Merge tag 'drm-amdkfd-next-2018-05-14' of git://people.freedesktop.org/~gabbayo/linux into drm-next

This is amdkfd pull for 4.18. The major new features are:

- Add support for GFXv9 dGPUs (VEGA)
- Add support for userptr memory mapping

In addition, there are a couple of small fixes and improvements, such as:
- Fix lock handling
- Fix rollback packet in kernel kfd_queue
- Optimize kfd signal handling
- Fix CP hang in APU
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180514070126.GA1827@odedg-x270

MAINTAINERS

@@ -767,12 +767,14 @@ F:	drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
 F:	drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h
 F:	drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c
 F:	drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c
+F:	drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c
 F:	drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c
 F:	drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
 F:	drivers/gpu/drm/amd/amdkfd/
 F:	drivers/gpu/drm/amd/include/cik_structs.h
 F:	drivers/gpu/drm/amd/include/kgd_kfd_interface.h
 F:	drivers/gpu/drm/amd/include/vi_structs.h
+F:	drivers/gpu/drm/amd/include/v9_structs.h
 F:	include/uapi/linux/kfd_ioctl.h
 
 AMD SEATTLE DEVICE TREE SUPPORT

drivers/gpu/drm/amd/amdgpu/Makefile

@@ -130,7 +130,8 @@ amdgpu-y += \
 	 amdgpu_amdkfd.o \
 	 amdgpu_amdkfd_fence.o \
 	 amdgpu_amdkfd_gpuvm.o \
-	 amdgpu_amdkfd_gfx_v8.o
+	 amdgpu_amdkfd_gfx_v8.o \
+	 amdgpu_amdkfd_gfx_v9.o
 
 # add cgs
 amdgpu-y += amdgpu_cgs.o

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c

@@ -92,6 +92,10 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev)
 	case CHIP_POLARIS11:
 		kfd2kgd = amdgpu_amdkfd_gfx_8_0_get_functions();
 		break;
+	case CHIP_VEGA10:
+	case CHIP_RAVEN:
+		kfd2kgd = amdgpu_amdkfd_gfx_9_0_get_functions();
+		break;
 	default:
 		dev_dbg(adev->dev, "kfd not supported on this ASIC\n");
 		return;
@@ -175,6 +179,28 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev)
 				&gpu_resources.doorbell_physical_address,
 				&gpu_resources.doorbell_aperture_size,
 				&gpu_resources.doorbell_start_offset);
+		if (adev->asic_type >= CHIP_VEGA10) {
+			/* On SOC15 the BIF is involved in routing
+			 * doorbells using the low 12 bits of the
+			 * address. Communicate the assignments to
+			 * KFD. KFD uses two doorbell pages per
+			 * process in case of 64-bit doorbells so we
+			 * can use each doorbell assignment twice.
+			 */
+			gpu_resources.sdma_doorbell[0][0] =
+				AMDGPU_DOORBELL64_sDMA_ENGINE0;
+			gpu_resources.sdma_doorbell[0][1] =
+				AMDGPU_DOORBELL64_sDMA_ENGINE0 + 0x200;
+			gpu_resources.sdma_doorbell[1][0] =
+				AMDGPU_DOORBELL64_sDMA_ENGINE1;
+			gpu_resources.sdma_doorbell[1][1] =
+				AMDGPU_DOORBELL64_sDMA_ENGINE1 + 0x200;
+			/* Doorbells 0x0f0-0ff and 0x2f0-2ff are reserved for
+			 * SDMA, IH and VCN. So don't use them for the CP.
+			 */
+			gpu_resources.reserved_doorbell_mask = 0x1f0;
+			gpu_resources.reserved_doorbell_val  = 0x0f0;
+		}
 
 		kgd2kfd->device_init(adev->kfd, &gpu_resources);
 	}

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h

@@ -28,6 +28,7 @@
 #include <linux/types.h>
 #include <linux/mm.h>
 #include <linux/mmu_context.h>
+#include <linux/workqueue.h>
 #include <kgd_kfd_interface.h>
 #include <drm/ttm/ttm_execbuf_util.h>
 #include "amdgpu_sync.h"
@@ -59,7 +60,9 @@ struct kgd_mem {
 
 	uint32_t mapping_flags;
 
+	atomic_t invalid;
 	struct amdkfd_process_info *process_info;
+	struct page **user_pages;
 
 	struct amdgpu_sync sync;
 
@@ -84,6 +87,9 @@ struct amdkfd_process_info {
 	struct list_head vm_list_head;
 	/* List head for all KFD BOs that belong to a KFD process. */
 	struct list_head kfd_bo_list;
+	/* List of userptr BOs that are valid or invalid */
+	struct list_head userptr_valid_list;
+	struct list_head userptr_inval_list;
 	/* Lock to protect kfd_bo_list */
 	struct mutex lock;
 
@@ -91,6 +97,11 @@ struct amdkfd_process_info {
 	unsigned int n_vms;
 	/* Eviction Fence */
 	struct amdgpu_amdkfd_fence *eviction_fence;
+
+	/* MMU-notifier related fields */
+	atomic_t evicted_bos;
+	struct delayed_work restore_userptr_work;
+	struct pid *pid;
 };
 
 int amdgpu_amdkfd_init(void);
@@ -104,12 +115,14 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev);
 void amdgpu_amdkfd_device_init(struct amdgpu_device *adev);
 void amdgpu_amdkfd_device_fini(struct amdgpu_device *adev);
 
+int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem, struct mm_struct *mm);
 int amdgpu_amdkfd_submit_ib(struct kgd_dev *kgd, enum kgd_engine_type engine,
 				uint32_t vmid, uint64_t gpu_addr,
 				uint32_t *ib_cmd, uint32_t ib_len);
 
 struct kfd2kgd_calls *amdgpu_amdkfd_gfx_7_get_functions(void);
 struct kfd2kgd_calls *amdgpu_amdkfd_gfx_8_0_get_functions(void);
+struct kfd2kgd_calls *amdgpu_amdkfd_gfx_9_0_get_functions(void);
 
 bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid);
 

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c

@@ -98,8 +98,6 @@ static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
 static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid,
 					unsigned int vmid);
 
-static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id,
-				uint32_t hpd_size, uint64_t hpd_gpu_addr);
 static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id);
 static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 			uint32_t queue_id, uint32_t __user *wptr,
@@ -183,7 +181,6 @@ static const struct kfd2kgd_calls kfd2kgd = {
 	.free_pasid = amdgpu_pasid_free,
 	.program_sh_mem_settings = kgd_program_sh_mem_settings,
 	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
-	.init_pipeline = kgd_init_pipeline,
 	.init_interrupts = kgd_init_interrupts,
 	.hqd_load = kgd_hqd_load,
 	.hqd_sdma_load = kgd_hqd_sdma_load,
@@ -309,13 +306,6 @@ static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid,
 	return 0;
 }
 
-static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id,
-				uint32_t hpd_size, uint64_t hpd_gpu_addr)
-{
-	/* amdgpu owns the per-pipe state */
-	return 0;
-}
-
 static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);

drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c

@@ -57,8 +57,6 @@ static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid,
 		uint32_t sh_mem_bases);
 static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid,
 		unsigned int vmid);
-static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id,
-		uint32_t hpd_size, uint64_t hpd_gpu_addr);
 static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id);
 static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id,
 			uint32_t queue_id, uint32_t __user *wptr,
@@ -141,7 +139,6 @@ static const struct kfd2kgd_calls kfd2kgd = {
 	.free_pasid = amdgpu_pasid_free,
 	.program_sh_mem_settings = kgd_program_sh_mem_settings,
 	.set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping,
-	.init_pipeline = kgd_init_pipeline,
 	.init_interrupts = kgd_init_interrupts,
 	.hqd_load = kgd_hqd_load,
 	.hqd_sdma_load = kgd_hqd_sdma_load,
@@ -270,13 +267,6 @@ static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid,
 	return 0;
 }
 
-static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id,
-				uint32_t hpd_size, uint64_t hpd_gpu_addr)
-{
-	/* amdgpu owns the per-pipe state */
-	return 0;
-}
-
 static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id)
 {
 	struct amdgpu_device *adev = get_amdgpu_device(kgd);

drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c

@@ -536,7 +536,7 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
 	if (p->bo_list) {
 		amdgpu_bo_list_get_list(p->bo_list, &p->validated);
 		if (p->bo_list->first_userptr != p->bo_list->num_entries)
-			p->mn = amdgpu_mn_get(p->adev);
+			p->mn = amdgpu_mn_get(p->adev, AMDGPU_MN_TYPE_GFX);
 	}
 
 	INIT_LIST_HEAD(&duplicates);

drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c

@@ -36,12 +36,14 @@
 #include <drm/drm.h>
 
 #include "amdgpu.h"
+#include "amdgpu_amdkfd.h"
 
 struct amdgpu_mn {
 	/* constant after initialisation */
 	struct amdgpu_device	*adev;
 	struct mm_struct	*mm;
 	struct mmu_notifier	mn;
+	enum amdgpu_mn_type	type;
 
 	/* only used on destruction */
 	struct work_struct	work;
@@ -185,7 +187,7 @@ static void amdgpu_mn_invalidate_node(struct amdgpu_mn_node *node,
 }
 
 /**
- * amdgpu_mn_invalidate_range_start - callback to notify about mm change
+ * amdgpu_mn_invalidate_range_start_gfx - callback to notify about mm change
  *
  * @mn: our notifier
  * @mn: the mm this callback is about
@@ -195,10 +197,10 @@ static void amdgpu_mn_invalidate_node(struct amdgpu_mn_node *node,
  * We block for all BOs between start and end to be idle and
  * unmap them by move them into system domain again.
  */
-static void amdgpu_mn_invalidate_range_start(struct mmu_notifier *mn,
-					     struct mm_struct *mm,
-					     unsigned long start,
-					     unsigned long end)
+static void amdgpu_mn_invalidate_range_start_gfx(struct mmu_notifier *mn,
+						 struct mm_struct *mm,
+						 unsigned long start,
+						 unsigned long end)
 {
 	struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn);
 	struct interval_tree_node *it;
@@ -219,6 +221,49 @@ static void amdgpu_mn_invalidate_range_start(struct mmu_notifier *mn,
 	}
 }
 
+/**
+ * amdgpu_mn_invalidate_range_start_hsa - callback to notify about mm change
+ *
+ * @mn: our notifier
+ * @mn: the mm this callback is about
+ * @start: start of updated range
+ * @end: end of updated range
+ *
+ * We temporarily evict all BOs between start and end. This
+ * necessitates evicting all user-mode queues of the process. The BOs
+ * are restorted in amdgpu_mn_invalidate_range_end_hsa.
+ */
+static void amdgpu_mn_invalidate_range_start_hsa(struct mmu_notifier *mn,
+						 struct mm_struct *mm,
+						 unsigned long start,
+						 unsigned long end)
+{
+	struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn);
+	struct interval_tree_node *it;
+
+	/* notification is exclusive, but interval is inclusive */
+	end -= 1;
+
+	amdgpu_mn_read_lock(rmn);
+
+	it = interval_tree_iter_first(&rmn->objects, start, end);
+	while (it) {
+		struct amdgpu_mn_node *node;
+		struct amdgpu_bo *bo;
+
+		node = container_of(it, struct amdgpu_mn_node, it);
+		it = interval_tree_iter_next(it, start, end);
+
+		list_for_each_entry(bo, &node->bos, mn_list) {
+			struct kgd_mem *mem = bo->kfd_bo;
+
+			if (amdgpu_ttm_tt_affect_userptr(bo->tbo.ttm,
+							 start, end))
+				amdgpu_amdkfd_evict_userptr(mem, mm);
+		}
+	}
+}
+
 /**
  * amdgpu_mn_invalidate_range_end - callback to notify about mm change
  *
@@ -239,23 +284,39 @@ static void amdgpu_mn_invalidate_range_end(struct mmu_notifier *mn,
 	amdgpu_mn_read_unlock(rmn);
 }
 
-static const struct mmu_notifier_ops amdgpu_mn_ops = {
-	.release = amdgpu_mn_release,
-	.invalidate_range_start = amdgpu_mn_invalidate_range_start,
-	.invalidate_range_end = amdgpu_mn_invalidate_range_end,
+static const struct mmu_notifier_ops amdgpu_mn_ops[] = {
+	[AMDGPU_MN_TYPE_GFX] = {
+		.release = amdgpu_mn_release,
+		.invalidate_range_start = amdgpu_mn_invalidate_range_start_gfx,
+		.invalidate_range_end = amdgpu_mn_invalidate_range_end,
+	},
+	[AMDGPU_MN_TYPE_HSA] = {
+		.release = amdgpu_mn_release,
+		.invalidate_range_start = amdgpu_mn_invalidate_range_start_hsa,
+		.invalidate_range_end = amdgpu_mn_invalidate_range_end,
+	},
 };
 
+/* Low bits of any reasonable mm pointer will be unused due to struct
+ * alignment. Use these bits to make a unique key from the mm pointer
+ * and notifier type.
+ */
+#define AMDGPU_MN_KEY(mm, type) ((unsigned long)(mm) + (type))
+
 /**
  * amdgpu_mn_get - create notifier context
  *
  * @adev: amdgpu device pointer
+ * @type: type of MMU notifier context
  *
  * Creates a notifier context for current->mm.
  */
-struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev)
+struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev,
+				enum amdgpu_mn_type type)
 {
 	struct mm_struct *mm = current->mm;
 	struct amdgpu_mn *rmn;
+	unsigned long key = AMDGPU_MN_KEY(mm, type);
 	int r;
 
 	mutex_lock(&adev->mn_lock);
@@ -264,8 +325,8 @@ struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev)
 		return ERR_PTR(-EINTR);
 	}
 
-	hash_for_each_possible(adev->mn_hash, rmn, node, (unsigned long)mm)
-		if (rmn->mm == mm)
+	hash_for_each_possible(adev->mn_hash, rmn, node, key)
+		if (AMDGPU_MN_KEY(rmn->mm, rmn->type) == key)
 			goto release_locks;
 
 	rmn = kzalloc(sizeof(*rmn), GFP_KERNEL);
@@ -276,8 +337,9 @@ struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev)
 
 	rmn->adev = adev;
 	rmn->mm = mm;
-	rmn->mn.ops = &amdgpu_mn_ops;
 	init_rwsem(&rmn->lock);
+	rmn->type = type;
+	rmn->mn.ops = &amdgpu_mn_ops[type];
 	rmn->objects = RB_ROOT_CACHED;
 	mutex_init(&rmn->read_lock);
 	atomic_set(&rmn->recursion, 0);
@@ -286,7 +348,7 @@ struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev)
 	if (r)
 		goto free_rmn;
 
-	hash_add(adev->mn_hash, &rmn->node, (unsigned long)mm);
+	hash_add(adev->mn_hash, &rmn->node, AMDGPU_MN_KEY(mm, type));
 
 release_locks:
 	up_write(&mm->mmap_sem);
@@ -315,15 +377,21 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
 {
 	unsigned long end = addr + amdgpu_bo_size(bo) - 1;
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
+	enum amdgpu_mn_type type =
+		bo->kfd_bo ? AMDGPU_MN_TYPE_HSA : AMDGPU_MN_TYPE_GFX;
 	struct amdgpu_mn *rmn;
-	struct amdgpu_mn_node *node = NULL;
+	struct amdgpu_mn_node *node = NULL, *new_node;
 	struct list_head bos;
 	struct interval_tree_node *it;
 
-	rmn = amdgpu_mn_get(adev);
+	rmn = amdgpu_mn_get(adev, type);
 	if (IS_ERR(rmn))
 		return PTR_ERR(rmn);
 
+	new_node = kmalloc(sizeof(*new_node), GFP_KERNEL);
+	if (!new_node)
+		return -ENOMEM;
+
 	INIT_LIST_HEAD(&bos);
 
 	down_write(&rmn->lock);
@@ -337,13 +405,10 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr)
 		list_splice(&node->bos, &bos);
 	}
 
-	if (!node) {
-		node = kmalloc(sizeof(struct amdgpu_mn_node), GFP_KERNEL);
-		if (!node) {
-			up_write(&rmn->lock);
-			return -ENOMEM;
-		}
-	}
+	if (!node)
+		node = new_node;
+	else
+		kfree(new_node);
 
 	bo->mn = rmn;
 

drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h

@@ -29,16 +29,23 @@
  */
 struct amdgpu_mn;
 
+enum amdgpu_mn_type {
+	AMDGPU_MN_TYPE_GFX,
+	AMDGPU_MN_TYPE_HSA,
+};
+
 #if defined(CONFIG_MMU_NOTIFIER)
 void amdgpu_mn_lock(struct amdgpu_mn *mn);
 void amdgpu_mn_unlock(struct amdgpu_mn *mn);
-struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev);
+struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev,
+				enum amdgpu_mn_type type);
 int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr);
 void amdgpu_mn_unregister(struct amdgpu_bo *bo);
 #else
 static inline void amdgpu_mn_lock(struct amdgpu_mn *mn) {}
 static inline void amdgpu_mn_unlock(struct amdgpu_mn *mn) {}
-static inline struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev)
+static inline struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev,
+					      enum amdgpu_mn_type type)
 {
 	return NULL;
 }

drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

@@ -695,7 +695,7 @@ struct amdgpu_ttm_tt {
 	struct ttm_dma_tt	ttm;
 	u64			offset;
 	uint64_t		userptr;
-	struct mm_struct	*usermm;
+	struct task_struct	*usertask;
 	uint32_t		userflags;
 	spinlock_t              guptasklock;
 	struct list_head        guptasks;
@@ -706,14 +706,18 @@ struct amdgpu_ttm_tt {
 int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages)
 {
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
+	struct mm_struct *mm = gtt->usertask->mm;
 	unsigned int flags = 0;
 	unsigned pinned = 0;
 	int r;
 
+	if (!mm) /* Happens during process shutdown */
+		return -ESRCH;
+
 	if (!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY))
 		flags |= FOLL_WRITE;
 
-	down_read(&current->mm->mmap_sem);
+	down_read(&mm->mmap_sem);
 
 	if (gtt->userflags & AMDGPU_GEM_USERPTR_ANONONLY) {
 		/* check that we only use anonymous memory
@@ -721,9 +725,9 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages)
 		unsigned long end = gtt->userptr + ttm->num_pages * PAGE_SIZE;
 		struct vm_area_struct *vma;
 
-		vma = find_vma(gtt->usermm, gtt->userptr);
+		vma = find_vma(mm, gtt->userptr);
 		if (!vma || vma->vm_file || vma->vm_end < end) {
-			up_read(&current->mm->mmap_sem);
+			up_read(&mm->mmap_sem);
 			return -EPERM;
 		}
 	}
@@ -739,7 +743,12 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages)
 		list_add(&guptask.list, &gtt->guptasks);
 		spin_unlock(&gtt->guptasklock);
 
-		r = get_user_pages(userptr, num_pages, flags, p, NULL);
+		if (mm == current->mm)
+			r = get_user_pages(userptr, num_pages, flags, p, NULL);
+		else
+			r = get_user_pages_remote(gtt->usertask,
+					mm, userptr, num_pages,
+					flags, p, NULL, NULL);
 
 		spin_lock(&gtt->guptasklock);
 		list_del(&guptask.list);
@@ -752,12 +761,12 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages)
 
 	} while (pinned < ttm->num_pages);
 
-	up_read(&current->mm->mmap_sem);
+	up_read(&mm->mmap_sem);
 	return 0;
 
 release_pages:
 	release_pages(pages, pinned);
-	up_read(&current->mm->mmap_sem);
+	up_read(&mm->mmap_sem);
 	return r;
 }
 
@@ -978,6 +987,9 @@ static void amdgpu_ttm_backend_destroy(struct ttm_tt *ttm)
 {
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
 
+	if (gtt->usertask)
+		put_task_struct(gtt->usertask);
+
 	ttm_dma_tt_fini(&gtt->ttm);
 	kfree(gtt);
 }
@@ -1079,8 +1091,13 @@ int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr,
 		return -EINVAL;
 
 	gtt->userptr = addr;
-	gtt->usermm = current->mm;
 	gtt->userflags = flags;
+
+	if (gtt->usertask)
+		put_task_struct(gtt->usertask);
+	gtt->usertask = current->group_leader;
+	get_task_struct(gtt->usertask);
+
 	spin_lock_init(&gtt->guptasklock);
 	INIT_LIST_HEAD(&gtt->guptasks);
 	atomic_set(&gtt->mmu_invalidations, 0);
@@ -1096,7 +1113,10 @@ struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm)
 	if (gtt == NULL)
 		return NULL;
 
-	return gtt->usermm;
+	if (gtt->usertask == NULL)
+		return NULL;
+
+	return gtt->usertask->mm;
 }
 
 bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start,

drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c

@@ -4686,6 +4686,7 @@ static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev,
 
 	cu_info->number = active_cu_number;
 	cu_info->ao_cu_mask = ao_cu_mask;
+	cu_info->simd_per_cu = NUM_SIMD_PER_CU;
 
 	return 0;
 }

drivers/gpu/drm/amd/amdgpu/soc15d.h

@@ -268,6 +268,11 @@
 			 * x=1: tmz_end
 			 */
 
+#define	PACKET3_INVALIDATE_TLBS				0x98
+#              define PACKET3_INVALIDATE_TLBS_DST_SEL(x)     ((x) << 0)
+#              define PACKET3_INVALIDATE_TLBS_ALL_HUB(x)     ((x) << 4)
+#              define PACKET3_INVALIDATE_TLBS_PASID(x)       ((x) << 5)
+#              define PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(x)  ((x) << 29)
 #define PACKET3_SET_RESOURCES				0xA0
 /* 1. header
  * 2. CONTROL

drivers/gpu/drm/amd/amdkfd/Makefile

@@ -30,12 +30,14 @@ amdkfd-y	:= kfd_module.o kfd_device.o kfd_chardev.o kfd_topology.o \
 		kfd_pasid.o kfd_doorbell.o kfd_flat_memory.o \
 		kfd_process.o kfd_queue.o kfd_mqd_manager.o \
 		kfd_mqd_manager_cik.o kfd_mqd_manager_vi.o \
+		kfd_mqd_manager_v9.o \
 		kfd_kernel_queue.o kfd_kernel_queue_cik.o \
-		kfd_kernel_queue_vi.o kfd_packet_manager.o \
-		kfd_process_queue_manager.o kfd_device_queue_manager.o \
-		kfd_device_queue_manager_cik.o kfd_device_queue_manager_vi.o \
+		kfd_kernel_queue_vi.o kfd_kernel_queue_v9.o \
+		kfd_packet_manager.o kfd_process_queue_manager.o \
+		kfd_device_queue_manager.o kfd_device_queue_manager_cik.o \
+		kfd_device_queue_manager_vi.o kfd_device_queue_manager_v9.o \
 		kfd_interrupt.o kfd_events.o cik_event_interrupt.o \
-		kfd_dbgdev.o kfd_dbgmgr.o kfd_crat.o
+		kfd_int_process_v9.o kfd_dbgdev.o kfd_dbgmgr.o kfd_crat.o
 
 ifneq ($(CONFIG_AMD_IOMMU_V2),)
 amdkfd-y += kfd_iommu.o

drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c

@@ -27,18 +27,28 @@
 static bool cik_event_interrupt_isr(struct kfd_dev *dev,
 					const uint32_t *ih_ring_entry)
 {
-	unsigned int pasid;
 	const struct cik_ih_ring_entry *ihre =
 			(const struct cik_ih_ring_entry *)ih_ring_entry;
+	unsigned int vmid, pasid;
+
+	/* Only handle interrupts from KFD VMIDs */
+	vmid  = (ihre->ring_id & 0x0000ff00) >> 8;
+	if (vmid < dev->vm_info.first_vmid_kfd ||
+	    vmid > dev->vm_info.last_vmid_kfd)
+		return 0;
 
+	/* If there is no valid PASID, it's likely a firmware bug */
 	pasid = (ihre->ring_id & 0xffff0000) >> 16;
+	if (WARN_ONCE(pasid == 0, "FW bug: No PASID in KFD interrupt"))
+		return 0;
 
-	/* Do not process in ISR, just request it to be forwarded to WQ. */
-	return (pasid != 0) &&
-		(ihre->source_id == CIK_INTSRC_CP_END_OF_PIPE ||
+	/* Interrupt types we care about: various signals and faults.
+	 * They will be forwarded to a work queue (see below).
+	 */
+	return ihre->source_id == CIK_INTSRC_CP_END_OF_PIPE ||
 		ihre->source_id == CIK_INTSRC_SDMA_TRAP ||
 		ihre->source_id == CIK_INTSRC_SQ_INTERRUPT_MSG ||
-		ihre->source_id == CIK_INTSRC_CP_BAD_OPCODE);
+		ihre->source_id == CIK_INTSRC_CP_BAD_OPCODE;
 }
 
 static void cik_event_interrupt_wq(struct kfd_dev *dev,

drivers/gpu/drm/amd/amdkfd/cik_regs.h

@@ -33,7 +33,8 @@
 #define	APE1_MTYPE(x)					((x) << 7)
 
 /* valid for both DEFAULT_MTYPE and APE1_MTYPE */
-#define	MTYPE_CACHED					0
+#define	MTYPE_CACHED_NV					0
+#define	MTYPE_CACHED					1
 #define	MTYPE_NONCACHED					3
 
 #define	DEFAULT_CP_HQD_PERSISTENT_STATE			(0x33U << 8)

drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx8.asm

@@ -20,9 +20,12 @@
  * OTHER DEALINGS IN THE SOFTWARE.
  */
 
-#if 0
-HW (VI) source code for CWSR trap handler
-#Version 18 + multiple trap handler
+/* To compile this assembly code:
+ * PROJECT=vi ./sp3 cwsr_trap_handler_gfx8.asm -hex tmp.hex
+ */
+
+/* HW (VI) source code for CWSR trap handler */
+/* Version 18 + multiple trap handler */
 
 // this performance-optimal version was originally from Seven Xu at SRDC
 
@@ -98,6 +101,7 @@ var SWIZZLE_EN                      =   0                   //whether we use swi
 /**************************************************************************/
 var SQ_WAVE_STATUS_INST_ATC_SHIFT  = 23
 var SQ_WAVE_STATUS_INST_ATC_MASK   = 0x00800000
+var SQ_WAVE_STATUS_SPI_PRIO_SHIFT  = 1
 var SQ_WAVE_STATUS_SPI_PRIO_MASK   = 0x00000006
 
 var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT    = 12
@@ -149,7 +153,7 @@ var s_save_spi_init_lo              =   exec_lo
 var s_save_spi_init_hi              =   exec_hi
 
                                                 //tba_lo and tba_hi need to be saved/restored
-var s_save_pc_lo            =   ttmp0           //{TTMP1, TTMP0} = {3??h0,pc_rewind[3:0], HT[0],trapID[7:0], PC[47:0]}
+var s_save_pc_lo            =   ttmp0           //{TTMP1, TTMP0} = {3'h0,pc_rewind[3:0], HT[0],trapID[7:0], PC[47:0]}
 var s_save_pc_hi            =   ttmp1
 var s_save_exec_lo          =   ttmp2
 var s_save_exec_hi          =   ttmp3
@@ -319,6 +323,10 @@ end
         s_sendmsg   sendmsg(MSG_SAVEWAVE)  //send SPI a message and wait for SPI's write to EXEC
     end
 
+    // Set SPI_PRIO=2 to avoid starving instruction fetch in the waves we're waiting for.
+    s_or_b32 s_save_tmp, s_save_status, (2 << SQ_WAVE_STATUS_SPI_PRIO_SHIFT)
+    s_setreg_b32 hwreg(HW_REG_STATUS), s_save_tmp
+
   L_SLEEP:
     s_sleep 0x2                // sleep 1 (64clk) is not enough for 8 waves per SIMD, which will cause SQ hang, since the 7,8th wave could not get arbit to exec inst, while other waves are stuck into the sleep-loop and waiting for wrexec!=0
 
@@ -1007,8 +1015,6 @@ end
 
     s_waitcnt       lgkmcnt(0)                                                                                      //from now on, it is safe to restore STATUS and IB_STS
 
-    s_and_b32 s_restore_pc_hi, s_restore_pc_hi, 0x0000ffff      //pc[47:32]        //Do it here in order not to affect STATUS
-
     //for normal save & restore, the saved PC points to the next inst to execute, no adjustment needs to be made, otherwise:
     if ((EMU_RUN_HACK) && (!EMU_RUN_HACK_RESTORE_NORMAL))
         s_add_u32 s_restore_pc_lo, s_restore_pc_lo, 8            //pc[31:0]+8     //two back-to-back s_trap are used (first for save and second for restore)
@@ -1044,6 +1050,7 @@ end
     s_lshr_b32      s_restore_m0, s_restore_m0, SQ_WAVE_STATUS_INST_ATC_SHIFT
     s_setreg_b32    hwreg(HW_REG_IB_STS),   s_restore_tmp
 
+    s_and_b32 s_restore_pc_hi, s_restore_pc_hi, 0x0000ffff      //pc[47:32]        //Do it here in order not to affect STATUS
     s_and_b64    exec, exec, exec  // Restore STATUS.EXECZ, not writable by s_setreg_b32
     s_and_b64    vcc, vcc, vcc  // Restore STATUS.VCCZ, not writable by s_setreg_b32
     s_setreg_b32    hwreg(HW_REG_STATUS),   s_restore_status     // SCC is included, which is changed by previous salu
@@ -1127,258 +1134,3 @@ end
 function get_hwreg_size_bytes
     return 128 //HWREG size 128 bytes
 end
-
-
-#endif
-
-static const uint32_t cwsr_trap_gfx8_hex[] = {
-	0xbf820001, 0xbf820123,
-	0xb8f4f802, 0x89748674,
-	0xb8f5f803, 0x8675ff75,
-	0x00000400, 0xbf850011,
-	0xc00a1e37, 0x00000000,
-	0xbf8c007f, 0x87777978,
-	0xbf840002, 0xb974f802,
-	0xbe801d78, 0xb8f5f803,
-	0x8675ff75, 0x000001ff,
-	0xbf850002, 0x80708470,
-	0x82718071, 0x8671ff71,
-	0x0000ffff, 0xb974f802,
-	0xbe801f70, 0xb8f5f803,
-	0x8675ff75, 0x00000100,
-	0xbf840006, 0xbefa0080,
-	0xb97a0203, 0x8671ff71,
-	0x0000ffff, 0x80f08870,
-	0x82f18071, 0xbefa0080,
-	0xb97a0283, 0xbef60068,
-	0xbef70069, 0xb8fa1c07,
-	0x8e7a9c7a, 0x87717a71,
-	0xb8fa03c7, 0x8e7a9b7a,
-	0x87717a71, 0xb8faf807,
-	0x867aff7a, 0x00007fff,
-	0xb97af807, 0xbef2007e,
-	0xbef3007f, 0xbefe0180,
-	0xbf900004, 0xbf8e0002,
-	0xbf88fffe, 0xbef8007e,
-	0x8679ff7f, 0x0000ffff,
-	0x8779ff79, 0x00040000,
-	0xbefa0080, 0xbefb00ff,
-	0x00807fac, 0x867aff7f,
-	0x08000000, 0x8f7a837a,
-	0x877b7a7b, 0x867aff7f,
-	0x70000000, 0x8f7a817a,
-	0x877b7a7b, 0xbeef007c,
-	0xbeee0080, 0xb8ee2a05,
-	0x806e816e, 0x8e6e8a6e,
-	0xb8fa1605, 0x807a817a,
-	0x8e7a867a, 0x806e7a6e,
-	0xbefa0084, 0xbefa00ff,
-	0x01000000, 0xbefe007c,
-	0xbefc006e, 0xc0611bfc,
-	0x0000007c, 0x806e846e,
-	0xbefc007e, 0xbefe007c,
-	0xbefc006e, 0xc0611c3c,
-	0x0000007c, 0x806e846e,
-	0xbefc007e, 0xbefe007c,
-	0xbefc006e, 0xc0611c7c,
-	0x0000007c, 0x806e846e,
-	0xbefc007e, 0xbefe007c,
-	0xbefc006e, 0xc0611cbc,
-	0x0000007c, 0x806e846e,
-	0xbefc007e, 0xbefe007c,
-	0xbefc006e, 0xc0611cfc,
-	0x0000007c, 0x806e846e,
-	0xbefc007e, 0xbefe007c,
-	0xbefc006e, 0xc0611d3c,
-	0x0000007c, 0x806e846e,
-	0xbefc007e, 0xb8f5f803,
-	0xbefe007c, 0xbefc006e,
-	0xc0611d7c, 0x0000007c,
-	0x806e846e, 0xbefc007e,
-	0xbefe007c, 0xbefc006e,
-	0xc0611dbc, 0x0000007c,
-	0x806e846e, 0xbefc007e,
-	0xbefe007c, 0xbefc006e,
-	0xc0611dfc, 0x0000007c,
-	0x806e846e, 0xbefc007e,
-	0xb8eff801, 0xbefe007c,
-	0xbefc006e, 0xc0611bfc,
-	0x0000007c, 0x806e846e,
-	0xbefc007e, 0xbefe007c,
-	0xbefc006e, 0xc0611b3c,
-	0x0000007c, 0x806e846e,
-	0xbefc007e, 0xbefe007c,
-	0xbefc006e, 0xc0611b7c,
-	0x0000007c, 0x806e846e,
-	0xbefc007e, 0x867aff7f,
-	0x04000000, 0xbef30080,
-	0x8773737a, 0xb8ee2a05,
-	0x806e816e, 0x8e6e8a6e,
-	0xb8f51605, 0x80758175,
-	0x8e758475, 0x8e7a8275,
-	0xbefa00ff, 0x01000000,
-	0xbef60178, 0x80786e78,
-	0x82798079, 0xbefc0080,
-	0xbe802b00, 0xbe822b02,
-	0xbe842b04, 0xbe862b06,
-	0xbe882b08, 0xbe8a2b0a,
-	0xbe8c2b0c, 0xbe8e2b0e,
-	0xc06b003c, 0x00000000,
-	0xc06b013c, 0x00000010,
-	0xc06b023c, 0x00000020,
-	0xc06b033c, 0x00000030,
-	0x8078c078, 0x82798079,
-	0x807c907c, 0xbf0a757c,
-	0xbf85ffeb, 0xbef80176,
-	0xbeee0080, 0xbefe00c1,
-	0xbeff00c1, 0xbefa00ff,
-	0x01000000, 0xe0724000,
-	0x6e1e0000, 0xe0724100,
-	0x6e1e0100, 0xe0724200,
-	0x6e1e0200, 0xe0724300,
-	0x6e1e0300, 0xbefe00c1,
-	0xbeff00c1, 0xb8f54306,
-	0x8675c175, 0xbf84002c,
-	0xbf8a0000, 0x867aff73,
-	0x04000000, 0xbf840028,
-	0x8e758675, 0x8e758275,
-	0xbefa0075, 0xb8ee2a05,
-	0x806e816e, 0x8e6e8a6e,
-	0xb8fa1605, 0x807a817a,
-	0x8e7a867a, 0x806e7a6e,
-	0x806eff6e, 0x00000080,
-	0xbefa00ff, 0x01000000,
-	0xbefc0080, 0xd28c0002,
-	0x000100c1, 0xd28d0003,
-	0x000204c1, 0xd1060002,
-	0x00011103, 0x7e0602ff,
-	0x00000200, 0xbefc00ff,
-	0x00010000, 0xbe80007b,
-	0x867bff7b, 0xff7fffff,
-	0x877bff7b, 0x00058000,
-	0xd8ec0000, 0x00000002,
-	0xbf8c007f, 0xe0765000,
-	0x6e1e0002, 0x32040702,
-	0xd0c9006a, 0x0000eb02,
-	0xbf87fff7, 0xbefb0000,
-	0xbeee00ff, 0x00000400,
-	0xbefe00c1, 0xbeff00c1,
-	0xb8f52a05, 0x80758175,
-	0x8e758275, 0x8e7a8875,
-	0xbefa00ff, 0x01000000,
-	0xbefc0084, 0xbf0a757c,
-	0xbf840015, 0xbf11017c,
-	0x8075ff75, 0x00001000,
-	0x7e000300, 0x7e020301,
-	0x7e040302, 0x7e060303,
-	0xe0724000, 0x6e1e0000,
-	0xe0724100, 0x6e1e0100,
-	0xe0724200, 0x6e1e0200,
-	0xe0724300, 0x6e1e0300,
-	0x807c847c, 0x806eff6e,
-	0x00000400, 0xbf0a757c,
-	0xbf85ffef, 0xbf9c0000,
-	0xbf8200ca, 0xbef8007e,
-	0x8679ff7f, 0x0000ffff,
-	0x8779ff79, 0x00040000,
-	0xbefa0080, 0xbefb00ff,
-	0x00807fac, 0x8676ff7f,
-	0x08000000, 0x8f768376,
-	0x877b767b, 0x8676ff7f,
-	0x70000000, 0x8f768176,
-	0x877b767b, 0x8676ff7f,
-	0x04000000, 0xbf84001e,
-	0xbefe00c1, 0xbeff00c1,
-	0xb8f34306, 0x8673c173,
-	0xbf840019, 0x8e738673,
-	0x8e738273, 0xbefa0073,
-	0xb8f22a05, 0x80728172,
-	0x8e728a72, 0xb8f61605,
-	0x80768176, 0x8e768676,
-	0x80727672, 0x8072ff72,
-	0x00000080, 0xbefa00ff,
-	0x01000000, 0xbefc0080,
-	0xe0510000, 0x721e0000,
-	0xe0510100, 0x721e0000,
-	0x807cff7c, 0x00000200,
-	0x8072ff72, 0x00000200,
-	0xbf0a737c, 0xbf85fff6,
-	0xbef20080, 0xbefe00c1,
-	0xbeff00c1, 0xb8f32a05,
-	0x80738173, 0x8e738273,
-	0x8e7a8873, 0xbefa00ff,
-	0x01000000, 0xbef60072,
-	0x8072ff72, 0x00000400,
-	0xbefc0084, 0xbf11087c,
-	0x8073ff73, 0x00008000,
-	0xe0524000, 0x721e0000,
-	0xe0524100, 0x721e0100,
-	0xe0524200, 0x721e0200,
-	0xe0524300, 0x721e0300,
-	0xbf8c0f70, 0x7e000300,
-	0x7e020301, 0x7e040302,
-	0x7e060303, 0x807c847c,
-	0x8072ff72, 0x00000400,
-	0xbf0a737c, 0xbf85ffee,
-	0xbf9c0000, 0xe0524000,
-	0x761e0000, 0xe0524100,
-	0x761e0100, 0xe0524200,
-	0x761e0200, 0xe0524300,
-	0x761e0300, 0xb8f22a05,
-	0x80728172, 0x8e728a72,
-	0xb8f61605, 0x80768176,
-	0x8e768676, 0x80727672,
-	0x80f2c072, 0xb8f31605,
-	0x80738173, 0x8e738473,
-	0x8e7a8273, 0xbefa00ff,
-	0x01000000, 0xbefc0073,
-	0xc031003c, 0x00000072,
-	0x80f2c072, 0xbf8c007f,
-	0x80fc907c, 0xbe802d00,
-	0xbe822d02, 0xbe842d04,
-	0xbe862d06, 0xbe882d08,
-	0xbe8a2d0a, 0xbe8c2d0c,
-	0xbe8e2d0e, 0xbf06807c,
-	0xbf84fff1, 0xb8f22a05,
-	0x80728172, 0x8e728a72,
-	0xb8f61605, 0x80768176,
-	0x8e768676, 0x80727672,
-	0xbefa0084, 0xbefa00ff,
-	0x01000000, 0xc0211cfc,
-	0x00000072, 0x80728472,
-	0xc0211c3c, 0x00000072,
-	0x80728472, 0xc0211c7c,
-	0x00000072, 0x80728472,
-	0xc0211bbc, 0x00000072,
-	0x80728472, 0xc0211bfc,
-	0x00000072, 0x80728472,
-	0xc0211d3c, 0x00000072,
-	0x80728472, 0xc0211d7c,
-	0x00000072, 0x80728472,
-	0xc0211a3c, 0x00000072,
-	0x80728472, 0xc0211a7c,
-	0x00000072, 0x80728472,
-	0xc0211dfc, 0x00000072,
-	0x80728472, 0xc0211b3c,
-	0x00000072, 0x80728472,
-	0xc0211b7c, 0x00000072,
-	0x80728472, 0xbf8c007f,
-	0x8671ff71, 0x0000ffff,
-	0xbefc0073, 0xbefe006e,
-	0xbeff006f, 0x867375ff,
-	0x000003ff, 0xb9734803,
-	0x867375ff, 0xfffff800,
-	0x8f738b73, 0xb973a2c3,
-	0xb977f801, 0x8673ff71,
-	0xf0000000, 0x8f739c73,
-	0x8e739073, 0xbef60080,
-	0x87767376, 0x8673ff71,
-	0x08000000, 0x8f739b73,
-	0x8e738f73, 0x87767376,
-	0x8673ff74, 0x00800000,
-	0x8f739773, 0xb976f807,
-	0x86fe7e7e, 0x86ea6a6a,
-	0xb974f802, 0xbf8a0000,
-	0x95807370, 0xbf810000,
-};
-

drivers/gpu/drm/amd/amdkfd/kfd_chardev.c

@@ -233,7 +233,7 @@ static int set_queue_properties_from_user(struct queue_properties *q_properties,
 	pr_debug("Queue Size: 0x%llX, %u\n",
 			q_properties->queue_size, args->ring_size);
 
-	pr_debug("Queue r/w Pointers: %p, %p\n",
+	pr_debug("Queue r/w Pointers: %px, %px\n",
 			q_properties->read_ptr,
 			q_properties->write_ptr);
 
@@ -292,8 +292,16 @@ static int kfd_ioctl_create_queue(struct file *filep, struct kfd_process *p,
 
 
 	/* Return gpu_id as doorbell offset for mmap usage */
-	args->doorbell_offset = (KFD_MMAP_DOORBELL_MASK | args->gpu_id);
+	args->doorbell_offset = KFD_MMAP_TYPE_DOORBELL;
+	args->doorbell_offset |= KFD_MMAP_GPU_ID(args->gpu_id);
 	args->doorbell_offset <<= PAGE_SHIFT;
+	if (KFD_IS_SOC15(dev->device_info->asic_family))
+		/* On SOC15 ASICs, doorbell allocation must be
+		 * per-device, and independent from the per-process
+		 * queue_id. Return the doorbell offset within the
+		 * doorbell aperture to user mode.
+		 */
+		args->doorbell_offset |= q_properties.doorbell_off;
 
 	mutex_unlock(&p->mutex);
 
@@ -1296,8 +1304,8 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep,
 		return -EINVAL;
 	}
 
-	devices_arr = kmalloc(args->n_devices * sizeof(*devices_arr),
-			      GFP_KERNEL);
+	devices_arr = kmalloc_array(args->n_devices, sizeof(*devices_arr),
+				    GFP_KERNEL);
 	if (!devices_arr)
 		return -ENOMEM;
 
@@ -1405,8 +1413,8 @@ static int kfd_ioctl_unmap_memory_from_gpu(struct file *filep,
 		return -EINVAL;
 	}
 
-	devices_arr = kmalloc(args->n_devices * sizeof(*devices_arr),
-			      GFP_KERNEL);
+	devices_arr = kmalloc_array(args->n_devices, sizeof(*devices_arr),
+				    GFP_KERNEL);
 	if (!devices_arr)
 		return -ENOMEM;
 
@@ -1645,23 +1653,33 @@ static long kfd_ioctl(struct file *filep, unsigned int cmd, unsigned long arg)
 static int kfd_mmap(struct file *filp, struct vm_area_struct *vma)
 {
 	struct kfd_process *process;
+	struct kfd_dev *dev = NULL;
+	unsigned long vm_pgoff;
+	unsigned int gpu_id;
 
 	process = kfd_get_process(current);
 	if (IS_ERR(process))
 		return PTR_ERR(process);
 
-	if ((vma->vm_pgoff & KFD_MMAP_DOORBELL_MASK) ==
-			KFD_MMAP_DOORBELL_MASK) {
-		vma->vm_pgoff = vma->vm_pgoff ^ KFD_MMAP_DOORBELL_MASK;
-		return kfd_doorbell_mmap(process, vma);
-	} else if ((vma->vm_pgoff & KFD_MMAP_EVENTS_MASK) ==
-			KFD_MMAP_EVENTS_MASK) {
-		vma->vm_pgoff = vma->vm_pgoff ^ KFD_MMAP_EVENTS_MASK;
+	vm_pgoff = vma->vm_pgoff;
+	vma->vm_pgoff = KFD_MMAP_OFFSET_VALUE_GET(vm_pgoff);
+	gpu_id = KFD_MMAP_GPU_ID_GET(vm_pgoff);
+	if (gpu_id)
+		dev = kfd_device_by_id(gpu_id);
+
+	switch (vm_pgoff & KFD_MMAP_TYPE_MASK) {
+	case KFD_MMAP_TYPE_DOORBELL:
+		if (!dev)
+			return -ENODEV;
+		return kfd_doorbell_mmap(dev, process, vma);
+
+	case KFD_MMAP_TYPE_EVENTS:
 		return kfd_event_mmap(process, vma);
-	} else if ((vma->vm_pgoff & KFD_MMAP_RESERVED_MEM_MASK) ==
-			KFD_MMAP_RESERVED_MEM_MASK) {
-		vma->vm_pgoff = vma->vm_pgoff ^ KFD_MMAP_RESERVED_MEM_MASK;
-		return kfd_reserved_mem_mmap(process, vma);
+
+	case KFD_MMAP_TYPE_RESERVED_MEM:
+		if (!dev)
+			return -ENODEV;
+		return kfd_reserved_mem_mmap(dev, process, vma);
 	}
 
 	return -EFAULT;

drivers/gpu/drm/amd/amdkfd/kfd_crat.c

@@ -132,6 +132,9 @@ static struct kfd_gpu_cache_info carrizo_cache_info[] = {
 #define fiji_cache_info  carrizo_cache_info
 #define polaris10_cache_info carrizo_cache_info
 #define polaris11_cache_info carrizo_cache_info
+/* TODO - check & update Vega10 cache details */
+#define vega10_cache_info carrizo_cache_info
+#define raven_cache_info carrizo_cache_info
 
 static void kfd_populated_cu_info_cpu(struct kfd_topology_device *dev,
 		struct crat_subtype_computeunit *cu)
@@ -603,6 +606,14 @@ static int kfd_fill_gpu_cache_info(struct kfd_dev *kdev,
 		pcache_info = polaris11_cache_info;
 		num_of_cache_types = ARRAY_SIZE(polaris11_cache_info);
 		break;
+	case CHIP_VEGA10:
+		pcache_info = vega10_cache_info;
+		num_of_cache_types = ARRAY_SIZE(vega10_cache_info);
+		break;
+	case CHIP_RAVEN:
+		pcache_info = raven_cache_info;
+		num_of_cache_types = ARRAY_SIZE(raven_cache_info);
+		break;
 	default:
 		return -EINVAL;
 	}

drivers/gpu/drm/amd/amdkfd/kfd_device.c

@@ -20,16 +20,13 @@
  * OTHER DEALINGS IN THE SOFTWARE.
  */
 
-#if defined(CONFIG_AMD_IOMMU_V2_MODULE) || defined(CONFIG_AMD_IOMMU_V2)
-#include <linux/amd-iommu.h>
-#endif
 #include <linux/bsearch.h>
 #include <linux/pci.h>
 #include <linux/slab.h>
 #include "kfd_priv.h"
 #include "kfd_device_queue_manager.h"
 #include "kfd_pm4_headers_vi.h"
-#include "cwsr_trap_handler_gfx8.asm"
+#include "cwsr_trap_handler.h"
 #include "kfd_iommu.h"
 
 #define MQD_SIZE_ALIGNED 768
@@ -41,6 +38,7 @@ static const struct kfd_device_info kaveri_device_info = {
 	.max_pasid_bits = 16,
 	/* max num of queues for KV.TODO should be a dynamic value */
 	.max_no_of_hqd	= 24,
+	.doorbell_size  = 4,
 	.ih_ring_entry_size = 4 * sizeof(uint32_t),
 	.event_interrupt_class = &event_interrupt_class_cik,
 	.num_of_watch_points = 4,
@@ -55,6 +53,7 @@ static const struct kfd_device_info carrizo_device_info = {
 	.max_pasid_bits = 16,
 	/* max num of queues for CZ.TODO should be a dynamic value */
 	.max_no_of_hqd	= 24,
+	.doorbell_size  = 4,
 	.ih_ring_entry_size = 4 * sizeof(uint32_t),
 	.event_interrupt_class = &event_interrupt_class_cik,
 	.num_of_watch_points = 4,
@@ -70,6 +69,7 @@ static const struct kfd_device_info hawaii_device_info = {
 	.max_pasid_bits = 16,
 	/* max num of queues for KV.TODO should be a dynamic value */
 	.max_no_of_hqd	= 24,
+	.doorbell_size  = 4,
 	.ih_ring_entry_size = 4 * sizeof(uint32_t),
 	.event_interrupt_class = &event_interrupt_class_cik,
 	.num_of_watch_points = 4,
@@ -83,6 +83,7 @@ static const struct kfd_device_info tonga_device_info = {
 	.asic_family = CHIP_TONGA,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
+	.doorbell_size  = 4,
 	.ih_ring_entry_size = 4 * sizeof(uint32_t),
 	.event_interrupt_class = &event_interrupt_class_cik,
 	.num_of_watch_points = 4,
@@ -96,6 +97,7 @@ static const struct kfd_device_info tonga_vf_device_info = {
 	.asic_family = CHIP_TONGA,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
+	.doorbell_size  = 4,
 	.ih_ring_entry_size = 4 * sizeof(uint32_t),
 	.event_interrupt_class = &event_interrupt_class_cik,
 	.num_of_watch_points = 4,
@@ -109,6 +111,7 @@ static const struct kfd_device_info fiji_device_info = {
 	.asic_family = CHIP_FIJI,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
+	.doorbell_size  = 4,
 	.ih_ring_entry_size = 4 * sizeof(uint32_t),
 	.event_interrupt_class = &event_interrupt_class_cik,
 	.num_of_watch_points = 4,
@@ -122,6 +125,7 @@ static const struct kfd_device_info fiji_vf_device_info = {
 	.asic_family = CHIP_FIJI,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
+	.doorbell_size  = 4,
 	.ih_ring_entry_size = 4 * sizeof(uint32_t),
 	.event_interrupt_class = &event_interrupt_class_cik,
 	.num_of_watch_points = 4,
@@ -136,6 +140,7 @@ static const struct kfd_device_info polaris10_device_info = {
 	.asic_family = CHIP_POLARIS10,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
+	.doorbell_size  = 4,
 	.ih_ring_entry_size = 4 * sizeof(uint32_t),
 	.event_interrupt_class = &event_interrupt_class_cik,
 	.num_of_watch_points = 4,
@@ -149,6 +154,7 @@ static const struct kfd_device_info polaris10_vf_device_info = {
 	.asic_family = CHIP_POLARIS10,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
+	.doorbell_size  = 4,
 	.ih_ring_entry_size = 4 * sizeof(uint32_t),
 	.event_interrupt_class = &event_interrupt_class_cik,
 	.num_of_watch_points = 4,
@@ -162,6 +168,7 @@ static const struct kfd_device_info polaris11_device_info = {
 	.asic_family = CHIP_POLARIS11,
 	.max_pasid_bits = 16,
 	.max_no_of_hqd  = 24,
+	.doorbell_size  = 4,
 	.ih_ring_entry_size = 4 * sizeof(uint32_t),
 	.event_interrupt_class = &event_interrupt_class_cik,
 	.num_of_watch_points = 4,
@@ -171,6 +178,34 @@ static const struct kfd_device_info polaris11_device_info = {
 	.needs_pci_atomics = true,
 };
 
+static const struct kfd_device_info vega10_device_info = {
+	.asic_family = CHIP_VEGA10,
+	.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,
+	.supports_cwsr = true,
+	.needs_iommu_device = false,
+	.needs_pci_atomics = false,
+};
+
+static const struct kfd_device_info vega10_vf_device_info = {
+	.asic_family = CHIP_VEGA10,
+	.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,
+	.supports_cwsr = true,
+	.needs_iommu_device = false,
+	.needs_pci_atomics = false,
+};
+
 
 struct kfd_deviceid {
 	unsigned short did;
@@ -250,6 +285,15 @@ static const struct kfd_deviceid supported_devices[] = {
 	{ 0x67EB, &polaris11_device_info },	/* Polaris11 */
 	{ 0x67EF, &polaris11_device_info },	/* Polaris11 */
 	{ 0x67FF, &polaris11_device_info },	/* Polaris11 */
+	{ 0x6860, &vega10_device_info },	/* Vega10 */
+	{ 0x6861, &vega10_device_info },	/* Vega10 */
+	{ 0x6862, &vega10_device_info },	/* Vega10 */
+	{ 0x6863, &vega10_device_info },	/* Vega10 */
+	{ 0x6864, &vega10_device_info },	/* Vega10 */
+	{ 0x6867, &vega10_device_info },	/* Vega10 */
+	{ 0x6868, &vega10_device_info },	/* Vega10 */
+	{ 0x686C, &vega10_vf_device_info },	/* Vega10  vf*/
+	{ 0x687F, &vega10_device_info },	/* Vega10 */
 };
 
 static int kfd_gtt_sa_init(struct kfd_dev *kfd, unsigned int buf_size,
@@ -279,7 +323,7 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd,
 	struct pci_dev *pdev, const struct kfd2kgd_calls *f2g)
 {
 	struct kfd_dev *kfd;
-
+	int ret;
 	const struct kfd_device_info *device_info =
 					lookup_device_info(pdev->device);
 
@@ -288,19 +332,18 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd,
 		return NULL;
 	}
 
-	if (device_info->needs_pci_atomics) {
-		/* Allow BIF to recode atomics to PCIe 3.0
-		 * AtomicOps. 32 and 64-bit requests are possible and
-		 * must be supported.
-		 */
-		if (pci_enable_atomic_ops_to_root(pdev,
-				PCI_EXP_DEVCAP2_ATOMIC_COMP32 |
-				PCI_EXP_DEVCAP2_ATOMIC_COMP64) < 0) {
-			dev_info(kfd_device,
-				"skipped device %x:%x, PCI rejects atomics",
-				 pdev->vendor, pdev->device);
-			return NULL;
-		}
+	/* Allow BIF to recode atomics to PCIe 3.0 AtomicOps.
+	 * 32 and 64-bit requests are possible and must be
+	 * supported.
+	 */
+	ret = pci_enable_atomic_ops_to_root(pdev,
+			PCI_EXP_DEVCAP2_ATOMIC_COMP32 |
+			PCI_EXP_DEVCAP2_ATOMIC_COMP64);
+	if (device_info->needs_pci_atomics && ret < 0) {
+		dev_info(kfd_device,
+			 "skipped device %x:%x, PCI rejects atomics\n",
+			 pdev->vendor, pdev->device);
+		return NULL;
 	}
 
 	kfd = kzalloc(sizeof(*kfd), GFP_KERNEL);
@@ -323,10 +366,16 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd,
 static void kfd_cwsr_init(struct kfd_dev *kfd)
 {
 	if (cwsr_enable && kfd->device_info->supports_cwsr) {
-		BUILD_BUG_ON(sizeof(cwsr_trap_gfx8_hex) > PAGE_SIZE);
+		if (kfd->device_info->asic_family < CHIP_VEGA10) {
+			BUILD_BUG_ON(sizeof(cwsr_trap_gfx8_hex) > PAGE_SIZE);
+			kfd->cwsr_isa = cwsr_trap_gfx8_hex;
+			kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx8_hex);
+		} else {
+			BUILD_BUG_ON(sizeof(cwsr_trap_gfx9_hex) > PAGE_SIZE);
+			kfd->cwsr_isa = cwsr_trap_gfx9_hex;
+			kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx9_hex);
+		}
 
-		kfd->cwsr_isa = cwsr_trap_gfx8_hex;
-		kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx8_hex);
 		kfd->cwsr_enabled = true;
 	}
 }
@@ -541,6 +590,44 @@ void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry)
 	spin_unlock(&kfd->interrupt_lock);
 }
 
+int kgd2kfd_quiesce_mm(struct mm_struct *mm)
+{
+	struct kfd_process *p;
+	int r;
+
+	/* Because we are called from arbitrary context (workqueue) as opposed
+	 * to process context, kfd_process could attempt to exit while we are
+	 * running so the lookup function increments the process ref count.
+	 */
+	p = kfd_lookup_process_by_mm(mm);
+	if (!p)
+		return -ESRCH;
+
+	r = kfd_process_evict_queues(p);
+
+	kfd_unref_process(p);
+	return r;
+}
+
+int kgd2kfd_resume_mm(struct mm_struct *mm)
+{
+	struct kfd_process *p;
+	int r;
+
+	/* Because we are called from arbitrary context (workqueue) as opposed
+	 * to process context, kfd_process could attempt to exit while we are
+	 * running so the lookup function increments the process ref count.
+	 */
+	p = kfd_lookup_process_by_mm(mm);
+	if (!p)
+		return -ESRCH;
+
+	r = kfd_process_restore_queues(p);
+
+	kfd_unref_process(p);
+	return r;
+}
+
 /** kgd2kfd_schedule_evict_and_restore_process - Schedules work queue that will
  *   prepare for safe eviction of KFD BOs that belong to the specified
  *   process.
@@ -652,7 +739,7 @@ int kfd_gtt_sa_allocate(struct kfd_dev *kfd, unsigned int size,
 	if (size > kfd->gtt_sa_num_of_chunks * kfd->gtt_sa_chunk_size)
 		return -ENOMEM;
 
-	*mem_obj = kmalloc(sizeof(struct kfd_mem_obj), GFP_KERNEL);
+	*mem_obj = kzalloc(sizeof(struct kfd_mem_obj), GFP_NOIO);
 	if ((*mem_obj) == NULL)
 		return -ENOMEM;
 

drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h

@@ -200,6 +200,8 @@ void device_queue_manager_init_vi(
 		struct device_queue_manager_asic_ops *asic_ops);
 void device_queue_manager_init_vi_tonga(
 		struct device_queue_manager_asic_ops *asic_ops);
+void device_queue_manager_init_v9(
+		struct device_queue_manager_asic_ops *asic_ops);
 void program_sh_mem_settings(struct device_queue_manager *dqm,
 					struct qcm_process_device *qpd);
 unsigned int get_queues_num(struct device_queue_manager *dqm);

drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c

+/*
+ * Copyright 2016-2018 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "kfd_device_queue_manager.h"
+#include "vega10_enum.h"
+#include "gc/gc_9_0_offset.h"
+#include "gc/gc_9_0_sh_mask.h"
+#include "sdma0/sdma0_4_0_sh_mask.h"
+
+static int update_qpd_v9(struct device_queue_manager *dqm,
+			 struct qcm_process_device *qpd);
+static void init_sdma_vm_v9(struct device_queue_manager *dqm, struct queue *q,
+			    struct qcm_process_device *qpd);
+
+void device_queue_manager_init_v9(
+	struct device_queue_manager_asic_ops *asic_ops)
+{
+	asic_ops->update_qpd = update_qpd_v9;
+	asic_ops->init_sdma_vm = init_sdma_vm_v9;
+}
+
+static uint32_t compute_sh_mem_bases_64bit(struct kfd_process_device *pdd)
+{
+	uint32_t shared_base = pdd->lds_base >> 48;
+	uint32_t private_base = pdd->scratch_base >> 48;
+
+	return (shared_base << SH_MEM_BASES__SHARED_BASE__SHIFT) |
+		private_base;
+}
+
+static int update_qpd_v9(struct device_queue_manager *dqm,
+			 struct qcm_process_device *qpd)
+{
+	struct kfd_process_device *pdd;
+
+	pdd = qpd_to_pdd(qpd);
+
+	/* check if sh_mem_config register already configured */
+	if (qpd->sh_mem_config == 0) {
+		qpd->sh_mem_config =
+				SH_MEM_ALIGNMENT_MODE_UNALIGNED <<
+					SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT;
+		if (vega10_noretry &&
+		    !dqm->dev->device_info->needs_iommu_device)
+			qpd->sh_mem_config |=
+				1 << SH_MEM_CONFIG__RETRY_DISABLE__SHIFT;
+
+		qpd->sh_mem_ape1_limit = 0;
+		qpd->sh_mem_ape1_base = 0;
+	}
+
+	qpd->sh_mem_bases = compute_sh_mem_bases_64bit(pdd);
+
+	pr_debug("sh_mem_bases 0x%X\n", qpd->sh_mem_bases);
+
+	return 0;
+}
+
+static void init_sdma_vm_v9(struct device_queue_manager *dqm, struct queue *q,
+			    struct qcm_process_device *qpd)
+{
+	/* Not needed on SDMAv4 any more */
+	q->properties.sdma_vm_addr = 0;
+}

drivers/gpu/drm/amd/amdkfd/kfd_events.c

@@ -345,7 +345,7 @@ int kfd_event_create(struct file *devkfd, struct kfd_process *p,
 	case KFD_EVENT_TYPE_DEBUG:
 		ret = create_signal_event(devkfd, p, ev);
 		if (!ret) {
-			*event_page_offset = KFD_MMAP_EVENTS_MASK;
+			*event_page_offset = KFD_MMAP_TYPE_EVENTS;
 			*event_page_offset <<= PAGE_SHIFT;
 			*event_slot_index = ev->event_id;
 		}
@@ -496,7 +496,7 @@ void kfd_signal_event_interrupt(unsigned int pasid, uint32_t partial_id,
 			pr_debug_ratelimited("Partial ID invalid: %u (%u valid bits)\n",
 					     partial_id, valid_id_bits);
 
-		if (p->signal_event_count < KFD_SIGNAL_EVENT_LIMIT/2) {
+		if (p->signal_event_count < KFD_SIGNAL_EVENT_LIMIT / 64) {
 			/* With relatively few events, it's faster to
 			 * iterate over the event IDR
 			 */

drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c

@@ -139,10 +139,12 @@ static void interrupt_wq(struct work_struct *work)
 {
 	struct kfd_dev *dev = container_of(work, struct kfd_dev,
 						interrupt_work);
+	uint32_t ih_ring_entry[KFD_MAX_RING_ENTRY_SIZE];
 
-	uint32_t ih_ring_entry[DIV_ROUND_UP(
-				dev->device_info->ih_ring_entry_size,
-				sizeof(uint32_t))];
+	if (dev->device_info->ih_ring_entry_size > sizeof(ih_ring_entry)) {
+		dev_err_once(kfd_chardev(), "Ring entry too small\n");
+		return;
+	}
 
 	while (dequeue_ih_ring_entry(dev, ih_ring_entry))
 		dev->device_info->event_interrupt_class->interrupt_wq(dev,

drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h

@@ -72,6 +72,7 @@ struct kernel_queue {
 	struct kfd_dev		*dev;
 	struct mqd_manager	*mqd;
 	struct queue		*queue;
+	uint64_t		pending_wptr64;
 	uint32_t		pending_wptr;
 	unsigned int		nop_packet;
 
@@ -79,7 +80,10 @@ struct kernel_queue {
 	uint32_t		*rptr_kernel;
 	uint64_t		rptr_gpu_addr;
 	struct kfd_mem_obj	*wptr_mem;
-	uint32_t		*wptr_kernel;
+	union {
+		uint64_t	*wptr64_kernel;
+		uint32_t	*wptr_kernel;
+	};
 	uint64_t		wptr_gpu_addr;
 	struct kfd_mem_obj	*pq;
 	uint64_t		pq_gpu_addr;
@@ -97,5 +101,6 @@ struct kernel_queue {
 
 void kernel_queue_init_cik(struct kernel_queue_ops *ops);
 void kernel_queue_init_vi(struct kernel_queue_ops *ops);
+void kernel_queue_init_v9(struct kernel_queue_ops *ops);
 
 #endif /* KFD_KERNEL_QUEUE_H_ */

drivers/gpu/drm/amd/amdkfd/kfd_module.c

@@ -43,6 +43,8 @@ static const struct kgd2kfd_calls kgd2kfd = {
 	.interrupt	= kgd2kfd_interrupt,
 	.suspend	= kgd2kfd_suspend,
 	.resume		= kgd2kfd_resume,
+	.quiesce_mm	= kgd2kfd_quiesce_mm,
+	.resume_mm	= kgd2kfd_resume_mm,
 	.schedule_evict_and_restore_process =
 			  kgd2kfd_schedule_evict_and_restore_process,
 };
@@ -81,6 +83,11 @@ module_param(ignore_crat, int, 0444);
 MODULE_PARM_DESC(ignore_crat,
 	"Ignore CRAT table during KFD initialization (0 = use CRAT (default), 1 = ignore CRAT)");
 
+int vega10_noretry;
+module_param_named(noretry, vega10_noretry, int, 0644);
+MODULE_PARM_DESC(noretry,
+	"Set sh_mem_config.retry_disable on Vega10 (0 = retry enabled (default), 1 = retry disabled)");
+
 static int amdkfd_init_completed;
 
 int kgd2kfd_init(unsigned int interface_version,