[PATCH v2] f2fs: use adjusted write range after f2fs_write_checks()

[PATCH v2] f2fs: use adjusted write range after f2fs_write_checks()

[Seongjae Jeong]

generic_write_checks() in f2fs_write_checks() can adjust iocb->ki_pos
for append writes and truncate the iterator to limit the number of bytes
to write.

In f2fs_file_write_iter(), the pinned-file overwrite check currently
uses the position and count saved before f2fs_write_checks(), so it
can check a range different from the actual write range.

The forced buffered I/O cleanup also uses orig_pos saved before
f2fs_write_checks(). For O_APPEND writes, this can make the cleanup
flush and invalidate the wrong page cache range.

Move the pinned-file overwrite check after f2fs_write_checks() and use
the adjusted iocb->ki_pos and iov_iter_count(from). Also save the
adjusted write position and use it for the forced buffered I/O cleanup.

Fixes: 3fdd89b452c2 ("f2fs: prevent writing without fallocate() for pinned files")
Fixes: 92318f20d703 ("f2fs: preserve direct write semantics when buffering is forced")
Signed-off-by: Seongjae Jeong <jsjlee1020@gmail.com>
---
V1 -> V2: Move the pinned-file overwrite check after f2fs_write_checks() and
use the adjusted write position and iterator count

 fs/f2fs/file.c | 17 +++++++++--------
 1 file changed, 9 insertions(+), 8 deletions(-)

diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index d82be8c1502a..d440231b8cb9 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -5641,9 +5641,8 @@ static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
 	bool dio;
 	bool may_need_sync = true;
 	int preallocated;
-	const loff_t pos = iocb->ki_pos;
-	const ssize_t count = iov_iter_count(from);
 	ssize_t ret;
+	loff_t bufio_start_pos;
 
 	if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
 		ret = -EIO;
@@ -5664,15 +5663,17 @@ static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
 		inode_lock(inode);
 	}
 
+	ret = f2fs_write_checks(iocb, from);
+	if (ret <= 0)
+		goto out_unlock;
+
 	if (f2fs_is_pinned_file(inode) &&
-	    !f2fs_overwrite_io(inode, pos, count)) {
+	    !f2fs_overwrite_io(inode, iocb->ki_pos, iov_iter_count(from))) {
 		ret = -EIO;
 		goto out_unlock;
 	}
 
-	ret = f2fs_write_checks(iocb, from);
-	if (ret <= 0)
-		goto out_unlock;
+	bufio_start_pos = iocb->ki_pos;
 
 	/* Determine whether we will do a direct write or a buffered write. */
 	dio = f2fs_should_use_dio(inode, iocb, from);
@@ -5727,8 +5728,8 @@ static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
 	 */
 	if (ret > 0 && !dio && (iocb->ki_flags & IOCB_DIRECT))
 		f2fs_flush_buffered_write(iocb->ki_filp->f_mapping,
-					  orig_pos,
-					  orig_pos + ret - 1);
+					  bufio_start_pos,
+					  bufio_start_pos + ret - 1);
 
 	return ret;
 }

base-commit: dafb84f092a6387748b2df4c8f647d46873bc1a0
-- 
2.53.0

Re: [PATCH v2] f2fs: use adjusted write range after f2fs_write_checks()

[Chao Yu]

On 8/24/26 10:32, Seongjae Jeong wrote:
> generic_write_checks() in f2fs_write_checks() can adjust iocb->ki_pos
> for append writes and truncate the iterator to limit the number of bytes
> to write.
> 
> In f2fs_file_write_iter(), the pinned-file overwrite check currently
> uses the position and count saved before f2fs_write_checks(), so it
> can check a range different from the actual write range.
> 
> The forced buffered I/O cleanup also uses orig_pos saved before
> f2fs_write_checks(). For O_APPEND writes, this can make the cleanup
> flush and invalidate the wrong page cache range.
> 
> Move the pinned-file overwrite check after f2fs_write_checks() and use
> the adjusted iocb->ki_pos and iov_iter_count(from). Also save the
> adjusted write position and use it for the forced buffered I/O cleanup.
> 
> Fixes: 3fdd89b452c2 ("f2fs: prevent writing without fallocate() for pinned files")
> Fixes: 92318f20d703 ("f2fs: preserve direct write semantics when buffering is forced")
> Signed-off-by: Seongjae Jeong <jsjlee1020@gmail.com>

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

Thanks,