Thanks Thomas and Christoph for testing and review.
I removed 'smp_wmb()' before up_write from the previous patch,
since up_write() should have necessary ordering constraints.
(I.e. the change of s_frozen is visible to others after up_write)
I'm quite sure the change is harmless but if you are uncomfortable
with Tested-by/Reviewed-by on the modified patch, please remove them.
If MS_RDONLY, freeze_bdev should just up_write(s_umount) instead of
deactivate_locked_super().
Also, keep sb->s_frozen consistent so that remount can check the frozen state.
Otherwise a crash reported here can happen:
http://lkml.org/lkml/2010/1/16/37
http://lkml.org/lkml/2010/1/28/53
This patch should be applied for 2.6.32 stable series, too.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Tested-by: Thomas Backlund <tmb@mandriva.org>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Cc: stable@kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
        if (!sb)
                goto out;
        if (sb->s_flags & MS_RDONLY) {
-               deactivate_locked_super(sb);
+               sb->s_frozen = SB_FREEZE_TRANS;
+               up_write(&sb->s_umount);
                mutex_unlock(&bdev->bd_fsfreeze_mutex);
                return sb;
        }
        BUG_ON(sb->s_bdev != bdev);
        down_write(&sb->s_umount);
        if (sb->s_flags & MS_RDONLY)
-               goto out_deactivate;
+               goto out_unfrozen;
 
        if (sb->s_op->unfreeze_fs) {
                error = sb->s_op->unfreeze_fs(sb);
                }
        }
 
+out_unfrozen:
        sb->s_frozen = SB_UNFROZEN;
        smp_wmb();
        wake_up(&sb->s_wait_unfrozen);
 
-out_deactivate:
        if (sb)
                deactivate_locked_super(sb);
 out_unlock: