When we call create(), lseek() and write() sequentially, offset != 0
cannot be used as a judgment condition for whether the file already
has extents.

Furthermore, when xfs_bmap_adjacent() has not given a better blkno,
it is not necessary to use exact EOF block allocation.

Suggested-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Jinliang Zheng <alexjlzheng@tencent.com>
---
Changelog:
- V3: use ap->eof to mark whether to use the EXACT allocation algorithm
- V2: https://lore.kernel.org/linux-xfs/Z1I74KeyZRv2pBBT@dread.disaster.area/
- V1: https://lore.kernel.org/linux-xfs/ZyFJm7xg7Msd6eVr@dread.disaster.area/T/#t
---
 fs/xfs/libxfs/xfs_bmap.c | 13 +++++++------
 1 file changed, 7 insertions(+), 6 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_bmap.c b/fs/xfs/libxfs/xfs_bmap.c
index 5255f93bae31..2b95279303d3 100644
--- a/fs/xfs/libxfs/xfs_bmap.c
+++ b/fs/xfs/libxfs/xfs_bmap.c
@@ -3566,12 +3566,12 @@ xfs_bmap_btalloc_at_eof(
 	int			error;
 
 	/*
-	 * If there are already extents in the file, try an exact EOF block
-	 * allocation to extend the file as a contiguous extent. If that fails,
-	 * or it's the first allocation in a file, just try for a stripe aligned
-	 * allocation.
+	 * If there are already extents in the file, and xfs_bmap_adjacent() has
+	 * given a better blkno, try an exact EOF block allocation to extend the
+	 * file as a contiguous extent. If that fails, or it's the first
+	 * allocation in a file, just try for a stripe aligned allocation.
 	 */
-	if (ap->offset) {
+	if (ap->eof) {
 		xfs_extlen_t	nextminlen = 0;
 
 		/*
@@ -3739,7 +3739,8 @@ xfs_bmap_btalloc_best_length(
 	int			error;
 
 	ap->blkno = XFS_INO_TO_FSB(args->mp, ap->ip->i_ino);
-	xfs_bmap_adjacent(ap);
+	if (!xfs_bmap_adjacent(ap))
+		ap->eof = false;
 
 	/*
 	 * Search for an allocation group with a single extent large enough for
-- 
2.41.1

On Wed, 15 Jan 2025 20:35:25 +0800, Jinliang Zheng wrote:
> When we call create(), lseek() and write() sequentially, offset != 0
> cannot be used as a judgment condition for whether the file already
> has extents.
> 
> Furthermore, when xfs_bmap_adjacent() has not given a better blkno,
> it is not necessary to use exact EOF block allocation.
> 
> [...]

Applied to for-next, thanks!

[1/1] xfs: fix the entry condition of exact EOF block allocation optimization
      commit: 915175b49f65d9edeb81659e82cbb27b621dbc17

Best regards,
-- 
Carlos Maiolino <cem@kernel.org>

On Wed, Jan 15, 2025 at 08:35:25PM +0800, Jinliang Zheng wrote:
> When we call create(), lseek() and write() sequentially, offset != 0
> cannot be used as a judgment condition for whether the file already
> has extents.
> 
> Furthermore, when xfs_bmap_adjacent() has not given a better blkno,
> it is not necessary to use exact EOF block allocation.
> 
> Suggested-by: Dave Chinner <david@fromorbit.com>
> Signed-off-by: Jinliang Zheng <alexjlzheng@tencent.com>
> ---
> Changelog:
> - V3: use ap->eof to mark whether to use the EXACT allocation algorithm
> - V2: https://lore.kernel.org/linux-xfs/Z1I74KeyZRv2pBBT@dread.disaster.area/
> - V1: https://lore.kernel.org/linux-xfs/ZyFJm7xg7Msd6eVr@dread.disaster.area/T/#t
> ---
>  fs/xfs/libxfs/xfs_bmap.c | 13 +++++++------
>  1 file changed, 7 insertions(+), 6 deletions(-)

Looks fine.

Reviewed-by: Dave Chinner <dchinner@redhat.com>

-- 
Dave Chinner
david@fromorbit.com