bail-early-when-clearing-inode-journal-flag-fails

   1 ext4: bail early when clearing inode journal flag fails
   2
   3 From: Jan Kara <jack@suse.cz>
   4
   5 When clearing inode journal flag, we call jbd2_journal_flush() to force
   6 all the journalled data to their final locations. Currently we ignore
   7 when this fails and continue clearing inode journal flag. This isn't a
   8 big problem because when jbd2_journal_flush() fails, journal is likely
   9 aborted anyway. But it can still lead to somewhat confusing results so
  10 rather bail out early.
  11
  12 Coverity-id: 989044
  13 Signed-off-by: Jan Kara <jack@suse.cz>
  14 Signed-off-by: Theodore Ts'o <tytso@mit.edu>
  15 ---
  16  fs/ext4/inode.c | 7 ++++++-
  17  1 file changed, 6 insertions(+), 1 deletion(-)
  18
  19 diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
  20 index e9777f9..3356ab5 100644
  21 --- a/fs/ext4/inode.c
  22 +++ b/fs/ext4/inode.c
  23 @@ -4959,7 +4959,12 @@ int ext4_change_inode_journal_flag(struct inode *inode, int val)
  24         if (val)
  25                 ext4_set_inode_flag(inode, EXT4_INODE_JOURNAL_DATA);
  26         else {
  27 -               jbd2_journal_flush(journal);
  28 +               err = jbd2_journal_flush(journal);
  29 +               if (err < 0) {
  30 +                       jbd2_journal_unlock_updates(journal);
  31 +                       ext4_inode_resume_unlocked_dio(inode);
  32 +                       return err;
  33 +               }
  34                 ext4_clear_inode_flag(inode, EXT4_INODE_JOURNAL_DATA);
  35         }
  36         ext4_set_aops(inode);