PERFORCE change 148494 for review

[Ed Schouten]

http://perforce.freebsd.org/chv.cgi?CH=148494

Change 148494 by ed at ed_dull on 2008/08/26 06:18:38

	We must check for zero-length queues when actually destroying
	the TTY, not when a close() is being processed.

Affected files ...

.. //depot/projects/mpsafetty/sys/kern/tty.c#32 edit

Differences ...

==== //depot/projects/mpsafetty/sys/kern/tty.c#32 (text+ko) ====

@@ -895,6 +895,10 @@
 	tty_list_count--;
 	sx_xunlock(&tty_list_sx);
 
+	/* Make sure we haven't leaked buffers. */
+	MPASS(ttyinq_getsize(&tp->t_inq) == 0);
+	MPASS(ttyoutq_getsize(&tp->t_outq) == 0);
+
 	knlist_destroy(&tp->t_inpoll.si_note);
 	knlist_destroy(&tp->t_outpoll.si_note);
 
@@ -923,9 +927,6 @@
 		return;
 	}
 
-	MPASS(ttyinq_getsize(&tp->t_inq) == 0);
-	MPASS(ttyoutq_getsize(&tp->t_outq) == 0);
-
 	/* TTY can be deallocated. */
 	dev = tp->t_dev;
 	tp->t_dev = NULL;