jbd: abort when failed to log metadata buffers
authorHidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
Sun, 19 Oct 2008 03:27:53 +0000 (20:27 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Mon, 20 Oct 2008 15:52:36 +0000 (08:52 -0700)
If we failed to write metadata buffers to the journal space and succeeded
to write the commit record, stale data can be written back to the
filesystem as metadata in the recovery phase.

To avoid this, when we failed to write out metadata buffers, abort the
journal before writing the commit record.

Signed-off-by: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
Acked-by: Jan Kara <jack@suse.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/jbd/commit.c

index ae08c05..f1ea861 100644 (file)
@@ -762,6 +762,9 @@ wait_for_iobuf:
                /* AKPM: bforget here */
        }
 
+       if (err)
+               journal_abort(journal, err);
+
        jbd_debug(3, "JBD: commit phase 6\n");
 
        if (journal_write_commit_record(journal, commit_transaction))