drm/amdgpu: keep the PTs validation list in the VM v2

This avoids allocating it on the fly.

v2: fix grammar in comment

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>

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

@@ -917,8 +917,8 @@ struct amdgpu_ring {
 #define AMDGPU_VM_FAULT_STOP_ALWAYS	2
 
 struct amdgpu_vm_pt {
-	struct amdgpu_bo	*bo;
-	uint64_t		addr;
+	struct amdgpu_bo_list_entry	entry;
+	uint64_t			addr;
 };
 
 struct amdgpu_vm_id {
@@ -983,8 +983,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm);
 void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
 			 struct list_head *validated,
 			 struct amdgpu_bo_list_entry *entry);
-struct amdgpu_bo_list_entry *amdgpu_vm_get_pt_bos(struct amdgpu_vm *vm,
-						  struct list_head *duplicates);
+void amdgpu_vm_get_pt_bos(struct amdgpu_vm *vm, struct list_head *duplicates);
 int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,
 		      struct amdgpu_sync *sync);
 void amdgpu_vm_flush(struct amdgpu_ring *ring,
@@ -1255,7 +1254,6 @@ struct amdgpu_cs_parser {
 	struct amdgpu_cs_chunk	*chunks;
 	/* relocations */
 	struct amdgpu_bo_list_entry	vm_pd;
-	struct amdgpu_bo_list_entry	*vm_bos;
 	struct list_head	validated;
 	struct fence		*fence;
 

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

@@ -396,11 +396,7 @@ static int amdgpu_cs_parser_relocs(struct amdgpu_cs_parser *p)
 	if (unlikely(r != 0))
 		goto error_reserve;
 
-	p->vm_bos = amdgpu_vm_get_pt_bos(&fpriv->vm, &duplicates);
-	if (!p->vm_bos) {
-		r = -ENOMEM;
-		goto error_validate;
-	}
+	amdgpu_vm_get_pt_bos(&fpriv->vm, &duplicates);
 
 	r = amdgpu_cs_list_validate(p->adev, &fpriv->vm, &p->validated);
 	if (r)
@@ -483,7 +479,6 @@ static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser, int error, bo
 	if (parser->bo_list)
 		amdgpu_bo_list_put(parser->bo_list);
 
-	drm_free_large(parser->vm_bos);
 	for (i = 0; i < parser->nchunks; i++)
 		drm_free_large(parser->chunks[i].kdata);
 	kfree(parser->chunks);

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

@@ -447,7 +447,6 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
 				    struct amdgpu_bo_va *bo_va, uint32_t operation)
 {
 	struct ttm_validate_buffer tv, *entry;
-	struct amdgpu_bo_list_entry *vm_bos;
 	struct amdgpu_bo_list_entry vm_pd;
 	struct ww_acquire_ctx ticket;
 	struct list_head list, duplicates;
@@ -468,12 +467,7 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
 	if (r)
 		goto error_print;
 
-	vm_bos = amdgpu_vm_get_pt_bos(bo_va->vm, &duplicates);
-	if (!vm_bos) {
-		r = -ENOMEM;
-		goto error_unreserve;
-	}
-
+	amdgpu_vm_get_pt_bos(bo_va->vm, &duplicates);
 	list_for_each_entry(entry, &list, head) {
 		domain = amdgpu_mem_type_to_domain(entry->bo->mem.mem_type);
 		/* if anything is swapped out don't swap it in here,
@@ -494,7 +488,6 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
 
 error_unreserve:
 	ttm_eu_backoff_reservation(&ticket, &list);
-	drm_free_large(vm_bos);
 
 error_print:
 	if (r && r != -ERESTARTSYS)

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

@@ -98,40 +98,27 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm,
 }
 
 /**
- * amdgpu_vm_get_bos - add the vm BOs to a validation list
+ * amdgpu_vm_get_bos - add the vm BOs to a duplicates list
  *
  * @vm: vm providing the BOs
  * @duplicates: head of duplicates list
  *
- * Add the page directory to the list of BOs to
- * validate for command submission (cayman+).
+ * Add the page directory to the BO duplicates list
+ * for command submission.
  */
-struct amdgpu_bo_list_entry *amdgpu_vm_get_pt_bos(struct amdgpu_vm *vm,
-						  struct list_head *duplicates)
+void amdgpu_vm_get_pt_bos(struct amdgpu_vm *vm, struct list_head *duplicates)
 {
-	struct amdgpu_bo_list_entry *list;
-	unsigned i, idx;
-
-	list = drm_malloc_ab(vm->max_pde_used + 1,
-			     sizeof(struct amdgpu_bo_list_entry));
-	if (!list)
-		return NULL;
+	unsigned i;
 
 	/* add the vm page table to the list */
-	for (i = 0, idx = 0; i <= vm->max_pde_used; i++) {
-		if (!vm->page_tables[i].bo)
+	for (i = 0; i <= vm->max_pde_used; ++i) {
+		struct amdgpu_bo_list_entry *entry = &vm->page_tables[i].entry;
+
+		if (!entry->robj)
 			continue;
 
-		list[idx].robj = vm->page_tables[i].bo;
-		list[idx].prefered_domains = AMDGPU_GEM_DOMAIN_VRAM;
-		list[idx].allowed_domains = AMDGPU_GEM_DOMAIN_VRAM;
-		list[idx].priority = 0;
-		list[idx].tv.bo = &list[idx].robj->tbo;
-		list[idx].tv.shared = true;
-		list_add(&list[idx++].tv.head, duplicates);
+		list_add(&entry->tv.head, duplicates);
 	}
-
-	return list;
 }
 
 /**
@@ -474,7 +461,7 @@ int amdgpu_vm_update_page_directory(struct amdgpu_device *adev,
 
 	/* walk over the address space and update the page directory */
 	for (pt_idx = 0; pt_idx <= vm->max_pde_used; ++pt_idx) {
-		struct amdgpu_bo *bo = vm->page_tables[pt_idx].bo;
+		struct amdgpu_bo *bo = vm->page_tables[pt_idx].entry.robj;
 		uint64_t pde, pt;
 
 		if (bo == NULL)
@@ -651,7 +638,7 @@ static int amdgpu_vm_update_ptes(struct amdgpu_device *adev,
 	/* walk over the address space and update the page tables */
 	for (addr = start; addr < end; ) {
 		uint64_t pt_idx = addr >> amdgpu_vm_block_size;
-		struct amdgpu_bo *pt = vm->page_tables[pt_idx].bo;
+		struct amdgpu_bo *pt = vm->page_tables[pt_idx].entry.robj;
 		unsigned nptes;
 		uint64_t pte;
 		int r;
@@ -1083,9 +1070,11 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 	/* walk over the address space and allocate the page tables */
 	for (pt_idx = saddr; pt_idx <= eaddr; ++pt_idx) {
 		struct reservation_object *resv = vm->page_directory->tbo.resv;
+		struct amdgpu_bo_list_entry *entry;
 		struct amdgpu_bo *pt;
 
-		if (vm->page_tables[pt_idx].bo)
+		entry = &vm->page_tables[pt_idx].entry;
+		if (entry->robj)
 			continue;
 
 		r = amdgpu_bo_create(adev, AMDGPU_VM_PTE_COUNT * 8,
@@ -1102,8 +1091,13 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 			goto error_free;
 		}
 
+		entry->robj = pt;
+		entry->prefered_domains = AMDGPU_GEM_DOMAIN_VRAM;
+		entry->allowed_domains = AMDGPU_GEM_DOMAIN_VRAM;
+		entry->priority = 0;
+		entry->tv.bo = &entry->robj->tbo;
+		entry->tv.shared = true;
 		vm->page_tables[pt_idx].addr = 0;
-		vm->page_tables[pt_idx].bo = pt;
 	}
 
 	return 0;
@@ -1334,7 +1328,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 	}
 
 	for (i = 0; i < amdgpu_vm_num_pdes(adev); i++)
-		amdgpu_bo_unref(&vm->page_tables[i].bo);
+		amdgpu_bo_unref(&vm->page_tables[i].entry.robj);
 	kfree(vm->page_tables);
 
 	amdgpu_bo_unref(&vm->page_directory);