Back out part of rev. 1.149. While adding a workaround in ptcopen() to

avoid leaked ptys works fine, this opens a possible security hole.

Submitted by:	bde
MFC after:	3 days

1 files changed, 2 insertions, 9 deletions

diff --git a/sys/kern/tty_pty.c b/sys/kern/tty_pty.c
index 2f1b2eb..a973f6a 100644
--- a/sys/kern/tty_pty.c
+++ b/sys/kern/tty_pty.c
@@ -335,15 +335,8 @@ ptcopen(struct cdev *dev, int flag, int devtype, struct thread *td)
 	}
 	tp = dev->si_tty;
 
-	if (tp->t_oproc) {
-		/*
-		 * Only return if we have a non empty
-		 * state to avoid leakage. Workaround for
-		 * vfs bugs and complications near revoke().
-		 */
-		if (tp->t_state != 0x0 || tp->t_refcnt > 1)
-			return (EIO);
-	}
+	if (tp->t_oproc)
+		return (EIO);
 	tp->t_timeout = -1;
 	tp->t_oproc = ptsstart;
 	tp->t_stop = ptsstop;