FIFOD5"},
+	{AMDGPU_GFX_TCX_DST_FIFOD6, "TCX_DST_FIFOD6"},
+	{AMDGPU_GFX_TCX_DST_FIFOD7, "TCX_DST_FIFOD7"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKB0, "TCX_DST_FIFO_ACKB0"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKB1, "TCX_DST_FIFO_ACKB1"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKB2, "TCX_DST_FIFO_ACKB2"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKB3, "TCX_DST_FIFO_ACKB3"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKB4, "TCX_DST_FIFO_ACKB4"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKB5, "TCX_DST_FIFO_ACKB5"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKB6, "TCX_DST_FIFO_ACKB6"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKB7, "TCX_DST_FIFO_ACKB7"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKD0, "TCX_DST_FIFO_ACKD0"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKD1, "TCX_DST_FIFO_ACKD1"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKD2, "TCX_DST_FIFO_ACKD2"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKD3, "TCX_DST_FIFO_ACKD3"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKD4, "TCX_DST_FIFO_ACKD4"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKD5, "TCX_DST_FIFO_ACKD5"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKD6, "TCX_DST_FIFO_ACKD6"},
+	{AMDGPU_GFX_TCX_DST_FIFO_ACKD7, "TCX_DST_FIFO_ACKD7"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_atc_l2_mem_list[] = {
+	{AMDGPU_GFX_ATC_L2_MEM, "ATC_L2_MEM"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_utcl2_mem_list[] = {
+	{AMDGPU_GFX_UTCL2_MEM, "UTCL2_MEM"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_vml2_mem_list[] = {
+	{AMDGPU_GFX_VML2_MEM, "VML2_MEM"},
+};
+
+static const struct amdgpu_ras_memory_id_entry gfx_v9_4_3_ras_vml2_walker_mem_list[] = {
+	{AMDGPU_GFX_VML2_WALKER_MEM, "VML2_WALKER_MEM"},
+};
+
+static const struct amdgpu_gfx_ras_mem_id_entry gfx_v9_4_3_ras_mem_list_array[AMDGPU_GFX_MEM_TYPE_NUM] = {
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_cp_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_gcea_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_gc_cane_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_gcutcl2_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_gds_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_lds_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_rlc_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_sp_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_spi_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_sqc_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_sq_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_ta_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_tcc_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_tca_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_tci_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_tcp_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_td_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_tcx_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_atc_l2_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_utcl2_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_vml2_mem_list)
+	AMDGPU_GFX_MEMID_ENT(gfx_v9_4_3_ras_vml2_walker_mem_list)
+};
+
+static const struct amdgpu_gfx_ras_reg_entry gfx_v9_4_3_ce_reg_list[] = {
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regRLC_CE_ERR_STATUS_LOW, regRLC_CE_ERR_STATUS_HIGH),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "RLC"},
+	    AMDGPU_GFX_RLC_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regCPC_CE_ERR_STATUS_LO, regCPC_CE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "CPC"},
+	    AMDGPU_GFX_CP_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regCPF_CE_ERR_STATUS_LO, regCPF_CE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "CPF"},
+	    AMDGPU_GFX_CP_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regCPG_CE_ERR_STATUS_LO, regCPG_CE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "CPG"},
+	    AMDGPU_GFX_CP_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regGDS_CE_ERR_STATUS_LO, regGDS_CE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "GDS"},
+	    AMDGPU_GFX_GDS_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regGC_CANE_CE_ERR_STATUS_LO, regGC_CANE_CE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "CANE"},
+	    AMDGPU_GFX_GC_CANE_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSPI_CE_ERR_STATUS_LO, regSPI_CE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SPI"},
+	    AMDGPU_GFX_SPI_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSP0_CE_ERR_STATUS_LO, regSP0_CE_ERR_STATUS_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SP0"},
+	    AMDGPU_GFX_SP_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSP1_CE_ERR_STATUS_LO, regSP1_CE_ERR_STATUS_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SP1"},
+	    AMDGPU_GFX_SP_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSQ_CE_ERR_STATUS_LO, regSQ_CE_ERR_STATUS_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SQ"},
+	    AMDGPU_GFX_SQ_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSQC_CE_EDC_LO, regSQC_CE_EDC_HI),
+	    5, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SQC"},
+	    AMDGPU_GFX_SQC_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCX_CE_ERR_STATUS_LO, regTCX_CE_ERR_STATUS_HI),
+	    2, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCX"},
+	    AMDGPU_GFX_TCX_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCC_CE_ERR_STATUS_LO, regTCC_CE_ERR_STATUS_HI),
+	    16, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCC"},
+	    AMDGPU_GFX_TCC_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTA_CE_EDC_LO, regTA_CE_EDC_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TA"},
+	    AMDGPU_GFX_TA_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCI_CE_EDC_LO_REG, regTCI_CE_EDC_HI_REG),
+	    27, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCI"},
+	    AMDGPU_GFX_TCI_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCP_CE_EDC_LO_REG, regTCP_CE_EDC_HI_REG),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCP"},
+	    AMDGPU_GFX_TCP_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTD_CE_EDC_LO, regTD_CE_EDC_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TD"},
+	    AMDGPU_GFX_TD_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regGCEA_CE_ERR_STATUS_LO, regGCEA_CE_ERR_STATUS_HI),
+	    16, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "GCEA"},
+	    AMDGPU_GFX_GCEA_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regLDS_CE_ERR_STATUS_LO, regLDS_CE_ERR_STATUS_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "LDS"},
+	    AMDGPU_GFX_LDS_MEM, 4},
+};
+
+static const struct amdgpu_gfx_ras_reg_entry gfx_v9_4_3_ue_reg_list[] = {
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regRLC_UE_ERR_STATUS_LOW, regRLC_UE_ERR_STATUS_HIGH),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "RLC"},
+	    AMDGPU_GFX_RLC_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regCPC_UE_ERR_STATUS_LO, regCPC_UE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "CPC"},
+	    AMDGPU_GFX_CP_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regCPF_UE_ERR_STATUS_LO, regCPF_UE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "CPF"},
+	    AMDGPU_GFX_CP_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regCPG_UE_ERR_STATUS_LO, regCPG_UE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "CPG"},
+	    AMDGPU_GFX_CP_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regGDS_UE_ERR_STATUS_LO, regGDS_UE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "GDS"},
+	    AMDGPU_GFX_GDS_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regGC_CANE_UE_ERR_STATUS_LO, regGC_CANE_UE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "CANE"},
+	    AMDGPU_GFX_GC_CANE_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSPI_UE_ERR_STATUS_LO, regSPI_UE_ERR_STATUS_HI),
+	    1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SPI"},
+	    AMDGPU_GFX_SPI_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSP0_UE_ERR_STATUS_LO, regSP0_UE_ERR_STATUS_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SP0"},
+	    AMDGPU_GFX_SP_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSP1_UE_ERR_STATUS_LO, regSP1_UE_ERR_STATUS_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SP1"},
+	    AMDGPU_GFX_SP_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSQ_UE_ERR_STATUS_LO, regSQ_UE_ERR_STATUS_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SQ"},
+	    AMDGPU_GFX_SQ_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regSQC_UE_EDC_LO, regSQC_UE_EDC_HI),
+	    5, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SQC"},
+	    AMDGPU_GFX_SQC_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCX_UE_ERR_STATUS_LO, regTCX_UE_ERR_STATUS_HI),
+	    2, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCX"},
+	    AMDGPU_GFX_TCX_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCC_UE_ERR_STATUS_LO, regTCC_UE_ERR_STATUS_HI),
+	    16, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCC"},
+	    AMDGPU_GFX_TCC_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTA_UE_EDC_LO, regTA_UE_EDC_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TA"},
+	    AMDGPU_GFX_TA_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCI_UE_EDC_LO_REG, regTCI_UE_EDC_HI_REG),
+	    27, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCI"},
+	    AMDGPU_GFX_TCI_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCP_UE_EDC_LO_REG, regTCP_UE_EDC_HI_REG),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCP"},
+	    AMDGPU_GFX_TCP_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTD_UE_EDC_LO, regTD_UE_EDC_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TD"},
+	    AMDGPU_GFX_TD_MEM, 4},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regTCA_UE_ERR_STATUS_LO, regTCA_UE_ERR_STATUS_HI),
+	    2, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "TCA"},
+	    AMDGPU_GFX_TCA_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regGCEA_UE_ERR_STATUS_LO, regGCEA_UE_ERR_STATUS_HI),
+	    16, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "GCEA"},
+	    AMDGPU_GFX_GCEA_MEM, 1},
+	{{AMDGPU_RAS_REG_ENTRY(GC, 0, regLDS_UE_ERR_STATUS_LO, regLDS_UE_ERR_STATUS_HI),
+	    10, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "LDS"},
+	    AMDGPU_GFX_LDS_MEM, 4},
+};
+
+static void gfx_v9_4_3_inst_query_ras_err_count(struct amdgpu_device *adev,
+					void *ras_error_status, int xcc_id)
+{
+	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
+	unsigned long ce_count = 0, ue_count = 0;
+	uint32_t i, j, k;
+
+	/* NOTE: convert xcc_id to physical XCD ID (XCD0 or XCD1) */
+	struct amdgpu_smuio_mcm_config_info mcm_info = {
+		.socket_id = adev->smuio.funcs->get_socket_id(adev),
+		.die_id = xcc_id & 0x01 ? 1 : 0,
+	};
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	for (i = 0; i < ARRAY_SIZE(gfx_v9_4_3_ce_reg_list); i++) {
+		for (j = 0; j < gfx_v9_4_3_ce_reg_list[i].se_num; j++) {
+			for (k = 0; k < gfx_v9_4_3_ce_reg_list[i].reg_entry.reg_inst; k++) {
+				/* no need to select if instance number is 1 */
+				if (gfx_v9_4_3_ce_reg_list[i].se_num > 1 ||
+				    gfx_v9_4_3_ce_reg_list[i].reg_entry.reg_inst > 1)
+					gfx_v9_4_3_xcc_select_se_sh(adev, j, 0, k, xcc_id);
+
+				amdgpu_ras_inst_query_ras_error_count(adev,
+					&(gfx_v9_4_3_ce_reg_list[i].reg_entry),
+					1,
+					gfx_v9_4_3_ras_mem_list_array[gfx_v9_4_3_ce_reg_list[i].mem_id_type].mem_id_ent,
+					gfx_v9_4_3_ras_mem_list_array[gfx_v9_4_3_ce_reg_list[i].mem_id_type].size,
+					GET_INST(GC, xcc_id),
+					AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE,
+					&ce_count);
+
+				amdgpu_ras_inst_query_ras_error_count(adev,
+					&(gfx_v9_4_3_ue_reg_list[i].reg_entry),
+					1,
+					gfx_v9_4_3_ras_mem_list_array[gfx_v9_4_3_ue_reg_list[i].mem_id_type].mem_id_ent,
+					gfx_v9_4_3_ras_mem_list_array[gfx_v9_4_3_ue_reg_list[i].mem_id_type].size,
+					GET_INST(GC, xcc_id),
+					AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
+					&ue_count);
+			}
+		}
+	}
+
+	/* handle extra register entries of UE */
+	for (; i < ARRAY_SIZE(gfx_v9_4_3_ue_reg_list); i++) {
+		for (j = 0; j < gfx_v9_4_3_ue_reg_list[i].se_num; j++) {
+			for (k = 0; k < gfx_v9_4_3_ue_reg_list[i].reg_entry.reg_inst; k++) {
+				/* no need to select if instance number is 1 */
+				if (gfx_v9_4_3_ue_reg_list[i].se_num > 1 ||
+					gfx_v9_4_3_ue_reg_list[i].reg_entry.reg_inst > 1)
+					gfx_v9_4_3_xcc_select_se_sh(adev, j, 0, k, xcc_id);
+
+				amdgpu_ras_inst_query_ras_error_count(adev,
+					&(gfx_v9_4_3_ue_reg_list[i].reg_entry),
+					1,
+					gfx_v9_4_3_ras_mem_list_array[gfx_v9_4_3_ue_reg_list[i].mem_id_type].mem_id_ent,
+					gfx_v9_4_3_ras_mem_list_array[gfx_v9_4_3_ue_reg_list[i].mem_id_type].size,
+					GET_INST(GC, xcc_id),
+					AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
+					&ue_count);
+			}
+		}
+	}
+
+	gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff,
+			xcc_id);
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	/* the caller should make sure initialize value of
+	 * err_data->ue_count and err_data->ce_count
+	 */
+	amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, ue_count);
+	amdgpu_ras_error_statistic_ce_count(err_data, &mcm_info, ce_count);
+}
+
+static void gfx_v9_4_3_inst_reset_ras_err_count(struct amdgpu_device *adev,
+					void *ras_error_status, int xcc_id)
+{
+	uint32_t i, j, k;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	for (i = 0; i < ARRAY_SIZE(gfx_v9_4_3_ce_reg_list); i++) {
+		for (j = 0; j < gfx_v9_4_3_ce_reg_list[i].se_num; j++) {
+			for (k = 0; k < gfx_v9_4_3_ce_reg_list[i].reg_entry.reg_inst; k++) {
+				/* no need to select if instance number is 1 */
+				if (gfx_v9_4_3_ce_reg_list[i].se_num > 1 ||
+				    gfx_v9_4_3_ce_reg_list[i].reg_entry.reg_inst > 1)
+					gfx_v9_4_3_xcc_select_se_sh(adev, j, 0, k, xcc_id);
+
+				amdgpu_ras_inst_reset_ras_error_count(adev,
+					&(gfx_v9_4_3_ce_reg_list[i].reg_entry),
+					1,
+					GET_INST(GC, xcc_id));
+
+				amdgpu_ras_inst_reset_ras_error_count(adev,
+					&(gfx_v9_4_3_ue_reg_list[i].reg_entry),
+					1,
+					GET_INST(GC, xcc_id));
+			}
+		}
+	}
+
+	/* handle extra register entries of UE */
+	for (; i < ARRAY_SIZE(gfx_v9_4_3_ue_reg_list); i++) {
+		for (j = 0; j < gfx_v9_4_3_ue_reg_list[i].se_num; j++) {
+			for (k = 0; k < gfx_v9_4_3_ue_reg_list[i].reg_entry.reg_inst; k++) {
+				/* no need to select if instance number is 1 */
+				if (gfx_v9_4_3_ue_reg_list[i].se_num > 1 ||
+					gfx_v9_4_3_ue_reg_list[i].reg_entry.reg_inst > 1)
+					gfx_v9_4_3_xcc_select_se_sh(adev, j, 0, k, xcc_id);
+
+				amdgpu_ras_inst_reset_ras_error_count(adev,
+					&(gfx_v9_4_3_ue_reg_list[i].reg_entry),
+					1,
+					GET_INST(GC, xcc_id));
+			}
+		}
+	}
+
+	gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff,
+			xcc_id);
+	mutex_unlock(&adev->grbm_idx_mutex);
+}
+
+static void gfx_v9_4_3_inst_enable_watchdog_timer(struct amdgpu_device *adev,
+					void *ras_error_status, int xcc_id)
+{
+	uint32_t i;
+	uint32_t data;
+
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	data = RREG32_SOC15(GC, GET_INST(GC, 0), regSQ_TIMEOUT_CONFIG);
+	data = REG_SET_FIELD(data, SQ_TIMEOUT_CONFIG, TIMEOUT_FATAL_DISABLE,
+			     amdgpu_watchdog_timer.timeout_fatal_disable ? 1 : 0);
+
+	if (amdgpu_watchdog_timer.timeout_fatal_disable &&
+	    (amdgpu_watchdog_timer.period < 1 ||
+	     amdgpu_watchdog_timer.period > 0x23)) {
+		dev_warn(adev->dev, "Watchdog period range is 1 to 0x23\n");
+		amdgpu_watchdog_timer.period = 0x23;
+	}
+	data = REG_SET_FIELD(data, SQ_TIMEOUT_CONFIG, PERIOD_SEL,
+			     amdgpu_watchdog_timer.period);
+
+	mutex_lock(&adev->grbm_idx_mutex);
+	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
+		gfx_v9_4_3_xcc_select_se_sh(adev, i, 0xffffffff, 0xffffffff, xcc_id);
+		WREG32_SOC15(GC, GET_INST(GC, xcc_id), regSQ_TIMEOUT_CONFIG, data);
+	}
+	gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff,
+			xcc_id);
+	mutex_unlock(&adev->grbm_idx_mutex);
+}
+
+static void gfx_v9_4_3_query_ras_error_count(struct amdgpu_device *adev,
+					void *ras_error_status)
+{
+	amdgpu_gfx_ras_error_func(adev, ras_error_status,
+			gfx_v9_4_3_inst_query_ras_err_count);
+}
+
+static void gfx_v9_4_3_reset_ras_error_count(struct amdgpu_device *adev)
+{
+	amdgpu_gfx_ras_error_func(adev, NULL, gfx_v9_4_3_inst_reset_ras_err_count);
+}
+
+static void gfx_v9_4_3_enable_watchdog_timer(struct amdgpu_device *adev)
+{
+	amdgpu_gfx_ras_error_func(adev, NULL, gfx_v9_4_3_inst_enable_watchdog_timer);
+}
+
+static void gfx_v9_4_3_ring_insert_nop(struct amdgpu_ring *ring, uint32_t num_nop)
+{
+	/* Header itself is a NOP packet */
+	if (num_nop == 1) {
+		amdgpu_ring_write(ring, ring->funcs->nop);
+		return;
+	}
+
+	/* Max HW optimization till 0x3ffe, followed by remaining one NOP at a time*/
+	amdgpu_ring_write(ring, PACKET3(PACKET3_NOP, min(num_nop - 2, 0x3ffe)));
+
+	/* Header is at index 0, followed by num_nops - 1 NOP packet's */
+	amdgpu_ring_insert_nop(ring, num_nop - 1);
+}
+
+static void gfx_v9_4_3_ip_print(struct amdgpu_ip_block *ip_block, struct drm_printer *p)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k;
+	uint32_t xcc_id, xcc_offset, inst_offset;
+	uint32_t num_xcc, reg, num_inst;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9_4_3);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	drm_printf(p, "Number of Instances:%d\n", num_xcc);
+	for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) {
+		xcc_offset = xcc_id * reg_count;
+		drm_printf(p, "\nInstance id:%d\n", xcc_id);
+		for (i = 0; i < reg_count; i++)
+			drm_printf(p, "%-50s \t 0x%08x\n",
+				   gc_reg_list_9_4_3[i].reg_name,
+				   adev->gfx.ip_dump_core[xcc_offset + i]);
+	}
+
+	/* print compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	num_inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec *
+		adev->gfx.mec.num_queue_per_pipe;
+
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_9_4_3);
+	drm_printf(p, "\nnum_xcc: %d num_mec: %d num_pipe: %d num_queue: %d\n",
+		   num_xcc,
+		   adev->gfx.mec.num_mec,
+		   adev->gfx.mec.num_pipe_per_mec,
+		   adev->gfx.mec.num_queue_per_pipe);
+
+	for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) {
+		xcc_offset = xcc_id * reg_count * num_inst;
+		inst_offset = 0;
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+					drm_printf(p,
+						   "\nxcc:%d mec:%d, pipe:%d, queue:%d\n",
+						    xcc_id, i, j, k);
+					for (reg = 0; reg < reg_count; reg++) {
+						if (i && gc_cp_reg_list_9_4_3[reg].reg_offset ==
+						    regCP_MEC_ME1_HEADER_DUMP)
+							drm_printf(p,
+								   "%-50s \t 0x%08x\n",
+								   "regCP_MEC_ME2_HEADER_DUMP",
+								   adev->gfx.ip_dump_compute_queues
+								   [xcc_offset + inst_offset +
+								    reg]);
+						else
+							drm_printf(p,
+								   "%-50s \t 0x%08x\n",
+								   gc_cp_reg_list_9_4_3[reg].reg_name,
+								   adev->gfx.ip_dump_compute_queues
+								   [xcc_offset + inst_offset +
+								    reg]);
+					}
+					inst_offset += reg_count;
+				}
+			}
+		}
+	}
+}
+
+static void gfx_v9_4_3_ip_dump(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	uint32_t i, j, k;
+	uint32_t num_xcc, reg, num_inst;
+	uint32_t xcc_id, xcc_offset, inst_offset;
+	uint32_t reg_count = ARRAY_SIZE(gc_reg_list_9_4_3);
+
+	if (!adev->gfx.ip_dump_core)
+		return;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+
+	for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) {
+		xcc_offset = xcc_id * reg_count;
+		for (i = 0; i < reg_count; i++)
+			adev->gfx.ip_dump_core[xcc_offset + i] =
+				RREG32(SOC15_REG_ENTRY_OFFSET_INST(gc_reg_list_9_4_3[i],
+								   GET_INST(GC, xcc_id)));
+	}
+
+	/* dump compute queue registers for all instances */
+	if (!adev->gfx.ip_dump_compute_queues)
+		return;
+
+	num_inst = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe_per_mec *
+		adev->gfx.mec.num_queue_per_pipe;
+	reg_count = ARRAY_SIZE(gc_cp_reg_list_9_4_3);
+	mutex_lock(&adev->srbm_mutex);
+	for (xcc_id = 0; xcc_id < num_xcc; xcc_id++) {
+		xcc_offset = xcc_id * reg_count * num_inst;
+		inst_offset = 0;
+		for (i = 0; i < adev->gfx.mec.num_mec; i++) {
+			for (j = 0; j < adev->gfx.mec.num_pipe_per_mec; j++) {
+				for (k = 0; k < adev->gfx.mec.num_queue_per_pipe; k++) {
+					/* ME0 is for GFX so start from 1 for CP */
+					soc15_grbm_select(adev, 1 + i, j, k, 0,
+							  GET_INST(GC, xcc_id));
+
+					for (reg = 0; reg < reg_count; reg++) {
+						if (i && gc_cp_reg_list_9_4_3[reg].reg_offset ==
+						    regCP_MEC_ME1_HEADER_DUMP)
+							adev->gfx.ip_dump_compute_queues
+								[xcc_offset +
+								 inst_offset + reg] =
+								RREG32(SOC15_REG_OFFSET(GC, GET_INST(GC, xcc_id),
+											regCP_MEC_ME2_HEADER_DUMP));
+						else
+							adev->gfx.ip_dump_compute_queues
+								[xcc_offset +
+								 inst_offset + reg] =
+								RREG32(SOC15_REG_ENTRY_OFFSET_INST(
+									       gc_cp_reg_list_9_4_3[reg],
+									       GET_INST(GC, xcc_id)));
+					}
+					inst_offset += reg_count;
+				}
+			}
+		}
+	}
+	soc15_grbm_select(adev, 0, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+
+static void gfx_v9_4_3_ring_emit_cleaner_shader(struct amdgpu_ring *ring)
+{
+	/* Emit the cleaner shader */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_RUN_CLEANER_SHADER, 0));
+	amdgpu_ring_write(ring, 0);  /* RESERVED field, programmed to zero */
+}
+
+static const struct amd_ip_funcs gfx_v9_4_3_ip_funcs = {
+	.name = "gfx_v9_4_3",
+	.early_init = gfx_v9_4_3_early_init,
+	.late_init = gfx_v9_4_3_late_init,
+	.sw_init = gfx_v9_4_3_sw_init,
+	.sw_fini = gfx_v9_4_3_sw_fini,
+	.hw_init = gfx_v9_4_3_hw_init,
+	.hw_fini = gfx_v9_4_3_hw_fini,
+	.suspend = gfx_v9_4_3_suspend,
+	.resume = gfx_v9_4_3_resume,
+	.is_idle = gfx_v9_4_3_is_idle,
+	.wait_for_idle = gfx_v9_4_3_wait_for_idle,
+	.soft_reset = gfx_v9_4_3_soft_reset,
+	.set_clockgating_state = gfx_v9_4_3_set_clockgating_state,
+	.set_powergating_state = gfx_v9_4_3_set_powergating_state,
+	.get_clockgating_state = gfx_v9_4_3_get_clockgating_state,
+	.dump_ip_state = gfx_v9_4_3_ip_dump,
+	.print_ip_state = gfx_v9_4_3_ip_print,
+};
+
+static const struct amdgpu_ring_funcs gfx_v9_4_3_ring_funcs_compute = {
+	.type = AMDGPU_RING_TYPE_COMPUTE,
+	.align_mask = 0xff,
+	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
+	.support_64bit_ptrs = true,
+	.get_rptr = gfx_v9_4_3_ring_get_rptr_compute,
+	.get_wptr = gfx_v9_4_3_ring_get_wptr_compute,
+	.set_wptr = gfx_v9_4_3_ring_set_wptr_compute,
+	.emit_frame_size =
+		20 + /* gfx_v9_4_3_ring_emit_gds_switch */
+		7 + /* gfx_v9_4_3_ring_emit_hdp_flush */
+		5 + /* hdp invalidate */
+		7 + /* gfx_v9_4_3_ring_emit_pipeline_sync */
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
+		2 + /* gfx_v9_4_3_ring_emit_vm_flush */
+		8 + 8 + 8 + /* gfx_v9_4_3_ring_emit_fence x3 for user fence, vm fence */
+		7 + /* gfx_v9_4_3_emit_mem_sync */
+		5 + /* gfx_v9_4_3_emit_wave_limit for updating regSPI_WCL_PIPE_PERCENT_GFX register */
+		15 + /* for updating 3 regSPI_WCL_PIPE_PERCENT_CS registers */
+		2, /* gfx_v9_4_3_ring_emit_cleaner_shader */
+	.emit_ib_size =	7, /* gfx_v9_4_3_ring_emit_ib_compute */
+	.emit_ib = gfx_v9_4_3_ring_emit_ib_compute,
+	.emit_fence = gfx_v9_4_3_ring_emit_fence,
+	.emit_pipeline_sync = gfx_v9_4_3_ring_emit_pipeline_sync,
+	.emit_vm_flush = gfx_v9_4_3_ring_emit_vm_flush,
+	.emit_gds_switch = gfx_v9_4_3_ring_emit_gds_switch,
+	.emit_hdp_flush = gfx_v9_4_3_ring_emit_hdp_flush,
+	.test_ring = gfx_v9_4_3_ring_test_ring,
+	.test_ib = gfx_v9_4_3_ring_test_ib,
+	.insert_nop = gfx_v9_4_3_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.emit_wreg = gfx_v9_4_3_ring_emit_wreg,
+	.emit_reg_wait = gfx_v9_4_3_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = gfx_v9_4_3_ring_emit_reg_write_reg_wait,
+	.soft_recovery = gfx_v9_4_3_ring_soft_recovery,
+	.emit_mem_sync = gfx_v9_4_3_emit_mem_sync,
+	.emit_wave_limit = gfx_v9_4_3_emit_wave_limit,
+	.reset = gfx_v9_4_3_reset_kcq,
+	.emit_cleaner_shader = gfx_v9_4_3_ring_emit_cleaner_shader,
+	.begin_use = amdgpu_gfx_enforce_isolation_ring_begin_use,
+	.end_use = amdgpu_gfx_enforce_isolation_ring_end_use,
+};
+
+static const struct amdgpu_ring_funcs gfx_v9_4_3_ring_funcs_kiq = {
+	.type = AMDGPU_RING_TYPE_KIQ,
+	.align_mask = 0xff,
+	.nop = PACKET3(PACKET3_NOP, 0x3FFF),
+	.support_64bit_ptrs = true,
+	.get_rptr = gfx_v9_4_3_ring_get_rptr_compute,
+	.get_wptr = gfx_v9_4_3_ring_get_wptr_compute,
+	.set_wptr = gfx_v9_4_3_ring_set_wptr_compute,
+	.emit_frame_size =
+		20 + /* gfx_v9_4_3_ring_emit_gds_switch */
+		7 + /* gfx_v9_4_3_ring_emit_hdp_flush */
+		5 + /* hdp invalidate */
+		7 + /* gfx_v9_4_3_ring_emit_pipeline_sync */
+		SOC15_FLUSH_GPU_TLB_NUM_WREG * 5 +
+		SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 +
+		2 + /* gfx_v9_4_3_ring_emit_vm_flush */
+		8 + 8 + 8, /* gfx_v9_4_3_ring_emit_fence_kiq x3 for user fence, vm fence */
+	.emit_ib_size =	7, /* gfx_v9_4_3_ring_emit_ib_compute */
+	.emit_fence = gfx_v9_4_3_ring_emit_fence_kiq,
+	.test_ring = gfx_v9_4_3_ring_test_ring,
+	.insert_nop = amdgpu_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
+	.emit_rreg = gfx_v9_4_3_ring_emit_rreg,
+	.emit_wreg = gfx_v9_4_3_ring_emit_wreg,
+	.emit_reg_wait = gfx_v9_4_3_ring_emit_reg_wait,
+	.emit_reg_write_reg_wait = gfx_v9_4_3_ring_emit_reg_write_reg_wait,
+	.emit_hdp_flush = gfx_v9_4_3_ring_emit_hdp_flush,
+};
+
+static void gfx_v9_4_3_set_ring_funcs(struct amdgpu_device *adev)
+{
+	int i, j, num_xcc;
+
+	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
+	for (i = 0; i < num_xcc; i++) {
+		adev->gfx.kiq[i].ring.funcs = &gfx_v9_4_3_ring_funcs_kiq;
+
+		for (j = 0; j < adev->gfx.num_compute_rings; j++)
+			adev->gfx.compute_ring[j + i * adev->gfx.num_compute_rings].funcs
+					= &gfx_v9_4_3_ring_funcs_compute;
+	}
+}
+
+static const struct amdgpu_irq_src_funcs gfx_v9_4_3_eop_irq_funcs = {
+	.set = gfx_v9_4_3_set_eop_interrupt_state,
+	.process = gfx_v9_4_3_eop_irq,
+};
+
+static const struct amdgpu_irq_src_funcs gfx_v9_4_3_priv_reg_irq_funcs = {
+	.set = gfx_v9_4_3_set_priv_reg_fault_state,
+	.process = gfx_v9_4_3_priv_reg_irq,
+};
+
+static const struct amdgpu_irq_src_funcs gfx_v9_4_3_bad_op_irq_funcs = {
+	.set = gfx_v9_4_3_set_bad_op_fault_state,
+	.process = gfx_v9_4_3_bad_op_irq,
+};
+
+static const struct amdgpu_irq_src_funcs gfx_v9_4_3_priv_inst_irq_funcs = {
+	.set = gfx_v9_4_3_set_priv_inst_fault_state,
+	.process = gfx_v9_4_3_priv_inst_irq,
+};
+
+static void gfx_v9_4_3_set_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->gfx.eop_irq.num_types = AMDGPU_CP_IRQ_LAST;
+	adev->gfx.eop_irq.funcs = &gfx_v9_4_3_eop_irq_funcs;
+
+	adev->gfx.priv_reg_irq.num_types = 1;
+	adev->gfx.priv_reg_irq.funcs = &gfx_v9_4_3_priv_reg_irq_funcs;
+
+	adev->gfx.bad_op_irq.num_types = 1;
+	adev->gfx.bad_op_irq.funcs = &gfx_v9_4_3_bad_op_irq_funcs;
+
+	adev->gfx.priv_inst_irq.num_types = 1;
+	adev->gfx.priv_inst_irq.funcs = &gfx_v9_4_3_priv_inst_irq_funcs;
+}
+
+static void gfx_v9_4_3_set_rlc_funcs(struct amdgpu_device *adev)
+{
+	adev->gfx.rlc.funcs = &gfx_v9_4_3_rlc_funcs;
+}
+
+
+static void gfx_v9_4_3_set_gds_init(struct amdgpu_device *adev)
+{
+	/* 9.4.3 variants removed all the GDS internal memory,
+	 * only support GWS opcode in kernel, like barrier
+	 * semaphore.etc */
+
+	/* init asic gds info */
+	adev->gds.gds_size = 0;
+	adev->gds.gds_compute_max_wave_id = 0;
+	adev->gds.gws_size = 64;
+	adev->gds.oa_size = 16;
+}
+
+static void gfx_v9_4_3_set_user_cu_inactive_bitmap(struct amdgpu_device *adev,
+						 u32 bitmap, int xcc_id)
+{
+	u32 data;
+
+	if (!bitmap)
+		return;
+
+	data = bitmap << GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_CUS__SHIFT;
+	data &= GC_USER_SHADER_ARRAY_CONFIG__INACTIVE_CUS_MASK;
+
+	WREG32_SOC15(GC, GET_INST(GC, xcc_id), regGC_USER_SHADER_ARRAY_CONFIG, data);
+}
+
+static u32 gfx_v9_4_3_get_cu_active_bitmap(struct amdgpu_device *adev, int xcc_id)
+{
+	u32 data, mask;
+
+	data = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCC_GC_SHADER_ARRAY_CONFIG);
+	data |= RREG32_SOC15(GC, GET_INST(GC, xcc_id), regGC_USER_SHADER_ARRAY_CONFIG);
+
+	data &= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_CUS_MASK;
+	data >>= CC_GC_SHADER_ARRAY_CONFIG__INACTIVE_CUS__SHIFT;
+
+	mask = amdgpu_gfx_create_bitmask(adev->gfx.config.max_cu_per_sh);
+
+	return (~data) & mask;
+}
+
+static int gfx_v9_4_3_get_cu_info(struct amdgpu_device *adev,
+				 struct amdgpu_cu_info *cu_info)
+{
+	int i, j, k, prev_counter, counter, xcc_id, active_cu_number = 0;
+	u32 mask, bitmap, ao_bitmap, ao_cu_mask = 0, tmp;
+	unsigned disable_masks[4 * 4];
+	bool is_symmetric_cus;
+
+	if (!adev || !cu_info)
+		return -EINVAL;
+
+	/*
+	 * 16 comes from bitmap array size 4*4, and it can cover all gfx9 ASICs
+	 */
+	if (adev->gfx.config.max_shader_engines *
+		adev->gfx.config.max_sh_per_se > 16)
+		return -EINVAL;
+
+	amdgpu_gfx_parse_disable_cu(disable_masks,
+				    adev->gfx.config.max_shader_engines,
+				    adev->gfx.config.max_sh_per_se);
+
+	mutex_lock(&adev->grbm_idx_mutex);
+	for (xcc_id = 0; xcc_id < NUM_XCC(adev->gfx.xcc_mask); xcc_id++) {
+		is_symmetric_cus = true;
+		for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
+			for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
+				mask = 1;
+				ao_bitmap = 0;
+				counter = 0;
+				gfx_v9_4_3_xcc_select_se_sh(adev, i, j, 0xffffffff, xcc_id);
+				gfx_v9_4_3_set_user_cu_inactive_bitmap(
+					adev,
+					disable_masks[i * adev->gfx.config.max_sh_per_se + j],
+					xcc_id);
+				bitmap = gfx_v9_4_3_get_cu_active_bitmap(adev, xcc_id);
+
+				cu_info->bitmap[xcc_id][i][j] = bitmap;
+
+				for (k = 0; k < adev->gfx.config.max_cu_per_sh; k++) {
+					if (bitmap & mask) {
+						if (counter < adev->gfx.config.max_cu_per_sh)
+							ao_bitmap |= mask;
+						counter++;
+					}
+					mask <<= 1;
+				}
+				active_cu_number += counter;
+				if (i < 2 && j < 2)
+					ao_cu_mask |= (ao_bitmap << (i * 16 + j * 8));
+				cu_info->ao_cu_bitmap[i][j] = ao_bitmap;
+			}
+			if (i && is_symmetric_cus && prev_counter != counter)
+				is_symmetric_cus = false;
+			prev_counter = counter;
+		}
+		if (is_symmetric_cus) {
+			tmp = RREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_CPC_DEBUG);
+			tmp = REG_SET_FIELD(tmp, CP_CPC_DEBUG, CPC_HARVESTING_RELAUNCH_DISABLE, 1);
+			tmp = REG_SET_FIELD(tmp, CP_CPC_DEBUG, CPC_HARVESTING_DISPATCH_DISABLE, 1);
+			WREG32_SOC15(GC, GET_INST(GC, xcc_id), regCP_CPC_DEBUG, tmp);
+		}
+		gfx_v9_4_3_xcc_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff,
+					    xcc_id);
+	}
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	cu_info->number = active_cu_number;
+	cu_info->ao_cu_mask = ao_cu_mask;
+	cu_info->simd_per_cu = NUM_SIMD_PER_CU;
+
+	return 0;
+}
+
+const struct amdgpu_ip_block_version gfx_v9_4_3_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_GFX,
+	.major = 9,
+	.minor = 4,
+	.rev = 3,
+	.funcs = &gfx_v9_4_3_ip_funcs,
+};
+
+static int gfx_v9_4_3_xcp_resume(void *handle, uint32_t inst_mask)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	uint32_t tmp_mask;
+	int i, r;
+
+	/* TODO : Initialize golden regs */
+	/* gfx_v9_4_3_init_golden_registers(adev); */
+
+	tmp_mask = inst_mask;
+	for_each_inst(i, tmp_mask)
+		gfx_v9_4_3_xcc_constants_init(adev, i);
+
+	if (!amdgpu_sriov_vf(adev)) {
+		tmp_mask = inst_mask;
+		for_each_inst(i, tmp_mask) {
+			r = gfx_v9_4_3_xcc_rlc_resume(adev, i);
+			if (r)
+				return r;
+		}
+	}
+
+	tmp_mask = inst_mask;
+	for_each_inst(i, tmp_mask) {
+		r = gfx_v9_4_3_xcc_cp_resume(adev, i);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static int gfx_v9_4_3_xcp_suspend(void *handle, uint32_t inst_mask)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	int i;
+
+	for_each_inst(i, inst_mask)
+		gfx_v9_4_3_xcc_fini(adev, i);
+
+	return 0;
+}
+
+struct amdgpu_xcp_ip_funcs gfx_v9_4_3_xcp_funcs = {
+	.suspend = &gfx_v9_4_3_xcp_suspend,
+	.resume = &gfx_v9_4_3_xcp_resume
+};
+
+struct amdgpu_ras_block_hw_ops  gfx_v9_4_3_ras_ops = {
+	.query_ras_error_count = &gfx_v9_4_3_query_ras_error_count,
+	.reset_ras_error_count = &gfx_v9_4_3_reset_ras_error_count,
+};
+
+static int gfx_v9_4_3_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
+{
+	int r;
+
+	r = amdgpu_ras_block_late_init(adev, ras_block);
+	if (r)
+		return r;
+
+	r = amdgpu_ras_bind_aca(adev, AMDGPU_RAS_BLOCK__GFX,
+				&gfx_v9_4_3_aca_info,
+				NULL);
+	if (r)
+		goto late_fini;
+
+	return 0;
+
+late_fini:
+	amdgpu_ras_block_late_fini(adev, ras_block);
+
+	return r;
+}
+
+struct amdgpu_gfx_ras gfx_v9_4_3_ras = {
+	.ras_block = {
+		.hw_ops = &gfx_v9_4_3_ras_ops,
+		.ras_late_init = &gfx_v9_4_3_ras_late_init,
+	},
+	.enable_watchdog_timer = &gfx_v9_4_3_enable_watchdog_timer,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.h
new file mode 100644
index 000000000000..42d67ee0e7ef
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.h
@@ -0,0 +1,31 @@
+/*
+ * Copyright 2022 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __GFX_V9_4_3_H__
+#define __GFX_V9_4_3_H__
+
+extern const struct amdgpu_ip_block_version gfx_v9_4_3_ip_block;
+
+extern struct amdgpu_xcp_ip_funcs gfx_v9_4_3_xcp_funcs;
+
+#endif /* __GFX_V9_4_3_H__ */
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3_cleaner_shader.asm b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3_cleaner_shader.asm
new file mode 100644
index 000000000000..d5325ef80ab0
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3_cleaner_shader.asm
@@ -0,0 +1,153 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+// This shader is to clean LDS, SGPRs and VGPRs. It is  first 64 Dwords or 256 bytes of 192 Dwords cleaner shader.
+//To turn this shader program on for complitaion change this to main and lower shader main to main_1
+ 
+// MI300 : Clear SGPRs, VGPRs and LDS
+//   Uses two kernels launched separately:
+//   1. Clean VGPRs, LDS, and lower SGPRs
+//        Launches one workgroup per CU, each workgroup with 4x wave64 per SIMD in the CU
+//        Waves are "wave64" and have 128 VGPRs each, which uses all 512 VGPRs per SIMD
+//        Waves in the workgroup share the 64KB of LDS
+//        Each wave clears SGPRs 0 - 95. Because there are 4 waves/SIMD, this is physical SGPRs 0-383
+//        Each wave clears 128 VGPRs, so all 512 in the SIMD
+//        The first wave of the workgroup clears its 64KB of LDS
+//        The shader starts with "S_BARRIER" to ensure SPI has launched all waves of the workgroup
+//          before any wave in the workgroup could end.  Without this, it is possible not all SGPRs get cleared.
+//    2. Clean remaining SGPRs
+//        Launches a workgroup with 24 waves per workgroup, yielding 6 waves per SIMD in each CU
+//        Waves are allocating 96 SGPRs
+//          CP sets up SPI_RESOURCE_RESERVE_* registers to prevent these waves from allocating SGPRs 0-223.
+//          As such, these 6 waves per SIMD are allocated physical SGPRs 224-799
+//        Barriers do not work for >16 waves per workgroup, so we cannot start with S_BARRIER
+//          Instead, the shader starts with an S_SETHALT 1. Once all waves are launched CP will send unhalt command
+//        The shader then clears all SGPRs allocated to it, cleaning out physical SGPRs 224-799
+ 
+shader main
+  asic(MI300)
+  type(CS)
+  wave_size(64)
+// Note: original source code from SQ team
+
+//   (theorhetical fastest = ~512clks vgpr + 1536 lds + ~128 sgpr  = 2176 clks)
+
+  s_cmp_eq_u32 s0, 1                                // Bit0 is set, sgpr0 is set then clear VGPRS and LDS as FW set COMPUTE_USER_DATA_3
+  s_cbranch_scc0  label_0023                        // Clean VGPRs and LDS if sgpr0 of wave is set, scc = (s3 == 1)
+  S_BARRIER
+
+  s_movk_i32    m0, 0x0000
+  s_mov_b32     s2, 0x00000078  // Loop 128/8=16 times  (loop unrolled for performance)
+  //
+  // CLEAR VGPRs
+  //
+  s_set_gpr_idx_on  s2, 0x8    // enable Dest VGPR indexing
+label_0005:
+  v_mov_b32     v0, 0
+  v_mov_b32     v1, 0
+  v_mov_b32     v2, 0
+  v_mov_b32     v3, 0
+  v_mov_b32     v4, 0
+  v_mov_b32     v5, 0
+  v_mov_b32     v6, 0
+  v_mov_b32     v7, 0
+  s_sub_u32     s2, s2, 8
+  s_set_gpr_idx_idx  s2
+  s_cbranch_scc0  label_0005
+  s_set_gpr_idx_off
+ 
+  //
+  //
+ 
+  s_mov_b32     s2, 0x80000000                      // Bit31 is first_wave
+  s_and_b32     s2, s2, s1                          // sgpr0 has tg_size (first_wave) term as in ucode only COMPUTE_PGM_RSRC2.tg_size_en is set
+  s_cbranch_scc0  label_clean_sgpr_1                // Clean LDS if its first wave of ThreadGroup/WorkGroup
+  // CLEAR LDS
+  //
+  s_mov_b32 exec_lo, 0xffffffff
+  s_mov_b32 exec_hi, 0xffffffff
+  v_mbcnt_lo_u32_b32  v1, exec_hi, 0          // Set V1 to thread-ID (0..63)
+  v_mbcnt_hi_u32_b32  v1, exec_lo, v1         // Set V1 to thread-ID (0..63)
+  v_mul_u32_u24  v1, 0x00000008, v1           // * 8, so each thread is a double-dword address (8byte)
+  s_mov_b32     s2, 0x00000003f               // 64 loop iteraions
+  s_mov_b32     m0, 0xffffffff
+  // Clear all of LDS space
+  // Each FirstWave of WorkGroup clears 64kbyte block
+ 
+label_001F:
+  ds_write2_b64  v1, v[2:3], v[2:3] offset1:32
+  ds_write2_b64  v1, v[4:5], v[4:5] offset0:64 offset1:96
+  v_add_co_u32     v1, vcc, 0x00000400, v1
+  s_sub_u32     s2, s2, 1
+  s_cbranch_scc0  label_001F
+  //
+  // CLEAR SGPRs
+  //
+label_clean_sgpr_1:
+  s_mov_b32     m0, 0x0000005c   // Loop 96/4=24 times  (loop unrolled for performance)
+  s_nop 0
+label_sgpr_loop:
+  s_movreld_b32     s0, 0
+  s_movreld_b32     s1, 0
+  s_movreld_b32     s2, 0
+  s_movreld_b32     s3, 0
+  s_sub_u32         m0, m0, 4
+  s_cbranch_scc0  label_sgpr_loop
+ 
+  //clear vcc, flat scratch
+  s_mov_b32 flat_scratch_lo, 0   //clear  flat scratch lo SGPR 
+  s_mov_b32 flat_scratch_hi, 0   //clear  flat scratch hi SGPR 
+  s_mov_b64 vcc, 0               //clear vcc
+  s_mov_b64 ttmp0, 0             //Clear ttmp0 and ttmp1 
+  s_mov_b64 ttmp2, 0             //Clear ttmp2 and ttmp3 
+  s_mov_b64 ttmp4, 0             //Clear ttmp4 and ttmp5 
+  s_mov_b64 ttmp6, 0             //Clear ttmp6 and ttmp7 
+  s_mov_b64 ttmp8, 0             //Clear ttmp8 and ttmp9 
+  s_mov_b64 ttmp10, 0            //Clear ttmp10 and ttmp11 
+  s_mov_b64 ttmp12, 0            //Clear ttmp12 and ttmp13 
+  s_mov_b64 ttmp14, 0            //Clear ttmp14 and ttmp15 
+s_endpgm
+
+label_0023:
+
+  s_sethalt 1
+
+  s_mov_b32     m0, 0x0000005c   // Loop 96/4=24 times  (loop unrolled for performance)
+  s_nop 0
+label_sgpr_loop1:
+
+  s_movreld_b32     s0, 0
+  s_movreld_b32     s1, 0
+  s_movreld_b32     s2, 0
+  s_movreld_b32     s3, 0
+  s_sub_u32         m0, m0, 4
+  s_cbranch_scc0  label_sgpr_loop1
+ 
+  //clear vcc, flat scratch
+  s_mov_b32 flat_scratch_lo, 0   //clear  flat scratch lo SGPR 
+  s_mov_b32 flat_scratch_hi, 0   //clear  flat scratch hi SGPR 
+  s_mov_b64 vcc, 0xee            //clear vcc
+
+s_endpgm
+end  
+  
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3_cleaner_shader.h b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3_cleaner_shader.h
new file mode 100644
index 000000000000..69aa567c6c1d
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3_cleaner_shader.h
@@ -0,0 +1,64 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright 2024 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/* Define the cleaner shader gfx_9_4_3 */
+static const u32 gfx_9_4_3_cleaner_shader_hex[] = {
+	0xbf068100, 0xbf84003b,
+	0xbf8a0000, 0xb07c0000,
+	0xbe8200ff, 0x00000078,
+	0xbf110802, 0x7e000280,
+	0x7e020280, 0x7e040280,
+	0x7e060280, 0x7e080280,
+	0x7e0a0280, 0x7e0c0280,
+	0x7e0e0280, 0x80828802,
+	0xbe803202, 0xbf84fff5,
+	0xbf9c0000, 0xbe8200ff,
+	0x80000000, 0x86020102,
+	0xbf840011, 0xbefe00c1,
+	0xbeff00c1, 0xd28c0001,
+	0x0001007f, 0xd28d0001,
+	0x0002027e, 0x10020288,
+	0xbe8200bf, 0xbefc00c1,
+	0xd89c2000, 0x00020201,
+	0xd89c6040, 0x00040401,
+	0x320202ff, 0x00000400,
+	0x80828102, 0xbf84fff8,
+	0xbefc00ff, 0x0000005c,
+	0xbf800000, 0xbe802c80,
+	0xbe812c80, 0xbe822c80,
+	0xbe832c80, 0x80fc847c,
+	0xbf84fffa, 0xbee60080,
+	0xbee70080, 0xbeea0180,
+	0xbeec0180, 0xbeee0180,
+	0xbef00180, 0xbef20180,
+	0xbef40180, 0xbef60180,
+	0xbef80180, 0xbefa0180,
+	0xbf810000, 0xbf8d0001,
+	0xbefc00ff, 0x0000005c,
+	0xbf800000, 0xbe802c80,
+	0xbe812c80, 0xbe822c80,
+	0xbe832c80, 0x80fc847c,
+	0xbf84fffa, 0xbee60080,
+	0xbee70080, 0xbeea01ff,
+	0x000000ee, 0xbf810000,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v11_5_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v11_5_0.c
new file mode 100644
index 000000000000..f9949fedfbb9
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v11_5_0.c
@@ -0,0 +1,516 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+#include "gfxhub_v11_5_0.h"
+
+#include "gc/gc_11_5_0_offset.h"
+#include "gc/gc_11_5_0_sh_mask.h"
+
+#include "navi10_enum.h"
+#include "soc15_common.h"
+
+#define regGCVM_L2_CNTL3_DEFAULT                0x80100007
+#define regGCVM_L2_CNTL4_DEFAULT                0x000000c1
+#define regGCVM_L2_CNTL5_DEFAULT                0x00003fe0
+
+
+static const char *gfxhub_client_ids[] = {
+	"CB/DB",
+	"Reserved",
+	"GE1",
+	"GE2",
+	"CPF",
+	"CPC",
+	"CPG",
+	"RLC",
+	"TCP",
+	"SQC (inst)",
+	"SQC (data)",
+	"SQG",
+	"Reserved",
+	"SDMA0",
+	"SDMA1",
+	"GCR",
+	"SDMA2",
+	"SDMA3",
+};
+
+static uint32_t gfxhub_v11_5_0_get_invalidate_req(unsigned int vmid,
+					       uint32_t flush_type)
+{
+	u32 req = 0;
+
+	/* invalidate using legacy mode on vmid*/
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ,
+			    PER_VMID_INVALIDATE_REQ, 1 << vmid);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, flush_type);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE2, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L1_PTES, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ,
+			    CLEAR_PROTECTION_FAULT_STATUS_ADDR,	0);
+
+	return req;
+}
+
+static void
+gfxhub_v11_5_0_print_l2_protection_fault_status(struct amdgpu_device *adev,
+					     uint32_t status)
+{
+	u32 cid = REG_GET_FIELD(status,
+				GCVM_L2_PROTECTION_FAULT_STATUS, CID);
+
+	dev_err(adev->dev,
+		"GCVM_L2_PROTECTION_FAULT_STATUS:0x%08X\n",
+		status);
+	dev_err(adev->dev, "\t Faulty UTCL2 client ID: %s (0x%x)\n",
+		cid >= ARRAY_SIZE(gfxhub_client_ids) ? "unknown" : gfxhub_client_ids[cid],
+		cid);
+	dev_err(adev->dev, "\t MORE_FAULTS: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS, MORE_FAULTS));
+	dev_err(adev->dev, "\t WALKER_ERROR: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS, WALKER_ERROR));
+	dev_err(adev->dev, "\t PERMISSION_FAULTS: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS, PERMISSION_FAULTS));
+	dev_err(adev->dev, "\t MAPPING_ERROR: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS, MAPPING_ERROR));
+	dev_err(adev->dev, "\t RW: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS, RW));
+}
+
+static u64 gfxhub_v11_5_0_get_fb_location(struct amdgpu_device *adev)
+{
+	u64 base = RREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_BASE);
+
+	base &= GCMC_VM_FB_LOCATION_BASE__FB_BASE_MASK;
+	base <<= 24;
+
+	return base;
+}
+
+static u64 gfxhub_v11_5_0_get_mc_fb_offset(struct amdgpu_device *adev)
+{
+	return (u64)RREG32_SOC15(GC, 0, regGCMC_VM_FB_OFFSET) << 24;
+}
+
+static void gfxhub_v11_5_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
+				uint64_t page_table_base)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+
+	WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
+			    hub->ctx_addr_distance * vmid,
+			    lower_32_bits(page_table_base));
+
+	WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
+			    hub->ctx_addr_distance * vmid,
+			    upper_32_bits(page_table_base));
+}
+
+static void gfxhub_v11_5_0_init_gart_aperture_regs(struct amdgpu_device *adev)
+{
+	uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo);
+
+	gfxhub_v11_5_0_setup_vm_pt_regs(adev, 0, pt_base);
+
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
+		     (u32)(adev->gmc.gart_start >> 12));
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
+		     (u32)(adev->gmc.gart_start >> 44));
+
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
+		     (u32)(adev->gmc.gart_end >> 12));
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
+		     (u32)(adev->gmc.gart_end >> 44));
+}
+
+static void gfxhub_v11_5_0_init_system_aperture_regs(struct amdgpu_device *adev)
+{
+	uint64_t value;
+
+	WREG32_SOC15(GC, 0, regGCMC_VM_AGP_BASE, 0);
+	WREG32_SOC15(GC, 0, regGCMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
+	WREG32_SOC15(GC, 0, regGCMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
+
+	/* Program the system aperture low logical page number. */
+	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_LOW_ADDR,
+		     min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
+
+	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+		     max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
+
+	/* Set default page address. */
+	value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
+	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
+		     (u32)(value >> 12));
+	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
+		     (u32)(value >> 44));
+
+	/* Program "protection fault". */
+	WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32,
+		     (u32)(adev->dummy_page_addr >> 12));
+	WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32,
+		     (u32)((u64)adev->dummy_page_addr >> 44));
+
+	WREG32_FIELD15_PREREG(GC, 0, GCVM_L2_PROTECTION_FAULT_CNTL2,
+		       ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1);
+}
+
+static void gfxhub_v11_5_0_init_tlb_regs(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	/* Setup TLB control */
+	tmp = RREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL);
+
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, 3);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL,
+			    ENABLE_ADVANCED_DRIVER_MODEL, 1);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL,
+			    SYSTEM_APERTURE_UNMAPPED_ACCESS, 0);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ECO_BITS, 0);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL,
+			    MTYPE, MTYPE_UC); /* UC, uncached */
+
+	WREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, tmp);
+}
+
+static void gfxhub_v11_5_0_init_cache_regs(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	/* These registers are not accessible to VF-SRIOV.
+	 * The PF will program them instead.
+	 */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	/* Setup L2 cache */
+	tmp = RREG32_SOC15(GC, 0, regGCVM_L2_CNTL);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, ENABLE_L2_CACHE, 1);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL,
+			    ENABLE_DEFAULT_PAGE_OUT_TO_SYSTEM_MEMORY, 1);
+	/* XXX for emulation, Refer to closed source code.*/
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL,
+			    L2_PDE0_CACHE_TAG_GENERATION_MODE, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL, tmp);
+
+	tmp = RREG32_SOC15(GC, 0, regGCVM_L2_CNTL2);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL2, INVALIDATE_L2_CACHE, 1);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL2, tmp);
+
+	tmp = regGCVM_L2_CNTL3_DEFAULT;
+	if (adev->gmc.translate_further) {
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3, BANK_SELECT, 12);
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3,
+				    L2_CACHE_BIGK_FRAGMENT_SIZE, 9);
+	} else {
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3, BANK_SELECT, 9);
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3,
+				    L2_CACHE_BIGK_FRAGMENT_SIZE, 6);
+	}
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL3, tmp);
+
+	tmp = regGCVM_L2_CNTL4_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL4, tmp);
+
+	tmp = regGCVM_L2_CNTL5_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL5, L2_CACHE_SMALLK_FRAGMENT_SIZE, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL5, tmp);
+}
+
+static void gfxhub_v11_5_0_enable_system_domain(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	tmp = RREG32_SOC15(GC, 0, regGCVM_CONTEXT0_CNTL);
+	tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
+	tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL,
+			    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_CNTL, tmp);
+}
+
+static void gfxhub_v11_5_0_disable_identity_aperture(struct amdgpu_device *adev)
+{
+	/* These registers are not accessible to VF-SRIOV.
+	 * The PF will program them instead.
+	 */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32,
+		     0xFFFFFFFF);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32,
+		     0x0000000F);
+
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32,
+		     0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32,
+		     0);
+
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, 0);
+
+}
+
+static void gfxhub_v11_5_0_setup_vmid_config(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	int i;
+	uint32_t tmp;
+
+	for (i = 0; i <= 14; i++) {
+		tmp = RREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL, i * hub->ctx_distance);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
+				    adev->vm_manager.num_level);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				PAGE_TABLE_BLOCK_SIZE,
+				adev->vm_manager.block_size - 9);
+		/* Send no-retry XNACK on fault to suppress VM fault storm. */
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT,
+				    !amdgpu_noretry);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL,
+				    i * hub->ctx_distance, tmp);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32,
+				    i * hub->ctx_addr_distance, 0);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32,
+				    i * hub->ctx_addr_distance, 0);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32,
+				    i * hub->ctx_addr_distance,
+				    lower_32_bits(adev->vm_manager.max_pfn - 1));
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32,
+				    i * hub->ctx_addr_distance,
+				    upper_32_bits(adev->vm_manager.max_pfn - 1));
+	}
+
+	hub->vm_cntx_cntl = tmp;
+}
+
+static void gfxhub_v11_5_0_program_invalidation(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	unsigned i;
+
+	for (i = 0 ; i < 18; ++i) {
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
+				    i * hub->eng_addr_distance, 0xffffffff);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ADDR_RANGE_HI32,
+				    i * hub->eng_addr_distance, 0x1f);
+	}
+}
+
+static int gfxhub_v11_5_0_gart_enable(struct amdgpu_device *adev)
+{
+	if (amdgpu_sriov_vf(adev)) {
+		/*
+		 * GCMC_VM_FB_LOCATION_BASE/TOP is NULL for VF, becuase they are
+		 * VF copy registers so vbios post doesn't program them, for
+		 * SRIOV driver need to program them
+		 */
+		WREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_BASE,
+			     adev->gmc.vram_start >> 24);
+		WREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_TOP,
+			     adev->gmc.vram_end >> 24);
+	}
+
+	/* GART Enable. */
+	gfxhub_v11_5_0_init_gart_aperture_regs(adev);
+	gfxhub_v11_5_0_init_system_aperture_regs(adev);
+	gfxhub_v11_5_0_init_tlb_regs(adev);
+	gfxhub_v11_5_0_init_cache_regs(adev);
+
+	gfxhub_v11_5_0_enable_system_domain(adev);
+	gfxhub_v11_5_0_disable_identity_aperture(adev);
+	gfxhub_v11_5_0_setup_vmid_config(adev);
+	gfxhub_v11_5_0_program_invalidation(adev);
+
+	return 0;
+}
+
+static void gfxhub_v11_5_0_gart_disable(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	u32 tmp;
+	u32 i;
+
+	/* Disable all tables */
+	for (i = 0; i < 16; i++)
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_CNTL,
+				    i * hub->ctx_distance, 0);
+
+	/* Setup TLB control */
+	tmp = RREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL,
+			    ENABLE_ADVANCED_DRIVER_MODEL, 0);
+	WREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, tmp);
+
+	/* Setup L2 cache */
+	WREG32_FIELD15_PREREG(GC, 0, GCVM_L2_CNTL, ENABLE_L2_CACHE, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL3, 0);
+}
+
+/**
+ * gfxhub_v11_5_0_set_fault_enable_default - update GART/VM fault handling
+ *
+ * @adev: amdgpu_device pointer
+ * @value: true redirects VM faults to the default page
+ */
+static void gfxhub_v11_5_0_set_fault_enable_default(struct amdgpu_device *adev,
+					  bool value)
+{
+	u32 tmp;
+
+	/* NO halt CP when page fault */
+	tmp = RREG32_SOC15(GC, 0, regCP_DEBUG);
+	tmp = REG_SET_FIELD(tmp, CP_DEBUG, CPG_UTCL1_ERROR_HALT_DISABLE, 1);
+	WREG32_SOC15(GC, 0, regCP_DEBUG, tmp);
+
+	/* These registers are not accessible to VF-SRIOV.
+	 * The PF will program them instead.
+	 */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	tmp = RREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT,
+			    value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    READ_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	if (!value) {
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_NO_RETRY_FAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_RETRY_FAULT, 1);
+	}
+	WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL, tmp);
+}
+
+static const struct amdgpu_vmhub_funcs gfxhub_v11_5_0_vmhub_funcs = {
+	.print_l2_protection_fault_status = gfxhub_v11_5_0_print_l2_protection_fault_status,
+	.get_invalidate_req = gfxhub_v11_5_0_get_invalidate_req,
+};
+
+static void gfxhub_v11_5_0_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+
+	hub->ctx0_ptb_addr_lo32 =
+		SOC15_REG_OFFSET(GC, 0,
+				 regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32);
+	hub->ctx0_ptb_addr_hi32 =
+		SOC15_REG_OFFSET(GC, 0,
+				 regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32);
+	hub->vm_inv_eng0_sem =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_SEM);
+	hub->vm_inv_eng0_req =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_REQ);
+	hub->vm_inv_eng0_ack =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ACK);
+	hub->vm_context0_cntl =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_CONTEXT0_CNTL);
+	hub->vm_l2_pro_fault_status =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_L2_PROTECTION_FAULT_STATUS);
+	hub->vm_l2_pro_fault_cntl =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL);
+
+	hub->ctx_distance = regGCVM_CONTEXT1_CNTL - regGCVM_CONTEXT0_CNTL;
+	hub->ctx_addr_distance = regGCVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 -
+		regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32;
+	hub->eng_distance = regGCVM_INVALIDATE_ENG1_REQ -
+		regGCVM_INVALIDATE_ENG0_REQ;
+	hub->eng_addr_distance = regGCVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 -
+		regGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32;
+
+	hub->vm_cntx_cntl_vm_fault = GCVM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__EXECUTE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK;
+
+	hub->vmhub_funcs = &gfxhub_v11_5_0_vmhub_funcs;
+}
+
+const struct amdgpu_gfxhub_funcs gfxhub_v11_5_0_funcs = {
+	.get_fb_location = gfxhub_v11_5_0_get_fb_location,
+	.get_mc_fb_offset = gfxhub_v11_5_0_get_mc_fb_offset,
+	.setup_vm_pt_regs = gfxhub_v11_5_0_setup_vm_pt_regs,
+	.gart_enable = gfxhub_v11_5_0_gart_enable,
+	.gart_disable = gfxhub_v11_5_0_gart_disable,
+	.set_fault_enable_default = gfxhub_v11_5_0_set_fault_enable_default,
+	.init = gfxhub_v11_5_0_init,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v11_5_0.h b/drivers/gpu/drm/amd/amdgpu/gfxhub_v11_5_0.h
new file mode 100644
index 000000000000..265ab631b3d0
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v11_5_0.h
@@ -0,0 +1,29 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __GFXHUB_V11_5_0_H__
+#define __GFXHUB_V11_5_0_H__
+
+extern const struct amdgpu_gfxhub_funcs gfxhub_v11_5_0_funcs;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v12_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v12_0.c
new file mode 100644
index 000000000000..7609b9cecae8
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v12_0.c
@@ -0,0 +1,521 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "amdgpu.h"
+#include "gfxhub_v12_0.h"
+
+#include "gc/gc_12_0_0_offset.h"
+#include "gc/gc_12_0_0_sh_mask.h"
+#include "soc24_enum.h"
+#include "soc15_common.h"
+
+#define regGCVM_L2_CNTL3_DEFAULT	0x80120007
+#define regGCVM_L2_CNTL4_DEFAULT	0x000000c1
+#define regGCVM_L2_CNTL5_DEFAULT	0x00003fe0
+#define regGRBM_GFX_INDEX_DEFAULT	0xe0000000
+
+static const char *gfxhub_client_ids[] = {
+	"CB",
+	"DB",
+	"GE1",
+	"GE2",
+	"CPF",
+	"CPC",
+	"CPG",
+	"RLC",
+	"TCP",
+	"SQC (inst)",
+	"SQC (data)",
+	"SQG/PC/SC",
+	"Reserved",
+	"SDMA0",
+	"SDMA1",
+	"GCR",
+	"Reserved",
+	"Reserved",
+	"WGS",
+	"DSM",
+	"PA"
+};
+
+static uint32_t gfxhub_v12_0_get_invalidate_req(unsigned int vmid,
+					        uint32_t flush_type)
+{
+	u32 req = 0;
+
+	/* invalidate using legacy mode on vmid*/
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ,
+			    PER_VMID_INVALIDATE_REQ, 1 << vmid);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, flush_type);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE2, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ, INVALIDATE_L1_PTES, 1);
+	req = REG_SET_FIELD(req, GCVM_INVALIDATE_ENG0_REQ,
+			    CLEAR_PROTECTION_FAULT_STATUS_ADDR,	0);
+
+	return req;
+}
+
+static void
+gfxhub_v12_0_print_l2_protection_fault_status(struct amdgpu_device *adev,
+					      uint32_t status)
+{
+	u32 cid = REG_GET_FIELD(status,
+				GCVM_L2_PROTECTION_FAULT_STATUS_LO32, CID);
+
+	dev_err(adev->dev,
+		"GCVM_L2_PROTECTION_FAULT_STATUS:0x%08X\n",
+		status);
+	dev_err(adev->dev, "\t Faulty UTCL2 client ID: %s (0x%x)\n",
+		cid >= ARRAY_SIZE(gfxhub_client_ids) ? "unknown" : gfxhub_client_ids[cid],
+		cid);
+	dev_err(adev->dev, "\t MORE_FAULTS: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS_LO32, MORE_FAULTS));
+	dev_err(adev->dev, "\t WALKER_ERROR: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS_LO32, WALKER_ERROR));
+	dev_err(adev->dev, "\t PERMISSION_FAULTS: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS_LO32, PERMISSION_FAULTS));
+	dev_err(adev->dev, "\t MAPPING_ERROR: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS_LO32, MAPPING_ERROR));
+	dev_err(adev->dev, "\t RW: 0x%lx\n",
+		REG_GET_FIELD(status,
+		GCVM_L2_PROTECTION_FAULT_STATUS_LO32, RW));
+}
+
+static u64 gfxhub_v12_0_get_fb_location(struct amdgpu_device *adev)
+{
+	u64 base = RREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_BASE);
+
+	base &= GCMC_VM_FB_LOCATION_BASE__FB_BASE_MASK;
+	base <<= 24;
+
+	return base;
+}
+
+static u64 gfxhub_v12_0_get_mc_fb_offset(struct amdgpu_device *adev)
+{
+	return (u64)RREG32_SOC15(GC, 0, regGCMC_VM_FB_OFFSET) << 24;
+}
+
+static void gfxhub_v12_0_setup_vm_pt_regs(struct amdgpu_device *adev,
+					  uint32_t vmid,
+					  uint64_t page_table_base)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+
+	WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
+			    hub->ctx_addr_distance * vmid,
+			    lower_32_bits(page_table_base));
+
+	WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
+			    hub->ctx_addr_distance * vmid,
+			    upper_32_bits(page_table_base));
+}
+
+static void gfxhub_v12_0_init_gart_aperture_regs(struct amdgpu_device *adev)
+{
+	uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo);
+
+	gfxhub_v12_0_setup_vm_pt_regs(adev, 0, pt_base);
+
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
+		     (u32)(adev->gmc.gart_start >> 12));
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
+		     (u32)(adev->gmc.gart_start >> 44));
+
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
+		     (u32)(adev->gmc.gart_end >> 12));
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
+		     (u32)(adev->gmc.gart_end >> 44));
+}
+
+static void gfxhub_v12_0_init_system_aperture_regs(struct amdgpu_device *adev)
+{
+	uint64_t value;
+
+	/* Program the AGP BAR */
+	WREG32_SOC15(GC, 0, regGCMC_VM_AGP_BASE, 0);
+	WREG32_SOC15(GC, 0, regGCMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
+	WREG32_SOC15(GC, 0, regGCMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
+
+	/* Program the system aperture low logical page number. */
+	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_LOW_ADDR,
+		     min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
+	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+		     max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
+
+	/* Set default page address. */
+	value = adev->mem_scratch.gpu_addr - adev->gmc.vram_start
+		+ adev->vm_manager.vram_base_offset;
+	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
+		     (u32)(value >> 12));
+	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
+		     (u32)(value >> 44));
+
+	/* Program "protection fault". */
+	WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32,
+		     (u32)(adev->dummy_page_addr >> 12));
+	WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32,
+		     (u32)((u64)adev->dummy_page_addr >> 44));
+
+	WREG32_FIELD15_PREREG(GC, 0, GCVM_L2_PROTECTION_FAULT_CNTL2,
+		       ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1);
+}
+
+
+static void gfxhub_v12_0_init_tlb_regs(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	/* Setup TLB control */
+	tmp = RREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL);
+
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, 3);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL,
+			    ENABLE_ADVANCED_DRIVER_MODEL, 1);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL,
+			    SYSTEM_APERTURE_UNMAPPED_ACCESS, 0);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ECO_BITS, 0);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL,
+			    MTYPE, MTYPE_UC); /* UC, uncached */
+
+	WREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, tmp);
+}
+
+static void gfxhub_v12_0_init_cache_regs(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	/* These registers are not accessible to VF-SRIOV.
+	 * The PF will program them instead.
+	 */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	/* Setup L2 cache */
+	tmp = RREG32_SOC15(GC, 0, regGCVM_L2_CNTL);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, ENABLE_L2_CACHE, 1);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL,
+			    ENABLE_DEFAULT_PAGE_OUT_TO_SYSTEM_MEMORY, 1);
+	/* XXX for emulation, Refer to closed source code.*/
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL,
+			    L2_PDE0_CACHE_TAG_GENERATION_MODE, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL, tmp);
+
+	tmp = RREG32_SOC15(GC, 0, regGCVM_L2_CNTL2);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL2, INVALIDATE_L2_CACHE, 1);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL2, tmp);
+
+	tmp = regGCVM_L2_CNTL3_DEFAULT;
+	if (adev->gmc.translate_further) {
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3, BANK_SELECT, 12);
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3,
+				    L2_CACHE_BIGK_FRAGMENT_SIZE, 9);
+	} else {
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3, BANK_SELECT, 9);
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL3,
+				    L2_CACHE_BIGK_FRAGMENT_SIZE, 6);
+	}
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL3, tmp);
+
+	tmp = regGCVM_L2_CNTL4_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL4, tmp);
+
+	tmp = regGCVM_L2_CNTL5_DEFAULT;
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_CNTL5, L2_CACHE_SMALLK_FRAGMENT_SIZE, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL5, tmp);
+}
+
+static void gfxhub_v12_0_enable_system_domain(struct amdgpu_device *adev)
+{
+	uint32_t tmp;
+
+	tmp = RREG32_SOC15(GC, 0, regGCVM_CONTEXT0_CNTL);
+	tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
+	tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH, 0);
+	tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT0_CNTL,
+			    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
+	WREG32_SOC15(GC, 0, regGCVM_CONTEXT0_CNTL, tmp);
+}
+
+static void gfxhub_v12_0_disable_identity_aperture(struct amdgpu_device *adev)
+{
+	/* These registers are not accessible to VF-SRIOV.
+	 * The PF will program them instead.
+	 */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32,
+		     0xFFFFFFFF);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32,
+		     0x0000000F);
+
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32,
+		     0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32,
+		     0);
+
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, 0);
+
+}
+
+static void gfxhub_v12_0_setup_vmid_config(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	int i;
+	uint32_t tmp;
+
+	for (i = 0; i <= 14; i++) {
+		tmp = RREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL, i);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
+				    adev->vm_manager.num_level);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				PAGE_TABLE_BLOCK_SIZE,
+				adev->vm_manager.block_size - 9);
+		/* Send no-retry XNACK on fault to suppress VM fault storm. */
+		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL,
+				    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT,
+				    !amdgpu_noretry);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL,
+				    i * hub->ctx_distance, tmp);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32,
+				    i * hub->ctx_addr_distance, 0);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32,
+				    i * hub->ctx_addr_distance, 0);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32,
+				    i * hub->ctx_addr_distance,
+				    lower_32_bits(adev->vm_manager.max_pfn - 1));
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32,
+				    i * hub->ctx_addr_distance,
+				    upper_32_bits(adev->vm_manager.max_pfn - 1));
+	}
+
+	hub->vm_cntx_cntl = tmp;
+}
+
+static void gfxhub_v12_0_program_invalidation(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	unsigned i;
+
+	for (i = 0 ; i < 18; ++i) {
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
+				    i * hub->eng_addr_distance, 0xffffffff);
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ADDR_RANGE_HI32,
+				    i * hub->eng_addr_distance, 0x1f);
+	}
+}
+
+static int gfxhub_v12_0_gart_enable(struct amdgpu_device *adev)
+{
+	if (amdgpu_sriov_vf(adev)) {
+		/*
+		 * GCMC_VM_FB_LOCATION_BASE/TOP is NULL for VF, becuase they are
+		 * VF copy registers so vbios post doesn't program them, for
+		 * SRIOV driver need to program them
+		 */
+		WREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_BASE,
+			     adev->gmc.vram_start >> 24);
+		WREG32_SOC15(GC, 0, regGCMC_VM_FB_LOCATION_TOP,
+			     adev->gmc.vram_end >> 24);
+	}
+
+	/* GART Enable. */
+	gfxhub_v12_0_init_gart_aperture_regs(adev);
+	gfxhub_v12_0_init_system_aperture_regs(adev);
+	gfxhub_v12_0_init_tlb_regs(adev);
+	gfxhub_v12_0_init_cache_regs(adev);
+
+	gfxhub_v12_0_enable_system_domain(adev);
+	gfxhub_v12_0_disable_identity_aperture(adev);
+	gfxhub_v12_0_setup_vmid_config(adev);
+	gfxhub_v12_0_program_invalidation(adev);
+
+	return 0;
+}
+
+static void gfxhub_v12_0_gart_disable(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	u32 tmp;
+	u32 i;
+
+	/* Disable all tables */
+	for (i = 0; i < 16; i++)
+		WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_CNTL,
+				    i * hub->ctx_distance, 0);
+
+	/* Setup TLB control */
+	tmp = RREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0);
+	tmp = REG_SET_FIELD(tmp, GCMC_VM_MX_L1_TLB_CNTL,
+			    ENABLE_ADVANCED_DRIVER_MODEL, 0);
+	WREG32_SOC15(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, tmp);
+
+	/* Setup L2 cache */
+	WREG32_FIELD15_PREREG(GC, 0, GCVM_L2_CNTL, ENABLE_L2_CACHE, 0);
+	WREG32_SOC15(GC, 0, regGCVM_L2_CNTL3, 0);
+}
+
+/**
+ * gfxhub_v12_0_set_fault_enable_default - update GART/VM fault handling
+ *
+ * @adev: amdgpu_device pointer
+ * @value: true redirects VM faults to the default page
+ */
+static void gfxhub_v12_0_set_fault_enable_default(struct amdgpu_device *adev,
+						  bool value)
+{
+	u32 tmp;
+
+	/* NO halt CP when page fault */
+	tmp = RREG32_SOC15(GC, 0, regCP_DEBUG);
+	tmp = REG_SET_FIELD(tmp, CP_DEBUG, CPG_UTCL1_ERROR_HALT_DISABLE, 1);
+	WREG32_SOC15(GC, 0, regCP_DEBUG, tmp);
+
+	/* These registers are not accessible to VF-SRIOV.
+	 * The PF will program them instead.
+	 */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
+	tmp = RREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT,
+			    value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    READ_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+			    EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+	if (!value) {
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_NO_RETRY_FAULT, 1);
+		tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
+				CRASH_ON_RETRY_FAULT, 1);
+	}
+	WREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL, tmp);
+}
+
+static const struct amdgpu_vmhub_funcs gfxhub_v12_0_vmhub_funcs = {
+	.print_l2_protection_fault_status = gfxhub_v12_0_print_l2_protection_fault_status,
+	.get_invalidate_req = gfxhub_v12_0_get_invalidate_req,
+};
+
+static void gfxhub_v12_0_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+
+	hub->ctx0_ptb_addr_lo32 =
+		SOC15_REG_OFFSET(GC, 0,
+				 regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32);
+	hub->ctx0_ptb_addr_hi32 =
+		SOC15_REG_OFFSET(GC, 0,
+				 regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32);
+	hub->vm_inv_eng0_sem =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_SEM);
+	hub->vm_inv_eng0_req =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_REQ);
+	hub->vm_inv_eng0_ack =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ACK);
+	hub->vm_context0_cntl =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_CONTEXT0_CNTL);
+	hub->vm_l2_pro_fault_status =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_L2_PROTECTION_FAULT_STATUS_LO32);
+	hub->vm_l2_pro_fault_cntl =
+		SOC15_REG_OFFSET(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL);
+
+	hub->ctx_distance = regGCVM_CONTEXT1_CNTL - regGCVM_CONTEXT0_CNTL;
+	hub->ctx_addr_distance = regGCVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 -
+		regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32;
+	hub->eng_distance = regGCVM_INVALIDATE_ENG1_REQ -
+		regGCVM_INVALIDATE_ENG0_REQ;
+	hub->eng_addr_distance = regGCVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 -
+		regGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32;
+
+	hub->vm_cntx_cntl_vm_fault = GCVM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		GCVM_CONTEXT1_CNTL__EXECUTE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK;
+
+	hub->vmhub_funcs = &gfxhub_v12_0_vmhub_funcs;
+}
+
+const struct amdgpu_gfxhub_funcs gfxhub_v12_0_funcs = {
+	.get_fb_location = gfxhub_v12_0_get_fb_location,
+	.get_mc_fb_offset = gfxhub_v12_0_get_mc_fb_offset,
+	.setup_vm_pt_regs = gfxhub_v12_0_setup_vm_pt_regs,
+	.gart_enable = gfxhub_v12_0_gart_enable,
+	.gart_disable = gfxhub_v12_0_gart_disable,
+	.set_fault_enable_default = gfxhub_v12_0_set_fault_enable_default,
+	.init = gfxhub_v12_0_init,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v12_0.h b/drivers/gpu/drm/amd/amdgpu/gfxhub_v12_0.h
new file mode 100644
index 000000000000..f1258265f802
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v12_0.h
@@ -0,0 +1,29 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __GFXHUB_V12_0_H__
+#define __GFXHUB_V12_0_H__
+
+extern const struct amdgpu_gfxhub_funcs gfxhub_v12_0_funcs;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c
index ab2325f6c7ac..a7bfc9f41d0e 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c
@@ -40,7 +40,7 @@ static void gfxhub_v1_0_setup_vm_pt_regs(struct amdgpu_device *adev,
 					 uint32_t vmid,
 					 uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	WREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			    hub->ctx_addr_distance * vmid,
@@ -92,18 +92,20 @@ static void gfxhub_v1_0_init_system_aperture_regs(struct amdgpu_device *adev)
 {
 	uint64_t value;
 
-	/* Program the AGP BAR */
-	WREG32_SOC15_RLC(GC, 0, mmMC_VM_AGP_BASE, 0);
-	WREG32_SOC15_RLC(GC, 0, mmMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
-	WREG32_SOC15_RLC(GC, 0, mmMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
-
 	if (!amdgpu_sriov_vf(adev) || adev->asic_type <= CHIP_VEGA10) {
+		/* Program the AGP BAR */
+		WREG32_SOC15_RLC(GC, 0, mmMC_VM_AGP_BASE, 0);
+		WREG32_SOC15_RLC(GC, 0, mmMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
+		WREG32_SOC15_RLC(GC, 0, mmMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
+
 		/* Program the system aperture low logical page number. */
 		WREG32_SOC15_RLC(GC, 0, mmMC_VM_SYSTEM_APERTURE_LOW_ADDR,
 			min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
 
-		if (adev->apu_flags & AMD_APU_IS_RAVEN2)
-			/*
+		if (adev->apu_flags & (AMD_APU_IS_RAVEN2 |
+				       AMD_APU_IS_RENOIR |
+				       AMD_APU_IS_GREEN_SARDINE))
+		       /*
 			* Raven2 has a HW issue that it is unable to use the
 			* vram which is out of MC_VM_SYSTEM_APERTURE_HIGH_ADDR.
 			* So here is the workaround that increase system
@@ -247,8 +249,8 @@ static void gfxhub_v1_0_disable_identity_aperture(struct amdgpu_device *adev)
 
 static void gfxhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	unsigned num_level, block_size;
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	unsigned int num_level, block_size;
 	uint32_t tmp;
 	int i;
 
@@ -260,7 +262,7 @@ static void gfxhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
 		block_size -= 9;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(GC, 0, mmVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    num_level);
@@ -307,8 +309,8 @@ static void gfxhub_v1_0_setup_vmid_config(struct amdgpu_device *adev)
 
 static void gfxhub_v1_0_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	unsigned i;
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	unsigned int i;
 
 	for (i = 0 ; i < 18; ++i) {
 		WREG32_SOC15_OFFSET(GC, 0, mmVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
@@ -338,7 +340,7 @@ static int gfxhub_v1_0_gart_enable(struct amdgpu_device *adev)
 
 static void gfxhub_v1_0_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	u32 tmp;
 	u32 i;
 
@@ -375,6 +377,7 @@ static void gfxhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev,
 						 bool value)
 {
 	u32 tmp;
+
 	tmp = RREG32_SOC15(GC, 0, mmVM_L2_PROTECTION_FAULT_CNTL);
 	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
 			RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
@@ -411,7 +414,7 @@ static void gfxhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev,
 
 static void gfxhub_v1_0_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(GC, 0,
@@ -440,7 +443,6 @@ static void gfxhub_v1_0_init(struct amdgpu_device *adev)
 		mmVM_INVALIDATE_ENG0_ADDR_RANGE_LO32;
 }
 
-
 const struct amdgpu_gfxhub_funcs gfxhub_v1_0_funcs = {
 	.get_mc_fb_offset = gfxhub_v1_0_get_mc_fb_offset,
 	.setup_vm_pt_regs = gfxhub_v1_0_setup_vm_pt_regs,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
index c59c6c85fbff..6c03bf9f1ae8 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.c
@@ -21,6 +21,7 @@
  *
  */
 #include "amdgpu.h"
+#include "amdgpu_xcp.h"
 #include "gfxhub_v1_2.h"
 #include "gfxhub_v1_1.h"
 
@@ -35,227 +36,302 @@
 
 static u64 gfxhub_v1_2_get_mc_fb_offset(struct amdgpu_device *adev)
 {
-	return (u64)RREG32_SOC15(GC, 0, regMC_VM_FB_OFFSET) << 24;
+	return (u64)RREG32_SOC15(GC, GET_INST(GC, 0), regMC_VM_FB_OFFSET) << 24;
+}
+
+static void gfxhub_v1_2_xcc_setup_vm_pt_regs(struct amdgpu_device *adev,
+					     uint32_t vmid,
+					     uint64_t page_table_base,
+					     uint32_t xcc_mask)
+{
+	struct amdgpu_vmhub *hub;
+	int i;
+
+	for_each_inst(i, xcc_mask) {
+		hub = &adev->vmhub[AMDGPU_GFXHUB(i)];
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, i),
+				    regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
+				    hub->ctx_addr_distance * vmid,
+				    lower_32_bits(page_table_base));
+
+		WREG32_SOC15_OFFSET(GC, GET_INST(GC, i),
+				    regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
+				    hub->ctx_addr_distance * vmid,
+				    upper_32_bits(page_table_base));
+	}
 }
 
 static void gfxhub_v1_2_setup_vm_pt_regs(struct amdgpu_device *adev,
 					 uint32_t vmid,
 					 uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-
-	WREG32_SOC15_OFFSET(GC, 0, regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
-			    hub->ctx_addr_distance * vmid,
-			    lower_32_bits(page_table_base));
+	uint32_t xcc_mask;
 
-	WREG32_SOC15_OFFSET(GC, 0, regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32,
-			    hub->ctx_addr_distance * vmid,
-			    upper_32_bits(page_table_base));
+	xcc_mask = GENMASK(NUM_XCC(adev->gfx.xcc_mask) - 1, 0);
+	gfxhub_v1_2_xcc_setup_vm_pt_regs(adev, vmid, page_table_base, xcc_mask);
 }
 
-static void gfxhub_v1_2_init_gart_aperture_regs(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_init_gart_aperture_regs(struct amdgpu_device *adev,
+						    uint32_t xcc_mask)
 {
+	uint64_t gart_start = amdgpu_virt_xgmi_migrate_enabled(adev) ?
+			adev->gmc.vram_start : adev->gmc.fb_start;
 	uint64_t pt_base;
+	int i;
 
 	if (adev->gmc.pdb0_bo)
 		pt_base = amdgpu_gmc_pd_addr(adev->gmc.pdb0_bo);
 	else
 		pt_base = amdgpu_gmc_pd_addr(adev->gart.bo);
 
-	gfxhub_v1_2_setup_vm_pt_regs(adev, 0, pt_base);
+	gfxhub_v1_2_xcc_setup_vm_pt_regs(adev, 0, pt_base, xcc_mask);
 
 	/* If use GART for FB translation, vmid0 page table covers both
 	 * vram and system memory (gart)
 	 */
-	if (adev->gmc.pdb0_bo) {
-		WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
-				(u32)(adev->gmc.fb_start >> 12));
-		WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
-				(u32)(adev->gmc.fb_start >> 44));
-
-		WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
-				(u32)(adev->gmc.gart_end >> 12));
-		WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
-				(u32)(adev->gmc.gart_end >> 44));
-	} else {
-		WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
-				(u32)(adev->gmc.gart_start >> 12));
-		WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
-				(u32)(adev->gmc.gart_start >> 44));
-
-		WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
-				(u32)(adev->gmc.gart_end >> 12));
-		WREG32_SOC15(GC, 0, regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
-				(u32)(adev->gmc.gart_end >> 44));
+	for_each_inst(i, xcc_mask) {
+		if (adev->gmc.pdb0_bo) {
+			WREG32_SOC15(GC, GET_INST(GC, i),
+				     regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
+				     (u32)(gart_start >> 12));
+			WREG32_SOC15(GC, GET_INST(GC, i),
+				     regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
+				     (u32)(gart_start >> 44));
+
+			WREG32_SOC15(GC, GET_INST(GC, i),
+				     regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
+				     (u32)(adev->gmc.gart_end >> 12));
+			WREG32_SOC15(GC, GET_INST(GC, i),
+				     regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
+				     (u32)(adev->gmc.gart_end >> 44));
+		} else {
+			WREG32_SOC15(GC, GET_INST(GC, i),
+				     regVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32,
+				     (u32)(adev->gmc.gart_start >> 12));
+			WREG32_SOC15(GC, GET_INST(GC, i),
+				     regVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32,
+				     (u32)(adev->gmc.gart_start >> 44));
+
+			WREG32_SOC15(GC, GET_INST(GC, i),
+				     regVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32,
+				     (u32)(adev->gmc.gart_end >> 12));
+			WREG32_SOC15(GC, GET_INST(GC, i),
+				     regVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32,
+				     (u32)(adev->gmc.gart_end >> 44));
+		}
 	}
 }
 
-static void gfxhub_v1_2_init_system_aperture_regs(struct amdgpu_device *adev)
+static void
+gfxhub_v1_2_xcc_init_system_aperture_regs(struct amdgpu_device *adev,
+					  uint32_t xcc_mask)
 {
 	uint64_t value;
 	uint32_t tmp;
+	int i;
 
-	/* Program the AGP BAR */
-	WREG32_SOC15_RLC(GC, 0, regMC_VM_AGP_BASE, 0);
-	WREG32_SOC15_RLC(GC, 0, regMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
-	WREG32_SOC15_RLC(GC, 0, regMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
-
-	if (!amdgpu_sriov_vf(adev) || adev->asic_type <= CHIP_VEGA10) {
-		/* Program the system aperture low logical page number. */
-		WREG32_SOC15_RLC(GC, 0, regMC_VM_SYSTEM_APERTURE_LOW_ADDR,
-			min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
-
-		if (adev->apu_flags & AMD_APU_IS_RAVEN2)
-			/*
-			* Raven2 has a HW issue that it is unable to use the
-			* vram which is out of MC_VM_SYSTEM_APERTURE_HIGH_ADDR.
-			* So here is the workaround that increase system
-			* aperture high address (add 1) to get rid of the VM
-			* fault and hardware hang.
-			*/
-			WREG32_SOC15_RLC(GC, 0,
-					 regMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
-					 max((adev->gmc.fb_end >> 18) + 0x1,
-					     adev->gmc.agp_end >> 18));
-		else
-			WREG32_SOC15_RLC(GC, 0,
-				regMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
-				max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
-
-		/* Set default page address. */
-		value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
-		WREG32_SOC15(GC, 0, regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
-			     (u32)(value >> 12));
-		WREG32_SOC15(GC, 0, regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
-			     (u32)(value >> 44));
-
-		/* Program "protection fault". */
-		WREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32,
-			     (u32)(adev->dummy_page_addr >> 12));
-		WREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32,
-			     (u32)((u64)adev->dummy_page_addr >> 44));
-
-		tmp = RREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL2);
-		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL2,
-				    ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1);
-		WREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL2, tmp);
-	}
-
-	/* In the case squeezing vram into GART aperture, we don't use
-	 * FB aperture and AGP aperture. Disable them.
-	 */
-	if (adev->gmc.pdb0_bo) {
-		WREG32_SOC15(GC, 0, regMC_VM_FB_LOCATION_TOP, 0);
-		WREG32_SOC15(GC, 0, regMC_VM_FB_LOCATION_BASE, 0x00FFFFFF);
-		WREG32_SOC15(GC, 0, regMC_VM_AGP_TOP, 0);
-		WREG32_SOC15(GC, 0, regMC_VM_AGP_BOT, 0xFFFFFF);
-		WREG32_SOC15(GC, 0, regMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x3FFFFFFF);
-		WREG32_SOC15(GC, 0, regMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0);
+	for_each_inst(i, xcc_mask) {
+		/* Program the AGP BAR */
+		WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_AGP_BASE, 0);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_AGP_TOP, adev->gmc.agp_end >> 24);
+
+		if (!amdgpu_sriov_vf(adev) || adev->asic_type <= CHIP_VEGA10) {
+			/* Program the system aperture low logical page number. */
+			WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_SYSTEM_APERTURE_LOW_ADDR,
+				min(adev->gmc.fb_start, adev->gmc.agp_start) >> 18);
+
+			if (adev->apu_flags & (AMD_APU_IS_RAVEN2 |
+					       AMD_APU_IS_RENOIR |
+					       AMD_APU_IS_GREEN_SARDINE))
+			       /*
+				* Raven2 has a HW issue that it is unable to use the
+				* vram which is out of MC_VM_SYSTEM_APERTURE_HIGH_ADDR.
+				* So here is the workaround that increase system
+				* aperture high address (add 1) to get rid of the VM
+				* fault and hardware hang.
+				*/
+				WREG32_SOC15_RLC(GC, GET_INST(GC, i),
+						 regMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+						 max((adev->gmc.fb_end >> 18) + 0x1,
+						     adev->gmc.agp_end >> 18));
+			else
+				WREG32_SOC15_RLC(GC, GET_INST(GC, i),
+					regMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+					max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
+
+			/* Set default page address. */
+			value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
+			WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
+				     (u32)(value >> 12));
+			WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
+				     (u32)(value >> 44));
+
+			/* Program "protection fault". */
+			WREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_LO32,
+				     (u32)(adev->dummy_page_addr >> 12));
+			WREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_DEFAULT_ADDR_HI32,
+				     (u32)((u64)adev->dummy_page_addr >> 44));
+
+			tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_CNTL2);
+			tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL2,
+					    ACTIVE_PAGE_MIGRATION_PTE_READ_RETRY, 1);
+			WREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_CNTL2, tmp);
+		}
+
+		/* In the case squeezing vram into GART aperture, we don't use
+		 * FB aperture and AGP aperture. Disable them.
+		 */
+		if (adev->gmc.pdb0_bo && adev->gmc.xgmi.connected_to_cpu) {
+			WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_FB_LOCATION_TOP, 0);
+			WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_FB_LOCATION_BASE, 0x00FFFFFF);
+			WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_AGP_TOP, 0);
+			WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_AGP_BOT, 0xFFFFFF);
+			WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x3FFFFFFF);
+			WREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0);
+		}
 	}
 }
 
-static void gfxhub_v1_2_init_tlb_regs(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_init_tlb_regs(struct amdgpu_device *adev,
+					  uint32_t xcc_mask)
 {
 	uint32_t tmp;
+	int i;
 
-	/* Setup TLB control */
-	tmp = RREG32_SOC15(GC, 0, regMC_VM_MX_L1_TLB_CNTL);
-
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1);
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, 3);
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
-			    ENABLE_ADVANCED_DRIVER_MODEL, 1);
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
-			    SYSTEM_APERTURE_UNMAPPED_ACCESS, 0);
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
-			    MTYPE, MTYPE_UC);/* XXX for emulation. */
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ATC_EN, 1);
-
-	WREG32_SOC15_RLC(GC, 0, regMC_VM_MX_L1_TLB_CNTL, tmp);
+	for_each_inst(i, xcc_mask) {
+		/* Setup TLB control */
+		tmp = RREG32_SOC15(GC, GET_INST(GC, i), regMC_VM_MX_L1_TLB_CNTL);
+
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+				    ENABLE_L1_TLB, 1);
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+				    SYSTEM_ACCESS_MODE, 3);
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+				    ENABLE_ADVANCED_DRIVER_MODEL, 1);
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+				    SYSTEM_APERTURE_UNMAPPED_ACCESS, 0);
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL,
+				    MTYPE, MTYPE_UC);/* XXX for emulation. */
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ATC_EN, 1);
+
+		WREG32_SOC15_RLC(GC, GET_INST(GC, i), regMC_VM_MX_L1_TLB_CNTL, tmp);
+	}
 }
 
-static void gfxhub_v1_2_init_cache_regs(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_init_cache_regs(struct amdgpu_device *adev,
+					    uint32_t xcc_mask)
 {
 	uint32_t tmp;
+	int i;
 
-	/* Setup L2 cache */
-	tmp = RREG32_SOC15(GC, 0, regVM_L2_CNTL);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 1);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 1);
-	/* XXX for emulation, Refer to closed source code.*/
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, L2_PDE0_CACHE_TAG_GENERATION_MODE,
-			    0);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0);
-	WREG32_SOC15_RLC(GC, 0, regVM_L2_CNTL, tmp);
-
-	tmp = RREG32_SOC15(GC, 0, regVM_L2_CNTL2);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1);
-	WREG32_SOC15_RLC(GC, 0, regVM_L2_CNTL2, tmp);
-
-	tmp = regVM_L2_CNTL3_DEFAULT;
-	if (adev->gmc.translate_further) {
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 12);
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
-				    L2_CACHE_BIGK_FRAGMENT_SIZE, 9);
-	} else {
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 9);
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
-				    L2_CACHE_BIGK_FRAGMENT_SIZE, 6);
+	for_each_inst(i, xcc_mask) {
+		/* Setup L2 cache */
+		tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_CNTL);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 1);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 1);
+		/* XXX for emulation, Refer to closed source code.*/
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, L2_PDE0_CACHE_TAG_GENERATION_MODE,
+				    0);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, PDE_FAULT_CLASSIFICATION, 0);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, IDENTITY_MODE_FRAGMENT_SIZE, 0);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, i), regVM_L2_CNTL, tmp);
+
+		tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_CNTL2);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1);
+		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, i), regVM_L2_CNTL2, tmp);
+
+		tmp = regVM_L2_CNTL3_DEFAULT;
+		if (adev->gmc.translate_further) {
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 12);
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
+					    L2_CACHE_BIGK_FRAGMENT_SIZE, 9);
+		} else {
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 9);
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3,
+					    L2_CACHE_BIGK_FRAGMENT_SIZE, 6);
+		}
+		WREG32_SOC15_RLC(GC, GET_INST(GC, i), regVM_L2_CNTL3, tmp);
+
+		tmp = regVM_L2_CNTL4_DEFAULT;
+		/* For AMD APP APUs setup WC memory */
+		if (adev->gmc.xgmi.connected_to_cpu || adev->gmc.is_app_apu) {
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 1);
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 1);
+		} else {
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0);
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0);
+		}
+		WREG32_SOC15_RLC(GC, GET_INST(GC, i), regVM_L2_CNTL4, tmp);
 	}
-	WREG32_SOC15_RLC(GC, 0, regVM_L2_CNTL3, tmp);
-
-	tmp = regVM_L2_CNTL4_DEFAULT;
-	if (adev->gmc.xgmi.connected_to_cpu) {
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 1);
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 1);
-	} else {
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PDE_REQUEST_PHYSICAL, 0);
-		tmp = REG_SET_FIELD(tmp, VM_L2_CNTL4, VMC_TAP_PTE_REQUEST_PHYSICAL, 0);
-	}
-	WREG32_SOC15_RLC(GC, 0, regVM_L2_CNTL4, tmp);
 }
 
-static void gfxhub_v1_2_enable_system_domain(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_enable_system_domain(struct amdgpu_device *adev,
+						 uint32_t xcc_mask)
 {
 	uint32_t tmp;
+	int i;
 
-	tmp = RREG32_SOC15(GC, 0, regVM_CONTEXT0_CNTL);
-	tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
-	tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH,
-			adev->gmc.vmid0_page_table_depth);
-	tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_BLOCK_SIZE,
-			adev->gmc.vmid0_page_table_block_size);
-	tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL,
-			    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
-	WREG32_SOC15(GC, 0, regVM_CONTEXT0_CNTL, tmp);
+	for_each_inst(i, xcc_mask) {
+		tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_CONTEXT0_CNTL);
+		tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
+		tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH,
+				adev->gmc.vmid0_page_table_depth);
+		tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_BLOCK_SIZE,
+				adev->gmc.vmid0_page_table_block_size);
+		tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL,
+				    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT, 0);
+		WREG32_SOC15(GC, GET_INST(GC, i), regVM_CONTEXT0_CNTL, tmp);
+	}
 }
 
-static void gfxhub_v1_2_disable_identity_aperture(struct amdgpu_device *adev)
+static void
+gfxhub_v1_2_xcc_disable_identity_aperture(struct amdgpu_device *adev,
+					  uint32_t xcc_mask)
 {
-	WREG32_SOC15(GC, 0, regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32,
-		     0XFFFFFFFF);
-	WREG32_SOC15(GC, 0, regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32,
-		     0x0000000F);
-
-	WREG32_SOC15(GC, 0, regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32,
-		     0);
-	WREG32_SOC15(GC, 0, regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32,
-		     0);
+	int i;
 
-	WREG32_SOC15(GC, 0, regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, 0);
-	WREG32_SOC15(GC, 0, regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, 0);
+	for_each_inst(i, xcc_mask) {
+		WREG32_SOC15(GC, GET_INST(GC, i),
+			     regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_LO32,
+			     0XFFFFFFFF);
+		WREG32_SOC15(GC, GET_INST(GC, i),
+			     regVM_L2_CONTEXT1_IDENTITY_APERTURE_LOW_ADDR_HI32,
+			     0x0000000F);
+
+		WREG32_SOC15(GC, GET_INST(GC, i),
+			     regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_LO32,
+			     0);
+		WREG32_SOC15(GC, GET_INST(GC, i),
+			     regVM_L2_CONTEXT1_IDENTITY_APERTURE_HIGH_ADDR_HI32,
+			     0);
+
+		WREG32_SOC15(GC, GET_INST(GC, i),
+			     regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_LO32, 0);
+		WREG32_SOC15(GC, GET_INST(GC, i),
+			     regVM_L2_CONTEXT_IDENTITY_PHYSICAL_OFFSET_HI32, 0);
+	}
+}
 
+static inline bool
+gfxhub_v1_2_per_process_xnack_support(struct amdgpu_device *adev)
+{
+	/*
+	 * TODO: Check if this function is really needed, so far only 9.4.3
+	 * variants use GFXHUB 1.2
+	 */
+	return !!adev->aid_mask;
 }
 
-static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_setup_vmid_config(struct amdgpu_device *adev,
+					      uint32_t xcc_mask)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	unsigned num_level, block_size;
+	struct amdgpu_vmhub *hub;
+	unsigned int num_level, block_size;
 	uint32_t tmp;
-	int i;
+	int i, j;
 
 	num_level = adev->vm_manager.num_level;
 	block_size = adev->vm_manager.block_size;
@@ -264,124 +340,190 @@ static void gfxhub_v1_2_setup_vmid_config(struct amdgpu_device *adev)
 	else
 		block_size -= 9;
 
-	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(GC, 0, regVM_CONTEXT1_CNTL, i);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
-				    num_level);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT,
-				    1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    PAGE_TABLE_BLOCK_SIZE,
-				    block_size);
-		/* Send no-retry XNACK on fault to suppress VM fault storm.
-		 * On Aldebaran, XNACK can be enabled in the SQ per-process.
-		 * Retry faults need to be enabled for that to work.
-		 */
-		tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
-				    RETRY_PERMISSION_OR_INVALID_PAGE_FAULT,
-				    !adev->gmc.noretry ||
-				    adev->asic_type == CHIP_ALDEBARAN);
-		WREG32_SOC15_OFFSET(GC, 0, regVM_CONTEXT1_CNTL,
-				    i * hub->ctx_distance, tmp);
-		WREG32_SOC15_OFFSET(GC, 0,
-				    regVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32,
-				    i * hub->ctx_addr_distance, 0);
-		WREG32_SOC15_OFFSET(GC, 0,
-				    regVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32,
-				    i * hub->ctx_addr_distance, 0);
-		WREG32_SOC15_OFFSET(GC, 0,
-				    regVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32,
-				    i * hub->ctx_addr_distance,
-				    lower_32_bits(adev->vm_manager.max_pfn - 1));
-		WREG32_SOC15_OFFSET(GC, 0,
-				    regVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32,
-				    i * hub->ctx_addr_distance,
-				    upper_32_bits(adev->vm_manager.max_pfn - 1));
+	for_each_inst(j, xcc_mask) {
+		hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
+		for (i = 0; i <= 14; i++) {
+			tmp = RREG32_SOC15_OFFSET(GC, GET_INST(GC, j), regVM_CONTEXT1_CNTL,
+					i * hub->ctx_distance);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
+					    num_level);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT,
+					    1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL,
+					    PAGE_TABLE_BLOCK_SIZE,
+					    block_size);
+			/* Send no-retry XNACK on fault to suppress VM fault storm.
+			 * On 9.4.3 variants, XNACK can be enabled in
+			 * the SQ per-process.
+			 * Retry faults need to be enabled for that to work.
+			 */
+			tmp = REG_SET_FIELD(
+				tmp, VM_CONTEXT1_CNTL,
+				RETRY_PERMISSION_OR_INVALID_PAGE_FAULT,
+				!adev->gmc.noretry ||
+					gfxhub_v1_2_per_process_xnack_support(
+						adev));
+			WREG32_SOC15_OFFSET(GC, GET_INST(GC, j), regVM_CONTEXT1_CNTL,
+					    i * hub->ctx_distance, tmp);
+			WREG32_SOC15_OFFSET(GC, GET_INST(GC, j),
+					    regVM_CONTEXT1_PAGE_TABLE_START_ADDR_LO32,
+					    i * hub->ctx_addr_distance, 0);
+			WREG32_SOC15_OFFSET(GC, GET_INST(GC, j),
+					    regVM_CONTEXT1_PAGE_TABLE_START_ADDR_HI32,
+					    i * hub->ctx_addr_distance, 0);
+			WREG32_SOC15_OFFSET(GC, GET_INST(GC, j),
+					    regVM_CONTEXT1_PAGE_TABLE_END_ADDR_LO32,
+					    i * hub->ctx_addr_distance,
+					    lower_32_bits(adev->vm_manager.max_pfn - 1));
+			WREG32_SOC15_OFFSET(GC, GET_INST(GC, j),
+					    regVM_CONTEXT1_PAGE_TABLE_END_ADDR_HI32,
+					    i * hub->ctx_addr_distance,
+					    upper_32_bits(adev->vm_manager.max_pfn - 1));
+		}
 	}
 }
 
-static void gfxhub_v1_2_program_invalidation(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_program_invalidation(struct amdgpu_device *adev,
+						 uint32_t xcc_mask)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	unsigned i;
-
-	for (i = 0 ; i < 18; ++i) {
-		WREG32_SOC15_OFFSET(GC, 0, regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
-				    i * hub->eng_addr_distance, 0xffffffff);
-		WREG32_SOC15_OFFSET(GC, 0, regVM_INVALIDATE_ENG0_ADDR_RANGE_HI32,
-				    i * hub->eng_addr_distance, 0x1f);
+	struct amdgpu_vmhub *hub;
+	unsigned int i, j;
+
+	for_each_inst(j, xcc_mask) {
+		hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
+
+		for (i = 0 ; i < 18; ++i) {
+			WREG32_SOC15_OFFSET(GC, GET_INST(GC, j), regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
+					    i * hub->eng_addr_distance, 0xffffffff);
+			WREG32_SOC15_OFFSET(GC, GET_INST(GC, j), regVM_INVALIDATE_ENG0_ADDR_RANGE_HI32,
+					    i * hub->eng_addr_distance, 0x1f);
+		}
 	}
 }
 
-static int gfxhub_v1_2_gart_enable(struct amdgpu_device *adev)
+static int gfxhub_v1_2_xcc_gart_enable(struct amdgpu_device *adev,
+				       uint32_t xcc_mask)
 {
-	if (amdgpu_sriov_vf(adev) && adev->asic_type != CHIP_ARCTURUS) {
-		/*
-		 * MC_VM_FB_LOCATION_BASE/TOP is NULL for VF, becuase they are
-		 * VF copy registers so vbios post doesn't program them, for
-		 * SRIOV driver need to program them
-		 */
-		WREG32_SOC15_RLC(GC, 0, regMC_VM_FB_LOCATION_BASE,
-			     adev->gmc.vram_start >> 24);
-		WREG32_SOC15_RLC(GC, 0, regMC_VM_FB_LOCATION_TOP,
-			     adev->gmc.vram_end >> 24);
-	}
-
 	/* GART Enable. */
-	gfxhub_v1_2_init_gart_aperture_regs(adev);
-	gfxhub_v1_2_init_system_aperture_regs(adev);
-	gfxhub_v1_2_init_tlb_regs(adev);
+	gfxhub_v1_2_xcc_init_gart_aperture_regs(adev, xcc_mask);
+	gfxhub_v1_2_xcc_init_system_aperture_regs(adev, xcc_mask);
+	gfxhub_v1_2_xcc_init_tlb_regs(adev, xcc_mask);
 	if (!amdgpu_sriov_vf(adev))
-		gfxhub_v1_2_init_cache_regs(adev);
+		gfxhub_v1_2_xcc_init_cache_regs(adev, xcc_mask);
 
-	gfxhub_v1_2_enable_system_domain(adev);
+	gfxhub_v1_2_xcc_enable_system_domain(adev, xcc_mask);
 	if (!amdgpu_sriov_vf(adev))
-		gfxhub_v1_2_disable_identity_aperture(adev);
-	gfxhub_v1_2_setup_vmid_config(adev);
-	gfxhub_v1_2_program_invalidation(adev);
+		gfxhub_v1_2_xcc_disable_identity_aperture(adev, xcc_mask);
+	gfxhub_v1_2_xcc_setup_vmid_config(adev, xcc_mask);
+	gfxhub_v1_2_xcc_program_invalidation(adev, xcc_mask);
 
 	return 0;
 }
 
+static int gfxhub_v1_2_gart_enable(struct amdgpu_device *adev)
+{
+	uint32_t xcc_mask;
+
+	xcc_mask = GENMASK(NUM_XCC(adev->gfx.xcc_mask) - 1, 0);
+	return gfxhub_v1_2_xcc_gart_enable(adev, xcc_mask);
+}
+
+static void gfxhub_v1_2_xcc_gart_disable(struct amdgpu_device *adev,
+					 uint32_t xcc_mask)
+{
+	struct amdgpu_vmhub *hub;
+	u32 tmp;
+	u32 i, j;
+
+	for_each_inst(j, xcc_mask) {
+		hub = &adev->vmhub[AMDGPU_GFXHUB(j)];
+		/* Disable all tables */
+		for (i = 0; i < 16; i++)
+			WREG32_SOC15_OFFSET(GC, GET_INST(GC, j), regVM_CONTEXT0_CNTL,
+					    i * hub->ctx_distance, 0);
+
+		/* Setup TLB control */
+		tmp = RREG32_SOC15(GC, GET_INST(GC, j), regMC_VM_MX_L1_TLB_CNTL);
+		tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0);
+		tmp = REG_SET_FIELD(tmp,
+					MC_VM_MX_L1_TLB_CNTL,
+					ENABLE_ADVANCED_DRIVER_MODEL,
+					0);
+		WREG32_SOC15_RLC(GC, GET_INST(GC, j), regMC_VM_MX_L1_TLB_CNTL, tmp);
+
+		/* Setup L2 cache */
+		if (!amdgpu_sriov_vf(adev)) {
+			tmp = RREG32_SOC15(GC, GET_INST(GC, j), regVM_L2_CNTL);
+			tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 0);
+			WREG32_SOC15(GC, GET_INST(GC, j), regVM_L2_CNTL, tmp);
+			WREG32_SOC15(GC, GET_INST(GC, j), regVM_L2_CNTL3, 0);
+		}
+	}
+}
+
 static void gfxhub_v1_2_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	uint32_t xcc_mask;
+
+	xcc_mask = GENMASK(NUM_XCC(adev->gfx.xcc_mask) - 1, 0);
+	gfxhub_v1_2_xcc_gart_disable(adev, xcc_mask);
+}
+
+static void gfxhub_v1_2_xcc_set_fault_enable_default(struct amdgpu_device *adev,
+						     bool value,
+						     uint32_t xcc_mask)
+{
 	u32 tmp;
-	u32 i;
-
-	/* Disable all tables */
-	for (i = 0; i < 16; i++)
-		WREG32_SOC15_OFFSET(GC, 0, regVM_CONTEXT0_CNTL,
-				    i * hub->ctx_distance, 0);
-
-	/* Setup TLB control */
-	tmp = RREG32_SOC15(GC, 0, regMC_VM_MX_L1_TLB_CNTL);
-	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0);
-	tmp = REG_SET_FIELD(tmp,
-				MC_VM_MX_L1_TLB_CNTL,
-				ENABLE_ADVANCED_DRIVER_MODEL,
-				0);
-	WREG32_SOC15_RLC(GC, 0, regMC_VM_MX_L1_TLB_CNTL, tmp);
-
-	/* Setup L2 cache */
-	tmp = RREG32_SOC15(GC, 0, regVM_L2_CNTL);
-	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 0);
-	WREG32_SOC15(GC, 0, regVM_L2_CNTL, tmp);
-	WREG32_SOC15(GC, 0, regVM_L2_CNTL3, 0);
+	int i;
+
+	for_each_inst(i, xcc_mask) {
+		tmp = RREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_CNTL);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp,
+				VM_L2_PROTECTION_FAULT_CNTL,
+				TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT,
+				value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				READ_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+				EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
+		if (!value) {
+			tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+					CRASH_ON_NO_RETRY_FAULT, 1);
+			tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
+					CRASH_ON_RETRY_FAULT, 1);
+		}
+		WREG32_SOC15(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_CNTL, tmp);
+	}
 }
 
 /**
@@ -393,72 +535,100 @@ static void gfxhub_v1_2_gart_disable(struct amdgpu_device *adev)
 static void gfxhub_v1_2_set_fault_enable_default(struct amdgpu_device *adev,
 						 bool value)
 {
-	u32 tmp;
-	tmp = RREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			PDE1_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			PDE2_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp,
-			VM_L2_PROTECTION_FAULT_CNTL,
-			TRANSLATE_FURTHER_PROTECTION_FAULT_ENABLE_DEFAULT,
-			value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			NACK_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			VALID_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			READ_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-			EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
-	if (!value) {
-		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-				CRASH_ON_NO_RETRY_FAULT, 1);
-		tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL,
-				CRASH_ON_RETRY_FAULT, 1);
-	}
-	WREG32_SOC15(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL, tmp);
+	uint32_t xcc_mask;
+
+	xcc_mask = GENMASK(NUM_XCC(adev->gfx.xcc_mask) - 1, 0);
+	gfxhub_v1_2_xcc_set_fault_enable_default(adev, value, xcc_mask);
 }
 
-static void gfxhub_v1_2_init(struct amdgpu_device *adev)
+static void gfxhub_v1_2_xcc_init(struct amdgpu_device *adev, uint32_t xcc_mask)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub;
+	int i;
+
+	for_each_inst(i, xcc_mask) {
+		hub = &adev->vmhub[AMDGPU_GFXHUB(i)];
 
-	hub->ctx0_ptb_addr_lo32 =
-		SOC15_REG_OFFSET(GC, 0,
+		hub->ctx0_ptb_addr_lo32 =
+			SOC15_REG_OFFSET(GC, GET_INST(GC, i),
 				regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32);
-	hub->ctx0_ptb_addr_hi32 =
-		SOC15_REG_OFFSET(GC, 0,
+		hub->ctx0_ptb_addr_hi32 =
+			SOC15_REG_OFFSET(GC, GET_INST(GC, i),
 				regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32);
-	hub->vm_inv_eng0_sem =
-		SOC15_REG_OFFSET(GC, 0, regVM_INVALIDATE_ENG0_SEM);
-	hub->vm_inv_eng0_req =
-		SOC15_REG_OFFSET(GC, 0, regVM_INVALIDATE_ENG0_REQ);
-	hub->vm_inv_eng0_ack =
-		SOC15_REG_OFFSET(GC, 0, regVM_INVALIDATE_ENG0_ACK);
-	hub->vm_context0_cntl =
-		SOC15_REG_OFFSET(GC, 0, regVM_CONTEXT0_CNTL);
-	hub->vm_l2_pro_fault_status =
-		SOC15_REG_OFFSET(GC, 0, regVM_L2_PROTECTION_FAULT_STATUS);
-	hub->vm_l2_pro_fault_cntl =
-		SOC15_REG_OFFSET(GC, 0, regVM_L2_PROTECTION_FAULT_CNTL);
-
-	hub->ctx_distance = regVM_CONTEXT1_CNTL - regVM_CONTEXT0_CNTL;
-	hub->ctx_addr_distance = regVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 -
-		regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32;
-	hub->eng_distance = regVM_INVALIDATE_ENG1_REQ - regVM_INVALIDATE_ENG0_REQ;
-	hub->eng_addr_distance = regVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 -
-		regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32;
+		hub->vm_inv_eng0_sem =
+			SOC15_REG_OFFSET(GC, GET_INST(GC, i), regVM_INVALIDATE_ENG0_SEM);
+		hub->vm_inv_eng0_req =
+			SOC15_REG_OFFSET(GC, GET_INST(GC, i), regVM_INVALIDATE_ENG0_REQ);
+		hub->vm_inv_eng0_ack =
+			SOC15_REG_OFFSET(GC, GET_INST(GC, i), regVM_INVALIDATE_ENG0_ACK);
+		hub->vm_context0_cntl =
+			SOC15_REG_OFFSET(GC, GET_INST(GC, i), regVM_CONTEXT0_CNTL);
+		hub->vm_l2_pro_fault_status =
+			SOC15_REG_OFFSET(GC, GET_INST(GC, i),
+				regVM_L2_PROTECTION_FAULT_STATUS);
+		hub->vm_l2_pro_fault_cntl =
+			SOC15_REG_OFFSET(GC, GET_INST(GC, i), regVM_L2_PROTECTION_FAULT_CNTL);
+
+		hub->ctx_distance = regVM_CONTEXT1_CNTL -
+				regVM_CONTEXT0_CNTL;
+		hub->ctx_addr_distance =
+				regVM_CONTEXT1_PAGE_TABLE_BASE_ADDR_LO32 -
+				regVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32;
+		hub->eng_distance = regVM_INVALIDATE_ENG1_REQ -
+				regVM_INVALIDATE_ENG0_REQ;
+		hub->eng_addr_distance =
+				regVM_INVALIDATE_ENG1_ADDR_RANGE_LO32 -
+				regVM_INVALIDATE_ENG0_ADDR_RANGE_LO32;
+	}
 }
 
+static void gfxhub_v1_2_init(struct amdgpu_device *adev)
+{
+	uint32_t xcc_mask;
+
+	xcc_mask = GENMASK(NUM_XCC(adev->gfx.xcc_mask) - 1, 0);
+	gfxhub_v1_2_xcc_init(adev, xcc_mask);
+}
+
+static int gfxhub_v1_2_get_xgmi_info(struct amdgpu_device *adev)
+{
+	u32 max_num_physical_nodes;
+	u32 max_physical_node_id;
+	u32 xgmi_lfb_cntl;
+	u32 max_region;
+	u64 seg_size;
+
+	xgmi_lfb_cntl = RREG32_SOC15(GC, GET_INST(GC, 0), regMC_VM_XGMI_LFB_CNTL);
+	seg_size = REG_GET_FIELD(
+		RREG32_SOC15(GC, GET_INST(GC, 0), regMC_VM_XGMI_LFB_SIZE),
+		MC_VM_XGMI_LFB_SIZE, PF_LFB_SIZE) << 24;
+	max_region =
+		REG_GET_FIELD(xgmi_lfb_cntl, MC_VM_XGMI_LFB_CNTL, PF_MAX_REGION);
+
+
+
+	max_num_physical_nodes   = 8;
+	max_physical_node_id     = 7;
+
+	/* PF_MAX_REGION=0 means xgmi is disabled */
+	if (max_region || adev->gmc.xgmi.connected_to_cpu) {
+		adev->gmc.xgmi.num_physical_nodes = max_region + 1;
+
+		if (adev->gmc.xgmi.num_physical_nodes > max_num_physical_nodes)
+			return -EINVAL;
+
+		adev->gmc.xgmi.physical_node_id =
+			REG_GET_FIELD(xgmi_lfb_cntl, MC_VM_XGMI_LFB_CNTL,
+					PF_LFB_REGION);
+
+		if (adev->gmc.xgmi.physical_node_id > max_physical_node_id)
+			return -EINVAL;
+
+		adev->gmc.xgmi.node_segment_size = seg_size;
+	}
+
+	return 0;
+}
 
 const struct amdgpu_gfxhub_funcs gfxhub_v1_2_funcs = {
 	.get_mc_fb_offset = gfxhub_v1_2_get_mc_fb_offset,
@@ -467,5 +637,38 @@ const struct amdgpu_gfxhub_funcs gfxhub_v1_2_funcs = {
 	.gart_disable = gfxhub_v1_2_gart_disable,
 	.set_fault_enable_default = gfxhub_v1_2_set_fault_enable_default,
 	.init = gfxhub_v1_2_init,
-	.get_xgmi_info = gfxhub_v1_1_get_xgmi_info,
+	.get_xgmi_info = gfxhub_v1_2_get_xgmi_info,
+};
+
+static int gfxhub_v1_2_xcp_resume(void *handle, uint32_t inst_mask)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	bool value;
+
+	if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS)
+		value = false;
+	else
+		value = true;
+
+	gfxhub_v1_2_xcc_set_fault_enable_default(adev, value, inst_mask);
+
+	if (!amdgpu_sriov_vf(adev))
+		return gfxhub_v1_2_xcc_gart_enable(adev, inst_mask);
+
+	return 0;
+}
+
+static int gfxhub_v1_2_xcp_suspend(void *handle, uint32_t inst_mask)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+	if (!amdgpu_sriov_vf(adev))
+		gfxhub_v1_2_xcc_gart_disable(adev, inst_mask);
+
+	return 0;
+}
+
+struct amdgpu_xcp_ip_funcs gfxhub_v1_2_xcp_funcs = {
+	.suspend = &gfxhub_v1_2_xcp_suspend,
+	.resume = &gfxhub_v1_2_xcp_resume
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.h b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.h
index e2d508f5a7ee..997e9f90c990 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.h
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_2.h
@@ -26,4 +26,6 @@
 
 extern const struct amdgpu_gfxhub_funcs gfxhub_v1_2_funcs;
 
+extern struct amdgpu_xcp_ip_funcs gfxhub_v1_2_xcp_funcs;
+
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c
index 9b3a02527318..793faf62cb07 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_0.c
@@ -31,7 +31,7 @@
 
 #include "soc15_common.h"
 
-static const char *gfxhub_client_ids[] = {
+static const char * const gfxhub_client_ids[] = {
 	"CB/DB",
 	"Reserved",
 	"GE1",
@@ -120,7 +120,7 @@ static u64 gfxhub_v2_0_get_mc_fb_offset(struct amdgpu_device *adev)
 static void gfxhub_v2_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
 				uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			    hub->ctx_addr_distance * vmid,
@@ -282,12 +282,12 @@ static void gfxhub_v2_0_disable_identity_aperture(struct amdgpu_device *adev)
 
 static void gfxhub_v2_0_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	int i;
 	uint32_t tmp;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    adev->vm_manager.num_level);
@@ -331,8 +331,8 @@ static void gfxhub_v2_0_setup_vmid_config(struct amdgpu_device *adev)
 
 static void gfxhub_v2_0_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	unsigned i;
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	unsigned int i;
 
 	for (i = 0 ; i < 18; ++i) {
 		WREG32_SOC15_OFFSET(GC, 0, mmGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
@@ -360,7 +360,7 @@ static int gfxhub_v2_0_gart_enable(struct amdgpu_device *adev)
 
 static void gfxhub_v2_0_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	u32 tmp;
 	u32 i;
 
@@ -393,6 +393,7 @@ static void gfxhub_v2_0_set_fault_enable_default(struct amdgpu_device *adev,
 					  bool value)
 {
 	u32 tmp;
+
 	tmp = RREG32_SOC15(GC, 0, mmGCVM_L2_PROTECTION_FAULT_CNTL);
 	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
 			    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
@@ -433,7 +434,7 @@ static const struct amdgpu_vmhub_funcs gfxhub_v2_0_vmhub_funcs = {
 
 static void gfxhub_v2_0_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(GC, 0,
@@ -470,6 +471,9 @@ static void gfxhub_v2_0_init(struct amdgpu_device *adev)
 		GCVM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
 		GCVM_CONTEXT1_CNTL__EXECUTE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK;
 
+	/* TODO: This is only needed on some Navi 1x revisions */
+	hub->sdma_invalidation_workaround = true;
+
 	hub->vmhub_funcs = &gfxhub_v2_0_vmhub_funcs;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c
index 4aacbbec31e2..deb95fab02df 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v2_1.c
@@ -34,7 +34,7 @@
 #define mmGCUTCL2_HARVEST_BYPASS_GROUPS_YELLOW_CARP				0x16f8
 #define mmGCUTCL2_HARVEST_BYPASS_GROUPS_YELLOW_CARP_BASE_IDX	0
 
-static const char *gfxhub_client_ids[] = {
+static const char * const gfxhub_client_ids[] = {
 	"CB/DB",
 	"Reserved",
 	"GE1",
@@ -123,7 +123,7 @@ static u64 gfxhub_v2_1_get_mc_fb_offset(struct amdgpu_device *adev)
 static void gfxhub_v2_1_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
 				uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	WREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			    hub->ctx_addr_distance * vmid,
@@ -155,6 +155,9 @@ static void gfxhub_v2_1_init_system_aperture_regs(struct amdgpu_device *adev)
 {
 	uint64_t value;
 
+	if (amdgpu_sriov_vf(adev))
+		return;
+
 	/* Program the AGP BAR */
 	WREG32_SOC15(GC, 0, mmGCMC_VM_AGP_BASE, 0);
 	WREG32_SOC15(GC, 0, mmGCMC_VM_AGP_BOT, adev->gmc.agp_start >> 24);
@@ -291,12 +294,12 @@ static void gfxhub_v2_1_disable_identity_aperture(struct amdgpu_device *adev)
 
 static void gfxhub_v2_1_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	int i;
 	uint32_t tmp;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(GC, 0, mmGCVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    adev->vm_manager.num_level);
@@ -340,8 +343,8 @@ static void gfxhub_v2_1_setup_vmid_config(struct amdgpu_device *adev)
 
 static void gfxhub_v2_1_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	unsigned i;
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	unsigned int i;
 
 	for (i = 0 ; i < 18; ++i) {
 		WREG32_SOC15_OFFSET(GC, 0, mmGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
@@ -381,7 +384,7 @@ static int gfxhub_v2_1_gart_enable(struct amdgpu_device *adev)
 
 static void gfxhub_v2_1_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	u32 tmp;
 	u32 i;
 
@@ -462,7 +465,7 @@ static const struct amdgpu_vmhub_funcs gfxhub_v2_1_vmhub_funcs = {
 
 static void gfxhub_v2_1_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(GC, 0,
@@ -502,42 +505,6 @@ static void gfxhub_v2_1_init(struct amdgpu_device *adev)
 	hub->vmhub_funcs = &gfxhub_v2_1_vmhub_funcs;
 }
 
-static int gfxhub_v2_1_get_xgmi_info(struct amdgpu_device *adev)
-{
-	u32 xgmi_lfb_cntl = RREG32_SOC15(GC, 0, mmGCMC_VM_XGMI_LFB_CNTL);
-	u32 max_region =
-		REG_GET_FIELD(xgmi_lfb_cntl, GCMC_VM_XGMI_LFB_CNTL, PF_MAX_REGION);
-	u32 max_num_physical_nodes   = 0;
-	u32 max_physical_node_id     = 0;
-
-	switch (adev->ip_versions[XGMI_HWIP][0]) {
-	case IP_VERSION(4, 8, 0):
-		max_num_physical_nodes   = 4;
-		max_physical_node_id     = 3;
-		break;
-	default:
-		return -EINVAL;
-	}
-
-	/* PF_MAX_REGION=0 means xgmi is disabled */
-	if (max_region) {
-		adev->gmc.xgmi.num_physical_nodes = max_region + 1;
-		if (adev->gmc.xgmi.num_physical_nodes > max_num_physical_nodes)
-			return -EINVAL;
-
-		adev->gmc.xgmi.physical_node_id =
-			REG_GET_FIELD(xgmi_lfb_cntl, GCMC_VM_XGMI_LFB_CNTL, PF_LFB_REGION);
-		if (adev->gmc.xgmi.physical_node_id > max_physical_node_id)
-			return -EINVAL;
-
-		adev->gmc.xgmi.node_segment_size = REG_GET_FIELD(
-			RREG32_SOC15(GC, 0, mmGCMC_VM_XGMI_LFB_SIZE),
-			GCMC_VM_XGMI_LFB_SIZE, PF_LFB_SIZE) << 24;
-	}
-
-	return 0;
-}
-
 static void gfxhub_v2_1_utcl2_harvest(struct amdgpu_device *adev)
 {
 	int i;
@@ -548,7 +515,7 @@ static void gfxhub_v2_1_utcl2_harvest(struct amdgpu_device *adev)
 		adev->gfx.config.max_sh_per_se *
 		adev->gfx.config.max_shader_engines);
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 1):
 	case IP_VERSION(10, 3, 3):
 		/* Get SA disabled bitmap from eFuse setting */
@@ -582,6 +549,7 @@ static void gfxhub_v2_1_utcl2_harvest(struct amdgpu_device *adev)
 static void gfxhub_v2_1_save_regs(struct amdgpu_device *adev)
 {
 	int i;
+
 	adev->gmc.VM_L2_CNTL = RREG32_SOC15(GC, 0, mmGCVM_L2_CNTL);
 	adev->gmc.VM_L2_CNTL2 = RREG32_SOC15(GC, 0, mmGCVM_L2_CNTL2);
 	adev->gmc.VM_DUMMY_PAGE_FAULT_CNTL = RREG32_SOC15(GC, 0, mmGCVM_DUMMY_PAGE_FAULT_CNTL);
@@ -616,6 +584,7 @@ static void gfxhub_v2_1_save_regs(struct amdgpu_device *adev)
 static void gfxhub_v2_1_restore_regs(struct amdgpu_device *adev)
 {
 	int i;
+
 	WREG32_SOC15(GC, 0, mmGCVM_L2_CNTL, adev->gmc.VM_L2_CNTL);
 	WREG32_SOC15(GC, 0, mmGCVM_L2_CNTL2, adev->gmc.VM_L2_CNTL2);
 	WREG32_SOC15(GC, 0, mmGCVM_DUMMY_PAGE_FAULT_CNTL, adev->gmc.VM_DUMMY_PAGE_FAULT_CNTL);
@@ -651,7 +620,7 @@ static void gfxhub_v2_1_restore_regs(struct amdgpu_device *adev)
 
 static void gfxhub_v2_1_halt(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	int i;
 	uint32_t tmp;
 	int time = 1000;
@@ -679,9 +648,8 @@ static void gfxhub_v2_1_halt(struct amdgpu_device *adev)
 		tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS2);
 	}
 
-	if (!time) {
+	if (!time)
 		DRM_WARN("failed to wait for GRBM(EA) idle\n");
-	}
 }
 
 const struct amdgpu_gfxhub_funcs gfxhub_v2_1_funcs = {
@@ -692,7 +660,6 @@ const struct amdgpu_gfxhub_funcs gfxhub_v2_1_funcs = {
 	.gart_disable = gfxhub_v2_1_gart_disable,
 	.set_fault_enable_default = gfxhub_v2_1_set_fault_enable_default,
 	.init = gfxhub_v2_1_init,
-	.get_xgmi_info = gfxhub_v2_1_get_xgmi_info,
 	.utcl2_harvest = gfxhub_v2_1_utcl2_harvest,
 	.mode2_save_regs = gfxhub_v2_1_save_regs,
 	.mode2_restore_regs = gfxhub_v2_1_restore_regs,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c
index be0d0f47415e..abe30c8bd2ba 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0.c
@@ -30,7 +30,7 @@
 #include "navi10_enum.h"
 #include "soc15_common.h"
 
-static const char *gfxhub_client_ids[] = {
+static const char * const gfxhub_client_ids[] = {
 	"CB/DB",
 	"Reserved",
 	"GE1",
@@ -119,7 +119,7 @@ static u64 gfxhub_v3_0_get_mc_fb_offset(struct amdgpu_device *adev)
 static void gfxhub_v3_0_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
 				uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			    hub->ctx_addr_distance * vmid,
@@ -164,8 +164,7 @@ static void gfxhub_v3_0_init_system_aperture_regs(struct amdgpu_device *adev)
 		     max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
 
 	/* Set default page address. */
-	value = adev->mem_scratch.gpu_addr - adev->gmc.vram_start
-		+ adev->vm_manager.vram_base_offset;
+	value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
 	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
 		     (u32)(value >> 12));
 	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
@@ -290,12 +289,12 @@ static void gfxhub_v3_0_disable_identity_aperture(struct amdgpu_device *adev)
 
 static void gfxhub_v3_0_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	int i;
 	uint32_t tmp;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    adev->vm_manager.num_level);
@@ -339,8 +338,8 @@ static void gfxhub_v3_0_setup_vmid_config(struct amdgpu_device *adev)
 
 static void gfxhub_v3_0_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	unsigned i;
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	unsigned int i;
 
 	for (i = 0 ; i < 18; ++i) {
 		WREG32_SOC15_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
@@ -380,7 +379,7 @@ static int gfxhub_v3_0_gart_enable(struct amdgpu_device *adev)
 
 static void gfxhub_v3_0_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	u32 tmp;
 	u32 i;
 
@@ -417,34 +416,12 @@ static void gfxhub_v3_0_set_fault_enable_default(struct amdgpu_device *adev,
 	tmp = REG_SET_FIELD(tmp, CP_DEBUG, CPG_UTCL1_ERROR_HALT_DISABLE, 1);
 	WREG32_SOC15(GC, 0, regCP_DEBUG, tmp);
 
-	/**
-	 * Set GRBM_GFX_INDEX in broad cast mode
-	 * before programming GL1C_UTCL0_CNTL1 and SQG_CONFIG
-	 */
-	WREG32_SOC15(GC, 0, regGRBM_GFX_INDEX, regGRBM_GFX_INDEX_DEFAULT);
-
-	/**
-	 * Retry respond mode: RETRY
-	 * Error (no retry) respond mode: SUCCESS
-	 */
-	tmp = RREG32_SOC15(GC, 0, regGL1C_UTCL0_CNTL1);
-	tmp = REG_SET_FIELD(tmp, GL1C_UTCL0_CNTL1, RESP_MODE, 0);
-	tmp = REG_SET_FIELD(tmp, GL1C_UTCL0_CNTL1, RESP_FAULT_MODE, 0x2);
-	WREG32_SOC15(GC, 0, regGL1C_UTCL0_CNTL1, tmp);
-
 	/* These registers are not accessible to VF-SRIOV.
 	 * The PF will program them instead.
 	 */
 	if (amdgpu_sriov_vf(adev))
 		return;
 
-	/* Disable SQ XNACK interrupt for all VMIDs */
-	tmp = RREG32_SOC15(GC, 0, regSQG_CONFIG);
-	tmp = REG_SET_FIELD(tmp, SQG_CONFIG, XNACK_INTR_MASK,
-			    SQG_CONFIG__XNACK_INTR_MASK_MASK >>
-			    SQG_CONFIG__XNACK_INTR_MASK__SHIFT);
-	WREG32_SOC15(GC, 0, regSQG_CONFIG, tmp);
-
 	tmp = RREG32_SOC15(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL);
 	tmp = REG_SET_FIELD(tmp, GCVM_L2_PROTECTION_FAULT_CNTL,
 			    RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, value);
@@ -485,7 +462,7 @@ static const struct amdgpu_vmhub_funcs gfxhub_v3_0_vmhub_funcs = {
 
 static void gfxhub_v3_0_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(GC, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c
index 6e0bd628c889..b3ef6e71811f 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v3_0_3.c
@@ -33,7 +33,7 @@
 #define regGCVM_L2_CNTL4_DEFAULT		0x000000c1
 #define regGCVM_L2_CNTL5_DEFAULT		0x00003fe0
 
-static const char *gfxhub_client_ids[] = {
+static const char * const gfxhub_client_ids[] = {
 	"CB/DB",
 	"Reserved",
 	"GE1",
@@ -122,7 +122,7 @@ static u64 gfxhub_v3_0_3_get_mc_fb_offset(struct amdgpu_device *adev)
 static void gfxhub_v3_0_3_setup_vm_pt_regs(struct amdgpu_device *adev, uint32_t vmid,
 				uint64_t page_table_base)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	WREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32,
 			    hub->ctx_addr_distance * vmid,
@@ -169,8 +169,7 @@ static void gfxhub_v3_0_3_init_system_aperture_regs(struct amdgpu_device *adev)
 		     max(adev->gmc.fb_end, adev->gmc.agp_end) >> 18);
 
 	/* Set default page address. */
-	value = adev->mem_scratch.gpu_addr - adev->gmc.vram_start
-		+ adev->vm_manager.vram_base_offset;
+	value = amdgpu_gmc_vram_mc2pa(adev, adev->mem_scratch.gpu_addr);
 	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB,
 		     (u32)(value >> 12));
 	WREG32_SOC15(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB,
@@ -295,12 +294,12 @@ static void gfxhub_v3_0_3_disable_identity_aperture(struct amdgpu_device *adev)
 
 static void gfxhub_v3_0_3_setup_vmid_config(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	int i;
 	uint32_t tmp;
 
 	for (i = 0; i <= 14; i++) {
-		tmp = RREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL, i);
+		tmp = RREG32_SOC15_OFFSET(GC, 0, regGCVM_CONTEXT1_CNTL, i * hub->ctx_distance);
 		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
 		tmp = REG_SET_FIELD(tmp, GCVM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH,
 				    adev->vm_manager.num_level);
@@ -344,8 +343,8 @@ static void gfxhub_v3_0_3_setup_vmid_config(struct amdgpu_device *adev)
 
 static void gfxhub_v3_0_3_program_invalidation(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
-	unsigned i;
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+	unsigned int i;
 
 	for (i = 0 ; i < 18; ++i) {
 		WREG32_SOC15_OFFSET(GC, 0, regGCVM_INVALIDATE_ENG0_ADDR_RANGE_LO32,
@@ -373,7 +372,7 @@ static int gfxhub_v3_0_3_gart_enable(struct amdgpu_device *adev)
 
 static void gfxhub_v3_0_3_gart_disable(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 	u32 tmp;
 	u32 i;
 
@@ -451,7 +450,7 @@ static const struct amdgpu_vmhub_funcs gfxhub_v3_0_3_vmhub_funcs = {
 
 static void gfxhub_v3_0_3_init(struct amdgpu_device *adev)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
 
 	hub->ctx0_ptb_addr_lo32 =
 		SOC15_REG_OFFSET(GC, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
index d99821692ba3..ce6e04242c52 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
@@ -51,18 +51,9 @@
 #include "athub_v2_0.h"
 #include "athub_v2_1.h"
 
-#include "amdgpu_reset.h"
-
-#if 0
-static const struct soc15_reg_golden golden_settings_navi10_hdp[] =
-{
-	/* TODO add golden setting for hdp */
-};
-#endif
-
 static int gmc_v10_0_ecc_interrupt_state(struct amdgpu_device *adev,
 					 struct amdgpu_irq_src *src,
-					 unsigned type,
+					 unsigned int type,
 					 enum amdgpu_interrupt_state state)
 {
 	return 0;
@@ -70,13 +61,13 @@ static int gmc_v10_0_ecc_interrupt_state(struct amdgpu_device *adev,
 
 static int
 gmc_v10_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
-				   struct amdgpu_irq_src *src, unsigned type,
+				   struct amdgpu_irq_src *src, unsigned int type,
 				   enum amdgpu_interrupt_state state)
 {
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
 		/* MM HUB */
-		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, false);
+		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), false);
 		/* GFX HUB */
 		/* This works because this interrupt is only
 		 * enabled at init/resume and disabled in
@@ -84,11 +75,11 @@ gmc_v10_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 		 * change over the course of suspend/resume.
 		 */
 		if (!adev->in_s0ix)
-			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, false);
+			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), false);
 		break;
 	case AMDGPU_IRQ_STATE_ENABLE:
 		/* MM HUB */
-		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, true);
+		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), true);
 		/* GFX HUB */
 		/* This works because this interrupt is only
 		 * enabled at init/resume and disabled in
@@ -96,7 +87,7 @@ gmc_v10_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 		 * change over the course of suspend/resume.
 		 */
 		if (!adev->in_s0ix)
-			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, true);
+			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), true);
 		break;
 	default:
 		break;
@@ -109,10 +100,14 @@ static int gmc_v10_0_process_interrupt(struct amdgpu_device *adev,
 				       struct amdgpu_irq_src *source,
 				       struct amdgpu_iv_entry *entry)
 {
-	bool retry_fault = !!(entry->src_data[1] & 0x80);
-	bool write_fault = !!(entry->src_data[1] & 0x20);
-	struct amdgpu_vmhub *hub = &adev->vmhub[entry->vmid_src];
-	struct amdgpu_task_info task_info;
+	uint32_t vmhub_index = entry->client_id == SOC15_IH_CLIENTID_VMC ?
+			       AMDGPU_MMHUB0(0) : AMDGPU_GFXHUB(0);
+	struct amdgpu_vmhub *hub = &adev->vmhub[vmhub_index];
+	bool retry_fault = !!(entry->src_data[1] &
+			      AMDGPU_GMC9_FAULT_SOURCE_DATA_RETRY);
+	bool write_fault = !!(entry->src_data[1] &
+			      AMDGPU_GMC9_FAULT_SOURCE_DATA_WRITE);
+	struct amdgpu_task_info *task_info;
 	uint32_t status = 0;
 	u64 addr;
 
@@ -139,7 +134,8 @@ static int gmc_v10_0_process_interrupt(struct amdgpu_device *adev,
 		/* Try to handle the recoverable page faults by filling page
 		 * tables
 		 */
-		if (amdgpu_vm_handle_fault(adev, entry->pasid, addr, write_fault))
+		if (amdgpu_vm_handle_fault(adev, entry->pasid, 0, 0, addr,
+					   entry->timestamp, write_fault))
 			return 1;
 	}
 
@@ -149,32 +145,39 @@ static int gmc_v10_0_process_interrupt(struct amdgpu_device *adev,
 		 * be updated to avoid reading an incorrect value due to
 		 * the new fast GRBM interface.
 		 */
-		if ((entry->vmid_src == AMDGPU_GFXHUB_0) &&
-		    (adev->ip_versions[GC_HWIP][0] < IP_VERSION(10, 3, 0)))
+		if ((entry->vmid_src == AMDGPU_GFXHUB(0)) &&
+		    (amdgpu_ip_version(adev, GC_HWIP, 0) <
+		     IP_VERSION(10, 3, 0)))
 			RREG32(hub->vm_l2_pro_fault_status);
 
 		status = RREG32(hub->vm_l2_pro_fault_status);
 		WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1);
+
+		amdgpu_vm_update_fault_cache(adev, entry->pasid, addr, status,
+					     entry->vmid_src ? AMDGPU_MMHUB0(0) : AMDGPU_GFXHUB(0));
 	}
 
 	if (!printk_ratelimit())
 		return 0;
 
-	memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-	amdgpu_vm_get_task_info(adev, entry->pasid, &task_info);
-
 	dev_err(adev->dev,
-		"[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u, "
-		"for process %s pid %d thread %s pid %d)\n",
+		"[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u)\n",
 		entry->vmid_src ? "mmhub" : "gfxhub",
-		entry->src_id, entry->ring_id, entry->vmid,
-		entry->pasid, task_info.process_name, task_info.tgid,
-		task_info.task_name, task_info.pid);
+		entry->src_id, entry->ring_id, entry->vmid, entry->pasid);
+	task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+	if (task_info) {
+		amdgpu_vm_print_task_info(adev, task_info);
+		amdgpu_vm_put_task_info(task_info);
+	}
+
 	dev_err(adev->dev, "  in page starting at address 0x%016llx from client 0x%x (%s)\n",
-		addr, entry->client_id,
-		soc15_ih_clientid_name[entry->client_id]);
+			addr, entry->client_id,
+			soc15_ih_clientid_name[entry->client_id]);
 
-	if (!amdgpu_sriov_vf(adev))
+	/* Only print L2 fault status if the status register could be read and
+	 * contains useful information
+	 */
+	if (status != 0)
 		hub->vmhub_funcs->print_l2_protection_fault_status(adev,
 								   status);
 
@@ -212,8 +215,7 @@ static void gmc_v10_0_set_irq_funcs(struct amdgpu_device *adev)
 static bool gmc_v10_0_use_invalidate_semaphore(struct amdgpu_device *adev,
 				       uint32_t vmhub)
 {
-	return ((vmhub == AMDGPU_MMHUB_0 ||
-		 vmhub == AMDGPU_MMHUB_1) &&
+	return ((vmhub == AMDGPU_MMHUB0(0)) &&
 		(!amdgpu_sriov_vf(adev)));
 }
 
@@ -237,20 +239,48 @@ static bool gmc_v10_0_get_atc_vmid_pasid_mapping_info(
  * by the amdgpu vm/hsa code.
  */
 
-static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
-				   unsigned int vmhub, uint32_t flush_type)
+/**
+ * gmc_v10_0_flush_gpu_tlb - gart tlb flush callback
+ *
+ * @adev: amdgpu_device pointer
+ * @vmid: vm instance to flush
+ * @vmhub: vmhub type
+ * @flush_type: the flush type
+ *
+ * Flush the TLB for the requested page table.
+ */
+static void gmc_v10_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
+					uint32_t vmhub, uint32_t flush_type)
 {
 	bool use_semaphore = gmc_v10_0_use_invalidate_semaphore(adev, vmhub);
 	struct amdgpu_vmhub *hub = &adev->vmhub[vmhub];
 	u32 inv_req = hub->vmhub_funcs->get_invalidate_req(vmid, flush_type);
-	u32 tmp;
 	/* Use register 17 for GART */
-	const unsigned eng = 17;
-	unsigned int i;
+	const unsigned int eng = 17;
 	unsigned char hub_ip = 0;
+	u32 sem, req, ack;
+	unsigned int i;
+	u32 tmp;
+
+	sem = hub->vm_inv_eng0_sem + hub->eng_distance * eng;
+	req = hub->vm_inv_eng0_req + hub->eng_distance * eng;
+	ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;
+
+	/* flush hdp cache */
+	amdgpu_device_flush_hdp(adev, NULL);
+
+	/* This is necessary for SRIOV as well as for GFXOFF to function
+	 * properly under bare metal
+	 */
+	if (adev->gfx.kiq[0].ring.sched.ready && !adev->enable_mes &&
+	    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) {
+		amdgpu_gmc_fw_reg_write_reg_wait(adev, req, ack, inv_req,
+						 1 << vmid, GET_INST(GC, 0));
+		return;
+	}
 
-	hub_ip = (vmhub == AMDGPU_GFXHUB_0) ?
-		   GC_HWIP : MMHUB_HWIP;
+	/* This path is needed before KIQ/MES/GFXOFF are set up */
+	hub_ip = (vmhub == AMDGPU_GFXHUB(0)) ? GC_HWIP : MMHUB_HWIP;
 
 	spin_lock(&adev->gmc.invalidate_lock);
 	/*
@@ -264,9 +294,7 @@ static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 	if (use_semaphore) {
 		for (i = 0; i < adev->usec_timeout; i++) {
 			/* a read return value of 1 means semaphore acuqire */
-			tmp = RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_sem +
-					 hub->eng_distance * eng, hub_ip);
-
+			tmp = RREG32_RLC_NO_KIQ(sem, hub_ip);
 			if (tmp & 0x1)
 				break;
 			udelay(1);
@@ -276,24 +304,19 @@ static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 			DRM_ERROR("Timeout waiting for sem acquire in VM flush!\n");
 	}
 
-	WREG32_RLC_NO_KIQ(hub->vm_inv_eng0_req +
-			  hub->eng_distance * eng,
-			  inv_req, hub_ip);
+	WREG32_RLC_NO_KIQ(req, inv_req, hub_ip);
 
 	/*
 	 * Issue a dummy read to wait for the ACK register to be cleared
 	 * to avoid a false ACK due to the new fast GRBM interface.
 	 */
-	if ((vmhub == AMDGPU_GFXHUB_0) &&
-	    (adev->ip_versions[GC_HWIP][0] < IP_VERSION(10, 3, 0)))
-		RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_req +
-				  hub->eng_distance * eng, hub_ip);
+	if ((vmhub == AMDGPU_GFXHUB(0)) &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(10, 3, 0)))
+		RREG32_RLC_NO_KIQ(req, hub_ip);
 
 	/* Wait for ACK with a delay.*/
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_ack +
-				  hub->eng_distance * eng, hub_ip);
-
+		tmp = RREG32_RLC_NO_KIQ(ack, hub_ip);
 		tmp &= 1 << vmid;
 		if (tmp)
 			break;
@@ -303,109 +326,13 @@ static void gmc_v10_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 
 	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
 	if (use_semaphore)
-		/*
-		 * add semaphore release after invalidation,
-		 * write with 0 means semaphore release
-		 */
-		WREG32_RLC_NO_KIQ(hub->vm_inv_eng0_sem +
-				  hub->eng_distance * eng, 0, hub_ip);
+		WREG32_RLC_NO_KIQ(sem, 0, hub_ip);
 
 	spin_unlock(&adev->gmc.invalidate_lock);
 
-	if (i < adev->usec_timeout)
-		return;
-
-	DRM_ERROR("Timeout waiting for VM flush hub: %d!\n", vmhub);
-}
-
-/**
- * gmc_v10_0_flush_gpu_tlb - gart tlb flush callback
- *
- * @adev: amdgpu_device pointer
- * @vmid: vm instance to flush
- * @vmhub: vmhub type
- * @flush_type: the flush type
- *
- * Flush the TLB for the requested page table.
- */
-static void gmc_v10_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
-					uint32_t vmhub, uint32_t flush_type)
-{
-	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
-	struct dma_fence *fence;
-	struct amdgpu_job *job;
-
-	int r;
-
-	/* flush hdp cache */
-	adev->hdp.funcs->flush_hdp(adev, NULL);
-
-	/* For SRIOV run time, driver shouldn't access the register through MMIO
-	 * Directly use kiq to do the vm invalidation instead
-	 */
-	if (adev->gfx.kiq.ring.sched.ready && !adev->enable_mes &&
-	    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev)) &&
-	    down_read_trylock(&adev->reset_domain->sem)) {
-		struct amdgpu_vmhub *hub = &adev->vmhub[vmhub];
-		const unsigned eng = 17;
-		u32 inv_req = hub->vmhub_funcs->get_invalidate_req(vmid, flush_type);
-		u32 req = hub->vm_inv_eng0_req + hub->eng_distance * eng;
-		u32 ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;
-
-		amdgpu_virt_kiq_reg_write_reg_wait(adev, req, ack, inv_req,
-				1 << vmid);
-
-		up_read(&adev->reset_domain->sem);
-		return;
-	}
-
-	mutex_lock(&adev->mman.gtt_window_lock);
-
-	if (vmhub == AMDGPU_MMHUB_0) {
-		gmc_v10_0_flush_vm_hub(adev, vmid, AMDGPU_MMHUB_0, 0);
-		mutex_unlock(&adev->mman.gtt_window_lock);
-		return;
-	}
-
-	BUG_ON(vmhub != AMDGPU_GFXHUB_0);
-
-	if (!adev->mman.buffer_funcs_enabled ||
-	    !adev->ib_pool_ready ||
-	    amdgpu_in_reset(adev) ||
-	    ring->sched.ready == false) {
-		gmc_v10_0_flush_vm_hub(adev, vmid, AMDGPU_GFXHUB_0, 0);
-		mutex_unlock(&adev->mman.gtt_window_lock);
-		return;
-	}
-
-	/* The SDMA on Navi has a bug which can theoretically result in memory
-	 * corruption if an invalidation happens at the same time as an VA
-	 * translation. Avoid this by doing the invalidation from the SDMA
-	 * itself.
-	 */
-	r = amdgpu_job_alloc_with_ib(ring->adev, &adev->mman.entity,
-				     AMDGPU_FENCE_OWNER_UNDEFINED,
-				     16 * 4, AMDGPU_IB_POOL_IMMEDIATE,
-				     &job);
-	if (r)
-		goto error_alloc;
-
-	job->vm_pd_addr = amdgpu_gmc_pd_addr(adev->gart.bo);
-	job->vm_needs_flush = true;
-	job->ibs->ptr[job->ibs->length_dw++] = ring->funcs->nop;
-	amdgpu_ring_pad_ib(ring, &job->ibs[0]);
-	fence = amdgpu_job_submit(job);
-
-	mutex_unlock(&adev->mman.gtt_window_lock);
-
-	dma_fence_wait(fence, false);
-	dma_fence_put(fence);
-
-	return;
-
-error_alloc:
-	mutex_unlock(&adev->mman.gtt_window_lock);
-	DRM_ERROR("Error flushing GPU TLB using the SDMA (%d)!\n", r);
+	if (i >= adev->usec_timeout)
+		dev_err(adev->dev, "Timeout waiting for VM flush hub: %d!\n",
+			vmhub);
 }
 
 /**
@@ -415,74 +342,44 @@ error_alloc:
  * @pasid: pasid to be flush
  * @flush_type: the flush type
  * @all_hub: Used with PACKET3_INVALIDATE_TLBS_ALL_HUB()
+ * @inst: is used to select which instance of KIQ to use for the invalidation
  *
  * Flush the TLB for the requested pasid.
  */
-static int gmc_v10_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
-					uint16_t pasid, uint32_t flush_type,
-					bool all_hub)
+static void gmc_v10_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
+					  uint16_t pasid, uint32_t flush_type,
+					  bool all_hub, uint32_t inst)
 {
+	uint16_t queried;
 	int vmid, i;
-	signed long r;
-	uint32_t seq;
-	uint16_t queried_pasid;
-	bool ret;
-	u32 usec_timeout = amdgpu_sriov_vf(adev) ? SRIOV_USEC_TIMEOUT : adev->usec_timeout;
-	struct amdgpu_ring *ring = &adev->gfx.kiq.ring;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-
-	if (amdgpu_emu_mode == 0 && ring->sched.ready) {
-		spin_lock(&adev->gfx.kiq.ring_lock);
-		/* 2 dwords flush + 8 dwords fence */
-		amdgpu_ring_alloc(ring, kiq->pmf->invalidate_tlbs_size + 8);
-		kiq->pmf->kiq_invalidate_tlbs(ring,
-					pasid, flush_type, all_hub);
-		r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
-		if (r) {
-			amdgpu_ring_undo(ring);
-			spin_unlock(&adev->gfx.kiq.ring_lock);
-			return -ETIME;
-		}
-
-		amdgpu_ring_commit(ring);
-		spin_unlock(&adev->gfx.kiq.ring_lock);
-		r = amdgpu_fence_wait_polling(ring, seq, usec_timeout);
-		if (r < 1) {
-			dev_err(adev->dev, "wait for kiq fence error: %ld.\n", r);
-			return -ETIME;
-		}
-
-		return 0;
-	}
 
 	for (vmid = 1; vmid < AMDGPU_NUM_VMID; vmid++) {
-
-		ret = gmc_v10_0_get_atc_vmid_pasid_mapping_info(adev, vmid,
-				&queried_pasid);
-		if (ret	&& queried_pasid == pasid) {
-			if (all_hub) {
-				for (i = 0; i < adev->num_vmhubs; i++)
-					gmc_v10_0_flush_gpu_tlb(adev, vmid,
-							i, flush_type);
-			} else {
-				gmc_v10_0_flush_gpu_tlb(adev, vmid,
-						AMDGPU_GFXHUB_0, flush_type);
-			}
-			if (!adev->enable_mes)
-				break;
+		bool valid;
+
+		valid = gmc_v10_0_get_atc_vmid_pasid_mapping_info(adev, vmid,
+								  &queried);
+		if (!valid || queried != pasid)
+			continue;
+
+		if (all_hub) {
+			for_each_set_bit(i, adev->vmhubs_mask,
+					 AMDGPU_MAX_VMHUBS)
+				gmc_v10_0_flush_gpu_tlb(adev, vmid, i,
+							flush_type);
+		} else {
+			gmc_v10_0_flush_gpu_tlb(adev, vmid, AMDGPU_GFXHUB(0),
+						flush_type);
 		}
 	}
-
-	return 0;
 }
 
 static uint64_t gmc_v10_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
-					     unsigned vmid, uint64_t pd_addr)
+					     unsigned int vmid, uint64_t pd_addr)
 {
-	bool use_semaphore = gmc_v10_0_use_invalidate_semaphore(ring->adev, ring->funcs->vmhub);
-	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
+	bool use_semaphore = gmc_v10_0_use_invalidate_semaphore(ring->adev, ring->vm_hub);
+	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
 	uint32_t req = hub->vmhub_funcs->get_invalidate_req(vmid, 0);
-	unsigned eng = ring->vm_inv_eng;
+	unsigned int eng = ring->vm_inv_eng;
 
 	/*
 	 * It may lose gpuvm invalidate acknowldege state across power-gating
@@ -524,17 +421,13 @@ static uint64_t gmc_v10_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 	return pd_addr;
 }
 
-static void gmc_v10_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
-					 unsigned pasid)
+static void gmc_v10_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned int vmid,
+					 unsigned int pasid)
 {
 	struct amdgpu_device *adev = ring->adev;
 	uint32_t reg;
 
-	/* MES fw manages IH_VMID_x_LUT updating */
-	if (ring->is_mes_queue)
-		return;
-
-	if (ring->funcs->vmhub == AMDGPU_GFXHUB_0)
+	if (ring->vm_hub == AMDGPU_GFXHUB(0))
 		reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT) + vmid;
 	else
 		reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT_MM) + vmid;
@@ -575,24 +468,6 @@ static void gmc_v10_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid
  * 0 valid
  */
 
-static uint64_t gmc_v10_0_map_mtype(struct amdgpu_device *adev, uint32_t flags)
-{
-	switch (flags) {
-	case AMDGPU_VM_MTYPE_DEFAULT:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_NC);
-	case AMDGPU_VM_MTYPE_NC:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_NC);
-	case AMDGPU_VM_MTYPE_WC:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_WC);
-	case AMDGPU_VM_MTYPE_CC:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_CC);
-	case AMDGPU_VM_MTYPE_UC:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_UC);
-	default:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_NC);
-	}
-}
-
 static void gmc_v10_0_get_vm_pde(struct amdgpu_device *adev, int level,
 				 uint64_t *addr, uint64_t *flags)
 {
@@ -617,21 +492,39 @@ static void gmc_v10_0_get_vm_pde(struct amdgpu_device *adev, int level,
 }
 
 static void gmc_v10_0_get_vm_pte(struct amdgpu_device *adev,
-				 struct amdgpu_bo_va_mapping *mapping,
+				 struct amdgpu_vm *vm,
+				 struct amdgpu_bo *bo,
+				 uint32_t vm_flags,
 				 uint64_t *flags)
 {
-	struct amdgpu_bo *bo = mapping->bo_va->base.bo;
-
-	*flags &= ~AMDGPU_PTE_EXECUTABLE;
-	*flags |= mapping->flags & AMDGPU_PTE_EXECUTABLE;
+	if (vm_flags & AMDGPU_VM_PAGE_EXECUTABLE)
+		*flags |= AMDGPU_PTE_EXECUTABLE;
+	else
+		*flags &= ~AMDGPU_PTE_EXECUTABLE;
 
-	*flags &= ~AMDGPU_PTE_MTYPE_NV10_MASK;
-	*flags |= (mapping->flags & AMDGPU_PTE_MTYPE_NV10_MASK);
+	switch (vm_flags & AMDGPU_VM_MTYPE_MASK) {
+	case AMDGPU_VM_MTYPE_DEFAULT:
+	case AMDGPU_VM_MTYPE_NC:
+	default:
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_NC);
+		break;
+	case AMDGPU_VM_MTYPE_WC:
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_WC);
+		break;
+	case AMDGPU_VM_MTYPE_CC:
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_CC);
+		break;
+	case AMDGPU_VM_MTYPE_UC:
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_UC);
+		break;
+	}
 
-	*flags &= ~AMDGPU_PTE_NOALLOC;
-	*flags |= (mapping->flags & AMDGPU_PTE_NOALLOC);
+	if (vm_flags & AMDGPU_VM_PAGE_NOALLOC)
+		*flags |= AMDGPU_PTE_NOALLOC;
+	else
+		*flags &= ~AMDGPU_PTE_NOALLOC;
 
-	if (mapping->flags & AMDGPU_PTE_PRT) {
+	if (vm_flags & AMDGPU_VM_PAGE_PRT) {
 		*flags |= AMDGPU_PTE_PRT;
 		*flags |= AMDGPU_PTE_SNOOPED;
 		*flags |= AMDGPU_PTE_LOG;
@@ -640,15 +533,15 @@ static void gmc_v10_0_get_vm_pte(struct amdgpu_device *adev,
 	}
 
 	if (bo && bo->flags & (AMDGPU_GEM_CREATE_COHERENT |
+			       AMDGPU_GEM_CREATE_EXT_COHERENT |
 			       AMDGPU_GEM_CREATE_UNCACHED))
-		*flags = (*flags & ~AMDGPU_PTE_MTYPE_NV10_MASK) |
-			 AMDGPU_PTE_MTYPE_NV10(MTYPE_UC);
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_UC);
 }
 
-static unsigned gmc_v10_0_get_vbios_fb_size(struct amdgpu_device *adev)
+static unsigned int gmc_v10_0_get_vbios_fb_size(struct amdgpu_device *adev)
 {
 	u32 d1vga_control = RREG32_SOC15(DCE, 0, mmD1VGA_CONTROL);
-	unsigned size;
+	unsigned int size;
 
 	if (REG_GET_FIELD(d1vga_control, D1VGA_CONTROL, D1VGA_MODE_ENABLE)) {
 		size = AMDGPU_VBIOS_VGA_ALLOCATION;
@@ -672,7 +565,6 @@ static const struct amdgpu_gmc_funcs gmc_v10_0_gmc_funcs = {
 	.flush_gpu_tlb_pasid = gmc_v10_0_flush_gpu_tlb_pasid,
 	.emit_flush_gpu_tlb = gmc_v10_0_emit_flush_gpu_tlb,
 	.emit_pasid_mapping = gmc_v10_0_emit_pasid_mapping,
-	.map_mtype = gmc_v10_0_map_mtype,
 	.get_vm_pde = gmc_v10_0_get_vm_pde,
 	.get_vm_pte = gmc_v10_0_get_vm_pte,
 	.get_vbios_fb_size = gmc_v10_0_get_vbios_fb_size,
@@ -686,7 +578,7 @@ static void gmc_v10_0_set_gmc_funcs(struct amdgpu_device *adev)
 
 static void gmc_v10_0_set_umc_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[UMC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, UMC_HWIP, 0)) {
 	case IP_VERSION(8, 7, 0):
 		adev->umc.max_ras_err_cnt_per_query = UMC_V8_7_TOTAL_CHANNEL_NUM;
 		adev->umc.channel_inst_num = UMC_V8_7_CHANNEL_INSTANCE_NUM;
@@ -703,7 +595,7 @@ static void gmc_v10_0_set_umc_funcs(struct amdgpu_device *adev)
 
 static void gmc_v10_0_set_mmhub_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(2, 3, 0):
 	case IP_VERSION(2, 4, 0):
 	case IP_VERSION(2, 4, 1):
@@ -717,7 +609,7 @@ static void gmc_v10_0_set_mmhub_funcs(struct amdgpu_device *adev)
 
 static void gmc_v10_0_set_gfxhub_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 	case IP_VERSION(10, 3, 2):
 	case IP_VERSION(10, 3, 1):
@@ -735,9 +627,9 @@ static void gmc_v10_0_set_gfxhub_funcs(struct amdgpu_device *adev)
 }
 
 
-static int gmc_v10_0_early_init(void *handle)
+static int gmc_v10_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v10_0_set_mmhub_funcs(adev);
 	gmc_v10_0_set_gfxhub_funcs(adev);
@@ -751,13 +643,14 @@ static int gmc_v10_0_early_init(void *handle)
 	adev->gmc.private_aperture_start = 0x1000000000000000ULL;
 	adev->gmc.private_aperture_end =
 		adev->gmc.private_aperture_start + (4ULL << 30) - 1;
+	adev->gmc.noretry_flags = AMDGPU_VM_NORETRY_FLAGS_TF;
 
 	return 0;
 }
 
-static int gmc_v10_0_late_init(void *handle)
+static int gmc_v10_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_gmc_allocate_vm_inv_eng(adev);
@@ -781,9 +674,11 @@ static void gmc_v10_0_vram_gtt_location(struct amdgpu_device *adev,
 	/* add the xgmi offset of the physical node */
 	base += adev->gmc.xgmi.physical_node_id * adev->gmc.xgmi.node_segment_size;
 
+	amdgpu_gmc_set_agp_default(adev, mc);
 	amdgpu_gmc_vram_location(adev, &adev->gmc, base);
-	amdgpu_gmc_gart_location(adev, mc);
-	amdgpu_gmc_agp_location(adev, mc);
+	amdgpu_gmc_gart_location(adev, mc, AMDGPU_GART_PLACEMENT_BEST_FIT);
+	if (!amdgpu_sriov_vf(adev) && (amdgpu_agp == 1))
+		amdgpu_gmc_agp_location(adev, mc);
 
 	/* base offset of vram pages */
 	adev->vm_manager.vram_base_offset = adev->gfxhub.funcs->get_mc_fb_offset(adev);
@@ -830,7 +725,7 @@ static int gmc_v10_0_mc_init(struct amdgpu_device *adev)
 
 	/* set the gart size */
 	if (amdgpu_gart_size == -1) {
-		switch (adev->ip_versions[GC_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 		default:
 			adev->gmc.gart_size = 512ULL << 20;
 			break;
@@ -865,16 +760,16 @@ static int gmc_v10_0_gart_init(struct amdgpu_device *adev)
 		return r;
 
 	adev->gart.table_size = adev->gart.num_gpu_pages * 8;
-	adev->gart.gart_pte_flags = AMDGPU_PTE_MTYPE_NV10(MTYPE_UC) |
+	adev->gart.gart_pte_flags = AMDGPU_PTE_MTYPE_NV10(0ULL, MTYPE_UC) |
 				 AMDGPU_PTE_EXECUTABLE;
 
 	return amdgpu_gart_table_vram_alloc(adev);
 }
 
-static int gmc_v10_0_sw_init(void *handle)
+static int gmc_v10_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, vram_width = 0, vram_type = 0, vram_vendor = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->gfxhub.funcs->init(adev);
 
@@ -897,7 +792,7 @@ static int gmc_v10_0_sw_init(void *handle)
 		adev->gmc.vram_vendor = vram_vendor;
 	}
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 3, 0):
 		adev->gmc.mall_size = 128 * 1024 * 1024;
 		break;
@@ -915,7 +810,7 @@ static int gmc_v10_0_sw_init(void *handle)
 		break;
 	}
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(10, 1, 10):
 	case IP_VERSION(10, 1, 1):
 	case IP_VERSION(10, 1, 2):
@@ -929,7 +824,8 @@ static int gmc_v10_0_sw_init(void *handle)
 	case IP_VERSION(10, 3, 6):
 	case IP_VERSION(10, 3, 3):
 	case IP_VERSION(10, 3, 7):
-		adev->num_vmhubs = 2;
+		set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+		set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
 		/*
 		 * To fulfill 4-level page support,
 		 * vm size is 256TB (48bit), maximum size of Navi10/Navi14/Navi12,
@@ -971,7 +867,7 @@ static int gmc_v10_0_sw_init(void *handle)
 
 	r = dma_set_mask_and_coherent(adev->dev, DMA_BIT_MASK(44));
 	if (r) {
-		printk(KERN_WARNING "amdgpu: No suitable DMA available.\n");
+		dev_warn(adev->dev, "amdgpu: No suitable DMA available.\n");
 		return r;
 	}
 
@@ -1021,9 +917,9 @@ static void gmc_v10_0_gart_fini(struct amdgpu_device *adev)
 	amdgpu_gart_table_vram_free(adev);
 }
 
-static int gmc_v10_0_sw_fini(void *handle)
+static int gmc_v10_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_vm_manager_fini(adev);
 	gmc_v10_0_gart_fini(adev);
@@ -1067,29 +963,30 @@ static int gmc_v10_0_gart_enable(struct amdgpu_device *adev)
 	adev->hdp.funcs->init_registers(adev);
 
 	/* Flush HDP after it is initialized */
-	adev->hdp.funcs->flush_hdp(adev, NULL);
+	amdgpu_device_flush_hdp(adev, NULL);
 
-	value = (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS) ?
-		false : true;
+	value = amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS;
 
 	if (!adev->in_s0ix)
 		adev->gfxhub.funcs->set_fault_enable_default(adev, value);
 	adev->mmhub.funcs->set_fault_enable_default(adev, value);
-	gmc_v10_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB_0, 0);
+	gmc_v10_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB0(0), 0);
 	if (!adev->in_s0ix)
-		gmc_v10_0_flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB_0, 0);
+		gmc_v10_0_flush_gpu_tlb(adev, 0, AMDGPU_GFXHUB(0), 0);
 
 	DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
-		 (unsigned)(adev->gmc.gart_size >> 20),
+		 (unsigned int)(adev->gmc.gart_size >> 20),
 		 (unsigned long long)amdgpu_bo_gpu_offset(adev->gart.bo));
 
 	return 0;
 }
 
-static int gmc_v10_0_hw_init(void *handle)
+static int gmc_v10_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+	adev->gmc.flush_pasid_uses_kiq = !amdgpu_emu_mode;
 
 	/* The sequence of these two function calls matters.*/
 	gmc_v10_0_init_golden_registers(adev);
@@ -1131,9 +1028,9 @@ static void gmc_v10_0_gart_disable(struct amdgpu_device *adev)
 	adev->mmhub.funcs->gart_disable(adev);
 }
 
-static int gmc_v10_0_hw_fini(void *handle)
+static int gmc_v10_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v10_0_gart_disable(adev);
 
@@ -1143,64 +1040,60 @@ static int gmc_v10_0_hw_fini(void *handle)
 		return 0;
 	}
 
-	amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
 	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
 
+	if (adev->gmc.ecc_irq.funcs &&
+		amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC))
+		amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
+
 	return 0;
 }
 
-static int gmc_v10_0_suspend(void *handle)
+static int gmc_v10_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	gmc_v10_0_hw_fini(adev);
+	gmc_v10_0_hw_fini(ip_block);
 
 	return 0;
 }
 
-static int gmc_v10_0_resume(void *handle)
+static int gmc_v10_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = gmc_v10_0_hw_init(adev);
+	r = gmc_v10_0_hw_init(ip_block);
 	if (r)
 		return r;
 
-	amdgpu_vmid_reset_all(adev);
+	amdgpu_vmid_reset_all(ip_block->adev);
 
 	return 0;
 }
 
-static bool gmc_v10_0_is_idle(void *handle)
+static bool gmc_v10_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* MC is always ready in GMC v10.*/
 	return true;
 }
 
-static int gmc_v10_0_wait_for_idle(void *handle)
+static int gmc_v10_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* There is no need to wait for MC idle in GMC v10.*/
 	return 0;
 }
 
-static int gmc_v10_0_soft_reset(void *handle)
-{
-	return 0;
-}
-
-static int gmc_v10_0_set_clockgating_state(void *handle,
+static int gmc_v10_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_clockgating_state state)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	/*
 	 * The issue mmhub can't disconnect from DF with MMHUB clock gating being disabled
 	 * is a new problem observed at DF 3.0.3, however with the same suspend sequence not
 	 * seen any issue on the DF 3.0.2 series platform.
 	 */
-	if (adev->in_s0ix && adev->ip_versions[DF_HWIP][0] > IP_VERSION(3, 0, 2)) {
+	if (adev->in_s0ix &&
+	    amdgpu_ip_version(adev, DF_HWIP, 0) > IP_VERSION(3, 0, 2)) {
 		dev_dbg(adev->dev, "keep mmhub clock gating being enabled for s0ix\n");
 		return 0;
 	}
@@ -1209,29 +1102,29 @@ static int gmc_v10_0_set_clockgating_state(void *handle,
 	if (r)
 		return r;
 
-	if (adev->ip_versions[ATHUB_HWIP][0] >= IP_VERSION(2, 1, 0))
+	if (amdgpu_ip_version(adev, ATHUB_HWIP, 0) >= IP_VERSION(2, 1, 0))
 		return athub_v2_1_set_clockgating(adev, state);
 	else
 		return athub_v2_0_set_clockgating(adev, state);
 }
 
-static void gmc_v10_0_get_clockgating_state(void *handle, u64 *flags)
+static void gmc_v10_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 3) ||
-	    adev->ip_versions[GC_HWIP][0] == IP_VERSION(10, 1, 4))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 3) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(10, 1, 4))
 		return;
 
 	adev->mmhub.funcs->get_clockgating(adev, flags);
 
-	if (adev->ip_versions[ATHUB_HWIP][0] >= IP_VERSION(2, 1, 0))
+	if (amdgpu_ip_version(adev, ATHUB_HWIP, 0) >= IP_VERSION(2, 1, 0))
 		athub_v2_1_get_clockgating(adev, flags);
 	else
 		athub_v2_0_get_clockgating(adev, flags);
 }
 
-static int gmc_v10_0_set_powergating_state(void *handle,
+static int gmc_v10_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_powergating_state state)
 {
 	return 0;
@@ -1249,14 +1142,12 @@ const struct amd_ip_funcs gmc_v10_0_ip_funcs = {
 	.resume = gmc_v10_0_resume,
 	.is_idle = gmc_v10_0_is_idle,
 	.wait_for_idle = gmc_v10_0_wait_for_idle,
-	.soft_reset = gmc_v10_0_soft_reset,
 	.set_clockgating_state = gmc_v10_0_set_clockgating_state,
 	.set_powergating_state = gmc_v10_0_set_powergating_state,
 	.get_clockgating_state = gmc_v10_0_get_clockgating_state,
 };
 
-const struct amdgpu_ip_block_version gmc_v10_0_ip_block =
-{
+const struct amdgpu_ip_block_version gmc_v10_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GMC,
 	.major = 10,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
index 9f4f28192c60..ba59ee8e398a 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c
@@ -31,6 +31,8 @@
 #include "umc_v8_10.h"
 #include "athub/athub_3_0_0_sh_mask.h"
 #include "athub/athub_3_0_0_offset.h"
+#include "dcn/dcn_3_2_0_offset.h"
+#include "dcn/dcn_3_2_0_sh_mask.h"
 #include "oss/osssys_6_0_0_offset.h"
 #include "ivsrcid/vmc/irqsrcs_vmc_1_0.h"
 #include "navi10_enum.h"
@@ -40,15 +42,17 @@
 #include "nbio_v4_3.h"
 #include "gfxhub_v3_0.h"
 #include "gfxhub_v3_0_3.h"
+#include "gfxhub_v11_5_0.h"
 #include "mmhub_v3_0.h"
 #include "mmhub_v3_0_1.h"
 #include "mmhub_v3_0_2.h"
+#include "mmhub_v3_3.h"
 #include "athub_v3_0.h"
 
 
 static int gmc_v11_0_ecc_interrupt_state(struct amdgpu_device *adev,
 					 struct amdgpu_irq_src *src,
-					 unsigned type,
+					 unsigned int type,
 					 enum amdgpu_interrupt_state state)
 {
 	return 0;
@@ -56,25 +60,26 @@ static int gmc_v11_0_ecc_interrupt_state(struct amdgpu_device *adev,
 
 static int
 gmc_v11_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
-				   struct amdgpu_irq_src *src, unsigned type,
+				   struct amdgpu_irq_src *src, unsigned int type,
 				   enum amdgpu_interrupt_state state)
 {
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
 		/* MM HUB */
-		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, false);
+		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), false);
 		/* GFX HUB */
 		/* This works because this interrupt is only
 		 * enabled at init/resume and disabled in
 		 * fini/suspend, so the overall state doesn't
 		 * change over the course of suspend/resume.
 		 */
-		if (!adev->in_s0ix)
-			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, false);
+		if (!adev->in_s0ix && (adev->in_runpm || adev->in_suspend ||
+							   amdgpu_in_reset(adev)))
+			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), false);
 		break;
 	case AMDGPU_IRQ_STATE_ENABLE:
 		/* MM HUB */
-		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB_0, true);
+		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), true);
 		/* GFX HUB */
 		/* This works because this interrupt is only
 		 * enabled at init/resume and disabled in
@@ -82,7 +87,7 @@ gmc_v11_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 		 * change over the course of suspend/resume.
 		 */
 		if (!adev->in_s0ix)
-			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB_0, true);
+			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), true);
 		break;
 	default:
 		break;
@@ -95,42 +100,80 @@ static int gmc_v11_0_process_interrupt(struct amdgpu_device *adev,
 				       struct amdgpu_irq_src *source,
 				       struct amdgpu_iv_entry *entry)
 {
-	struct amdgpu_vmhub *hub = &adev->vmhub[entry->vmid_src];
+	uint32_t vmhub_index = entry->client_id == SOC21_IH_CLIENTID_VMC ?
+			       AMDGPU_MMHUB0(0) : AMDGPU_GFXHUB(0);
+	struct amdgpu_vmhub *hub = &adev->vmhub[vmhub_index];
+	bool retry_fault = !!(entry->src_data[1] &
+			      AMDGPU_GMC9_FAULT_SOURCE_DATA_RETRY);
+	bool write_fault = !!(entry->src_data[1] &
+			      AMDGPU_GMC9_FAULT_SOURCE_DATA_WRITE);
 	uint32_t status = 0;
 	u64 addr;
 
 	addr = (u64)entry->src_data[0] << 12;
 	addr |= ((u64)entry->src_data[1] & 0xf) << 44;
 
+	if (retry_fault) {
+		/* Returning 1 here also prevents sending the IV to the KFD */
+
+		/* Process it only if it's the first fault for this address */
+		if (entry->ih != &adev->irq.ih_soft &&
+		    amdgpu_gmc_filter_faults(adev, entry->ih, addr, entry->pasid,
+					     entry->timestamp))
+			return 1;
+
+		/* Delegate it to a different ring if the hardware hasn't
+		 * already done it.
+		 */
+		if (entry->ih == &adev->irq.ih) {
+			amdgpu_irq_delegate(adev, entry, 8);
+			return 1;
+		}
+
+		/* Try to handle the recoverable page faults by filling page
+		 * tables
+		 */
+		if (amdgpu_vm_handle_fault(adev, entry->pasid, 0, 0, addr,
+					   entry->timestamp, write_fault))
+			return 1;
+	}
+
 	if (!amdgpu_sriov_vf(adev)) {
 		/*
 		 * Issue a dummy read to wait for the status register to
 		 * be updated to avoid reading an incorrect value due to
 		 * the new fast GRBM interface.
 		 */
-		if (entry->vmid_src == AMDGPU_GFXHUB_0)
+		if (entry->vmid_src == AMDGPU_GFXHUB(0))
 			RREG32(hub->vm_l2_pro_fault_status);
 
 		status = RREG32(hub->vm_l2_pro_fault_status);
 		WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1);
+
+		amdgpu_vm_update_fault_cache(adev, entry->pasid, addr, status,
+					     entry->vmid_src ? AMDGPU_MMHUB0(0) : AMDGPU_GFXHUB(0));
 	}
 
 	if (printk_ratelimit()) {
-		struct amdgpu_task_info task_info;
-
-		memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-		amdgpu_vm_get_task_info(adev, entry->pasid, &task_info);
+		struct amdgpu_task_info *task_info;
 
 		dev_err(adev->dev,
-			"[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u, "
-			"for process %s pid %d thread %s pid %d)\n",
+			"[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u)\n",
 			entry->vmid_src ? "mmhub" : "gfxhub",
-			entry->src_id, entry->ring_id, entry->vmid,
-			entry->pasid, task_info.process_name, task_info.tgid,
-			task_info.task_name, task_info.pid);
+			entry->src_id, entry->ring_id, entry->vmid, entry->pasid);
+		task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+		if (task_info) {
+			amdgpu_vm_print_task_info(adev, task_info);
+			amdgpu_vm_put_task_info(task_info);
+		}
+
 		dev_err(adev->dev, "  in page starting at address 0x%016llx from client %d\n",
-			addr, entry->client_id);
-		if (!amdgpu_sriov_vf(adev))
+				addr, entry->client_id);
+
+		/* Only print L2 fault status if the status register could be read and
+		 * contains useful information
+		 */
+		if (status != 0)
 			hub->vmhub_funcs->print_l2_protection_fault_status(adev, status);
 	}
 
@@ -168,7 +211,7 @@ static void gmc_v11_0_set_irq_funcs(struct amdgpu_device *adev)
 static bool gmc_v11_0_use_invalidate_semaphore(struct amdgpu_device *adev,
 				       uint32_t vmhub)
 {
-	return ((vmhub == AMDGPU_MMHUB_0) &&
+	return ((vmhub == AMDGPU_MMHUB0(0)) &&
 		(!amdgpu_sriov_vf(adev)));
 }
 
@@ -181,27 +224,51 @@ static bool gmc_v11_0_get_vmid_pasid_mapping_info(
 	return !!(*p_pasid);
 }
 
-/*
- * GART
- * VMID 0 is the physical GPU addresses as used by the kernel.
- * VMIDs 1-15 are used for userspace clients and are handled
- * by the amdgpu vm/hsa code.
+/**
+ * gmc_v11_0_flush_gpu_tlb - gart tlb flush callback
+ *
+ * @adev: amdgpu_device pointer
+ * @vmid: vm instance to flush
+ * @vmhub: which hub to flush
+ * @flush_type: the flush type
+ *
+ * Flush the TLB for the requested page table.
  */
-
-static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
-				   unsigned int vmhub, uint32_t flush_type)
+static void gmc_v11_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
+					uint32_t vmhub, uint32_t flush_type)
 {
 	bool use_semaphore = gmc_v11_0_use_invalidate_semaphore(adev, vmhub);
 	struct amdgpu_vmhub *hub = &adev->vmhub[vmhub];
 	u32 inv_req = hub->vmhub_funcs->get_invalidate_req(vmid, flush_type);
-	u32 tmp;
 	/* Use register 17 for GART */
-	const unsigned eng = 17;
+	const unsigned int eng = 17;
+	unsigned char hub_ip;
+	u32 sem, req, ack;
 	unsigned int i;
-	unsigned char hub_ip = 0;
+	u32 tmp;
+
+	if ((vmhub == AMDGPU_GFXHUB(0)) && !adev->gfx.is_poweron)
+		return;
 
-	hub_ip = (vmhub == AMDGPU_GFXHUB_0) ?
-		   GC_HWIP : MMHUB_HWIP;
+	sem = hub->vm_inv_eng0_sem + hub->eng_distance * eng;
+	req = hub->vm_inv_eng0_req + hub->eng_distance * eng;
+	ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;
+
+	/* flush hdp cache */
+	amdgpu_device_flush_hdp(adev, NULL);
+
+	/* This is necessary for SRIOV as well as for GFXOFF to function
+	 * properly under bare metal
+	 */
+	if ((adev->gfx.kiq[0].ring.sched.ready || adev->mes.ring[0].sched.ready) &&
+	    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) {
+		amdgpu_gmc_fw_reg_write_reg_wait(adev, req, ack, inv_req,
+						 1 << vmid, GET_INST(GC, 0));
+		return;
+	}
+
+	/* This path is needed before KIQ/MES/GFXOFF are set up */
+	hub_ip = (vmhub == AMDGPU_GFXHUB(0)) ? GC_HWIP : MMHUB_HWIP;
 
 	spin_lock(&adev->gmc.invalidate_lock);
 	/*
@@ -215,8 +282,7 @@ static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 	if (use_semaphore) {
 		for (i = 0; i < adev->usec_timeout; i++) {
 			/* a read return value of 1 means semaphore acuqire */
-			tmp = RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_sem +
-					    hub->eng_distance * eng, hub_ip);
+			tmp = RREG32_RLC_NO_KIQ(sem, hub_ip);
 			if (tmp & 0x1)
 				break;
 			udelay(1);
@@ -226,12 +292,11 @@ static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 			DRM_ERROR("Timeout waiting for sem acquire in VM flush!\n");
 	}
 
-	WREG32_RLC_NO_KIQ(hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req, hub_ip);
+	WREG32_RLC_NO_KIQ(req, inv_req, hub_ip);
 
 	/* Wait for ACK with a delay.*/
 	for (i = 0; i < adev->usec_timeout; i++) {
-		tmp = RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_ack +
-				    hub->eng_distance * eng, hub_ip);
+		tmp = RREG32_RLC_NO_KIQ(ack, hub_ip);
 		tmp &= 1 << vmid;
 		if (tmp)
 			break;
@@ -241,15 +306,10 @@ static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 
 	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
 	if (use_semaphore)
-		/*
-		 * add semaphore release after invalidation,
-		 * write with 0 means semaphore release
-		 */
-		WREG32_RLC_NO_KIQ(hub->vm_inv_eng0_sem +
-			      hub->eng_distance * eng, 0, hub_ip);
+		WREG32_RLC_NO_KIQ(sem, 0, hub_ip);
 
 	/* Issue additional private vm invalidation to MMHUB */
-	if ((vmhub != AMDGPU_GFXHUB_0) &&
+	if ((vmhub != AMDGPU_GFXHUB(0)) &&
 	    (hub->vm_l2_bank_select_reserved_cid2) &&
 		!amdgpu_sriov_vf(adev)) {
 		inv_req = RREG32_NO_KIQ(hub->vm_l2_bank_select_reserved_cid2);
@@ -263,51 +323,8 @@ static void gmc_v11_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
 
 	spin_unlock(&adev->gmc.invalidate_lock);
 
-	if (i < adev->usec_timeout)
-		return;
-
-	DRM_ERROR("Timeout waiting for VM flush ACK!\n");
-}
-
-/**
- * gmc_v11_0_flush_gpu_tlb - gart tlb flush callback
- *
- * @adev: amdgpu_device pointer
- * @vmid: vm instance to flush
- * @vmhub: which hub to flush
- * @flush_type: the flush type
- *
- * Flush the TLB for the requested page table.
- */
-static void gmc_v11_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
-					uint32_t vmhub, uint32_t flush_type)
-{
-	if ((vmhub == AMDGPU_GFXHUB_0) && !adev->gfx.is_poweron)
-		return;
-
-	/* flush hdp cache */
-	adev->hdp.funcs->flush_hdp(adev, NULL);
-
-	/* For SRIOV run time, driver shouldn't access the register through MMIO
-	 * Directly use kiq to do the vm invalidation instead
-	 */
-	if ((adev->gfx.kiq.ring.sched.ready || adev->mes.ring.sched.ready) &&
-	    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) {
-		struct amdgpu_vmhub *hub = &adev->vmhub[vmhub];
-		const unsigned eng = 17;
-		u32 inv_req = hub->vmhub_funcs->get_invalidate_req(vmid, flush_type);
-		u32 req = hub->vm_inv_eng0_req + hub->eng_distance * eng;
-		u32 ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;
-
-		amdgpu_virt_kiq_reg_write_reg_wait(adev, req, ack, inv_req,
-				1 << vmid);
-		return;
-	}
-
-	mutex_lock(&adev->mman.gtt_window_lock);
-	gmc_v11_0_flush_vm_hub(adev, vmid, vmhub, 0);
-	mutex_unlock(&adev->mman.gtt_window_lock);
-	return;
+	if (i >= adev->usec_timeout)
+		dev_err(adev->dev, "Timeout waiting for VM flush ACK!\n");
 }
 
 /**
@@ -317,71 +334,44 @@ static void gmc_v11_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
  * @pasid: pasid to be flush
  * @flush_type: the flush type
  * @all_hub: flush all hubs
+ * @inst: is used to select which instance of KIQ to use for the invalidation
  *
  * Flush the TLB for the requested pasid.
  */
-static int gmc_v11_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
-					uint16_t pasid, uint32_t flush_type,
-					bool all_hub)
+static void gmc_v11_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
+					  uint16_t pasid, uint32_t flush_type,
+					  bool all_hub, uint32_t inst)
 {
+	uint16_t queried;
 	int vmid, i;
-	signed long r;
-	uint32_t seq;
-	uint16_t queried_pasid;
-	bool ret;
-	struct amdgpu_ring *ring = &adev->gfx.kiq.ring;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-
-	if (amdgpu_emu_mode == 0 && ring->sched.ready) {
-		spin_lock(&adev->gfx.kiq.ring_lock);
-		/* 2 dwords flush + 8 dwords fence */
-		amdgpu_ring_alloc(ring, kiq->pmf->invalidate_tlbs_size + 8);
-		kiq->pmf->kiq_invalidate_tlbs(ring,
-					pasid, flush_type, all_hub);
-		r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
-		if (r) {
-			amdgpu_ring_undo(ring);
-			spin_unlock(&adev->gfx.kiq.ring_lock);
-			return -ETIME;
-		}
-
-		amdgpu_ring_commit(ring);
-		spin_unlock(&adev->gfx.kiq.ring_lock);
-		r = amdgpu_fence_wait_polling(ring, seq, adev->usec_timeout);
-		if (r < 1) {
-			dev_err(adev->dev, "wait for kiq fence error: %ld.\n", r);
-			return -ETIME;
-		}
-
-		return 0;
-	}
 
 	for (vmid = 1; vmid < 16; vmid++) {
-
-		ret = gmc_v11_0_get_vmid_pasid_mapping_info(adev, vmid,
-				&queried_pasid);
-		if (ret	&& queried_pasid == pasid) {
-			if (all_hub) {
-				for (i = 0; i < adev->num_vmhubs; i++)
-					gmc_v11_0_flush_gpu_tlb(adev, vmid,
-							i, flush_type);
-			} else {
-				gmc_v11_0_flush_gpu_tlb(adev, vmid,
-						AMDGPU_GFXHUB_0, flush_type);
-			}
+		bool valid;
+
+		valid = gmc_v11_0_get_vmid_pasid_mapping_info(adev, vmid,
+							      &queried);
+		if (!valid || queried != pasid)
+			continue;
+
+		if (all_hub) {
+			for_each_set_bit(i, adev->vmhubs_mask,
+					 AMDGPU_MAX_VMHUBS)
+				gmc_v11_0_flush_gpu_tlb(adev, vmid, i,
+							flush_type);
+		} else {
+			gmc_v11_0_flush_gpu_tlb(adev, vmid, AMDGPU_GFXHUB(0),
+						flush_type);
 		}
 	}
-
-	return 0;
 }
 
 static uint64_t gmc_v11_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
-					     unsigned vmid, uint64_t pd_addr)
+					     unsigned int vmid, uint64_t pd_addr)
 {
-	bool use_semaphore = gmc_v11_0_use_invalidate_semaphore(ring->adev, ring->funcs->vmhub);
-	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub];
+	bool use_semaphore = gmc_v11_0_use_invalidate_semaphore(ring->adev, ring->vm_hub);
+	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
 	uint32_t req = hub->vmhub_funcs->get_invalidate_req(vmid, 0);
-	unsigned eng = ring->vm_inv_eng;
+	unsigned int eng = ring->vm_inv_eng;
 
 	/*
 	 * It may lose gpuvm invalidate acknowldege state across power-gating
@@ -423,17 +413,13 @@ static uint64_t gmc_v11_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 	return pd_addr;
 }
 
-static void gmc_v11_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
-					 unsigned pasid)
+static void gmc_v11_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned int vmid,
+					 unsigned int pasid)
 {
 	struct amdgpu_device *adev = ring->adev;
 	uint32_t reg;
 
-	/* MES fw manages IH_VMID_x_LUT updating */
-	if (ring->is_mes_queue)
-		return;
-
-	if (ring->funcs->vmhub == AMDGPU_GFXHUB_0)
+	if (ring->vm_hub == AMDGPU_GFXHUB(0))
 		reg = SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT) + vmid;
 	else
 		reg = SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT_MM) + vmid;
@@ -473,30 +459,11 @@ static void gmc_v11_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid
  * 0 valid
  */
 
-static uint64_t gmc_v11_0_map_mtype(struct amdgpu_device *adev, uint32_t flags)
-{
-	switch (flags) {
-	case AMDGPU_VM_MTYPE_DEFAULT:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_NC);
-	case AMDGPU_VM_MTYPE_NC:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_NC);
-	case AMDGPU_VM_MTYPE_WC:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_WC);
-	case AMDGPU_VM_MTYPE_CC:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_CC);
-	case AMDGPU_VM_MTYPE_UC:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_UC);
-	default:
-		return AMDGPU_PTE_MTYPE_NV10(MTYPE_NC);
-	}
-}
-
 static void gmc_v11_0_get_vm_pde(struct amdgpu_device *adev, int level,
 				 uint64_t *addr, uint64_t *flags)
 {
 	if (!(*flags & AMDGPU_PDE_PTE) && !(*flags & AMDGPU_PTE_SYSTEM))
-		*addr = adev->vm_manager.vram_base_offset + *addr -
-			adev->gmc.vram_start;
+		*addr = amdgpu_gmc_vram_mc2pa(adev, *addr);
 	BUG_ON(*addr & 0xFFFF00000000003FULL);
 
 	if (!adev->gmc.translate_further)
@@ -516,21 +483,39 @@ static void gmc_v11_0_get_vm_pde(struct amdgpu_device *adev, int level,
 }
 
 static void gmc_v11_0_get_vm_pte(struct amdgpu_device *adev,
-				 struct amdgpu_bo_va_mapping *mapping,
+				 struct amdgpu_vm *vm,
+				 struct amdgpu_bo *bo,
+				 uint32_t vm_flags,
 				 uint64_t *flags)
 {
-	struct amdgpu_bo *bo = mapping->bo_va->base.bo;
-
-	*flags &= ~AMDGPU_PTE_EXECUTABLE;
-	*flags |= mapping->flags & AMDGPU_PTE_EXECUTABLE;
+	if (vm_flags & AMDGPU_VM_PAGE_EXECUTABLE)
+		*flags |= AMDGPU_PTE_EXECUTABLE;
+	else
+		*flags &= ~AMDGPU_PTE_EXECUTABLE;
 
-	*flags &= ~AMDGPU_PTE_MTYPE_NV10_MASK;
-	*flags |= (mapping->flags & AMDGPU_PTE_MTYPE_NV10_MASK);
+	switch (vm_flags & AMDGPU_VM_MTYPE_MASK) {
+	case AMDGPU_VM_MTYPE_DEFAULT:
+	case AMDGPU_VM_MTYPE_NC:
+	default:
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_NC);
+		break;
+	case AMDGPU_VM_MTYPE_WC:
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_WC);
+		break;
+	case AMDGPU_VM_MTYPE_CC:
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_CC);
+		break;
+	case AMDGPU_VM_MTYPE_UC:
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_UC);
+		break;
+	}
 
-	*flags &= ~AMDGPU_PTE_NOALLOC;
-	*flags |= (mapping->flags & AMDGPU_PTE_NOALLOC);
+	if (vm_flags & AMDGPU_VM_PAGE_NOALLOC)
+		*flags |= AMDGPU_PTE_NOALLOC;
+	else
+		*flags &= ~AMDGPU_PTE_NOALLOC;
 
-	if (mapping->flags & AMDGPU_PTE_PRT) {
+	if (vm_flags & AMDGPU_VM_PAGE_PRT) {
 		*flags |= AMDGPU_PTE_PRT;
 		*flags |= AMDGPU_PTE_SNOOPED;
 		*flags |= AMDGPU_PTE_LOG;
@@ -539,14 +524,31 @@ static void gmc_v11_0_get_vm_pte(struct amdgpu_device *adev,
 	}
 
 	if (bo && bo->flags & (AMDGPU_GEM_CREATE_COHERENT |
+			       AMDGPU_GEM_CREATE_EXT_COHERENT |
 			       AMDGPU_GEM_CREATE_UNCACHED))
-		*flags = (*flags & ~AMDGPU_PTE_MTYPE_NV10_MASK) |
-			 AMDGPU_PTE_MTYPE_NV10(MTYPE_UC);
+		*flags = AMDGPU_PTE_MTYPE_NV10(*flags, MTYPE_UC);
 }
 
-static unsigned gmc_v11_0_get_vbios_fb_size(struct amdgpu_device *adev)
+static unsigned int gmc_v11_0_get_vbios_fb_size(struct amdgpu_device *adev)
 {
-	return 0;
+	u32 d1vga_control = RREG32_SOC15(DCE, 0, regD1VGA_CONTROL);
+	unsigned int size;
+
+	if (REG_GET_FIELD(d1vga_control, D1VGA_CONTROL, D1VGA_MODE_ENABLE)) {
+		size = AMDGPU_VBIOS_VGA_ALLOCATION;
+	} else {
+		u32 viewport;
+		u32 pitch;
+
+		viewport = RREG32_SOC15(DCE, 0, regHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION);
+		pitch = RREG32_SOC15(DCE, 0, regHUBPREQ0_DCSURF_SURFACE_PITCH);
+		size = (REG_GET_FIELD(viewport,
+					HUBP0_DCSURF_PRI_VIEWPORT_DIMENSION, PRI_VIEWPORT_HEIGHT) *
+				REG_GET_FIELD(pitch, HUBPREQ0_DCSURF_SURFACE_PITCH, PITCH) *
+				4);
+	}
+
+	return size;
 }
 
 static const struct amdgpu_gmc_funcs gmc_v11_0_gmc_funcs = {
@@ -554,7 +556,6 @@ static const struct amdgpu_gmc_funcs gmc_v11_0_gmc_funcs = {
 	.flush_gpu_tlb_pasid = gmc_v11_0_flush_gpu_tlb_pasid,
 	.emit_flush_gpu_tlb = gmc_v11_0_emit_flush_gpu_tlb,
 	.emit_pasid_mapping = gmc_v11_0_emit_pasid_mapping,
-	.map_mtype = gmc_v11_0_map_mtype,
 	.get_vm_pde = gmc_v11_0_get_vm_pde,
 	.get_vm_pte = gmc_v11_0_get_vm_pte,
 	.get_vbios_fb_size = gmc_v11_0_get_vbios_fb_size,
@@ -567,7 +568,7 @@ static void gmc_v11_0_set_gmc_funcs(struct amdgpu_device *adev)
 
 static void gmc_v11_0_set_umc_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[UMC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, UMC_HWIP, 0)) {
 	case IP_VERSION(8, 10, 0):
 		adev->umc.channel_inst_num = UMC_V8_10_CHANNEL_INSTANCE_NUM;
 		adev->umc.umc_inst_num = UMC_V8_10_UMC_INSTANCE_NUM;
@@ -590,13 +591,18 @@ static void gmc_v11_0_set_umc_funcs(struct amdgpu_device *adev)
 
 static void gmc_v11_0_set_mmhub_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(3, 0, 1):
 		adev->mmhub.funcs = &mmhub_v3_0_1_funcs;
 		break;
 	case IP_VERSION(3, 0, 2):
 		adev->mmhub.funcs = &mmhub_v3_0_2_funcs;
 		break;
+	case IP_VERSION(3, 3, 0):
+	case IP_VERSION(3, 3, 1):
+	case IP_VERSION(3, 3, 2):
+		adev->mmhub.funcs = &mmhub_v3_3_funcs;
+		break;
 	default:
 		adev->mmhub.funcs = &mmhub_v3_0_funcs;
 		break;
@@ -605,19 +611,25 @@ static void gmc_v11_0_set_mmhub_funcs(struct amdgpu_device *adev)
 
 static void gmc_v11_0_set_gfxhub_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 3):
 		adev->gfxhub.funcs = &gfxhub_v3_0_3_funcs;
 		break;
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
+		adev->gfxhub.funcs = &gfxhub_v11_5_0_funcs;
+		break;
 	default:
 		adev->gfxhub.funcs = &gfxhub_v3_0_funcs;
 		break;
 	}
 }
 
-static int gmc_v11_0_early_init(void *handle)
+static int gmc_v11_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v11_0_set_gfxhub_funcs(adev);
 	gmc_v11_0_set_mmhub_funcs(adev);
@@ -631,13 +643,14 @@ static int gmc_v11_0_early_init(void *handle)
 	adev->gmc.private_aperture_start = 0x1000000000000000ULL;
 	adev->gmc.private_aperture_end =
 		adev->gmc.private_aperture_start + (4ULL << 30) - 1;
+	adev->gmc.noretry_flags = AMDGPU_VM_NORETRY_FLAGS_TF;
 
 	return 0;
 }
 
-static int gmc_v11_0_late_init(void *handle)
+static int gmc_v11_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_gmc_allocate_vm_inv_eng(adev);
@@ -658,9 +671,13 @@ static void gmc_v11_0_vram_gtt_location(struct amdgpu_device *adev,
 
 	base = adev->mmhub.funcs->get_fb_location(adev);
 
+	amdgpu_gmc_set_agp_default(adev, mc);
 	amdgpu_gmc_vram_location(adev, &adev->gmc, base);
-	amdgpu_gmc_gart_location(adev, mc);
-	amdgpu_gmc_agp_location(adev, mc);
+	amdgpu_gmc_gart_location(adev, mc, AMDGPU_GART_PLACEMENT_HIGH);
+	if (!amdgpu_sriov_vf(adev) &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(11, 5, 0)) &&
+	    (amdgpu_agp == 1))
+		amdgpu_gmc_agp_location(adev, mc);
 
 	/* base offset of vram pages */
 	if (amdgpu_sriov_vf(adev))
@@ -707,9 +724,9 @@ static int gmc_v11_0_mc_init(struct amdgpu_device *adev)
 		adev->gmc.visible_vram_size = adev->gmc.real_vram_size;
 
 	/* set the gart size */
-	if (amdgpu_gart_size == -1) {
+	if (amdgpu_gart_size == -1)
 		adev->gmc.gart_size = 512ULL << 20;
-	} else
+	else
 		adev->gmc.gart_size = (u64)amdgpu_gart_size << 20;
 
 	gmc_v11_0_vram_gtt_location(adev, &adev->gmc);
@@ -732,19 +749,21 @@ static int gmc_v11_0_gart_init(struct amdgpu_device *adev)
 		return r;
 
 	adev->gart.table_size = adev->gart.num_gpu_pages * 8;
-	adev->gart.gart_pte_flags = AMDGPU_PTE_MTYPE_NV10(MTYPE_UC) |
+	adev->gart.gart_pte_flags = AMDGPU_PTE_MTYPE_NV10(0ULL, MTYPE_UC) |
 				 AMDGPU_PTE_EXECUTABLE;
 
 	return amdgpu_gart_table_vram_alloc(adev);
 }
 
-static int gmc_v11_0_sw_init(void *handle)
+static int gmc_v11_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, vram_width = 0, vram_type = 0, vram_vendor = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->mmhub.funcs->init(adev);
 
+	adev->gfxhub.funcs->init(adev);
+
 	spin_lock_init(&adev->gmc.invalidate_lock);
 
 	r = amdgpu_atomfirmware_get_vram_info(adev,
@@ -754,13 +773,30 @@ static int gmc_v11_0_sw_init(void *handle)
 	adev->gmc.vram_type = vram_type;
 	adev->gmc.vram_vendor = vram_vendor;
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	/* The mall_size is already calculated as mall_size_per_umc * num_umc.
+	 * However, for gfx1151, which features a 2-to-1 UMC mapping,
+	 * the result must be multiplied by 2 to determine the actual mall size.
+	 */
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(11, 5, 1):
+		adev->gmc.mall_size *= 2;
+		break;
+	default:
+		break;
+	}
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
 	case IP_VERSION(11, 0, 1):
 	case IP_VERSION(11, 0, 2):
 	case IP_VERSION(11, 0, 3):
 	case IP_VERSION(11, 0, 4):
-		adev->num_vmhubs = 2;
+	case IP_VERSION(11, 5, 0):
+	case IP_VERSION(11, 5, 1):
+	case IP_VERSION(11, 5, 2):
+	case IP_VERSION(11, 5, 3):
+		set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+		set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
 		/*
 		 * To fulfill 4-level page support,
 		 * vm size is 256TB (48bit), maximum size,
@@ -802,7 +838,7 @@ static int gmc_v11_0_sw_init(void *handle)
 
 	r = dma_set_mask_and_coherent(adev->dev, DMA_BIT_MASK(44));
 	if (r) {
-		printk(KERN_WARNING "amdgpu: No suitable DMA available.\n");
+		dev_warn(adev->dev, "amdgpu: No suitable DMA available.\n");
 		return r;
 	}
 
@@ -829,7 +865,7 @@ static int gmc_v11_0_sw_init(void *handle)
 	 * amdgpu graphics/compute will use VMIDs 1-7
 	 * amdkfd will use VMIDs 8-15
 	 */
-	adev->vm_manager.first_kfd_vmid = 8;
+	adev->vm_manager.first_kfd_vmid = adev->gfx.disable_kq ? 1 : 8;
 
 	amdgpu_vm_manager_init(adev);
 
@@ -852,9 +888,9 @@ static void gmc_v11_0_gart_fini(struct amdgpu_device *adev)
 	amdgpu_gart_table_vram_free(adev);
 }
 
-static int gmc_v11_0_sw_fini(void *handle)
+static int gmc_v11_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_vm_manager_fini(adev);
 	gmc_v11_0_gart_fini(adev);
@@ -867,7 +903,7 @@ static int gmc_v11_0_sw_fini(void *handle)
 static void gmc_v11_0_init_golden_registers(struct amdgpu_device *adev)
 {
 	if (amdgpu_sriov_vf(adev)) {
-		struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB_0];
+		struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
 
 		WREG32(hub->vm_contexts_disable, 0);
 		return;
@@ -896,25 +932,26 @@ static int gmc_v11_0_gart_enable(struct amdgpu_device *adev)
 		return r;
 
 	/* Flush HDP after it is initialized */
-	adev->hdp.funcs->flush_hdp(adev, NULL);
+	amdgpu_device_flush_hdp(adev, NULL);
 
-	value = (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS) ?
-		false : true;
+	value = amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS;
 
 	adev->mmhub.funcs->set_fault_enable_default(adev, value);
-	gmc_v11_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB_0, 0);
+	gmc_v11_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB0(0), 0);
 
 	DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
-		 (unsigned)(adev->gmc.gart_size >> 20),
+		 (unsigned int)(adev->gmc.gart_size >> 20),
 		 (unsigned long long)amdgpu_bo_gpu_offset(adev->gart.bo));
 
 	return 0;
 }
 
-static int gmc_v11_0_hw_init(void *handle)
+static int gmc_v11_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+
+	adev->gmc.flush_pasid_uses_kiq = !amdgpu_emu_mode;
 
 	/* The sequence of these two function calls matters.*/
 	gmc_v11_0_init_golden_registers(adev);
@@ -941,9 +978,9 @@ static void gmc_v11_0_gart_disable(struct amdgpu_device *adev)
 	adev->mmhub.funcs->gart_disable(adev);
 }
 
-static int gmc_v11_0_hw_fini(void *handle)
+static int gmc_v11_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev)) {
 		/* full access mode, so don't touch any GMC register */
@@ -951,58 +988,54 @@ static int gmc_v11_0_hw_fini(void *handle)
 		return 0;
 	}
 
-	amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
 	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
+
+	if (adev->gmc.ecc_irq.funcs &&
+		amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC))
+		amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
+
 	gmc_v11_0_gart_disable(adev);
 
 	return 0;
 }
 
-static int gmc_v11_0_suspend(void *handle)
+static int gmc_v11_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	gmc_v11_0_hw_fini(adev);
+	gmc_v11_0_hw_fini(ip_block);
 
 	return 0;
 }
 
-static int gmc_v11_0_resume(void *handle)
+static int gmc_v11_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = gmc_v11_0_hw_init(adev);
+	r = gmc_v11_0_hw_init(ip_block);
 	if (r)
 		return r;
 
-	amdgpu_vmid_reset_all(adev);
+	amdgpu_vmid_reset_all(ip_block->adev);
 
 	return 0;
 }
 
-static bool gmc_v11_0_is_idle(void *handle)
+static bool gmc_v11_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* MC is always ready in GMC v11.*/
 	return true;
 }
 
-static int gmc_v11_0_wait_for_idle(void *handle)
+static int gmc_v11_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* There is no need to wait for MC idle in GMC v11.*/
 	return 0;
 }
 
-static int gmc_v11_0_soft_reset(void *handle)
-{
-	return 0;
-}
-
-static int gmc_v11_0_set_clockgating_state(void *handle,
+static int gmc_v11_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_clockgating_state state)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = adev->mmhub.funcs->set_clockgating(adev, state);
 	if (r)
@@ -1011,16 +1044,16 @@ static int gmc_v11_0_set_clockgating_state(void *handle,
 	return athub_v3_0_set_clockgating(adev, state);
 }
 
-static void gmc_v11_0_get_clockgating_state(void *handle, u64 *flags)
+static void gmc_v11_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->mmhub.funcs->get_clockgating(adev, flags);
 
 	athub_v3_0_get_clockgating(adev, flags);
 }
 
-static int gmc_v11_0_set_powergating_state(void *handle,
+static int gmc_v11_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_powergating_state state)
 {
 	return 0;
@@ -1038,7 +1071,6 @@ const struct amd_ip_funcs gmc_v11_0_ip_funcs = {
 	.resume = gmc_v11_0_resume,
 	.is_idle = gmc_v11_0_is_idle,
 	.wait_for_idle = gmc_v11_0_wait_for_idle,
-	.soft_reset = gmc_v11_0_soft_reset,
 	.set_clockgating_state = gmc_v11_0_set_clockgating_state,
 	.set_powergating_state = gmc_v11_0_set_powergating_state,
 	.get_clockgating_state = gmc_v11_0_get_clockgating_state,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c
new file mode 100644
index 000000000000..7a9d6894e321
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c
@@ -0,0 +1,1070 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include <linux/firmware.h>
+#include <linux/pci.h>
+
+#include <drm/drm_cache.h>
+
+#include "amdgpu.h"
+#include "amdgpu_atomfirmware.h"
+#include "gmc_v12_0.h"
+#include "athub/athub_4_1_0_sh_mask.h"
+#include "athub/athub_4_1_0_offset.h"
+#include "oss/osssys_7_0_0_offset.h"
+#include "ivsrcid/vmc/irqsrcs_vmc_1_0.h"
+#include "soc24_enum.h"
+#include "soc24.h"
+#include "soc15d.h"
+#include "soc15_common.h"
+#include "nbif_v6_3_1.h"
+#include "gfxhub_v12_0.h"
+#include "mmhub_v4_1_0.h"
+#include "athub_v4_1_0.h"
+#include "umc_v8_14.h"
+
+static int gmc_v12_0_ecc_interrupt_state(struct amdgpu_device *adev,
+					 struct amdgpu_irq_src *src,
+					 unsigned type,
+					 enum amdgpu_interrupt_state state)
+{
+	return 0;
+}
+
+static int gmc_v12_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
+					      struct amdgpu_irq_src *src, unsigned type,
+					      enum amdgpu_interrupt_state state)
+{
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		/* MM HUB */
+		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), false);
+		/* GFX HUB */
+		/* This works because this interrupt is only
+		 * enabled at init/resume and disabled in
+		 * fini/suspend, so the overall state doesn't
+		 * change over the course of suspend/resume.
+		 */
+		if (!adev->in_s0ix)
+			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), false);
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		/* MM HUB */
+		amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_MMHUB0(0), true);
+		/* GFX HUB */
+		/* This works because this interrupt is only
+		 * enabled at init/resume and disabled in
+		 * fini/suspend, so the overall state doesn't
+		 * change over the course of suspend/resume.
+		 */
+		if (!adev->in_s0ix)
+			amdgpu_gmc_set_vm_fault_masks(adev, AMDGPU_GFXHUB(0), true);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int gmc_v12_0_process_interrupt(struct amdgpu_device *adev,
+				       struct amdgpu_irq_src *source,
+				       struct amdgpu_iv_entry *entry)
+{
+	struct amdgpu_vmhub *hub;
+	bool retry_fault = !!(entry->src_data[1] &
+			      AMDGPU_GMC9_FAULT_SOURCE_DATA_RETRY);
+	bool write_fault = !!(entry->src_data[1] &
+			      AMDGPU_GMC9_FAULT_SOURCE_DATA_WRITE);
+	uint32_t status = 0;
+	u64 addr;
+
+	addr = (u64)entry->src_data[0] << 12;
+	addr |= ((u64)entry->src_data[1] & 0xf) << 44;
+
+	if (entry->client_id == SOC21_IH_CLIENTID_VMC)
+		hub = &adev->vmhub[AMDGPU_MMHUB0(0)];
+	else
+		hub = &adev->vmhub[AMDGPU_GFXHUB(0)];
+
+	if (retry_fault) {
+		/* Returning 1 here also prevents sending the IV to the KFD */
+
+		/* Process it only if it's the first fault for this address */
+		if (entry->ih != &adev->irq.ih_soft &&
+		    amdgpu_gmc_filter_faults(adev, entry->ih, addr, entry->pasid,
+					     entry->timestamp))
+			return 1;
+
+		/* Delegate it to a different ring if the hardware hasn't
+		 * already done it.
+		 */
+		if (entry->ih == &adev->irq.ih) {
+			amdgpu_irq_delegate(adev, entry, 8);
+			return 1;
+		}
+
+		/* Try to handle the recoverable page faults by filling page
+		 * tables
+		 */
+		if (amdgpu_vm_handle_fault(adev, entry->pasid, 0, 0, addr,
+					   entry->timestamp, write_fault))
+			return 1;
+	}
+
+	if (!amdgpu_sriov_vf(adev)) {
+		/*
+		 * Issue a dummy read to wait for the status register to
+		 * be updated to avoid reading an incorrect value due to
+		 * the new fast GRBM interface.
+		 */
+		if (entry->vmid_src == AMDGPU_GFXHUB(0))
+			RREG32(hub->vm_l2_pro_fault_status);
+
+		status = RREG32(hub->vm_l2_pro_fault_status);
+		WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1);
+
+		amdgpu_vm_update_fault_cache(adev, entry->pasid, addr, status,
+					     entry->vmid_src ? AMDGPU_MMHUB0(0) : AMDGPU_GFXHUB(0));
+	}
+
+	if (printk_ratelimit()) {
+		struct amdgpu_task_info *task_info;
+
+		dev_err(adev->dev,
+			"[%s] page fault (src_id:%u ring:%u vmid:%u pasid:%u)\n",
+			entry->vmid_src ? "mmhub" : "gfxhub",
+			entry->src_id, entry->ring_id, entry->vmid, entry->pasid);
+		task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+		if (task_info) {
+			amdgpu_vm_print_task_info(adev, task_info);
+			amdgpu_vm_put_task_info(task_info);
+		}
+
+		dev_err(adev->dev, "  in page starting at address 0x%016llx from client %d\n",
+				addr, entry->client_id);
+
+		/* Only print L2 fault status if the status register could be read and
+		 * contains useful information
+		 */
+		if (status != 0)
+			hub->vmhub_funcs->print_l2_protection_fault_status(adev, status);
+	}
+
+	return 0;
+}
+
+static const struct amdgpu_irq_src_funcs gmc_v12_0_irq_funcs = {
+	.set = gmc_v12_0_vm_fault_interrupt_state,
+	.process = gmc_v12_0_process_interrupt,
+};
+
+static const struct amdgpu_irq_src_funcs gmc_v12_0_ecc_funcs = {
+	.set = gmc_v12_0_ecc_interrupt_state,
+	.process = amdgpu_umc_process_ecc_irq,
+};
+
+static void gmc_v12_0_set_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->gmc.vm_fault.num_types = 1;
+	adev->gmc.vm_fault.funcs = &gmc_v12_0_irq_funcs;
+
+	if (!amdgpu_sriov_vf(adev)) {
+		adev->gmc.ecc_irq.num_types = 1;
+		adev->gmc.ecc_irq.funcs = &gmc_v12_0_ecc_funcs;
+	}
+}
+
+/**
+ * gmc_v12_0_use_invalidate_semaphore - judge whether to use semaphore
+ *
+ * @adev: amdgpu_device pointer
+ * @vmhub: vmhub type
+ *
+ */
+static bool gmc_v12_0_use_invalidate_semaphore(struct amdgpu_device *adev,
+				       uint32_t vmhub)
+{
+	return ((vmhub == AMDGPU_MMHUB0(0)) &&
+		(!amdgpu_sriov_vf(adev)));
+}
+
+static bool gmc_v12_0_get_vmid_pasid_mapping_info(
+					struct amdgpu_device *adev,
+					uint8_t vmid, uint16_t *p_pasid)
+{
+	*p_pasid = RREG32(SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT) + vmid) & 0xffff;
+
+	return !!(*p_pasid);
+}
+
+/*
+ * GART
+ * VMID 0 is the physical GPU addresses as used by the kernel.
+ * VMIDs 1-15 are used for userspace clients and are handled
+ * by the amdgpu vm/hsa code.
+ */
+
+static void gmc_v12_0_flush_vm_hub(struct amdgpu_device *adev, uint32_t vmid,
+				   unsigned int vmhub, uint32_t flush_type)
+{
+	bool use_semaphore = gmc_v12_0_use_invalidate_semaphore(adev, vmhub);
+	struct amdgpu_vmhub *hub = &adev->vmhub[vmhub];
+	u32 inv_req = hub->vmhub_funcs->get_invalidate_req(vmid, flush_type);
+	u32 tmp;
+	/* Use register 17 for GART */
+	const unsigned eng = 17;
+	unsigned int i;
+	unsigned char hub_ip = 0;
+
+	hub_ip = (vmhub == AMDGPU_GFXHUB(0)) ?
+		   GC_HWIP : MMHUB_HWIP;
+
+	spin_lock(&adev->gmc.invalidate_lock);
+	/*
+	 * It may lose gpuvm invalidate acknowldege state across power-gating
+	 * off cycle, add semaphore acquire before invalidation and semaphore
+	 * release after invalidation to avoid entering power gated state
+	 * to WA the Issue
+	 */
+
+	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
+	if (use_semaphore) {
+		for (i = 0; i < adev->usec_timeout; i++) {
+			/* a read return value of 1 means semaphore acuqire */
+			tmp = RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_sem +
+					    hub->eng_distance * eng, hub_ip);
+			if (tmp & 0x1)
+				break;
+			udelay(1);
+		}
+
+		if (i >= adev->usec_timeout)
+			dev_err(adev->dev,
+				"Timeout waiting for sem acquire in VM flush!\n");
+	}
+
+	WREG32_RLC_NO_KIQ(hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req, hub_ip);
+
+	/* Wait for ACK with a delay.*/
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32_RLC_NO_KIQ(hub->vm_inv_eng0_ack +
+				    hub->eng_distance * eng, hub_ip);
+		tmp &= 1 << vmid;
+		if (tmp)
+			break;
+
+		udelay(1);
+	}
+
+	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
+	if (use_semaphore)
+		/*
+		 * add semaphore release after invalidation,
+		 * write with 0 means semaphore release
+		 */
+		WREG32_RLC_NO_KIQ(hub->vm_inv_eng0_sem +
+			      hub->eng_distance * eng, 0, hub_ip);
+
+	/* Issue additional private vm invalidation to MMHUB */
+	if ((vmhub != AMDGPU_GFXHUB(0)) &&
+	    (hub->vm_l2_bank_select_reserved_cid2) &&
+		!amdgpu_sriov_vf(adev)) {
+		inv_req = RREG32_NO_KIQ(hub->vm_l2_bank_select_reserved_cid2);
+		/* bit 25: RSERVED_CACHE_PRIVATE_INVALIDATION */
+		inv_req |= (1 << 25);
+		/* Issue private invalidation */
+		WREG32_NO_KIQ(hub->vm_l2_bank_select_reserved_cid2, inv_req);
+		/* Read back to ensure invalidation is done*/
+		RREG32_NO_KIQ(hub->vm_l2_bank_select_reserved_cid2);
+	}
+
+	spin_unlock(&adev->gmc.invalidate_lock);
+
+	if (i < adev->usec_timeout)
+		return;
+
+	dev_err(adev->dev, "Timeout waiting for VM flush ACK!\n");
+}
+
+/**
+ * gmc_v12_0_flush_gpu_tlb - gart tlb flush callback
+ *
+ * @adev: amdgpu_device pointer
+ * @vmid: vm instance to flush
+ * @vmhub: which hub to flush
+ * @flush_type: the flush type
+ *
+ * Flush the TLB for the requested page table.
+ */
+static void gmc_v12_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
+					uint32_t vmhub, uint32_t flush_type)
+{
+	if ((vmhub == AMDGPU_GFXHUB(0)) && !adev->gfx.is_poweron)
+		return;
+
+	/* flush hdp cache */
+	amdgpu_device_flush_hdp(adev, NULL);
+
+	/* This is necessary for SRIOV as well as for GFXOFF to function
+	 * properly under bare metal
+	 */
+	if ((adev->gfx.kiq[0].ring.sched.ready || adev->mes.ring[0].sched.ready) &&
+	    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) {
+		struct amdgpu_vmhub *hub = &adev->vmhub[vmhub];
+		const unsigned eng = 17;
+		u32 inv_req = hub->vmhub_funcs->get_invalidate_req(vmid, flush_type);
+		u32 req = hub->vm_inv_eng0_req + hub->eng_distance * eng;
+		u32 ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;
+
+		amdgpu_gmc_fw_reg_write_reg_wait(adev, req, ack, inv_req,
+				1 << vmid, GET_INST(GC, 0));
+		return;
+	}
+
+	gmc_v12_0_flush_vm_hub(adev, vmid, vmhub, 0);
+	return;
+}
+
+/**
+ * gmc_v12_0_flush_gpu_tlb_pasid - tlb flush via pasid
+ *
+ * @adev: amdgpu_device pointer
+ * @pasid: pasid to be flush
+ * @flush_type: the flush type
+ * @all_hub: flush all hubs
+ * @inst: is used to select which instance of KIQ to use for the invalidation
+ *
+ * Flush the TLB for the requested pasid.
+ */
+static void gmc_v12_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
+					  uint16_t pasid, uint32_t flush_type,
+					  bool all_hub, uint32_t inst)
+{
+	uint16_t queried;
+	int vmid, i;
+
+	if (adev->enable_uni_mes && adev->mes.ring[AMDGPU_MES_SCHED_PIPE].sched.ready &&
+	    (adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) >= 0x84) {
+		struct mes_inv_tlbs_pasid_input input = {0};
+		input.pasid = pasid;
+		input.flush_type = flush_type;
+		input.hub_id = AMDGPU_GFXHUB(0);
+		/* MES will invalidate all gc_hub for the device from master */
+		adev->mes.funcs->invalidate_tlbs_pasid(&adev->mes, &input);
+		if (all_hub) {
+			/* Only need to invalidate mm_hub now, gfx12 only support one mmhub */
+			input.hub_id = AMDGPU_MMHUB0(0);
+			adev->mes.funcs->invalidate_tlbs_pasid(&adev->mes, &input);
+		}
+		return;
+	}
+
+	for (vmid = 1; vmid < 16; vmid++) {
+		bool valid;
+
+		valid = gmc_v12_0_get_vmid_pasid_mapping_info(adev, vmid,
+							      &queried);
+		if (!valid || queried != pasid)
+			continue;
+
+		if (all_hub) {
+			for_each_set_bit(i, adev->vmhubs_mask,
+					 AMDGPU_MAX_VMHUBS)
+				gmc_v12_0_flush_gpu_tlb(adev, vmid, i,
+							flush_type);
+		} else {
+			gmc_v12_0_flush_gpu_tlb(adev, vmid, AMDGPU_GFXHUB(0),
+						flush_type);
+		}
+	}
+}
+
+static uint64_t gmc_v12_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
+					     unsigned vmid, uint64_t pd_addr)
+{
+	bool use_semaphore = gmc_v12_0_use_invalidate_semaphore(ring->adev, ring->vm_hub);
+	struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->vm_hub];
+	uint32_t req = hub->vmhub_funcs->get_invalidate_req(vmid, 0);
+	unsigned eng = ring->vm_inv_eng;
+
+	/*
+	 * It may lose gpuvm invalidate acknowldege state across power-gating
+	 * off cycle, add semaphore acquire before invalidation and semaphore
+	 * release after invalidation to avoid entering power gated state
+	 * to WA the Issue
+	 */
+
+	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
+	if (use_semaphore)
+		/* a read return value of 1 means semaphore acuqire */
+		amdgpu_ring_emit_reg_wait(ring,
+					  hub->vm_inv_eng0_sem +
+					  hub->eng_distance * eng, 0x1, 0x1);
+
+	amdgpu_ring_emit_wreg(ring, hub->ctx0_ptb_addr_lo32 +
+			      (hub->ctx_addr_distance * vmid),
+			      lower_32_bits(pd_addr));
+
+	amdgpu_ring_emit_wreg(ring, hub->ctx0_ptb_addr_hi32 +
+			      (hub->ctx_addr_distance * vmid),
+			      upper_32_bits(pd_addr));
+
+	amdgpu_ring_emit_reg_write_reg_wait(ring, hub->vm_inv_eng0_req +
+					    hub->eng_distance * eng,
+					    hub->vm_inv_eng0_ack +
+					    hub->eng_distance * eng,
+					    req, 1 << vmid);
+
+	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
+	if (use_semaphore)
+		/*
+		 * add semaphore release after invalidation,
+		 * write with 0 means semaphore release
+		 */
+		amdgpu_ring_emit_wreg(ring, hub->vm_inv_eng0_sem +
+				      hub->eng_distance * eng, 0);
+
+	return pd_addr;
+}
+
+static void gmc_v12_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
+					 unsigned pasid)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t reg;
+
+	if (ring->vm_hub == AMDGPU_GFXHUB(0))
+		reg = SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT) + vmid;
+	else
+		reg = SOC15_REG_OFFSET(OSSSYS, 0, regIH_VMID_0_LUT_MM) + vmid;
+
+	amdgpu_ring_emit_wreg(ring, reg, pasid);
+}
+
+/*
+ * PTE format:
+ * 63 P
+ * 62:59 reserved
+ * 58 D
+ * 57 G
+ * 56 T
+ * 55:54 M
+ * 53:52 SW
+ * 51:48 reserved for future
+ * 47:12 4k physical page base address
+ * 11:7 fragment
+ * 6 write
+ * 5 read
+ * 4 exe
+ * 3 Z
+ * 2 snooped
+ * 1 system
+ * 0 valid
+ *
+ * PDE format:
+ * 63 P
+ * 62:58 block fragment size
+ * 57 reserved
+ * 56 A
+ * 55:54 M
+ * 53:52 reserved
+ * 51:48 reserved for future
+ * 47:6 physical base address of PD or PTE
+ * 5:3 reserved
+ * 2 C
+ * 1 system
+ * 0 valid
+ */
+
+static void gmc_v12_0_get_vm_pde(struct amdgpu_device *adev, int level,
+				 uint64_t *addr, uint64_t *flags)
+{
+	if (!(*flags & AMDGPU_PDE_PTE_GFX12) && !(*flags & AMDGPU_PTE_SYSTEM))
+		*addr = adev->vm_manager.vram_base_offset + *addr -
+			adev->gmc.vram_start;
+	BUG_ON(*addr & 0xFFFF00000000003FULL);
+
+	if (!adev->gmc.translate_further)
+		return;
+
+	if (level == AMDGPU_VM_PDB1) {
+		/* Set the block fragment size */
+		if (!(*flags & AMDGPU_PDE_PTE_GFX12))
+			*flags |= AMDGPU_PDE_BFS_GFX12(0x9);
+
+	} else if (level == AMDGPU_VM_PDB0) {
+		if (*flags & AMDGPU_PDE_PTE_GFX12)
+			*flags &= ~AMDGPU_PDE_PTE_GFX12;
+	}
+}
+
+static void gmc_v12_0_get_vm_pte(struct amdgpu_device *adev,
+				 struct amdgpu_vm *vm,
+				 struct amdgpu_bo *bo,
+				 uint32_t vm_flags,
+				 uint64_t *flags)
+{
+	if (vm_flags & AMDGPU_VM_PAGE_EXECUTABLE)
+		*flags |= AMDGPU_PTE_EXECUTABLE;
+	else
+		*flags &= ~AMDGPU_PTE_EXECUTABLE;
+
+	switch (vm_flags & AMDGPU_VM_MTYPE_MASK) {
+	case AMDGPU_VM_MTYPE_DEFAULT:
+		*flags = AMDGPU_PTE_MTYPE_GFX12(*flags, MTYPE_NC);
+		break;
+	case AMDGPU_VM_MTYPE_NC:
+	default:
+		*flags = AMDGPU_PTE_MTYPE_GFX12(*flags, MTYPE_NC);
+		break;
+	case AMDGPU_VM_MTYPE_UC:
+		*flags = AMDGPU_PTE_MTYPE_GFX12(*flags, MTYPE_UC);
+		break;
+	}
+
+	if (vm_flags & AMDGPU_VM_PAGE_NOALLOC)
+		*flags |= AMDGPU_PTE_NOALLOC;
+	else
+		*flags &= ~AMDGPU_PTE_NOALLOC;
+
+	if (vm_flags & AMDGPU_VM_PAGE_PRT) {
+		*flags |= AMDGPU_PTE_PRT_GFX12;
+		*flags |= AMDGPU_PTE_SNOOPED;
+		*flags |= AMDGPU_PTE_SYSTEM;
+		*flags |= AMDGPU_PTE_IS_PTE;
+		*flags &= ~AMDGPU_PTE_VALID;
+	}
+
+	if (bo && bo->flags & AMDGPU_GEM_CREATE_GFX12_DCC)
+		*flags |= AMDGPU_PTE_DCC;
+
+	if (bo && bo->flags & AMDGPU_GEM_CREATE_UNCACHED)
+		*flags = AMDGPU_PTE_MTYPE_GFX12(*flags, MTYPE_UC);
+}
+
+static unsigned gmc_v12_0_get_vbios_fb_size(struct amdgpu_device *adev)
+{
+	return 0;
+}
+
+static unsigned int gmc_v12_0_get_dcc_alignment(struct amdgpu_device *adev)
+{
+	unsigned int max_tex_channel_caches, alignment;
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(12, 0, 0) &&
+	    amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(12, 0, 1))
+		return 0;
+
+	max_tex_channel_caches = adev->gfx.config.max_texture_channel_caches;
+	if (is_power_of_2(max_tex_channel_caches))
+		alignment = (unsigned int)(max_tex_channel_caches / SZ_4);
+	else
+		alignment = roundup_pow_of_two(max_tex_channel_caches);
+
+	return (unsigned int)(alignment * max_tex_channel_caches * SZ_1K);
+}
+
+static const struct amdgpu_gmc_funcs gmc_v12_0_gmc_funcs = {
+	.flush_gpu_tlb = gmc_v12_0_flush_gpu_tlb,
+	.flush_gpu_tlb_pasid = gmc_v12_0_flush_gpu_tlb_pasid,
+	.emit_flush_gpu_tlb = gmc_v12_0_emit_flush_gpu_tlb,
+	.emit_pasid_mapping = gmc_v12_0_emit_pasid_mapping,
+	.get_vm_pde = gmc_v12_0_get_vm_pde,
+	.get_vm_pte = gmc_v12_0_get_vm_pte,
+	.get_vbios_fb_size = gmc_v12_0_get_vbios_fb_size,
+	.get_dcc_alignment = gmc_v12_0_get_dcc_alignment,
+};
+
+static void gmc_v12_0_set_gmc_funcs(struct amdgpu_device *adev)
+{
+	adev->gmc.gmc_funcs = &gmc_v12_0_gmc_funcs;
+}
+
+static void gmc_v12_0_set_umc_funcs(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, UMC_HWIP, 0)) {
+	case IP_VERSION(8, 14, 0):
+		adev->umc.channel_inst_num = UMC_V8_14_CHANNEL_INSTANCE_NUM;
+		adev->umc.umc_inst_num = UMC_V8_14_UMC_INSTANCE_NUM(adev);
+		adev->umc.node_inst_num = 0;
+		adev->umc.max_ras_err_cnt_per_query = UMC_V8_14_TOTAL_CHANNEL_NUM(adev);
+		adev->umc.channel_offs = UMC_V8_14_PER_CHANNEL_OFFSET;
+		adev->umc.ras = &umc_v8_14_ras;
+		break;
+	default:
+		break;
+	}
+}
+
+
+static void gmc_v12_0_set_mmhub_funcs(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
+	case IP_VERSION(4, 1, 0):
+		adev->mmhub.funcs = &mmhub_v4_1_0_funcs;
+		break;
+	default:
+		break;
+	}
+}
+
+static void gmc_v12_0_set_gfxhub_funcs(struct amdgpu_device *adev)
+{
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		adev->gfxhub.funcs = &gfxhub_v12_0_funcs;
+		break;
+	default:
+		break;
+	}
+}
+
+static int gmc_v12_0_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	gmc_v12_0_set_gfxhub_funcs(adev);
+	gmc_v12_0_set_mmhub_funcs(adev);
+	gmc_v12_0_set_gmc_funcs(adev);
+	gmc_v12_0_set_irq_funcs(adev);
+	gmc_v12_0_set_umc_funcs(adev);
+
+	adev->gmc.shared_aperture_start = 0x2000000000000000ULL;
+	adev->gmc.shared_aperture_end =
+		adev->gmc.shared_aperture_start + (4ULL << 30) - 1;
+	adev->gmc.private_aperture_start = 0x1000000000000000ULL;
+	adev->gmc.private_aperture_end =
+		adev->gmc.private_aperture_start + (4ULL << 30) - 1;
+
+	return 0;
+}
+
+static int gmc_v12_0_late_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int r;
+
+	r = amdgpu_gmc_allocate_vm_inv_eng(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_gmc_ras_late_init(adev);
+	if (r)
+		return r;
+
+	return amdgpu_irq_get(adev, &adev->gmc.vm_fault, 0);
+}
+
+static void gmc_v12_0_vram_gtt_location(struct amdgpu_device *adev,
+					struct amdgpu_gmc *mc)
+{
+	u64 base = 0;
+
+	base = adev->mmhub.funcs->get_fb_location(adev);
+
+	amdgpu_gmc_set_agp_default(adev, mc);
+	amdgpu_gmc_vram_location(adev, &adev->gmc, base);
+	amdgpu_gmc_gart_location(adev, mc, AMDGPU_GART_PLACEMENT_LOW);
+	if (!amdgpu_sriov_vf(adev) && (amdgpu_agp == 1))
+		amdgpu_gmc_agp_location(adev, mc);
+
+	/* base offset of vram pages */
+	if (amdgpu_sriov_vf(adev))
+		adev->vm_manager.vram_base_offset = 0;
+	else
+		adev->vm_manager.vram_base_offset = adev->mmhub.funcs->get_mc_fb_offset(adev);
+}
+
+/**
+ * gmc_v12_0_mc_init - initialize the memory controller driver params
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Look up the amount of vram, vram width, and decide how to place
+ * vram and gart within the GPU's physical address space.
+ * Returns 0 for success.
+ */
+static int gmc_v12_0_mc_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	/* size in MB on si */
+	adev->gmc.mc_vram_size =
+		adev->nbio.funcs->get_memsize(adev) * 1024ULL * 1024ULL;
+	adev->gmc.real_vram_size = adev->gmc.mc_vram_size;
+
+	if (!(adev->flags & AMD_IS_APU)) {
+		r = amdgpu_device_resize_fb_bar(adev);
+		if (r)
+			return r;
+	}
+
+	adev->gmc.aper_base = pci_resource_start(adev->pdev, 0);
+	adev->gmc.aper_size = pci_resource_len(adev->pdev, 0);
+
+#ifdef CONFIG_X86_64
+	if ((adev->flags & AMD_IS_APU) && !amdgpu_passthrough(adev)) {
+		adev->gmc.aper_base = adev->mmhub.funcs->get_mc_fb_offset(adev);
+		adev->gmc.aper_size = adev->gmc.real_vram_size;
+	}
+#endif
+	/* In case the PCI BAR is larger than the actual amount of vram */
+	adev->gmc.visible_vram_size = adev->gmc.aper_size;
+	if (adev->gmc.visible_vram_size > adev->gmc.real_vram_size)
+		adev->gmc.visible_vram_size = adev->gmc.real_vram_size;
+
+	/* set the gart size */
+	if (amdgpu_gart_size == -1) {
+		adev->gmc.gart_size = 512ULL << 20;
+	} else
+		adev->gmc.gart_size = (u64)amdgpu_gart_size << 20;
+
+	gmc_v12_0_vram_gtt_location(adev, &adev->gmc);
+
+	return 0;
+}
+
+static int gmc_v12_0_gart_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (adev->gart.bo) {
+		WARN(1, "PCIE GART already initialized\n");
+		return 0;
+	}
+
+	/* Initialize common gart structure */
+	r = amdgpu_gart_init(adev);
+	if (r)
+		return r;
+
+	adev->gart.table_size = adev->gart.num_gpu_pages * 8;
+	adev->gart.gart_pte_flags = AMDGPU_PTE_MTYPE_GFX12(0ULL, MTYPE_UC) |
+				    AMDGPU_PTE_EXECUTABLE |
+				    AMDGPU_PTE_IS_PTE;
+
+	return amdgpu_gart_table_vram_alloc(adev);
+}
+
+static int gmc_v12_0_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r, vram_width = 0, vram_type = 0, vram_vendor = 0;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	adev->mmhub.funcs->init(adev);
+
+	adev->gfxhub.funcs->init(adev);
+
+	spin_lock_init(&adev->gmc.invalidate_lock);
+
+	r = amdgpu_atomfirmware_get_vram_info(adev,
+					      &vram_width, &vram_type, &vram_vendor);
+	adev->gmc.vram_width = vram_width;
+
+	adev->gmc.vram_type = vram_type;
+	adev->gmc.vram_vendor = vram_vendor;
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+		set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
+		/*
+		 * To fulfill 4-level page support,
+		 * vm size is 256TB (48bit), maximum size,
+		 * block size 512 (9bit)
+		 */
+		amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48);
+		break;
+	default:
+		break;
+	}
+
+	/* This interrupt is VMC page fault.*/
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_VMC,
+			      VMC_1_0__SRCID__VM_FAULT,
+			      &adev->gmc.vm_fault);
+
+	if (r)
+		return r;
+
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_GFX,
+			      UTCL2_1_0__SRCID__FAULT,
+			      &adev->gmc.vm_fault);
+	if (r)
+		return r;
+
+	if (!amdgpu_sriov_vf(adev)) {
+		/* interrupt sent to DF. */
+		r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_DF, 0,
+				      &adev->gmc.ecc_irq);
+		if (r)
+			return r;
+	}
+
+	/*
+	 * Set the internal MC address mask This is the max address of the GPU's
+	 * internal address space.
+	 */
+	adev->gmc.mc_mask = 0xffffffffffffULL; /* 48 bit MC */
+
+	r = dma_set_mask_and_coherent(adev->dev, DMA_BIT_MASK(44));
+	if (r) {
+		printk(KERN_WARNING "amdgpu: No suitable DMA available.\n");
+		return r;
+	}
+
+	adev->need_swiotlb = drm_need_swiotlb(44);
+
+	r = gmc_v12_0_mc_init(adev);
+	if (r)
+		return r;
+
+	amdgpu_gmc_get_vbios_allocations(adev);
+
+	/* Memory manager */
+	r = amdgpu_bo_init(adev);
+	if (r)
+		return r;
+
+	r = gmc_v12_0_gart_init(adev);
+	if (r)
+		return r;
+
+	/*
+	 * number of VMs
+	 * VMID 0 is reserved for System
+	 * amdgpu graphics/compute will use VMIDs 1-7
+	 * amdkfd will use VMIDs 8-15
+	 */
+	adev->vm_manager.first_kfd_vmid = adev->gfx.disable_kq ? 1 : 8;
+
+	amdgpu_vm_manager_init(adev);
+
+	r = amdgpu_gmc_ras_sw_init(adev);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+/**
+ * gmc_v12_0_gart_fini - vm fini callback
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Tears down the driver GART/VM setup (CIK).
+ */
+static void gmc_v12_0_gart_fini(struct amdgpu_device *adev)
+{
+	amdgpu_gart_table_vram_free(adev);
+}
+
+static int gmc_v12_0_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	amdgpu_vm_manager_fini(adev);
+	gmc_v12_0_gart_fini(adev);
+	amdgpu_gem_force_release(adev);
+	amdgpu_bo_fini(adev);
+
+	return 0;
+}
+
+static void gmc_v12_0_init_golden_registers(struct amdgpu_device *adev)
+{
+}
+
+/**
+ * gmc_v12_0_gart_enable - gart enable
+ *
+ * @adev: amdgpu_device pointer
+ */
+static int gmc_v12_0_gart_enable(struct amdgpu_device *adev)
+{
+	int r;
+	bool value;
+
+	if (adev->gart.bo == NULL) {
+		dev_err(adev->dev, "No VRAM object for PCIE GART.\n");
+		return -EINVAL;
+	}
+
+	amdgpu_gtt_mgr_recover(&adev->mman.gtt_mgr);
+
+	r = adev->mmhub.funcs->gart_enable(adev);
+	if (r)
+		return r;
+
+	/* Flush HDP after it is initialized */
+	amdgpu_device_flush_hdp(adev, NULL);
+
+	value = amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS;
+
+	adev->mmhub.funcs->set_fault_enable_default(adev, value);
+	gmc_v12_0_flush_gpu_tlb(adev, 0, AMDGPU_MMHUB0(0), 0);
+
+	dev_info(adev->dev, "PCIE GART of %uM enabled (table at 0x%016llX).\n",
+		 (unsigned)(adev->gmc.gart_size >> 20),
+		 (unsigned long long)amdgpu_bo_gpu_offset(adev->gart.bo));
+
+	return 0;
+}
+
+static int gmc_v12_0_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	/* The sequence of these two function calls matters.*/
+	gmc_v12_0_init_golden_registers(adev);
+
+	r = gmc_v12_0_gart_enable(adev);
+	if (r)
+		return r;
+
+	if (adev->umc.funcs && adev->umc.funcs->init_registers)
+		adev->umc.funcs->init_registers(adev);
+
+	return 0;
+}
+
+/**
+ * gmc_v12_0_gart_disable - gart disable
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * This disables all VM page table.
+ */
+static void gmc_v12_0_gart_disable(struct amdgpu_device *adev)
+{
+	adev->mmhub.funcs->gart_disable(adev);
+}
+
+static int gmc_v12_0_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (amdgpu_sriov_vf(adev)) {
+		/* full access mode, so don't touch any GMC register */
+		DRM_DEBUG("For SRIOV client, shouldn't do anything.\n");
+		return 0;
+	}
+
+	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
+
+	if (adev->gmc.ecc_irq.funcs &&
+		amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC))
+		amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
+
+	gmc_v12_0_gart_disable(adev);
+
+	return 0;
+}
+
+static int gmc_v12_0_suspend(struct amdgpu_ip_block *ip_block)
+{
+	gmc_v12_0_hw_fini(ip_block);
+
+	return 0;
+}
+
+static int gmc_v12_0_resume(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+
+	r = gmc_v12_0_hw_init(ip_block);
+	if (r)
+		return r;
+
+	amdgpu_vmid_reset_all(ip_block->adev);
+
+	return 0;
+}
+
+static bool gmc_v12_0_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	/* MC is always ready in GMC v11.*/
+	return true;
+}
+
+static int gmc_v12_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	/* There is no need to wait for MC idle in GMC v11.*/
+	return 0;
+}
+
+static int gmc_v12_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					   enum amd_clockgating_state state)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	r = adev->mmhub.funcs->set_clockgating(adev, state);
+	if (r)
+		return r;
+
+	return athub_v4_1_0_set_clockgating(adev, state);
+}
+
+static void gmc_v12_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	adev->mmhub.funcs->get_clockgating(adev, flags);
+
+	athub_v4_1_0_get_clockgating(adev, flags);
+}
+
+static int gmc_v12_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					   enum amd_powergating_state state)
+{
+	return 0;
+}
+
+const struct amd_ip_funcs gmc_v12_0_ip_funcs = {
+	.name = "gmc_v12_0",
+	.early_init = gmc_v12_0_early_init,
+	.sw_init = gmc_v12_0_sw_init,
+	.hw_init = gmc_v12_0_hw_init,
+	.late_init = gmc_v12_0_late_init,
+	.sw_fini = gmc_v12_0_sw_fini,
+	.hw_fini = gmc_v12_0_hw_fini,
+	.suspend = gmc_v12_0_suspend,
+	.resume = gmc_v12_0_resume,
+	.is_idle = gmc_v12_0_is_idle,
+	.wait_for_idle = gmc_v12_0_wait_for_idle,
+	.set_clockgating_state = gmc_v12_0_set_clockgating_state,
+	.set_powergating_state = gmc_v12_0_set_powergating_state,
+	.get_clockgating_state = gmc_v12_0_get_clockgating_state,
+};
+
+const struct amdgpu_ip_block_version gmc_v12_0_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_GMC,
+	.major = 12,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &gmc_v12_0_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.h b/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.h
new file mode 100644
index 000000000000..deca93e4a156
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.h
@@ -0,0 +1,30 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __GMC_V12_0_H__
+#define __GMC_V12_0_H__
+
+extern const struct amd_ip_funcs gmc_v12_0_ip_funcs;
+extern const struct amdgpu_ip_block_version gmc_v12_0_ip_block;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
index b7dad4e67813..a8ec95f42926 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c
@@ -43,7 +43,7 @@
 
 static void gmc_v6_0_set_gmc_funcs(struct amdgpu_device *adev);
 static void gmc_v6_0_set_irq_funcs(struct amdgpu_device *adev);
-static int gmc_v6_0_wait_for_idle(void *handle);
+static int gmc_v6_0_wait_for_idle(struct amdgpu_ip_block *ip_block);
 
 MODULE_FIRMWARE("amdgpu/tahiti_mc.bin");
 MODULE_FIRMWARE("amdgpu/pitcairn_mc.bin");
@@ -64,8 +64,13 @@ MODULE_FIRMWARE("amdgpu/si58_mc.bin");
 static void gmc_v6_0_mc_stop(struct amdgpu_device *adev)
 {
 	u32 blackout;
+	struct amdgpu_ip_block *ip_block;
 
-	gmc_v6_0_wait_for_idle((void *)adev);
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GMC);
+	if (!ip_block)
+		return;
+
+	gmc_v6_0_wait_for_idle(ip_block);
 
 	blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL);
 	if (REG_GET_FIELD(blackout, MC_SHARED_BLACKOUT_CNTL, BLACKOUT_MODE) != 1) {
@@ -98,9 +103,7 @@ static void gmc_v6_0_mc_resume(struct amdgpu_device *adev)
 static int gmc_v6_0_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err;
-	bool is_58_fw = false;
 
 	DRM_DEBUG("\n");
 
@@ -120,22 +123,20 @@ static int gmc_v6_0_init_microcode(struct amdgpu_device *adev)
 	case CHIP_HAINAN:
 		chip_name = "hainan";
 		break;
-	default: BUG();
+	default:
+		BUG();
 	}
 
 	/* this memory configuration requires special firmware */
 	if (((RREG32(mmMC_SEQ_MISC0) & 0xff000000) >> 24) == 0x58)
-		is_58_fw = true;
+		chip_name = "si58";
 
-	if (is_58_fw)
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/si58_mc.bin");
-	else
-		snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mc.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gmc.fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gmc.fw, AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_mc.bin", chip_name);
 	if (err) {
 		dev_err(adev->dev,
-		       "si_mc: Failed to load firmware \"%s\"\n",
-		       fw_name);
+		       "si_mc: Failed to load firmware \"%s_mc.bin\"\n",
+		       chip_name);
 		amdgpu_ucode_release(&adev->gmc.fw);
 	}
 	return err;
@@ -178,9 +179,8 @@ static int gmc_v6_0_mc_load_microcode(struct amdgpu_device *adev)
 			WREG32(mmMC_SEQ_IO_DEBUG_DATA, le32_to_cpup(new_io_mc_regs++));
 		}
 		/* load the MC ucode */
-		for (i = 0; i < ucode_size; i++) {
+		for (i = 0; i < ucode_size; i++)
 			WREG32(mmMC_SEQ_SUP_PGM, le32_to_cpup(new_fw_data++));
-		}
 
 		/* put the engine back into the active state */
 		WREG32(mmMC_SEQ_SUP_CNTL, 0x00000008);
@@ -208,15 +208,19 @@ static void gmc_v6_0_vram_gtt_location(struct amdgpu_device *adev,
 				       struct amdgpu_gmc *mc)
 {
 	u64 base = RREG32(mmMC_VM_FB_LOCATION) & 0xFFFF;
+
 	base <<= 24;
 
+	amdgpu_gmc_set_agp_default(adev, mc);
 	amdgpu_gmc_vram_location(adev, mc, base);
-	amdgpu_gmc_gart_location(adev, mc);
+	amdgpu_gmc_gart_location(adev, mc, AMDGPU_GART_PLACEMENT_LOW);
 }
 
 static void gmc_v6_0_mc_program(struct amdgpu_device *adev)
 {
 	int i, j;
+	struct amdgpu_ip_block *ip_block;
+
 
 	/* Initialize HDP */
 	for (i = 0, j = 0; i < 32; i++, j += 0x6) {
@@ -228,9 +232,12 @@ static void gmc_v6_0_mc_program(struct amdgpu_device *adev)
 	}
 	WREG32(mmHDP_REG_COHERENCY_FLUSH_CNTL, 0);
 
-	if (gmc_v6_0_wait_for_idle((void *)adev)) {
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GMC);
+	if (!ip_block)
+		return;
+
+	if (gmc_v6_0_wait_for_idle(ip_block))
 		dev_warn(adev->dev, "Wait for MC idle timedout !\n");
-	}
 
 	if (adev->mode_info.num_crtc) {
 		u32 tmp;
@@ -242,7 +249,7 @@ static void gmc_v6_0_mc_program(struct amdgpu_device *adev)
 
 		/* disable VGA render */
 		tmp = RREG32(mmVGA_RENDER_CONTROL);
-		tmp &= ~VGA_VSTATUS_CNTL;
+		tmp &= VGA_RENDER_CONTROL__VGA_VSTATUS_CNTL_MASK;
 		WREG32(mmVGA_RENDER_CONTROL, tmp);
 	}
 	/* Update configuration */
@@ -253,12 +260,11 @@ static void gmc_v6_0_mc_program(struct amdgpu_device *adev)
 	WREG32(mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR,
 	       adev->mem_scratch.gpu_addr >> 12);
 	WREG32(mmMC_VM_AGP_BASE, 0);
-	WREG32(mmMC_VM_AGP_TOP, 0x0FFFFFFF);
-	WREG32(mmMC_VM_AGP_BOT, 0x0FFFFFFF);
+	WREG32(mmMC_VM_AGP_TOP, adev->gmc.agp_end >> 22);
+	WREG32(mmMC_VM_AGP_BOT, adev->gmc.agp_start >> 22);
 
-	if (gmc_v6_0_wait_for_idle((void *)adev)) {
+	if (gmc_v6_0_wait_for_idle(ip_block))
 		dev_warn(adev->dev, "Wait for MC idle timedout !\n");
-	}
 }
 
 static int gmc_v6_0_mc_init(struct amdgpu_device *adev)
@@ -269,13 +275,13 @@ static int gmc_v6_0_mc_init(struct amdgpu_device *adev)
 	int r;
 
 	tmp = RREG32(mmMC_ARB_RAMCFG);
-	if (tmp & (1 << 11)) {
+	if (tmp & (1 << 11))
 		chansize = 16;
-	} else if (tmp & MC_ARB_RAMCFG__CHANSIZE_MASK) {
+	else if (tmp & MC_ARB_RAMCFG__CHANSIZE_MASK)
 		chansize = 64;
-	} else {
+	else
 		chansize = 32;
-	}
+
 	tmp = RREG32(mmMC_SHARED_CHMAP);
 	switch ((tmp & MC_SHARED_CHMAP__NOOFCHAN_MASK) >> MC_SHARED_CHMAP__NOOFCHAN__SHIFT) {
 	case 0:
@@ -352,7 +358,7 @@ static void gmc_v6_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 }
 
 static uint64_t gmc_v6_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
-					    unsigned vmid, uint64_t pd_addr)
+					    unsigned int vmid, uint64_t pd_addr)
 {
 	uint32_t reg;
 
@@ -376,7 +382,9 @@ static void gmc_v6_0_get_vm_pde(struct amdgpu_device *adev, int level,
 }
 
 static void gmc_v6_0_get_vm_pte(struct amdgpu_device *adev,
-				struct amdgpu_bo_va_mapping *mapping,
+				struct amdgpu_vm *vm,
+				struct amdgpu_bo *bo,
+				uint32_t vm_flags,
 				uint64_t *flags)
 {
 	*flags &= ~AMDGPU_PTE_EXECUTABLE;
@@ -405,11 +413,11 @@ static void gmc_v6_0_set_fault_enable_default(struct amdgpu_device *adev,
 }
 
  /**
-   + * gmc_v8_0_set_prt - set PRT VM fault
-   + *
-   + * @adev: amdgpu_device pointer
-   + * @enable: enable/disable VM fault handling for PRT
-   +*/
+  * gmc_v8_0_set_prt() - set PRT VM fault
+  *
+  * @adev: amdgpu_device pointer
+  * @enable: enable/disable VM fault handling for PRT
+  */
 static void gmc_v6_0_set_prt(struct amdgpu_device *adev, bool enable)
 {
 	u32 tmp;
@@ -435,9 +443,10 @@ static void gmc_v6_0_set_prt(struct amdgpu_device *adev, bool enable)
 	WREG32(mmVM_PRT_CNTL, tmp);
 
 	if (enable) {
-		uint32_t low = AMDGPU_VA_RESERVED_SIZE >> AMDGPU_GPU_PAGE_SHIFT;
+		uint32_t low = AMDGPU_VA_RESERVED_BOTTOM >>
+			AMDGPU_GPU_PAGE_SHIFT;
 		uint32_t high = adev->vm_manager.max_pfn -
-			(AMDGPU_VA_RESERVED_SIZE >> AMDGPU_GPU_PAGE_SHIFT);
+			(AMDGPU_VA_RESERVED_TOP >> AMDGPU_GPU_PAGE_SHIFT);
 
 		WREG32(mmVM_PRT_APERTURE0_LOW_ADDR, low);
 		WREG32(mmVM_PRT_APERTURE1_LOW_ADDR, low);
@@ -547,7 +556,7 @@ static int gmc_v6_0_gart_enable(struct amdgpu_device *adev)
 
 	gmc_v6_0_flush_gpu_tlb(adev, 0, 0, 0);
 	dev_info(adev->dev, "PCIE GART of %uM enabled (table at 0x%016llX).\n",
-		 (unsigned)(adev->gmc.gart_size >> 20),
+		 (unsigned int)(adev->gmc.gart_size >> 20),
 		 (unsigned long long)table_addr);
 	return 0;
 }
@@ -601,36 +610,33 @@ static void gmc_v6_0_gart_disable(struct amdgpu_device *adev)
 }
 
 static void gmc_v6_0_vm_decode_fault(struct amdgpu_device *adev,
-				     u32 status, u32 addr, u32 mc_client)
+				     u32 status, u32 addr)
 {
 	u32 mc_id;
 	u32 vmid = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, VMID);
 	u32 protections = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
 					PROTECTIONS);
-	char block[5] = { mc_client >> 24, (mc_client >> 16) & 0xff,
-		(mc_client >> 8) & 0xff, mc_client & 0xff, 0 };
 
 	mc_id = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
 			      MEMORY_CLIENT_ID);
 
-	dev_err(adev->dev, "VM fault (0x%02x, vmid %d) at page %u, %s from '%s' (0x%08x) (%d)\n",
+	dev_err(adev->dev, "VM fault (0x%02x, vmid %d) at page %u, %s from %d\n",
 	       protections, vmid, addr,
 	       REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
 			     MEMORY_CLIENT_RW) ?
-	       "write" : "read", block, mc_client, mc_id);
+	       "write" : "read", mc_id);
 }
 
-/*
 static const u32 mc_cg_registers[] = {
-	MC_HUB_MISC_HUB_CG,
-	MC_HUB_MISC_SIP_CG,
-	MC_HUB_MISC_VM_CG,
-	MC_XPB_CLK_GAT,
-	ATC_MISC_CG,
-	MC_CITF_MISC_WR_CG,
-	MC_CITF_MISC_RD_CG,
-	MC_CITF_MISC_VM_CG,
-	VM_L2_CG,
+	mmMC_HUB_MISC_HUB_CG,
+	mmMC_HUB_MISC_SIP_CG,
+	mmMC_HUB_MISC_VM_CG,
+	mmMC_XPB_CLK_GAT,
+	mmATC_MISC_CG,
+	mmMC_CITF_MISC_WR_CG,
+	mmMC_CITF_MISC_RD_CG,
+	mmMC_CITF_MISC_VM_CG,
+	mmVM_L2_CG,
 };
 
 static const u32 mc_cg_ls_en[] = {
@@ -665,7 +671,7 @@ static void gmc_v6_0_enable_mc_ls(struct amdgpu_device *adev,
 
 	for (i = 0; i < ARRAY_SIZE(mc_cg_registers); i++) {
 		orig = data = RREG32(mc_cg_registers[i]);
-		if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_MC_LS))
+		if (enable && (adev->cg_flags & AMD_CG_SUPPORT_MC_LS))
 			data |= mc_cg_ls_en[i];
 		else
 			data &= ~mc_cg_ls_en[i];
@@ -682,7 +688,7 @@ static void gmc_v6_0_enable_mc_mgcg(struct amdgpu_device *adev,
 
 	for (i = 0; i < ARRAY_SIZE(mc_cg_registers); i++) {
 		orig = data = RREG32(mc_cg_registers[i]);
-		if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_MC_MGCG))
+		if (enable && (adev->cg_flags & AMD_CG_SUPPORT_MC_MGCG))
 			data |= mc_cg_en[i];
 		else
 			data &= ~mc_cg_en[i];
@@ -698,7 +704,7 @@ static void gmc_v6_0_enable_bif_mgls(struct amdgpu_device *adev,
 
 	orig = data = RREG32_PCIE(ixPCIE_CNTL2);
 
-	if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_BIF_LS)) {
+	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_LS)) {
 		data = REG_SET_FIELD(data, PCIE_CNTL2, SLV_MEM_LS_EN, 1);
 		data = REG_SET_FIELD(data, PCIE_CNTL2, MST_MEM_LS_EN, 1);
 		data = REG_SET_FIELD(data, PCIE_CNTL2, REPLAY_MEM_LS_EN, 1);
@@ -721,7 +727,7 @@ static void gmc_v6_0_enable_hdp_mgcg(struct amdgpu_device *adev,
 
 	orig = data = RREG32(mmHDP_HOST_PATH_CNTL);
 
-	if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_HDP_MGCG))
+	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_HDP_MGCG))
 		data = REG_SET_FIELD(data, HDP_HOST_PATH_CNTL, CLOCK_GATING_DIS, 0);
 	else
 		data = REG_SET_FIELD(data, HDP_HOST_PATH_CNTL, CLOCK_GATING_DIS, 1);
@@ -737,7 +743,7 @@ static void gmc_v6_0_enable_hdp_ls(struct amdgpu_device *adev,
 
 	orig = data = RREG32(mmHDP_MEM_POWER_LS);
 
-	if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_HDP_LS))
+	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_HDP_LS))
 		data = REG_SET_FIELD(data, HDP_MEM_POWER_LS, LS_ENABLE, 1);
 	else
 		data = REG_SET_FIELD(data, HDP_MEM_POWER_LS, LS_ENABLE, 0);
@@ -745,7 +751,6 @@ static void gmc_v6_0_enable_hdp_ls(struct amdgpu_device *adev,
 	if (orig != data)
 		WREG32(mmHDP_MEM_POWER_LS, data);
 }
-*/
 
 static int gmc_v6_0_convert_vram_type(int mc_seq_vram_type)
 {
@@ -767,9 +772,9 @@ static int gmc_v6_0_convert_vram_type(int mc_seq_vram_type)
 	}
 }
 
-static int gmc_v6_0_early_init(void *handle)
+static int gmc_v6_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v6_0_set_gmc_funcs(adev);
 	gmc_v6_0_set_irq_funcs(adev);
@@ -777,9 +782,9 @@ static int gmc_v6_0_early_init(void *handle)
 	return 0;
 }
 
-static int gmc_v6_0_late_init(void *handle)
+static int gmc_v6_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS)
 		return amdgpu_irq_get(adev, &adev->gmc.vm_fault, 0);
@@ -787,15 +792,16 @@ static int gmc_v6_0_late_init(void *handle)
 		return 0;
 }
 
-static unsigned gmc_v6_0_get_vbios_fb_size(struct amdgpu_device *adev)
+static unsigned int gmc_v6_0_get_vbios_fb_size(struct amdgpu_device *adev)
 {
 	u32 d1vga_control = RREG32(mmD1VGA_CONTROL);
-	unsigned size;
+	unsigned int size;
 
 	if (REG_GET_FIELD(d1vga_control, D1VGA_CONTROL, D1VGA_MODE_ENABLE)) {
 		size = AMDGPU_VBIOS_VGA_ALLOCATION;
 	} else {
 		u32 viewport = RREG32(mmVIEWPORT_SIZE);
+
 		size = (REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_HEIGHT) *
 			REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_WIDTH) *
 			4);
@@ -803,17 +809,18 @@ static unsigned gmc_v6_0_get_vbios_fb_size(struct amdgpu_device *adev)
 	return size;
 }
 
-static int gmc_v6_0_sw_init(void *handle)
+static int gmc_v6_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	adev->num_vmhubs = 1;
+	set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
 
 	if (adev->flags & AMD_IS_APU) {
 		adev->gmc.vram_type = AMDGPU_VRAM_TYPE_UNKNOWN;
 	} else {
 		u32 tmp = RREG32(mmMC_SEQ_MISC0);
+
 		tmp &= MC_SEQ_MISC0__MT__MASK;
 		adev->gmc.vram_type = gmc_v6_0_convert_vram_type(tmp);
 	}
@@ -879,9 +886,9 @@ static int gmc_v6_0_sw_init(void *handle)
 	return 0;
 }
 
-static int gmc_v6_0_sw_fini(void *handle)
+static int gmc_v6_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_gem_force_release(adev);
 	amdgpu_vm_manager_fini(adev);
@@ -892,10 +899,10 @@ static int gmc_v6_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int gmc_v6_0_hw_init(void *handle)
+static int gmc_v6_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v6_0_mc_program(adev);
 
@@ -913,13 +920,13 @@ static int gmc_v6_0_hw_init(void *handle)
 
 	if (amdgpu_emu_mode == 1)
 		return amdgpu_gmc_vram_checking(adev);
-	else
-		return r;
+
+	return 0;
 }
 
-static int gmc_v6_0_hw_fini(void *handle)
+static int gmc_v6_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
 	gmc_v6_0_gart_disable(adev);
@@ -927,21 +934,19 @@ static int gmc_v6_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int gmc_v6_0_suspend(void *handle)
+static int gmc_v6_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	gmc_v6_0_hw_fini(adev);
+	gmc_v6_0_hw_fini(ip_block);
 
 	return 0;
 }
 
-static int gmc_v6_0_resume(void *handle)
+static int gmc_v6_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	r = gmc_v6_0_hw_init(adev);
+	r = gmc_v6_0_hw_init(ip_block);
 	if (r)
 		return r;
 
@@ -950,9 +955,10 @@ static int gmc_v6_0_resume(void *handle)
 	return 0;
 }
 
-static bool gmc_v6_0_is_idle(void *handle)
+static bool gmc_v6_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (tmp & (SRBM_STATUS__MCB_BUSY_MASK | SRBM_STATUS__MCB_NON_DISPLAY_BUSY_MASK |
@@ -962,13 +968,13 @@ static bool gmc_v6_0_is_idle(void *handle)
 	return true;
 }
 
-static int gmc_v6_0_wait_for_idle(void *handle)
+static int gmc_v6_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	unsigned i;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	unsigned int i;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		if (gmc_v6_0_is_idle(handle))
+		if (gmc_v6_0_is_idle(ip_block))
 			return 0;
 		udelay(1);
 	}
@@ -976,9 +982,10 @@ static int gmc_v6_0_wait_for_idle(void *handle)
 
 }
 
-static int gmc_v6_0_soft_reset(void *handle)
+static int gmc_v6_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+
 	u32 srbm_soft_reset = 0;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
@@ -995,10 +1002,9 @@ static int gmc_v6_0_soft_reset(void *handle)
 
 	if (srbm_soft_reset) {
 		gmc_v6_0_mc_stop(adev);
-		if (gmc_v6_0_wait_for_idle(adev)) {
-			dev_warn(adev->dev, "Wait for GMC idle timed out !\n");
-		}
 
+		if (gmc_v6_0_wait_for_idle(ip_block))
+			dev_warn(adev->dev, "Wait for GMC idle timed out !\n");
 
 		tmp = RREG32(mmSRBM_SOFT_RESET);
 		tmp |= srbm_soft_reset;
@@ -1023,7 +1029,7 @@ static int gmc_v6_0_soft_reset(void *handle)
 
 static int gmc_v6_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 					     struct amdgpu_irq_src *src,
-					     unsigned type,
+					     unsigned int type,
 					     enum amdgpu_interrupt_state state)
 {
 	u32 tmp;
@@ -1064,6 +1070,12 @@ static int gmc_v6_0_process_interrupt(struct amdgpu_device *adev,
 {
 	u32 addr, status;
 
+	/* Delegate to the soft IRQ handler ring */
+	if (adev->irq.ih_soft.enabled && entry->ih != &adev->irq.ih_soft) {
+		amdgpu_irq_delegate(adev, entry, 4);
+		return 1;
+	}
+
 	addr = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_ADDR);
 	status = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_STATUS);
 	WREG32_P(mmVM_CONTEXT1_CNTL2, 1, ~1);
@@ -1071,6 +1083,10 @@ static int gmc_v6_0_process_interrupt(struct amdgpu_device *adev,
 	if (!addr && !status)
 		return 0;
 
+	amdgpu_vm_update_fault_cache(adev, entry->pasid,
+				     ((u64)addr) << AMDGPU_GPU_PAGE_SHIFT,
+				     status, AMDGPU_GFXHUB(0));
+
 	if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_FIRST)
 		gmc_v6_0_set_fault_enable_default(adev, false);
 
@@ -1081,19 +1097,33 @@ static int gmc_v6_0_process_interrupt(struct amdgpu_device *adev,
 			addr);
 		dev_err(adev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n",
 			status);
-		gmc_v6_0_vm_decode_fault(adev, status, addr, 0);
+		gmc_v6_0_vm_decode_fault(adev, status, addr);
 	}
 
 	return 0;
 }
 
-static int gmc_v6_0_set_clockgating_state(void *handle,
+static int gmc_v6_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
+	struct amdgpu_device *adev = ip_block->adev;
+	bool gate = false;
+
+	if (state == AMD_CG_STATE_GATE)
+		gate = true;
+
+	if (!(adev->flags & AMD_IS_APU)) {
+		gmc_v6_0_enable_mc_mgcg(adev, gate);
+		gmc_v6_0_enable_mc_ls(adev, gate);
+	}
+	gmc_v6_0_enable_bif_mgls(adev, gate);
+	gmc_v6_0_enable_hdp_mgcg(adev, gate);
+	gmc_v6_0_enable_hdp_ls(adev, gate);
+
 	return 0;
 }
 
-static int gmc_v6_0_set_powergating_state(void *handle,
+static int gmc_v6_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -1141,8 +1171,7 @@ static void gmc_v6_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->gmc.vm_fault.funcs = &gmc_v6_0_irq_funcs;
 }
 
-const struct amdgpu_ip_block_version gmc_v6_0_ip_block =
-{
+const struct amdgpu_ip_block_version gmc_v6_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GMC,
 	.major = 6,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
index 402960b0174e..fbd0bf147f50 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c
@@ -52,22 +52,20 @@
 
 static void gmc_v7_0_set_gmc_funcs(struct amdgpu_device *adev);
 static void gmc_v7_0_set_irq_funcs(struct amdgpu_device *adev);
-static int gmc_v7_0_wait_for_idle(void *handle);
+static int gmc_v7_0_wait_for_idle(struct amdgpu_ip_block *ip_block);
 
 MODULE_FIRMWARE("amdgpu/bonaire_mc.bin");
 MODULE_FIRMWARE("amdgpu/hawaii_mc.bin");
 MODULE_FIRMWARE("amdgpu/topaz_mc.bin");
 
-static const u32 golden_settings_iceland_a11[] =
-{
+static const u32 golden_settings_iceland_a11[] = {
 	mmVM_PRT_APERTURE0_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE1_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE2_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE3_LOW_ADDR, 0x0fffffff, 0x0fffffff
 };
 
-static const u32 iceland_mgcg_cgcg_init[] =
-{
+static const u32 iceland_mgcg_cgcg_init[] = {
 	mmMC_MEM_POWER_LS, 0xffffffff, 0x00000104
 };
 
@@ -89,9 +87,14 @@ static void gmc_v7_0_init_golden_registers(struct amdgpu_device *adev)
 
 static void gmc_v7_0_mc_stop(struct amdgpu_device *adev)
 {
+	struct amdgpu_ip_block *ip_block;
 	u32 blackout;
 
-	gmc_v7_0_wait_for_idle((void *)adev);
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GMC);
+	if (!ip_block)
+		return;
+
+	gmc_v7_0_wait_for_idle(ip_block);
 
 	blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL);
 	if (REG_GET_FIELD(blackout, MC_SHARED_BLACKOUT_CNTL, BLACKOUT_MODE) != 1) {
@@ -132,7 +135,6 @@ static void gmc_v7_0_mc_resume(struct amdgpu_device *adev)
 static int gmc_v7_0_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err;
 
 	DRM_DEBUG("\n");
@@ -151,14 +153,14 @@ static int gmc_v7_0_init_microcode(struct amdgpu_device *adev)
 	case CHIP_KABINI:
 	case CHIP_MULLINS:
 		return 0;
-	default: BUG();
+	default:
+		return -EINVAL;
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mc.bin", chip_name);
-
-	err = amdgpu_ucode_request(adev, &adev->gmc.fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gmc.fw, AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_mc.bin", chip_name);
 	if (err) {
-		pr_err("cik_mc: Failed to load firmware \"%s\"\n", fw_name);
+		pr_err("cik_mc: Failed to load firmware \"%s_mc.bin\"\n", chip_name);
 		amdgpu_ucode_release(&adev->gmc.fw);
 	}
 	return err;
@@ -237,10 +239,12 @@ static void gmc_v7_0_vram_gtt_location(struct amdgpu_device *adev,
 				       struct amdgpu_gmc *mc)
 {
 	u64 base = RREG32(mmMC_VM_FB_LOCATION) & 0xFFFF;
+
 	base <<= 24;
 
+	amdgpu_gmc_set_agp_default(adev, mc);
 	amdgpu_gmc_vram_location(adev, mc, base);
-	amdgpu_gmc_gart_location(adev, mc);
+	amdgpu_gmc_gart_location(adev, mc, AMDGPU_GART_PLACEMENT_BEST_FIT);
 }
 
 /**
@@ -253,9 +257,14 @@ static void gmc_v7_0_vram_gtt_location(struct amdgpu_device *adev,
  */
 static void gmc_v7_0_mc_program(struct amdgpu_device *adev)
 {
+	struct amdgpu_ip_block *ip_block;
 	u32 tmp;
 	int i, j;
 
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GMC);
+	if (!ip_block)
+		return;
+
 	/* Initialize HDP */
 	for (i = 0, j = 0; i < 32; i++, j += 0x6) {
 		WREG32((0xb05 + j), 0x00000000);
@@ -266,9 +275,9 @@ static void gmc_v7_0_mc_program(struct amdgpu_device *adev)
 	}
 	WREG32(mmHDP_REG_COHERENCY_FLUSH_CNTL, 0);
 
-	if (gmc_v7_0_wait_for_idle((void *)adev)) {
+	if (gmc_v7_0_wait_for_idle(ip_block))
 		dev_warn(adev->dev, "Wait for MC idle timedout !\n");
-	}
+
 	if (adev->mode_info.num_crtc) {
 		/* Lockout access through VGA aperture*/
 		tmp = RREG32(mmVGA_HDP_CONTROL);
@@ -288,11 +297,10 @@ static void gmc_v7_0_mc_program(struct amdgpu_device *adev)
 	WREG32(mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR,
 	       adev->mem_scratch.gpu_addr >> 12);
 	WREG32(mmMC_VM_AGP_BASE, 0);
-	WREG32(mmMC_VM_AGP_TOP, 0x0FFFFFFF);
-	WREG32(mmMC_VM_AGP_BOT, 0x0FFFFFFF);
-	if (gmc_v7_0_wait_for_idle((void *)adev)) {
+	WREG32(mmMC_VM_AGP_TOP, adev->gmc.agp_end >> 22);
+	WREG32(mmMC_VM_AGP_BOT, adev->gmc.agp_start >> 22);
+	if (gmc_v7_0_wait_for_idle(ip_block))
 		dev_warn(adev->dev, "Wait for MC idle timedout !\n");
-	}
 
 	WREG32(mmBIF_FB_EN, BIF_FB_EN__FB_READ_EN_MASK | BIF_FB_EN__FB_WRITE_EN_MASK);
 
@@ -324,11 +332,11 @@ static int gmc_v7_0_mc_init(struct amdgpu_device *adev)
 
 		/* Get VRAM informations */
 		tmp = RREG32(mmMC_ARB_RAMCFG);
-		if (REG_GET_FIELD(tmp, MC_ARB_RAMCFG, CHANSIZE)) {
+		if (REG_GET_FIELD(tmp, MC_ARB_RAMCFG, CHANSIZE))
 			chansize = 64;
-		} else {
+		else
 			chansize = 32;
-		}
+
 		tmp = RREG32(mmMC_SHARED_CHMAP);
 		switch (REG_GET_FIELD(tmp, MC_SHARED_CHMAP, NOOFCHAN)) {
 		case 0:
@@ -419,31 +427,27 @@ static int gmc_v7_0_mc_init(struct amdgpu_device *adev)
  * @pasid: pasid to be flush
  * @flush_type: type of flush
  * @all_hub: flush all hubs
+ * @inst: is used to select which instance of KIQ to use for the invalidation
  *
  * Flush the TLB for the requested pasid.
  */
-static int gmc_v7_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
-					uint16_t pasid, uint32_t flush_type,
-					bool all_hub)
+static void gmc_v7_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
+					 uint16_t pasid, uint32_t flush_type,
+					 bool all_hub, uint32_t inst)
 {
+	u32 mask = 0x0;
 	int vmid;
-	unsigned int tmp;
-
-	if (amdgpu_in_reset(adev))
-		return -EIO;
 
 	for (vmid = 1; vmid < 16; vmid++) {
+		u32 tmp = RREG32(mmATC_VMID0_PASID_MAPPING + vmid);
 
-		tmp = RREG32(mmATC_VMID0_PASID_MAPPING + vmid);
 		if ((tmp & ATC_VMID0_PASID_MAPPING__VALID_MASK) &&
-			(tmp & ATC_VMID0_PASID_MAPPING__PASID_MASK) == pasid) {
-			WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid);
-			RREG32(mmVM_INVALIDATE_RESPONSE);
-			break;
-		}
+		    (tmp & ATC_VMID0_PASID_MAPPING__PASID_MASK) == pasid)
+			mask |= 1 << vmid;
 	}
 
-	return 0;
+	WREG32(mmVM_INVALIDATE_REQUEST, mask);
+	RREG32(mmVM_INVALIDATE_RESPONSE);
 }
 
 /*
@@ -471,7 +475,7 @@ static void gmc_v7_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 }
 
 static uint64_t gmc_v7_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
-					    unsigned vmid, uint64_t pd_addr)
+					    unsigned int vmid, uint64_t pd_addr)
 {
 	uint32_t reg;
 
@@ -487,8 +491,8 @@ static uint64_t gmc_v7_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 	return pd_addr;
 }
 
-static void gmc_v7_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
-					unsigned pasid)
+static void gmc_v7_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned int vmid,
+					unsigned int pasid)
 {
 	amdgpu_ring_emit_wreg(ring, mmIH_VMID_0_LUT + vmid, pasid);
 }
@@ -500,7 +504,9 @@ static void gmc_v7_0_get_vm_pde(struct amdgpu_device *adev, int level,
 }
 
 static void gmc_v7_0_get_vm_pte(struct amdgpu_device *adev,
-				struct amdgpu_bo_va_mapping *mapping,
+				struct amdgpu_vm *vm,
+				struct amdgpu_bo *bo,
+				uint32_t vm_flags,
 				uint64_t *flags)
 {
 	*flags &= ~AMDGPU_PTE_EXECUTABLE;
@@ -567,9 +573,10 @@ static void gmc_v7_0_set_prt(struct amdgpu_device *adev, bool enable)
 	WREG32(mmVM_PRT_CNTL, tmp);
 
 	if (enable) {
-		uint32_t low = AMDGPU_VA_RESERVED_SIZE >> AMDGPU_GPU_PAGE_SHIFT;
+		uint32_t low = AMDGPU_VA_RESERVED_BOTTOM >>
+			AMDGPU_GPU_PAGE_SHIFT;
 		uint32_t high = adev->vm_manager.max_pfn -
-			(AMDGPU_VA_RESERVED_SIZE >> AMDGPU_GPU_PAGE_SHIFT);
+			(AMDGPU_VA_RESERVED_TOP >> AMDGPU_GPU_PAGE_SHIFT);
 
 		WREG32(mmVM_PRT_APERTURE0_LOW_ADDR, low);
 		WREG32(mmVM_PRT_APERTURE1_LOW_ADDR, low);
@@ -699,7 +706,7 @@ static int gmc_v7_0_gart_enable(struct amdgpu_device *adev)
 
 	gmc_v7_0_flush_gpu_tlb(adev, 0, 0, 0);
 	DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
-		 (unsigned)(adev->gmc.gart_size >> 20),
+		 (unsigned int)(adev->gmc.gart_size >> 20),
 		 (unsigned long long)table_addr);
 	return 0;
 }
@@ -760,7 +767,7 @@ static void gmc_v7_0_gart_disable(struct amdgpu_device *adev)
  * Print human readable fault information (CIK).
  */
 static void gmc_v7_0_vm_decode_fault(struct amdgpu_device *adev, u32 status,
-				     u32 addr, u32 mc_client, unsigned pasid)
+				     u32 addr, u32 mc_client, unsigned int pasid)
 {
 	u32 vmid = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, VMID);
 	u32 protections = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
@@ -927,9 +934,9 @@ static int gmc_v7_0_convert_vram_type(int mc_seq_vram_type)
 	}
 }
 
-static int gmc_v7_0_early_init(void *handle)
+static int gmc_v7_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v7_0_set_gmc_funcs(adev);
 	gmc_v7_0_set_irq_funcs(adev);
@@ -941,13 +948,14 @@ static int gmc_v7_0_early_init(void *handle)
 		adev->gmc.shared_aperture_end + 1;
 	adev->gmc.private_aperture_end =
 		adev->gmc.private_aperture_start + (4ULL << 30) - 1;
+	adev->gmc.noretry_flags = AMDGPU_VM_NORETRY_FLAGS_TF;
 
 	return 0;
 }
 
-static int gmc_v7_0_late_init(void *handle)
+static int gmc_v7_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS)
 		return amdgpu_irq_get(adev, &adev->gmc.vm_fault, 0);
@@ -955,15 +963,16 @@ static int gmc_v7_0_late_init(void *handle)
 		return 0;
 }
 
-static unsigned gmc_v7_0_get_vbios_fb_size(struct amdgpu_device *adev)
+static unsigned int gmc_v7_0_get_vbios_fb_size(struct amdgpu_device *adev)
 {
 	u32 d1vga_control = RREG32(mmD1VGA_CONTROL);
-	unsigned size;
+	unsigned int size;
 
 	if (REG_GET_FIELD(d1vga_control, D1VGA_CONTROL, D1VGA_MODE_ENABLE)) {
 		size = AMDGPU_VBIOS_VGA_ALLOCATION;
 	} else {
 		u32 viewport = RREG32(mmVIEWPORT_SIZE);
+
 		size = (REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_HEIGHT) *
 			REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_WIDTH) *
 			4);
@@ -972,17 +981,18 @@ static unsigned gmc_v7_0_get_vbios_fb_size(struct amdgpu_device *adev)
 	return size;
 }
 
-static int gmc_v7_0_sw_init(void *handle)
+static int gmc_v7_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	adev->num_vmhubs = 1;
+	set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
 
 	if (adev->flags & AMD_IS_APU) {
 		adev->gmc.vram_type = AMDGPU_VRAM_TYPE_UNKNOWN;
 	} else {
 		u32 tmp = RREG32(mmMC_SEQ_MISC0);
+
 		tmp &= MC_SEQ_MISC0__MT__MASK;
 		adev->gmc.vram_type = gmc_v7_0_convert_vram_type(tmp);
 	}
@@ -1058,14 +1068,14 @@ static int gmc_v7_0_sw_init(void *handle)
 					GFP_KERNEL);
 	if (!adev->gmc.vm_fault_info)
 		return -ENOMEM;
-	atomic_set(&adev->gmc.vm_fault_info_updated, 0);
+	atomic_set_release(&adev->gmc.vm_fault_info_updated, 0);
 
 	return 0;
 }
 
-static int gmc_v7_0_sw_fini(void *handle)
+static int gmc_v7_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_gem_force_release(adev);
 	amdgpu_vm_manager_fini(adev);
@@ -1077,10 +1087,10 @@ static int gmc_v7_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int gmc_v7_0_hw_init(void *handle)
+static int gmc_v7_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v7_0_init_golden_registers(adev);
 
@@ -1100,13 +1110,13 @@ static int gmc_v7_0_hw_init(void *handle)
 
 	if (amdgpu_emu_mode == 1)
 		return amdgpu_gmc_vram_checking(adev);
-	else
-		return r;
+
+	return 0;
 }
 
-static int gmc_v7_0_hw_fini(void *handle)
+static int gmc_v7_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
 	gmc_v7_0_gart_disable(adev);
@@ -1114,32 +1124,29 @@ static int gmc_v7_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int gmc_v7_0_suspend(void *handle)
+static int gmc_v7_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	gmc_v7_0_hw_fini(adev);
+	gmc_v7_0_hw_fini(ip_block);
 
 	return 0;
 }
 
-static int gmc_v7_0_resume(void *handle)
+static int gmc_v7_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = gmc_v7_0_hw_init(adev);
+	r = gmc_v7_0_hw_init(ip_block);
 	if (r)
 		return r;
 
-	amdgpu_vmid_reset_all(adev);
+	amdgpu_vmid_reset_all(ip_block->adev);
 
 	return 0;
 }
 
-static bool gmc_v7_0_is_idle(void *handle)
+static bool gmc_v7_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (tmp & (SRBM_STATUS__MCB_BUSY_MASK | SRBM_STATUS__MCB_NON_DISPLAY_BUSY_MASK |
@@ -1149,20 +1156,13 @@ static bool gmc_v7_0_is_idle(void *handle)
 	return true;
 }
 
-static int gmc_v7_0_wait_for_idle(void *handle)
+static int gmc_v7_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	unsigned i;
-	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	unsigned int i;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
-		/* read MC_STATUS */
-		tmp = RREG32(mmSRBM_STATUS) & (SRBM_STATUS__MCB_BUSY_MASK |
-					       SRBM_STATUS__MCB_NON_DISPLAY_BUSY_MASK |
-					       SRBM_STATUS__MCC_BUSY_MASK |
-					       SRBM_STATUS__MCD_BUSY_MASK |
-					       SRBM_STATUS__VMC_BUSY_MASK);
-		if (!tmp)
+		if (gmc_v7_0_is_idle(ip_block))
 			return 0;
 		udelay(1);
 	}
@@ -1170,9 +1170,9 @@ static int gmc_v7_0_wait_for_idle(void *handle)
 
 }
 
-static int gmc_v7_0_soft_reset(void *handle)
+static int gmc_v7_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 srbm_soft_reset = 0;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
@@ -1189,10 +1189,8 @@ static int gmc_v7_0_soft_reset(void *handle)
 
 	if (srbm_soft_reset) {
 		gmc_v7_0_mc_stop(adev);
-		if (gmc_v7_0_wait_for_idle((void *)adev)) {
+		if (gmc_v7_0_wait_for_idle(ip_block))
 			dev_warn(adev->dev, "Wait for GMC idle timed out !\n");
-		}
-
 
 		tmp = RREG32(mmSRBM_SOFT_RESET);
 		tmp |= srbm_soft_reset;
@@ -1218,7 +1216,7 @@ static int gmc_v7_0_soft_reset(void *handle)
 
 static int gmc_v7_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 					     struct amdgpu_irq_src *src,
-					     unsigned type,
+					     unsigned int type,
 					     enum amdgpu_interrupt_state state)
 {
 	u32 tmp;
@@ -1263,6 +1261,12 @@ static int gmc_v7_0_process_interrupt(struct amdgpu_device *adev,
 {
 	u32 addr, status, mc_client, vmid;
 
+	/* Delegate to the soft IRQ handler ring */
+	if (adev->irq.ih_soft.enabled && entry->ih != &adev->irq.ih_soft) {
+		amdgpu_irq_delegate(adev, entry, 4);
+		return 1;
+	}
+
 	addr = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_ADDR);
 	status = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_STATUS);
 	mc_client = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_MCCLIENT);
@@ -1272,6 +1276,9 @@ static int gmc_v7_0_process_interrupt(struct amdgpu_device *adev,
 	if (!addr && !status)
 		return 0;
 
+	amdgpu_vm_update_fault_cache(adev, entry->pasid,
+				     ((u64)addr) << AMDGPU_GPU_PAGE_SHIFT, status, AMDGPU_GFXHUB(0));
+
 	if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_FIRST)
 		gmc_v7_0_set_fault_enable_default(adev, false);
 
@@ -1289,7 +1296,7 @@ static int gmc_v7_0_process_interrupt(struct amdgpu_device *adev,
 	vmid = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
 			     VMID);
 	if (amdgpu_amdkfd_is_kfd_vmid(adev, vmid)
-		&& !atomic_read(&adev->gmc.vm_fault_info_updated)) {
+		&& !atomic_read_acquire(&adev->gmc.vm_fault_info_updated)) {
 		struct kfd_vm_fault_info *info = adev->gmc.vm_fault_info;
 		u32 protections = REG_GET_FIELD(status,
 					VM_CONTEXT1_PROTECTION_FAULT_STATUS,
@@ -1305,18 +1312,17 @@ static int gmc_v7_0_process_interrupt(struct amdgpu_device *adev,
 		info->prot_read = protections & 0x8 ? true : false;
 		info->prot_write = protections & 0x10 ? true : false;
 		info->prot_exec = protections & 0x20 ? true : false;
-		mb();
-		atomic_set(&adev->gmc.vm_fault_info_updated, 1);
+		atomic_set_release(&adev->gmc.vm_fault_info_updated, 1);
 	}
 
 	return 0;
 }
 
-static int gmc_v7_0_set_clockgating_state(void *handle,
+static int gmc_v7_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	bool gate = false;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (state == AMD_CG_STATE_GATE)
 		gate = true;
@@ -1332,7 +1338,7 @@ static int gmc_v7_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int gmc_v7_0_set_powergating_state(void *handle,
+static int gmc_v7_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -1382,8 +1388,7 @@ static void gmc_v7_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->gmc.vm_fault.funcs = &gmc_v7_0_irq_funcs;
 }
 
-const struct amdgpu_ip_block_version gmc_v7_0_ip_block =
-{
+const struct amdgpu_ip_block_version gmc_v7_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GMC,
 	.major = 7,
 	.minor = 0,
@@ -1391,8 +1396,7 @@ const struct amdgpu_ip_block_version gmc_v7_0_ip_block =
 	.funcs = &gmc_v7_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version gmc_v7_4_ip_block =
-{
+const struct amdgpu_ip_block_version gmc_v7_4_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GMC,
 	.major = 7,
 	.minor = 4,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
index 504c1b34dab7..6551b60f2584 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
@@ -53,7 +53,7 @@
 
 static void gmc_v8_0_set_gmc_funcs(struct amdgpu_device *adev);
 static void gmc_v8_0_set_irq_funcs(struct amdgpu_device *adev);
-static int gmc_v8_0_wait_for_idle(void *handle);
+static int gmc_v8_0_wait_for_idle(struct amdgpu_ip_block *ip_block);
 
 MODULE_FIRMWARE("amdgpu/tonga_mc.bin");
 MODULE_FIRMWARE("amdgpu/polaris11_mc.bin");
@@ -64,8 +64,7 @@ MODULE_FIRMWARE("amdgpu/polaris11_k_mc.bin");
 MODULE_FIRMWARE("amdgpu/polaris10_k_mc.bin");
 MODULE_FIRMWARE("amdgpu/polaris12_k_mc.bin");
 
-static const u32 golden_settings_tonga_a11[] =
-{
+static const u32 golden_settings_tonga_a11[] = {
 	mmMC_ARB_WTM_GRPWT_RD, 0x00000003, 0x00000000,
 	mmMC_HUB_RDREQ_DMIF_LIMIT, 0x0000007f, 0x00000028,
 	mmMC_HUB_WDP_UMC, 0x00007fb6, 0x00000991,
@@ -75,34 +74,29 @@ static const u32 golden_settings_tonga_a11[] =
 	mmVM_PRT_APERTURE3_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 };
 
-static const u32 tonga_mgcg_cgcg_init[] =
-{
+static const u32 tonga_mgcg_cgcg_init[] = {
 	mmMC_MEM_POWER_LS, 0xffffffff, 0x00000104
 };
 
-static const u32 golden_settings_fiji_a10[] =
-{
+static const u32 golden_settings_fiji_a10[] = {
 	mmVM_PRT_APERTURE0_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE1_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE2_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE3_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 };
 
-static const u32 fiji_mgcg_cgcg_init[] =
-{
+static const u32 fiji_mgcg_cgcg_init[] = {
 	mmMC_MEM_POWER_LS, 0xffffffff, 0x00000104
 };
 
-static const u32 golden_settings_polaris11_a11[] =
-{
+static const u32 golden_settings_polaris11_a11[] = {
 	mmVM_PRT_APERTURE0_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE1_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE2_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE3_LOW_ADDR, 0x0fffffff, 0x0fffffff
 };
 
-static const u32 golden_settings_polaris10_a11[] =
-{
+static const u32 golden_settings_polaris10_a11[] = {
 	mmMC_ARB_WTM_GRPWT_RD, 0x00000003, 0x00000000,
 	mmVM_PRT_APERTURE0_LOW_ADDR, 0x0fffffff, 0x0fffffff,
 	mmVM_PRT_APERTURE1_LOW_ADDR, 0x0fffffff, 0x0fffffff,
@@ -110,19 +104,16 @@ static const u32 golden_settings_polaris10_a11[] =
 	mmVM_PRT_APERTURE3_LOW_ADDR, 0x0fffffff, 0x0fffffff
 };
 
-static const u32 cz_mgcg_cgcg_init[] =
-{
+static const u32 cz_mgcg_cgcg_init[] = {
 	mmMC_MEM_POWER_LS, 0xffffffff, 0x00000104
 };
 
-static const u32 stoney_mgcg_cgcg_init[] =
-{
+static const u32 stoney_mgcg_cgcg_init[] = {
 	mmATC_MISC_CG, 0xffffffff, 0x000c0200,
 	mmMC_MEM_POWER_LS, 0xffffffff, 0x00000104
 };
 
-static const u32 golden_settings_stoney_common[] =
-{
+static const u32 golden_settings_stoney_common[] = {
 	mmMC_HUB_RDREQ_UVD, MC_HUB_RDREQ_UVD__PRESCALE_MASK, 0x00000004,
 	mmMC_RD_GRP_OTH, MC_RD_GRP_OTH__UVD_MASK, 0x00600000
 };
@@ -179,8 +170,13 @@ static void gmc_v8_0_init_golden_registers(struct amdgpu_device *adev)
 static void gmc_v8_0_mc_stop(struct amdgpu_device *adev)
 {
 	u32 blackout;
+	struct amdgpu_ip_block *ip_block;
+
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GMC);
+	if (!ip_block)
+		return;
 
-	gmc_v8_0_wait_for_idle(adev);
+	gmc_v8_0_wait_for_idle(ip_block);
 
 	blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL);
 	if (REG_GET_FIELD(blackout, MC_SHARED_BLACKOUT_CNTL, BLACKOUT_MODE) != 1) {
@@ -221,7 +217,6 @@ static void gmc_v8_0_mc_resume(struct amdgpu_device *adev)
 static int gmc_v8_0_init_microcode(struct amdgpu_device *adev)
 {
 	const char *chip_name;
-	char fw_name[30];
 	int err;
 
 	DRM_DEBUG("\n");
@@ -260,13 +255,14 @@ static int gmc_v8_0_init_microcode(struct amdgpu_device *adev)
 	case CHIP_STONEY:
 	case CHIP_VEGAM:
 		return 0;
-	default: BUG();
+	default:
+		return -EINVAL;
 	}
 
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mc.bin", chip_name);
-	err = amdgpu_ucode_request(adev, &adev->gmc.fw, fw_name);
+	err = amdgpu_ucode_request(adev, &adev->gmc.fw, AMDGPU_UCODE_REQUIRED,
+				   "amdgpu/%s_mc.bin", chip_name);
 	if (err) {
-		pr_err("mc: Failed to load firmware \"%s\"\n", fw_name);
+		pr_err("mc: Failed to load firmware \"%s_mc.bin\"\n", chip_name);
 		amdgpu_ucode_release(&adev->gmc.fw);
 	}
 	return err;
@@ -421,8 +417,9 @@ static void gmc_v8_0_vram_gtt_location(struct amdgpu_device *adev,
 		base = RREG32(mmMC_VM_FB_LOCATION) & 0xFFFF;
 	base <<= 24;
 
+	amdgpu_gmc_set_agp_default(adev, mc);
 	amdgpu_gmc_vram_location(adev, mc, base);
-	amdgpu_gmc_gart_location(adev, mc);
+	amdgpu_gmc_gart_location(adev, mc, AMDGPU_GART_PLACEMENT_BEST_FIT);
 }
 
 /**
@@ -435,6 +432,7 @@ static void gmc_v8_0_vram_gtt_location(struct amdgpu_device *adev,
  */
 static void gmc_v8_0_mc_program(struct amdgpu_device *adev)
 {
+	struct amdgpu_ip_block *ip_block;
 	u32 tmp;
 	int i, j;
 
@@ -448,9 +446,13 @@ static void gmc_v8_0_mc_program(struct amdgpu_device *adev)
 	}
 	WREG32(mmHDP_REG_COHERENCY_FLUSH_CNTL, 0);
 
-	if (gmc_v8_0_wait_for_idle((void *)adev)) {
+	ip_block = amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GMC);
+	if (!ip_block)
+		return;
+
+	if (gmc_v8_0_wait_for_idle(ip_block))
 		dev_warn(adev->dev, "Wait for MC idle timedout !\n");
-	}
+
 	if (adev->mode_info.num_crtc) {
 		/* Lockout access through VGA aperture*/
 		tmp = RREG32(mmVGA_HDP_CONTROL);
@@ -481,11 +483,10 @@ static void gmc_v8_0_mc_program(struct amdgpu_device *adev)
 	}
 
 	WREG32(mmMC_VM_AGP_BASE, 0);
-	WREG32(mmMC_VM_AGP_TOP, 0x0FFFFFFF);
-	WREG32(mmMC_VM_AGP_BOT, 0x0FFFFFFF);
-	if (gmc_v8_0_wait_for_idle((void *)adev)) {
+	WREG32(mmMC_VM_AGP_TOP, adev->gmc.agp_end >> 22);
+	WREG32(mmMC_VM_AGP_BOT, adev->gmc.agp_start >> 22);
+	if (gmc_v8_0_wait_for_idle(ip_block))
 		dev_warn(adev->dev, "Wait for MC idle timedout !\n");
-	}
 
 	WREG32(mmBIF_FB_EN, BIF_FB_EN__FB_READ_EN_MASK | BIF_FB_EN__FB_WRITE_EN_MASK);
 
@@ -517,11 +518,11 @@ static int gmc_v8_0_mc_init(struct amdgpu_device *adev)
 
 		/* Get VRAM informations */
 		tmp = RREG32(mmMC_ARB_RAMCFG);
-		if (REG_GET_FIELD(tmp, MC_ARB_RAMCFG, CHANSIZE)) {
+		if (REG_GET_FIELD(tmp, MC_ARB_RAMCFG, CHANSIZE))
 			chansize = 64;
-		} else {
+		else
 			chansize = 32;
-		}
+
 		tmp = RREG32(mmMC_SHARED_CHMAP);
 		switch (REG_GET_FIELD(tmp, MC_SHARED_CHMAP, NOOFCHAN)) {
 		case 0:
@@ -617,32 +618,27 @@ static int gmc_v8_0_mc_init(struct amdgpu_device *adev)
  * @pasid: pasid to be flush
  * @flush_type: type of flush
  * @all_hub: flush all hubs
+ * @inst: is used to select which instance of KIQ to use for the invalidation
  *
  * Flush the TLB for the requested pasid.
  */
-static int gmc_v8_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
-					uint16_t pasid, uint32_t flush_type,
-					bool all_hub)
+static void gmc_v8_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
+					 uint16_t pasid, uint32_t flush_type,
+					 bool all_hub, uint32_t inst)
 {
+	u32 mask = 0x0;
 	int vmid;
-	unsigned int tmp;
-
-	if (amdgpu_in_reset(adev))
-		return -EIO;
 
 	for (vmid = 1; vmid < 16; vmid++) {
+		u32 tmp = RREG32(mmATC_VMID0_PASID_MAPPING + vmid);
 
-		tmp = RREG32(mmATC_VMID0_PASID_MAPPING + vmid);
 		if ((tmp & ATC_VMID0_PASID_MAPPING__VALID_MASK) &&
-			(tmp & ATC_VMID0_PASID_MAPPING__PASID_MASK) == pasid) {
-			WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid);
-			RREG32(mmVM_INVALIDATE_RESPONSE);
-			break;
-		}
+		    (tmp & ATC_VMID0_PASID_MAPPING__PASID_MASK) == pasid)
+			mask |= 1 << vmid;
 	}
 
-	return 0;
-
+	WREG32(mmVM_INVALIDATE_REQUEST, mask);
+	RREG32(mmVM_INVALIDATE_RESPONSE);
 }
 
 /*
@@ -670,7 +666,7 @@ static void gmc_v8_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 }
 
 static uint64_t gmc_v8_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
-					    unsigned vmid, uint64_t pd_addr)
+					    unsigned int vmid, uint64_t pd_addr)
 {
 	uint32_t reg;
 
@@ -686,8 +682,8 @@ static uint64_t gmc_v8_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 	return pd_addr;
 }
 
-static void gmc_v8_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
-					unsigned pasid)
+static void gmc_v8_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned int vmid,
+					unsigned int pasid)
 {
 	amdgpu_ring_emit_wreg(ring, mmIH_VMID_0_LUT + vmid, pasid);
 }
@@ -720,11 +716,15 @@ static void gmc_v8_0_get_vm_pde(struct amdgpu_device *adev, int level,
 }
 
 static void gmc_v8_0_get_vm_pte(struct amdgpu_device *adev,
-				struct amdgpu_bo_va_mapping *mapping,
+				struct amdgpu_vm *vm,
+				struct amdgpu_bo *bo,
+				uint32_t vm_flags,
 				uint64_t *flags)
 {
-	*flags &= ~AMDGPU_PTE_EXECUTABLE;
-	*flags |= mapping->flags & AMDGPU_PTE_EXECUTABLE;
+	if (vm_flags & AMDGPU_VM_PAGE_EXECUTABLE)
+		*flags |= AMDGPU_PTE_EXECUTABLE;
+	else
+		*flags &= ~AMDGPU_PTE_EXECUTABLE;
 	*flags &= ~AMDGPU_PTE_PRT;
 }
 
@@ -758,11 +758,11 @@ static void gmc_v8_0_set_fault_enable_default(struct amdgpu_device *adev,
 }
 
 /**
- * gmc_v8_0_set_prt - set PRT VM fault
+ * gmc_v8_0_set_prt() - set PRT VM fault
  *
  * @adev: amdgpu_device pointer
  * @enable: enable/disable VM fault handling for PRT
-*/
+ */
 static void gmc_v8_0_set_prt(struct amdgpu_device *adev, bool enable)
 {
 	u32 tmp;
@@ -790,9 +790,10 @@ static void gmc_v8_0_set_prt(struct amdgpu_device *adev, bool enable)
 	WREG32(mmVM_PRT_CNTL, tmp);
 
 	if (enable) {
-		uint32_t low = AMDGPU_VA_RESERVED_SIZE >> AMDGPU_GPU_PAGE_SHIFT;
+		uint32_t low = AMDGPU_VA_RESERVED_BOTTOM >>
+			AMDGPU_GPU_PAGE_SHIFT;
 		uint32_t high = adev->vm_manager.max_pfn -
-			(AMDGPU_VA_RESERVED_SIZE >> AMDGPU_GPU_PAGE_SHIFT);
+			(AMDGPU_VA_RESERVED_TOP >> AMDGPU_GPU_PAGE_SHIFT);
 
 		WREG32(mmVM_PRT_APERTURE0_LOW_ADDR, low);
 		WREG32(mmVM_PRT_APERTURE1_LOW_ADDR, low);
@@ -939,7 +940,7 @@ static int gmc_v8_0_gart_enable(struct amdgpu_device *adev)
 
 	gmc_v8_0_flush_gpu_tlb(adev, 0, 0, 0);
 	DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
-		 (unsigned)(adev->gmc.gart_size >> 20),
+		 (unsigned int)(adev->gmc.gart_size >> 20),
 		 (unsigned long long)table_addr);
 	return 0;
 }
@@ -1000,7 +1001,7 @@ static void gmc_v8_0_gart_disable(struct amdgpu_device *adev)
  * Print human readable fault information (VI).
  */
 static void gmc_v8_0_vm_decode_fault(struct amdgpu_device *adev, u32 status,
-				     u32 addr, u32 mc_client, unsigned pasid)
+				     u32 addr, u32 mc_client, unsigned int pasid)
 {
 	u32 vmid = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, VMID);
 	u32 protections = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
@@ -1041,9 +1042,9 @@ static int gmc_v8_0_convert_vram_type(int mc_seq_vram_type)
 	}
 }
 
-static int gmc_v8_0_early_init(void *handle)
+static int gmc_v8_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v8_0_set_gmc_funcs(adev);
 	gmc_v8_0_set_irq_funcs(adev);
@@ -1055,13 +1056,14 @@ static int gmc_v8_0_early_init(void *handle)
 		adev->gmc.shared_aperture_end + 1;
 	adev->gmc.private_aperture_end =
 		adev->gmc.private_aperture_start + (4ULL << 30) - 1;
+	adev->gmc.noretry_flags = AMDGPU_VM_NORETRY_FLAGS_TF;
 
 	return 0;
 }
 
-static int gmc_v8_0_late_init(void *handle)
+static int gmc_v8_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS)
 		return amdgpu_irq_get(adev, &adev->gmc.vm_fault, 0);
@@ -1069,15 +1071,16 @@ static int gmc_v8_0_late_init(void *handle)
 		return 0;
 }
 
-static unsigned gmc_v8_0_get_vbios_fb_size(struct amdgpu_device *adev)
+static unsigned int gmc_v8_0_get_vbios_fb_size(struct amdgpu_device *adev)
 {
 	u32 d1vga_control = RREG32(mmD1VGA_CONTROL);
-	unsigned size;
+	unsigned int size;
 
 	if (REG_GET_FIELD(d1vga_control, D1VGA_CONTROL, D1VGA_MODE_ENABLE)) {
 		size = AMDGPU_VBIOS_VGA_ALLOCATION;
 	} else {
 		u32 viewport = RREG32(mmVIEWPORT_SIZE);
+
 		size = (REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_HEIGHT) *
 			REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_WIDTH) *
 			4);
@@ -1088,12 +1091,12 @@ static unsigned gmc_v8_0_get_vbios_fb_size(struct amdgpu_device *adev)
 
 #define mmMC_SEQ_MISC0_FIJI 0xA71
 
-static int gmc_v8_0_sw_init(void *handle)
+static int gmc_v8_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	adev->num_vmhubs = 1;
+	set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
 
 	if (adev->flags & AMD_IS_APU) {
 		adev->gmc.vram_type = AMDGPU_VRAM_TYPE_UNKNOWN;
@@ -1180,14 +1183,14 @@ static int gmc_v8_0_sw_init(void *handle)
 					GFP_KERNEL);
 	if (!adev->gmc.vm_fault_info)
 		return -ENOMEM;
-	atomic_set(&adev->gmc.vm_fault_info_updated, 0);
+	atomic_set_release(&adev->gmc.vm_fault_info_updated, 0);
 
 	return 0;
 }
 
-static int gmc_v8_0_sw_fini(void *handle)
+static int gmc_v8_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_gem_force_release(adev);
 	amdgpu_vm_manager_fini(adev);
@@ -1199,10 +1202,10 @@ static int gmc_v8_0_sw_fini(void *handle)
 	return 0;
 }
 
-static int gmc_v8_0_hw_init(void *handle)
+static int gmc_v8_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v8_0_init_golden_registers(adev);
 
@@ -1230,13 +1233,13 @@ static int gmc_v8_0_hw_init(void *handle)
 
 	if (amdgpu_emu_mode == 1)
 		return amdgpu_gmc_vram_checking(adev);
-	else
-		return r;
+
+	return 0;
 }
 
-static int gmc_v8_0_hw_fini(void *handle)
+static int gmc_v8_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
 	gmc_v8_0_gart_disable(adev);
@@ -1244,32 +1247,29 @@ static int gmc_v8_0_hw_fini(void *handle)
 	return 0;
 }
 
-static int gmc_v8_0_suspend(void *handle)
+static int gmc_v8_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	gmc_v8_0_hw_fini(adev);
+	gmc_v8_0_hw_fini(ip_block);
 
 	return 0;
 }
 
-static int gmc_v8_0_resume(void *handle)
+static int gmc_v8_0_resume(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = gmc_v8_0_hw_init(adev);
+	r = gmc_v8_0_hw_init(ip_block);
 	if (r)
 		return r;
 
-	amdgpu_vmid_reset_all(adev);
+	amdgpu_vmid_reset_all(ip_block->adev);
 
 	return 0;
 }
 
-static bool gmc_v8_0_is_idle(void *handle)
+static bool gmc_v8_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (tmp & (SRBM_STATUS__MCB_BUSY_MASK | SRBM_STATUS__MCB_NON_DISPLAY_BUSY_MASK |
@@ -1279,11 +1279,11 @@ static bool gmc_v8_0_is_idle(void *handle)
 	return true;
 }
 
-static int gmc_v8_0_wait_for_idle(void *handle)
+static int gmc_v8_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
-	unsigned i;
+	unsigned int i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		/* read MC_STATUS */
@@ -1301,10 +1301,10 @@ static int gmc_v8_0_wait_for_idle(void *handle)
 
 }
 
-static bool gmc_v8_0_check_soft_reset(void *handle)
+static bool gmc_v8_0_check_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 srbm_soft_reset = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (tmp & SRBM_STATUS__VMC_BUSY_MASK)
@@ -1317,33 +1317,34 @@ static bool gmc_v8_0_check_soft_reset(void *handle)
 			srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset,
 							SRBM_SOFT_RESET, SOFT_RESET_MC, 1);
 	}
+
 	if (srbm_soft_reset) {
 		adev->gmc.srbm_soft_reset = srbm_soft_reset;
 		return true;
-	} else {
-		adev->gmc.srbm_soft_reset = 0;
-		return false;
 	}
+
+	adev->gmc.srbm_soft_reset = 0;
+
+	return false;
 }
 
-static int gmc_v8_0_pre_soft_reset(void *handle)
+static int gmc_v8_0_pre_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (!adev->gmc.srbm_soft_reset)
 		return 0;
 
 	gmc_v8_0_mc_stop(adev);
-	if (gmc_v8_0_wait_for_idle(adev)) {
+	if (gmc_v8_0_wait_for_idle(ip_block))
 		dev_warn(adev->dev, "Wait for GMC idle timed out !\n");
-	}
 
 	return 0;
 }
 
-static int gmc_v8_0_soft_reset(void *handle)
+static int gmc_v8_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 srbm_soft_reset;
 
 	if (!adev->gmc.srbm_soft_reset)
@@ -1372,9 +1373,9 @@ static int gmc_v8_0_soft_reset(void *handle)
 	return 0;
 }
 
-static int gmc_v8_0_post_soft_reset(void *handle)
+static int gmc_v8_0_post_soft_reset(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (!adev->gmc.srbm_soft_reset)
 		return 0;
@@ -1385,7 +1386,7 @@ static int gmc_v8_0_post_soft_reset(void *handle)
 
 static int gmc_v8_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 					     struct amdgpu_irq_src *src,
-					     unsigned type,
+					     unsigned int type,
 					     enum amdgpu_interrupt_state state)
 {
 	u32 tmp;
@@ -1438,6 +1439,12 @@ static int gmc_v8_0_process_interrupt(struct amdgpu_device *adev,
 		return 0;
 	}
 
+	/* Delegate to the soft IRQ handler ring */
+	if (adev->irq.ih_soft.enabled && entry->ih != &adev->irq.ih_soft) {
+		amdgpu_irq_delegate(adev, entry, 4);
+		return 1;
+	}
+
 	addr = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_ADDR);
 	status = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_STATUS);
 	mc_client = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_MCCLIENT);
@@ -1447,22 +1454,29 @@ static int gmc_v8_0_process_interrupt(struct amdgpu_device *adev,
 	if (!addr && !status)
 		return 0;
 
+	amdgpu_vm_update_fault_cache(adev, entry->pasid,
+				     ((u64)addr) << AMDGPU_GPU_PAGE_SHIFT, status, AMDGPU_GFXHUB(0));
+
 	if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_FIRST)
 		gmc_v8_0_set_fault_enable_default(adev, false);
 
 	if (printk_ratelimit()) {
-		struct amdgpu_task_info task_info;
+		struct amdgpu_task_info *task_info;
+
+		dev_err(adev->dev, "GPU fault detected: %d 0x%08x\n",
+			entry->src_id, entry->src_data[0]);
 
-		memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-		amdgpu_vm_get_task_info(adev, entry->pasid, &task_info);
+		task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+		if (task_info) {
+			amdgpu_vm_print_task_info(adev, task_info);
+			amdgpu_vm_put_task_info(task_info);
+		}
 
-		dev_err(adev->dev, "GPU fault detected: %d 0x%08x for process %s pid %d thread %s pid %d\n",
-			entry->src_id, entry->src_data[0], task_info.process_name,
-			task_info.tgid, task_info.task_name, task_info.pid);
 		dev_err(adev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x%08X\n",
-			addr);
+				addr);
 		dev_err(adev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n",
 			status);
+
 		gmc_v8_0_vm_decode_fault(adev, status, addr, mc_client,
 					 entry->pasid);
 	}
@@ -1470,7 +1484,7 @@ static int gmc_v8_0_process_interrupt(struct amdgpu_device *adev,
 	vmid = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
 			     VMID);
 	if (amdgpu_amdkfd_is_kfd_vmid(adev, vmid)
-		&& !atomic_read(&adev->gmc.vm_fault_info_updated)) {
+		&& !atomic_read_acquire(&adev->gmc.vm_fault_info_updated)) {
 		struct kfd_vm_fault_info *info = adev->gmc.vm_fault_info;
 		u32 protections = REG_GET_FIELD(status,
 					VM_CONTEXT1_PROTECTION_FAULT_STATUS,
@@ -1486,8 +1500,7 @@ static int gmc_v8_0_process_interrupt(struct amdgpu_device *adev,
 		info->prot_read = protections & 0x8 ? true : false;
 		info->prot_write = protections & 0x10 ? true : false;
 		info->prot_exec = protections & 0x20 ? true : false;
-		mb();
-		atomic_set(&adev->gmc.vm_fault_info_updated, 1);
+		atomic_set_release(&adev->gmc.vm_fault_info_updated, 1);
 	}
 
 	return 0;
@@ -1653,10 +1666,10 @@ static void fiji_update_mc_light_sleep(struct amdgpu_device *adev,
 	}
 }
 
-static int gmc_v8_0_set_clockgating_state(void *handle,
+static int gmc_v8_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
@@ -1674,15 +1687,15 @@ static int gmc_v8_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static int gmc_v8_0_set_powergating_state(void *handle,
+static int gmc_v8_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
 }
 
-static void gmc_v8_0_get_clockgating_state(void *handle, u64 *flags)
+static void gmc_v8_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int data;
 
 	if (amdgpu_sriov_vf(adev))
@@ -1746,8 +1759,7 @@ static void gmc_v8_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->gmc.vm_fault.funcs = &gmc_v8_0_irq_funcs;
 }
 
-const struct amdgpu_ip_block_version gmc_v8_0_ip_block =
-{
+const struct amdgpu_ip_block_version gmc_v8_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GMC,
 	.major = 8,
 	.minor = 0,
@@ -1755,8 +1767,7 @@ const struct amdgpu_ip_block_version gmc_v8_0_ip_block =
 	.funcs = &gmc_v8_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version gmc_v8_1_ip_block =
-{
+const struct amdgpu_ip_block_version gmc_v8_1_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GMC,
 	.major = 8,
 	.minor = 1,
@@ -1764,8 +1775,7 @@ const struct amdgpu_ip_block_version gmc_v8_1_ip_block =
 	.funcs = &gmc_v8_0_ip_funcs,
 };
 
-const struct amdgpu_ip_block_version gmc_v8_5_ip_block =
-{
+const struct amdgpu_ip_block_version gmc_v8_5_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GMC,
 	.major = 8,
 	.minor = 5,
diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index 67ad92097a65..8ad7519f7b58 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
@@ -56,6 +56,7 @@
 #include "umc_v6_1.h"
 #include "umc_v6_0.h"
 #include "umc_v6_7.h"
+#include "umc_v12_0.h"
 #include "hdp_v4_0.h"
 #include "mca_v3_0.h"
 
@@ -64,8 +65,6 @@
 #include "amdgpu_ras.h"
 #include "amdgpu_xgmi.h"
 
-#include "amdgpu_reset.h"
-
 /* add these here since we already include dce12 headers and these are for DCN */
 #define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION                                                          0x055d
 #define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION_BASE_IDX                                                 2
@@ -79,8 +78,7 @@
 #define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION_DCN2                                                          0x05ea
 #define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION_DCN2_BASE_IDX                                                 2
 
-
-static const char *gfxhub_client_ids[] = {
+static const char * const gfxhub_client_ids[] = {
 	"CB",
 	"DB",
 	"IA",
@@ -331,14 +329,12 @@ static const char *mmhub_client_ids_aldebaran[][2] = {
 	[384+0][1] = "OSS",
 };
 
-static const struct soc15_reg_golden golden_settings_mmhub_1_0_0[] =
-{
+static const struct soc15_reg_golden golden_settings_mmhub_1_0_0[] = {
 	SOC15_REG_GOLDEN_VALUE(MMHUB, 0, mmDAGB1_WRCLI2, 0x00000007, 0xfe5fe0fa),
 	SOC15_REG_GOLDEN_VALUE(MMHUB, 0, mmMMEA1_DRAM_WR_CLI2GRP_MAP0, 0x00000030, 0x55555565)
 };
 
-static const struct soc15_reg_golden golden_settings_athub_1_0_0[] =
-{
+static const struct soc15_reg_golden golden_settings_athub_1_0_0[] = {
 	SOC15_REG_GOLDEN_VALUE(ATHUB, 0, mmRPB_ARB_CNTL, 0x0000ff00, 0x00000800),
 	SOC15_REG_GOLDEN_VALUE(ATHUB, 0, mmRPB_ARB_CNTL2, 0x00ff00ff, 0x00080008)
 };
@@ -415,13 +411,14 @@ static const uint32_t ecc_umc_mcumc_ctrl_mask_addrs[] = {
 
 static int gmc_v9_0_ecc_interrupt_state(struct amdgpu_device *adev,
 		struct amdgpu_irq_src *src,
-		unsigned type,
+		unsigned int type,
 		enum amdgpu_interrupt_state state)
 {
 	u32 bits, i, tmp, reg;
 
 	/* Devices newer then VEGA10/12 shall have these programming
-	     sequences performed by PSP BL */
+	 * sequences performed by PSP BL
+	 */
 	if (adev->asic_type >= CHIP_VEGA20)
 		return 0;
 
@@ -465,7 +462,7 @@ static int gmc_v9_0_ecc_interrupt_state(struct amdgpu_device *adev,
 
 static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 					struct amdgpu_irq_src *src,
-					unsigned type,
+					unsigned int type,
 					enum amdgpu_interrupt_state state)
 {
 	struct amdgpu_vmhub *hub;
@@ -481,7 +478,7 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 
 	switch (state) {
 	case AMDGPU_IRQ_STATE_DISABLE:
-		for (j = 0; j < adev->num_vmhubs; j++) {
+		for_each_set_bit(j, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS) {
 			hub = &adev->vmhub[j];
 			for (i = 0; i < 16; i++) {
 				reg = hub->vm_context0_cntl + i;
@@ -491,25 +488,25 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 				 * fini/suspend, so the overall state doesn't
 				 * change over the course of suspend/resume.
 				 */
-				if (adev->in_s0ix && (j == AMDGPU_GFXHUB_0))
+				if (adev->in_s0ix && (j == AMDGPU_GFXHUB(0)))
 					continue;
 
-				if (j == AMDGPU_GFXHUB_0)
-					tmp = RREG32_SOC15_IP(GC, reg);
-				else
+				if (j >= AMDGPU_MMHUB0(0))
 					tmp = RREG32_SOC15_IP(MMHUB, reg);
+				else
+					tmp = RREG32_XCC(reg, j);
 
 				tmp &= ~bits;
 
-				if (j == AMDGPU_GFXHUB_0)
-					WREG32_SOC15_IP(GC, reg, tmp);
-				else
+				if (j >= AMDGPU_MMHUB0(0))
 					WREG32_SOC15_IP(MMHUB, reg, tmp);
+				else
+					WREG32_XCC(reg, tmp, j);
 			}
 		}
 		break;
 	case AMDGPU_IRQ_STATE_ENABLE:
-		for (j = 0; j < adev->num_vmhubs; j++) {
+		for_each_set_bit(j, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS) {
 			hub = &adev->vmhub[j];
 			for (i = 0; i < 16; i++) {
 				reg = hub->vm_context0_cntl + i;
@@ -519,20 +516,20 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
 				 * fini/suspend, so the overall state doesn't
 				 * change over the course of suspend/resume.
 				 */
-				if (adev->in_s0ix && (j == AMDGPU_GFXHUB_0))
+				if (adev->in_s0ix && (j == AMDGPU_GFXHUB(0)))
 					continue;
 
-				if (j == AMDGPU_GFXHUB_0)
-					tmp = RREG32_SOC15_IP(GC, reg);
-				else
+				if (j >= AMDGPU_MMHUB0(0))
 					tmp = RREG32_SOC15_IP(MMHUB, reg);
+				else
+					tmp = RREG32_XCC(reg, j);
 
 				tmp |= bits;
 
-				if (j == AMDGPU_GFXHUB_0)
-					WREG32_SOC15_IP(GC, reg, tmp);
-				else
+				if (j >= AMDGPU_MMHUB0(0))
 					WREG32_SOC15_IP(MMHUB, reg, tmp);
+				else
+					WREG32_XCC(reg, tmp, j);
 			}
 		}
 		break;
@@ -547,70 +544,111 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev,
 				      struct amdgpu_irq_src *source,
 				      struct amdgpu_iv_entry *entry)
 {
-	bool retry_fault = !!(entry->src_data[1] & 0x80);
-	bool write_fault = !!(entry->src_data[1] & 0x20);
-	uint32_t status = 0, cid = 0, rw = 0;
-	struct amdgpu_task_info task_info;
+	bool retry_fault = !!(entry->src_data[1] &
+			      AMDGPU_GMC9_FAULT_SOURCE_DATA_RETRY);
+	bool write_fault = !!(entry->src_data[1] &
+			      AMDGPU_GMC9_FAULT_SOURCE_DATA_WRITE);
+	uint32_t status = 0, cid = 0, rw = 0, fed = 0;
+	struct amdgpu_task_info *task_info;
 	struct amdgpu_vmhub *hub;
 	const char *mmhub_cid;
 	const char *hub_name;
+	unsigned int vmhub;
 	u64 addr;
+	uint32_t cam_index = 0;
+	int ret, xcc_id = 0;
+	uint32_t node_id;
+
+	node_id = entry->node_id;
 
 	addr = (u64)entry->src_data[0] << 12;
 	addr |= ((u64)entry->src_data[1] & 0xf) << 44;
 
+	if (entry->client_id == SOC15_IH_CLIENTID_VMC) {
+		hub_name = "mmhub0";
+		vmhub = AMDGPU_MMHUB0(node_id / 4);
+	} else if (entry->client_id == SOC15_IH_CLIENTID_VMC1) {
+		hub_name = "mmhub1";
+		vmhub = AMDGPU_MMHUB1(0);
+	} else {
+		hub_name = "gfxhub0";
+		if (adev->gfx.funcs->ih_node_to_logical_xcc) {
+			xcc_id = adev->gfx.funcs->ih_node_to_logical_xcc(adev,
+				node_id);
+			if (xcc_id < 0)
+				xcc_id = 0;
+		}
+		vmhub = xcc_id;
+	}
+	hub = &adev->vmhub[vmhub];
+
 	if (retry_fault) {
-		/* Returning 1 here also prevents sending the IV to the KFD */
+		if (adev->irq.retry_cam_enabled) {
+			/* Delegate it to a different ring if the hardware hasn't
+			 * already done it.
+			 */
+			if (entry->ih == &adev->irq.ih) {
+				amdgpu_irq_delegate(adev, entry, 8);
+				return 1;
+			}
+
+			cam_index = entry->src_data[2] & 0x3ff;
 
-		/* Process it onyl if it's the first fault for this address */
-		if (entry->ih != &adev->irq.ih_soft &&
-		    amdgpu_gmc_filter_faults(adev, entry->ih, addr, entry->pasid,
+			ret = amdgpu_vm_handle_fault(adev, entry->pasid, entry->vmid, node_id,
+						     addr, entry->timestamp, write_fault);
+			WDOORBELL32(adev->irq.retry_cam_doorbell_index, cam_index);
+			if (ret)
+				return 1;
+		} else {
+			/* Process it onyl if it's the first fault for this address */
+			if (entry->ih != &adev->irq.ih_soft &&
+			    amdgpu_gmc_filter_faults(adev, entry->ih, addr, entry->pasid,
 					     entry->timestamp))
-			return 1;
+				return 1;
 
-		/* Delegate it to a different ring if the hardware hasn't
-		 * already done it.
-		 */
-		if (entry->ih == &adev->irq.ih) {
-			amdgpu_irq_delegate(adev, entry, 8);
-			return 1;
-		}
+			/* Delegate it to a different ring if the hardware hasn't
+			 * already done it.
+			 */
+			if (entry->ih == &adev->irq.ih) {
+				amdgpu_irq_delegate(adev, entry, 8);
+				return 1;
+			}
 
-		/* Try to handle the recoverable page faults by filling page
-		 * tables
-		 */
-		if (amdgpu_vm_handle_fault(adev, entry->pasid, addr, write_fault))
-			return 1;
+			/* Try to handle the recoverable page faults by filling page
+			 * tables
+			 */
+			if (amdgpu_vm_handle_fault(adev, entry->pasid, entry->vmid, node_id,
+						   addr, entry->timestamp, write_fault))
+				return 1;
+		}
 	}
 
+	if (kgd2kfd_vmfault_fast_path(adev, entry, retry_fault))
+		return 1;
+
 	if (!printk_ratelimit())
 		return 0;
 
-	if (entry->client_id == SOC15_IH_CLIENTID_VMC) {
-		hub_name = "mmhub0";
-		hub = &adev->vmhub[AMDGPU_MMHUB_0];
-	} else if (entry->client_id == SOC15_IH_CLIENTID_VMC1) {
-		hub_name = "mmhub1";
-		hub = &adev->vmhub[AMDGPU_MMHUB_1];
-	} else {
-		hub_name = "gfxhub0";
-		hub = &adev->vmhub[AMDGPU_GFXHUB_0];
+	dev_err(adev->dev,
+		"[%s] %s page fault (src_id:%u ring:%u vmid:%u pasid:%u)\n", hub_name,
+		retry_fault ? "retry" : "no-retry",
+		entry->src_id, entry->ring_id, entry->vmid, entry->pasid);
+
+	task_info = amdgpu_vm_get_task_info_pasid(adev, entry->pasid);
+	if (task_info) {
+		amdgpu_vm_print_task_info(adev, task_info);
+		amdgpu_vm_put_task_info(task_info);
 	}
 
-	memset(&task_info, 0, sizeof(struct amdgpu_task_info));
-	amdgpu_vm_get_task_info(adev, entry->pasid, &task_info);
-
-	dev_err(adev->dev,
-		"[%s] %s page fault (src_id:%u ring:%u vmid:%u "
-		"pasid:%u, for process %s pid %d thread %s pid %d)\n",
-		hub_name, retry_fault ? "retry" : "no-retry",
-		entry->src_id, entry->ring_id, entry->vmid,
-		entry->pasid, task_info.process_name, task_info.tgid,
-		task_info.task_name, task_info.pid);
 	dev_err(adev->dev, "  in page starting at address 0x%016llx from IH client 0x%x (%s)\n",
 		addr, entry->client_id,
 		soc15_ih_clientid_name[entry->client_id]);
 
+	if (amdgpu_is_multi_aid(adev))
+		dev_err(adev->dev, "  cookie node_id %d fault from die %s%d%s\n",
+			node_id, node_id % 4 == 3 ? "RSV" : "AID", node_id / 4,
+			node_id % 4 == 1 ? ".XCD0" : node_id % 4 == 2 ? ".XCD1" : "");
+
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
@@ -619,26 +657,41 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev,
 	 * be updated to avoid reading an incorrect value due to
 	 * the new fast GRBM interface.
 	 */
-	if ((entry->vmid_src == AMDGPU_GFXHUB_0) &&
-	    (adev->ip_versions[GC_HWIP][0] < IP_VERSION(9, 4, 2)))
+	if ((entry->vmid_src == AMDGPU_GFXHUB(0)) &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(9, 4, 2)))
 		RREG32(hub->vm_l2_pro_fault_status);
 
 	status = RREG32(hub->vm_l2_pro_fault_status);
 	cid = REG_GET_FIELD(status, VM_L2_PROTECTION_FAULT_STATUS, CID);
 	rw = REG_GET_FIELD(status, VM_L2_PROTECTION_FAULT_STATUS, RW);
-	WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1);
+	fed = REG_GET_FIELD(status, VM_L2_PROTECTION_FAULT_STATUS, FED);
 
+	/* for fed error, kfd will handle it, return directly */
+	if (fed && amdgpu_ras_is_poison_mode_supported(adev) &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(9, 4, 2)))
+		return 0;
+
+	/* Only print L2 fault status if the status register could be read and
+	 * contains useful information
+	 */
+	if (!status)
+		return 0;
+
+	if (!amdgpu_sriov_vf(adev))
+		WREG32_P(hub->vm_l2_pro_fault_cntl, 1, ~1);
+
+	amdgpu_vm_update_fault_cache(adev, entry->pasid, addr, status, vmhub);
 
 	dev_err(adev->dev,
 		"VM_L2_PROTECTION_FAULT_STATUS:0x%08X\n",
 		status);
-	if (hub == &adev->vmhub[AMDGPU_GFXHUB_0]) {
+	if (entry->vmid_src == AMDGPU_GFXHUB(0)) {
 		dev_err(adev->dev, "\t Faulty UTCL2 client ID: %s (0x%x)\n",
 			cid >= ARRAY_SIZE(gfxhub_client_ids) ? "unknown" :
 			gfxhub_client_ids[cid],
 			cid);
 	} else {
-		switch (adev->ip_versions[MMHUB_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 		case IP_VERSION(9, 0, 0):
 			mmhub_cid = mmhub_client_ids_vega10[cid][rw];
 			break;
@@ -703,7 +756,8 @@ static void gmc_v9_0_set_irq_funcs(struct amdgpu_device *adev)
 	adev->gmc.vm_fault.funcs = &gmc_v9_0_irq_funcs;
 
 	if (!amdgpu_sriov_vf(adev) &&
-	    !adev->gmc.xgmi.connected_to_cpu) {
+	    !adev->gmc.xgmi.connected_to_cpu &&
+	    !adev->gmc.is_app_apu) {
 		adev->gmc.ecc_irq.num_types = 1;
 		adev->gmc.ecc_irq.funcs = &gmc_v9_0_ecc_funcs;
 	}
@@ -738,12 +792,12 @@ static uint32_t gmc_v9_0_get_invalidate_req(unsigned int vmid,
 static bool gmc_v9_0_use_invalidate_semaphore(struct amdgpu_device *adev,
 				       uint32_t vmhub)
 {
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) ||
-	    adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3))
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2) ||
+	    amdgpu_is_multi_aid(adev))
 		return false;
 
-	return ((vmhub == AMDGPU_MMHUB_0 ||
-		 vmhub == AMDGPU_MMHUB_1) &&
+	return ((vmhub == AMDGPU_MMHUB0(0) ||
+		 vmhub == AMDGPU_MMHUB1(0)) &&
 		(!amdgpu_sriov_vf(adev)) &&
 		(!(!(adev->apu_flags & AMD_APU_IS_RAVEN2) &&
 		   (adev->apu_flags & AMD_APU_IS_PICASSO))));
@@ -782,43 +836,37 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 					uint32_t vmhub, uint32_t flush_type)
 {
 	bool use_semaphore = gmc_v9_0_use_invalidate_semaphore(adev, vmhub);
-	const unsigned eng = 17;
-	u32 j, inv_req, inv_req2, tmp;
+	u32 j, inv_req, tmp, sem, req, ack, inst;
+	const unsigned int eng = 17;
 	struct amdgpu_vmhub *hub;
 
-	BUG_ON(vmhub >= adev->num_vmhubs);
+	BUG_ON(vmhub >= AMDGPU_MAX_VMHUBS);
 
 	hub = &adev->vmhub[vmhub];
-	if (adev->gmc.xgmi.num_physical_nodes &&
-	    adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 0)) {
-		/* Vega20+XGMI caches PTEs in TC and TLB. Add a
-		 * heavy-weight TLB flush (type 2), which flushes
-		 * both. Due to a race condition with concurrent
-		 * memory accesses using the same TLB cache line, we
-		 * still need a second TLB flush after this.
-		 */
-		inv_req = gmc_v9_0_get_invalidate_req(vmid, 2);
-		inv_req2 = gmc_v9_0_get_invalidate_req(vmid, flush_type);
-	} else {
-		inv_req = gmc_v9_0_get_invalidate_req(vmid, flush_type);
-		inv_req2 = 0;
-	}
+	inv_req = gmc_v9_0_get_invalidate_req(vmid, flush_type);
+	sem = hub->vm_inv_eng0_sem + hub->eng_distance * eng;
+	req = hub->vm_inv_eng0_req + hub->eng_distance * eng;
+	ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;
+
+	if (vmhub >= AMDGPU_MMHUB0(0))
+		inst = 0;
+	else
+		inst = vmhub;
 
-	/* This is necessary for a HW workaround under SRIOV as well
-	 * as GFXOFF under bare metal
+	/* This is necessary for SRIOV as well as for GFXOFF to function
+	 * properly under bare metal
 	 */
-	if (adev->gfx.kiq.ring.sched.ready &&
-	    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev)) &&
-	    down_read_trylock(&adev->reset_domain->sem)) {
+	if (adev->gfx.kiq[inst].ring.sched.ready &&
+	    (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) {
 		uint32_t req = hub->vm_inv_eng0_req + hub->eng_distance * eng;
 		uint32_t ack = hub->vm_inv_eng0_ack + hub->eng_distance * eng;
 
-		amdgpu_virt_kiq_reg_write_reg_wait(adev, req, ack, inv_req,
-						   1 << vmid);
-		up_read(&adev->reset_domain->sem);
+		amdgpu_gmc_fw_reg_write_reg_wait(adev, req, ack, inv_req,
+						 1 << vmid, inst);
 		return;
 	}
 
+	/* This path is needed before KIQ/MES/GFXOFF are set up */
 	spin_lock(&adev->gmc.invalidate_lock);
 
 	/*
@@ -832,11 +880,10 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 	if (use_semaphore) {
 		for (j = 0; j < adev->usec_timeout; j++) {
 			/* a read return value of 1 means semaphore acquire */
-			if (vmhub == AMDGPU_GFXHUB_0)
-				tmp = RREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_sem + hub->eng_distance * eng);
+			if (vmhub >= AMDGPU_MMHUB0(0))
+				tmp = RREG32_SOC15_IP_NO_KIQ(MMHUB, sem, GET_INST(GC, inst));
 			else
-				tmp = RREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_sem + hub->eng_distance * eng);
-
+				tmp = RREG32_SOC15_IP_NO_KIQ(GC, sem, GET_INST(GC, inst));
 			if (tmp & 0x1)
 				break;
 			udelay(1);
@@ -846,36 +893,29 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 			DRM_ERROR("Timeout waiting for sem acquire in VM flush!\n");
 	}
 
-	do {
-		if (vmhub == AMDGPU_GFXHUB_0)
-			WREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req);
-		else
-			WREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_req + hub->eng_distance * eng, inv_req);
-
-		/*
-		 * Issue a dummy read to wait for the ACK register to
-		 * be cleared to avoid a false ACK due to the new fast
-		 * GRBM interface.
-		 */
-		if ((vmhub == AMDGPU_GFXHUB_0) &&
-		    (adev->ip_versions[GC_HWIP][0] < IP_VERSION(9, 4, 2)))
-			RREG32_NO_KIQ(hub->vm_inv_eng0_req +
-				      hub->eng_distance * eng);
-
-		for (j = 0; j < adev->usec_timeout; j++) {
-			if (vmhub == AMDGPU_GFXHUB_0)
-				tmp = RREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_ack + hub->eng_distance * eng);
-			else
-				tmp = RREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_ack + hub->eng_distance * eng);
+	if (vmhub >= AMDGPU_MMHUB0(0))
+		WREG32_SOC15_IP_NO_KIQ(MMHUB, req, inv_req, GET_INST(GC, inst));
+	else
+		WREG32_SOC15_IP_NO_KIQ(GC, req, inv_req, GET_INST(GC, inst));
 
-			if (tmp & (1 << vmid))
-				break;
-			udelay(1);
-		}
+	/*
+	 * Issue a dummy read to wait for the ACK register to
+	 * be cleared to avoid a false ACK due to the new fast
+	 * GRBM interface.
+	 */
+	if ((vmhub == AMDGPU_GFXHUB(0)) &&
+	    (amdgpu_ip_version(adev, GC_HWIP, 0) < IP_VERSION(9, 4, 2)))
+		RREG32_NO_KIQ(req);
 
-		inv_req = inv_req2;
-		inv_req2 = 0;
-	} while (inv_req);
+	for (j = 0; j < adev->usec_timeout; j++) {
+		if (vmhub >= AMDGPU_MMHUB0(0))
+			tmp = RREG32_SOC15_IP_NO_KIQ(MMHUB, ack, GET_INST(GC, inst));
+		else
+			tmp = RREG32_SOC15_IP_NO_KIQ(GC, ack, GET_INST(GC, inst));
+		if (tmp & (1 << vmid))
+			break;
+		udelay(1);
+	}
 
 	/* TODO: It needs to continue working on debugging with semaphore for GFXHUB as well. */
 	if (use_semaphore) {
@@ -883,10 +923,10 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
 		 * add semaphore release after invalidation,
 		 * write with 0 means semaphore release
 		 */
-		if (vmhub == AMDGPU_GFXHUB_0)
-			WREG32_SOC15_IP_NO_KIQ(GC, hub->vm_inv_eng0_sem + hub->eng_distance * eng, 0);
+		if (vmhub >= AMDGPU_MMHUB0(0))
+			WREG32_SOC15_IP_NO_KIQ(MMHUB, sem, 0, GET_INST(GC, inst));
 		else
-			WREG32_SOC15_IP_NO_KIQ(MMHUB, hub->vm_inv_eng0_sem + hub->eng_distance * eng, 0);
+			WREG32_SOC15_IP_NO_KIQ(GC, sem, 0, GET_INST(GC, inst));
 	}
 
 	spin_unlock(&adev->gmc.invalidate_lock);
@@ -904,97 +944,46 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid,
  * @pasid: pasid to be flush
  * @flush_type: the flush type
  * @all_hub: flush all hubs
+ * @inst: is used to select which instance of KIQ to use for the invalidation
  *
  * Flush the TLB for the requested pasid.
  */
-static int gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
-					uint16_t pasid, uint32_t flush_type,
-					bool all_hub)
+static void gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
+					 uint16_t pasid, uint32_t flush_type,
+					 bool all_hub, uint32_t inst)
 {
-	int vmid, i;
-	signed long r;
-	uint32_t seq;
-	uint16_t queried_pasid;
-	bool ret;
-	u32 usec_timeout = amdgpu_sriov_vf(adev) ? SRIOV_USEC_TIMEOUT : adev->usec_timeout;
-	struct amdgpu_ring *ring = &adev->gfx.kiq.ring;
-	struct amdgpu_kiq *kiq = &adev->gfx.kiq;
-
-	if (amdgpu_in_reset(adev))
-		return -EIO;
-
-	if (ring->sched.ready && down_read_trylock(&adev->reset_domain->sem)) {
-		/* Vega20+XGMI caches PTEs in TC and TLB. Add a
-		 * heavy-weight TLB flush (type 2), which flushes
-		 * both. Due to a race condition with concurrent
-		 * memory accesses using the same TLB cache line, we
-		 * still need a second TLB flush after this.
-		 */
-		bool vega20_xgmi_wa = (adev->gmc.xgmi.num_physical_nodes &&
-				       adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 0));
-		/* 2 dwords flush + 8 dwords fence */
-		unsigned int ndw = kiq->pmf->invalidate_tlbs_size + 8;
-
-		if (vega20_xgmi_wa)
-			ndw += kiq->pmf->invalidate_tlbs_size;
-
-		spin_lock(&adev->gfx.kiq.ring_lock);
-		/* 2 dwords flush + 8 dwords fence */
-		amdgpu_ring_alloc(ring, ndw);
-		if (vega20_xgmi_wa)
-			kiq->pmf->kiq_invalidate_tlbs(ring,
-						      pasid, 2, all_hub);
-		kiq->pmf->kiq_invalidate_tlbs(ring,
-					pasid, flush_type, all_hub);
-		r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
-		if (r) {
-			amdgpu_ring_undo(ring);
-			spin_unlock(&adev->gfx.kiq.ring_lock);
-			up_read(&adev->reset_domain->sem);
-			return -ETIME;
-		}
-
-		amdgpu_ring_commit(ring);
-		spin_unlock(&adev->gfx.kiq.ring_lock);
-		r = amdgpu_fence_wait_polling(ring, seq, usec_timeout);
-		if (r < 1) {
-			dev_err(adev->dev, "wait for kiq fence error: %ld.\n", r);
-			up_read(&adev->reset_domain->sem);
-			return -ETIME;
-		}
-		up_read(&adev->reset_domain->sem);
-		return 0;
-	}
+	uint16_t queried;
+	int i, vmid;
 
 	for (vmid = 1; vmid < 16; vmid++) {
+		bool valid;
 
-		ret = gmc_v9_0_get_atc_vmid_pasid_mapping_info(adev, vmid,
-				&queried_pasid);
-		if (ret && queried_pasid == pasid) {
-			if (all_hub) {
-				for (i = 0; i < adev->num_vmhubs; i++)
-					gmc_v9_0_flush_gpu_tlb(adev, vmid,
-							i, flush_type);
-			} else {
-				gmc_v9_0_flush_gpu_tlb(adev, vmid,
-						AMDGPU_GFXHUB_0, flush_type);
-			}
-			break;
+		valid = gmc_v9_0_get_atc_vmid_pasid_mapping_info(adev, vmid,
+								 &queried);
+		if (!valid || queried != pasid)
+			continue;
+
+		if (all_hub) {
+			for_each_set_bit(i, adev->vmhubs_mask,
+					 AMDGPU_MAX_VMHUBS)
+				gmc_v9_0_flush_gpu_tlb(adev, vmid, i,
+						       flush_type);
+		} else {
+			gmc_v9_0_flush_gpu_tlb(adev, vmid,
+					       AMDGPU_GFXHUB(0),
+					       flush_type);
 		}
 	}
-
-	return 0;
-
 }
 
 static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
-					    unsigned vmid, uint64_t pd_addr)
+					    unsigned int vmid, uint64_t pd_addr)
 {
-	bool use_semaphore = gmc_v9_0_use_invalidate_semaphore(ring->adev, ring->funcs->vmhub);
+	bool use_semaphore = gmc_v9_0_use_invalidate_semaphore(ring->adev, ring->vm_hub);
 	struct amdgpu_device *adev = ring->adev;
-	struct amdgpu_vmhub *hub = &adev->vmhub[ring->funcs->vmhub];
+	struct amdgpu_vmhub *hub = &adev->vmhub[ring->vm_hub];
 	uint32_t req = gmc_v9_0_get_invalidate_req(vmid, 0);
-	unsigned eng = ring->vm_inv_eng;
+	unsigned int eng = ring->vm_inv_eng;
 
 	/*
 	 * It may lose gpuvm invalidate acknowldege state across power-gating
@@ -1036,17 +1025,17 @@ static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring,
 	return pd_addr;
 }
 
-static void gmc_v9_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
-					unsigned pasid)
+static void gmc_v9_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned int vmid,
+					unsigned int pasid)
 {
 	struct amdgpu_device *adev = ring->adev;
 	uint32_t reg;
 
 	/* Do nothing because there's no lut register for mmhub1. */
-	if (ring->funcs->vmhub == AMDGPU_MMHUB_1)
+	if (ring->vm_hub == AMDGPU_MMHUB1(0))
 		return;
 
-	if (ring->funcs->vmhub == AMDGPU_GFXHUB_0)
+	if (ring->vm_hub == AMDGPU_GFXHUB(0))
 		reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT) + vmid;
 	else
 		reg = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT_MM) + vmid;
@@ -1086,27 +1075,6 @@ static void gmc_v9_0_emit_pasid_mapping(struct amdgpu_ring *ring, unsigned vmid,
  * 0 valid
  */
 
-static uint64_t gmc_v9_0_map_mtype(struct amdgpu_device *adev, uint32_t flags)
-
-{
-	switch (flags) {
-	case AMDGPU_VM_MTYPE_DEFAULT:
-		return AMDGPU_PTE_MTYPE_VG10(MTYPE_NC);
-	case AMDGPU_VM_MTYPE_NC:
-		return AMDGPU_PTE_MTYPE_VG10(MTYPE_NC);
-	case AMDGPU_VM_MTYPE_WC:
-		return AMDGPU_PTE_MTYPE_VG10(MTYPE_WC);
-	case AMDGPU_VM_MTYPE_RW:
-		return AMDGPU_PTE_MTYPE_VG10(MTYPE_RW);
-	case AMDGPU_VM_MTYPE_CC:
-		return AMDGPU_PTE_MTYPE_VG10(MTYPE_CC);
-	case AMDGPU_VM_MTYPE_UC:
-		return AMDGPU_PTE_MTYPE_VG10(MTYPE_UC);
-	default:
-		return AMDGPU_PTE_MTYPE_VG10(MTYPE_NC);
-	}
-}
-
 static void gmc_v9_0_get_vm_pde(struct amdgpu_device *adev, int level,
 				uint64_t *addr, uint64_t *flags)
 {
@@ -1134,21 +1102,28 @@ static void gmc_v9_0_get_vm_pde(struct amdgpu_device *adev, int level,
 }
 
 static void gmc_v9_0_get_coherence_flags(struct amdgpu_device *adev,
+					 struct amdgpu_vm *vm,
 					 struct amdgpu_bo *bo,
-					 struct amdgpu_bo_va_mapping *mapping,
+					 uint32_t vm_flags,
 					 uint64_t *flags)
 {
 	struct amdgpu_device *bo_adev = amdgpu_ttm_adev(bo->tbo.bdev);
-	bool is_vram = bo->tbo.resource->mem_type == TTM_PL_VRAM;
-	bool coherent = bo->flags & AMDGPU_GEM_CREATE_COHERENT;
+	bool is_vram = bo->tbo.resource &&
+		bo->tbo.resource->mem_type == TTM_PL_VRAM;
+	bool coherent = bo->flags & (AMDGPU_GEM_CREATE_COHERENT |
+				     AMDGPU_GEM_CREATE_EXT_COHERENT);
+	bool ext_coherent = bo->flags & AMDGPU_GEM_CREATE_EXT_COHERENT;
 	bool uncached = bo->flags & AMDGPU_GEM_CREATE_UNCACHED;
-	unsigned int mtype;
+	unsigned int mtype_local, mtype;
+	uint32_t gc_ip_version = amdgpu_ip_version(adev, GC_HWIP, 0);
 	bool snoop = false;
+	bool is_local;
+
+	dma_resv_assert_held(bo->tbo.base.resv);
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (gc_ip_version) {
 	case IP_VERSION(9, 4, 1):
 	case IP_VERSION(9, 4, 2):
-	case IP_VERSION(9, 4, 3):
 		if (is_vram) {
 			if (bo_adev == adev) {
 				if (uncached)
@@ -1160,8 +1135,7 @@ static void gmc_v9_0_get_coherence_flags(struct amdgpu_device *adev,
 				/* FIXME: is this still needed? Or does
 				 * amdgpu_ttm_tt_pde_flags already handle this?
 				 */
-				if ((adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) ||
-				     adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3)) &&
+				if (gc_ip_version == IP_VERSION(9, 4, 2) &&
 				    adev->gmc.xgmi.connected_to_cpu)
 					snoop = true;
 			} else {
@@ -1169,7 +1143,7 @@ static void gmc_v9_0_get_coherence_flags(struct amdgpu_device *adev,
 					mtype = MTYPE_UC;
 				else
 					mtype = MTYPE_NC;
-				if (mapping->bo_va->is_xgmi)
+				if (amdgpu_xgmi_same_hive(adev, bo_adev))
 					snoop = true;
 			}
 		} else {
@@ -1183,6 +1157,47 @@ static void gmc_v9_0_get_coherence_flags(struct amdgpu_device *adev,
 			snoop = true;
 		}
 		break;
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
+		/* Only local VRAM BOs or system memory on non-NUMA APUs
+		 * can be assumed to be local in their entirety. Choose
+		 * MTYPE_NC as safe fallback for all system memory BOs on
+		 * NUMA systems. Their MTYPE can be overridden per-page in
+		 * gmc_v9_0_override_vm_pte_flags.
+		 */
+		mtype_local = MTYPE_RW;
+		if (amdgpu_mtype_local == 1) {
+			DRM_INFO_ONCE("Using MTYPE_NC for local memory\n");
+			mtype_local = MTYPE_NC;
+		} else if (amdgpu_mtype_local == 2) {
+			DRM_INFO_ONCE("Using MTYPE_CC for local memory\n");
+			mtype_local = MTYPE_CC;
+		} else {
+			DRM_INFO_ONCE("Using MTYPE_RW for local memory\n");
+		}
+		is_local = (!is_vram && (adev->flags & AMD_IS_APU) &&
+			    num_possible_nodes() <= 1) ||
+			   (is_vram && adev == bo_adev &&
+			    KFD_XCP_MEM_ID(adev, bo->xcp_id) == vm->mem_id);
+		snoop = true;
+		if (uncached) {
+			mtype = MTYPE_UC;
+		} else if (ext_coherent) {
+			mtype = is_local ? MTYPE_CC : MTYPE_UC;
+		} else if (adev->flags & AMD_IS_APU) {
+			mtype = is_local ? mtype_local : MTYPE_NC;
+		} else {
+			/* dGPU */
+			if (is_local)
+				mtype = mtype_local;
+			else if (gc_ip_version < IP_VERSION(9, 5, 0) && !is_vram)
+				mtype = MTYPE_UC;
+			else
+				mtype = MTYPE_NC;
+		}
+
+		break;
 	default:
 		if (uncached || coherent)
 			mtype = MTYPE_UC;
@@ -1197,37 +1212,131 @@ static void gmc_v9_0_get_coherence_flags(struct amdgpu_device *adev,
 	}
 
 	if (mtype != MTYPE_NC)
-		*flags = (*flags & ~AMDGPU_PTE_MTYPE_VG10_MASK) |
-			 AMDGPU_PTE_MTYPE_VG10(mtype);
+		*flags = AMDGPU_PTE_MTYPE_VG10(*flags, mtype);
+
 	*flags |= snoop ? AMDGPU_PTE_SNOOPED : 0;
 }
 
 static void gmc_v9_0_get_vm_pte(struct amdgpu_device *adev,
-				struct amdgpu_bo_va_mapping *mapping,
+				struct amdgpu_vm *vm,
+				struct amdgpu_bo *bo,
+				uint32_t vm_flags,
 				uint64_t *flags)
 {
-	struct amdgpu_bo *bo = mapping->bo_va->base.bo;
-
-	*flags &= ~AMDGPU_PTE_EXECUTABLE;
-	*flags |= mapping->flags & AMDGPU_PTE_EXECUTABLE;
+	if (vm_flags & AMDGPU_VM_PAGE_EXECUTABLE)
+		*flags |= AMDGPU_PTE_EXECUTABLE;
+	else
+		*flags &= ~AMDGPU_PTE_EXECUTABLE;
 
-	*flags &= ~AMDGPU_PTE_MTYPE_VG10_MASK;
-	*flags |= mapping->flags & AMDGPU_PTE_MTYPE_VG10_MASK;
+	switch (vm_flags & AMDGPU_VM_MTYPE_MASK) {
+	case AMDGPU_VM_MTYPE_DEFAULT:
+	case AMDGPU_VM_MTYPE_NC:
+	default:
+		*flags = AMDGPU_PTE_MTYPE_VG10(*flags, MTYPE_NC);
+		break;
+	case AMDGPU_VM_MTYPE_WC:
+		*flags |= AMDGPU_PTE_MTYPE_VG10(*flags, MTYPE_WC);
+		break;
+	case AMDGPU_VM_MTYPE_RW:
+		*flags |= AMDGPU_PTE_MTYPE_VG10(*flags, MTYPE_RW);
+		break;
+	case AMDGPU_VM_MTYPE_CC:
+		*flags |= AMDGPU_PTE_MTYPE_VG10(*flags, MTYPE_CC);
+		break;
+	case AMDGPU_VM_MTYPE_UC:
+		*flags |= AMDGPU_PTE_MTYPE_VG10(*flags, MTYPE_UC);
+		break;
+	}
 
-	if (mapping->flags & AMDGPU_PTE_PRT) {
+	if (vm_flags & AMDGPU_VM_PAGE_PRT) {
 		*flags |= AMDGPU_PTE_PRT;
 		*flags &= ~AMDGPU_PTE_VALID;
 	}
 
-	if (bo && bo->tbo.resource)
-		gmc_v9_0_get_coherence_flags(adev, mapping->bo_va->base.bo,
-					     mapping, flags);
+	if ((*flags & AMDGPU_PTE_VALID) && bo)
+		gmc_v9_0_get_coherence_flags(adev, vm, bo, vm_flags, flags);
 }
 
-static unsigned gmc_v9_0_get_vbios_fb_size(struct amdgpu_device *adev)
+static void gmc_v9_0_override_vm_pte_flags(struct amdgpu_device *adev,
+					   struct amdgpu_vm *vm,
+					   uint64_t addr, uint64_t *flags)
+{
+	int local_node, nid;
+
+	/* Only GFX 9.4.3 APUs associate GPUs with NUMA nodes. Local system
+	 * memory can use more efficient MTYPEs.
+	 */
+	if (!(adev->flags & AMD_IS_APU) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(9, 4, 3))
+		return;
+
+	/* Only direct-mapped memory allows us to determine the NUMA node from
+	 * the DMA address.
+	 */
+	if (!adev->ram_is_direct_mapped) {
+		dev_dbg_ratelimited(adev->dev, "RAM is not direct mapped\n");
+		return;
+	}
+
+	/* MTYPE_NC is the same default and can be overridden.
+	 * MTYPE_UC will be present if the memory is extended-coherent
+	 * and can also be overridden.
+	 */
+	if ((*flags & AMDGPU_PTE_MTYPE_VG10_MASK) !=
+	    AMDGPU_PTE_MTYPE_VG10(0ULL, MTYPE_NC) &&
+	    (*flags & AMDGPU_PTE_MTYPE_VG10_MASK) !=
+	    AMDGPU_PTE_MTYPE_VG10(0ULL, MTYPE_UC)) {
+		dev_dbg_ratelimited(adev->dev, "MTYPE is not NC or UC\n");
+		return;
+	}
+
+	/* FIXME: Only supported on native mode for now. For carve-out, the
+	 * NUMA affinity of the GPU/VM needs to come from the PCI info because
+	 * memory partitions are not associated with different NUMA nodes.
+	 */
+	if (adev->gmc.is_app_apu && vm->mem_id >= 0) {
+		local_node = adev->gmc.mem_partitions[vm->mem_id].numa.node;
+	} else {
+		dev_dbg_ratelimited(adev->dev, "Only native mode APU is supported.\n");
+		return;
+	}
+
+	/* Only handle real RAM. Mappings of PCIe resources don't have struct
+	 * page or NUMA nodes.
+	 */
+	if (!page_is_ram(addr >> PAGE_SHIFT)) {
+		dev_dbg_ratelimited(adev->dev, "Page is not RAM.\n");
+		return;
+	}
+	nid = pfn_to_nid(addr >> PAGE_SHIFT);
+	dev_dbg_ratelimited(adev->dev, "vm->mem_id=%d, local_node=%d, nid=%d\n",
+			    vm->mem_id, local_node, nid);
+	if (nid == local_node) {
+		uint64_t old_flags = *flags;
+		if ((*flags & AMDGPU_PTE_MTYPE_VG10_MASK) ==
+			AMDGPU_PTE_MTYPE_VG10(0ULL, MTYPE_NC)) {
+			unsigned int mtype_local = MTYPE_RW;
+
+			if (amdgpu_mtype_local == 1)
+				mtype_local = MTYPE_NC;
+			else if (amdgpu_mtype_local == 2)
+				mtype_local = MTYPE_CC;
+
+			*flags = AMDGPU_PTE_MTYPE_VG10(*flags, mtype_local);
+		} else {
+			/* MTYPE_UC case */
+			*flags = AMDGPU_PTE_MTYPE_VG10(*flags, MTYPE_CC);
+		}
+
+		dev_dbg_ratelimited(adev->dev, "flags updated from %llx to %llx\n",
+				    old_flags, *flags);
+	}
+}
+
+static unsigned int gmc_v9_0_get_vbios_fb_size(struct amdgpu_device *adev)
 {
 	u32 d1vga_control = RREG32_SOC15(DCE, 0, mmD1VGA_CONTROL);
-	unsigned size;
+	unsigned int size;
 
 	/* TODO move to DC so GMC doesn't need to hard-code DCN registers */
 
@@ -1236,7 +1345,7 @@ static unsigned gmc_v9_0_get_vbios_fb_size(struct amdgpu_device *adev)
 	} else {
 		u32 viewport;
 
-		switch (adev->ip_versions[DCE_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, DCE_HWIP, 0)) {
 		case IP_VERSION(1, 0, 0):
 		case IP_VERSION(1, 0, 1):
 			viewport = RREG32_SOC15(DCE, 0, mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION);
@@ -1266,15 +1375,29 @@ static unsigned gmc_v9_0_get_vbios_fb_size(struct amdgpu_device *adev)
 	return size;
 }
 
+static bool gmc_v9_0_need_reset_on_init(struct amdgpu_device *adev)
+{
+	if (adev->nbio.funcs && adev->nbio.funcs->is_nps_switch_requested &&
+	    adev->nbio.funcs->is_nps_switch_requested(adev)) {
+		adev->gmc.reset_flags |= AMDGPU_GMC_INIT_RESET_NPS;
+		return true;
+	}
+
+	return false;
+}
+
 static const struct amdgpu_gmc_funcs gmc_v9_0_gmc_funcs = {
 	.flush_gpu_tlb = gmc_v9_0_flush_gpu_tlb,
 	.flush_gpu_tlb_pasid = gmc_v9_0_flush_gpu_tlb_pasid,
 	.emit_flush_gpu_tlb = gmc_v9_0_emit_flush_gpu_tlb,
 	.emit_pasid_mapping = gmc_v9_0_emit_pasid_mapping,
-	.map_mtype = gmc_v9_0_map_mtype,
 	.get_vm_pde = gmc_v9_0_get_vm_pde,
 	.get_vm_pte = gmc_v9_0_get_vm_pte,
+	.override_vm_pte_flags = gmc_v9_0_override_vm_pte_flags,
 	.get_vbios_fb_size = gmc_v9_0_get_vbios_fb_size,
+	.query_mem_partition_mode = &amdgpu_gmc_query_memory_partition,
+	.request_mem_partition_mode = &amdgpu_gmc_request_memory_partition,
+	.need_reset_on_init = &gmc_v9_0_need_reset_on_init,
 };
 
 static void gmc_v9_0_set_gmc_funcs(struct amdgpu_device *adev)
@@ -1284,7 +1407,7 @@ static void gmc_v9_0_set_gmc_funcs(struct amdgpu_device *adev)
 
 static void gmc_v9_0_set_umc_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[UMC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, UMC_HWIP, 0)) {
 	case IP_VERSION(6, 0, 0):
 		adev->umc.funcs = &umc_v6_0_funcs;
 		break;
@@ -1320,6 +1443,17 @@ static void gmc_v9_0_set_umc_funcs(struct amdgpu_device *adev)
 		else
 			adev->umc.channel_idx_tbl = &umc_v6_7_channel_idx_tbl_second[0][0];
 		break;
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 5, 0):
+		adev->umc.max_ras_err_cnt_per_query =
+			UMC_V12_0_TOTAL_CHANNEL_NUM(adev) * UMC_V12_0_BAD_PAGE_NUM_PER_CHANNEL;
+		adev->umc.channel_inst_num = UMC_V12_0_CHANNEL_INSTANCE_NUM;
+		adev->umc.umc_inst_num = UMC_V12_0_UMC_INSTANCE_NUM;
+		adev->umc.node_inst_num /= UMC_V12_0_UMC_INSTANCE_NUM;
+		adev->umc.channel_offs = UMC_V12_0_PER_CHANNEL_OFFSET;
+		if (!adev->gmc.xgmi.connected_to_cpu && !adev->gmc.is_app_apu)
+			adev->umc.ras = &umc_v12_0_ras;
+		break;
 	default:
 		break;
 	}
@@ -1327,7 +1461,7 @@ static void gmc_v9_0_set_umc_funcs(struct amdgpu_device *adev)
 
 static void gmc_v9_0_set_mmhub_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(9, 4, 1):
 		adev->mmhub.funcs = &mmhub_v9_4_funcs;
 		break;
@@ -1335,6 +1469,7 @@ static void gmc_v9_0_set_mmhub_funcs(struct amdgpu_device *adev)
 		adev->mmhub.funcs = &mmhub_v1_7_funcs;
 		break;
 	case IP_VERSION(1, 8, 0):
+	case IP_VERSION(1, 8, 1):
 		adev->mmhub.funcs = &mmhub_v1_8_funcs;
 		break;
 	default:
@@ -1345,7 +1480,7 @@ static void gmc_v9_0_set_mmhub_funcs(struct amdgpu_device *adev)
 
 static void gmc_v9_0_set_mmhub_ras_funcs(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(9, 4, 0):
 		adev->mmhub.ras = &mmhub_v1_0_ras;
 		break;
@@ -1355,6 +1490,10 @@ static void gmc_v9_0_set_mmhub_ras_funcs(struct amdgpu_device *adev)
 	case IP_VERSION(9, 4, 2):
 		adev->mmhub.ras = &mmhub_v1_7_ras;
 		break;
+	case IP_VERSION(1, 8, 0):
+	case IP_VERSION(1, 8, 1):
+		adev->mmhub.ras = &mmhub_v1_8_ras;
+		break;
 	default:
 		/* mmhub ras is not available */
 		break;
@@ -1363,7 +1502,7 @@ static void gmc_v9_0_set_mmhub_ras_funcs(struct amdgpu_device *adev)
 
 static void gmc_v9_0_set_gfxhub_funcs(struct amdgpu_device *adev)
 {
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3))
+	if (amdgpu_is_multi_aid(adev))
 		adev->gfxhub.funcs = &gfxhub_v1_2_funcs;
 	else
 		adev->gfxhub.funcs = &gfxhub_v1_0_funcs;
@@ -1379,7 +1518,7 @@ static void gmc_v9_0_set_mca_ras_funcs(struct amdgpu_device *adev)
 	struct amdgpu_mca *mca = &adev->mca;
 
 	/* is UMC the right IP to check for MCA?  Maybe DF? */
-	switch (adev->ip_versions[UMC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, UMC_HWIP, 0)) {
 	case IP_VERSION(6, 7, 0):
 		if (!adev->gmc.xgmi.connected_to_cpu) {
 			mca->mp0.ras = &mca_v3_0_mp0_ras;
@@ -1398,21 +1537,76 @@ static void gmc_v9_0_set_xgmi_ras_funcs(struct amdgpu_device *adev)
 		adev->gmc.xgmi.ras = &xgmi_ras;
 }
 
-static int gmc_v9_0_early_init(void *handle)
+static void gmc_v9_0_init_nps_details(struct amdgpu_device *adev)
+{
+	enum amdgpu_memory_partition mode;
+	uint32_t supp_modes;
+	int i;
+
+	adev->gmc.supported_nps_modes = 0;
+
+	if (amdgpu_sriov_vf(adev) || (adev->flags & AMD_IS_APU))
+		return;
+
+	mode = amdgpu_gmc_get_memory_partition(adev, &supp_modes);
+
+	/* Mode detected by hardware and supported modes available */
+	if ((mode != UNKNOWN_MEMORY_PARTITION_MODE) && supp_modes) {
+		while ((i = ffs(supp_modes))) {
+			if (AMDGPU_ALL_NPS_MASK & BIT(i))
+				adev->gmc.supported_nps_modes |= BIT(i);
+			supp_modes &= supp_modes - 1;
+		}
+	} else {
+		/*TODO: Check PSP version also which supports NPS switch. Otherwise keep
+	 * supported modes as 0.
+	 */
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+		case IP_VERSION(9, 4, 3):
+		case IP_VERSION(9, 4, 4):
+			adev->gmc.supported_nps_modes =
+				BIT(AMDGPU_NPS1_PARTITION_MODE) |
+				BIT(AMDGPU_NPS4_PARTITION_MODE);
+			break;
+		default:
+			break;
+		}
+	}
+}
+
+static int gmc_v9_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
-	/* ARCT and VEGA20 don't have XGMI defined in their IP discovery tables */
-	if (adev->asic_type == CHIP_VEGA20 ||
-	    adev->asic_type == CHIP_ARCTURUS)
+	/*
+	 * 9.4.0, 9.4.1 and 9.4.3 don't have XGMI defined
+	 * in their IP discovery tables
+	 */
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 0) ||
+	    amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1) ||
+	    amdgpu_is_multi_aid(adev))
 		adev->gmc.xgmi.supported = true;
 
-	if (adev->ip_versions[XGMI_HWIP][0] == IP_VERSION(6, 1, 0)) {
+	if (amdgpu_ip_version(adev, XGMI_HWIP, 0) == IP_VERSION(6, 1, 0)) {
 		adev->gmc.xgmi.supported = true;
 		adev->gmc.xgmi.connected_to_cpu =
 			adev->smuio.funcs->is_host_gpu_xgmi_supported(adev);
 	}
 
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3)) {
+		enum amdgpu_pkg_type pkg_type =
+			adev->smuio.funcs->get_pkg_type(adev);
+		/* On GFXIP 9.4.3. APU, there is no physical VRAM domain present
+		 * and the APU, can be in used two possible modes:
+		 *  - carveout mode
+		 *  - native APU mode
+		 * "is_app_apu" can be used to identify the APU in the native
+		 * mode.
+		 */
+		adev->gmc.is_app_apu = (pkg_type == AMDGPU_PKG_TYPE_APU &&
+					!pci_resource_len(adev->pdev, 0));
+	}
+
 	gmc_v9_0_set_gmc_funcs(adev);
 	gmc_v9_0_set_irq_funcs(adev);
 	gmc_v9_0_set_umc_funcs(adev);
@@ -1429,13 +1623,14 @@ static int gmc_v9_0_early_init(void *handle)
 	adev->gmc.private_aperture_start = 0x1000000000000000ULL;
 	adev->gmc.private_aperture_end =
 		adev->gmc.private_aperture_start + (4ULL << 30) - 1;
+	adev->gmc.noretry_flags = AMDGPU_VM_NORETRY_FLAGS_TF;
 
 	return 0;
 }
 
-static int gmc_v9_0_late_init(void *handle)
+static int gmc_v9_0_late_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
 
 	r = amdgpu_gmc_allocate_vm_inv_eng(adev);
@@ -1447,7 +1642,7 @@ static int gmc_v9_0_late_init(void *handle)
 	 * writes, while disables HBM ECC for vega10.
 	 */
 	if (!amdgpu_sriov_vf(adev) &&
-	    (adev->ip_versions[UMC_HWIP][0] == IP_VERSION(6, 0, 0))) {
+	    (amdgpu_ip_version(adev, UMC_HWIP, 0) == IP_VERSION(6, 0, 0))) {
 		if (!(adev->ras_enabled & (1 << AMDGPU_RAS_BLOCK__UMC))) {
 			if (adev->df.funcs &&
 			    adev->df.funcs->enable_ecc_force_par_wr_rmw)
@@ -1456,13 +1651,8 @@ static int gmc_v9_0_late_init(void *handle)
 	}
 
 	if (!amdgpu_persistent_edc_harvesting_supported(adev)) {
-		if (adev->mmhub.ras && adev->mmhub.ras->ras_block.hw_ops &&
-		    adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count)
-			adev->mmhub.ras->ras_block.hw_ops->reset_ras_error_count(adev);
-
-		if (adev->hdp.ras && adev->hdp.ras->ras_block.hw_ops &&
-		    adev->hdp.ras->ras_block.hw_ops->reset_ras_error_count)
-			adev->hdp.ras->ras_block.hw_ops->reset_ras_error_count(adev);
+		amdgpu_ras_reset_error_count(adev, AMDGPU_RAS_BLOCK__MMHUB);
+		amdgpu_ras_reset_error_count(adev, AMDGPU_RAS_BLOCK__HDP);
 	}
 
 	r = amdgpu_gmc_ras_late_init(adev);
@@ -1477,14 +1667,17 @@ static void gmc_v9_0_vram_gtt_location(struct amdgpu_device *adev,
 {
 	u64 base = adev->mmhub.funcs->get_fb_location(adev);
 
+	amdgpu_gmc_set_agp_default(adev, mc);
+
 	/* add the xgmi offset of the physical node */
 	base += adev->gmc.xgmi.physical_node_id * adev->gmc.xgmi.node_segment_size;
-	if (adev->gmc.xgmi.connected_to_cpu) {
+	if (amdgpu_gmc_is_pdb0_enabled(adev)) {
 		amdgpu_gmc_sysvm_location(adev, mc);
 	} else {
 		amdgpu_gmc_vram_location(adev, mc, base);
-		amdgpu_gmc_gart_location(adev, mc);
-		amdgpu_gmc_agp_location(adev, mc);
+		amdgpu_gmc_gart_location(adev, mc, AMDGPU_GART_PLACEMENT_BEST_FIT);
+		if (!amdgpu_sriov_vf(adev) && (amdgpu_agp == 1))
+			amdgpu_gmc_agp_location(adev, mc);
 	}
 	/* base offset of vram pages */
 	adev->vm_manager.vram_base_offset = adev->gfxhub.funcs->get_mc_fb_offset(adev);
@@ -1508,8 +1701,13 @@ static int gmc_v9_0_mc_init(struct amdgpu_device *adev)
 	int r;
 
 	/* size in MB on si */
-	adev->gmc.mc_vram_size =
-		adev->nbio.funcs->get_memsize(adev) * 1024ULL * 1024ULL;
+	if (!adev->gmc.is_app_apu) {
+		adev->gmc.mc_vram_size =
+			adev->nbio.funcs->get_memsize(adev) * 1024ULL * 1024ULL;
+	} else {
+		DRM_DEBUG("Set mc_vram_size = 0 for APP APU\n");
+		adev->gmc.mc_vram_size = 0;
+	}
 	adev->gmc.real_vram_size = adev->gmc.mc_vram_size;
 
 	if (!(adev->flags & AMD_IS_APU) &&
@@ -1534,7 +1732,8 @@ static int gmc_v9_0_mc_init(struct amdgpu_device *adev)
 	 */
 
 	/* check whether both host-gpu and gpu-gpu xgmi links exist */
-	if (((adev->flags & AMD_IS_APU) && !amdgpu_passthrough(adev)) ||
+	if ((!amdgpu_sriov_vf(adev) &&
+		(adev->flags & AMD_IS_APU) && !amdgpu_passthrough(adev)) ||
 	    (adev->gmc.xgmi.supported &&
 	     adev->gmc.xgmi.connected_to_cpu)) {
 		adev->gmc.aper_base =
@@ -1549,13 +1748,15 @@ static int gmc_v9_0_mc_init(struct amdgpu_device *adev)
 
 	/* set the gart size */
 	if (amdgpu_gart_size == -1) {
-		switch (adev->ip_versions[GC_HWIP][0]) {
+		switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 		case IP_VERSION(9, 0, 1):  /* all engines support GPUVM */
 		case IP_VERSION(9, 2, 1):  /* all engines support GPUVM */
 		case IP_VERSION(9, 4, 0):
 		case IP_VERSION(9, 4, 1):
 		case IP_VERSION(9, 4, 2):
 		case IP_VERSION(9, 4, 3):
+		case IP_VERSION(9, 4, 4):
+		case IP_VERSION(9, 5, 0):
 		default:
 			adev->gmc.gart_size = 512ULL << 20;
 			break;
@@ -1585,7 +1786,7 @@ static int gmc_v9_0_gart_init(struct amdgpu_device *adev)
 		return 0;
 	}
 
-	if (adev->gmc.xgmi.connected_to_cpu) {
+	if (amdgpu_gmc_is_pdb0_enabled(adev)) {
 		adev->gmc.vmid0_page_table_depth = 1;
 		adev->gmc.vmid0_page_table_block_size = 12;
 	} else {
@@ -1598,15 +1799,21 @@ static int gmc_v9_0_gart_init(struct amdgpu_device *adev)
 	if (r)
 		return r;
 	adev->gart.table_size = adev->gart.num_gpu_pages * 8;
-	adev->gart.gart_pte_flags = AMDGPU_PTE_MTYPE_VG10(MTYPE_UC) |
+	adev->gart.gart_pte_flags = AMDGPU_PTE_MTYPE_VG10(0ULL, MTYPE_UC) |
 				 AMDGPU_PTE_EXECUTABLE;
 
-	r = amdgpu_gart_table_vram_alloc(adev);
-	if (r)
-		return r;
+	if (!adev->gmc.real_vram_size) {
+		dev_info(adev->dev, "Put GART in system memory for APU\n");
+		r = amdgpu_gart_table_ram_alloc(adev);
+		if (r)
+			dev_err(adev->dev, "Failed to allocate GART in system memory\n");
+	} else {
+		r = amdgpu_gart_table_vram_alloc(adev);
+		if (r)
+			return r;
 
-	if (adev->gmc.xgmi.connected_to_cpu) {
-		r = amdgpu_gmc_pdb0_alloc(adev);
+		if (amdgpu_gmc_is_pdb0_enabled(adev))
+			r = amdgpu_gmc_pdb0_alloc(adev);
 	}
 
 	return r;
@@ -1622,15 +1829,37 @@ static int gmc_v9_0_gart_init(struct amdgpu_device *adev)
  */
 static void gmc_v9_0_save_registers(struct amdgpu_device *adev)
 {
-	if ((adev->ip_versions[DCE_HWIP][0] == IP_VERSION(1, 0, 0)) ||
-	    (adev->ip_versions[DCE_HWIP][0] == IP_VERSION(1, 0, 1)))
+	if ((amdgpu_ip_version(adev, DCE_HWIP, 0) == IP_VERSION(1, 0, 0)) ||
+	    (amdgpu_ip_version(adev, DCE_HWIP, 0) == IP_VERSION(1, 0, 1)))
 		adev->gmc.sdpif_register = RREG32_SOC15(DCE, 0, mmDCHUBBUB_SDPIF_MMIO_CNTRL_0);
 }
 
-static int gmc_v9_0_sw_init(void *handle)
+static void gmc_v9_4_3_init_vram_info(struct amdgpu_device *adev)
+{
+	static const u32 regBIF_BIOS_SCRATCH_4 = 0x50;
+	u32 vram_info;
+
+	adev->gmc.vram_type = AMDGPU_VRAM_TYPE_HBM;
+	adev->gmc.vram_width = 128 * 64;
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 5, 0))
+		adev->gmc.vram_type = AMDGPU_VRAM_TYPE_HBM3E;
+
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 4) &&
+		adev->rev_id == 0x3)
+		adev->gmc.vram_type = AMDGPU_VRAM_TYPE_HBM3E;
+
+	if (!(adev->flags & AMD_IS_APU) && !amdgpu_sriov_vf(adev)) {
+		vram_info = RREG32(regBIF_BIOS_SCRATCH_4);
+		adev->gmc.vram_vendor = vram_info & 0xF;
+	}
+}
+
+static int gmc_v9_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r, vram_width = 0, vram_type = 0, vram_vendor = 0, dma_addr_bits;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+	unsigned long inst_mask = adev->aid_mask;
 
 	adev->gfxhub.funcs->init(adev);
 
@@ -1638,38 +1867,51 @@ static int gmc_v9_0_sw_init(void *handle)
 
 	spin_lock_init(&adev->gmc.invalidate_lock);
 
-	r = amdgpu_atomfirmware_get_vram_info(adev,
-		&vram_width, &vram_type, &vram_vendor);
-	if (amdgpu_sriov_vf(adev))
-		/* For Vega10 SR-IOV, vram_width can't be read from ATOM as RAVEN,
-		 * and DF related registers is not readable, seems hardcord is the
-		 * only way to set the correct vram_width
-		 */
-		adev->gmc.vram_width = 2048;
-	else if (amdgpu_emu_mode != 1)
-		adev->gmc.vram_width = vram_width;
+	if (amdgpu_is_multi_aid(adev)) {
+		gmc_v9_4_3_init_vram_info(adev);
+	} else if (!adev->bios) {
+		if (adev->flags & AMD_IS_APU) {
+			adev->gmc.vram_type = AMDGPU_VRAM_TYPE_DDR4;
+			adev->gmc.vram_width = 64 * 64;
+		} else {
+			adev->gmc.vram_type = AMDGPU_VRAM_TYPE_HBM;
+			adev->gmc.vram_width = 128 * 64;
+		}
+	} else {
+		r = amdgpu_atomfirmware_get_vram_info(adev,
+			&vram_width, &vram_type, &vram_vendor);
+		if (amdgpu_sriov_vf(adev))
+			/* For Vega10 SR-IOV, vram_width can't be read from ATOM as RAVEN,
+			 * and DF related registers is not readable, seems hardcord is the
+			 * only way to set the correct vram_width
+			 */
+			adev->gmc.vram_width = 2048;
+		else if (amdgpu_emu_mode != 1)
+			adev->gmc.vram_width = vram_width;
 
-	if (!adev->gmc.vram_width) {
-		int chansize, numchan;
+		if (!adev->gmc.vram_width) {
+			int chansize, numchan;
 
-		/* hbm memory channel size */
-		if (adev->flags & AMD_IS_APU)
-			chansize = 64;
-		else
-			chansize = 128;
-		if (adev->df.funcs &&
-		    adev->df.funcs->get_hbm_channel_number) {
-			numchan = adev->df.funcs->get_hbm_channel_number(adev);
-			adev->gmc.vram_width = numchan * chansize;
+			/* hbm memory channel size */
+			if (adev->flags & AMD_IS_APU)
+				chansize = 64;
+			else
+				chansize = 128;
+			if (adev->df.funcs &&
+			    adev->df.funcs->get_hbm_channel_number) {
+				numchan = adev->df.funcs->get_hbm_channel_number(adev);
+				adev->gmc.vram_width = numchan * chansize;
+			}
 		}
-	}
 
-	adev->gmc.vram_type = vram_type;
-	adev->gmc.vram_vendor = vram_vendor;
-	switch (adev->ip_versions[GC_HWIP][0]) {
+		adev->gmc.vram_type = vram_type;
+		adev->gmc.vram_vendor = vram_vendor;
+	}
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(9, 1, 0):
 	case IP_VERSION(9, 2, 2):
-		adev->num_vmhubs = 2;
+		set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+		set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
 
 		if (adev->rev_id == 0x0 || adev->rev_id == 0x1) {
 			amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48);
@@ -1685,30 +1927,40 @@ static int gmc_v9_0_sw_init(void *handle)
 	case IP_VERSION(9, 4, 0):
 	case IP_VERSION(9, 3, 0):
 	case IP_VERSION(9, 4, 2):
-	case IP_VERSION(9, 4, 3):
-		adev->num_vmhubs = 2;
-
+		set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+		set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
 
 		/*
 		 * To fulfill 4-level page support,
 		 * vm size is 256TB (48bit), maximum size of Vega10,
 		 * block size 512 (9bit)
 		 */
-		/* sriov restrict max_pfn below AMDGPU_GMC_HOLE */
-		if (amdgpu_sriov_vf(adev))
-			amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 47);
-		else
-			amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48);
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2))
+
+		amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48);
+		if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2))
 			adev->gmc.translate_further = adev->vm_manager.num_level > 1;
 		break;
 	case IP_VERSION(9, 4, 1):
-		adev->num_vmhubs = 3;
+		set_bit(AMDGPU_GFXHUB(0), adev->vmhubs_mask);
+		set_bit(AMDGPU_MMHUB0(0), adev->vmhubs_mask);
+		set_bit(AMDGPU_MMHUB1(0), adev->vmhubs_mask);
 
 		/* Keep the vm size same with Vega20 */
 		amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48);
 		adev->gmc.translate_further = adev->vm_manager.num_level > 1;
 		break;
+	case IP_VERSION(9, 4, 3):
+	case IP_VERSION(9, 4, 4):
+	case IP_VERSION(9, 5, 0):
+		bitmap_set(adev->vmhubs_mask, AMDGPU_GFXHUB(0),
+				  NUM_XCC(adev->gfx.xcc_mask));
+
+		inst_mask <<= AMDGPU_MMHUB0(0);
+		bitmap_or(adev->vmhubs_mask, adev->vmhubs_mask, &inst_mask, 32);
+
+		amdgpu_vm_adjust_size(adev, 256 * 1024, 9, 3, 48);
+		adev->gmc.translate_further = adev->vm_manager.num_level > 1;
+		break;
 	default:
 		break;
 	}
@@ -1719,7 +1971,7 @@ static int gmc_v9_0_sw_init(void *handle)
 	if (r)
 		return r;
 
-	if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1)) {
+	if (amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1)) {
 		r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VMC1, VMC_1_0__SRCID__VM_FAULT,
 					&adev->gmc.vm_fault);
 		if (r)
@@ -1733,7 +1985,8 @@ static int gmc_v9_0_sw_init(void *handle)
 		return r;
 
 	if (!amdgpu_sriov_vf(adev) &&
-	    !adev->gmc.xgmi.connected_to_cpu) {
+	    !adev->gmc.xgmi.connected_to_cpu &&
+	    !adev->gmc.is_app_apu) {
 		/* interrupt sent to DF. */
 		r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_DF, 0,
 				      &adev->gmc.ecc_irq);
@@ -1747,10 +2000,13 @@ static int gmc_v9_0_sw_init(void *handle)
 	 */
 	adev->gmc.mc_mask = 0xffffffffffffULL; /* 48 bit MC */
 
-	dma_addr_bits = adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) ? 48:44;
+	dma_addr_bits = amdgpu_ip_version(adev, GC_HWIP, 0) >=
+					IP_VERSION(9, 4, 2) ?
+				48 :
+				44;
 	r = dma_set_mask_and_coherent(adev->dev, DMA_BIT_MASK(dma_addr_bits));
 	if (r) {
-		printk(KERN_WARNING "amdgpu: No suitable DMA available.\n");
+		dev_warn(adev->dev, "amdgpu: No suitable DMA available.\n");
 		return r;
 	}
 	adev->need_swiotlb = drm_need_swiotlb(dma_addr_bits);
@@ -1761,6 +2017,12 @@ static int gmc_v9_0_sw_init(void *handle)
 
 	amdgpu_gmc_get_vbios_allocations(adev);
 
+	if (amdgpu_is_multi_aid(adev)) {
+		r = amdgpu_gmc_init_mem_ranges(adev);
+		if (r)
+			return r;
+	}
+
 	/* Memory manager */
 	r = amdgpu_bo_init(adev);
 	if (r)
@@ -1770,6 +2032,7 @@ static int gmc_v9_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	gmc_v9_0_init_nps_details(adev);
 	/*
 	 * number of VMs
 	 * VMID 0 is reserved for System
@@ -1781,9 +2044,11 @@ static int gmc_v9_0_sw_init(void *handle)
 	 * for video processing.
 	 */
 	adev->vm_manager.first_kfd_vmid =
-		(adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 1) ||
-		 adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 2) ||
-		 adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3)) ? 3 : 8;
+		(amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 1) ||
+		 amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 2) ||
+		 amdgpu_is_multi_aid(adev)) ?
+			3 :
+			8;
 
 	amdgpu_vm_manager_init(adev);
 
@@ -1793,27 +2058,40 @@ static int gmc_v9_0_sw_init(void *handle)
 	if (r)
 		return r;
 
+	if (amdgpu_is_multi_aid(adev))
+		amdgpu_gmc_sysfs_init(adev);
+
 	return 0;
 }
 
-static int gmc_v9_0_sw_fini(void *handle)
+static int gmc_v9_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (amdgpu_is_multi_aid(adev))
+		amdgpu_gmc_sysfs_fini(adev);
 
 	amdgpu_gmc_ras_fini(adev);
 	amdgpu_gem_force_release(adev);
 	amdgpu_vm_manager_fini(adev);
-	amdgpu_gart_table_vram_free(adev);
+	if (!adev->gmc.real_vram_size) {
+		dev_info(adev->dev, "Put GART in system memory for APU free\n");
+		amdgpu_gart_table_ram_free(adev);
+	} else {
+		amdgpu_gart_table_vram_free(adev);
+	}
 	amdgpu_bo_free_kernel(&adev->gmc.pdb0_bo, NULL, &adev->gmc.ptr_pdb0);
 	amdgpu_bo_fini(adev);
 
+	adev->gmc.num_mem_partitions = 0;
+	kfree(adev->gmc.mem_partitions);
+
 	return 0;
 }
 
 static void gmc_v9_0_init_golden_registers(struct amdgpu_device *adev)
 {
-
-	switch (adev->ip_versions[MMHUB_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, MMHUB_HWIP, 0)) {
 	case IP_VERSION(9, 0, 0):
 		if (amdgpu_sriov_vf(adev))
 			break;
@@ -1847,8 +2125,8 @@ static void gmc_v9_0_init_golden_registers(struct amdgpu_device *adev)
  */
 void gmc_v9_0_restore_registers(struct amdgpu_device *adev)
 {
-	if ((adev->ip_versions[DCE_HWIP][0] == IP_VERSION(1, 0, 0)) ||
-	    (adev->ip_versions[DCE_HWIP][0] == IP_VERSION(1, 0, 1))) {
+	if ((amdgpu_ip_version(adev, DCE_HWIP, 0) == IP_VERSION(1, 0, 0)) ||
+	    (amdgpu_ip_version(adev, DCE_HWIP, 0) == IP_VERSION(1, 0, 1))) {
 		WREG32_SOC15(DCE, 0, mmDCHUBBUB_SDPIF_MMIO_CNTRL_0, adev->gmc.sdpif_register);
 		WARN_ON(adev->gmc.sdpif_register !=
 			RREG32_SOC15(DCE, 0, mmDCHUBBUB_SDPIF_MMIO_CNTRL_0));
@@ -1864,7 +2142,7 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev)
 {
 	int r;
 
-	if (adev->gmc.xgmi.connected_to_cpu)
+	if (amdgpu_gmc_is_pdb0_enabled(adev))
 		amdgpu_gmc_init_pdb0(adev);
 
 	if (adev->gart.bo == NULL) {
@@ -1885,7 +2163,7 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev)
 		return r;
 
 	DRM_INFO("PCIE GART of %uM enabled.\n",
-		 (unsigned)(adev->gmc.gart_size >> 20));
+		 (unsigned int)(adev->gmc.gart_size >> 20));
 	if (adev->gmc.pdb0_bo)
 		DRM_INFO("PDB0 located at 0x%016llX\n",
 				(unsigned long long)amdgpu_bo_gpu_offset(adev->gmc.pdb0_bo));
@@ -1895,12 +2173,23 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev)
 	return 0;
 }
 
-static int gmc_v9_0_hw_init(void *handle)
+static int gmc_v9_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool value;
 	int i, r;
 
+	adev->gmc.flush_pasid_uses_kiq = true;
+
+	/* Vega20+XGMI caches PTEs in TC and TLB. Add a heavy-weight TLB flush
+	 * (type 2), which flushes both. Due to a race condition with
+	 * concurrent memory accesses using the same TLB cache line, we still
+	 * need a second TLB flush after this.
+	 */
+	adev->gmc.flush_tlb_needs_extra_type_2 =
+		amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 0) &&
+		adev->gmc.xgmi.num_physical_nodes;
+
 	/* The sequence of these two function calls matters.*/
 	gmc_v9_0_init_golden_registers(adev);
 
@@ -1917,7 +2206,7 @@ static int gmc_v9_0_hw_init(void *handle)
 	adev->hdp.funcs->init_registers(adev);
 
 	/* After HDP is initialized, flush HDP.*/
-	adev->hdp.funcs->flush_hdp(adev, NULL);
+	amdgpu_device_flush_hdp(adev, NULL);
 
 	if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS)
 		value = false;
@@ -1929,8 +2218,8 @@ static int gmc_v9_0_hw_init(void *handle)
 			adev->gfxhub.funcs->set_fault_enable_default(adev, value);
 		adev->mmhub.funcs->set_fault_enable_default(adev, value);
 	}
-	for (i = 0; i < adev->num_vmhubs; ++i) {
-		if (adev->in_s0ix && (i == AMDGPU_GFXHUB_0))
+	for_each_set_bit(i, adev->vmhubs_mask, AMDGPU_MAX_VMHUBS) {
+		if (adev->in_s0ix && (i == AMDGPU_GFXHUB(0)))
 			continue;
 		gmc_v9_0_flush_gpu_tlb(adev, 0, i, 0);
 	}
@@ -1944,8 +2233,8 @@ static int gmc_v9_0_hw_init(void *handle)
 
 	if (amdgpu_emu_mode == 1)
 		return amdgpu_gmc_vram_checking(adev);
-	else
-		return r;
+
+	return 0;
 }
 
 /**
@@ -1962,9 +2251,9 @@ static void gmc_v9_0_gart_disable(struct amdgpu_device *adev)
 	adev->mmhub.funcs->gart_disable(adev);
 }
 
-static int gmc_v9_0_hw_fini(void *handle)
+static int gmc_v9_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	gmc_v9_0_gart_disable(adev);
 
@@ -1982,55 +2271,70 @@ static int gmc_v9_0_hw_fini(void *handle)
 	if (adev->mmhub.funcs->update_power_gating)
 		adev->mmhub.funcs->update_power_gating(adev, false);
 
-	amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
-	amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
+	/*
+	 * For minimal init, late_init is not called, hence VM fault/RAS irqs
+	 * are not enabled.
+	 */
+	if (adev->init_lvl->level != AMDGPU_INIT_LEVEL_MINIMAL_XGMI) {
+		amdgpu_irq_put(adev, &adev->gmc.vm_fault, 0);
+
+		if (adev->gmc.ecc_irq.funcs &&
+		    amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__UMC))
+			amdgpu_irq_put(adev, &adev->gmc.ecc_irq, 0);
+	}
 
 	return 0;
 }
 
-static int gmc_v9_0_suspend(void *handle)
+static int gmc_v9_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return gmc_v9_0_hw_fini(adev);
+	return gmc_v9_0_hw_fini(ip_block);
 }
 
-static int gmc_v9_0_resume(void *handle)
+static int gmc_v9_0_resume(struct amdgpu_ip_block *ip_block)
 {
+	struct amdgpu_device *adev = ip_block->adev;
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = gmc_v9_0_hw_init(adev);
+	/* If a reset is done for NPS mode switch, read the memory range
+	 * information again.
+	 */
+	if (adev->gmc.reset_flags & AMDGPU_GMC_INIT_RESET_NPS) {
+		amdgpu_gmc_init_sw_mem_ranges(adev, adev->gmc.mem_partitions);
+		adev->gmc.reset_flags &= ~AMDGPU_GMC_INIT_RESET_NPS;
+	}
+
+	r = gmc_v9_0_hw_init(ip_block);
 	if (r)
 		return r;
 
-	amdgpu_vmid_reset_all(adev);
+	amdgpu_vmid_reset_all(ip_block->adev);
 
 	return 0;
 }
 
-static bool gmc_v9_0_is_idle(void *handle)
+static bool gmc_v9_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* MC is always ready in GMC v9.*/
 	return true;
 }
 
-static int gmc_v9_0_wait_for_idle(void *handle)
+static int gmc_v9_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* There is no need to wait for MC idle in GMC v9.*/
 	return 0;
 }
 
-static int gmc_v9_0_soft_reset(void *handle)
+static int gmc_v9_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	/* XXX for emulation.*/
 	return 0;
 }
 
-static int gmc_v9_0_set_clockgating_state(void *handle,
+static int gmc_v9_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->mmhub.funcs->set_clockgating(adev, state);
 
@@ -2039,16 +2343,16 @@ static int gmc_v9_0_set_clockgating_state(void *handle,
 	return 0;
 }
 
-static void gmc_v9_0_get_clockgating_state(void *handle, u64 *flags)
+static void gmc_v9_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	adev->mmhub.funcs->get_clockgating(adev, flags);
 
 	athub_v1_0_get_clockgating(adev, flags);
 }
 
-static int gmc_v9_0_set_powergating_state(void *handle,
+static int gmc_v9_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					enum amd_powergating_state state)
 {
 	return 0;
@@ -2072,8 +2376,7 @@ const struct amd_ip_funcs gmc_v9_0_ip_funcs = {
 	.get_clockgating_state = gmc_v9_0_get_clockgating_state,
 };
 
-const struct amdgpu_ip_block_version gmc_v9_0_ip_block =
-{
+const struct amdgpu_ip_block_version gmc_v9_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_GMC,
 	.major = 9,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c
index 71d1a2e3bac9..e6c0d86d3486 100644
--- a/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/hdp_v4_0.c
@@ -21,7 +21,6 @@
  *
  */
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "hdp_v4_0.h"
 #include "amdgpu_ras.h"
 
@@ -37,27 +36,21 @@
 #define HDP_MEM_POWER_CTRL__RC_MEM_POWER_LS_EN_MASK     0x00020000L
 #define mmHDP_MEM_POWER_CTRL_BASE_IDX   0
 
-static void hdp_v4_0_flush_hdp(struct amdgpu_device *adev,
-				struct amdgpu_ring *ring)
-{
-	if (!ring || !ring->funcs->emit_wreg)
-		WREG32_NO_KIQ((adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, 0);
-	else
-		amdgpu_ring_emit_wreg(ring, (adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, 0);
-}
-
 static void hdp_v4_0_invalidate_hdp(struct amdgpu_device *adev,
 				    struct amdgpu_ring *ring)
 {
-	if (adev->ip_versions[HDP_HWIP][0] == IP_VERSION(4, 4, 0) ||
-	    adev->ip_versions[HDP_HWIP][0] == IP_VERSION(4, 4, 2))
+	if (amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 4, 0) ||
+	    amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 4, 2) ||
+	    amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 4, 5))
 		return;
 
-	if (!ring || !ring->funcs->emit_wreg)
+	if (!ring || !ring->funcs->emit_wreg) {
 		WREG32_SOC15_NO_KIQ(HDP, 0, mmHDP_READ_CACHE_INVALIDATE, 1);
-	else
+		RREG32_SOC15_NO_KIQ(HDP, 0, mmHDP_READ_CACHE_INVALIDATE);
+	} else {
 		amdgpu_ring_emit_wreg(ring, SOC15_REG_OFFSET(
 			HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 1);
+	}
 }
 
 static void hdp_v4_0_query_ras_error_count(struct amdgpu_device *adev,
@@ -80,7 +73,7 @@ static void hdp_v4_0_reset_ras_error_count(struct amdgpu_device *adev)
 	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__HDP))
 		return;
 
-	if (adev->ip_versions[HDP_HWIP][0] >= IP_VERSION(4, 4, 0))
+	if (amdgpu_ip_version(adev, HDP_HWIP, 0) >= IP_VERSION(4, 4, 0))
 		WREG32_SOC15(HDP, 0, mmHDP_EDC_CNT, 0);
 	else
 		/*read back hdp ras counter to reset it to 0 */
@@ -92,10 +85,10 @@ static void hdp_v4_0_update_clock_gating(struct amdgpu_device *adev,
 {
 	uint32_t def, data;
 
-	if (adev->ip_versions[HDP_HWIP][0] == IP_VERSION(4, 0, 0) ||
-	    adev->ip_versions[HDP_HWIP][0] == IP_VERSION(4, 0, 1) ||
-	    adev->ip_versions[HDP_HWIP][0] == IP_VERSION(4, 1, 1) ||
-	    adev->ip_versions[HDP_HWIP][0] == IP_VERSION(4, 1, 0)) {
+	if (amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 0, 0) ||
+	    amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 0, 1) ||
+	    amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 1, 1) ||
+	    amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 1, 0)) {
 		def = data = RREG32(SOC15_REG_OFFSET(HDP, 0, mmHDP_MEM_POWER_LS));
 
 		if (enable && (adev->cg_flags & AMD_CG_SUPPORT_HDP_LS))
@@ -129,6 +122,12 @@ static void hdp_v4_0_get_clockgating_state(struct amdgpu_device *adev,
 {
 	int data;
 
+	if (amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 4, 2) ||
+	    amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 4, 5)) {
+		/* Default enabled */
+		*flags |= AMD_CG_SUPPORT_HDP_MGCG;
+		return;
+	}
 	/* AMD_CG_SUPPORT_HDP_LS */
 	data = RREG32(SOC15_REG_OFFSET(HDP, 0, mmHDP_MEM_POWER_LS));
 	if (data & HDP_MEM_POWER_LS__LS_ENABLE_MASK)
@@ -137,7 +136,7 @@ static void hdp_v4_0_get_clockgating_state(struct amdgpu_device *adev,
 
 static void hdp_v4_0_init_registers(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[HDP_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, HDP_HWIP, 0)) {
 	case IP_VERSION(4, 2, 1):
 		WREG32_FIELD15(HDP, 0, HDP_MMHUB_CNTL, HDP_MMHUB_GCC, 1);
 		break;
@@ -145,9 +144,13 @@ static void hdp_v4_0_init_registers(struct amdgpu_device *adev)
 		break;
 	}
 
+	/* Do not program registers if VF */
+	if (amdgpu_sriov_vf(adev))
+		return;
+
 	WREG32_FIELD15(HDP, 0, HDP_MISC_CNTL, FLUSH_INVALIDATE_CACHE, 1);
 
-	if (adev->ip_versions[HDP_HWIP][0] == IP_VERSION(4, 4, 0))
+	if (amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(4, 4, 0))
 		WREG32_FIELD15(HDP, 0, HDP_MISC_CNTL, READ_BUFFER_WATERMARK, 2);
 
 	WREG32_SOC15(HDP, 0, mmHDP_NONSURFACE_BASE, (adev->gmc.vram_start >> 8));
@@ -166,7 +169,7 @@ struct amdgpu_hdp_ras hdp_v4_0_ras = {
 };
 
 const struct amdgpu_hdp_funcs hdp_v4_0_funcs = {
-	.flush_hdp = hdp_v4_0_flush_hdp,
+	.flush_hdp = amdgpu_hdp_generic_flush,
 	.invalidate_hdp = hdp_v4_0_invalidate_hdp,
 	.update_clock_gating = hdp_v4_0_update_clock_gating,
 	.get_clock_gating_state = hdp_v4_0_get_clockgating_state,
diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v5_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v5_0.c
index a9ea23fa0def..8bc001dc9f63 100644
--- a/drivers/gpu/drm/amd/amdgpu/hdp_v5_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/hdp_v5_0.c
@@ -21,27 +21,18 @@
  *
  */
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "hdp_v5_0.h"
 
 #include "hdp/hdp_5_0_0_offset.h"
 #include "hdp/hdp_5_0_0_sh_mask.h"
 #include <uapi/linux/kfd_ioctl.h>
 
-static void hdp_v5_0_flush_hdp(struct amdgpu_device *adev,
-				struct amdgpu_ring *ring)
-{
-	if (!ring || !ring->funcs->emit_wreg)
-		WREG32_NO_KIQ((adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, 0);
-	else
-		amdgpu_ring_emit_wreg(ring, (adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, 0);
-}
-
 static void hdp_v5_0_invalidate_hdp(struct amdgpu_device *adev,
 				    struct amdgpu_ring *ring)
 {
 	if (!ring || !ring->funcs->emit_wreg) {
 		WREG32_SOC15_NO_KIQ(HDP, 0, mmHDP_READ_CACHE_INVALIDATE, 1);
+		RREG32_SOC15_NO_KIQ(HDP, 0, mmHDP_READ_CACHE_INVALIDATE);
 	} else {
 		amdgpu_ring_emit_wreg(ring, SOC15_REG_OFFSET(
 					HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 1);
@@ -215,7 +206,7 @@ static void hdp_v5_0_init_registers(struct amdgpu_device *adev)
 }
 
 const struct amdgpu_hdp_funcs hdp_v5_0_funcs = {
-	.flush_hdp = hdp_v5_0_flush_hdp,
+	.flush_hdp = amdgpu_hdp_generic_flush,
 	.invalidate_hdp = hdp_v5_0_invalidate_hdp,
 	.update_clock_gating = hdp_v5_0_update_clock_gating,
 	.get_clock_gating_state = hdp_v5_0_get_clockgating_state,
diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v5_2.c b/drivers/gpu/drm/amd/amdgpu/hdp_v5_2.c
index 29c3484ae1f1..40940b4ab400 100644
--- a/drivers/gpu/drm/amd/amdgpu/hdp_v5_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/hdp_v5_2.c
@@ -21,7 +21,6 @@
  *
  */
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "hdp_v5_2.h"
 
 #include "hdp/hdp_5_2_1_offset.h"
@@ -31,13 +30,25 @@
 static void hdp_v5_2_flush_hdp(struct amdgpu_device *adev,
 				struct amdgpu_ring *ring)
 {
-	if (!ring || !ring->funcs->emit_wreg)
+	if (!ring || !ring->funcs->emit_wreg) {
 		WREG32_NO_KIQ((adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2,
 			0);
-	else
+		if (amdgpu_sriov_vf(adev)) {
+			/* this is fine because SR_IOV doesn't remap the register */
+			RREG32_NO_KIQ((adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2);
+		} else {
+			/* We just need to read back a register to post the write.
+			 * Reading back the remapped register causes problems on
+			 * some platforms so just read back the memory size register.
+			 */
+			if (adev->nbio.funcs->get_memsize)
+				adev->nbio.funcs->get_memsize(adev);
+		}
+	} else {
 		amdgpu_ring_emit_wreg(ring,
 			(adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2,
 			0);
+	}
 }
 
 static void hdp_v5_2_update_mem_power_gating(struct amdgpu_device *adev,
diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v6_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v6_0.c
index 063eba619f2f..ec20daf4272c 100644
--- a/drivers/gpu/drm/amd/amdgpu/hdp_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/hdp_v6_0.c
@@ -21,26 +21,19 @@
  *
  */
 #include "amdgpu.h"
-#include "amdgpu_atombios.h"
 #include "hdp_v6_0.h"
 
 #include "hdp/hdp_6_0_0_offset.h"
 #include "hdp/hdp_6_0_0_sh_mask.h"
 #include <uapi/linux/kfd_ioctl.h>
 
-static void hdp_v6_0_flush_hdp(struct amdgpu_device *adev,
-				struct amdgpu_ring *ring)
-{
-	if (!ring || !ring->funcs->emit_wreg)
-		WREG32_NO_KIQ((adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, 0);
-	else
-		amdgpu_ring_emit_wreg(ring, (adev->rmmio_remap.reg_offset + KFD_MMIO_REMAP_HDP_MEM_FLUSH_CNTL) >> 2, 0);
-}
+#define regHDP_CLK_CNTL_V6_1	0xd5
+#define regHDP_CLK_CNTL_V6_1_BASE_IDX 0
 
 static void hdp_v6_0_update_clock_gating(struct amdgpu_device *adev,
 					 bool enable)
 {
-	uint32_t hdp_clk_cntl, hdp_clk_cntl1;
+	uint32_t hdp_clk_cntl;
 	uint32_t hdp_mem_pwr_cntl;
 
 	if (!(adev->cg_flags & (AMD_CG_SUPPORT_HDP_LS |
@@ -48,14 +41,20 @@ static void hdp_v6_0_update_clock_gating(struct amdgpu_device *adev,
 				AMD_CG_SUPPORT_HDP_SD)))
 		return;
 
-	hdp_clk_cntl = hdp_clk_cntl1 = RREG32_SOC15(HDP, 0,regHDP_CLK_CNTL);
+	if (amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(6, 1, 0))
+		hdp_clk_cntl = RREG32_SOC15(HDP, 0, regHDP_CLK_CNTL_V6_1);
+	else
+		hdp_clk_cntl = RREG32_SOC15(HDP, 0, regHDP_CLK_CNTL);
 	hdp_mem_pwr_cntl = RREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL);
 
 	/* Before doing clock/power mode switch,
 	 * forced on IPH & RC clock */
 	hdp_clk_cntl = REG_SET_FIELD(hdp_clk_cntl, HDP_CLK_CNTL,
 				     RC_MEM_CLK_SOFT_OVERRIDE, 1);
-	WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl);
+	if (amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(6, 1, 0))
+		WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL_V6_1, hdp_clk_cntl);
+	else
+		WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl);
 
 	/* disable clock and power gating before any changing */
 	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
@@ -117,7 +116,10 @@ static void hdp_v6_0_update_clock_gating(struct amdgpu_device *adev,
 	/* disable IPH & RC clock override after clock/power mode changing */
 	hdp_clk_cntl = REG_SET_FIELD(hdp_clk_cntl, HDP_CLK_CNTL,
 				     RC_MEM_CLK_SOFT_OVERRIDE, 0);
-	WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl);
+	if (amdgpu_ip_version(adev, HDP_HWIP, 0) == IP_VERSION(6, 1, 0))
+		WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL_V6_1, hdp_clk_cntl);
+	else
+		WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl);
 }
 
 static void hdp_v6_0_get_clockgating_state(struct amdgpu_device *adev,
@@ -136,7 +138,7 @@ static void hdp_v6_0_get_clockgating_state(struct amdgpu_device *adev,
 }
 
 const struct amdgpu_hdp_funcs hdp_v6_0_funcs = {
-	.flush_hdp = hdp_v6_0_flush_hdp,
+	.flush_hdp = amdgpu_hdp_generic_flush,
 	.update_clock_gating = hdp_v6_0_update_clock_gating,
 	.get_clock_gating_state = hdp_v6_0_get_clockgating_state,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v7_0.c b/drivers/gpu/drm/amd/amdgpu/hdp_v7_0.c
new file mode 100644
index 000000000000..ed1debc03507
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/hdp_v7_0.c
@@ -0,0 +1,132 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "amdgpu.h"
+#include "hdp_v7_0.h"
+
+#include "hdp/hdp_7_0_0_offset.h"
+#include "hdp/hdp_7_0_0_sh_mask.h"
+#include <uapi/linux/kfd_ioctl.h>
+
+static void hdp_v7_0_update_clock_gating(struct amdgpu_device *adev,
+					 bool enable)
+{
+	uint32_t hdp_clk_cntl, hdp_clk_cntl1;
+	uint32_t hdp_mem_pwr_cntl;
+
+	if (!(adev->cg_flags & (AMD_CG_SUPPORT_HDP_LS |
+				AMD_CG_SUPPORT_HDP_DS |
+				AMD_CG_SUPPORT_HDP_SD)))
+		return;
+
+	hdp_clk_cntl = hdp_clk_cntl1 = RREG32_SOC15(HDP, 0,regHDP_CLK_CNTL);
+	hdp_mem_pwr_cntl = RREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL);
+
+	/* Before doing clock/power mode switch,
+	 * forced on IPH & RC clock */
+	hdp_clk_cntl = REG_SET_FIELD(hdp_clk_cntl, HDP_CLK_CNTL,
+				     RC_MEM_CLK_SOFT_OVERRIDE, 1);
+	WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl);
+
+	/* disable clock and power gating before any changing */
+	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+					 ATOMIC_MEM_POWER_CTRL_EN, 0);
+	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+					 ATOMIC_MEM_POWER_LS_EN, 0);
+	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+					 ATOMIC_MEM_POWER_DS_EN, 0);
+	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+					 ATOMIC_MEM_POWER_SD_EN, 0);
+	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+					 RC_MEM_POWER_CTRL_EN, 0);
+	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+					 RC_MEM_POWER_LS_EN, 0);
+	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+					 RC_MEM_POWER_DS_EN, 0);
+	hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+					 RC_MEM_POWER_SD_EN, 0);
+	WREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL, hdp_mem_pwr_cntl);
+
+	/* Already disabled above. The actions below are for "enabled" only */
+	if (enable) {
+		/* only one clock gating mode (LS/DS/SD) can be enabled */
+		if (adev->cg_flags & AMD_CG_SUPPORT_HDP_SD) {
+			hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl,
+							 HDP_MEM_POWER_CTRL,
+							 ATOMIC_MEM_POWER_SD_EN, 1);
+			hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl,
+							 HDP_MEM_POWER_CTRL,
+							 RC_MEM_POWER_SD_EN, 1);
+		} else if (adev->cg_flags & AMD_CG_SUPPORT_HDP_LS) {
+			hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl,
+							 HDP_MEM_POWER_CTRL,
+							 ATOMIC_MEM_POWER_LS_EN, 1);
+			hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl,
+							 HDP_MEM_POWER_CTRL,
+							 RC_MEM_POWER_LS_EN, 1);
+		} else if (adev->cg_flags & AMD_CG_SUPPORT_HDP_DS) {
+			hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl,
+							 HDP_MEM_POWER_CTRL,
+							 ATOMIC_MEM_POWER_DS_EN, 1);
+			hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl,
+							 HDP_MEM_POWER_CTRL,
+							 RC_MEM_POWER_DS_EN, 1);
+		}
+
+		/* confirmed that IPH_MEM_POWER_CTRL_EN and RC_MEM_POWER_CTRL_EN have to
+		 * be set for SRAM LS/DS/SD */
+		if (adev->cg_flags & (AMD_CG_SUPPORT_HDP_LS | AMD_CG_SUPPORT_HDP_DS |
+				      AMD_CG_SUPPORT_HDP_SD)) {
+			hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+							 ATOMIC_MEM_POWER_CTRL_EN, 1);
+			hdp_mem_pwr_cntl = REG_SET_FIELD(hdp_mem_pwr_cntl, HDP_MEM_POWER_CTRL,
+							 RC_MEM_POWER_CTRL_EN, 1);
+			WREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL, hdp_mem_pwr_cntl);
+		}
+	}
+
+	/* disable IPH & RC clock override after clock/power mode changing */
+	hdp_clk_cntl = REG_SET_FIELD(hdp_clk_cntl, HDP_CLK_CNTL,
+				     RC_MEM_CLK_SOFT_OVERRIDE, 0);
+	WREG32_SOC15(HDP, 0, regHDP_CLK_CNTL, hdp_clk_cntl);
+}
+
+static void hdp_v7_0_get_clockgating_state(struct amdgpu_device *adev,
+					    u64 *flags)
+{
+	uint32_t tmp;
+
+	/* AMD_CG_SUPPORT_HDP_LS/DS/SD */
+	tmp = RREG32_SOC15(HDP, 0, regHDP_MEM_POWER_CTRL);
+	if (tmp & HDP_MEM_POWER_CTRL__ATOMIC_MEM_POWER_LS_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_HDP_LS;
+	else if (tmp & HDP_MEM_POWER_CTRL__ATOMIC_MEM_POWER_DS_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_HDP_DS;
+	else if (tmp & HDP_MEM_POWER_CTRL__ATOMIC_MEM_POWER_SD_EN_MASK)
+		*flags |= AMD_CG_SUPPORT_HDP_SD;
+}
+
+const struct amdgpu_hdp_funcs hdp_v7_0_funcs = {
+	.flush_hdp = amdgpu_hdp_generic_flush,
+	.update_clock_gating = hdp_v7_0_update_clock_gating,
+	.get_clock_gating_state = hdp_v7_0_get_clockgating_state,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/hdp_v7_0.h b/drivers/gpu/drm/amd/amdgpu/hdp_v7_0.h
new file mode 100644
index 000000000000..25b69201402d
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/hdp_v7_0.h
@@ -0,0 +1,31 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __HDP_V7_0_H__
+#define __HDP_V7_0_H__
+
+#include "soc15_common.h"
+
+extern const struct amdgpu_hdp_funcs hdp_v7_0_funcs;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c
index aecad530b10a..01cadf898c00 100644
--- a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c
@@ -157,6 +157,9 @@ static int iceland_ih_irq_init(struct amdgpu_device *adev)
 	/* enable interrupts */
 	iceland_ih_enable_interrupts(adev);
 
+	if (adev->irq.ih_soft.ring_size)
+		adev->irq.ih_soft.enabled = true;
+
 	return 0;
 }
 
@@ -194,6 +197,9 @@ static u32 iceland_ih_get_wptr(struct amdgpu_device *adev,
 
 	wptr = le32_to_cpu(*ih->wptr_cpu);
 
+	if (ih == &adev->irq.ih_soft)
+		goto out;
+
 	if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
 		goto out;
 
@@ -215,6 +221,11 @@ static u32 iceland_ih_get_wptr(struct amdgpu_device *adev,
 	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
 	WREG32(mmIH_RB_CNTL, tmp);
 
+	/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+	 * can be detected.
+	 */
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	WREG32(mmIH_RB_CNTL, tmp);
 
 out:
 	return (wptr & ih->ptr_mask);
@@ -268,9 +279,9 @@ static void iceland_ih_set_rptr(struct amdgpu_device *adev,
 	WREG32(mmIH_RB_RPTR, ih->rptr);
 }
 
-static int iceland_ih_early_init(void *handle)
+static int iceland_ih_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	int ret;
 
 	ret = amdgpu_irq_add_domain(adev);
@@ -282,23 +293,27 @@ static int iceland_ih_early_init(void *handle)
 	return 0;
 }
 
-static int iceland_ih_sw_init(void *handle)
+static int iceland_ih_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, false);
 	if (r)
 		return r;
 
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, IH_SW_RING_SIZE, true);
+	if (r)
+		return r;
+
 	r = amdgpu_irq_init(adev);
 
 	return r;
 }
 
-static int iceland_ih_sw_fini(void *handle)
+static int iceland_ih_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_fini_sw(adev);
 	amdgpu_irq_remove_domain(adev);
@@ -306,39 +321,33 @@ static int iceland_ih_sw_fini(void *handle)
 	return 0;
 }
 
-static int iceland_ih_hw_init(void *handle)
+static int iceland_ih_hw_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	return iceland_ih_irq_init(adev);
 }
 
-static int iceland_ih_hw_fini(void *handle)
+static int iceland_ih_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	iceland_ih_irq_disable(adev);
+	iceland_ih_irq_disable(ip_block->adev);
 
 	return 0;
 }
 
-static int iceland_ih_suspend(void *handle)
+static int iceland_ih_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return iceland_ih_hw_fini(adev);
+	return iceland_ih_hw_fini(ip_block);
 }
 
-static int iceland_ih_resume(void *handle)
+static int iceland_ih_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return iceland_ih_hw_init(adev);
+	return iceland_ih_hw_init(ip_block);
 }
 
-static bool iceland_ih_is_idle(void *handle)
+static bool iceland_ih_is_idle(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (REG_GET_FIELD(tmp, SRBM_STATUS, IH_BUSY))
@@ -347,11 +356,11 @@ static bool iceland_ih_is_idle(void *handle)
 	return true;
 }
 
-static int iceland_ih_wait_for_idle(void *handle)
+static int iceland_ih_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	unsigned i;
 	u32 tmp;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	for (i = 0; i < adev->usec_timeout; i++) {
 		/* read MC_STATUS */
@@ -363,10 +372,10 @@ static int iceland_ih_wait_for_idle(void *handle)
 	return -ETIMEDOUT;
 }
 
-static int iceland_ih_soft_reset(void *handle)
+static int iceland_ih_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	u32 srbm_soft_reset = 0;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	u32 tmp = RREG32(mmSRBM_STATUS);
 
 	if (tmp & SRBM_STATUS__IH_BUSY_MASK)
@@ -393,13 +402,13 @@ static int iceland_ih_soft_reset(void *handle)
 	return 0;
 }
 
-static int iceland_ih_set_clockgating_state(void *handle,
+static int iceland_ih_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_clockgating_state state)
 {
 	return 0;
 }
 
-static int iceland_ih_set_powergating_state(void *handle,
+static int iceland_ih_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					  enum amd_powergating_state state)
 {
 	return 0;
@@ -408,7 +417,6 @@ static int iceland_ih_set_powergating_state(void *handle,
 static const struct amd_ip_funcs iceland_ih_ip_funcs = {
 	.name = "iceland_ih",
 	.early_init = iceland_ih_early_init,
-	.late_init = NULL,
 	.sw_init = iceland_ih_sw_init,
 	.sw_fini = iceland_ih_sw_fini,
 	.hw_init = iceland_ih_hw_init,
diff --git a/drivers/gpu/drm/amd/amdgpu/ih_v6_0.c b/drivers/gpu/drm/amd/amdgpu/ih_v6_0.c
index b02e1cef78a7..333e9c30c091 100644
--- a/drivers/gpu/drm/amd/amdgpu/ih_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/ih_v6_0.c
@@ -135,6 +135,34 @@ static int ih_v6_0_toggle_ring_interrupts(struct amdgpu_device *adev,
 
 	tmp = RREG32(ih_regs->ih_rb_cntl);
 	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_ENABLE, (enable ? 1 : 0));
+
+	if (enable) {
+		/* Unset the CLEAR_OVERFLOW bit to make sure the next step
+		 * is switching the bit from 0 to 1
+		 */
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+		if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+			if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp))
+				return -ETIMEDOUT;
+		} else {
+			WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+		}
+
+		/* Clear RB_OVERFLOW bit */
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
+		if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+			if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp))
+				return -ETIMEDOUT;
+		} else {
+			WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+		}
+
+		/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+		 * can be detected.
+		 */
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	}
+
 	/* enable_intr field is only valid in ring0 */
 	if (ih == &adev->irq.ih)
 		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, ENABLE_INTR, (enable ? 1 : 0));
@@ -321,6 +349,7 @@ static int ih_v6_0_irq_init(struct amdgpu_device *adev)
 			if (ret)
 				return ret;
 		}
+		ih[i]->overflow = false;
 	}
 
 	/* update doorbell range for ih ring 0 */
@@ -346,6 +375,21 @@ static int ih_v6_0_irq_init(struct amdgpu_device *adev)
 			    DELAY, 3);
 	WREG32_SOC15(OSSSYS, 0, regIH_MSI_STORM_CTRL, tmp);
 
+	/* Redirect the interrupts to IH RB1 for dGPU */
+	if (adev->irq.ih1.ring_size) {
+		tmp = RREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_INDEX);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_INDEX, INDEX, 0);
+		WREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_INDEX, tmp);
+
+		tmp = RREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_DATA);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA, CLIENT_ID, 0xa);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA, SOURCE_ID, 0x0);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA,
+				    SOURCE_ID_MATCH_ENABLE, 0x1);
+
+		WREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_DATA, tmp);
+	}
+
 	pci_set_master(adev->pdev);
 
 	/* enable interrupts */
@@ -403,7 +447,10 @@ static u32 ih_v6_0_get_wptr(struct amdgpu_device *adev,
 	wptr = RREG32_NO_KIQ(ih_regs->ih_rb_wptr);
 	if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
 		goto out;
-	wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0);
+	if (!amdgpu_sriov_vf(adev))
+		wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0);
+	else
+		ih->overflow = true;
 
 	/* When a ring buffer overflow happen start parsing interrupt
 	 * from the last not overwritten vector (wptr + 32). Hopefully
@@ -418,6 +465,12 @@ static u32 ih_v6_0_get_wptr(struct amdgpu_device *adev,
 	tmp = RREG32_NO_KIQ(ih_regs->ih_rb_cntl);
 	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
 	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+
+	/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+	 * can be detected.
+	 */
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
 out:
 	return (wptr & ih->ptr_mask);
 }
@@ -494,7 +547,8 @@ static int ih_v6_0_self_irq(struct amdgpu_device *adev,
 		*adev->irq.ih1.wptr_cpu = wptr;
 		schedule_work(&adev->irq.ih1_work);
 		break;
-	default: break;
+	default:
+		break;
 	}
 	return 0;
 }
@@ -509,19 +563,19 @@ static void ih_v6_0_set_self_irq_funcs(struct amdgpu_device *adev)
 	adev->irq.self_irq.funcs = &ih_v6_0_self_irq_funcs;
 }
 
-static int ih_v6_0_early_init(void *handle)
+static int ih_v6_0_early_init(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	ih_v6_0_set_interrupt_funcs(adev);
 	ih_v6_0_set_self_irq_funcs(adev);
 	return 0;
 }
 
-static int ih_v6_0_sw_init(void *handle)
+static int ih_v6_0_sw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool use_bus_addr;
 
 	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_IH, 0,
@@ -533,22 +587,28 @@ static int ih_v6_0_sw_init(void *handle)
 	/* use gpu virtual address for ih ring
 	 * until ih_checken is programmed to allow
 	 * use bus address for ih ring by psp bl */
-	use_bus_addr =
-		(adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) ? false : true;
-	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 256 * 1024, use_bus_addr);
+	use_bus_addr = adev->firmware.load_type != AMDGPU_FW_LOAD_PSP;
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, IH_RING_SIZE, use_bus_addr);
 	if (r)
 		return r;
 
 	adev->irq.ih.use_doorbell = true;
 	adev->irq.ih.doorbell_index = adev->doorbell_index.ih << 1;
 
-	adev->irq.ih1.ring_size = 0;
-	adev->irq.ih2.ring_size = 0;
+	if (!(adev->flags & AMD_IS_APU)) {
+		r = amdgpu_ih_ring_init(adev, &adev->irq.ih1, IH_RING_SIZE,
+					use_bus_addr);
+		if (r)
+			return r;
+
+		adev->irq.ih1.use_doorbell = true;
+		adev->irq.ih1.doorbell_index = (adev->doorbell_index.ih + 1) << 1;
+	}
 
 	/* initialize ih control register offset */
 	ih_v6_0_init_register_offset(adev);
 
-	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, PAGE_SIZE, true);
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, IH_SW_RING_SIZE, true);
 	if (r)
 		return r;
 
@@ -557,19 +617,19 @@ static int ih_v6_0_sw_init(void *handle)
 	return r;
 }
 
-static int ih_v6_0_sw_fini(void *handle)
+static int ih_v6_0_sw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	amdgpu_irq_fini_sw(adev);
 
 	return 0;
 }
 
-static int ih_v6_0_hw_init(void *handle)
+static int ih_v6_0_hw_init(struct amdgpu_ip_block *ip_block)
 {
 	int r;
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	r = ih_v6_0_irq_init(adev);
 	if (r)
@@ -578,42 +638,36 @@ static int ih_v6_0_hw_init(void *handle)
 	return 0;
 }
 
-static int ih_v6_0_hw_fini(void *handle)
+static int ih_v6_0_hw_fini(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	ih_v6_0_irq_disable(adev);
+	ih_v6_0_irq_disable(ip_block->adev);
 
 	return 0;
 }
 
-static int ih_v6_0_suspend(void *handle)
+static int ih_v6_0_suspend(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return ih_v6_0_hw_fini(adev);
+	return ih_v6_0_hw_fini(ip_block);
 }
 
-static int ih_v6_0_resume(void *handle)
+static int ih_v6_0_resume(struct amdgpu_ip_block *ip_block)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
-
-	return ih_v6_0_hw_init(adev);
+	return ih_v6_0_hw_init(ip_block);
 }
 
-static bool ih_v6_0_is_idle(void *handle)
+static bool ih_v6_0_is_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 	return true;
 }
 
-static int ih_v6_0_wait_for_idle(void *handle)
+static int ih_v6_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 	return -ETIMEDOUT;
 }
 
-static int ih_v6_0_soft_reset(void *handle)
+static int ih_v6_0_soft_reset(struct amdgpu_ip_block *ip_block)
 {
 	/* todo */
 	return 0;
@@ -640,14 +694,12 @@ static void ih_v6_0_update_clockgating_state(struct amdgpu_device *adev,
 		if (def != data)
 			WREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL, data);
 	}
-
-	return;
 }
 
-static int ih_v6_0_set_clockgating_state(void *handle,
+static int ih_v6_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
 					   enum amd_clockgating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	ih_v6_0_update_clockgating_state(adev,
 				state == AMD_CG_STATE_GATE);
@@ -707,10 +759,10 @@ static void ih_v6_0_update_ih_mem_power_gating(struct amdgpu_device *adev,
 	WREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL, ih_mem_pwr_cntl);
 }
 
-static int ih_v6_0_set_powergating_state(void *handle,
+static int ih_v6_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
 					 enum amd_powergating_state state)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 	bool enable = (state == AMD_PG_STATE_GATE);
 
 	if (adev->pg_flags & AMD_PG_SUPPORT_IH_SRAM_PG)
@@ -719,20 +771,17 @@ static int ih_v6_0_set_powergating_state(void *handle,
 	return 0;
 }
 
-static void ih_v6_0_get_clockgating_state(void *handle, u64 *flags)
+static void ih_v6_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	struct amdgpu_device *adev = ip_block->adev;
 
 	if (!RREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL))
 		*flags |= AMD_CG_SUPPORT_IH_CG;
-
-	return;
 }
 
 static const struct amd_ip_funcs ih_v6_0_ip_funcs = {
 	.name = "ih_v6_0",
 	.early_init = ih_v6_0_early_init,
-	.late_init = NULL,
 	.sw_init = ih_v6_0_sw_init,
 	.sw_fini = ih_v6_0_sw_fini,
 	.hw_init = ih_v6_0_hw_init,
@@ -759,8 +808,7 @@ static void ih_v6_0_set_interrupt_funcs(struct amdgpu_device *adev)
 	adev->irq.ih_funcs = &ih_v6_0_funcs;
 }
 
-const struct amdgpu_ip_block_version ih_v6_0_ip_block =
-{
+const struct amdgpu_ip_block_version ih_v6_0_ip_block = {
 	.type = AMD_IP_BLOCK_TYPE_IH,
 	.major = 6,
 	.minor = 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/ih_v6_1.c b/drivers/gpu/drm/amd/amdgpu/ih_v6_1.c
new file mode 100644
index 000000000000..95b3f4e55ec3
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/ih_v6_1.c
@@ -0,0 +1,796 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/pci.h>
+
+#include "amdgpu.h"
+#include "amdgpu_ih.h"
+
+#include "oss/osssys_6_1_0_offset.h"
+#include "oss/osssys_6_1_0_sh_mask.h"
+
+#include "soc15_common.h"
+#include "ih_v6_1.h"
+
+#define MAX_REARM_RETRY 10
+
+static void ih_v6_1_set_interrupt_funcs(struct amdgpu_device *adev);
+
+/**
+ * ih_v6_1_init_register_offset - Initialize register offset for ih rings
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Initialize register offset ih rings (IH_V6_0).
+ */
+static void ih_v6_1_init_register_offset(struct amdgpu_device *adev)
+{
+	struct amdgpu_ih_regs *ih_regs;
+
+	/* ih ring 2 is removed
+	 * ih ring and ih ring 1 are available */
+	if (adev->irq.ih.ring_size) {
+		ih_regs = &adev->irq.ih.ih_regs;
+		ih_regs->ih_rb_base = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE);
+		ih_regs->ih_rb_base_hi = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE_HI);
+		ih_regs->ih_rb_cntl = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_CNTL);
+		ih_regs->ih_rb_wptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR);
+		ih_regs->ih_rb_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_RPTR);
+		ih_regs->ih_doorbell_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_DOORBELL_RPTR);
+		ih_regs->ih_rb_wptr_addr_lo = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR_ADDR_LO);
+		ih_regs->ih_rb_wptr_addr_hi = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR_ADDR_HI);
+		ih_regs->psp_reg_id = PSP_REG_IH_RB_CNTL;
+	}
+
+	if (adev->irq.ih1.ring_size) {
+		ih_regs = &adev->irq.ih1.ih_regs;
+		ih_regs->ih_rb_base = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE_RING1);
+		ih_regs->ih_rb_base_hi = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE_HI_RING1);
+		ih_regs->ih_rb_cntl = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_CNTL_RING1);
+		ih_regs->ih_rb_wptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR_RING1);
+		ih_regs->ih_rb_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_RPTR_RING1);
+		ih_regs->ih_doorbell_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_DOORBELL_RPTR_RING1);
+		ih_regs->psp_reg_id = PSP_REG_IH_RB_CNTL_RING1;
+	}
+}
+
+/**
+ * force_update_wptr_for_self_int - Force update the wptr for self interrupt
+ *
+ * @adev: amdgpu_device pointer
+ * @threshold: threshold to trigger the wptr reporting
+ * @timeout: timeout to trigger the wptr reporting
+ * @enabled: Enable/disable timeout flush mechanism
+ *
+ * threshold input range: 0 ~ 15, default 0,
+ * real_threshold = 2^threshold
+ * timeout input range: 0 ~ 20, default 8,
+ * real_timeout = (2^timeout) * 1024 / (socclk_freq)
+ *
+ * Force update wptr for self interrupt ( >= SIENNA_CICHLID).
+ */
+static void
+force_update_wptr_for_self_int(struct amdgpu_device *adev,
+			       u32 threshold, u32 timeout, bool enabled)
+{
+	u32 ih_cntl, ih_rb_cntl;
+
+	ih_cntl = RREG32_SOC15(OSSSYS, 0, regIH_CNTL2);
+	ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, regIH_RB_CNTL_RING1);
+
+	ih_cntl = REG_SET_FIELD(ih_cntl, IH_CNTL2,
+				SELF_IV_FORCE_WPTR_UPDATE_TIMEOUT, timeout);
+	ih_cntl = REG_SET_FIELD(ih_cntl, IH_CNTL2,
+				SELF_IV_FORCE_WPTR_UPDATE_ENABLE, enabled);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL_RING1,
+				   RB_USED_INT_THRESHOLD, threshold);
+
+	if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+		if (psp_reg_program(&adev->psp, PSP_REG_IH_RB_CNTL_RING1, ih_rb_cntl))
+			return;
+	} else {
+		WREG32_SOC15(OSSSYS, 0, regIH_RB_CNTL_RING1, ih_rb_cntl);
+	}
+
+	WREG32_SOC15(OSSSYS, 0, regIH_CNTL2, ih_cntl);
+}
+
+/**
+ * ih_v6_1_toggle_ring_interrupts - toggle the interrupt ring buffer
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: amdgpu_ih_ring pointer
+ * @enable: true - enable the interrupts, false - disable the interrupts
+ *
+ * Toggle the interrupt ring buffer (IH_V6_0)
+ */
+static int ih_v6_1_toggle_ring_interrupts(struct amdgpu_device *adev,
+					  struct amdgpu_ih_ring *ih,
+					  bool enable)
+{
+	struct amdgpu_ih_regs *ih_regs;
+	uint32_t tmp;
+
+	ih_regs = &ih->ih_regs;
+
+	tmp = RREG32(ih_regs->ih_rb_cntl);
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_ENABLE, (enable ? 1 : 0));
+	/* enable_intr field is only valid in ring0 */
+	if (ih == &adev->irq.ih)
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, ENABLE_INTR, (enable ? 1 : 0));
+
+	if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+		if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp))
+			return -ETIMEDOUT;
+	} else {
+		WREG32(ih_regs->ih_rb_cntl, tmp);
+	}
+
+	if (enable) {
+		ih->enabled = true;
+	} else {
+		/* set rptr, wptr to 0 */
+		WREG32(ih_regs->ih_rb_rptr, 0);
+		WREG32(ih_regs->ih_rb_wptr, 0);
+		ih->enabled = false;
+		ih->rptr = 0;
+	}
+
+	return 0;
+}
+
+/**
+ * ih_v6_1_toggle_interrupts - Toggle all the available interrupt ring buffers
+ *
+ * @adev: amdgpu_device pointer
+ * @enable: enable or disable interrupt ring buffers
+ *
+ * Toggle all the available interrupt ring buffers (IH_V6_0).
+ */
+static int ih_v6_1_toggle_interrupts(struct amdgpu_device *adev, bool enable)
+{
+	struct amdgpu_ih_ring *ih[] = {&adev->irq.ih, &adev->irq.ih1};
+	int i;
+	int r;
+
+	for (i = 0; i < ARRAY_SIZE(ih); i++) {
+		if (ih[i]->ring_size) {
+			r = ih_v6_1_toggle_ring_interrupts(adev, ih[i], enable);
+			if (r)
+				return r;
+		}
+	}
+
+	return 0;
+}
+
+static uint32_t ih_v6_1_rb_cntl(struct amdgpu_ih_ring *ih, uint32_t ih_rb_cntl)
+{
+	int rb_bufsz = order_base_2(ih->ring_size / 4);
+
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+				   MC_SPACE, ih->use_bus_addr ? 2 : 4);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+				   WPTR_OVERFLOW_CLEAR, 1);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+				   WPTR_OVERFLOW_ENABLE, 1);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, RB_SIZE, rb_bufsz);
+	/* Ring Buffer write pointer writeback. If enabled, IH_RB_WPTR register
+	 * value is written to memory
+	 */
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+				   WPTR_WRITEBACK_ENABLE, 1);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_SNOOP, 1);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_RO, 0);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_VMID, 0);
+
+	return ih_rb_cntl;
+}
+
+static uint32_t ih_v6_1_doorbell_rptr(struct amdgpu_ih_ring *ih)
+{
+	u32 ih_doorbell_rtpr = 0;
+
+	if (ih->use_doorbell) {
+		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+						 IH_DOORBELL_RPTR, OFFSET,
+						 ih->doorbell_index);
+		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+						 IH_DOORBELL_RPTR,
+						 ENABLE, 1);
+	} else {
+		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+						 IH_DOORBELL_RPTR,
+						 ENABLE, 0);
+	}
+	return ih_doorbell_rtpr;
+}
+
+/**
+ * ih_v6_1_enable_ring - enable an ih ring buffer
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: amdgpu_ih_ring pointer
+ *
+ * Enable an ih ring buffer (IH_V6_0)
+ */
+static int ih_v6_1_enable_ring(struct amdgpu_device *adev,
+				      struct amdgpu_ih_ring *ih)
+{
+	struct amdgpu_ih_regs *ih_regs;
+	uint32_t tmp;
+
+	ih_regs = &ih->ih_regs;
+
+	/* Ring Buffer base. [39:8] of 40-bit address of the beginning of the ring buffer*/
+	WREG32(ih_regs->ih_rb_base, ih->gpu_addr >> 8);
+	WREG32(ih_regs->ih_rb_base_hi, (ih->gpu_addr >> 40) & 0xff);
+
+	tmp = RREG32(ih_regs->ih_rb_cntl);
+	tmp = ih_v6_1_rb_cntl(ih, tmp);
+	if (ih == &adev->irq.ih)
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RPTR_REARM, !!adev->irq.msi_enabled);
+	if (ih == &adev->irq.ih1) {
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_ENABLE, 0);
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_FULL_DRAIN_ENABLE, 1);
+	}
+
+	if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+		if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp)) {
+			DRM_ERROR("PSP program IH_RB_CNTL failed!\n");
+			return -ETIMEDOUT;
+		}
+	} else {
+		WREG32(ih_regs->ih_rb_cntl, tmp);
+	}
+
+	if (ih == &adev->irq.ih) {
+		/* set the ih ring 0 writeback address whether it's enabled or not */
+		WREG32(ih_regs->ih_rb_wptr_addr_lo, lower_32_bits(ih->wptr_addr));
+		WREG32(ih_regs->ih_rb_wptr_addr_hi, upper_32_bits(ih->wptr_addr) & 0xFFFF);
+	}
+
+	/* set rptr, wptr to 0 */
+	WREG32(ih_regs->ih_rb_wptr, 0);
+	WREG32(ih_regs->ih_rb_rptr, 0);
+
+	WREG32(ih_regs->ih_doorbell_rptr, ih_v6_1_doorbell_rptr(ih));
+
+	return 0;
+}
+
+/**
+ * ih_v6_1_irq_init - init and enable the interrupt ring
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Allocate a ring buffer for the interrupt controller,
+ * enable the RLC, disable interrupts, enable the IH
+ * ring buffer and enable it.
+ * Called at device load and reume.
+ * Returns 0 for success, errors for failure.
+ */
+static int ih_v6_1_irq_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_ih_ring *ih[] = {&adev->irq.ih, &adev->irq.ih1};
+	u32 ih_chicken;
+	u32 tmp;
+	int ret;
+	int i;
+
+	/* disable irqs */
+	ret = ih_v6_1_toggle_interrupts(adev, false);
+	if (ret)
+		return ret;
+
+	adev->nbio.funcs->ih_control(adev);
+
+	if (unlikely((adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) ||
+		     (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO))) {
+		if (ih[0]->use_bus_addr) {
+			ih_chicken = RREG32_SOC15(OSSSYS, 0, regIH_CHICKEN);
+			ih_chicken = REG_SET_FIELD(ih_chicken,
+					IH_CHICKEN, MC_SPACE_GPA_ENABLE, 1);
+			WREG32_SOC15(OSSSYS, 0, regIH_CHICKEN, ih_chicken);
+		}
+	}
+
+	for (i = 0; i < ARRAY_SIZE(ih); i++) {
+		if (ih[i]->ring_size) {
+			ret = ih_v6_1_enable_ring(adev, ih[i]);
+			if (ret)
+				return ret;
+		}
+	}
+
+	/* update doorbell range for ih ring 0 */
+	adev->nbio.funcs->ih_doorbell_range(adev, ih[0]->use_doorbell,
+					    ih[0]->doorbell_index);
+
+	tmp = RREG32_SOC15(OSSSYS, 0, regIH_STORM_CLIENT_LIST_CNTL);
+	tmp = REG_SET_FIELD(tmp, IH_STORM_CLIENT_LIST_CNTL,
+			    CLIENT18_IS_STORM_CLIENT, 1);
+	WREG32_SOC15(OSSSYS, 0, regIH_STORM_CLIENT_LIST_CNTL, tmp);
+
+	tmp = RREG32_SOC15(OSSSYS, 0, regIH_INT_FLOOD_CNTL);
+	tmp = REG_SET_FIELD(tmp, IH_INT_FLOOD_CNTL, FLOOD_CNTL_ENABLE, 1);
+	WREG32_SOC15(OSSSYS, 0, regIH_INT_FLOOD_CNTL, tmp);
+
+	/* GC/MMHUB UTCL2 page fault interrupts are configured as
+	 * MSI storm capable interrupts by deafult. The delay is
+	 * used to avoid ISR being called too frequently
+	 * when page fault happens on several continuous page
+	 * and thus avoid MSI storm */
+	tmp = RREG32_SOC15(OSSSYS, 0, regIH_MSI_STORM_CTRL);
+	tmp = REG_SET_FIELD(tmp, IH_MSI_STORM_CTRL,
+			    DELAY, 3);
+	WREG32_SOC15(OSSSYS, 0, regIH_MSI_STORM_CTRL, tmp);
+
+	/* Redirect the interrupts to IH RB1 for dGPU */
+	if (adev->irq.ih1.ring_size) {
+		tmp = RREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_INDEX);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_INDEX, INDEX, 0);
+		WREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_INDEX, tmp);
+
+		tmp = RREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_DATA);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA, CLIENT_ID, 0xa);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA, SOURCE_ID, 0x0);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA,
+				    SOURCE_ID_MATCH_ENABLE, 0x1);
+
+		WREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_DATA, tmp);
+	}
+
+	pci_set_master(adev->pdev);
+
+	/* enable interrupts */
+	ret = ih_v6_1_toggle_interrupts(adev, true);
+	if (ret)
+		return ret;
+	/* enable wptr force update for self int */
+	force_update_wptr_for_self_int(adev, 0, 8, true);
+
+	if (adev->irq.ih_soft.ring_size)
+		adev->irq.ih_soft.enabled = true;
+
+	return 0;
+}
+
+/**
+ * ih_v6_1_irq_disable - disable interrupts
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Disable interrupts on the hw.
+ */
+static void ih_v6_1_irq_disable(struct amdgpu_device *adev)
+{
+	force_update_wptr_for_self_int(adev, 0, 8, false);
+	ih_v6_1_toggle_interrupts(adev, false);
+
+	/* Wait and acknowledge irq */
+	mdelay(1);
+}
+
+/**
+ * ih_v6_1_get_wptr - get the IH ring buffer wptr
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: amdgpu_ih_ring pointer
+ *
+ * Get the IH ring buffer wptr from either the register
+ * or the writeback memory buffer.  Also check for
+ * ring buffer overflow and deal with it.
+ * Returns the value of the wptr.
+ */
+static u32 ih_v6_1_get_wptr(struct amdgpu_device *adev,
+			      struct amdgpu_ih_ring *ih)
+{
+	u32 wptr, tmp;
+	struct amdgpu_ih_regs *ih_regs;
+
+	wptr = le32_to_cpu(*ih->wptr_cpu);
+	ih_regs = &ih->ih_regs;
+
+	if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
+		goto out;
+
+	wptr = RREG32_NO_KIQ(ih_regs->ih_rb_wptr);
+	if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
+		goto out;
+	wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0);
+
+	/* When a ring buffer overflow happen start parsing interrupt
+	 * from the last not overwritten vector (wptr + 32). Hopefully
+	 * this should allow us to catch up.
+	 */
+	tmp = (wptr + 32) & ih->ptr_mask;
+	dev_warn(adev->dev, "IH ring buffer overflow "
+		 "(0x%08X, 0x%08X, 0x%08X)\n",
+		 wptr, ih->rptr, tmp);
+	ih->rptr = tmp;
+
+	tmp = RREG32_NO_KIQ(ih_regs->ih_rb_cntl);
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
+	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+
+	/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+	 * can be detected.
+	 */
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+
+out:
+	return (wptr & ih->ptr_mask);
+}
+
+/**
+ * ih_v6_1_irq_rearm - rearm IRQ if lost
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: amdgpu_ih_ring pointer
+ *
+ */
+static void ih_v6_1_irq_rearm(struct amdgpu_device *adev,
+			       struct amdgpu_ih_ring *ih)
+{
+	uint32_t v = 0;
+	uint32_t i = 0;
+	struct amdgpu_ih_regs *ih_regs;
+
+	ih_regs = &ih->ih_regs;
+
+	/* Rearm IRQ / re-write doorbell if doorbell write is lost */
+	for (i = 0; i < MAX_REARM_RETRY; i++) {
+		v = RREG32_NO_KIQ(ih_regs->ih_rb_rptr);
+		if ((v < ih->ring_size) && (v != ih->rptr))
+			WDOORBELL32(ih->doorbell_index, ih->rptr);
+		else
+			break;
+	}
+}
+
+/**
+ * ih_v6_1_set_rptr - set the IH ring buffer rptr
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: amdgpu_ih_ring pointer
+ *
+ * Set the IH ring buffer rptr.
+ */
+static void ih_v6_1_set_rptr(struct amdgpu_device *adev,
+			       struct amdgpu_ih_ring *ih)
+{
+	struct amdgpu_ih_regs *ih_regs;
+
+	if (ih->use_doorbell) {
+		/* XXX check if swapping is necessary on BE */
+		*ih->rptr_cpu = ih->rptr;
+		WDOORBELL32(ih->doorbell_index, ih->rptr);
+
+		if (amdgpu_sriov_vf(adev))
+			ih_v6_1_irq_rearm(adev, ih);
+	} else {
+		ih_regs = &ih->ih_regs;
+		WREG32(ih_regs->ih_rb_rptr, ih->rptr);
+	}
+}
+
+/**
+ * ih_v6_1_self_irq - dispatch work for ring 1
+ *
+ * @adev: amdgpu_device pointer
+ * @source: irq source
+ * @entry: IV with WPTR update
+ *
+ * Update the WPTR from the IV and schedule work to handle the entries.
+ */
+static int ih_v6_1_self_irq(struct amdgpu_device *adev,
+			      struct amdgpu_irq_src *source,
+			      struct amdgpu_iv_entry *entry)
+{
+	uint32_t wptr = cpu_to_le32(entry->src_data[0]);
+
+	switch (entry->ring_id) {
+	case 1:
+		*adev->irq.ih1.wptr_cpu = wptr;
+		schedule_work(&adev->irq.ih1_work);
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
+static const struct amdgpu_irq_src_funcs ih_v6_1_self_irq_funcs = {
+	.process = ih_v6_1_self_irq,
+};
+
+static void ih_v6_1_set_self_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->irq.self_irq.num_types = 0;
+	adev->irq.self_irq.funcs = &ih_v6_1_self_irq_funcs;
+}
+
+static int ih_v6_1_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	int ret;
+
+	ret = amdgpu_irq_add_domain(adev);
+	if (ret) {
+		return ret;
+	}
+
+	ih_v6_1_set_interrupt_funcs(adev);
+	ih_v6_1_set_self_irq_funcs(adev);
+	return 0;
+}
+
+static int ih_v6_1_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+	bool use_bus_addr;
+
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_IH, 0,
+			      &adev->irq.self_irq);
+
+	if (r)
+		return r;
+
+	/* use gpu virtual address for ih ring
+	 * until ih_checken is programmed to allow
+	 * use bus address for ih ring by psp bl */
+	use_bus_addr = adev->firmware.load_type != AMDGPU_FW_LOAD_PSP;
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 256 * 1024, use_bus_addr);
+	if (r)
+		return r;
+
+	adev->irq.ih.use_doorbell = true;
+	adev->irq.ih.doorbell_index = adev->doorbell_index.ih << 1;
+
+	if (!(adev->flags & AMD_IS_APU)) {
+		r = amdgpu_ih_ring_init(adev, &adev->irq.ih1, IH_RING_SIZE,
+					use_bus_addr);
+		if (r)
+			return r;
+
+		adev->irq.ih1.use_doorbell = true;
+		adev->irq.ih1.doorbell_index = (adev->doorbell_index.ih + 1) << 1;
+	}
+
+	/* initialize ih control register offset */
+	ih_v6_1_init_register_offset(adev);
+
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, PAGE_SIZE, true);
+	if (r)
+		return r;
+
+	r = amdgpu_irq_init(adev);
+
+	return r;
+}
+
+static int ih_v6_1_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	amdgpu_irq_fini_sw(adev);
+
+	return 0;
+}
+
+static int ih_v6_1_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	r = ih_v6_1_irq_init(adev);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static int ih_v6_1_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	ih_v6_1_irq_disable(ip_block->adev);
+
+	return 0;
+}
+
+static int ih_v6_1_suspend(struct amdgpu_ip_block *ip_block)
+{
+	return ih_v6_1_hw_fini(ip_block);
+}
+
+static int ih_v6_1_resume(struct amdgpu_ip_block *ip_block)
+{
+	return ih_v6_1_hw_init(ip_block);
+}
+
+static bool ih_v6_1_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	/* todo */
+	return true;
+}
+
+static int ih_v6_1_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	/* todo */
+	return -ETIMEDOUT;
+}
+
+static int ih_v6_1_soft_reset(struct amdgpu_ip_block *ip_block)
+{
+	/* todo */
+	return 0;
+}
+
+static void ih_v6_1_update_clockgating_state(struct amdgpu_device *adev,
+					       bool enable)
+{
+	uint32_t data, def, field_val;
+
+	if (adev->cg_flags & AMD_CG_SUPPORT_IH_CG) {
+		def = data = RREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL);
+		field_val = enable ? 0 : 1;
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     DBUS_MUX_CLK_SOFT_OVERRIDE, field_val);
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     OSSSYS_SHARE_CLK_SOFT_OVERRIDE, field_val);
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     LIMIT_SMN_CLK_SOFT_OVERRIDE, field_val);
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     DYN_CLK_SOFT_OVERRIDE, field_val);
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     REG_CLK_SOFT_OVERRIDE, field_val);
+		if (def != data)
+			WREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL, data);
+	}
+
+	return;
+}
+
+static int ih_v6_1_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					   enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	ih_v6_1_update_clockgating_state(adev,
+				state == AMD_CG_STATE_GATE);
+	return 0;
+}
+
+static void ih_v6_1_update_ih_mem_power_gating(struct amdgpu_device *adev,
+					       bool enable)
+{
+	uint32_t ih_mem_pwr_cntl;
+
+	/* Disable ih sram power cntl before switch powergating mode */
+	ih_mem_pwr_cntl = RREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL);
+	ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+					IH_BUFFER_MEM_POWER_CTRL_EN, 0);
+	WREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL, ih_mem_pwr_cntl);
+
+	/* It is recommended to set mem powergating mode to DS mode */
+	if (enable) {
+		/* mem power mode */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_LS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_DS_EN, 1);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_SD_EN, 0);
+		/* cam mem power mode */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_LS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_DS_EN, 1);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_SD_EN, 0);
+		/* re-enable power cntl */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_CTRL_EN, 1);
+	} else {
+		/* mem power mode */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_LS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_DS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_SD_EN, 0);
+		/* cam mem power mode */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_LS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_DS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_SD_EN, 0);
+		/* re-enable power cntl*/
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_CTRL_EN, 1);
+	}
+
+	WREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL, ih_mem_pwr_cntl);
+}
+
+static int ih_v6_1_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					 enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = (state == AMD_PG_STATE_GATE);
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_IH_SRAM_PG)
+		ih_v6_1_update_ih_mem_power_gating(adev, enable);
+
+	return 0;
+}
+
+static void ih_v6_1_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (!RREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL))
+		*flags |= AMD_CG_SUPPORT_IH_CG;
+
+	return;
+}
+
+static const struct amd_ip_funcs ih_v6_1_ip_funcs = {
+	.name = "ih_v6_1",
+	.early_init = ih_v6_1_early_init,
+	.sw_init = ih_v6_1_sw_init,
+	.sw_fini = ih_v6_1_sw_fini,
+	.hw_init = ih_v6_1_hw_init,
+	.hw_fini = ih_v6_1_hw_fini,
+	.suspend = ih_v6_1_suspend,
+	.resume = ih_v6_1_resume,
+	.is_idle = ih_v6_1_is_idle,
+	.wait_for_idle = ih_v6_1_wait_for_idle,
+	.soft_reset = ih_v6_1_soft_reset,
+	.set_clockgating_state = ih_v6_1_set_clockgating_state,
+	.set_powergating_state = ih_v6_1_set_powergating_state,
+	.get_clockgating_state = ih_v6_1_get_clockgating_state,
+};
+
+static const struct amdgpu_ih_funcs ih_v6_1_funcs = {
+	.get_wptr = ih_v6_1_get_wptr,
+	.decode_iv = amdgpu_ih_decode_iv_helper,
+	.decode_iv_ts = amdgpu_ih_decode_iv_ts_helper,
+	.set_rptr = ih_v6_1_set_rptr
+};
+
+static void ih_v6_1_set_interrupt_funcs(struct amdgpu_device *adev)
+{
+	adev->irq.ih_funcs = &ih_v6_1_funcs;
+}
+
+const struct amdgpu_ip_block_version ih_v6_1_ip_block = {
+	.type = AMD_IP_BLOCK_TYPE_IH,
+	.major = 6,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &ih_v6_1_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/ih_v6_1.h b/drivers/gpu/drm/amd/amdgpu/ih_v6_1.h
new file mode 100644
index 000000000000..2232bc5cbd09
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/ih_v6_1.h
@@ -0,0 +1,28 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __IH_V6_1_IH_H__
+#define __IH_V6_1_IH_H__
+
+extern const struct amdgpu_ip_block_version ih_v6_1_ip_block;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/ih_v7_0.c b/drivers/gpu/drm/amd/amdgpu/ih_v7_0.c
new file mode 100644
index 000000000000..b32ea4129c61
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/ih_v7_0.c
@@ -0,0 +1,787 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/pci.h>
+
+#include "amdgpu.h"
+#include "amdgpu_ih.h"
+
+#include "oss/osssys_7_0_0_offset.h"
+#include "oss/osssys_7_0_0_sh_mask.h"
+
+#include "soc15_common.h"
+#include "ih_v7_0.h"
+
+#define MAX_REARM_RETRY 10
+
+static void ih_v7_0_set_interrupt_funcs(struct amdgpu_device *adev);
+
+/**
+ * ih_v7_0_init_register_offset - Initialize register offset for ih rings
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Initialize register offset ih rings (IH_V7_0).
+ */
+static void ih_v7_0_init_register_offset(struct amdgpu_device *adev)
+{
+	struct amdgpu_ih_regs *ih_regs;
+
+	/* ih ring 2 is removed
+	 * ih ring and ih ring 1 are available */
+	if (adev->irq.ih.ring_size) {
+		ih_regs = &adev->irq.ih.ih_regs;
+		ih_regs->ih_rb_base = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE);
+		ih_regs->ih_rb_base_hi = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE_HI);
+		ih_regs->ih_rb_cntl = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_CNTL);
+		ih_regs->ih_rb_wptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR);
+		ih_regs->ih_rb_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_RPTR);
+		ih_regs->ih_doorbell_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_DOORBELL_RPTR);
+		ih_regs->ih_rb_wptr_addr_lo = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR_ADDR_LO);
+		ih_regs->ih_rb_wptr_addr_hi = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR_ADDR_HI);
+		ih_regs->psp_reg_id = PSP_REG_IH_RB_CNTL;
+	}
+
+	if (adev->irq.ih1.ring_size) {
+		ih_regs = &adev->irq.ih1.ih_regs;
+		ih_regs->ih_rb_base = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE_RING1);
+		ih_regs->ih_rb_base_hi = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_BASE_HI_RING1);
+		ih_regs->ih_rb_cntl = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_CNTL_RING1);
+		ih_regs->ih_rb_wptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_WPTR_RING1);
+		ih_regs->ih_rb_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_RB_RPTR_RING1);
+		ih_regs->ih_doorbell_rptr = SOC15_REG_OFFSET(OSSSYS, 0, regIH_DOORBELL_RPTR_RING1);
+		ih_regs->psp_reg_id = PSP_REG_IH_RB_CNTL_RING1;
+	}
+}
+
+/**
+ * force_update_wptr_for_self_int - Force update the wptr for self interrupt
+ *
+ * @adev: amdgpu_device pointer
+ * @threshold: threshold to trigger the wptr reporting
+ * @timeout: timeout to trigger the wptr reporting
+ * @enabled: Enable/disable timeout flush mechanism
+ *
+ * threshold input range: 0 ~ 15, default 0,
+ * real_threshold = 2^threshold
+ * timeout input range: 0 ~ 20, default 8,
+ * real_timeout = (2^timeout) * 1024 / (socclk_freq)
+ *
+ * Force update wptr for self interrupt ( >= SIENNA_CICHLID).
+ */
+static void
+force_update_wptr_for_self_int(struct amdgpu_device *adev,
+			       u32 threshold, u32 timeout, bool enabled)
+{
+	u32 ih_cntl, ih_rb_cntl;
+
+	ih_cntl = RREG32_SOC15(OSSSYS, 0, regIH_CNTL2);
+	ih_rb_cntl = RREG32_SOC15(OSSSYS, 0, regIH_RB_CNTL_RING1);
+
+	ih_cntl = REG_SET_FIELD(ih_cntl, IH_CNTL2,
+				SELF_IV_FORCE_WPTR_UPDATE_TIMEOUT, timeout);
+	ih_cntl = REG_SET_FIELD(ih_cntl, IH_CNTL2,
+				SELF_IV_FORCE_WPTR_UPDATE_ENABLE, enabled);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL_RING1,
+				   RB_USED_INT_THRESHOLD, threshold);
+
+	if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+		if (psp_reg_program(&adev->psp, PSP_REG_IH_RB_CNTL_RING1, ih_rb_cntl))
+			return;
+	} else {
+		WREG32_SOC15(OSSSYS, 0, regIH_RB_CNTL_RING1, ih_rb_cntl);
+	}
+
+	WREG32_SOC15(OSSSYS, 0, regIH_CNTL2, ih_cntl);
+}
+
+/**
+ * ih_v7_0_toggle_ring_interrupts - toggle the interrupt ring buffer
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: amdgpu_ih_ring pointet
+ * @enable: true - enable the interrupts, false - disable the interrupts
+ *
+ * Toggle the interrupt ring buffer (IH_V7_0)
+ */
+static int ih_v7_0_toggle_ring_interrupts(struct amdgpu_device *adev,
+					  struct amdgpu_ih_ring *ih,
+					  bool enable)
+{
+	struct amdgpu_ih_regs *ih_regs;
+	uint32_t tmp;
+
+	ih_regs = &ih->ih_regs;
+
+	tmp = RREG32(ih_regs->ih_rb_cntl);
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_ENABLE, (enable ? 1 : 0));
+	/* enable_intr field is only valid in ring0 */
+	if (ih == &adev->irq.ih)
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, ENABLE_INTR, (enable ? 1 : 0));
+
+	if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+		if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp))
+			return -ETIMEDOUT;
+	} else {
+		WREG32(ih_regs->ih_rb_cntl, tmp);
+	}
+
+	if (enable) {
+		ih->enabled = true;
+	} else {
+		/* set rptr, wptr to 0 */
+		WREG32(ih_regs->ih_rb_rptr, 0);
+		WREG32(ih_regs->ih_rb_wptr, 0);
+		ih->enabled = false;
+		ih->rptr = 0;
+	}
+
+	return 0;
+}
+
+/**
+ * ih_v7_0_toggle_interrupts - Toggle all the available interrupt ring buffers
+ *
+ * @adev: amdgpu_device pointer
+ * @enable: enable or disable interrupt ring buffers
+ *
+ * Toggle all the available interrupt ring buffers (IH_V7_0).
+ */
+static int ih_v7_0_toggle_interrupts(struct amdgpu_device *adev, bool enable)
+{
+	struct amdgpu_ih_ring *ih[] = {&adev->irq.ih, &adev->irq.ih1};
+	int i;
+	int r;
+
+	for (i = 0; i < ARRAY_SIZE(ih); i++) {
+		if (ih[i]->ring_size) {
+			r = ih_v7_0_toggle_ring_interrupts(adev, ih[i], enable);
+			if (r)
+				return r;
+		}
+	}
+
+	return 0;
+}
+
+static uint32_t ih_v7_0_rb_cntl(struct amdgpu_ih_ring *ih, uint32_t ih_rb_cntl)
+{
+	int rb_bufsz = order_base_2(ih->ring_size / 4);
+
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+				   MC_SPACE, ih->use_bus_addr ? 2 : 4);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+				   WPTR_OVERFLOW_CLEAR, 1);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+				   WPTR_OVERFLOW_ENABLE, 1);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, RB_SIZE, rb_bufsz);
+	/* Ring Buffer write pointer writeback. If enabled, IH_RB_WPTR register
+	 * value is written to memory
+	 */
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL,
+				   WPTR_WRITEBACK_ENABLE, 1);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_SNOOP, 1);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_RO, 0);
+	ih_rb_cntl = REG_SET_FIELD(ih_rb_cntl, IH_RB_CNTL, MC_VMID, 0);
+
+	return ih_rb_cntl;
+}
+
+static uint32_t ih_v7_0_doorbell_rptr(struct amdgpu_ih_ring *ih)
+{
+	u32 ih_doorbell_rtpr = 0;
+
+	if (ih->use_doorbell) {
+		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+						 IH_DOORBELL_RPTR, OFFSET,
+						 ih->doorbell_index);
+		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+						 IH_DOORBELL_RPTR,
+						 ENABLE, 1);
+	} else {
+		ih_doorbell_rtpr = REG_SET_FIELD(ih_doorbell_rtpr,
+						 IH_DOORBELL_RPTR,
+						 ENABLE, 0);
+	}
+	return ih_doorbell_rtpr;
+}
+
+/**
+ * ih_v7_0_enable_ring - enable an ih ring buffer
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: amdgpu_ih_ring pointer
+ *
+ * Enable an ih ring buffer (IH_V7_0)
+ */
+static int ih_v7_0_enable_ring(struct amdgpu_device *adev,
+				      struct amdgpu_ih_ring *ih)
+{
+	struct amdgpu_ih_regs *ih_regs;
+	uint32_t tmp;
+
+	ih_regs = &ih->ih_regs;
+
+	/* Ring Buffer base. [39:8] of 40-bit address of the beginning of the ring buffer*/
+	WREG32(ih_regs->ih_rb_base, ih->gpu_addr >> 8);
+	WREG32(ih_regs->ih_rb_base_hi, (ih->gpu_addr >> 40) & 0xff);
+
+	tmp = RREG32(ih_regs->ih_rb_cntl);
+	tmp = ih_v7_0_rb_cntl(ih, tmp);
+	if (ih == &adev->irq.ih)
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RPTR_REARM, !!adev->irq.msi_enabled);
+	if (ih == &adev->irq.ih1) {
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_ENABLE, 0);
+		tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, RB_FULL_DRAIN_ENABLE, 1);
+	}
+
+	if (amdgpu_sriov_vf(adev) && amdgpu_sriov_reg_indirect_ih(adev)) {
+		if (psp_reg_program(&adev->psp, ih_regs->psp_reg_id, tmp)) {
+			DRM_ERROR("PSP program IH_RB_CNTL failed!\n");
+			return -ETIMEDOUT;
+		}
+	} else {
+		WREG32(ih_regs->ih_rb_cntl, tmp);
+	}
+
+	if (ih == &adev->irq.ih) {
+		/* set the ih ring 0 writeback address whether it's enabled or not */
+		WREG32(ih_regs->ih_rb_wptr_addr_lo, lower_32_bits(ih->wptr_addr));
+		WREG32(ih_regs->ih_rb_wptr_addr_hi, upper_32_bits(ih->wptr_addr) & 0xFFFF);
+	}
+
+	/* set rptr, wptr to 0 */
+	WREG32(ih_regs->ih_rb_wptr, 0);
+	WREG32(ih_regs->ih_rb_rptr, 0);
+
+	WREG32(ih_regs->ih_doorbell_rptr, ih_v7_0_doorbell_rptr(ih));
+
+	return 0;
+}
+
+/**
+ * ih_v7_0_irq_init - init and enable the interrupt ring
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Allocate a ring buffer for the interrupt controller,
+ * enable the RLC, disable interrupts, enable the IH
+ * ring buffer and enable it.
+ * Called at device load and reume.
+ * Returns 0 for success, errors for failure.
+ */
+static int ih_v7_0_irq_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_ih_ring *ih[] = {&adev->irq.ih, &adev->irq.ih1};
+	u32 ih_chicken;
+	u32 tmp;
+	int ret;
+	int i;
+
+	/* disable irqs */
+	ret = ih_v7_0_toggle_interrupts(adev, false);
+	if (ret)
+		return ret;
+
+	adev->nbio.funcs->ih_control(adev);
+
+	if (unlikely((adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) ||
+		     (adev->firmware.load_type == AMDGPU_FW_LOAD_RLC_BACKDOOR_AUTO))) {
+		if (ih[0]->use_bus_addr) {
+			ih_chicken = RREG32_SOC15(OSSSYS, 0, regIH_CHICKEN);
+			ih_chicken = REG_SET_FIELD(ih_chicken,
+					IH_CHICKEN, MC_SPACE_GPA_ENABLE, 1);
+			WREG32_SOC15(OSSSYS, 0, regIH_CHICKEN, ih_chicken);
+		}
+	}
+
+	for (i = 0; i < ARRAY_SIZE(ih); i++) {
+		if (ih[i]->ring_size) {
+			ret = ih_v7_0_enable_ring(adev, ih[i]);
+			if (ret)
+				return ret;
+		}
+	}
+
+	/* update doorbell range for ih ring 0 */
+	adev->nbio.funcs->ih_doorbell_range(adev, ih[0]->use_doorbell,
+					    ih[0]->doorbell_index);
+
+	tmp = RREG32_SOC15(OSSSYS, 0, regIH_STORM_CLIENT_LIST_CNTL);
+	tmp = REG_SET_FIELD(tmp, IH_STORM_CLIENT_LIST_CNTL,
+			    CLIENT18_IS_STORM_CLIENT, 1);
+	WREG32_SOC15(OSSSYS, 0, regIH_STORM_CLIENT_LIST_CNTL, tmp);
+
+	tmp = RREG32_SOC15(OSSSYS, 0, regIH_INT_FLOOD_CNTL);
+	tmp = REG_SET_FIELD(tmp, IH_INT_FLOOD_CNTL, FLOOD_CNTL_ENABLE, 1);
+	WREG32_SOC15(OSSSYS, 0, regIH_INT_FLOOD_CNTL, tmp);
+
+	/* GC/MMHUB UTCL2 page fault interrupts are configured as
+	 * MSI storm capable interrupts by deafult. The delay is
+	 * used to avoid ISR being called too frequently
+	 * when page fault happens on several continuous page
+	 * and thus avoid MSI storm */
+	tmp = RREG32_SOC15(OSSSYS, 0, regIH_MSI_STORM_CTRL);
+	tmp = REG_SET_FIELD(tmp, IH_MSI_STORM_CTRL,
+			    DELAY, 3);
+	WREG32_SOC15(OSSSYS, 0, regIH_MSI_STORM_CTRL, tmp);
+
+	/* Redirect the interrupts to IH RB1 for dGPU */
+	if (adev->irq.ih1.ring_size) {
+		tmp = RREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_INDEX);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_INDEX, INDEX, 0);
+		WREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_INDEX, tmp);
+
+		tmp = RREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_DATA);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA, CLIENT_ID, 0xa);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA, SOURCE_ID, 0x0);
+		tmp = REG_SET_FIELD(tmp, IH_RING1_CLIENT_CFG_DATA,
+				    SOURCE_ID_MATCH_ENABLE, 0x1);
+
+		WREG32_SOC15(OSSSYS, 0, regIH_RING1_CLIENT_CFG_DATA, tmp);
+	}
+
+	pci_set_master(adev->pdev);
+
+	/* enable interrupts */
+	ret = ih_v7_0_toggle_interrupts(adev, true);
+	if (ret)
+		return ret;
+	/* enable wptr force update for self int */
+	force_update_wptr_for_self_int(adev, 0, 8, true);
+
+	if (adev->irq.ih_soft.ring_size)
+		adev->irq.ih_soft.enabled = true;
+
+	return 0;
+}
+
+/**
+ * ih_v7_0_irq_disable - disable interrupts
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Disable interrupts on the hw.
+ */
+static void ih_v7_0_irq_disable(struct amdgpu_device *adev)
+{
+	force_update_wptr_for_self_int(adev, 0, 8, false);
+	ih_v7_0_toggle_interrupts(adev, false);
+
+	/* Wait and acknowledge irq */
+	mdelay(1);
+}
+
+/**
+ * ih_v7_0_get_wptr() - get the IH ring buffer wptr
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: IH ring buffer to fetch wptr
+ *
+ * Get the IH ring buffer wptr from either the register
+ * or the writeback memory buffer.  Also check for
+ * ring buffer overflow and deal with it.
+ * Returns the value of the wptr.
+ */
+static u32 ih_v7_0_get_wptr(struct amdgpu_device *adev,
+			      struct amdgpu_ih_ring *ih)
+{
+	u32 wptr, tmp;
+	struct amdgpu_ih_regs *ih_regs;
+
+	wptr = le32_to_cpu(*ih->wptr_cpu);
+	ih_regs = &ih->ih_regs;
+
+	if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
+		goto out;
+
+	wptr = RREG32_NO_KIQ(ih_regs->ih_rb_wptr);
+	if (!REG_GET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW))
+		goto out;
+	wptr = REG_SET_FIELD(wptr, IH_RB_WPTR, RB_OVERFLOW, 0);
+
+	/* When a ring buffer overflow happen start parsing interrupt
+	 * from the last not overwritten vector (wptr + 32). Hopefully
+	 * this should allow us to catch up.
+	 */
+	tmp = (wptr + 32) & ih->ptr_mask;
+	dev_warn(adev->dev, "IH ring buffer overflow "
+		 "(0x%08X, 0x%08X, 0x%08X)\n",
+		 wptr, ih->rptr, tmp);
+	ih->rptr = tmp;
+
+	tmp = RREG32_NO_KIQ(ih_regs->ih_rb_cntl);
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1);
+	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+
+	/* Unset the CLEAR_OVERFLOW bit immediately so new overflows
+	 * can be detected.
+	 */
+	tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 0);
+	WREG32_NO_KIQ(ih_regs->ih_rb_cntl, tmp);
+out:
+	return (wptr & ih->ptr_mask);
+}
+
+/**
+ * ih_v7_0_irq_rearm - rearm IRQ if lost
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: IH ring to match
+ *
+ */
+static void ih_v7_0_irq_rearm(struct amdgpu_device *adev,
+			       struct amdgpu_ih_ring *ih)
+{
+	uint32_t v = 0;
+	uint32_t i = 0;
+	struct amdgpu_ih_regs *ih_regs;
+
+	ih_regs = &ih->ih_regs;
+
+	/* Rearm IRQ / re-write doorbell if doorbell write is lost */
+	for (i = 0; i < MAX_REARM_RETRY; i++) {
+		v = RREG32_NO_KIQ(ih_regs->ih_rb_rptr);
+		if ((v < ih->ring_size) && (v != ih->rptr))
+			WDOORBELL32(ih->doorbell_index, ih->rptr);
+		else
+			break;
+	}
+}
+
+/**
+ * ih_v7_0_set_rptr - set the IH ring buffer rptr
+ *
+ * @adev: amdgpu_device pointer
+ * @ih: IH ring buffer to set rptr
+ */
+static void ih_v7_0_set_rptr(struct amdgpu_device *adev,
+			       struct amdgpu_ih_ring *ih)
+{
+	struct amdgpu_ih_regs *ih_regs;
+
+	if (ih->use_doorbell) {
+		/* XXX check if swapping is necessary on BE */
+		*ih->rptr_cpu = ih->rptr;
+		WDOORBELL32(ih->doorbell_index, ih->rptr);
+
+		if (amdgpu_sriov_vf(adev))
+			ih_v7_0_irq_rearm(adev, ih);
+	} else {
+		ih_regs = &ih->ih_regs;
+		WREG32(ih_regs->ih_rb_rptr, ih->rptr);
+	}
+}
+
+/**
+ * ih_v7_0_self_irq - dispatch work for ring 1
+ *
+ * @adev: amdgpu_device pointer
+ * @source: irq source
+ * @entry: IV with WPTR update
+ *
+ * Update the WPTR from the IV and schedule work to handle the entries.
+ */
+static int ih_v7_0_self_irq(struct amdgpu_device *adev,
+			      struct amdgpu_irq_src *source,
+			      struct amdgpu_iv_entry *entry)
+{
+	uint32_t wptr = cpu_to_le32(entry->src_data[0]);
+
+	switch (entry->ring_id) {
+	case 1:
+		*adev->irq.ih1.wptr_cpu = wptr;
+		schedule_work(&adev->irq.ih1_work);
+		break;
+	default: break;
+	}
+	return 0;
+}
+
+static const struct amdgpu_irq_src_funcs ih_v7_0_self_irq_funcs = {
+	.process = ih_v7_0_self_irq,
+};
+
+static void ih_v7_0_set_self_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->irq.self_irq.num_types = 0;
+	adev->irq.self_irq.funcs = &ih_v7_0_self_irq_funcs;
+}
+
+static int ih_v7_0_early_init(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	ih_v7_0_set_interrupt_funcs(adev);
+	ih_v7_0_set_self_irq_funcs(adev);
+	return 0;
+}
+
+static int ih_v7_0_sw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+	bool use_bus_addr;
+
+	r = amdgpu_irq_add_id(adev, SOC21_IH_CLIENTID_IH, 0,
+			      &adev->irq.self_irq);
+
+	if (r)
+		return r;
+
+	/* use gpu virtual address for ih ring
+	 * until ih_checken is programmed to allow
+	 * use bus address for ih ring by psp bl */
+	use_bus_addr = adev->firmware.load_type != AMDGPU_FW_LOAD_PSP;
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 256 * 1024, use_bus_addr);
+	if (r)
+		return r;
+
+	adev->irq.ih.use_doorbell = true;
+	adev->irq.ih.doorbell_index = adev->doorbell_index.ih << 1;
+
+	if (!(adev->flags & AMD_IS_APU)) {
+		r = amdgpu_ih_ring_init(adev, &adev->irq.ih1, IH_RING_SIZE,
+					use_bus_addr);
+		if (r)
+			return r;
+
+		adev->irq.ih1.use_doorbell = true;
+		adev->irq.ih1.doorbell_index = (adev->doorbell_index.ih + 1) << 1;
+	}
+
+	/* initialize ih control register offset */
+	ih_v7_0_init_register_offset(adev);
+
+	r = amdgpu_ih_ring_init(adev, &adev->irq.ih_soft, PAGE_SIZE, true);
+	if (r)
+		return r;
+
+	r = amdgpu_irq_init(adev);
+
+	return r;
+}
+
+static int ih_v7_0_sw_fini(struct amdgpu_ip_block *ip_block)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	amdgpu_irq_fini_sw(adev);
+
+	return 0;
+}
+
+static int ih_v7_0_hw_init(struct amdgpu_ip_block *ip_block)
+{
+	int r;
+	struct amdgpu_device *adev = ip_block->adev;
+
+	r = ih_v7_0_irq_init(adev);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static int ih_v7_0_hw_fini(struct amdgpu_ip_block *ip_block)
+{
+	ih_v7_0_irq_disable(ip_block->adev);
+
+	return 0;
+}
+
+static int ih_v7_0_suspend(struct amdgpu_ip_block *ip_block)
+{
+	return ih_v7_0_hw_fini(ip_block);
+}
+
+static int ih_v7_0_resume(struct amdgpu_ip_block *ip_block)
+{
+	return ih_v7_0_hw_init(ip_block);
+}
+
+static bool ih_v7_0_is_idle(struct amdgpu_ip_block *ip_block)
+{
+	/* todo */
+	return true;
+}
+
+static int ih_v7_0_wait_for_idle(struct amdgpu_ip_block *ip_block)
+{
+	/* todo */
+	return -ETIMEDOUT;
+}
+
+static int ih_v7_0_soft_reset(struct amdgpu_ip_block *ip_block)
+{
+	/* todo */
+	return 0;
+}
+
+static void ih_v7_0_update_clockgating_state(struct amdgpu_device *adev,
+					       bool enable)
+{
+	uint32_t data, def, field_val;
+
+	if (adev->cg_flags & AMD_CG_SUPPORT_IH_CG) {
+		def = data = RREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL);
+		field_val = enable ? 0 : 1;
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     DBUS_MUX_CLK_SOFT_OVERRIDE, field_val);
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     OSSSYS_SHARE_CLK_SOFT_OVERRIDE, field_val);
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     LIMIT_SMN_CLK_SOFT_OVERRIDE, field_val);
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     DYN_CLK_SOFT_OVERRIDE, field_val);
+		data = REG_SET_FIELD(data, IH_CLK_CTRL,
+				     REG_CLK_SOFT_OVERRIDE, field_val);
+		if (def != data)
+			WREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL, data);
+	}
+
+	return;
+}
+
+static int ih_v7_0_set_clockgating_state(struct amdgpu_ip_block *ip_block,
+					   enum amd_clockgating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	ih_v7_0_update_clockgating_state(adev,
+				state == AMD_CG_STATE_GATE);
+	return 0;
+}
+
+static void ih_v7_0_update_ih_mem_power_gating(struct amdgpu_device *adev,
+					       bool enable)
+{
+	uint32_t ih_mem_pwr_cntl;
+
+	/* Disable ih sram power cntl before switch powergating mode */
+	ih_mem_pwr_cntl = RREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL);
+	ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+					IH_BUFFER_MEM_POWER_CTRL_EN, 0);
+	WREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL, ih_mem_pwr_cntl);
+
+	/* It is recommended to set mem powergating mode to DS mode */
+	if (enable) {
+		/* mem power mode */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_LS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_DS_EN, 1);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_SD_EN, 0);
+		/* cam mem power mode */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_LS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_DS_EN, 1);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_SD_EN, 0);
+		/* re-enable power cntl */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_CTRL_EN, 1);
+	} else {
+		/* mem power mode */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_LS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_DS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_SD_EN, 0);
+		/* cam mem power mode */
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_LS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_DS_EN, 0);
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_RETRY_INT_CAM_MEM_POWER_SD_EN, 0);
+		/* re-enable power cntl*/
+		ih_mem_pwr_cntl = REG_SET_FIELD(ih_mem_pwr_cntl, IH_MEM_POWER_CTRL,
+						IH_BUFFER_MEM_POWER_CTRL_EN, 1);
+	}
+
+	WREG32_SOC15(OSSSYS, 0, regIH_MEM_POWER_CTRL, ih_mem_pwr_cntl);
+}
+
+static int ih_v7_0_set_powergating_state(struct amdgpu_ip_block *ip_block,
+					 enum amd_powergating_state state)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+	bool enable = (state == AMD_PG_STATE_GATE);
+
+	if (adev->pg_flags & AMD_PG_SUPPORT_IH_SRAM_PG)
+		ih_v7_0_update_ih_mem_power_gating(adev, enable);
+
+	return 0;
+}
+
+static void ih_v7_0_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags)
+{
+	struct amdgpu_device *adev = ip_block->adev;
+
+	if (!RREG32_SOC15(OSSSYS, 0, regIH_CLK_CTRL))
+		*flags |= AMD_CG_SUPPORT_IH_CG;
+
+	return;
+}
+
+static const struct amd_ip_funcs ih_v7_0_ip_funcs = {
+	.name = "ih_v7_0",
+	.early_init = ih_v7_0_early_init,
+	.sw_init = ih_v7_0_sw_init,
+	.sw_fini = ih_v7_0_sw_fini,
+	.hw_init = ih_v7_0_hw_init,
+	.hw_fini = ih_v7_0_hw_fini,
+	.suspend = ih_v7_0_suspend,
+	.resume = ih_v7_0_resume,
+	.is_idle = ih_v7_0_is_idle,
+	.wait_for_idle = ih_v7_0_wait_for_idle,
+	.soft_reset = ih_v7_0_soft_reset,
+	.set_clockgating_state = ih_v7_0_set_clockgating_state,
+	.set_powergating_state = ih_v7_0_set_powergating_state,
+	.get_clockgating_state = ih_v7_0_get_clockgating_state,
+};
+
+static const struct amdgpu_ih_funcs ih_v7_0_funcs = {
+	.get_wptr = ih_v7_0_get_wptr,
+	.decode_iv = amdgpu_ih_decode_iv_helper,
+	.decode_iv_ts = amdgpu_ih_decode_iv_ts_helper,
+	.set_rptr = ih_v7_0_set_rptr
+};
+
+static void ih_v7_0_set_interrupt_funcs(struct amdgpu_device *adev)
+{
+	adev->irq.ih_funcs = &ih_v7_0_funcs;
+}
+
+const struct amdgpu_ip_block_version ih_v7_0_ip_block =
+{
+	.type = AMD_IP_BLOCK_TYPE_IH,
+	.major = 7,
+	.minor = 0,
+	.rev = 0,
+	.funcs = &ih_v7_0_ip_funcs,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/ih_v7_0.h b/drivers/gpu/drm/amd/amdgpu/ih_v7_0.h
new file mode 100644
index 000000000000..af9dcbc451fd
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/ih_v7_0.h
@@ -0,0 +1,28 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __IH_V7_0_IH_H__
+#define __IH_V7_0_IH_H__
+
+extern const struct amdgpu_ip_block_version ih_v7_0_ip_block;
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/imu_v11_0.c b/drivers/gpu/drm/amd/amdgpu/imu_v11_0.c
index 4ab90c7852c3..cc626036ed9c 100644
--- a/drivers/gpu/drm/amd/amdgpu/imu_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/imu_v11_0.c
@@ -32,14 +32,18 @@
 #include "gc/gc_11_0_0_sh_mask.h"
 
 MODULE_FIRMWARE("amdgpu/gc_11_0_0_imu.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_0_0_imu_kicker.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_1_imu.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_2_imu.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_3_imu.bin");
 MODULE_FIRMWARE("amdgpu/gc_11_0_4_imu.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_0_imu.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_1_imu.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_2_imu.bin");
+MODULE_FIRMWARE("amdgpu/gc_11_5_3_imu.bin");
 
 static int imu_v11_0_init_microcode(struct amdgpu_device *adev)
 {
-	char fw_name[40];
 	char ucode_prefix[30];
 	int err;
 	const struct imu_firmware_header_v1_0 *imu_hdr;
@@ -48,13 +52,16 @@ static int imu_v11_0_init_microcode(struct amdgpu_device *adev)
 	DRM_DEBUG("\n");
 
 	amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix));
-
-	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_imu.bin", ucode_prefix);
-	err = amdgpu_ucode_request(adev, &adev->gfx.imu_fw, fw_name);
+	if (amdgpu_is_kicker_fw(adev))
+		err = amdgpu_ucode_request(adev, &adev->gfx.imu_fw, AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_imu_kicker.bin", ucode_prefix);
+	else
+		err = amdgpu_ucode_request(adev, &adev->gfx.imu_fw, AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_imu.bin", ucode_prefix);
 	if (err)
 		goto out;
+
 	imu_hdr = (const struct imu_firmware_header_v1_0 *)adev->gfx.imu_fw->data;
-	adev->gfx.imu_fw_version = le32_to_cpu(imu_hdr->header.ucode_version);
 	//adev->gfx.imu_feature_version = le32_to_cpu(imu_hdr->ucode_feature_version);
 	
 	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
@@ -68,13 +75,14 @@ static int imu_v11_0_init_microcode(struct amdgpu_device *adev)
 		info->fw = adev->gfx.imu_fw;
 		adev->firmware.fw_size +=
 			ALIGN(le32_to_cpu(imu_hdr->imu_dram_ucode_size_bytes), PAGE_SIZE);
-	}
+	} else
+		adev->gfx.imu_fw_version = le32_to_cpu(imu_hdr->header.ucode_version);
 
 out:
 	if (err) {
 		dev_err(adev->dev,
-			"gfx11: Failed to load firmware \"%s\"\n",
-			fw_name);
+			"gfx11: Failed to load firmware \"%s_imu.bin\"\n",
+			ucode_prefix);
 		amdgpu_ucode_release(&adev->gfx.imu_fw);
 	}
 
@@ -152,7 +160,7 @@ static void imu_v11_0_setup(struct amdgpu_device *adev)
 		WREG32_SOC15(GC, 0, regGFX_IMU_C2PMSG_16, imu_reg_val);
 	}
 
-	//disble imu Rtavfs, SmsRepair, DfllBTC, and ClkB
+	//disable imu Rtavfs, SmsRepair, DfllBTC, and ClkB
 	imu_reg_val = RREG32_SOC15(GC, 0, regGFX_IMU_SCRATCH_10);
 	imu_reg_val |= 0x10007;
 	WREG32_SOC15(GC, 0, regGFX_IMU_SCRATCH_10, imu_reg_val);
@@ -352,7 +360,7 @@ static void imu_v11_0_program_rlc_ram(struct amdgpu_device *adev)
 
 	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_INDEX, 0x2);
 
-	switch (adev->ip_versions[GC_HWIP][0]) {
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
 	case IP_VERSION(11, 0, 0):
 		program_imu_rlc_ram(adev, imu_rlc_ram_golden_11,
 				(const u32)ARRAY_SIZE(imu_rlc_ram_golden_11));
diff --git a/drivers/gpu/drm/amd/amdgpu/imu_v12_0.c b/drivers/gpu/drm/amd/amdgpu/imu_v12_0.c
new file mode 100644
index 000000000000..58cd87db8061
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/imu_v12_0.c
@@ -0,0 +1,406 @@
+/*
+ * Copyright 2023 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/firmware.h>
+#include "amdgpu.h"
+#include "amdgpu_imu.h"
+#include "amdgpu_dpm.h"
+
+#include "imu_v12_0.h"
+
+#include "gc/gc_12_0_0_offset.h"
+#include "gc/gc_12_0_0_sh_mask.h"
+#include "mmhub/mmhub_4_1_0_offset.h"
+
+MODULE_FIRMWARE("amdgpu/gc_12_0_0_imu.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_imu.bin");
+MODULE_FIRMWARE("amdgpu/gc_12_0_1_imu_kicker.bin");
+
+#define TRANSFER_RAM_MASK	0x001c0000
+
+static int imu_v12_0_init_microcode(struct amdgpu_device *adev)
+{
+	char ucode_prefix[30];
+	int err;
+	const struct imu_firmware_header_v1_0 *imu_hdr;
+	struct amdgpu_firmware_info *info = NULL;
+
+	DRM_DEBUG("\n");
+
+	amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix));
+	if (amdgpu_is_kicker_fw(adev))
+		err = amdgpu_ucode_request(adev, &adev->gfx.imu_fw, AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_imu_kicker.bin", ucode_prefix);
+	else
+		err = amdgpu_ucode_request(adev, &adev->gfx.imu_fw, AMDGPU_UCODE_REQUIRED,
+					   "amdgpu/%s_imu.bin", ucode_prefix);
+	if (err)
+		goto out;
+
+	imu_hdr = (const struct imu_firmware_header_v1_0 *)adev->gfx.imu_fw->data;
+	adev->gfx.imu_fw_version = le32_to_cpu(imu_hdr->header.ucode_version);
+
+	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
+		info = &adev->firmware.ucode[AMDGPU_UCODE_ID_IMU_I];
+		info->ucode_id = AMDGPU_UCODE_ID_IMU_I;
+		info->fw = adev->gfx.imu_fw;
+		adev->firmware.fw_size +=
+			ALIGN(le32_to_cpu(imu_hdr->imu_iram_ucode_size_bytes), PAGE_SIZE);
+		info = &adev->firmware.ucode[AMDGPU_UCODE_ID_IMU_D];
+		info->ucode_id = AMDGPU_UCODE_ID_IMU_D;
+		info->fw = adev->gfx.imu_fw;
+		adev->firmware.fw_size +=
+			ALIGN(le32_to_cpu(imu_hdr->imu_dram_ucode_size_bytes), PAGE_SIZE);
+	}
+
+out:
+	if (err) {
+		dev_err(adev->dev,
+			"gfx12: Failed to load firmware \"%s_imu.bin\"\n",
+			ucode_prefix);
+		amdgpu_ucode_release(&adev->gfx.imu_fw);
+	}
+
+	return err;
+}
+
+static int imu_v12_0_load_microcode(struct amdgpu_device *adev)
+{
+	const struct imu_firmware_header_v1_0 *hdr;
+	const __le32 *fw_data;
+	unsigned i, fw_size;
+
+	if (!adev->gfx.imu_fw)
+		return -EINVAL;
+
+	hdr = (const struct imu_firmware_header_v1_0 *)adev->gfx.imu_fw->data;
+
+	fw_data = (const __le32 *)(adev->gfx.imu_fw->data +
+			le32_to_cpu(hdr->header.ucode_array_offset_bytes));
+	fw_size = le32_to_cpu(hdr->imu_iram_ucode_size_bytes) / 4;
+
+	WREG32_SOC15(GC, 0, regGFX_IMU_I_RAM_ADDR, 0);
+
+	for (i = 0; i < fw_size; i++)
+		WREG32_SOC15(GC, 0, regGFX_IMU_I_RAM_DATA, le32_to_cpup(fw_data++));
+
+	WREG32_SOC15(GC, 0, regGFX_IMU_I_RAM_ADDR, adev->gfx.imu_fw_version);
+
+	fw_data = (const __le32 *)(adev->gfx.imu_fw->data +
+			le32_to_cpu(hdr->header.ucode_array_offset_bytes) +
+			le32_to_cpu(hdr->imu_iram_ucode_size_bytes));
+	fw_size = le32_to_cpu(hdr->imu_dram_ucode_size_bytes) / 4;
+
+	WREG32_SOC15(GC, 0, regGFX_IMU_D_RAM_ADDR, 0);
+
+	for (i = 0; i < fw_size; i++)
+		WREG32_SOC15(GC, 0, regGFX_IMU_D_RAM_DATA, le32_to_cpup(fw_data++));
+
+	WREG32_SOC15(GC, 0, regGFX_IMU_D_RAM_ADDR, adev->gfx.imu_fw_version);
+
+	return 0;
+}
+
+static int imu_v12_0_wait_for_reset_status(struct amdgpu_device *adev)
+{
+	u32 imu_reg_val = 0;
+	int i;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		imu_reg_val = RREG32_SOC15(GC, 0, regGFX_IMU_GFX_RESET_CTRL);
+		if ((imu_reg_val & 0x1f) == 0x1f)
+			break;
+		udelay(1);
+	}
+
+	if (i >= adev->usec_timeout) {
+		dev_err(adev->dev, "init imu: IMU start timeout\n");
+		return -ETIMEDOUT;
+	}
+
+	return 0;
+}
+
+static void imu_v12_0_setup(struct amdgpu_device *adev)
+{
+	u32 imu_reg_val;
+
+	WREG32_SOC15(GC, 0, regGFX_IMU_C2PMSG_ACCESS_CTRL0, 0xffffff);
+	WREG32_SOC15(GC, 0, regGFX_IMU_C2PMSG_ACCESS_CTRL1, 0xffff);
+
+	if (adev->gfx.imu.mode == DEBUG_MODE) {
+		imu_reg_val = RREG32_SOC15(GC, 0, regGFX_IMU_C2PMSG_16);
+		imu_reg_val |= 0x1;
+		WREG32_SOC15(GC, 0, regGFX_IMU_C2PMSG_16, imu_reg_val);
+
+		imu_reg_val = RREG32_SOC15(GC, 0, regGFX_IMU_SCRATCH_10);
+		imu_reg_val |= 0x20010007;
+		WREG32_SOC15(GC, 0, regGFX_IMU_SCRATCH_10, imu_reg_val);
+
+	}
+}
+
+static int imu_v12_0_start(struct amdgpu_device *adev)
+{
+	u32 imu_reg_val;
+
+	imu_reg_val = RREG32_SOC15(GC, 0, regGFX_IMU_CORE_CTRL);
+	imu_reg_val &= 0xfffffffe;
+	WREG32_SOC15(GC, 0, regGFX_IMU_CORE_CTRL, imu_reg_val);
+
+	if (adev->flags & AMD_IS_APU)
+		amdgpu_dpm_set_gfx_power_up_by_imu(adev);
+
+	return imu_v12_0_wait_for_reset_status(adev);
+}
+
+static const struct imu_rlc_ram_golden imu_rlc_ram_golden_12_0_1[] = {
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCH_PIPE_STEER, 0x1e4, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL1X_PIPE_STEER, 0x1e4, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL1_PIPE_STEER, 0x1e4, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2_PIPE_STEER_0, 0x13571357, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2_PIPE_STEER_1, 0x64206420, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2_PIPE_STEER_2, 0x2460246, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2_PIPE_STEER_3, 0x75317531, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGL2C_CTRL3, 0xc0d41183, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regSDMA0_CHICKEN_BITS, 0x507d1c0, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regSDMA1_CHICKEN_BITS, 0x507d1c0, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCP_RB_WPTR_POLL_CNTL, 0x600100, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_CREDITS, 0x3f7fff, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_CREDITS, 0x3f7ebf, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_TAG_RESERVE0, 0x2e00000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_TAG_RESERVE1, 0x1a078, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_TAG_RESERVE2, 0x0, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_TAG_RESERVE0, 0x0, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_TAG_RESERVE1, 0x12030, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_TAG_RESERVE2, 0x0, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_VCC_RESERVE0, 0x19041000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_VCC_RESERVE1, 0x80000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_VCC_RESERVE0, 0x1e080000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_VCC_RESERVE1, 0x80000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_PRIORITY, 0x880, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_PRIORITY, 0x8880, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_ARB_FINAL, 0x17, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_ARB_FINAL, 0x77, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_SDP_ENABLE, 0x00000001, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_SDP_ENABLE, 0x00000001, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL2, 0x20000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_APT_CNTL, 0x0c, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_CACHEABLE_DRAM_ADDRESS_END, 0xfffff, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_CPWD_MISC, 0x0091, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGC_EA_SE_MISC, 0x0091, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGRBM_GFX_INDEX, 0xe0000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCR_GENERAL_CNTL, 0x00008500, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regPA_CL_ENHANCE, 0x00880007, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regTD_CNTL, 0x00000001, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGRBM_GFX_INDEX, 0x00000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regRMI_GENERAL_CNTL, 0x01e00000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGRBM_GFX_INDEX, 0x00000001, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regRMI_GENERAL_CNTL, 0x01e00000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGRBM_GFX_INDEX, 0x00000100, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regRMI_GENERAL_CNTL, 0x01e00000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGRBM_GFX_INDEX, 0x00000101, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regRMI_GENERAL_CNTL, 0x01e00000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGRBM_GFX_INDEX, 0xe0000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGB_ADDR_CONFIG, 0x08200545, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGRBMH_CP_PERFMON_CNTL, 0x00000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCB_PERFCOUNTER0_SELECT1, 0x000fffff, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCP_DEBUG_2, 0x00020000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regCP_CPC_DEBUG, 0x00500010, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, 0x00000500, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x00000001, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0x00000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_LOCAL_FB_ADDRESS_START, 0x00000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_LOCAL_FB_ADDRESS_END, 0x0000000f, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_FB_LOCATION_BASE, 0x00006000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_FB_LOCATION_TOP, 0x0000600f, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_CONTEXT0_CNTL, 0x00000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_CONTEXT1_CNTL, 0x00000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_NB_TOP_OF_DRAM_SLOT1, 0xff800000, 0xe0000000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_NB_LOWER_TOP_OF_DRAM2, 0x00000001, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_NB_UPPER_TOP_OF_DRAM2, 0x0000ffff, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_AGP_BASE, 0x00000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_AGP_BOT, 0x00000002, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_AGP_TOP, 0x00000000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_PROTECTION_FAULT_CNTL, 0x00001ffc, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL, 0x00000551, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL, 0x00080603, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL2, 0x00000003, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL3, 0x00100003, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCVM_L2_CNTL5, 0x00003fe0, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_LOW_ADDR, 0x0003d000, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_HIGH_ADDR, 0x0003d7ff, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB, 0, 0x1c0000),
+	IMU_RLC_RAM_GOLDEN_VALUE(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB, 0, 0x1c0000)
+};
+
+static void program_imu_rlc_ram_old(struct amdgpu_device *adev,
+				    const struct imu_rlc_ram_golden *regs,
+				    const u32 array_size)
+{
+	const struct imu_rlc_ram_golden *entry;
+	u32 reg, data;
+	int i;
+
+	for (i = 0; i < array_size; ++i) {
+		entry = &regs[i];
+		reg =  adev->reg_offset[entry->hwip][entry->instance][entry->segment] + entry->reg;
+		reg |= entry->addr_mask;
+		data = entry->data;
+		if (entry->reg == regGCMC_VM_AGP_BASE)
+			data = 0x00ffffff;
+		else if (entry->reg == regGCMC_VM_AGP_TOP)
+			data = 0x0;
+		else if (entry->reg == regGCMC_VM_FB_LOCATION_BASE)
+			data = adev->gmc.vram_start >> 24;
+		else if (entry->reg == regGCMC_VM_FB_LOCATION_TOP)
+			data = adev->gmc.vram_end >> 24;
+
+		WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_HIGH, 0);
+		WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_LOW, reg);
+		WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_DATA, data);
+	}
+}
+
+static u32 imu_v12_0_grbm_gfx_index_remap(struct amdgpu_device *adev,
+					  u32 data, bool high)
+{
+	u32 val, inst_index;
+
+	inst_index = REG_GET_FIELD(data, GRBM_GFX_INDEX, INSTANCE_INDEX);
+
+	if (high)
+		val = inst_index >> 5;
+	else
+		val = REG_GET_FIELD(data, GRBM_GFX_INDEX, SE_BROADCAST_WRITES) << 18 |
+		      REG_GET_FIELD(data, GRBM_GFX_INDEX, SA_BROADCAST_WRITES) << 19 |
+		      REG_GET_FIELD(data, GRBM_GFX_INDEX, INSTANCE_BROADCAST_WRITES) << 20 |
+		      REG_GET_FIELD(data, GRBM_GFX_INDEX, SE_INDEX) << 21 |
+		      REG_GET_FIELD(data, GRBM_GFX_INDEX, SA_INDEX) << 25 |
+		      (inst_index & 0x1f);
+
+	return val;
+}
+
+static u32 imu_v12_init_gfxhub_settings(struct amdgpu_device *adev,
+					u32 reg, u32 data)
+{
+	if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_FB_LOCATION_BASE))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_FB_LOCATION_BASE);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_FB_LOCATION_TOP))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_FB_LOCATION_TOP);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_FB_OFFSET))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_FB_OFFSET);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_AGP_BASE))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_BASE);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_AGP_BOT))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_BOT);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_AGP_TOP))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_AGP_TOP);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_MX_L1_TLB_CNTL))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_MX_L1_TLB_CNTL);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_SYSTEM_APERTURE_LOW_ADDR))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_LOW_ADDR);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_SYSTEM_APERTURE_HIGH_ADDR))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_HIGH_ADDR);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_LOCAL_FB_ADDRESS_START))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_LOCAL_FB_ADDRESS_START);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_LOCAL_FB_ADDRESS_END))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_LOCAL_FB_ADDRESS_END);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_LOCAL_SYSMEM_ADDRESS_START))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_LOCAL_SYSMEM_ADDRESS_START);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_LOCAL_SYSMEM_ADDRESS_END))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_LOCAL_SYSMEM_ADDRESS_END);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_LSB);
+	else if (reg == SOC15_REG_OFFSET(GC, 0, regGCMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB))
+		return RREG32_SOC15(MMHUB, 0, regMMMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR_MSB);
+	else
+		return data;
+}
+
+static void program_imu_rlc_ram(struct amdgpu_device *adev,
+				const u32 *regs,
+				const u32 array_size)
+{
+	u32 reg, data, val_h = 0, val_l = TRANSFER_RAM_MASK;
+	int i;
+
+	if (array_size % 3)
+		return;
+
+	for (i = 0; i < array_size; i += 3) {
+		reg = regs[i + 0];
+		data = regs[i + 2];
+		data = imu_v12_init_gfxhub_settings(adev, reg, data);
+		if (reg == SOC15_REG_OFFSET(GC, 0, regGRBM_GFX_INDEX)) {
+			val_l = imu_v12_0_grbm_gfx_index_remap(adev, data, false);
+			val_h = imu_v12_0_grbm_gfx_index_remap(adev, data, true);
+		} else {
+			WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_HIGH, val_h);
+			WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_LOW, reg | val_l);
+			WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_DATA, data);
+		}
+	}
+}
+
+static void imu_v12_0_program_rlc_ram(struct amdgpu_device *adev)
+{
+	u32 reg_data, size = 0;
+	const u32 *data = NULL;
+	int r = -EINVAL;
+
+	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_INDEX, 0x2);
+
+	switch (amdgpu_ip_version(adev, GC_HWIP, 0)) {
+	case IP_VERSION(12, 0, 0):
+	case IP_VERSION(12, 0, 1):
+		if (!r)
+			program_imu_rlc_ram(adev, data, (const u32)size);
+		else
+			program_imu_rlc_ram_old(adev, imu_rlc_ram_golden_12_0_1,
+				(const u32)ARRAY_SIZE(imu_rlc_ram_golden_12_0_1));
+		break;
+	default:
+		BUG();
+		break;
+	}
+
+	//Indicate the latest entry
+	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_HIGH, 0);
+	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_ADDR_LOW, 0);
+	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_DATA, 0);
+
+	reg_data = RREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_INDEX);
+	reg_data |= GFX_IMU_RLC_RAM_INDEX__RAM_VALID_MASK;
+	WREG32_SOC15(GC, 0, regGFX_IMU_RLC_RAM_INDEX, reg_data);
+}
+
+const struct amdgpu_imu_funcs gfx_v12_0_imu_funcs = {
+	.init_microcode = imu_v12_0_init_microcode,
+	.load_microcode = imu_v12_0_load_microcode,
+	.setup_imu = imu_v12_0_setup,
+	.start_imu = imu_v12_0_start,
+	.program_rlc_ram = imu_v12_0_program_rlc_ram,
+	.wait_for_reset_status = imu_v12_0_wait_for_reset_status,
+};
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v10_1.h b/drivers/gpu/drm/amd/amdgpu/imu_v12_0.h
index 9afd6ddb01e9..a1f50cb1aeab 100644
--- a/drivers/gpu/drm/amd/amdgpu/mes_v10_1.h
+++ b/drivers/gpu/drm/amd/amdgpu/imu_v12_0.h
@@ -1,5 +1,5 @@
 /*
- * Copyright 2019 Advanced Micro Devices, Inc.
+ * Copyright 2021 Advanced Micro Devices, Inc.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -21,9 +21,10 @@
  *
  */
 
-#ifndef __MES_V10_1_H__
-#define __MES_V10_1_H__
+#ifndef __IMU_V12_0_H__
+#define __IMU_V12_0_H__
 
-extern const struct amdgpu_ip_block_version mes_v10_1_ip_block;
+extern const struct amdgpu_imu_funcs gfx_v12_0_imu_funcs;
 
 #endif
+
diff --git a/drivers/gpu/drm/amd/amdgpu/isp_v4_1_0.c b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_0.c
new file mode 100644
index 000000000000..0027a639c7e6
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_0.c
@@ -0,0 +1,191 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright (C) 2024 Advanced Micro Devices, Inc. All rights reserved.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ */
+
+#include "amdgpu.h"
+#include "isp_v4_1_0.h"
+
+static const unsigned int isp_4_1_0_int_srcid[MAX_ISP410_INT_SRC] = {
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT9,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT10,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT11,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT12,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT13,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT14,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT15,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT16
+};
+
+static int isp_v4_1_0_hw_init(struct amdgpu_isp *isp)
+{
+	struct amdgpu_device *adev = isp->adev;
+	int idx, int_idx, num_res, r;
+	u64 isp_base;
+
+	if (adev->rmmio_size == 0 || adev->rmmio_size < 0x5289)
+		return -EINVAL;
+
+	isp_base = adev->rmmio_base;
+
+	isp->isp_cell = kcalloc(3, sizeof(struct mfd_cell), GFP_KERNEL);
+	if (!isp->isp_cell) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev,
+			"%s: isp mfd cell alloc failed\n", __func__);
+		goto failure;
+	}
+
+	num_res = MAX_ISP410_MEM_RES + MAX_ISP410_INT_SRC;
+	isp->isp_res = kcalloc(num_res, sizeof(struct resource),
+			       GFP_KERNEL);
+	if (!isp->isp_res) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev,
+			"%s: isp mfd res alloc failed\n", __func__);
+		goto failure;
+	}
+
+	isp->isp_pdata = kzalloc(sizeof(*isp->isp_pdata), GFP_KERNEL);
+	if (!isp->isp_pdata) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev,
+			"%s: isp platform data alloc failed\n", __func__);
+		goto failure;
+	}
+
+	/* initialize isp platform data */
+	isp->isp_pdata->adev = (void *)adev;
+	isp->isp_pdata->asic_type = adev->asic_type;
+	isp->isp_pdata->base_rmmio_size = adev->rmmio_size;
+
+	isp->isp_res[0].name = "isp_4_1_0_reg";
+	isp->isp_res[0].flags = IORESOURCE_MEM;
+	isp->isp_res[0].start = isp_base;
+	isp->isp_res[0].end = isp_base + ISP_REGS_OFFSET_END;
+
+	isp->isp_res[1].name = "isp_4_1_phy0_reg";
+	isp->isp_res[1].flags = IORESOURCE_MEM;
+	isp->isp_res[1].start = isp_base + ISP410_PHY0_OFFSET;
+	isp->isp_res[1].end = isp_base + ISP410_PHY0_OFFSET + ISP410_PHY0_SIZE;
+
+	for (idx = MAX_ISP410_MEM_RES, int_idx = 0; idx < num_res; idx++, int_idx++) {
+		isp->isp_res[idx].name = "isp_4_1_0_irq";
+		isp->isp_res[idx].flags = IORESOURCE_IRQ;
+		isp->isp_res[idx].start =
+			amdgpu_irq_create_mapping(adev, isp_4_1_0_int_srcid[int_idx]);
+		isp->isp_res[idx].end =
+			isp->isp_res[idx].start;
+	}
+
+	isp->isp_cell[0].name = "amd_isp_capture";
+	isp->isp_cell[0].num_resources = num_res;
+	isp->isp_cell[0].resources = &isp->isp_res[0];
+	isp->isp_cell[0].platform_data = isp->isp_pdata;
+	isp->isp_cell[0].pdata_size = sizeof(struct isp_platform_data);
+
+	/* initialize isp i2c platform data */
+	isp->isp_i2c_res = kcalloc(1, sizeof(struct resource), GFP_KERNEL);
+	if (!isp->isp_i2c_res) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev,
+			"%s: isp mfd res alloc failed\n", __func__);
+		goto failure;
+	}
+
+	isp->isp_i2c_res[0].name = "isp_i2c0_reg";
+	isp->isp_i2c_res[0].flags = IORESOURCE_MEM;
+	isp->isp_i2c_res[0].start = isp_base + ISP410_I2C0_OFFSET;
+	isp->isp_i2c_res[0].end = isp_base + ISP410_I2C0_OFFSET + ISP410_I2C0_SIZE;
+
+	isp->isp_cell[1].name = "amd_isp_i2c_designware";
+	isp->isp_cell[1].num_resources = 1;
+	isp->isp_cell[1].resources = &isp->isp_i2c_res[0];
+	isp->isp_cell[1].platform_data = isp->isp_pdata;
+	isp->isp_cell[1].pdata_size = sizeof(struct isp_platform_data);
+
+	/* initialize isp gpiochip platform data */
+	isp->isp_gpio_res = kcalloc(1, sizeof(struct resource), GFP_KERNEL);
+	if (!isp->isp_gpio_res) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev,
+			"%s: isp gpio res alloc failed\n", __func__);
+		goto failure;
+	}
+
+	isp->isp_gpio_res[0].name = "isp_gpio_reg";
+	isp->isp_gpio_res[0].flags = IORESOURCE_MEM;
+	isp->isp_gpio_res[0].start = isp_base + ISP410_GPIO_SENSOR_OFFSET;
+	isp->isp_gpio_res[0].end = isp_base + ISP410_GPIO_SENSOR_OFFSET +
+				   ISP410_GPIO_SENSOR_SIZE;
+
+	isp->isp_cell[2].name = "amdisp-pinctrl";
+	isp->isp_cell[2].num_resources = 1;
+	isp->isp_cell[2].resources = &isp->isp_gpio_res[0];
+	isp->isp_cell[2].platform_data = isp->isp_pdata;
+	isp->isp_cell[2].pdata_size = sizeof(struct isp_platform_data);
+
+	r = mfd_add_hotplug_devices(isp->parent, isp->isp_cell, 3);
+	if (r) {
+		drm_err(&adev->ddev,
+			"%s: add mfd hotplug device failed\n", __func__);
+		goto failure;
+	}
+
+	return 0;
+
+failure:
+
+	kfree(isp->isp_pdata);
+	kfree(isp->isp_res);
+	kfree(isp->isp_cell);
+	kfree(isp->isp_i2c_res);
+	kfree(isp->isp_gpio_res);
+
+	return r;
+}
+
+static int isp_v4_1_0_hw_fini(struct amdgpu_isp *isp)
+{
+	mfd_remove_devices(isp->parent);
+
+	kfree(isp->isp_res);
+	kfree(isp->isp_cell);
+	kfree(isp->isp_pdata);
+	kfree(isp->isp_i2c_res);
+	kfree(isp->isp_gpio_res);
+
+	return 0;
+}
+
+static const struct isp_funcs isp_v4_1_0_funcs = {
+	.hw_init = isp_v4_1_0_hw_init,
+	.hw_fini = isp_v4_1_0_hw_fini,
+};
+
+void isp_v4_1_0_set_isp_funcs(struct amdgpu_isp *isp)
+{
+	isp->funcs = &isp_v4_1_0_funcs;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/isp_v4_1_0.h b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_0.h
new file mode 100644
index 000000000000..4d239198edd0
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_0.h
@@ -0,0 +1,50 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright (C) 2024 Advanced Micro Devices, Inc. All rights reserved.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ */
+
+#ifndef __ISP_V4_1_0_H__
+#define __ISP_V4_1_0_H__
+
+#include "amdgpu_isp.h"
+
+#include "ivsrcid/isp/irqsrcs_isp_4_1.h"
+
+#define MAX_ISP410_MEM_RES 2
+#define MAX_ISP410_SENSOR_RES 1
+#define MAX_ISP410_INT_SRC 8
+
+#define ISP410_PHY0_OFFSET 0x66700
+#define ISP410_PHY0_SIZE   0xD30
+
+#define ISP410_I2C0_OFFSET 0x66400
+#define ISP410_I2C0_SIZE 0x100
+
+#define ISP410_GPIO_SENSOR_OFFSET 0x6613C
+#define ISP410_GPIO_SENSOR_SIZE 0x54
+
+void isp_v4_1_0_set_isp_funcs(struct amdgpu_isp *isp);
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c
new file mode 100644
index 000000000000..4258d3e0b706
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c
@@ -0,0 +1,377 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright (C) 2024 Advanced Micro Devices, Inc. All rights reserved.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ */
+
+#include <linux/gpio/machine.h>
+#include "amdgpu.h"
+#include "isp_v4_1_1.h"
+
+MODULE_FIRMWARE("amdgpu/isp_4_1_1.bin");
+
+#define ISP_PERFORMANCE_STATE_LOW 0
+#define ISP_PERFORMANCE_STATE_HIGH 1
+
+#define ISP_HIGH_PERFORMANC_XCLK 788
+#define ISP_HIGH_PERFORMANC_ICLK 788
+
+static const unsigned int isp_4_1_1_int_srcid[MAX_ISP411_INT_SRC] = {
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT9,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT10,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT11,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT12,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT13,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT14,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT15,
+	ISP_4_1__SRCID__ISP_RINGBUFFER_WPT16
+};
+
+static struct gpiod_lookup_table isp_gpio_table = {
+	.dev_id = "amd_isp_capture",
+	.table = {
+		GPIO_LOOKUP("AMDI0030:00", 85, "enable_isp", GPIO_ACTIVE_HIGH),
+		{ }
+	},
+};
+
+static struct gpiod_lookup_table isp_sensor_gpio_table = {
+	.dev_id = "i2c-ov05c10",
+	.table = {
+		GPIO_LOOKUP("amdisp-pinctrl", 0, "enable", GPIO_ACTIVE_HIGH),
+		{ }
+	},
+};
+
+static int isp_poweroff(struct generic_pm_domain *genpd)
+{
+	struct amdgpu_isp *isp = container_of(genpd, struct amdgpu_isp, ispgpd);
+	struct amdgpu_device *adev = isp->adev;
+
+	return amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ISP, true, 0);
+}
+
+static int isp_poweron(struct generic_pm_domain *genpd)
+{
+	struct amdgpu_isp *isp = container_of(genpd, struct amdgpu_isp, ispgpd);
+	struct amdgpu_device *adev = isp->adev;
+
+	return amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ISP, false, 0);
+}
+
+static int isp_set_performance_state(struct generic_pm_domain *genpd,
+				     unsigned int state)
+{
+	struct amdgpu_isp *isp = container_of(genpd, struct amdgpu_isp, ispgpd);
+	struct amdgpu_device *adev = isp->adev;
+	u32 iclk, xclk;
+	int ret;
+
+	switch (state) {
+	case ISP_PERFORMANCE_STATE_HIGH:
+		xclk = ISP_HIGH_PERFORMANC_XCLK;
+		iclk = ISP_HIGH_PERFORMANC_ICLK;
+		break;
+	case ISP_PERFORMANCE_STATE_LOW:
+		/* isp runs at default lowest clock-rate on power-on, do nothing */
+		return 0;
+	default:
+		return -EINVAL;
+	}
+
+	ret = amdgpu_dpm_set_soft_freq_range(adev, PP_ISPXCLK, xclk, 0);
+	if (ret) {
+		drm_err(&adev->ddev, "failed to set xclk %u to %u: %d\n",
+			xclk, state, ret);
+		return ret;
+	}
+
+	ret = amdgpu_dpm_set_soft_freq_range(adev, PP_ISPICLK, iclk, 0);
+	if (ret) {
+		drm_err(&adev->ddev, "failed to set iclk %u to %u: %d\n",
+			iclk, state, ret);
+		return ret;
+	}
+
+	return 0;
+}
+
+static int isp_genpd_add_device(struct device *dev, void *data)
+{
+	struct generic_pm_domain *gpd = data;
+	struct platform_device *pdev = container_of(dev, struct platform_device, dev);
+	struct amdgpu_isp *isp = container_of(gpd, struct amdgpu_isp, ispgpd);
+	struct amdgpu_device *adev = isp->adev;
+	int ret;
+
+	if (!pdev)
+		return -EINVAL;
+
+	if (!dev->type->name) {
+		drm_dbg(&adev->ddev, "Invalid device type to add\n");
+		goto exit;
+	}
+
+	if (strcmp(dev->type->name, "mfd_device")) {
+		drm_dbg(&adev->ddev, "Invalid isp mfd device %s to add\n", pdev->mfd_cell->name);
+		goto exit;
+	}
+
+	ret = pm_genpd_add_device(gpd, dev);
+	if (ret) {
+		drm_err(&adev->ddev, "Failed to add dev %s to genpd %d\n",
+			pdev->mfd_cell->name, ret);
+		return -ENODEV;
+	}
+
+exit:
+	/* Continue to add */
+	return 0;
+}
+
+static int isp_genpd_remove_device(struct device *dev, void *data)
+{
+	struct generic_pm_domain *gpd = data;
+	struct platform_device *pdev = container_of(dev, struct platform_device, dev);
+	struct amdgpu_isp *isp = container_of(gpd, struct amdgpu_isp, ispgpd);
+	struct amdgpu_device *adev = isp->adev;
+	int ret;
+
+	if (!pdev)
+		return -EINVAL;
+
+	if (!dev->type->name) {
+		drm_dbg(&adev->ddev, "Invalid device type to remove\n");
+		goto exit;
+	}
+
+	if (strcmp(dev->type->name, "mfd_device")) {
+		drm_dbg(&adev->ddev, "Invalid isp mfd device %s to remove\n",
+			pdev->mfd_cell->name);
+		goto exit;
+	}
+
+	ret = pm_genpd_remove_device(dev);
+	if (ret) {
+		drm_err(&adev->ddev, "Failed to remove dev from genpd %d\n", ret);
+		return -ENODEV;
+	}
+
+exit:
+	/* Continue to remove */
+	return 0;
+}
+
+static int isp_v4_1_1_hw_init(struct amdgpu_isp *isp)
+{
+	const struct software_node *amd_camera_node, *isp4_node;
+	struct amdgpu_device *adev = isp->adev;
+	struct acpi_device *acpi_dev;
+	int idx, int_idx, num_res, r;
+	u64 isp_base;
+
+	if (adev->rmmio_size == 0 || adev->rmmio_size < 0x5289)
+		return -EINVAL;
+
+	r = amdgpu_acpi_get_isp4_dev(&acpi_dev);
+	if (r) {
+		drm_dbg(&adev->ddev, "Invalid isp platform detected (%d)", r);
+		/* allow GPU init to progress */
+		return 0;
+	}
+
+	/* add GPIO resources required for OMNI5C10 sensor */
+	if (!strcmp("OMNI5C10", acpi_device_hid(acpi_dev))) {
+		gpiod_add_lookup_table(&isp_gpio_table);
+		gpiod_add_lookup_table(&isp_sensor_gpio_table);
+	}
+
+	isp_base = adev->rmmio_base;
+
+	isp->ispgpd.name = "ISP_v_4_1_1";
+	isp->ispgpd.power_off = isp_poweroff;
+	isp->ispgpd.power_on = isp_poweron;
+	isp->ispgpd.set_performance_state = isp_set_performance_state;
+
+	r = pm_genpd_init(&isp->ispgpd, NULL, true);
+	if (r) {
+		drm_err(&adev->ddev, "failed to initialize genpd (%d)\n", r);
+		return -EINVAL;
+	}
+
+	isp->isp_cell = kcalloc(3, sizeof(struct mfd_cell), GFP_KERNEL);
+	if (!isp->isp_cell) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev, "isp mfd cell alloc failed (%d)\n", r);
+		goto failure;
+	}
+
+	num_res = MAX_ISP411_MEM_RES + MAX_ISP411_INT_SRC;
+
+	isp->isp_res = kcalloc(num_res, sizeof(struct resource),
+			       GFP_KERNEL);
+	if (!isp->isp_res) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev, "isp mfd resource alloc failed (%d)\n", r);
+		goto failure;
+	}
+
+	isp->isp_pdata = kzalloc(sizeof(*isp->isp_pdata), GFP_KERNEL);
+	if (!isp->isp_pdata) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev, "isp platform data alloc failed (%d)\n", r);
+		goto failure;
+	}
+
+	amd_camera_node = (const struct software_node *)acpi_dev->driver_data;
+	isp4_node = software_node_find_by_name(amd_camera_node, "isp4");
+
+	/* initialize isp platform data */
+	isp->isp_pdata->adev = (void *)adev;
+	isp->isp_pdata->asic_type = adev->asic_type;
+	isp->isp_pdata->base_rmmio_size = adev->rmmio_size;
+
+	isp->isp_res[0].name = "isp_4_1_1_reg";
+	isp->isp_res[0].flags = IORESOURCE_MEM;
+	isp->isp_res[0].start = isp_base;
+	isp->isp_res[0].end = isp_base + ISP_REGS_OFFSET_END;
+
+	isp->isp_res[1].name = "isp_4_1_1_phy0_reg";
+	isp->isp_res[1].flags = IORESOURCE_MEM;
+	isp->isp_res[1].start = isp_base + ISP411_PHY0_OFFSET;
+	isp->isp_res[1].end = isp_base + ISP411_PHY0_OFFSET + ISP411_PHY0_SIZE;
+
+	for (idx = MAX_ISP411_MEM_RES, int_idx = 0; idx < num_res; idx++, int_idx++) {
+		isp->isp_res[idx].name = "isp_4_1_1_irq";
+		isp->isp_res[idx].flags = IORESOURCE_IRQ;
+		isp->isp_res[idx].start =
+			amdgpu_irq_create_mapping(adev, isp_4_1_1_int_srcid[int_idx]);
+		isp->isp_res[idx].end =
+			isp->isp_res[idx].start;
+	}
+
+	isp->isp_cell[0].name = "amd_isp_capture";
+	isp->isp_cell[0].num_resources = num_res;
+	isp->isp_cell[0].resources = &isp->isp_res[0];
+	isp->isp_cell[0].platform_data = isp->isp_pdata;
+	isp->isp_cell[0].swnode = isp4_node;
+	isp->isp_cell[0].pdata_size = sizeof(struct isp_platform_data);
+
+	/* initialize isp i2c platform data */
+	isp->isp_i2c_res = kcalloc(1, sizeof(struct resource), GFP_KERNEL);
+	if (!isp->isp_i2c_res) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev, "isp mfd res alloc failed (%d)\n", r);
+		goto failure;
+	}
+
+	isp->isp_i2c_res[0].name = "isp_i2c0_reg";
+	isp->isp_i2c_res[0].flags = IORESOURCE_MEM;
+	isp->isp_i2c_res[0].start = isp_base + ISP411_I2C0_OFFSET;
+	isp->isp_i2c_res[0].end = isp_base + ISP411_I2C0_OFFSET + ISP411_I2C0_SIZE;
+
+	isp->isp_cell[1].name = "amd_isp_i2c_designware";
+	isp->isp_cell[1].num_resources = 1;
+	isp->isp_cell[1].resources = &isp->isp_i2c_res[0];
+	isp->isp_cell[1].platform_data = isp->isp_pdata;
+	isp->isp_cell[1].pdata_size = sizeof(struct isp_platform_data);
+
+	/* initialize isp gpiochip platform data */
+	isp->isp_gpio_res = kcalloc(1, sizeof(struct resource), GFP_KERNEL);
+	if (!isp->isp_gpio_res) {
+		r = -ENOMEM;
+		drm_err(&adev->ddev, "isp gpio resource alloc failed (%d)\n", r);
+		goto failure;
+	}
+
+	isp->isp_gpio_res[0].name = "isp_gpio_reg";
+	isp->isp_gpio_res[0].flags = IORESOURCE_MEM;
+	isp->isp_gpio_res[0].start = isp_base + ISP411_GPIO_SENSOR_OFFSET;
+	isp->isp_gpio_res[0].end = isp_base + ISP411_GPIO_SENSOR_OFFSET +
+				   ISP411_GPIO_SENSOR_SIZE;
+
+	isp->isp_cell[2].name = "amdisp-pinctrl";
+	isp->isp_cell[2].num_resources = 1;
+	isp->isp_cell[2].resources = &isp->isp_gpio_res[0];
+	isp->isp_cell[2].platform_data = isp->isp_pdata;
+	isp->isp_cell[2].pdata_size = sizeof(struct isp_platform_data);
+
+	/* add only amd_isp_capture and amd_isp_i2c_designware to genpd */
+	r = mfd_add_hotplug_devices(isp->parent, isp->isp_cell, 2);
+	if (r) {
+		drm_err(&adev->ddev, "add mfd hotplug device failed (%d)\n", r);
+		goto failure;
+	}
+
+	r = device_for_each_child(isp->parent, &isp->ispgpd,
+				  isp_genpd_add_device);
+	if (r) {
+		drm_err(&adev->ddev, "failed to add devices to genpd (%d)\n", r);
+		goto failure;
+	}
+
+	r = mfd_add_hotplug_devices(isp->parent, &isp->isp_cell[2], 1);
+	if (r) {
+		drm_err(&adev->ddev, "add pinctl hotplug device failed (%d)\n", r);
+		goto failure;
+	}
+
+	return 0;
+
+failure:
+
+	kfree(isp->isp_pdata);
+	kfree(isp->isp_res);
+	kfree(isp->isp_cell);
+	kfree(isp->isp_i2c_res);
+	kfree(isp->isp_gpio_res);
+
+	return r;
+}
+
+static int isp_v4_1_1_hw_fini(struct amdgpu_isp *isp)
+{
+	device_for_each_child(isp->parent, NULL,
+			      isp_genpd_remove_device);
+
+	mfd_remove_devices(isp->parent);
+
+	kfree(isp->isp_res);
+	kfree(isp->isp_cell);
+	kfree(isp->isp_pdata);
+	kfree(isp->isp_i2c_res);
+	kfree(isp->isp_gpio_res);
+
+	return 0;
+}
+
+static const struct isp_funcs isp_v4_1_1_funcs = {
+	.hw_init = isp_v4_1_1_hw_init,
+	.hw_fini = isp_v4_1_1_hw_fini,
+};
+
+void isp_v4_1_1_set_isp_funcs(struct amdgpu_isp *isp)
+{
+	isp->funcs = &isp_v4_1_1_funcs;
+}
diff --git a/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.h b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.h
new file mode 100644
index 000000000000..fe45d70d87f1
--- /dev/null
+++ b/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.h
@@ -0,0 +1,49 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright (C) 2024 Advanced Micro Devices, Inc. All rights reserved.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ */
+
+#ifndef __ISP_V4_1_1_H__
+#define __ISP_V4_1_1_H__
+
+#include "amdgpu_isp.h"
+
+#include "ivsrcid/isp/irqsrcs_isp_4_1.h"
+
+#define MAX_ISP411_MEM_RES 2
+#define MAX_ISP411_INT_SRC 8
+
+#define ISP411_PHY0_OFFSET 0x66700
+#define ISP411_PHY0_SIZE   0xD30
+
+#define ISP411_I2C0_OFFSET 0x66400
+#define ISP411_I2C0_SIZE 0x100
+
+#define ISP411_GPIO_SENSOR_OFFSET 0x6613C
+#define ISP411_GPIO_SENSOR_SIZE 0x54
+
+void isp_v4_1_1_set_isp_funcs(struct amdgpu_isp *isp);
+
+#endif
diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c
index 9360204da7fb..b5bb7f4d607c 100644
--- a/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c
author	Michal Marek <mmarek@suse.cz>	2015-01-09 14:08:26 +0100
committer	Michal Marek <mmarek@suse.cz>	2015-04-07 21:27:45 +0200
commit	7e114bbf51fbb015dc25d8123e090afcce5b5048 (patch)
tree	4a859c5f6756e2893a00c7f283529e7abc99875f /tools/perf/scripts/python
parent	c517d838eb7d07bbe9507871fab3931deccff539 (diff)