From owner-freebsd-current  Thu Mar 12 09:57:43 1998
Return-Path: <owner-freebsd-current@FreeBSD.ORG>
Received: (from majordom@localhost)
          by hub.freebsd.org (8.8.8/8.8.8) id JAA24807
          for freebsd-current-outgoing; Thu, 12 Mar 1998 09:57:43 -0800 (PST)
          (envelope-from owner-freebsd-current@FreeBSD.ORG)
Received: from pat.idi.ntnu.no (0@pat.idi.ntnu.no [129.241.103.5])
          by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id JAA24758;
          Thu, 12 Mar 1998 09:57:31 -0800 (PST)
          (envelope-from Tor.Egge@idi.ntnu.no)
Received: from idi.ntnu.no (tegge@presis.idi.ntnu.no [129.241.111.173])
	by pat.idi.ntnu.no (8.8.8/8.8.8) with ESMTP id SAA05731;
	Thu, 12 Mar 1998 18:53:43 +0100 (MET)
Message-Id: <199803121753.SAA05731@pat.idi.ntnu.no>
To: root@mantar.slip.netcom.com
Cc: current@FreeBSD.ORG, smp@FreeBSD.ORG
Subject: Re: spkrtest locks up system
In-Reply-To: Your message of "Wed, 11 Mar 1998 18:31:52 -0800 (PST)"
References: <Pine.BSF.3.96.980311181947.343A-100000@mantar.slip.netcom.com>
X-Mailer: Mew version 1.70 on Emacs 19.34.1
Mime-Version: 1.0
Content-Type: Text/Plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Date: Thu, 12 Mar 1998 18:53:42 +0100
From: Tor Egge <Tor.Egge@idi.ntnu.no>
Sender: owner-freebsd-current@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.ORG

> On a current SMP system when I run spkrtest the system locks up.
> On a kernel built around the 26th of Feb no problem. I have a old version
> of spkrtest   v1.0 by Eric S. Raymond (Feb 1990) that I use as sort of
> a backup alarm clock. I was wondering why for the past couple of weeks
> I'd wake up and the machine would be locked up. The current graphical
> version does the same thing. I have no sound card installed. One time
> it dropped into the debugger , now it just locks up. There was something 
> about _mplock if i remember right. There was alot of changes to current
> around the end of Feb. (SMP and locking) and I'm not sure which is causing 
> the problem. Sorry I can't be of more help.
> Manfred  

The following patch might be appropiate.  

Without this patch, the kernel can deadlock or panic when attempting
to enter a critical region protected by disable_intr()/enable_intr(),
since the mpintr_lock might be held by the last cpu having prezeroed
a page.

Index: sys/i386/i386/vm_machdep.c
===================================================================
RCS file: /home/ncvs/src/sys/i386/i386/vm_machdep.c,v
retrieving revision 1.101
diff -u -r1.101 vm_machdep.c
--- vm_machdep.c	1998/02/25 03:56:09	1.101
+++ vm_machdep.c	1998/03/08 06:00:49
@@ -950,7 +967,7 @@
 	if (try_mplock()) {
 #endif
 		s = splvm();
-		enable_intr();
+		__asm __volatile("sti" : : : "memory");
 		m = vm_page_list_find(PQ_FREE, free_rover);
 		if (m != NULL) {
 			--(*vm_page_queues[m->queue].lcnt);
@@ -973,7 +990,7 @@
 			++cnt_prezero;
 		}
 		splx(s);
-		disable_intr();
+		__asm __volatile("cli" : : : "memory");
 #ifdef SMP
 		rel_mplock();
 		return (1);


- Tor Egge

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message