Revert "drm/amdgpu: replace get_user_pages with HMM mirror helpers"

This reverts commit 915d3ee.

This depends on an HMM fix which is not upstream yet.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

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

@@ -61,6 +61,7 @@ struct kgd_mem {
 
 	atomic_t invalid;
 	struct amdkfd_process_info *process_info;
+	struct page **user_pages;
 
 	struct amdgpu_sync sync;
 

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

@@ -491,12 +491,28 @@ static int init_user_pages(struct kgd_mem *mem, struct mm_struct *mm,
 		goto out;
 	}
 
-	ret = amdgpu_ttm_tt_get_user_pages(bo->tbo.ttm, bo->tbo.ttm->pages);
+	/* If no restore worker is running concurrently, user_pages
+	 * should not be allocated
+	 */
+	WARN(mem->user_pages, "Leaking user_pages array");
+
+	mem->user_pages = kvmalloc_array(bo->tbo.ttm->num_pages,
+					   sizeof(struct page *),
+					   GFP_KERNEL | __GFP_ZERO);
+	if (!mem->user_pages) {
+		pr_err("%s: Failed to allocate pages array\n", __func__);
+		ret = -ENOMEM;
+		goto unregister_out;
+	}
+
+	ret = amdgpu_ttm_tt_get_user_pages(bo->tbo.ttm, mem->user_pages);
 	if (ret) {
 		pr_err("%s: Failed to get user pages: %d\n", __func__, ret);
-		goto unregister_out;
+		goto free_out;
 	}
 
+	amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, mem->user_pages);
+
 	ret = amdgpu_bo_reserve(bo, true);
 	if (ret) {
 		pr_err("%s: Failed to reserve BO\n", __func__);
@@ -509,7 +525,11 @@ static int init_user_pages(struct kgd_mem *mem, struct mm_struct *mm,
 	amdgpu_bo_unreserve(bo);
 
 release_out:
-	amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm);
+	if (ret)
+		release_pages(mem->user_pages, bo->tbo.ttm->num_pages);
+free_out:
+	kvfree(mem->user_pages);
+	mem->user_pages = NULL;
 unregister_out:
 	if (ret)
 		amdgpu_mn_unregister(bo);
@@ -568,6 +588,7 @@ static int reserve_bo_and_vm(struct kgd_mem *mem,
 	ctx->kfd_bo.priority = 0;
 	ctx->kfd_bo.tv.bo = &bo->tbo;
 	ctx->kfd_bo.tv.num_shared = 1;
+	ctx->kfd_bo.user_pages = NULL;
 	list_add(&ctx->kfd_bo.tv.head, &ctx->list);
 
 	amdgpu_vm_get_pd_bo(vm, &ctx->list, &ctx->vm_pd[0]);
@@ -631,6 +652,7 @@ static int reserve_bo_and_cond_vms(struct kgd_mem *mem,
 	ctx->kfd_bo.priority = 0;
 	ctx->kfd_bo.tv.bo = &bo->tbo;
 	ctx->kfd_bo.tv.num_shared = 1;
+	ctx->kfd_bo.user_pages = NULL;
 	list_add(&ctx->kfd_bo.tv.head, &ctx->list);
 
 	i = 0;
@@ -1240,6 +1262,15 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
 	list_del(&bo_list_entry->head);
 	mutex_unlock(&process_info->lock);
 
+	/* Free user pages if necessary */
+	if (mem->user_pages) {
+		pr_debug("%s: Freeing user_pages array\n", __func__);
+		if (mem->user_pages[0])
+			release_pages(mem->user_pages,
+					mem->bo->tbo.ttm->num_pages);
+		kvfree(mem->user_pages);
+	}
+
 	ret = reserve_bo_and_cond_vms(mem, NULL, BO_VM_ALL, &ctx);
 	if (unlikely(ret))
 		return ret;
@@ -1713,11 +1744,25 @@ static int update_invalid_user_pages(struct amdkfd_process_info *process_info,
 
 		bo = mem->bo;
 
+		if (!mem->user_pages) {
+			mem->user_pages =
+				kvmalloc_array(bo->tbo.ttm->num_pages,
+						 sizeof(struct page *),
+						 GFP_KERNEL | __GFP_ZERO);
+			if (!mem->user_pages) {
+				pr_err("%s: Failed to allocate pages array\n",
+				       __func__);
+				return -ENOMEM;
+			}
+		} else if (mem->user_pages[0]) {
+			release_pages(mem->user_pages, bo->tbo.ttm->num_pages);
+		}
+
 		/* Get updated user pages */
 		ret = amdgpu_ttm_tt_get_user_pages(bo->tbo.ttm,
-						   bo->tbo.ttm->pages);
+						   mem->user_pages);
 		if (ret) {
-			bo->tbo.ttm->pages[0] = NULL;
+			mem->user_pages[0] = NULL;
 			pr_info("%s: Failed to get user pages: %d\n",
 				__func__, ret);
 			/* Pretend it succeeded. It will fail later
@@ -1726,6 +1771,12 @@ static int update_invalid_user_pages(struct amdkfd_process_info *process_info,
 			 * stalled user mode queues.
 			 */
 		}
+
+		/* Mark the BO as valid unless it was invalidated
+		 * again concurrently
+		 */
+		if (atomic_cmpxchg(&mem->invalid, invalid, 0) != invalid)
+			return -EAGAIN;
 	}
 
 	return 0;
@@ -1755,8 +1806,7 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info)
 				     GFP_KERNEL);
 	if (!pd_bo_list_entries) {
 		pr_err("%s: Failed to allocate PD BO list entries\n", __func__);
-		ret = -ENOMEM;
-		goto out_no_mem;
+		return -ENOMEM;
 	}
 
 	INIT_LIST_HEAD(&resv_list);
@@ -1780,7 +1830,7 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info)
 	ret = ttm_eu_reserve_buffers(&ticket, &resv_list, false, &duplicates);
 	WARN(!list_empty(&duplicates), "Duplicates should be empty");
 	if (ret)
-		goto out_free;
+		goto out;
 
 	amdgpu_sync_create(&sync);
 
@@ -1796,8 +1846,10 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info)
 
 		bo = mem->bo;
 
-		/* Validate the BO if we got user pages */
-		if (bo->tbo.ttm->pages[0]) {
+		/* Copy pages array and validate the BO if we got user pages */
+		if (mem->user_pages[0]) {
+			amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm,
+						     mem->user_pages);
 			amdgpu_bo_placement_from_domain(bo, mem->domain);
 			ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx);
 			if (ret) {
@@ -1806,16 +1858,16 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info)
 			}
 		}
 
+		/* Validate succeeded, now the BO owns the pages, free
+		 * our copy of the pointer array. Put this BO back on
+		 * the userptr_valid_list. If we need to revalidate
+		 * it, we need to start from scratch.
+		 */
+		kvfree(mem->user_pages);
+		mem->user_pages = NULL;
 		list_move_tail(&mem->validate_list.head,
 			       &process_info->userptr_valid_list);
 
-		/* Stop HMM track the userptr update. We dont check the return
-		 * value for concurrent CPU page table update because we will
-		 * reschedule the restore worker if process_info->evicted_bos
-		 * is updated.
-		 */
-		amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm);
-
 		/* Update mapping. If the BO was not validated
 		 * (because we couldn't get user pages), this will
 		 * clear the page table entries, which will result in
@@ -1845,15 +1897,8 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info)
 	ttm_eu_backoff_reservation(&ticket, &resv_list);
 	amdgpu_sync_wait(&sync, false);
 	amdgpu_sync_free(&sync);
-out_free:
+out:
 	kfree(pd_bo_list_entries);
-out_no_mem:
-	list_for_each_entry_safe(mem, tmp_mem,
-				 &process_info->userptr_inval_list,
-				 validate_list.head) {
-		bo = mem->bo;
-		amdgpu_ttm_tt_get_user_pages_done(bo->tbo.ttm);
-	}
 
 	return ret;
 }

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

@@ -36,7 +36,7 @@ struct amdgpu_bo_list_entry {
 	struct amdgpu_bo_va		*bo_va;
 	uint32_t			priority;
 	struct page			**user_pages;
-	bool				user_invalidated;
+	int				user_invalidated;
 };
 
 struct amdgpu_bo_list {