[for,5.15.x] btrfs: fix lost file sync on direct IO write with nowait and dsync iocb

Commit Message

Filipe Manana Nov. 7, 2022, 9:41 a.m. UTC

From: Filipe Manana <[e-mail]>

commit 8184620ae21213d51eaf2e0bd4186baacb928172 upstream.

When doing a direct IO write using a iocb with nowait and dsync set, we
end up not syncing the file once the write completes.

This is because we tell iomap to not call generic_write_sync(), which
would result in calling btrfs_sync_file(), in order to avoid a deadlock
since iomap can call it while we are holding the inode's lock and
btrfs_sync_file() needs to acquire the inode's lock. The deadlock happens
only if the write happens synchronously, when iomap_dio_rw() calls
iomap_dio_complete() before it returns. Instead we do the sync ourselves
at btrfs_do_write_iter().

For a nowait write however we can end up not doing the sync ourselves at
at btrfs_do_write_iter() because the write could have been queued, and
therefore we get -EIOCBQUEUED returned from iomap in such case. That makes
us skip the sync call at btrfs_do_write_iter(), as we don't do it for
any error returned from btrfs_direct_write(). We can't simply do the call
even if -EIOCBQUEUED is returned, since that would block the task waiting
for IO, both for the data since there are bios still in progress as well
as potentially blocking when joining a log transaction and when syncing
the log (writing log trees, super blocks, etc).

So let iomap do the sync call itself and in order to avoid deadlocks for
the case of synchronous writes (without nowait), use __iomap_dio_rw() and
have ourselves call iomap_dio_complete() after unlocking the inode.

A test case will later be sent for fstests, after this is fixed in Linus'
tree.

Fixes: 51bd9563b678 ("btrfs: fix deadlock due to page faults during direct IO reads and writes")
Reported-by: Марк Коренберг <[e-mail]>
Link: https://lore.kernel.org/linux-btrfs/[e-mail]/
CC: [e-mail] # 6.0+
Signed-off-by: Filipe Manana <[e-mail]>
Signed-off-by: David Sterba <[e-mail]>
---

The commit in the Fixes tag was backported to 5.15 stable releases, so
this patch is meant for 5.15.x and was tested on top of 5.15.77.

 fs/btrfs/file.c | 39 ++++++++++++++++-----------------------
 1 file changed, 16 insertions(+), 23 deletions(-)

Comments

Greg KH Nov. 7, 2022, 12:22 p.m. UTC | #1

On Mon, Nov 07, 2022 at 09:41:35AM +0000, [e-mail] wrote:
> From: Filipe Manana <[e-mail]>
> 
> commit 8184620ae21213d51eaf2e0bd4186baacb928172 upstream.
> 
> When doing a direct IO write using a iocb with nowait and dsync set, we
> end up not syncing the file once the write completes.
> 
> This is because we tell iomap to not call generic_write_sync(), which
> would result in calling btrfs_sync_file(), in order to avoid a deadlock
> since iomap can call it while we are holding the inode's lock and
> btrfs_sync_file() needs to acquire the inode's lock. The deadlock happens
> only if the write happens synchronously, when iomap_dio_rw() calls
> iomap_dio_complete() before it returns. Instead we do the sync ourselves
> at btrfs_do_write_iter().
> 
> For a nowait write however we can end up not doing the sync ourselves at
> at btrfs_do_write_iter() because the write could have been queued, and
> therefore we get -EIOCBQUEUED returned from iomap in such case. That makes
> us skip the sync call at btrfs_do_write_iter(), as we don't do it for
> any error returned from btrfs_direct_write(). We can't simply do the call
> even if -EIOCBQUEUED is returned, since that would block the task waiting
> for IO, both for the data since there are bios still in progress as well
> as potentially blocking when joining a log transaction and when syncing
> the log (writing log trees, super blocks, etc).
> 
> So let iomap do the sync call itself and in order to avoid deadlocks for
> the case of synchronous writes (without nowait), use __iomap_dio_rw() and
> have ourselves call iomap_dio_complete() after unlocking the inode.
> 
> A test case will later be sent for fstests, after this is fixed in Linus'
> tree.
> 
> Fixes: 51bd9563b678 ("btrfs: fix deadlock due to page faults during direct IO reads and writes")
> Reported-by: Марк Коренберг <[e-mail]>
> Link: https://lore.kernel.org/linux-btrfs/[e-mail]/
> CC: [e-mail] # 6.0+
> Signed-off-by: Filipe Manana <[e-mail]>
> Signed-off-by: David Sterba <[e-mail]>
> ---
> 
> The commit in the Fixes tag was backported to 5.15 stable releases, so
> this patch is meant for 5.15.x and was tested on top of 5.15.77.

Now queued up, thanks.

greg k-h

Patch

diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 1c597cd6c024..90934711dcf0 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -1906,7 +1906,6 @@  static ssize_t check_direct_IO(struct btrfs_fs_info *fs_info,
 
 static ssize_t btrfs_direct_write(struct kiocb *iocb, struct iov_iter *from)
 {
-	const bool is_sync_write = (iocb->ki_flags & IOCB_DSYNC);
 	struct file *file = iocb->ki_filp;
 	struct inode *inode = file_inode(file);
 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
@@ -1917,6 +1916,7 @@  static ssize_t btrfs_direct_write(struct kiocb *iocb, struct iov_iter *from)
 	loff_t endbyte;
 	ssize_t err;
 	unsigned int ilock_flags = 0;
+	struct iomap_dio *dio;
 
 	if (iocb->ki_flags & IOCB_NOWAIT)
 		ilock_flags |= BTRFS_ILOCK_TRY;
@@ -1959,15 +1959,6 @@  static ssize_t btrfs_direct_write(struct kiocb *iocb, struct iov_iter *from)
 		goto buffered;
 	}
 
-	/*
-	 * We remove IOCB_DSYNC so that we don't deadlock when iomap_dio_rw()
-	 * calls generic_write_sync() (through iomap_dio_complete()), because
-	 * that results in calling fsync (btrfs_sync_file()) which will try to
-	 * lock the inode in exclusive/write mode.
-	 */
-	if (is_sync_write)
-		iocb->ki_flags &= ~IOCB_DSYNC;
-
 	/*
 	 * The iov_iter can be mapped to the same file range we are writing to.
 	 * If that's the case, then we will deadlock in the iomap code, because
@@ -1986,12 +1977,23 @@  static ssize_t btrfs_direct_write(struct kiocb *iocb, struct iov_iter *from)
 	 * So here we disable page faults in the iov_iter and then retry if we
 	 * got -EFAULT, faulting in the pages before the retry.
 	 */
-again:
 	from->nofault = true;
-	err = iomap_dio_rw(iocb, from, &btrfs_dio_iomap_ops, &btrfs_dio_ops,
-			   IOMAP_DIO_PARTIAL, written);
+	dio = __iomap_dio_rw(iocb, from, &btrfs_dio_iomap_ops, &btrfs_dio_ops,
+			     IOMAP_DIO_PARTIAL, written);
 	from->nofault = false;
 
+	/*
+	 * iomap_dio_complete() will call btrfs_sync_file() if we have a dsync
+	 * iocb, and that needs to lock the inode. So unlock it before calling
+	 * iomap_dio_complete() to avoid a deadlock.
+	 */
+	btrfs_inode_unlock(inode, ilock_flags);
+
+	if (IS_ERR_OR_NULL(dio))
+		err = PTR_ERR_OR_ZERO(dio);
+	else
+		err = iomap_dio_complete(dio);
+
 	/* No increment (+=) because iomap returns a cumulative value. */
 	if (err > 0)
 		written = err;
@@ -2017,19 +2019,10 @@  static ssize_t btrfs_direct_write(struct kiocb *iocb, struct iov_iter *from)
 		} else {
 			fault_in_iov_iter_readable(from, left);
 			prev_left = left;
-			goto again;
+			goto relock;
 		}
 	}
 
-	btrfs_inode_unlock(inode, ilock_flags);
-
-	/*
-	 * Add back IOCB_DSYNC. Our caller, btrfs_file_write_iter(), will do
-	 * the fsync (call generic_write_sync()).
-	 */
-	if (is_sync_write)
-		iocb->ki_flags |= IOCB_DSYNC;
-
 	/* If 'err' is -ENOTBLK then it means we must fallback to buffered IO. */
 	if ((err < 0 && err != -ENOTBLK) || !iov_iter_count(from))
 		goto out;