drm/ttm: Allow page allocations w/o triggering OOM..

This to allow drivers to choose to avoid OOM invocation and handle
page allocation failures instead.

v2:
Remove extra new lines.

Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Roger He <Hongbo.He@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

5 files changed, 27 insertions, 2 deletions

diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index 2fef09a56d16..d33a6bb742a1 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -235,6 +235,9 @@ static int ttm_bo_add_ttm(struct ttm_buffer_object *bo, bool zero_alloc)
 	if (bdev->need_dma32)
 		page_flags |= TTM_PAGE_FLAG_DMA32;
 
+	if (bdev->no_retry)
+		page_flags |= TTM_PAGE_FLAG_NO_RETRY;
+
 	switch (bo->type) {
 	case ttm_bo_type_device:
 		if (zero_alloc)
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c
index 2b12c55a3bff..c84da14311ec 100644
--- a/drivers/gpu/drm/ttm/ttm_page_alloc.c
+++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c
@@ -741,6 +741,9 @@ out:
 		if (ttm_flags & TTM_PAGE_FLAG_ZERO_ALLOC)
 			gfp_flags |= __GFP_ZERO;
 
+		if (ttm_flags & TTM_PAGE_FLAG_NO_RETRY)
+			gfp_flags |= __GFP_RETRY_MAYFAIL;
+
 		/* ttm_alloc_new_pages doesn't reference pool so we can run
 		 * multiple requests in parallel.
 		 **/
@@ -893,6 +896,9 @@ static int ttm_get_pages(struct page **pages, unsigned npages, int flags,
 		if (flags & TTM_PAGE_FLAG_ZERO_ALLOC)
 			gfp_flags |= __GFP_ZERO;
 
+		if (flags & TTM_PAGE_FLAG_NO_RETRY)
+			gfp_flags |= __GFP_RETRY_MAYFAIL;
+
 		if (flags & TTM_PAGE_FLAG_DMA32)
 			gfp_flags |= GFP_DMA32;
 		else
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
index a88051552ace..9e90d0ebc773 100644
--- a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
+++ b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
@@ -920,6 +920,9 @@ static gfp_t ttm_dma_pool_gfp_flags(struct ttm_dma_tt *ttm_dma, bool huge)
 		gfp_flags &= ~__GFP_COMP;
 	}
 
+	if (ttm->page_flags & TTM_PAGE_FLAG_NO_RETRY)
+		gfp_flags |= __GFP_RETRY_MAYFAIL;
+
 	return gfp_flags;
 }
 
diff --git a/drivers/gpu/drm/ttm/ttm_tt.c b/drivers/gpu/drm/ttm/ttm_tt.c
index 5a046a3c543a..9e4d43d68e91 100644
--- a/drivers/gpu/drm/ttm/ttm_tt.c
+++ b/drivers/gpu/drm/ttm/ttm_tt.c
@@ -301,7 +301,11 @@ int ttm_tt_swapin(struct ttm_tt *ttm)
 	swap_space = swap_storage->f_mapping;
 
 	for (i = 0; i < ttm->num_pages; ++i) {
-		from_page = shmem_read_mapping_page(swap_space, i);
+		gfp_t gfp_mask = mapping_gfp_mask(swap_space);
+
+		gfp_mask |= (ttm->page_flags & TTM_PAGE_FLAG_NO_RETRY ? __GFP_RETRY_MAYFAIL : 0);
+		from_page = shmem_read_mapping_page_gfp(swap_space, i, gfp_mask);
+
 		if (IS_ERR(from_page)) {
 			ret = PTR_ERR(from_page);
 			goto out_err;
@@ -350,10 +354,15 @@ int ttm_tt_swapout(struct ttm_tt *ttm, struct file *persistent_swap_storage)
 	swap_space = swap_storage->f_mapping;
 
 	for (i = 0; i < ttm->num_pages; ++i) {
+		gfp_t gfp_mask = mapping_gfp_mask(swap_space);
+
+		gfp_mask |= (ttm->page_flags & TTM_PAGE_FLAG_NO_RETRY ? __GFP_RETRY_MAYFAIL : 0);
+
 		from_page = ttm->pages[i];
 		if (unlikely(from_page == NULL))
 			continue;
-		to_page = shmem_read_mapping_page(swap_space, i);
+
+		to_page = shmem_read_mapping_page_gfp(swap_space, i, gfp_mask);
 		if (IS_ERR(to_page)) {
 			ret = PTR_ERR(to_page);
 			goto out_err;
diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h
index 94064b126e8e..9b417eb2df20 100644
--- a/include/drm/ttm/ttm_bo_driver.h
+++ b/include/drm/ttm/ttm_bo_driver.h
@@ -86,6 +86,7 @@ struct ttm_backend_func {
 #define TTM_PAGE_FLAG_ZERO_ALLOC      (1 << 6)
 #define TTM_PAGE_FLAG_DMA32           (1 << 7)
 #define TTM_PAGE_FLAG_SG              (1 << 8)
+#define TTM_PAGE_FLAG_NO_RETRY	      (1 << 9)
 
 enum ttm_caching_state {
 	tt_uncached,
@@ -556,6 +557,7 @@ struct ttm_bo_global {
  * @dev_mapping: A pointer to the struct address_space representing the
  * device address space.
  * @wq: Work queue structure for the delayed delete workqueue.
+ * @no_retry: Don't retry allocation if it fails
  *
  */
 
@@ -592,6 +594,8 @@ struct ttm_bo_device {
 	struct delayed_work wq;
 
 	bool need_dma32;
+
+	bool no_retry;
 };
 
 /**