Merge tag 'drm-next-2021-08-31-1' of git://anongit.freedesktop.org/drm/drm

Pull drm updates from Dave Airlie:
 "Highlights:

   - i915 has seen a lot of refactoring and uAPI cleanups due to a
     change in the upstream direction going forward

     This has all been audited with known userspace, but there may be
     some pitfalls that were missed.

   - i915 now uses common TTM to enable discrete memory on DG1/2 GPUs

   - i915 enables Jasper and Elkhart Lake by default and has preliminary
     XeHP/DG2 support

   - amdgpu adds support for Cyan Skillfish

   - lots of implicit fencing rules documented and fixed up in drivers

   - msm now uses the core scheduler

   - the irq midlayer has been removed for non-legacy drivers

   - the sysfb code now works on more than x86.

  Otherwise the usual smattering of stuff everywhere, panels, bridges,
  refactorings.

  Detailed summary:

  core:
   - extract i915 eDP backlight into core
   - DP aux bus support
   - drm_device.irq_enabled removed
   - port drivers to native irq interfaces
   - export gem shadow plane handling for vgem
   - print proper driver name in framebuffer registration
   - driver fixes for implicit fencing rules
   - ARM fixed rate compression modifier added
   - updated fb damage handling
   - rmfb ioctl logging/docs
   - drop drm_gem_object_put_locked
   - define DRM_FORMAT_MAX_PLANES
   - add gem fb vmap/vunmap helpers
   - add lockdep_assert(once) helpers
   - mark drm irq midlayer as legacy
   - use offset adjusted bo mapping conversion

  vgaarb:
   - cleanups

  fbdev:
   - extend efifb handling to all arches
   - div by 0 fixes for multiple drivers

  udmabuf:
   - add hugepage mapping support

  dma-buf:
   - non-dynamic exporter fixups
   - document implicit fencing rules

  amdgpu:
   - Initial Cyan Skillfish support
   - switch virtual DCE over to vkms based atomic
   - VCN/JPEG power down fixes
   - NAVI PCIE link handling fixes
   - AMD HDMI freesync fixes
   - Yellow Carp + Beige Goby fixes
   - Clockgating/S0ix/SMU/EEPROM fixes
   - embed hw fence in job
   - rework dma-resv handling
   - ensure eviction to system ram

  amdkfd:
   - uapi: SVM address range query added
   - sysfs leak fix
   - GPUVM TLB optimizations
   - vmfault/migration counters

  i915:
   - Enable JSL and EHL by default
   - preliminary XeHP/DG2 support
   - remove all CNL support (never shipped)
   - move to TTM for discrete memory support
   - allow mixed object mmap handling
   - GEM uAPI spring cleaning
       - add I915_MMAP_OBJECT_FIXED
       - reinstate ADL-P mmap ioctls
       - drop a bunch of unused by userspace features
       - disable and remove GPU relocations
   - revert some i915 misfeatures
   - major refactoring of GuC for Gen11+
   - execbuffer object locking separate step
   - reject caching/set-domain on discrete
   - Enable pipe DMC loading on XE-LPD and ADL-P
   - add PSF GV point support
   - Refactor and fix DDI buffer translations
   - Clean up FBC CFB allocation code
   - Finish INTEL_GEN() and friends macro conversions

  nouveau:
   - add eDP backlight support
   - implicit fence fix

  msm:
   - a680/7c3 support
   - drm/scheduler conversion

  panfrost:
   - rework GPU reset

  virtio:
   - fix fencing for planes

  ast:
   - add detect support

  bochs:
   - move to tiny GPU driver

  vc4:
   - use hotplug irqs
   - HDMI codec support

  vmwgfx:
   - use internal vmware device headers

  ingenic:
   - demidlayering irq

  rcar-du:
   - shutdown fixes
   - convert to bridge connector helpers

  zynqmp-dsub:
   - misc fixes

  mgag200:
   - convert PLL handling to atomic

  mediatek:
   - MT8133 AAL support
   - gem mmap object support
   - MT8167 support

  etnaviv:
   - NXP Layerscape LS1028A SoC support
   - GEM mmap cleanups

  tegra:
   - new user API

  exynos:
   - missing unlock fix
   - build warning fix
   - use refcount_t"

* tag 'drm-next-2021-08-31-1' of git://anongit.freedesktop.org/drm/drm: (1318 commits)
  drm/amd/display: Move AllowDRAMSelfRefreshOrDRAMClockChangeInVblank to bounding box
  drm/amd/display: Remove duplicate dml init
  drm/amd/display: Update bounding box states (v2)
  drm/amd/display: Update number of DCN3 clock states
  drm/amdgpu: disable GFX CGCG in aldebaran
  drm/amdgpu: Clear RAS interrupt status on aldebaran
  drm/amdgpu: Add support for RAS XGMI err query
  drm/amdkfd: Account for SH/SE count when setting up cu masks.
  drm/amdgpu: rename amdgpu_bo_get_preferred_pin_domain
  drm/amdgpu: drop redundant cancel_delayed_work_sync call
  drm/amdgpu: add missing cleanups for more ASICs on UVD/VCE suspend
  drm/amdgpu: add missing cleanups for Polaris12 UVD/VCE on suspend
  drm/amdkfd: map SVM range with correct access permission
  drm/amdkfd: check access permisson to restore retry fault
  drm/amdgpu: Update RAS XGMI Error Query
  drm/amdgpu: Add driver infrastructure for MCA RAS
  drm/amd/display: Add Logging for HDMI color depth information
  drm/amd/amdgpu: consolidate PSP TA init shared buf functions
  drm/amd/amdgpu: add name field back to ras_common_if
  drm/amdgpu: Fix build with missing pm_suspend_target_state module export
  ...

17 files changed, 381 insertions, 142 deletions

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
index e48acdd03c1a..86afd37b098d 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
@@ -1393,6 +1393,7 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep,
 	long err = 0;
 	int i;
 	uint32_t *devices_arr = NULL;
+	bool table_freed = false;
 
 	dev = kfd_device_by_id(GET_GPU_ID(args->handle));
 	if (!dev)
@@ -1450,7 +1451,8 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep,
 			goto get_mem_obj_from_handle_failed;
 		}
 		err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(
-			peer->kgd, (struct kgd_mem *)mem, peer_pdd->drm_priv);
+			peer->kgd, (struct kgd_mem *)mem,
+			peer_pdd->drm_priv, &table_freed);
 		if (err) {
 			pr_err("Failed to map to gpu %d/%d\n",
 			       i, args->n_devices);
@@ -1468,16 +1470,17 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep,
 	}
 
 	/* Flush TLBs after waiting for the page table updates to complete */
-	for (i = 0; i < args->n_devices; i++) {
-		peer = kfd_device_by_id(devices_arr[i]);
-		if (WARN_ON_ONCE(!peer))
-			continue;
-		peer_pdd = kfd_get_process_device_data(peer, p);
-		if (WARN_ON_ONCE(!peer_pdd))
-			continue;
-		kfd_flush_tlb(peer_pdd, TLB_FLUSH_LEGACY);
+	if (table_freed) {
+		for (i = 0; i < args->n_devices; i++) {
+			peer = kfd_device_by_id(devices_arr[i]);
+			if (WARN_ON_ONCE(!peer))
+				continue;
+			peer_pdd = kfd_get_process_device_data(peer, p);
+			if (WARN_ON_ONCE(!peer_pdd))
+				continue;
+			kfd_flush_tlb(peer_pdd, TLB_FLUSH_LEGACY);
+		}
 	}
-
 	kfree(devices_arr);
 
 	return err;
@@ -1565,10 +1568,29 @@ static int kfd_ioctl_unmap_memory_from_gpu(struct file *filep,
 		}
 		args->n_success = i+1;
 	}
-	kfree(devices_arr);
-
 	mutex_unlock(&p->mutex);
 
+	if (dev->device_info->asic_family == CHIP_ALDEBARAN) {
+		err = amdgpu_amdkfd_gpuvm_sync_memory(dev->kgd,
+				(struct kgd_mem *) mem, true);
+		if (err) {
+			pr_debug("Sync memory failed, wait interrupted by user signal\n");
+			goto sync_memory_failed;
+		}
+
+		/* Flush TLBs after waiting for the page table updates to complete */
+		for (i = 0; i < args->n_devices; i++) {
+			peer = kfd_device_by_id(devices_arr[i]);
+			if (WARN_ON_ONCE(!peer))
+				continue;
+			peer_pdd = kfd_get_process_device_data(peer, p);
+			if (WARN_ON_ONCE(!peer_pdd))
+				continue;
+			kfd_flush_tlb(peer_pdd, TLB_FLUSH_HEAVYWEIGHT);
+		}
+	}
+	kfree(devices_arr);
+
 	return 0;
 
 bind_process_to_device_failed:
@@ -1576,6 +1598,7 @@ get_mem_obj_from_handle_failed:
 unmap_memory_from_gpu_failed:
 	mutex_unlock(&p->mutex);
 copy_from_user_failed:
+sync_memory_failed:
 	kfree(devices_arr);
 	return err;
 }
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
index c6b02aee4993..cfedfb1e8596 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c
@@ -1404,6 +1404,7 @@ static int kfd_fill_gpu_cache_info(struct kfd_dev *kdev,
 		break;
 	case CHIP_NAVI10:
 	case CHIP_NAVI12:
+	case CHIP_CYAN_SKILLFISH:
 		pcache_info = navi10_cache_info;
 		num_of_cache_types = ARRAY_SIZE(navi10_cache_info);
 		break;
@@ -1989,8 +1990,19 @@ static int kfd_fill_gpu_direct_io_link_to_cpu(int *avail_size,
 		sub_type_hdr->flags |= CRAT_IOLINK_FLAGS_BI_DIRECTIONAL;
 		sub_type_hdr->io_interface_type = CRAT_IOLINK_TYPE_XGMI;
 		sub_type_hdr->num_hops_xgmi = 1;
+		if (adev->asic_type == CHIP_ALDEBARAN) {
+			sub_type_hdr->minimum_bandwidth_mbs =
+					amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(
+							kdev->kgd, NULL, true);
+			sub_type_hdr->maximum_bandwidth_mbs =
+					sub_type_hdr->minimum_bandwidth_mbs;
+		}
 	} else {
 		sub_type_hdr->io_interface_type = CRAT_IOLINK_TYPE_PCIEXPRESS;
+		sub_type_hdr->minimum_bandwidth_mbs =
+				amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->kgd, true);
+		sub_type_hdr->maximum_bandwidth_mbs =
+				amdgpu_amdkfd_get_pcie_bandwidth_mbytes(kdev->kgd, false);
 	}
 
 	sub_type_hdr->proximity_domain_from = proximity_domain;
@@ -2033,6 +2045,11 @@ static int kfd_fill_gpu_xgmi_link_to_gpu(int *avail_size,
 	sub_type_hdr->proximity_domain_to = proximity_domain_to;
 	sub_type_hdr->num_hops_xgmi =
 		amdgpu_amdkfd_get_xgmi_hops_count(kdev->kgd, peer_kdev->kgd);
+	sub_type_hdr->maximum_bandwidth_mbs =
+		amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->kgd, peer_kdev->kgd, false);
+	sub_type_hdr->minimum_bandwidth_mbs = sub_type_hdr->maximum_bandwidth_mbs ?
+		amdgpu_amdkfd_get_xgmi_bandwidth_mbytes(kdev->kgd, NULL, true) : 0;
+
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index 6b57dfd2cd2a..16a57b70cc1a 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -84,12 +84,14 @@ static const struct kfd2kgd_calls *kfd2kgd_funcs[] = {
 	[CHIP_DIMGREY_CAVEFISH] = &gfx_v10_3_kfd2kgd,
 	[CHIP_BEIGE_GOBY] = &gfx_v10_3_kfd2kgd,
 	[CHIP_YELLOW_CARP] = &gfx_v10_3_kfd2kgd,
+	[CHIP_CYAN_SKILLFISH] = &gfx_v10_kfd2kgd,
 };
 
 #ifdef KFD_SUPPORT_IOMMU_V2
 static const struct kfd_device_info kaveri_device_info = {
 	.asic_family = CHIP_KAVERI,
 	.asic_name = "kaveri",
+	.gfx_target_version = 70000,
 	.max_pasid_bits = 16,
 	/* max num of queues for KV.TODO should be a dynamic value */
 	.max_no_of_hqd	= 24,
@@ -109,6 +111,7 @@ static const struct kfd_device_info kaveri_device_info = {
 static const struct kfd_device_info carrizo_device_info = {
 	.asic_family = CHIP_CARRIZO,
 	.asic_name = "carrizo",
+	.gfx_target_version = 80001,
 	.max_pasid_bits = 16,
 	/* max num of queues for CZ.TODO should be a dynamic value */
 	.max_no_of_hqd	= 24,
@@ -129,6 +132,7 @@ static const struct kfd_device_info carrizo_device_info = {
 static const struct kfd_device_info raven_device_info = {
 	.asic_family = CHIP_RAVEN,
 	.asic_name = "raven",
+	.gfx_target_version = 90002,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 8,
@@ -147,6 +151,7 @@ static const struct kfd_device_info raven_device_info = {
 static const struct kfd_device_info hawaii_device_info = {
 	.asic_family = CHIP_HAWAII,
 	.asic_name = "hawaii",
+	.gfx_target_version = 70001,
 	.max_pasid_bits = 16,
 	/* max num of queues for KV.TODO should be a dynamic value */
 	.max_no_of_hqd	= 24,
@@ -166,6 +171,7 @@ static const struct kfd_device_info hawaii_device_info = {
 static const struct kfd_device_info tonga_device_info = {
 	.asic_family = CHIP_TONGA,
 	.asic_name = "tonga",
+	.gfx_target_version = 80002,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 4,
@@ -184,6 +190,7 @@ static const struct kfd_device_info tonga_device_info = {
 static const struct kfd_device_info fiji_device_info = {
 	.asic_family = CHIP_FIJI,
 	.asic_name = "fiji",
+	.gfx_target_version = 80003,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 4,
@@ -202,6 +209,7 @@ static const struct kfd_device_info fiji_device_info = {
 static const struct kfd_device_info fiji_vf_device_info = {
 	.asic_family = CHIP_FIJI,
 	.asic_name = "fiji",
+	.gfx_target_version = 80003,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 4,
@@ -221,6 +229,7 @@ static const struct kfd_device_info fiji_vf_device_info = {
 static const struct kfd_device_info polaris10_device_info = {
 	.asic_family = CHIP_POLARIS10,
 	.asic_name = "polaris10",
+	.gfx_target_version = 80003,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 4,
@@ -239,6 +248,7 @@ static const struct kfd_device_info polaris10_device_info = {
 static const struct kfd_device_info polaris10_vf_device_info = {
 	.asic_family = CHIP_POLARIS10,
 	.asic_name = "polaris10",
+	.gfx_target_version = 80003,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 4,
@@ -257,6 +267,7 @@ static const struct kfd_device_info polaris10_vf_device_info = {
 static const struct kfd_device_info polaris11_device_info = {
 	.asic_family = CHIP_POLARIS11,
 	.asic_name = "polaris11",
+	.gfx_target_version = 80003,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 4,
@@ -275,6 +286,7 @@ static const struct kfd_device_info polaris11_device_info = {
 static const struct kfd_device_info polaris12_device_info = {
 	.asic_family = CHIP_POLARIS12,
 	.asic_name = "polaris12",
+	.gfx_target_version = 80003,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 4,
@@ -293,6 +305,7 @@ static const struct kfd_device_info polaris12_device_info = {
 static const struct kfd_device_info vegam_device_info = {
 	.asic_family = CHIP_VEGAM,
 	.asic_name = "vegam",
+	.gfx_target_version = 80003,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 4,
@@ -311,6 +324,7 @@ static const struct kfd_device_info vegam_device_info = {
 static const struct kfd_device_info vega10_device_info = {
 	.asic_family = CHIP_VEGA10,
 	.asic_name = "vega10",
+	.gfx_target_version = 90000,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 8,
@@ -329,6 +343,7 @@ static const struct kfd_device_info vega10_device_info = {
 static const struct kfd_device_info vega10_vf_device_info = {
 	.asic_family = CHIP_VEGA10,
 	.asic_name = "vega10",
+	.gfx_target_version = 90000,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 8,
@@ -347,6 +362,7 @@ static const struct kfd_device_info vega10_vf_device_info = {
 static const struct kfd_device_info vega12_device_info = {
 	.asic_family = CHIP_VEGA12,
 	.asic_name = "vega12",
+	.gfx_target_version = 90004,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 8,
@@ -365,6 +381,7 @@ static const struct kfd_device_info vega12_device_info = {
 static const struct kfd_device_info vega20_device_info = {
 	.asic_family = CHIP_VEGA20,
 	.asic_name = "vega20",
+	.gfx_target_version = 90006,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd	= 24,
 	.doorbell_size	= 8,
@@ -383,6 +400,7 @@ static const struct kfd_device_info vega20_device_info = {
 static const struct kfd_device_info arcturus_device_info = {
 	.asic_family = CHIP_ARCTURUS,
 	.asic_name = "arcturus",
+	.gfx_target_version = 90008,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd	= 24,
 	.doorbell_size	= 8,
@@ -401,6 +419,7 @@ static const struct kfd_device_info arcturus_device_info = {
 static const struct kfd_device_info aldebaran_device_info = {
 	.asic_family = CHIP_ALDEBARAN,
 	.asic_name = "aldebaran",
+	.gfx_target_version = 90010,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd	= 24,
 	.doorbell_size	= 8,
@@ -419,6 +438,7 @@ static const struct kfd_device_info aldebaran_device_info = {
 static const struct kfd_device_info renoir_device_info = {
 	.asic_family = CHIP_RENOIR,
 	.asic_name = "renoir",
+	.gfx_target_version = 90002,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 8,
@@ -437,6 +457,7 @@ static const struct kfd_device_info renoir_device_info = {
 static const struct kfd_device_info navi10_device_info = {
 	.asic_family = CHIP_NAVI10,
 	.asic_name = "navi10",
+	.gfx_target_version = 100100,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 8,
@@ -455,6 +476,7 @@ static const struct kfd_device_info navi10_device_info = {
 static const struct kfd_device_info navi12_device_info = {
 	.asic_family = CHIP_NAVI12,
 	.asic_name = "navi12",
+	.gfx_target_version = 100101,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 8,
@@ -473,6 +495,7 @@ static const struct kfd_device_info navi12_device_info = {
 static const struct kfd_device_info navi14_device_info = {
 	.asic_family = CHIP_NAVI14,
 	.asic_name = "navi14",
+	.gfx_target_version = 100102,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 8,
@@ -491,6 +514,7 @@ static const struct kfd_device_info navi14_device_info = {
 static const struct kfd_device_info sienna_cichlid_device_info = {
 	.asic_family = CHIP_SIENNA_CICHLID,
 	.asic_name = "sienna_cichlid",
+	.gfx_target_version = 100300,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 8,
@@ -509,6 +533,7 @@ static const struct kfd_device_info sienna_cichlid_device_info = {
 static const struct kfd_device_info navy_flounder_device_info = {
 	.asic_family = CHIP_NAVY_FLOUNDER,
 	.asic_name = "navy_flounder",
+	.gfx_target_version = 100301,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 8,
@@ -527,6 +552,7 @@ static const struct kfd_device_info navy_flounder_device_info = {
 static const struct kfd_device_info vangogh_device_info = {
 	.asic_family = CHIP_VANGOGH,
 	.asic_name = "vangogh",
+	.gfx_target_version = 100303,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 8,
@@ -545,6 +571,7 @@ static const struct kfd_device_info vangogh_device_info = {
 static const struct kfd_device_info dimgrey_cavefish_device_info = {
 	.asic_family = CHIP_DIMGREY_CAVEFISH,
 	.asic_name = "dimgrey_cavefish",
+	.gfx_target_version = 100302,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 8,
@@ -563,6 +590,7 @@ static const struct kfd_device_info dimgrey_cavefish_device_info = {
 static const struct kfd_device_info beige_goby_device_info = {
 	.asic_family = CHIP_BEIGE_GOBY,
 	.asic_name = "beige_goby",
+	.gfx_target_version = 100304,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 8,
@@ -581,6 +609,7 @@ static const struct kfd_device_info beige_goby_device_info = {
 static const struct kfd_device_info yellow_carp_device_info = {
 	.asic_family = CHIP_YELLOW_CARP,
 	.asic_name = "yellow_carp",
+	.gfx_target_version = 100305,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
 	.doorbell_size  = 8,
@@ -596,6 +625,25 @@ static const struct kfd_device_info yellow_carp_device_info = {
 	.num_sdma_queues_per_engine = 2,
 };
 
+static const struct kfd_device_info cyan_skillfish_device_info = {
+	.asic_family = CHIP_CYAN_SKILLFISH,
+	.asic_name = "cyan_skillfish",
+	.gfx_target_version = 100103,
+	.max_pasid_bits = 16,
+	.max_no_of_hqd  = 24,
+	.doorbell_size  = 8,
+	.ih_ring_entry_size = 8 * sizeof(uint32_t),
+	.event_interrupt_class = &event_interrupt_class_v9,
+	.num_of_watch_points = 4,
+	.mqd_size_aligned = MQD_SIZE_ALIGNED,
+	.needs_iommu_device = false,
+	.supports_cwsr = true,
+	.needs_pci_atomics = true,
+	.num_sdma_engines = 2,
+	.num_xgmi_sdma_engines = 0,
+	.num_sdma_queues_per_engine = 8,
+};
+
 /* For each entry, [0] is regular and [1] is virtualisation device. */
 static const struct kfd_device_info *kfd_supported_devices[][2] = {
 #ifdef KFD_SUPPORT_IOMMU_V2
@@ -625,6 +673,7 @@ static const struct kfd_device_info *kfd_supported_devices[][2] = {
 	[CHIP_DIMGREY_CAVEFISH] = {&dimgrey_cavefish_device_info, &dimgrey_cavefish_device_info},
 	[CHIP_BEIGE_GOBY] = {&beige_goby_device_info, &beige_goby_device_info},
 	[CHIP_YELLOW_CARP] = {&yellow_carp_device_info, NULL},
+	[CHIP_CYAN_SKILLFISH] = {&cyan_skillfish_device_info, NULL},
 };
 
 static int kfd_gtt_sa_init(struct kfd_dev *kfd, unsigned int buf_size,
@@ -1369,7 +1418,7 @@ void kfd_dec_compute_active(struct kfd_dev *kfd)
 	WARN_ONCE(count < 0, "Compute profile ref. count error");
 }
 
-void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint32_t throttle_bitmask)
+void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint64_t throttle_bitmask)
 {
 	if (kfd && kfd->init_complete)
 		kfd_smi_event_update_thermal_throttling(kfd, throttle_bitmask);
@@ -1382,18 +1431,12 @@ void kgd2kfd_smi_event_throttle(struct kfd_dev *kfd, uint32_t throttle_bitmask)
  */
 int kfd_debugfs_hang_hws(struct kfd_dev *dev)
 {
-	int r = 0;
-
 	if (dev->dqm->sched_policy != KFD_SCHED_POLICY_HWS) {
 		pr_err("HWS is not enabled");
 		return -EINVAL;
 	}
 
-	r = pm_debugfs_hang_hws(&dev->dqm->packets);
-	if (!r)
-		r = dqm_debugfs_execute_queues(dev->dqm);
-
-	return r;
+	return dqm_debugfs_hang_hws(dev->dqm);
 }
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index 16a1713808c2..f8fce9d05f50 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -211,6 +211,15 @@ static void deallocate_doorbell(struct qcm_process_device *qpd,
 	WARN_ON(!old);
 }
 
+static void program_trap_handler_settings(struct device_queue_manager *dqm,
+				struct qcm_process_device *qpd)
+{
+	if (dqm->dev->kfd2kgd->program_trap_handler_settings)
+		dqm->dev->kfd2kgd->program_trap_handler_settings(
+						dqm->dev->kgd, qpd->vmid,
+						qpd->tba_addr, qpd->tma_addr);
+}
+
 static int allocate_vmid(struct device_queue_manager *dqm,
 			struct qcm_process_device *qpd,
 			struct queue *q)
@@ -241,6 +250,10 @@ static int allocate_vmid(struct device_queue_manager *dqm,
 
 	program_sh_mem_settings(dqm, qpd);
 
+	if (dqm->dev->device_info->asic_family >= CHIP_VEGA10 &&
+	    dqm->dev->cwsr_enabled)
+		program_trap_handler_settings(dqm, qpd);
+
 	/* qpd->page_table_base is set earlier when register_process()
 	 * is called, i.e. when the first queue is created.
 	 */
@@ -260,7 +273,7 @@ static int allocate_vmid(struct device_queue_manager *dqm,
 static int flush_texture_cache_nocpsch(struct kfd_dev *kdev,
 				struct qcm_process_device *qpd)
 {
-	const struct packet_manager_funcs *pmf = qpd->dqm->packets.pmf;
+	const struct packet_manager_funcs *pmf = qpd->dqm->packet_mgr.pmf;
 	int ret;
 
 	if (!qpd->ib_kaddr)
@@ -582,7 +595,9 @@ static int update_queue(struct device_queue_manager *dqm, struct queue *q)
 		}
 
 		retval = mqd_mgr->destroy_mqd(mqd_mgr, q->mqd,
-				KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN,
+				(dqm->dev->cwsr_enabled?
+				 KFD_PREEMPT_TYPE_WAVEFRONT_SAVE:
+				KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN),
 				KFD_UNMAP_LATENCY_MS, q->pipe, q->queue);
 		if (retval) {
 			pr_err("destroy mqd failed\n");
@@ -675,7 +690,9 @@ static int evict_process_queues_nocpsch(struct device_queue_manager *dqm,
 			continue;
 
 		retval = mqd_mgr->destroy_mqd(mqd_mgr, q->mqd,
-				KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN,
+				(dqm->dev->cwsr_enabled?
+				 KFD_PREEMPT_TYPE_WAVEFRONT_SAVE:
+				KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN),
 				KFD_UNMAP_LATENCY_MS, q->pipe, q->queue);
 		if (retval && !ret)
 			/* Return the first error, but keep going to
@@ -1000,7 +1017,7 @@ static int start_nocpsch(struct device_queue_manager *dqm)
 	init_interrupts(dqm);
 	
 	if (dqm->dev->device_info->asic_family == CHIP_HAWAII)
-		return pm_init(&dqm->packets, dqm);
+		return pm_init(&dqm->packet_mgr, dqm);
 	dqm->sched_running = true;
 
 	return 0;
@@ -1009,7 +1026,7 @@ static int start_nocpsch(struct device_queue_manager *dqm)
 static int stop_nocpsch(struct device_queue_manager *dqm)
 {
 	if (dqm->dev->device_info->asic_family == CHIP_HAWAII)
-		pm_uninit(&dqm->packets, false);
+		pm_uninit(&dqm->packet_mgr, false);
 	dqm->sched_running = false;
 
 	return 0;
@@ -1124,7 +1141,7 @@ static int set_sched_resources(struct device_queue_manager *dqm)
 			"queue mask: 0x%8llX\n",
 			res.vmid_mask, res.queue_mask);
 
-	return pm_send_set_resources(&dqm->packets, &res);
+	return pm_send_set_resources(&dqm->packet_mgr, &res);
 }
 
 static int initialize_cpsch(struct device_queue_manager *dqm)
@@ -1164,7 +1181,8 @@ static int start_cpsch(struct device_queue_manager *dqm)
 
 	retval = 0;
 
-	retval = pm_init(&dqm->packets, dqm);
+	dqm_lock(dqm);
+	retval = pm_init(&dqm->packet_mgr, dqm);
 	if (retval)
 		goto fail_packet_manager_init;
 
@@ -1186,7 +1204,6 @@ static int start_cpsch(struct device_queue_manager *dqm)
 
 	init_interrupts(dqm);
 
-	dqm_lock(dqm);
 	/* clear hang status when driver try to start the hw scheduler */
 	dqm->is_hws_hang = false;
 	dqm->is_resetting = false;
@@ -1197,8 +1214,9 @@ static int start_cpsch(struct device_queue_manager *dqm)
 	return 0;
 fail_allocate_vidmem:
 fail_set_sched_resources:
-	pm_uninit(&dqm->packets, false);
+	pm_uninit(&dqm->packet_mgr, false);
 fail_packet_manager_init:
+	dqm_unlock(dqm);
 	return retval;
 }
 
@@ -1211,12 +1229,12 @@ static int stop_cpsch(struct device_queue_manager *dqm)
 		unmap_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0);
 	hanging = dqm->is_hws_hang || dqm->is_resetting;
 	dqm->sched_running = false;
-	dqm_unlock(dqm);
 
-	pm_release_ib(&dqm->packets);
+	pm_release_ib(&dqm->packet_mgr);
 
 	kfd_gtt_sa_free(dqm->dev, dqm->fence_mem);
-	pm_uninit(&dqm->packets, hanging);
+	pm_uninit(&dqm->packet_mgr, hanging);
+	dqm_unlock(dqm);
 
 	return 0;
 }
@@ -1390,7 +1408,7 @@ static int map_queues_cpsch(struct device_queue_manager *dqm)
 	if (dqm->active_runlist)
 		return 0;
 
-	retval = pm_send_runlist(&dqm->packets, &dqm->queues);
+	retval = pm_send_runlist(&dqm->packet_mgr, &dqm->queues);
 	pr_debug("%s sent runlist\n", __func__);
 	if (retval) {
 		pr_err("failed to execute runlist\n");
@@ -1416,13 +1434,13 @@ static int unmap_queues_cpsch(struct device_queue_manager *dqm,
 	if (!dqm->active_runlist)
 		return retval;
 
-	retval = pm_send_unmap_queue(&dqm->packets, KFD_QUEUE_TYPE_COMPUTE,
+	retval = pm_send_unmap_queue(&dqm->packet_mgr, KFD_QUEUE_TYPE_COMPUTE,
 			filter, filter_param, false, 0);
 	if (retval)
 		return retval;
 
 	*dqm->fence_addr = KFD_FENCE_INIT;
-	pm_send_query_status(&dqm->packets, dqm->fence_gpu_addr,
+	pm_send_query_status(&dqm->packet_mgr, dqm->fence_gpu_addr,
 				KFD_FENCE_COMPLETED);
 	/* should be timed out */
 	retval = amdkfd_fence_wait_timeout(dqm->fence_addr, KFD_FENCE_COMPLETED,
@@ -1448,14 +1466,14 @@ static int unmap_queues_cpsch(struct device_queue_manager *dqm,
 	 * check those fields
 	 */
 	mqd_mgr = dqm->mqd_mgrs[KFD_MQD_TYPE_HIQ];
-	if (mqd_mgr->read_doorbell_id(dqm->packets.priv_queue->queue->mqd)) {
+	if (mqd_mgr->read_doorbell_id(dqm->packet_mgr.priv_queue->queue->mqd)) {
 		pr_err("HIQ MQD's queue_doorbell_id0 is not 0, Queue preemption time out\n");
 		while (halt_if_hws_hang)
 			schedule();
 		return -ETIME;
 	}
 
-	pm_release_ib(&dqm->packets);
+	pm_release_ib(&dqm->packet_mgr);
 	dqm->active_runlist = false;
 
 	return retval;
@@ -1946,6 +1964,7 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev)
 	case CHIP_DIMGREY_CAVEFISH:
 	case CHIP_BEIGE_GOBY:
 	case CHIP_YELLOW_CARP:
+	case CHIP_CYAN_SKILLFISH:
 		device_queue_manager_init_v10_navi10(&dqm->asic_ops);
 		break;
 	default:
@@ -2099,11 +2118,16 @@ int dqm_debugfs_hqds(struct seq_file *m, void *data)
 	return r;
 }
 
-int dqm_debugfs_execute_queues(struct device_queue_manager *dqm)
+int dqm_debugfs_hang_hws(struct device_queue_manager *dqm)
 {
 	int r = 0;
 
 	dqm_lock(dqm);
+	r = pm_debugfs_hang_hws(&dqm->packet_mgr);
+	if (r) {
+		dqm_unlock(dqm);
+		return r;
+	}
 	dqm->active_runlist = true;
 	r = execute_queues_cpsch(dqm, KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES, 0);
 	dqm_unlock(dqm);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
index 71e2fde56b2b..c8719682c4da 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h
@@ -169,7 +169,7 @@ struct device_queue_manager {
 	struct device_queue_manager_asic_ops asic_ops;
 
 	struct mqd_manager	*mqd_mgrs[KFD_MQD_TYPE_MAX];
-	struct packet_manager	packets;
+	struct packet_manager	packet_mgr;
 	struct kfd_dev		*dev;
 	struct mutex		lock_hidden; /* use dqm_lock/unlock(dqm) */
 	struct list_head	queues;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c
index a9b329f0f862..2e86692def19 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c
@@ -422,6 +422,7 @@ int kfd_init_apertures(struct kfd_process *process)
 			case CHIP_DIMGREY_CAVEFISH:
 			case CHIP_BEIGE_GOBY:
 			case CHIP_YELLOW_CARP:
+			case CHIP_CYAN_SKILLFISH:
 				kfd_init_apertures_v9(pdd, id);
 				break;
 			default:
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
index 88813dad731f..c021519af810 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
@@ -98,36 +98,78 @@ void mqd_symmetrically_map_cu_mask(struct mqd_manager *mm,
 		uint32_t *se_mask)
 {
 	struct kfd_cu_info cu_info;
-	uint32_t cu_per_se[KFD_MAX_NUM_SE] = {0};
-	int i, se, sh, cu = 0;
-
+	uint32_t cu_per_sh[KFD_MAX_NUM_SE][KFD_MAX_NUM_SH_PER_SE] = {0};
+	int i, se, sh, cu;
 	amdgpu_amdkfd_get_cu_info(mm->dev->kgd, &cu_info);
 
 	if (cu_mask_count > cu_info.cu_active_number)
 		cu_mask_count = cu_info.cu_active_number;
 
+	/* Exceeding these bounds corrupts the stack and indicates a coding error.
+	 * Returning with no CU's enabled will hang the queue, which should be
+	 * attention grabbing.
+	 */
+	if (cu_info.num_shader_engines > KFD_MAX_NUM_SE) {
+		pr_err("Exceeded KFD_MAX_NUM_SE, chip reports %d\n", cu_info.num_shader_engines);
+		return;
+	}
+	if (cu_info.num_shader_arrays_per_engine > KFD_MAX_NUM_SH_PER_SE) {
+		pr_err("Exceeded KFD_MAX_NUM_SH, chip reports %d\n",
+			cu_info.num_shader_arrays_per_engine * cu_info.num_shader_engines);
+		return;
+	}
+	/* Count active CUs per SH.
+	 *
+	 * Some CUs in an SH may be disabled.	HW expects disabled CUs to be
+	 * represented in the high bits of each SH's enable mask (the upper and lower
+	 * 16 bits of se_mask) and will take care of the actual distribution of
+	 * disabled CUs within each SH automatically.
+	 * Each half of se_mask must be filled only on bits 0-cu_per_sh[se][sh]-1.
+	 *
+	 * See note on Arcturus cu_bitmap layout in gfx_v9_0_get_cu_info.
+	 */
 	for (se = 0; se < cu_info.num_shader_engines; se++)
 		for (sh = 0; sh < cu_info.num_shader_arrays_per_engine; sh++)
-			cu_per_se[se] += hweight32(cu_info.cu_bitmap[se % 4][sh + (se / 4)]);
-
-	/* Symmetrically map cu_mask to all SEs:
-	 * cu_mask[0] bit0 -> se_mask[0] bit0;
-	 * cu_mask[0] bit1 -> se_mask[1] bit0;
-	 * ... (if # SE is 4)
-	 * cu_mask[0] bit4 -> se_mask[0] bit1;
+			cu_per_sh[se][sh] = hweight32(cu_info.cu_bitmap[se % 4][sh + (se / 4)]);
+
+	/* Symmetrically map cu_mask to all SEs & SHs:
+	 * se_mask programs up to 2 SH in the upper and lower 16 bits.
+	 *
+	 * Examples
+	 * Assuming 1 SH/SE, 4 SEs:
+	 * cu_mask[0] bit0 -> se_mask[0] bit0
+	 * cu_mask[0] bit1 -> se_mask[1] bit0
+	 * ...
+	 * cu_mask[0] bit4 -> se_mask[0] bit1
+	 * ...
+	 *
+	 * Assuming 2 SH/SE, 4 SEs
+	 * cu_mask[0] bit0 -> se_mask[0] bit0 (SE0,SH0,CU0)
+	 * cu_mask[0] bit1 -> se_mask[1] bit0 (SE1,SH0,CU0)
+	 * ...
+	 * cu_mask[0] bit4 -> se_mask[0] bit16 (SE0,SH1,CU0)
+	 * cu_mask[0] bit5 -> se_mask[1] bit16 (SE1,SH1,CU0)
+	 * ...
+	 * cu_mask[0] bit8 -> se_mask[0] bit1 (SE0,SH0,CU1)
 	 * ...
+	 *
+	 * First ensure all CUs are disabled, then enable user specified CUs.
 	 */
-	se = 0;
-	for (i = 0; i < cu_mask_count; i++) {
-		if (cu_mask[i / 32] & (1 << (i % 32)))
-			se_mask[se] |= 1 << cu;
-
-		do {
-			se++;
-			if (se == cu_info.num_shader_engines) {
-				se = 0;
-				cu++;
+	for (i = 0; i < cu_info.num_shader_engines; i++)
+		se_mask[i] = 0;
+
+	i = 0;
+	for (cu = 0; cu < 16; cu++) {
+		for (sh = 0; sh < cu_info.num_shader_arrays_per_engine; sh++) {
+			for (se = 0; se < cu_info.num_shader_engines; se++) {
+				if (cu_per_sh[se][sh] > cu) {
+					if (cu_mask[i / 32] & (1 << (i % 32)))
+						se_mask[se] |= 1 << (cu + sh * 16);
+					i++;
+					if (i == cu_mask_count)
+						return;
+				}
 			}
-		} while (cu >= cu_per_se[se] && cu < 32);
+		}
 	}
 }
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h
index b5e2ea7550d4..6e6918ccedfd 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.h
@@ -27,6 +27,7 @@
 #include "kfd_priv.h"
 
 #define KFD_MAX_NUM_SE 8
+#define KFD_MAX_NUM_SH_PER_SE 2
 
 /**
  * struct mqd_manager
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
index d8e940f03102..e547f1f8c49f 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c
@@ -251,6 +251,7 @@ int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm)
 	case CHIP_DIMGREY_CAVEFISH:
 	case CHIP_BEIGE_GOBY:
 	case CHIP_YELLOW_CARP:
+	case CHIP_CYAN_SKILLFISH:
 		pm->pmf = &kfd_v9_pm_funcs;
 		break;
 	case CHIP_ALDEBARAN:
@@ -278,6 +279,7 @@ void pm_uninit(struct packet_manager *pm, bool hanging)
 {
 	mutex_destroy(&pm->lock);
 	kernel_queue_uninit(pm->priv_queue, hanging);
+	pm->priv_queue = NULL;
 }
 
 int pm_send_set_resources(struct packet_manager *pm,
@@ -447,6 +449,9 @@ int pm_debugfs_hang_hws(struct packet_manager *pm)
 	uint32_t *buffer, size;
 	int r = 0;
 
+	if (!pm->priv_queue)
+		return -EAGAIN;
+
 	size = pm->pmf->query_status_size;
 	mutex_lock(&pm->lock);
 	kq_acquire_packet_buffer(pm->priv_queue,
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 3426743ed228..ab83b0de6b22 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -196,6 +196,7 @@ struct kfd_event_interrupt_class {
 struct kfd_device_info {
 	enum amd_asic_type asic_family;
 	const char *asic_name;
+	uint32_t gfx_target_version;
 	const struct kfd_event_interrupt_class *event_interrupt_class;
 	unsigned int max_pasid_bits;
 	unsigned int max_no_of_hqd;
@@ -1194,7 +1195,7 @@ int pm_debugfs_runlist(struct seq_file *m, void *data);
 
 int kfd_debugfs_hang_hws(struct kfd_dev *dev);
 int pm_debugfs_hang_hws(struct packet_manager *pm);
-int dqm_debugfs_execute_queues(struct device_queue_manager *dqm);
+int dqm_debugfs_hang_hws(struct device_queue_manager *dqm);
 
 #else
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 8a2c6fc438c0..21ec8a18cad2 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -714,7 +714,8 @@ static int kfd_process_alloc_gpuvm(struct kfd_process_device *pdd,
 	if (err)
 		goto err_alloc_mem;
 
-	err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(kdev->kgd, mem, pdd->drm_priv);
+	err = amdgpu_amdkfd_gpuvm_map_memory_to_gpu(kdev->kgd, mem,
+			pdd->drm_priv, NULL);
 	if (err)
 		goto err_map_mem;
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
index 246522423559..ed4bc5f844ce 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c
@@ -205,23 +205,23 @@ void kfd_smi_event_update_gpu_reset(struct kfd_dev *dev, bool post_reset)
 }
 
 void kfd_smi_event_update_thermal_throttling(struct kfd_dev *dev,
-					     uint32_t throttle_bitmask)
+					     uint64_t throttle_bitmask)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)dev->kgd;
 	/*
 	 * ThermalThrottle msg = throttle_bitmask(8):
 	 * 			 thermal_interrupt_count(16):
-	 * 1 byte event + 1 byte space + 8 byte throttle_bitmask +
+	 * 1 byte event + 1 byte space + 16 byte throttle_bitmask +
 	 * 1 byte : + 16 byte thermal_interupt_counter + 1 byte \n +
-	 * 1 byte \0 = 29
+	 * 1 byte \0 = 37
 	 */
-	char fifo_in[29];
+	char fifo_in[37];
 	int len;
 
 	if (list_empty(&dev->smi_clients))
 		return;
 
-	len = snprintf(fifo_in, sizeof(fifo_in), "%x %x:%llx\n",
+	len = snprintf(fifo_in, sizeof(fifo_in), "%x %llx:%llx\n",
 		       KFD_SMI_EVENT_THERMAL_THROTTLE, throttle_bitmask,
 		       atomic64_read(&adev->smu.throttle_int_counter));
 
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.h b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.h
index b9b0438202e2..bffd0c32b060 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.h
@@ -26,7 +26,7 @@
 int kfd_smi_event_open(struct kfd_dev *dev, uint32_t *fd);
 void kfd_smi_event_update_vmfault(struct kfd_dev *dev, uint16_t pasid);
 void kfd_smi_event_update_thermal_throttling(struct kfd_dev *dev,
-					     uint32_t throttle_bitmask);
+					     uint64_t throttle_bitmask);
 void kfd_smi_event_update_gpu_reset(struct kfd_dev *dev, bool post_reset);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
index e883731c3f8f..491373fcdb38 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
@@ -120,6 +120,7 @@ static void svm_range_remove_notifier(struct svm_range *prange)
 
 static int
 svm_range_dma_map_dev(struct amdgpu_device *adev, struct svm_range *prange,
+		      unsigned long offset, unsigned long npages,
 		      unsigned long *hmm_pfns, uint32_t gpuidx)
 {
 	enum dma_data_direction dir = DMA_BIDIRECTIONAL;
@@ -136,7 +137,8 @@ svm_range_dma_map_dev(struct amdgpu_device *adev, struct svm_range *prange,
 		prange->dma_addr[gpuidx] = addr;
 	}
 
-	for (i = 0; i < prange->npages; i++) {
+	addr += offset;
+	for (i = 0; i < npages; i++) {
 		if (WARN_ONCE(addr[i] && !dma_mapping_error(dev, addr[i]),
 			      "leaking dma mapping\n"))
 			dma_unmap_page(dev, addr[i], PAGE_SIZE, dir);
@@ -167,6 +169,7 @@ svm_range_dma_map_dev(struct amdgpu_device *adev, struct svm_range *prange,
 
 static int
 svm_range_dma_map(struct svm_range *prange, unsigned long *bitmap,
+		  unsigned long offset, unsigned long npages,
 		  unsigned long *hmm_pfns)
 {
 	struct kfd_process *p;
@@ -187,7 +190,8 @@ svm_range_dma_map(struct svm_range *prange, unsigned long *bitmap,
 		}
 		adev = (struct amdgpu_device *)pdd->dev->kgd;
 
-		r = svm_range_dma_map_dev(adev, prange, hmm_pfns, gpuidx);
+		r = svm_range_dma_map_dev(adev, prange, offset, npages,
+					  hmm_pfns, gpuidx);
 		if (r)
 			break;
 	}
@@ -1088,11 +1092,6 @@ svm_range_get_pte_flags(struct amdgpu_device *adev, struct svm_range *prange,
 	pte_flags |= snoop ? AMDGPU_PTE_SNOOPED : 0;
 
 	pte_flags |= amdgpu_gem_va_map_flags(adev, mapping_flags);
-
-	pr_debug("svms 0x%p [0x%lx 0x%lx] vram %d PTE 0x%llx mapping 0x%x\n",
-		 prange->svms, prange->start, prange->last,
-		 (domain == SVM_RANGE_VRAM_DOMAIN) ? 1:0, pte_flags, mapping_flags);
-
 	return pte_flags;
 }
 
@@ -1156,7 +1155,8 @@ svm_range_unmap_from_gpus(struct svm_range *prange, unsigned long start,
 
 static int
 svm_range_map_to_gpu(struct amdgpu_device *adev, struct amdgpu_vm *vm,
-		     struct svm_range *prange, dma_addr_t *dma_addr,
+		     struct svm_range *prange, unsigned long offset,
+		     unsigned long npages, bool readonly, dma_addr_t *dma_addr,
 		     struct amdgpu_device *bo_adev, struct dma_fence **fence)
 {
 	struct amdgpu_bo_va bo_va;
@@ -1167,14 +1167,15 @@ svm_range_map_to_gpu(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	int r = 0;
 	int64_t i;
 
-	pr_debug("svms 0x%p [0x%lx 0x%lx]\n", prange->svms, prange->start,
-		 prange->last);
+	last_start = prange->start + offset;
+
+	pr_debug("svms 0x%p [0x%lx 0x%lx] readonly %d\n", prange->svms,
+		 last_start, last_start + npages - 1, readonly);
 
 	if (prange->svm_bo && prange->ttm_res)
 		bo_va.is_xgmi = amdgpu_xgmi_same_hive(adev, bo_adev);
 
-	last_start = prange->start;
-	for (i = 0; i < prange->npages; i++) {
+	for (i = offset; i < offset + npages; i++) {
 		last_domain = dma_addr[i] & SVM_RANGE_VRAM_DOMAIN;
 		dma_addr[i] &= ~SVM_RANGE_VRAM_DOMAIN;
 		if ((prange->start + i) < prange->last &&
@@ -1183,13 +1184,21 @@ svm_range_map_to_gpu(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 
 		pr_debug("Mapping range [0x%lx 0x%llx] on domain: %s\n",
 			 last_start, prange->start + i, last_domain ? "GPU" : "CPU");
+
 		pte_flags = svm_range_get_pte_flags(adev, prange, last_domain);
-		r = amdgpu_vm_bo_update_mapping(adev, bo_adev, vm, false, false, NULL,
-						last_start,
+		if (readonly)
+			pte_flags &= ~AMDGPU_PTE_WRITEABLE;
+
+		pr_debug("svms 0x%p map [0x%lx 0x%llx] vram %d PTE 0x%llx\n",
+			 prange->svms, last_start, prange->start + i,
+			 (last_domain == SVM_RANGE_VRAM_DOMAIN) ? 1 : 0,
+			 pte_flags);
+
+		r = amdgpu_vm_bo_update_mapping(adev, bo_adev, vm, false, false,
+						NULL, last_start,
 						prange->start + i, pte_flags,
 						last_start - prange->start,
-						NULL,
-						dma_addr,
+						NULL, dma_addr,
 						&vm->last_update,
 						&table_freed);
 		if (r) {
@@ -1220,8 +1229,10 @@ out:
 	return r;
 }
 
-static int svm_range_map_to_gpus(struct svm_range *prange,
-				 unsigned long *bitmap, bool wait)
+static int
+svm_range_map_to_gpus(struct svm_range *prange, unsigned long offset,
+		      unsigned long npages, bool readonly,
+		      unsigned long *bitmap, bool wait)
 {
 	struct kfd_process_device *pdd;
 	struct amdgpu_device *bo_adev;
@@ -1257,7 +1268,8 @@ static int svm_range_map_to_gpus(struct svm_range *prange,
 		}
 
 		r = svm_range_map_to_gpu(adev, drm_priv_to_vm(pdd->drm_priv),
-					 prange, prange->dma_addr[gpuidx],
+					 prange, offset, npages, readonly,
+					 prange->dma_addr[gpuidx],
 					 bo_adev, wait ? &fence : NULL);
 		if (r)
 			break;
@@ -1390,7 +1402,7 @@ static int svm_range_validate_and_map(struct mm_struct *mm,
 				      int32_t gpuidx, bool intr, bool wait)
 {
 	struct svm_validate_context ctx;
-	struct hmm_range *hmm_range;
+	unsigned long start, end, addr;
 	struct kfd_process *p;
 	void *owner;
 	int32_t idx;
@@ -1448,40 +1460,66 @@ static int svm_range_validate_and_map(struct mm_struct *mm,
 			break;
 		}
 	}
-	r = amdgpu_hmm_range_get_pages(&prange->notifier, mm, NULL,
-				       prange->start << PAGE_SHIFT,
-				       prange->npages, &hmm_range,
-				       false, true, owner);
-	if (r) {
-		pr_debug("failed %d to get svm range pages\n", r);
-		goto unreserve_out;
-	}
 
-	r = svm_range_dma_map(prange, ctx.bitmap,
-			      hmm_range->hmm_pfns);
-	if (r) {
-		pr_debug("failed %d to dma map range\n", r);
-		goto unreserve_out;
-	}
+	start = prange->start << PAGE_SHIFT;
+	end = (prange->last + 1) << PAGE_SHIFT;
+	for (addr = start; addr < end && !r; ) {
+		struct hmm_range *hmm_range;
+		struct vm_area_struct *vma;
+		unsigned long next;
+		unsigned long offset;
+		unsigned long npages;
+		bool readonly;
 
-	prange->validated_once = true;
+		vma = find_vma(mm, addr);
+		if (!vma || addr < vma->vm_start) {
+			r = -EFAULT;
+			goto unreserve_out;
+		}
+		readonly = !(vma->vm_flags & VM_WRITE);
 
-	svm_range_lock(prange);
-	if (amdgpu_hmm_range_get_pages_done(hmm_range)) {
-		pr_debug("hmm update the range, need validate again\n");
-		r = -EAGAIN;
-		goto unlock_out;
-	}
-	if (!list_empty(&prange->child_list)) {
-		pr_debug("range split by unmap in parallel, validate again\n");
-		r = -EAGAIN;
-		goto unlock_out;
-	}
+		next = min(vma->vm_end, end);
+		npages = (next - addr) >> PAGE_SHIFT;
+		r = amdgpu_hmm_range_get_pages(&prange->notifier, mm, NULL,
+					       addr, npages, &hmm_range,
+					       readonly, true, owner);
+		if (r) {
+			pr_debug("failed %d to get svm range pages\n", r);
+			goto unreserve_out;
+		}
 
-	r = svm_range_map_to_gpus(prange, ctx.bitmap, wait);
+		offset = (addr - start) >> PAGE_SHIFT;
+		r = svm_range_dma_map(prange, ctx.bitmap, offset, npages,
+				      hmm_range->hmm_pfns);
+		if (r) {
+			pr_debug("failed %d to dma map range\n", r);
+			goto unreserve_out;
+		}
+
+		svm_range_lock(prange);
+		if (amdgpu_hmm_range_get_pages_done(hmm_range)) {
+			pr_debug("hmm update the range, need validate again\n");
+			r = -EAGAIN;
+			goto unlock_out;
+		}
+		if (!list_empty(&prange->child_list)) {
+			pr_debug("range split by unmap in parallel, validate again\n");
+			r = -EAGAIN;
+			goto unlock_out;
+		}
+
+		r = svm_range_map_to_gpus(prange, offset, npages, readonly,
+					  ctx.bitmap, wait);
 
 unlock_out:
-	svm_range_unlock(prange);
+		svm_range_unlock(prange);
+
+		addr = next;
+	}
+
+	if (addr == end)
+		prange->validated_once = true;
+
 unreserve_out:
 	svm_range_unreserve_bos(&ctx);
 
@@ -2400,9 +2438,29 @@ svm_range_count_fault(struct amdgpu_device *adev, struct kfd_process *p,
 		WRITE_ONCE(pdd->faults, pdd->faults + 1);
 }
 
+static bool
+svm_fault_allowed(struct mm_struct *mm, uint64_t addr, bool write_fault)
+{
+	unsigned long requested = VM_READ;
+	struct vm_area_struct *vma;
+
+	if (write_fault)
+		requested |= VM_WRITE;
+
+	vma = find_vma(mm, addr << PAGE_SHIFT);
+	if (!vma || (addr << PAGE_SHIFT) < vma->vm_start) {
+		pr_debug("address 0x%llx VMA is removed\n", addr);
+		return true;
+	}
+
+	pr_debug("requested 0x%lx, vma permission flags 0x%lx\n", requested,
+		vma->vm_flags);
+	return (vma->vm_flags & requested) == requested;
+}
+
 int
 svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid,
-			uint64_t addr)
+			uint64_t addr, bool write_fault)
 {
 	struct mm_struct *mm = NULL;
 	struct svm_range_list *svms;
@@ -2484,6 +2542,13 @@ retry_write_locked:
 		goto out_unlock_range;
 	}
 
+	if (!svm_fault_allowed(mm, addr, write_fault)) {
+		pr_debug("fault addr 0x%llx no %s permission\n", addr,
+			write_fault ? "write" : "read");
+		r = -EPERM;
+		goto out_unlock_range;
+	}
+
 	best_loc = svm_range_best_restore_location(prange, adev, &gpuidx);
 	if (best_loc == -1) {
 		pr_debug("svms %p failed get best restore loc [0x%lx 0x%lx]\n",
@@ -2675,22 +2740,26 @@ svm_range_add(struct kfd_process *p, uint64_t start, uint64_t size,
 	return 0;
 }
 
-/* svm_range_best_prefetch_location - decide the best prefetch location
+/**
+ * svm_range_best_prefetch_location - decide the best prefetch location
  * @prange: svm range structure
  *
  * For xnack off:
- * If range map to single GPU, the best acutal location is prefetch loc, which
+ * If range map to single GPU, the best prefetch location is prefetch_loc, which
  * can be CPU or GPU.
  *
- * If range map to multiple GPUs, only if mGPU connection on xgmi same hive,
- * the best actual location could be prefetch_loc GPU. If mGPU connection on
- * PCIe, the best actual location is always CPU, because GPU cannot access vram
- * of other GPUs, assuming PCIe small bar (large bar support is not upstream).
+ * If range is ACCESS or ACCESS_IN_PLACE by mGPUs, only if mGPU connection on
+ * XGMI same hive, the best prefetch location is prefetch_loc GPU, othervise
+ * the best prefetch location is always CPU, because GPU can not have coherent
+ * mapping VRAM of other GPUs even with large-BAR PCIe connection.
  *
  * For xnack on:
- * The best actual location is prefetch location. If mGPU connection on xgmi
- * same hive, range map to multiple GPUs. Otherwise, the range only map to
- * actual location GPU. Other GPU access vm fault will trigger migration.
+ * If range is not ACCESS_IN_PLACE by mGPUs, the best prefetch location is
+ * prefetch_loc, other GPU access will generate vm fault and trigger migration.
+ *
+ * If range is ACCESS_IN_PLACE by mGPUs, only if mGPU connection on XGMI same
+ * hive, the best prefetch location is prefetch_loc GPU, otherwise the best
+ * prefetch location is always CPU.
  *
  * Context: Process context
  *
@@ -2710,11 +2779,6 @@ svm_range_best_prefetch_location(struct svm_range *prange)
 
 	p = container_of(prange->svms, struct kfd_process, svms);
 
-	/* xnack on */
-	if (p->xnack_enabled)
-		goto out;
-
-	/* xnack off */
 	if (!best_loc || best_loc == KFD_IOCTL_SVM_LOCATION_UNDEFINED)
 		goto out;
 
@@ -2724,8 +2788,12 @@ svm_range_best_prefetch_location(struct svm_range *prange)
 		best_loc = 0;
 		goto out;
 	}
-	bitmap_or(bitmap, prange->bitmap_access, prange->bitmap_aip,
-		  MAX_GPU_INSTANCE);
+
+	if (p->xnack_enabled)
+		bitmap_copy(bitmap, prange->bitmap_aip, MAX_GPU_INSTANCE);
+	else
+		bitmap_or(bitmap, prange->bitmap_access, prange->bitmap_aip,
+			  MAX_GPU_INSTANCE);
 
 	for_each_set_bit(gpuidx, bitmap, MAX_GPU_INSTANCE) {
 		pdd = kfd_process_device_from_gpuidx(p, gpuidx);
@@ -3019,7 +3087,8 @@ svm_range_get_attr(struct kfd_process *p, uint64_t start, uint64_t size,
 	struct svm_range *prange;
 	uint32_t prefetch_loc = KFD_IOCTL_SVM_LOCATION_UNDEFINED;
 	uint32_t location = KFD_IOCTL_SVM_LOCATION_UNDEFINED;
-	uint32_t flags = 0xffffffff;
+	uint32_t flags_and = 0xffffffff;
+	uint32_t flags_or = 0;
 	int gpuidx;
 	uint32_t i;
 
@@ -3054,12 +3123,12 @@ svm_range_get_attr(struct kfd_process *p, uint64_t start, uint64_t size,
 			get_accessible = true;
 			break;
 		case KFD_IOCTL_SVM_ATTR_SET_FLAGS:
+		case KFD_IOCTL_SVM_ATTR_CLR_FLAGS:
 			get_flags = true;
 			break;
 		case KFD_IOCTL_SVM_ATTR_GRANULARITY:
 			get_granularity = true;
 			break;
-		case KFD_IOCTL_SVM_ATTR_CLR_FLAGS:
 		case KFD_IOCTL_SVM_ATTR_ACCESS_IN_PLACE:
 		case KFD_IOCTL_SVM_ATTR_NO_ACCESS:
 			fallthrough;
@@ -3077,7 +3146,8 @@ svm_range_get_attr(struct kfd_process *p, uint64_t start, uint64_t size,
 	if (!node) {
 		pr_debug("range attrs not found return default values\n");
 		svm_range_set_default_attributes(&location, &prefetch_loc,
-						 &granularity, &flags);
+						 &granularity, &flags_and);
+		flags_or = flags_and;
 		if (p->xnack_enabled)
 			bitmap_copy(bitmap_access, svms->bitmap_supported,
 				    MAX_GPU_INSTANCE);
@@ -3123,8 +3193,10 @@ svm_range_get_attr(struct kfd_process *p, uint64_t start, uint64_t size,
 			bitmap_and(bitmap_aip, bitmap_aip,
 				   prange->bitmap_aip, MAX_GPU_INSTANCE);
 		}
-		if (get_flags)
-			flags &= prange->flags;
+		if (get_flags) {
+			flags_and &= prange->flags;
+			flags_or |= prange->flags;
+		}
 
 		if (get_granularity && prange->granularity < granularity)
 			granularity = prange->granularity;
@@ -3158,7 +3230,10 @@ fill_values:
 				attrs[i].type = KFD_IOCTL_SVM_ATTR_NO_ACCESS;
 			break;
 		case KFD_IOCTL_SVM_ATTR_SET_FLAGS:
-			attrs[i].value = flags;
+			attrs[i].value = flags_and;
+			break;
+		case KFD_IOCTL_SVM_ATTR_CLR_FLAGS:
+			attrs[i].value = ~flags_or;
 			break;
 		case KFD_IOCTL_SVM_ATTR_GRANULARITY:
 			attrs[i].value = (uint32_t)granularity;
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.h b/drivers/gpu/drm/amd/amdkfd/kfd_svm.h
index 3fc1fd8b4fbc..c6ec55354c7b 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.h
@@ -175,7 +175,7 @@ int svm_range_split_by_granularity(struct kfd_process *p, struct mm_struct *mm,
 			       unsigned long addr, struct svm_range *parent,
 			       struct svm_range *prange);
 int svm_range_restore_pages(struct amdgpu_device *adev,
-			    unsigned int pasid, uint64_t addr);
+			    unsigned int pasid, uint64_t addr, bool write_fault);
 int svm_range_schedule_evict_svm_bo(struct amdgpu_amdkfd_fence *fence);
 void svm_range_add_list_work(struct svm_range_list *svms,
 			     struct svm_range *prange, struct mm_struct *mm,
@@ -209,7 +209,8 @@ static inline void svm_range_list_fini(struct kfd_process *p)
 }
 
 static inline int svm_range_restore_pages(struct amdgpu_device *adev,
-					  unsigned int pasid, uint64_t addr)
+					  unsigned int pasid, uint64_t addr,
+					  bool write_fault)
 {
 	return -EFAULT;
 }
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
index b1ce072aa20b..98cca5f2b27f 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c
@@ -478,6 +478,8 @@ static ssize_t node_show(struct kobject *kobj, struct attribute *attr,
 			      dev->node_props.simd_per_cu);
 	sysfs_show_32bit_prop(buffer, offs, "max_slots_scratch_cu",
 			      dev->node_props.max_slots_scratch_cu);
+	sysfs_show_32bit_prop(buffer, offs, "gfx_target_version",
+			      dev->node_props.gfx_target_version);
 	sysfs_show_32bit_prop(buffer, offs, "vendor_id",
 			      dev->node_props.vendor_id);
 	sysfs_show_32bit_prop(buffer, offs, "device_id",
@@ -1360,6 +1362,7 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
 	dev->node_props.simd_arrays_per_engine =
 		cu_info.num_shader_arrays_per_engine;
 
+	dev->node_props.gfx_target_version = gpu->device_info->gfx_target_version;
 	dev->node_props.vendor_id = gpu->pdev->vendor;
 	dev->node_props.device_id = gpu->pdev->device;
 	dev->node_props.capability |=
@@ -1424,6 +1427,7 @@ int kfd_topology_add_device(struct kfd_dev *gpu)
 	case CHIP_DIMGREY_CAVEFISH:
 	case CHIP_BEIGE_GOBY:
 	case CHIP_YELLOW_CARP:
+	case CHIP_CYAN_SKILLFISH:
 		dev->node_props.capability |= ((HSA_CAP_DOORBELL_TYPE_2_0 <<
 			HSA_CAP_DOORBELL_TYPE_TOTALBITS_SHIFT) &
 			HSA_CAP_DOORBELL_TYPE_TOTALBITS_MASK);
@@ -1630,7 +1634,7 @@ int kfd_debugfs_rls_by_device(struct seq_file *m, void *data)
 		}
 
 		seq_printf(m, "Node %u, gpu_id %x:\n", i++, dev->gpu->id);
-		r = pm_debugfs_runlist(m, &dev->gpu->dqm->packets);
+		r = pm_debugfs_runlist(m, &dev->gpu->dqm->packet_mgr);
 		if (r)
 			break;
 	}
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h
index 8b48c6692007..a8db017c9b8e 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h
@@ -78,6 +78,7 @@ struct kfd_node_properties {
 	uint32_t simd_per_cu;
 	uint32_t max_slots_scratch_cu;
 	uint32_t engine_id;
+	uint32_t gfx_target_version;
 	uint32_t vendor_id;
 	uint32_t device_id;
 	uint32_t location_id;