fuse: invalidate the correct range after O_APPEND direct write
Subsystems (2)
PATCH 1/1 fuse: invalidate the correct range after O_APPEND direct write
fuse_direct_write_iter() captures pos before generic_write_checks(), which moves ki_pos to EOF for O_APPEND writes:
fuse_direct_write_iter()
{
pos = iocb->ki_pos; /* 0 (user-supplied) */
generic_write_checks(); /* ki_pos -> EOF */
fuse_direct_io(); /* writes at EOF, correct */
invalidate(pos, pos + res); /* [0, res) -- wrong */
}The post-write invalidation targets a stale range instead of the actual written range at EOF.
This can cause data inconsistency when the file size is not page-aligned. The tail page straddling EOF has a valid portion before EOF that concurrent readers can fault back in during the DIO write window:
Tail page (file size X not page-aligned):
page_start X (EOF) page_end
|--- valid data ----|-- stale --|
CPU0 (O_APPEND DIO writer) CPU1 (buffered reader)
-------------------------- ----------------------
invalidate [X, X+len)
tail page evicted
FUSE_WRITE in flight ...
read [page_start, X)
tail page re-faulted [X, page_end) = stale
FUSE_WRITE completes
i_size = X + len
invalidate [0, len) <- WRONG
tail page still cached
read [X, X+len)
hits stale tail page returns old data
Fix by reading pos back from iocb->ki_pos after generic_write_checks(), as generic_file_direct_write() does.
Also fix a typo in the comment ("may have" -> "may have competed").
--- fs/fuse/file.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
fs/fuse/file.c +3 -2
@@ -1789,13 +1789,14 @@ static ssize_t fuse_direct_write_iter(struct kiocb *iocb, struct iov_iter *from) {struct inode *inode =file_inode (iocb->ki_filp);struct address_space *mapping = inode->i_mapping; -loff_t pos = iocb->ki_pos;ssize_t res;bool exclusive;fuse_dio_lock (iocb, from, &exclusive); res =generic_write_checks (iocb, from);if (res >0 ) { +loff_t pos = iocb->ki_pos; +task_io_account_write (res);if (!is_sync_kiocb (iocb)) { res =fuse_direct_IO (iocb, from);
@@ -1810,7 +1811,7 @@ static ssize_t fuse_direct_write_iter(struct kiocb *iocb, struct iov_iter *from)/* * As in generic_file_direct_write(), invalidate after * write, to invalidate read-ahead cache that may have -* with the write. +* competed with the write. */ invalidate_inode_pages2_range(mapping, pos >> PAGE_SHIFT,
Signature · 1 line
2.43.7
Quoted 72 lines · Baokun Li
fuse_direct_write_iter()captures pos beforegeneric_write_checks(), which moveski_posto EOF forO_APPENDwrites:fuse_direct_write_iter() { pos = iocb->ki_pos; /* 0 (user-supplied) */ generic_write_checks(); /* ki_pos -> EOF */ fuse_direct_io(); /* writes at EOF, correct */ invalidate(pos, pos + res); /* [0, res) -- wrong */ }The post-write invalidation targets a stale range instead of the actual written range at EOF.
This can cause data inconsistency when the file size is not page-aligned. The tail page straddling EOF has a valid portion before EOF that concurrent readers can fault back in during the DIO write window:
Tail page (file size X not page-aligned):
page_startX (EOF)page_end
|--- valid data ----|-- stale --|CPU0 (
O_APPENDDIO writer) CPU1 (buffered reader)
-------------------------- ----------------------
invalidate [X, X+len)
tail page evicted
FUSE_WRITEin flight ...
read [page_start, X)
tail page re-faulted [X,page_end) = stale
FUSE_WRITEcompletes
i_size= X + len
invalidate [0, len) <- WRONG
tail page still cached
read [X, X+len)
hits stale tail page returns old dataFix by reading pos back from
iocb->ki_posaftergeneric_write_checks(), asgeneric_file_direct_write()does.Also fix a typo in the comment ("may have" -> "may have competed").
Fixes 2b0408d0284f ("fuse: invalidate page cache after DIO and async DIO writes")Signed-off-by Baokun Li <[email protected]>--- fs/fuse/file.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)fs/fuse/file.c +2
-1@@ -1789,13 +1789,14 @@ static ssize_t fuse_direct_write_iter(struct kiocb *iocb, struct iov_iter *from) {struct inode *inode =file_inode (iocb->ki_filp);struct address_space *mapping = inode->i_mapping; -loff_t pos = iocb->ki_pos;ssize_t res;bool exclusive;fuse_dio_lock (iocb, from, &exclusive); res =generic_write_checks (iocb, from);if (res >0 ) { +loff_t pos = iocb->ki_pos; +task_io_account_write (res);if (!is_sync_kiocb (iocb)) { res =fuse_direct_IO (iocb, from);
fs/fuse/file.c static ssize_t fuse_direct_write_iter(struct kiocb *iocb, struct iov_iter *from) · line 1817
Quoted 8 lines
fs/fuse/file.c +1
-1@@ -1810,7 +1811,7 @@ static ssize_t fuse_direct_write_iter(struct kiocb *iocb, struct iov_iter *from)/* * As in generic_file_direct_write(), invalidate after * write, to invalidate read-ahead cache that may have -* with the write. +* competed with the write. */ invalidate_inode_pages2_range(mapping, pos >> PAGE_SHIFT,
Good catch! Looks like there is a missing xfstest scenario.
Quoted 7 lines · Baokun Li
Fix by reading pos back from
iocb->ki_posaftergeneric_write_checks(), asgeneric_file_direct_write()does.Also fix a typo in the comment ("may have" -> "may have competed").
Fixes 2b0408d0284f ("fuse: invalidate page cache after DIO and async DIO writes")Signed-off-by Baokun Li <[email protected]>
Applied, thanks.
Miklos