From: Jiucheng Xu <jiucheng.xu@amlogic.com>

On 32-bit architectures, when GFP_NOFS is used, the file cache for write
operations cannot be allocated from the highmem and CMA.

Since mapping->gfp_mask is set to GFP_HIGHUSER_MOVABLE during inode
allocation, using mapping_gfp_mask(mapping) as the GFP flag of getting file
cache for writing is more efficient for 32-bit architectures.

Additionally, use FGP_NOFS to avoid potential deadlock issues caused by
GFP_FS in GFP_HIGHUSER_MOVABLE

Signed-off-by: Jiucheng Xu <jiucheng.xu@amlogic.com>
---
Changes in v2:
- Add FGP_NOFS to keep original GFP_NOFS flag.
- Link to v1: https://lore.kernel.org/r/20250918-origin-dev-v1-1-8d9877df9e77@amlogic.com
---
 fs/f2fs/data.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 7961e0ddfca3aaa332b7dbd4985ae7766551834f..b449edb519d9069f58572685d607a2502e4c9473 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -3587,7 +3587,8 @@ static int f2fs_write_begin(const struct kiocb *iocb,
 	 * Will wait that below with our IO control.
 	 */
 	folio = __filemap_get_folio(mapping, index,
-				FGP_LOCK | FGP_WRITE | FGP_CREAT, GFP_NOFS);
+				FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_NOFS,
+				mapping_gfp_mask(mapping));
 	if (IS_ERR(folio)) {
 		err = PTR_ERR(folio);
 		goto fail;

---
base-commit: c872b6279cd26762339ff02513e2a3f16149a6f1
change-id: 20250910-origin-dev-8a5ff6bee1f2

Best regards,
-- 
Jiucheng Xu <jiucheng.xu@amlogic.com>

On 10/10/25 18:45, Jiucheng Xu via B4 Relay wrote:
> From: Jiucheng Xu <jiucheng.xu@amlogic.com>
> 
> On 32-bit architectures, when GFP_NOFS is used, the file cache for write
> operations cannot be allocated from the highmem and CMA.
> 
> Since mapping->gfp_mask is set to GFP_HIGHUSER_MOVABLE during inode
> allocation, using mapping_gfp_mask(mapping) as the GFP flag of getting file
> cache for writing is more efficient for 32-bit architectures.
> 
> Additionally, use FGP_NOFS to avoid potential deadlock issues caused by
> GFP_FS in GFP_HIGHUSER_MOVABLE
> 
> Signed-off-by: Jiucheng Xu <jiucheng.xu@amlogic.com>

Reviewed-by: Chao Yu <chao@kernel.org>

Thanks,