Merge tag 'amd-drm-next-6.5-2023-06-09' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-next-6.5-2023-06-02:

amdgpu:
- SR-IOV fixes
- Warning fixes
- Misc code cleanups and spelling fixes
- DCN 3.2 updates
- Improved DC FAMS support for better power management
- Improved DC SubVP support for better power management
- DCN 3.1.x fixes
- Max IB size query
- DC GPU reset fixes
- RAS updates
- DCN 3.0.x fixes
- S/G display fixes
- CP shadow buffer support
- Implement connector force callback
- Z8 power improvements
- PSP 13.0.10 vbflash support
- Mode2 reset fixes
- Store MQDs in VRAM to improve queue switch latency
- VCN 3.x fixes
- JPEG 3.x fixes
- Enable DC_FP on LoongArch
- GFXOFF fixes
- GC 9.4.3 partition support
- SDMA 4.4.2 partition support
- VCN/JPEG 4.0.3 partition support
- VCN 4.0.3 updates
- NBIO 7.9 updates
- GC 9.4.3 updates
- Take NUMA into account when allocating memory
- Handle NUMA for partitions
- SMU 13.0.6 updates
- GC 9.4.3 RAS updates
- Stop including unused swiotlb.h
- SMU 13.0.7 fixes
- Fix clock output ordering on some APUs
- Clean up DC FPGA code
- GFX9 preemption fixes
- Misc irq fixes
- S0ix fixes
- Add new DRM_AMDGPU_WERROR config parameter to help with CI
- PCIe fix for RDNA2
- kdoc fixes
- Documentation updates

amdkfd:
- Query TTM mem limit rather than hardcoding it
- GC 9.4.3 partition support
- Handle NUMA for partitions

radeon:
- Fix possible double free
- Stop including unused swiotlb.h
- Fix possible division by zero

ttm:
- Add query for TTM mem limit
- Add NUMA awareness to pools
- Export ttm_pool_fini()

UAPI:
- Add new ctx query flag to better handle GPU resets
  Mesa MR: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22290
- Add new interface to query and set shadow buffer for RDNA3
  Mesa MR: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21986
- Add new INFO query for max IB size
  Proposed userspace: https://gitlab.freedesktop.org/bnieuwenhuizen/mesa/-/commits/ib-rejection-v3

amd-drm-next-6.5-2023-06-09:

amdgpu:
- S0ix fixes
- Initial SMU13 Overdrive support
- kdoc fixes
- Misc clode cleanups
- Flexible array fixes
- Display OTG fixes
- SMU 13.0.6 updates
- Revert some broken clock counter updates
- Misc display fixes
- GFX9 preemption fixes
- Add support for newer EEPROM bad page table format
- Add missing radeon secondary id
- Add support for new colorspace KMS API
- CSA fix
- Stable pstate fixes for APUs
- make vbl interface admin only
- Handle PCI accelerator class

amdkfd:
- Add debugger support for gdb

radeon:
- Fix possible UAF

drm:
- Add Colorspace functionality

UAPI:
- Add debugger interface for enabling gdb
  Proposed userspace: https://github.com/ROCm-Developer-Tools/ROCdbgapi/tree/wip-dbgapi
- Add KMS colorspace API
  Discussion: https://lists.freedesktop.org/archives/dri-devel/2023-June/408128.html

From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230609174817.7764-1-alexander.deucher@amd.com

1 files changed, 186 insertions, 117 deletions

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
index 96a138a39515..5ff1a5a89d96 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_svm.c
@@ -170,12 +170,11 @@ svm_range_dma_map_dev(struct amdgpu_device *adev, struct svm_range *prange,
 
 		page = hmm_pfn_to_page(hmm_pfns[i]);
 		if (is_zone_device_page(page)) {
-			struct amdgpu_device *bo_adev =
-					amdgpu_ttm_adev(prange->svm_bo->bo->tbo.bdev);
+			struct amdgpu_device *bo_adev = prange->svm_bo->node->adev;
 
 			addr[i] = (hmm_pfns[i] << PAGE_SHIFT) +
 				   bo_adev->vm_manager.vram_base_offset -
-				   bo_adev->kfd.dev->pgmap.range.start;
+				   bo_adev->kfd.pgmap.range.start;
 			addr[i] |= SVM_RANGE_VRAM_DOMAIN;
 			pr_debug_ratelimited("vram address: 0x%llx\n", addr[i]);
 			continue;
@@ -281,7 +280,7 @@ static void svm_range_free(struct svm_range *prange, bool update_mem_usage)
 	if (update_mem_usage && !p->xnack_enabled) {
 		pr_debug("unreserve prange 0x%p size: 0x%llx\n", prange, size);
 		amdgpu_amdkfd_unreserve_mem_limit(NULL, size,
-					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 	}
 	mutex_destroy(&prange->lock);
 	mutex_destroy(&prange->migrate_mutex);
@@ -314,7 +313,7 @@ svm_range *svm_range_new(struct svm_range_list *svms, uint64_t start,
 	p = container_of(svms, struct kfd_process, svms);
 	if (!p->xnack_enabled && update_mem_usage &&
 	    amdgpu_amdkfd_reserve_mem_limit(NULL, size << PAGE_SHIFT,
-					    KFD_IOC_ALLOC_MEM_FLAGS_USERPTR)) {
+				    KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0)) {
 		pr_info("SVM mapping failed, exceeds resident system memory limit\n");
 		kfree(prange);
 		return NULL;
@@ -424,10 +423,8 @@ static void svm_range_bo_unref(struct svm_range_bo *svm_bo)
 }
 
 static bool
-svm_range_validate_svm_bo(struct amdgpu_device *adev, struct svm_range *prange)
+svm_range_validate_svm_bo(struct kfd_node *node, struct svm_range *prange)
 {
-	struct amdgpu_device *bo_adev;
-
 	mutex_lock(&prange->lock);
 	if (!prange->svm_bo) {
 		mutex_unlock(&prange->lock);
@@ -440,12 +437,11 @@ svm_range_validate_svm_bo(struct amdgpu_device *adev, struct svm_range *prange)
 	}
 	if (svm_bo_ref_unless_zero(prange->svm_bo)) {
 		/*
-		 * Migrate from GPU to GPU, remove range from source bo_adev
-		 * svm_bo range list, and return false to allocate svm_bo from
-		 * destination adev.
+		 * Migrate from GPU to GPU, remove range from source svm_bo->node
+		 * range list, and return false to allocate svm_bo from destination
+		 * node.
 		 */
-		bo_adev = amdgpu_ttm_adev(prange->svm_bo->bo->tbo.bdev);
-		if (bo_adev != adev) {
+		if (prange->svm_bo->node != node) {
 			mutex_unlock(&prange->lock);
 
 			spin_lock(&prange->svm_bo->list_lock);
@@ -513,7 +509,7 @@ static struct svm_range_bo *svm_range_bo_new(void)
 }
 
 int
-svm_range_vram_node_new(struct amdgpu_device *adev, struct svm_range *prange,
+svm_range_vram_node_new(struct kfd_node *node, struct svm_range *prange,
 			bool clear)
 {
 	struct amdgpu_bo_param bp;
@@ -528,7 +524,7 @@ svm_range_vram_node_new(struct amdgpu_device *adev, struct svm_range *prange,
 	pr_debug("pasid: %x svms 0x%p [0x%lx 0x%lx]\n", p->pasid, prange->svms,
 		 prange->start, prange->last);
 
-	if (svm_range_validate_svm_bo(adev, prange))
+	if (svm_range_validate_svm_bo(node, prange))
 		return 0;
 
 	svm_bo = svm_range_bo_new();
@@ -542,6 +538,7 @@ svm_range_vram_node_new(struct amdgpu_device *adev, struct svm_range *prange,
 		kfree(svm_bo);
 		return -ESRCH;
 	}
+	svm_bo->node = node;
 	svm_bo->eviction_fence =
 		amdgpu_amdkfd_fence_create(dma_fence_context_alloc(1),
 					   mm,
@@ -558,13 +555,20 @@ svm_range_vram_node_new(struct amdgpu_device *adev, struct svm_range *prange,
 	bp.flags |= AMDGPU_GEM_CREATE_DISCARDABLE;
 	bp.type = ttm_bo_type_device;
 	bp.resv = NULL;
+	if (node->xcp)
+		bp.xcp_id_plus1 = node->xcp->id + 1;
 
-	r = amdgpu_bo_create_user(adev, &bp, &ubo);
+	r = amdgpu_bo_create_user(node->adev, &bp, &ubo);
 	if (r) {
 		pr_debug("failed %d to create bo\n", r);
 		goto create_bo_failed;
 	}
 	bo = &ubo->bo;
+
+	pr_debug("alloc bo at offset 0x%lx size 0x%lx on partition %d\n",
+		 bo->tbo.resource->start << PAGE_SHIFT, bp.size,
+		 bp.xcp_id_plus1 - 1);
+
 	r = amdgpu_bo_reserve(bo, true);
 	if (r) {
 		pr_debug("failed %d to reserve bo\n", r);
@@ -617,45 +621,30 @@ void svm_range_vram_node_free(struct svm_range *prange)
 	prange->ttm_res = NULL;
 }
 
-struct amdgpu_device *
-svm_range_get_adev_by_id(struct svm_range *prange, uint32_t gpu_id)
+struct kfd_node *
+svm_range_get_node_by_id(struct svm_range *prange, uint32_t gpu_id)
 {
-	struct kfd_process_device *pdd;
 	struct kfd_process *p;
-	int32_t gpu_idx;
+	struct kfd_process_device *pdd;
 
 	p = container_of(prange->svms, struct kfd_process, svms);
-
-	gpu_idx = kfd_process_gpuidx_from_gpuid(p, gpu_id);
-	if (gpu_idx < 0) {
-		pr_debug("failed to get device by id 0x%x\n", gpu_id);
-		return NULL;
-	}
-	pdd = kfd_process_device_from_gpuidx(p, gpu_idx);
+	pdd = kfd_process_device_data_by_id(p, gpu_id);
 	if (!pdd) {
-		pr_debug("failed to get device by idx 0x%x\n", gpu_idx);
+		pr_debug("failed to get kfd process device by id 0x%x\n", gpu_id);
 		return NULL;
 	}
 
-	return pdd->dev->adev;
+	return pdd->dev;
 }
 
 struct kfd_process_device *
-svm_range_get_pdd_by_adev(struct svm_range *prange, struct amdgpu_device *adev)
+svm_range_get_pdd_by_node(struct svm_range *prange, struct kfd_node *node)
 {
 	struct kfd_process *p;
-	int32_t gpu_idx, gpuid;
-	int r;
 
 	p = container_of(prange->svms, struct kfd_process, svms);
 
-	r = kfd_process_gpuid_from_adev(p, adev, &gpuid, &gpu_idx);
-	if (r) {
-		pr_debug("failed to get device id by adev %p\n", adev);
-		return NULL;
-	}
-
-	return kfd_process_device_from_gpuidx(p, gpu_idx);
+	return kfd_get_process_device_data(node, p);
 }
 
 static int svm_range_bo_validate(void *param, struct amdgpu_bo *bo)
@@ -735,7 +724,9 @@ svm_range_apply_attrs(struct kfd_process *p, struct svm_range *prange,
 		case KFD_IOCTL_SVM_ATTR_ACCESS:
 		case KFD_IOCTL_SVM_ATTR_ACCESS_IN_PLACE:
 		case KFD_IOCTL_SVM_ATTR_NO_ACCESS:
-			*update_mapping = true;
+			if (!p->xnack_enabled)
+				*update_mapping = true;
+
 			gpuidx = kfd_process_gpuidx_from_gpuid(p,
 							       attrs[i].value);
 			if (attrs[i].type == KFD_IOCTL_SVM_ATTR_NO_ACCESS) {
@@ -818,7 +809,7 @@ svm_range_is_same_attrs(struct kfd_process *p, struct svm_range *prange,
 		}
 	}
 
-	return true;
+	return !prange->is_error_flag;
 }
 
 /**
@@ -1146,31 +1137,39 @@ svm_range_split_by_granularity(struct kfd_process *p, struct mm_struct *mm,
 	}
 	return 0;
 }
+static bool
+svm_nodes_in_same_hive(struct kfd_node *node_a, struct kfd_node *node_b)
+{
+	return (node_a->adev == node_b->adev ||
+		amdgpu_xgmi_same_hive(node_a->adev, node_b->adev));
+}
 
 static uint64_t
-svm_range_get_pte_flags(struct amdgpu_device *adev, struct svm_range *prange,
-			int domain)
+svm_range_get_pte_flags(struct kfd_node *node,
+			struct svm_range *prange, int domain)
 {
-	struct amdgpu_device *bo_adev;
+	struct kfd_node *bo_node;
 	uint32_t flags = prange->flags;
 	uint32_t mapping_flags = 0;
 	uint64_t pte_flags;
 	bool snoop = (domain != SVM_RANGE_VRAM_DOMAIN);
 	bool coherent = flags & KFD_IOCTL_SVM_FLAG_COHERENT;
+	bool uncached = false; /*flags & KFD_IOCTL_SVM_FLAG_UNCACHED;*/
+	unsigned int mtype_local;
 
 	if (domain == SVM_RANGE_VRAM_DOMAIN)
-		bo_adev = amdgpu_ttm_adev(prange->svm_bo->bo->tbo.bdev);
+		bo_node = prange->svm_bo->node;
 
-	switch (KFD_GC_VERSION(adev->kfd.dev)) {
+	switch (node->adev->ip_versions[GC_HWIP][0]) {
 	case IP_VERSION(9, 4, 1):
 		if (domain == SVM_RANGE_VRAM_DOMAIN) {
-			if (bo_adev == adev) {
+			if (bo_node == node) {
 				mapping_flags |= coherent ?
 					AMDGPU_VM_MTYPE_CC : AMDGPU_VM_MTYPE_RW;
 			} else {
 				mapping_flags |= coherent ?
 					AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
-				if (amdgpu_xgmi_same_hive(adev, bo_adev))
+				if (svm_nodes_in_same_hive(node, bo_node))
 					snoop = true;
 			}
 		} else {
@@ -1180,15 +1179,15 @@ svm_range_get_pte_flags(struct amdgpu_device *adev, struct svm_range *prange,
 		break;
 	case IP_VERSION(9, 4, 2):
 		if (domain == SVM_RANGE_VRAM_DOMAIN) {
-			if (bo_adev == adev) {
+			if (bo_node == node) {
 				mapping_flags |= coherent ?
 					AMDGPU_VM_MTYPE_CC : AMDGPU_VM_MTYPE_RW;
-				if (adev->gmc.xgmi.connected_to_cpu)
+				if (node->adev->gmc.xgmi.connected_to_cpu)
 					snoop = true;
 			} else {
 				mapping_flags |= coherent ?
 					AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
-				if (amdgpu_xgmi_same_hive(adev, bo_adev))
+				if (svm_nodes_in_same_hive(node, bo_node))
 					snoop = true;
 			}
 		} else {
@@ -1196,6 +1195,37 @@ svm_range_get_pte_flags(struct amdgpu_device *adev, struct svm_range *prange,
 				AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
 		}
 		break;
+	case IP_VERSION(9, 4, 3):
+		mtype_local = amdgpu_mtype_local == 1 ? AMDGPU_VM_MTYPE_NC :
+			     (amdgpu_mtype_local == 2 ? AMDGPU_VM_MTYPE_CC : AMDGPU_VM_MTYPE_RW);
+		snoop = true;
+		if (uncached) {
+			mapping_flags |= AMDGPU_VM_MTYPE_UC;
+		} else if (domain == SVM_RANGE_VRAM_DOMAIN) {
+			/* local HBM region close to partition */
+			if (bo_node->adev == node->adev &&
+			    (!bo_node->xcp || !node->xcp || bo_node->xcp->mem_id == node->xcp->mem_id))
+				mapping_flags |= mtype_local;
+			/* local HBM region far from partition or remote XGMI GPU */
+			else if (svm_nodes_in_same_hive(bo_node, node))
+				mapping_flags |= AMDGPU_VM_MTYPE_NC;
+			/* PCIe P2P */
+			else
+				mapping_flags |= AMDGPU_VM_MTYPE_UC;
+		/* system memory accessed by the APU */
+		} else if (node->adev->flags & AMD_IS_APU) {
+			/* On NUMA systems, locality is determined per-page
+			 * in amdgpu_gmc_override_vm_pte_flags
+			 */
+			if (num_possible_nodes() <= 1)
+				mapping_flags |= mtype_local;
+			else
+				mapping_flags |= AMDGPU_VM_MTYPE_NC;
+		/* system memory accessed by the dGPU */
+		} else {
+			mapping_flags |= AMDGPU_VM_MTYPE_UC;
+		}
+		break;
 	default:
 		mapping_flags |= coherent ?
 			AMDGPU_VM_MTYPE_UC : AMDGPU_VM_MTYPE_NC;
@@ -1212,7 +1242,7 @@ svm_range_get_pte_flags(struct amdgpu_device *adev, struct svm_range *prange,
 	pte_flags |= (domain == SVM_RANGE_VRAM_DOMAIN) ? 0 : AMDGPU_PTE_SYSTEM;
 	pte_flags |= snoop ? AMDGPU_PTE_SNOOPED : 0;
 
-	pte_flags |= amdgpu_gem_va_map_flags(adev, mapping_flags);
+	pte_flags |= amdgpu_gem_va_map_flags(node->adev, mapping_flags);
 	return pte_flags;
 }
 
@@ -1319,7 +1349,7 @@ svm_range_map_to_gpu(struct kfd_process_device *pdd, struct svm_range *prange,
 		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);
+		pte_flags = svm_range_get_pte_flags(pdd->dev, prange, last_domain);
 		if (readonly)
 			pte_flags &= ~AMDGPU_PTE_WRITEABLE;
 
@@ -1328,6 +1358,10 @@ svm_range_map_to_gpu(struct kfd_process_device *pdd, struct svm_range *prange,
 			 (last_domain == SVM_RANGE_VRAM_DOMAIN) ? 1 : 0,
 			 pte_flags);
 
+		/* For dGPU mode, we use same vm_manager to allocate VRAM for
+		 * different memory partition based on fpfn/lpfn, we should use
+		 * same vm_manager.vram_base_offset regardless memory partition.
+		 */
 		r = amdgpu_vm_update_range(adev, vm, false, false, flush_tlb, NULL,
 					   last_start, prange->start + i,
 					   pte_flags,
@@ -1365,16 +1399,14 @@ svm_range_map_to_gpus(struct svm_range *prange, unsigned long offset,
 		      unsigned long *bitmap, bool wait, bool flush_tlb)
 {
 	struct kfd_process_device *pdd;
-	struct amdgpu_device *bo_adev;
+	struct amdgpu_device *bo_adev = NULL;
 	struct kfd_process *p;
 	struct dma_fence *fence = NULL;
 	uint32_t gpuidx;
 	int r = 0;
 
 	if (prange->svm_bo && prange->ttm_res)
-		bo_adev = amdgpu_ttm_adev(prange->svm_bo->bo->tbo.bdev);
-	else
-		bo_adev = NULL;
+		bo_adev = prange->svm_bo->node->adev;
 
 	p = container_of(prange->svms, struct kfd_process, svms);
 	for_each_set_bit(gpuidx, bitmap, MAX_GPU_INSTANCE) {
@@ -1522,48 +1554,54 @@ static int svm_range_validate_and_map(struct mm_struct *mm,
 				      struct svm_range *prange, int32_t gpuidx,
 				      bool intr, bool wait, bool flush_tlb)
 {
-	struct svm_validate_context ctx;
+	struct svm_validate_context *ctx;
 	unsigned long start, end, addr;
 	struct kfd_process *p;
 	void *owner;
 	int32_t idx;
 	int r = 0;
 
-	ctx.process = container_of(prange->svms, struct kfd_process, svms);
-	ctx.prange = prange;
-	ctx.intr = intr;
+	ctx = kzalloc(sizeof(struct svm_validate_context), GFP_KERNEL);
+	if (!ctx)
+		return -ENOMEM;
+	ctx->process = container_of(prange->svms, struct kfd_process, svms);
+	ctx->prange = prange;
+	ctx->intr = intr;
 
 	if (gpuidx < MAX_GPU_INSTANCE) {
-		bitmap_zero(ctx.bitmap, MAX_GPU_INSTANCE);
-		bitmap_set(ctx.bitmap, gpuidx, 1);
-	} else if (ctx.process->xnack_enabled) {
-		bitmap_copy(ctx.bitmap, prange->bitmap_aip, MAX_GPU_INSTANCE);
+		bitmap_zero(ctx->bitmap, MAX_GPU_INSTANCE);
+		bitmap_set(ctx->bitmap, gpuidx, 1);
+	} else if (ctx->process->xnack_enabled) {
+		bitmap_copy(ctx->bitmap, prange->bitmap_aip, MAX_GPU_INSTANCE);
 
 		/* If prefetch range to GPU, or GPU retry fault migrate range to
 		 * GPU, which has ACCESS attribute to the range, create mapping
 		 * on that GPU.
 		 */
 		if (prange->actual_loc) {
-			gpuidx = kfd_process_gpuidx_from_gpuid(ctx.process,
+			gpuidx = kfd_process_gpuidx_from_gpuid(ctx->process,
 							prange->actual_loc);
 			if (gpuidx < 0) {
 				WARN_ONCE(1, "failed get device by id 0x%x\n",
 					 prange->actual_loc);
-				return -EINVAL;
+				r = -EINVAL;
+				goto free_ctx;
 			}
 			if (test_bit(gpuidx, prange->bitmap_access))
-				bitmap_set(ctx.bitmap, gpuidx, 1);
+				bitmap_set(ctx->bitmap, gpuidx, 1);
 		}
 	} else {
-		bitmap_or(ctx.bitmap, prange->bitmap_access,
+		bitmap_or(ctx->bitmap, prange->bitmap_access,
 			  prange->bitmap_aip, MAX_GPU_INSTANCE);
 	}
 
-	if (bitmap_empty(ctx.bitmap, MAX_GPU_INSTANCE)) {
-		if (!prange->mapped_to_gpu)
-			return 0;
+	if (bitmap_empty(ctx->bitmap, MAX_GPU_INSTANCE)) {
+		if (!prange->mapped_to_gpu) {
+			r = 0;
+			goto free_ctx;
+		}
 
-		bitmap_copy(ctx.bitmap, prange->bitmap_access, MAX_GPU_INSTANCE);
+		bitmap_copy(ctx->bitmap, prange->bitmap_access, MAX_GPU_INSTANCE);
 	}
 
 	if (prange->actual_loc && !prange->ttm_res) {
@@ -1571,15 +1609,16 @@ static int svm_range_validate_and_map(struct mm_struct *mm,
 		 * svm_migrate_ram_to_vram after allocating a BO.
 		 */
 		WARN_ONCE(1, "VRAM BO missing during validation\n");
-		return -EINVAL;
+		r = -EINVAL;
+		goto free_ctx;
 	}
 
-	svm_range_reserve_bos(&ctx);
+	svm_range_reserve_bos(ctx);
 
 	p = container_of(prange->svms, struct kfd_process, svms);
-	owner = kfd_svm_page_owner(p, find_first_bit(ctx.bitmap,
+	owner = kfd_svm_page_owner(p, find_first_bit(ctx->bitmap,
 						MAX_GPU_INSTANCE));
-	for_each_set_bit(idx, ctx.bitmap, MAX_GPU_INSTANCE) {
+	for_each_set_bit(idx, ctx->bitmap, MAX_GPU_INSTANCE) {
 		if (kfd_svm_page_owner(p, idx) != owner) {
 			owner = NULL;
 			break;
@@ -1616,7 +1655,7 @@ static int svm_range_validate_and_map(struct mm_struct *mm,
 		}
 
 		offset = (addr - start) >> PAGE_SHIFT;
-		r = svm_range_dma_map(prange, ctx.bitmap, offset, npages,
+		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);
@@ -1636,7 +1675,7 @@ static int svm_range_validate_and_map(struct mm_struct *mm,
 		}
 
 		r = svm_range_map_to_gpus(prange, offset, npages, readonly,
-					  ctx.bitmap, wait, flush_tlb);
+					  ctx->bitmap, wait, flush_tlb);
 
 unlock_out:
 		svm_range_unlock(prange);
@@ -1650,11 +1689,15 @@ unlock_out:
 	}
 
 unreserve_out:
-	svm_range_unreserve_bos(&ctx);
+	svm_range_unreserve_bos(ctx);
 
+	prange->is_error_flag = !!r;
 	if (!r)
 		prange->validate_timestamp = ktime_get_boottime();
 
+free_ctx:
+	kfree(ctx);
+
 	return r;
 }
 
@@ -1783,6 +1826,7 @@ out_reschedule:
  * @mm: current process mm_struct
  * @start: starting process queue number
  * @last: last process queue number
+ * @event: mmu notifier event when range is evicted or migrated
  *
  * Stop all queues of the process to ensure GPU doesn't access the memory, then
  * return to let CPU evict the buffer and proceed CPU pagetable update.
@@ -1906,14 +1950,23 @@ void svm_range_set_max_pages(struct amdgpu_device *adev)
 {
 	uint64_t max_pages;
 	uint64_t pages, _pages;
+	uint64_t min_pages = 0;
+	int i, id;
+
+	for (i = 0; i < adev->kfd.dev->num_nodes; i++) {
+		if (adev->kfd.dev->nodes[i]->xcp)
+			id = adev->kfd.dev->nodes[i]->xcp->id;
+		else
+			id = -1;
+		pages = KFD_XCP_MEMORY_SIZE(adev, id) >> 17;
+		pages = clamp(pages, 1ULL << 9, 1ULL << 18);
+		pages = rounddown_pow_of_two(pages);
+		min_pages = min_not_zero(min_pages, pages);
+	}
 
-	/* 1/32 VRAM size in pages */
-	pages = adev->gmc.real_vram_size >> 17;
-	pages = clamp(pages, 1ULL << 9, 1ULL << 18);
-	pages = rounddown_pow_of_two(pages);
 	do {
 		max_pages = READ_ONCE(max_svm_range_pages);
-		_pages = min_not_zero(max_pages, pages);
+		_pages = min_not_zero(max_pages, min_pages);
 	} while (cmpxchg(&max_svm_range_pages, max_pages, _pages) != max_pages);
 }
 
@@ -2507,29 +2560,31 @@ svm_range_from_addr(struct svm_range_list *svms, unsigned long addr,
  */
 static int32_t
 svm_range_best_restore_location(struct svm_range *prange,
-				struct amdgpu_device *adev,
+				struct kfd_node *node,
 				int32_t *gpuidx)
 {
-	struct amdgpu_device *bo_adev, *preferred_adev;
+	struct kfd_node *bo_node, *preferred_node;
 	struct kfd_process *p;
 	uint32_t gpuid;
 	int r;
 
 	p = container_of(prange->svms, struct kfd_process, svms);
 
-	r = kfd_process_gpuid_from_adev(p, adev, &gpuid, gpuidx);
+	r = kfd_process_gpuid_from_node(p, node, &gpuid, gpuidx);
 	if (r < 0) {
 		pr_debug("failed to get gpuid from kgd\n");
 		return -1;
 	}
 
+	if (node->adev->gmc.is_app_apu)
+		return 0;
+
 	if (prange->preferred_loc == gpuid ||
 	    prange->preferred_loc == KFD_IOCTL_SVM_LOCATION_SYSMEM) {
 		return prange->preferred_loc;
 	} else if (prange->preferred_loc != KFD_IOCTL_SVM_LOCATION_UNDEFINED) {
-		preferred_adev = svm_range_get_adev_by_id(prange,
-							prange->preferred_loc);
-		if (amdgpu_xgmi_same_hive(adev, preferred_adev))
+		preferred_node = svm_range_get_node_by_id(prange, prange->preferred_loc);
+		if (preferred_node && svm_nodes_in_same_hive(node, preferred_node))
 			return prange->preferred_loc;
 		/* fall through */
 	}
@@ -2541,8 +2596,8 @@ svm_range_best_restore_location(struct svm_range *prange,
 		if (!prange->actual_loc)
 			return 0;
 
-		bo_adev = svm_range_get_adev_by_id(prange, prange->actual_loc);
-		if (amdgpu_xgmi_same_hive(adev, bo_adev))
+		bo_node = svm_range_get_node_by_id(prange, prange->actual_loc);
+		if (bo_node && svm_nodes_in_same_hive(node, bo_node))
 			return prange->actual_loc;
 		else
 			return 0;
@@ -2659,7 +2714,7 @@ svm_range_check_vm_userptr(struct kfd_process *p, uint64_t start, uint64_t last,
 }
 
 static struct
-svm_range *svm_range_create_unregistered_range(struct amdgpu_device *adev,
+svm_range *svm_range_create_unregistered_range(struct kfd_node *node,
 						struct kfd_process *p,
 						struct mm_struct *mm,
 						int64_t addr)
@@ -2694,7 +2749,7 @@ svm_range *svm_range_create_unregistered_range(struct amdgpu_device *adev,
 		pr_debug("Failed to create prange in address [0x%llx]\n", addr);
 		return NULL;
 	}
-	if (kfd_process_gpuid_from_adev(p, adev, &gpuid, &gpuidx)) {
+	if (kfd_process_gpuid_from_node(p, node, &gpuid, &gpuidx)) {
 		pr_debug("failed to get gpuid from kgd\n");
 		svm_range_free(prange, true);
 		return NULL;
@@ -2748,7 +2803,7 @@ static bool svm_range_skip_recover(struct svm_range *prange)
 }
 
 static void
-svm_range_count_fault(struct amdgpu_device *adev, struct kfd_process *p,
+svm_range_count_fault(struct kfd_node *node, struct kfd_process *p,
 		      int32_t gpuidx)
 {
 	struct kfd_process_device *pdd;
@@ -2761,7 +2816,7 @@ svm_range_count_fault(struct amdgpu_device *adev, struct kfd_process *p,
 		uint32_t gpuid;
 		int r;
 
-		r = kfd_process_gpuid_from_adev(p, adev, &gpuid, &gpuidx);
+		r = kfd_process_gpuid_from_node(p, node, &gpuid, &gpuidx);
 		if (r < 0)
 			return;
 	}
@@ -2789,6 +2844,7 @@ svm_fault_allowed(struct vm_area_struct *vma, bool write_fault)
 
 int
 svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid,
+			uint32_t vmid, uint32_t node_id,
 			uint64_t addr, bool write_fault)
 {
 	struct mm_struct *mm = NULL;
@@ -2796,6 +2852,7 @@ svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid,
 	struct svm_range *prange;
 	struct kfd_process *p;
 	ktime_t timestamp = ktime_get_boottime();
+	struct kfd_node *node;
 	int32_t best_loc;
 	int32_t gpuidx = MAX_GPU_INSTANCE;
 	bool write_locked = false;
@@ -2803,7 +2860,7 @@ svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid,
 	bool migration = false;
 	int r = 0;
 
-	if (!KFD_IS_SVM_API_SUPPORTED(adev->kfd.dev)) {
+	if (!KFD_IS_SVM_API_SUPPORTED(adev)) {
 		pr_debug("device does not support SVM\n");
 		return -EFAULT;
 	}
@@ -2839,6 +2896,13 @@ svm_range_restore_pages(struct amdgpu_device *adev, unsigned int pasid,
 		goto out;
 	}
 
+	node = kfd_node_by_irq_ids(adev, node_id, vmid);
+	if (!node) {
+		pr_debug("kfd node does not exist node_id: %d, vmid: %d\n", node_id,
+			 vmid);
+		r = -EFAULT;
+		goto out;
+	}
 	mmap_read_lock(mm);
 retry_write_locked:
 	mutex_lock(&svms->lock);
@@ -2857,7 +2921,7 @@ retry_write_locked:
 			write_locked = true;
 			goto retry_write_locked;
 		}
-		prange = svm_range_create_unregistered_range(adev, p, mm, addr);
+		prange = svm_range_create_unregistered_range(node, p, mm, addr);
 		if (!prange) {
 			pr_debug("failed to create unregistered range svms 0x%p address [0x%llx]\n",
 				 svms, addr);
@@ -2872,7 +2936,7 @@ retry_write_locked:
 	mutex_lock(&prange->migrate_mutex);
 
 	if (svm_range_skip_recover(prange)) {
-		amdgpu_gmc_filter_faults_remove(adev, addr, pasid);
+		amdgpu_gmc_filter_faults_remove(node->adev, addr, pasid);
 		r = 0;
 		goto out_unlock_range;
 	}
@@ -2903,7 +2967,7 @@ retry_write_locked:
 		goto out_unlock_range;
 	}
 
-	best_loc = svm_range_best_restore_location(prange, adev, &gpuidx);
+	best_loc = svm_range_best_restore_location(prange, node, &gpuidx);
 	if (best_loc == -1) {
 		pr_debug("svms %p failed get best restore loc [0x%lx 0x%lx]\n",
 			 svms, prange->start, prange->last);
@@ -2915,7 +2979,7 @@ retry_write_locked:
 		 svms, prange->start, prange->last, best_loc,
 		 prange->actual_loc);
 
-	kfd_smi_event_page_fault_start(adev->kfd.dev, p->lead_thread->pid, addr,
+	kfd_smi_event_page_fault_start(node, p->lead_thread->pid, addr,
 				       write_fault, timestamp);
 
 	if (prange->actual_loc != best_loc) {
@@ -2953,7 +3017,7 @@ retry_write_locked:
 		pr_debug("failed %d to map svms 0x%p [0x%lx 0x%lx] to gpus\n",
 			 r, svms, prange->start, prange->last);
 
-	kfd_smi_event_page_fault_end(adev->kfd.dev, p->lead_thread->pid, addr,
+	kfd_smi_event_page_fault_end(node, p->lead_thread->pid, addr,
 				     migration);
 
 out_unlock_range:
@@ -2962,7 +3026,7 @@ out_unlock_svms:
 	mutex_unlock(&svms->lock);
 	mmap_read_unlock(mm);
 
-	svm_range_count_fault(adev, p, gpuidx);
+	svm_range_count_fault(node, p, gpuidx);
 
 	mmput(mm);
 out:
@@ -2970,7 +3034,7 @@ out:
 
 	if (r == -EAGAIN) {
 		pr_debug("recover vm fault later\n");
-		amdgpu_gmc_filter_faults_remove(adev, addr, pasid);
+		amdgpu_gmc_filter_faults_remove(node->adev, addr, pasid);
 		r = 0;
 	}
 	return r;
@@ -2994,10 +3058,10 @@ svm_range_switch_xnack_reserve_mem(struct kfd_process *p, bool xnack_enabled)
 			size = (pchild->last - pchild->start + 1) << PAGE_SHIFT;
 			if (xnack_enabled) {
 				amdgpu_amdkfd_unreserve_mem_limit(NULL, size,
-						KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 			} else {
 				r = amdgpu_amdkfd_reserve_mem_limit(NULL, size,
-						KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 				if (r)
 					goto out_unlock;
 				reserved_size += size;
@@ -3007,10 +3071,10 @@ svm_range_switch_xnack_reserve_mem(struct kfd_process *p, bool xnack_enabled)
 		size = (prange->last - prange->start + 1) << PAGE_SHIFT;
 		if (xnack_enabled) {
 			amdgpu_amdkfd_unreserve_mem_limit(NULL, size,
-						KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 		} else {
 			r = amdgpu_amdkfd_reserve_mem_limit(NULL, size,
-						KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 			if (r)
 				goto out_unlock;
 			reserved_size += size;
@@ -3023,7 +3087,7 @@ out_unlock:
 
 	if (r)
 		amdgpu_amdkfd_unreserve_mem_limit(NULL, reserved_size,
-						KFD_IOC_ALLOC_MEM_FLAGS_USERPTR);
+					KFD_IOC_ALLOC_MEM_FLAGS_USERPTR, 0);
 	else
 		/* Change xnack mode must be inside svms lock, to avoid race with
 		 * svm_range_deferred_list_work unreserve memory in parallel.
@@ -3081,7 +3145,7 @@ int svm_range_list_init(struct kfd_process *p)
 	spin_lock_init(&svms->deferred_list_lock);
 
 	for (i = 0; i < p->n_pdds; i++)
-		if (KFD_IS_SVM_API_SUPPORTED(p->pdds[i]->dev))
+		if (KFD_IS_SVM_API_SUPPORTED(p->pdds[i]->dev->adev))
 			bitmap_set(svms->bitmap_supported, i, 1);
 
 	return 0;
@@ -3212,7 +3276,7 @@ svm_range_best_prefetch_location(struct svm_range *prange)
 	DECLARE_BITMAP(bitmap, MAX_GPU_INSTANCE);
 	uint32_t best_loc = prange->prefetch_loc;
 	struct kfd_process_device *pdd;
-	struct amdgpu_device *bo_adev;
+	struct kfd_node *bo_node;
 	struct kfd_process *p;
 	uint32_t gpuidx;
 
@@ -3221,9 +3285,14 @@ svm_range_best_prefetch_location(struct svm_range *prange)
 	if (!best_loc || best_loc == KFD_IOCTL_SVM_LOCATION_UNDEFINED)
 		goto out;
 
-	bo_adev = svm_range_get_adev_by_id(prange, best_loc);
-	if (!bo_adev) {
-		WARN_ONCE(1, "failed to get device by id 0x%x\n", best_loc);
+	bo_node = svm_range_get_node_by_id(prange, best_loc);
+	if (!bo_node) {
+		WARN_ONCE(1, "failed to get valid kfd node at id%x\n", best_loc);
+		best_loc = 0;
+		goto out;
+	}
+
+	if (bo_node->adev->gmc.is_app_apu) {
 		best_loc = 0;
 		goto out;
 	}
@@ -3241,10 +3310,10 @@ svm_range_best_prefetch_location(struct svm_range *prange)
 			continue;
 		}
 
-		if (pdd->dev->adev == bo_adev)
+		if (pdd->dev->adev == bo_node->adev)
 			continue;
 
-		if (!amdgpu_xgmi_same_hive(pdd->dev->adev, bo_adev)) {
+		if (!svm_nodes_in_same_hive(pdd->dev, bo_node)) {
 			best_loc = 0;
 			break;
 		}