Merge tag 'drm-next-2024-11-21' of https://gitlab.freedesktop.org/drm/kernel

Pull drm updates from Dave Airlie:
 "There's a lot of rework, the panic helper support is being added to
  more drivers, v3d gets support for HW superpages, scheduler
  documentation, drm client and video aperture reworks, some new
  MAINTAINERS added, amdgpu has the usual lots of IP refactors, Intel
  has some Pantherlake enablement and xe is getting some SRIOV bits, but
  just lots of stuff everywhere.

  core:
   - split DSC helpers from DP helpers
   - clang build fixes for drm/mm test
   - drop simple pipeline support for gem vram
   - document submission error signaling
   - move drm_rect to drm core module from kms helper
   - add default client setup to most drivers
   - move to video aperture helpers instead of drm ones

  tests:
   - new framebuffer tests

  ttm:
   - remove swapped and pinned BOs from TTM lru

  panic:
   - fix uninit spinlock
   - add ABGR2101010 support

  bridge:
   - add TI TDP158 support
   - use standard PM OPS

  dma-fence:
   - use read_trylock instead of read_lock to help lockdep

  scheduler:
   - add errno to sched start to report different errors
   - add locking to drm_sched_entity_modify_sched
   - improve documentation

  xe:
   - add drm_line_printer
   - lots of refactoring
   - Enable Xe2 + PES disaggregation
   - add new ARL PCI ID
   - SRIOV development work
   - fix exec unnecessary implicit fence
   - define and parse OA sync props
   - forcewake refactoring

  i915:
   - Enable BMG/LNL ultra joiner
   - Enable 10bpx + CCS scanout on ICL+, fp16/CCS on TGL+
   - use DSB for plane/color mgmt
   - Arrow lake PCI IDs
   - lots of i915/xe display refactoring
   - enable PXP GuC autoteardown
   - Pantherlake (PTL) Xe3 LPD display enablement
   - Allow fastset HDR infoframe changes
   - write DP source OUI for non-eDP sinks
   - share PCI IDs between i915 and xe

  amdgpu:
   - SDMA queue reset support
   - SMU 13.0.6, JPEG 4.0.3 updates
   - Initial runtime repartitioning support
   - rework IP structs for multiple IP instances
   - Fetch EDID from _DDC if available
   - SMU13 zero rpm user control
   - lots of fixes/cleanups

  amdkfd:
   - Increase event FIFO size
   - add topology cap flag for per queue reset

  msm:
   - DPU:
      - SA8775P support
      - (disabled by default) MSM8917, MSM8937, MSM8953 and MSM8996 support
      - Enable large framebuffer support
      - Drop MSM8998 and SDM845
   - DP:
      - SA8775P support
   - GPU:
      - a7xx preemption support
      - Adreno A663 support

  ast:
   - warn about unsupported TX chips

  ivpu:
   - add coredump
   - add pantherlake support

  rockchip:
   - 4K@60Hz display enablement
   - generate pll programming tables

  panthor:
   - add timestamp query API
   - add realtime group priority
   - add fdinfo support

  etnaviv:
   - improve handling of DMA address limits
   - improve GPU hangcheck

  exynos:
   - Decon Exynos7870 support

  mediatek:
   - add OF graph support

  omap:
   - locking fixes

  bochs:
   - convert to gem/shmem from simpledrm

  v3d:
   - support big/super pages
   - add gemfs

  vc4:
   - BCM2712 support refactoring
   - add YUV444 format support

  udmabuf:
   - folio related fixes

  nouveau:
   - add panic support on nv50+"

* tag 'drm-next-2024-11-21' of https://gitlab.freedesktop.org/drm/kernel: (1583 commits)
  drm/xe/guc: Fix dereference before NULL check
  drm/amd: Fix initialization mistake for NBIO 7.7.0
  Revert "drm/amd/display: parse umc_info or vram_info based on ASIC"
  drm/amd/display: Fix failure to read vram info due to static BP_RESULT
  drm/amdgpu: enable GTT fallback handling for dGPUs only
  drm/amd/amdgpu: limit single process inside MES
  drm/fourcc: add AMD_FMT_MOD_TILE_GFX9_4K_D_X
  drm/amdgpu/mes12: correct kiq unmap latency
  drm/amdgpu: Support vcn and jpeg error info parsing
  drm/amd : Update MES API header file for v11 & v12
  drm/amd/amdkfd: add/remove kfd queues on start/stop KFD scheduling
  drm/amdkfd: change kfd process kref count at creation
  drm/amdgpu: Cleanup shift coding style
  drm/amd/amdgpu: Increase MES log buffer to dump mes scratch data
  drm/amdgpu: Implement virt req_ras_err_count
  drm/amdgpu: VF Query RAS Caps from Host if supported
  drm/amdgpu: Add msg handlers for SRIOV RAS Telemetry
  drm/amdgpu: Update SRIOV Exchange Headers for RAS Telemetry Support
  drm/amd/display: 3.2.309
  drm/amd/display: Adjust VSDB parser for replay feature
  ...

1 files changed, 385 insertions, 64 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index f1ffab5a1eae..f57cc72c43cf 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -87,16 +87,6 @@ int amdgpu_gfx_me_queue_to_bit(struct amdgpu_device *adev,
 	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)
 {
@@ -415,7 +405,7 @@ int amdgpu_gfx_mqd_sw_init(struct amdgpu_device *adev,
 		}
 
 		/* prepare MQD backup */
-		kiq->mqd_backup = kmalloc(mqd_size, GFP_KERNEL);
+		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);
@@ -438,7 +428,7 @@ int amdgpu_gfx_mqd_sw_init(struct amdgpu_device *adev,
 
 				ring->mqd_size = mqd_size;
 				/* prepare MQD backup */
-				adev->gfx.me.mqd_backup[i] = kmalloc(mqd_size, GFP_KERNEL);
+				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;
@@ -462,7 +452,7 @@ int amdgpu_gfx_mqd_sw_init(struct amdgpu_device *adev,
 
 			ring->mqd_size = mqd_size;
 			/* prepare MQD backup */
-			adev->gfx.mec.mqd_backup[j] = kmalloc(mqd_size, GFP_KERNEL);
+			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;
@@ -525,6 +515,9 @@ int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev, int xcc_id)
 	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
 		return -EINVAL;
 
+	if (!kiq_ring->sched.ready || adev->job_hang || 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)) {
@@ -538,20 +531,15 @@ int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev, int xcc_id)
 					   &adev->gfx.compute_ring[j],
 					   RESET_QUEUES, 0, 0);
 	}
-
-	/**
-	 * This is workaround: only skip kiq_ring test
-	 * during ras recovery in suspend stage for gfx9.4.3
+	/* 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.
 	 */
-	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_ras_in_recovery(adev)) {
-		spin_unlock(&kiq->ring_lock);
-		return 0;
-	}
+	r = amdgpu_ring_test_helper(kiq_ring);
 
-	if (kiq_ring->sched.ready && !adev->job_hang)
-		r = amdgpu_ring_test_helper(kiq_ring);
 	spin_unlock(&kiq->ring_lock);
 
 	return r;
@@ -579,8 +567,11 @@ int amdgpu_gfx_disable_kgq(struct amdgpu_device *adev, int xcc_id)
 	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
 		return -EINVAL;
 
-	spin_lock(&kiq->ring_lock);
+	if (!adev->gfx.kiq[0].ring.sched.ready || adev->job_hang)
+		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);
@@ -593,11 +584,17 @@ int amdgpu_gfx_disable_kgq(struct amdgpu_device *adev, int xcc_id)
 						   &adev->gfx.gfx_ring[j],
 						   PREEMPT_QUEUES, 0, 0);
 		}
-	}
+		/* Submit unmap queue packet */
+		amdgpu_ring_commit(kiq_ring);
 
-	if (adev->gfx.kiq[0].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(&kiq->ring_lock);
+		spin_unlock(&kiq->ring_lock);
+	}
 
 	return r;
 }
@@ -702,7 +699,13 @@ int amdgpu_gfx_enable_kcq(struct amdgpu_device *adev, int xcc_id)
 		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(&kiq->ring_lock);
 	if (r)
@@ -753,7 +756,13 @@ int amdgpu_gfx_enable_kgq(struct amdgpu_device *adev, int xcc_id)
 						 &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)
@@ -895,6 +904,9 @@ int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *r
 		if (r)
 			return r;
 
+		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)
@@ -1363,35 +1375,35 @@ static ssize_t amdgpu_gfx_set_compute_partition(struct device *dev,
 	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);
-	char *supported_partition;
+	struct amdgpu_xcp_mgr *xcp_mgr = adev->xcp_mgr;
+	int size = 0, mode;
+	char *sep = "";
 
-	/* TBD */
-	switch (NUM_XCC(adev->gfx.xcc_mask)) {
-	case 8:
-		supported_partition = "SPX, DPX, QPX, CPX";
-		break;
-	case 6:
-		supported_partition = "SPX, TPX, CPX";
-		break;
-	case 4:
-		supported_partition = "SPX, DPX, CPX";
-		break;
-	/* this seems only existing in emulation phase */
-	case 2:
-		supported_partition = "SPX, CPX";
-		break;
-	default:
-		supported_partition = "Not supported";
-		break;
+	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 = ", ";
 	}
 
-	return sysfs_emit(buf, "%s\n", supported_partition);
+	size += sysfs_emit_at(buf, size, "\n");
+
+	return size;
 }
 
 static int amdgpu_gfx_run_cleaner_shader_job(struct amdgpu_ring *ring)
@@ -1586,9 +1598,11 @@ static ssize_t amdgpu_gfx_set_enforce_isolation(struct device *dev,
 		if (adev->enforce_isolation[i] && !partition_values[i]) {
 			/* Going from enabled to disabled */
 			amdgpu_vmid_free_reserved(adev, AMDGPU_GFXHUB(i));
+			amdgpu_mes_set_enforce_isolation(adev, i, false);
 		} else if (!adev->enforce_isolation[i] && partition_values[i]) {
 			/* Going from disabled to enabled */
 			amdgpu_vmid_alloc_reserved(adev, AMDGPU_GFXHUB(i));
+			amdgpu_mes_set_enforce_isolation(adev, i, true);
 		}
 		adev->enforce_isolation[i] = partition_values[i];
 	}
@@ -1598,6 +1612,32 @@ static ssize_t amdgpu_gfx_set_enforce_isolation(struct device *dev,
 	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);
 
@@ -1611,45 +1651,136 @@ static DEVICE_ATTR(current_compute_partition, 0644,
 
 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);
 
-int amdgpu_gfx_sysfs_init(struct amdgpu_device *adev)
+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;
 
-	r = device_create_file(adev->dev, &dev_attr_available_compute_partition);
+	if (xcp_switch_supported)
+		r = device_create_file(adev->dev,
+				       &dev_attr_available_compute_partition);
 
 	return r;
 }
 
-void amdgpu_gfx_sysfs_fini(struct amdgpu_device *adev)
+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);
-	device_remove_file(adev->dev, &dev_attr_available_compute_partition);
+
+	if (xcp_switch_supported)
+		device_remove_file(adev->dev,
+				   &dev_attr_available_compute_partition);
 }
 
-int amdgpu_gfx_sysfs_isolation_shader_init(struct amdgpu_device *adev)
+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);
 
-	r = device_create_file(adev->dev, &dev_attr_run_cleaner_shader);
-	if (r)
+	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;
 
-	return 0;
+	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;
 }
 
-void amdgpu_gfx_sysfs_isolation_shader_fini(struct amdgpu_device *adev)
+static void amdgpu_gfx_sysfs_reset_mask_fini(struct amdgpu_device *adev)
 {
-	device_remove_file(adev->dev, &dev_attr_enforce_isolation);
-	device_remove_file(adev->dev, &dev_attr_run_cleaner_shader);
+	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)
+{
+	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,
@@ -1737,7 +1868,7 @@ static void amdgpu_gfx_kfd_sch_ctrl(struct amdgpu_device *adev, u32 idx,
 		if (adev->gfx.kfd_sch_req_count[idx] == 0 &&
 		    adev->gfx.kfd_sch_inactive[idx]) {
 			schedule_delayed_work(&adev->gfx.enforce_isolation[idx].work,
-					      GFX_SLICE_PERIOD);
+					      msecs_to_jiffies(adev->gfx.enforce_isolation_time[idx]));
 		}
 	} else {
 		if (adev->gfx.kfd_sch_req_count[idx] == 0) {
@@ -1792,8 +1923,9 @@ void amdgpu_gfx_enforce_isolation_handler(struct work_struct *work)
 			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,
-				      GFX_SLICE_PERIOD);
+				      msecs_to_jiffies(1));
 	} else {
 		/* Tell KFD to resume the runqueue */
 		if (adev->kfd.init_complete) {
@@ -1806,6 +1938,51 @@ void amdgpu_gfx_enforce_isolation_handler(struct work_struct *work)
 	mutex_unlock(&adev->enforce_isolation_mutex);
 }
 
+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]) {
+		/* 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);
+}
+
 void amdgpu_gfx_enforce_isolation_ring_begin_use(struct amdgpu_ring *ring)
 {
 	struct amdgpu_device *adev = ring->adev;
@@ -1822,6 +1999,9 @@ void amdgpu_gfx_enforce_isolation_ring_begin_use(struct amdgpu_ring *ring)
 	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]) {
 		if (adev->kfd.init_complete)
@@ -1853,3 +2033,144 @@ void amdgpu_gfx_enforce_isolation_ring_end_use(struct amdgpu_ring *ring)
 	}
 	mutex_unlock(&adev->enforce_isolation_mutex);
 }
+
+/*
+ * 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 = (1 << 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 |= 1 << 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 = (1 << 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 |= 1 << 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
+}