Skip to contentSkip to conversation

fs: sync_filesystem should not discard sync_fs return code

PATCH In review Jeff Layton 2 participants 2 messages May 30, 2017 – May 31, 2017 VFS

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 fs: sync_filesystem should not discard sync_fs return code

sync_filesystem is called from several kernel subsystems to write out a superblock. After writing out the inodes, it calls the superblock sync_fs operation, which is an int return operation. That error code is then discarded. It then calls __sync_blockdev and returns an error if that fails.

Most filesystems that have a sync_fs operation do most of the work inside that operation and then return an error if it fails. With those filesystems, the sync_blockdev is effectively a no-op and may very well not return an error even though writeback failed.

Have __sync_filesystem preserve the error from sync_fs and return that if it fails, or the return from __sync_blockdev if it succeeds.

---
 fs/sync.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
fs/sync.c +6 -2
index 45ecf6854595..ec93aac4feb9 100644
@@ -29,14 +29,18 @@
  */
 static int __sync_filesystem(struct super_block *sb, int wait)
 {
+	int fs_ret = 0, bd_ret;
+
 	if (wait)
 		sync_inodes_sb(sb);
 	else
 		writeback_inodes_sb(sb, WB_REASON_SYNC);
 
 	if (sb->s_op->sync_fs)
-		sb->s_op->sync_fs(sb, wait);
-	return __sync_blockdev(sb->s_bdev, wait);
+		fs_ret = sb->s_op->sync_fs(sb, wait);
+	bd_ret = __sync_blockdev(sb->s_bdev, wait);
+
+	return fs_ret ? fs_ret : bd_ret;
 }
 
 /*
Signature · 1 line
2.9.4
reviewer Reviewed-by on v1 1/1
#1 lore raw
Quoted 11 lines · Jeff Layton

sync_filesystem is called from several kernel subsystems to write out a superblock. After writing out the inodes, it calls the superblock sync_fs operation, which is an int return operation. That error code is then discarded. It then calls __sync_blockdev and returns an error if that fails.

Most filesystems that have a sync_fs operation do most of the work inside that operation and then return an error if it fails. With those filesystems, the sync_blockdev is effectively a no-op and may very well not return an error even though writeback failed.

Quoted 4 lines

Have __sync_filesystem preserve the error from sync_fs and return that if it fails, or the return from __sync_blockdev if it succeeds.

Makes sense. Although this will likely have only a small impact since most of the errors are likely to happen in the inode writeback where they just get ignored. Anyway you can add:

Honza

Quoted 32 lines
---
 fs/sync.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
fs/sync.c +6 -2
index 45ecf6854595..ec93aac4feb9 100644
@@ -29,14 +29,18 @@
  */
 static int __sync_filesystem(struct super_block *sb, int wait)
 {
+	int fs_ret = 0, bd_ret;
+
 	if (wait)
 		sync_inodes_sb(sb);
 	else
 		writeback_inodes_sb(sb, WB_REASON_SYNC);
 
 	if (sb->s_op->sync_fs)
-		sb->s_op->sync_fs(sb, wait);
-	return __sync_blockdev(sb->s_bdev, wait);
+		fs_ret = sb->s_op->sync_fs(sb, wait);
+	bd_ret = __sync_blockdev(sb->s_bdev, wait);
+
+	return fs_ret ? fs_ret : bd_ret;
 }
 
 /*
Signature · 1 line
2.9.4
Signature · 2 lines
Jan Kara <[email protected]>
SUSE Labs, CR