ext4: fix deadlock in ext4_remount() and orphan list handling

ext4_orphan_add() and ext4_orphan_del() functions lock sb->s_lock with a
transaction started with ext4_mark_recovery_complete() waits for a transaction
holding sb->s_lock, thus leading to a possible deadlock.  At the moment we
call ext4_mark_recovery_complete() from ext4_remount() we have done all the
work needed for remounting and thus we are safe to drop sb->s_lock before we
wait for transactions to commit.  Note that at this moment we are still
guarded by s_umount lock against other remounts/umounts.

Signed-off-by: Jan Kara <jack@suse.cz>
Cc: Eric Sandeen <sandeen@sandeen.net>
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/ext4/super.c

@@ -2222,12 +2222,14 @@ static void ext4_mark_recovery_complete(struct super_block * sb,
 
 	jbd2_journal_lock_updates(journal);
 	jbd2_journal_flush(journal);
+	lock_super(sb);
 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
 	    sb->s_flags & MS_RDONLY) {
 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
 		sb->s_dirt = 0;
 		ext4_commit_super(sb, es, 1);
 	}
+	unlock_super(sb);
 	jbd2_journal_unlock_updates(journal);
 }
 
@@ -2416,7 +2418,13 @@ static int ext4_remount (struct super_block * sb, int * flags, char * data)
 			    (sbi->s_mount_state & EXT4_VALID_FS))
 				es->s_state = cpu_to_le16(sbi->s_mount_state);
 
+			/*
+			 * We have to unlock super so that we can wait for
+			 * transactions.
+			 */
+			unlock_super(sb);
 			ext4_mark_recovery_complete(sb, es);
+			lock_super(sb);
 		} else {
 			__le32 ret;
 			if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,