tty_open can return to userspace holding tty_mutex

__tty_open could return (to userspace) holding the tty_mutex thanks to a
regression introduced by 4a2b5fd ("Move
tty lookup/reopen to caller").

This was found by bisecting an fsfuzzer problem.  Admittedly I have no
idea how it managed to tickle this 100% reliably, but it is clearly a
regression and when hit leaves the box in a completely unusable state.
This patch lets the fsfuzzer test complete every time.

Signed-off-by: Eric Paris <eparis@redhat.com>
Signed-off-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/char/tty_io.c

@@ -1817,8 +1817,10 @@ static int __tty_open(struct inode *inode, struct file *filp)
 		/* check whether we're reopening an existing tty */
 		tty = tty_driver_lookup_tty(driver, inode, index);
 
-		if (IS_ERR(tty))
+		if (IS_ERR(tty)) {
+			mutex_unlock(&tty_mutex);
 			return PTR_ERR(tty);
+		}
 	}
 
 	if (tty) {