ext4: fix mmp use after free during unmount

In ext4_put_super, we call brelse on the buffer head containing
the ext4 superblock, but then try to use it when we stop the
mmp thread, because when the thread shuts down it does:

write_mmp_block
  ext4_mmp_csum_set
    ext4_has_metadata_csum
      WARN_ON_ONCE(ext4_has_feature_metadata_csum(sb)...)

which reaches into sb->s_fs_info->s_es->s_feature_ro_compat,
which lives in the superblock buffer s_sbh which we just released.

Fix this by moving the brelse down to a point where we are no
longer using it.

Reported-by: Wang Shu <shuwang@redhat.com>
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>

diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index bb3a8edc75db94c2d4c6079f045f75ede54cd69f..a526956e49e789ba8187944f339311fe2b30aaab 100644 (file)
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -863,7 +863,6 @@ static void ext4_put_super(struct super_block *sb)
        percpu_counter_destroy(&sbi->s_dirs_counter);
        percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
        percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
-       brelse(sbi->s_sbh);
 #ifdef CONFIG_QUOTA
        for (i = 0; i < EXT4_MAXQUOTAS; i++)
                kfree(sbi->s_qf_names[i]);
@@ -895,6 +894,7 @@ static void ext4_put_super(struct super_block *sb)
        }
        if (sbi->s_mmp_tsk)
                kthread_stop(sbi->s_mmp_tsk);
+       brelse(sbi->s_sbh);
        sb->s_fs_info = NULL;
        /*
         * Now that we are completely done shutting down the