From owner-freebsd-current@FreeBSD.ORG Tue Jul 20 15:38:02 2004 Return-Path: Delivered-To: freebsd-current@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 031E216A4CE; Tue, 20 Jul 2004 15:38:02 +0000 (GMT) Received: from postal3.es.net (postal3.es.net [198.128.3.207]) by mx1.FreeBSD.org (Postfix) with ESMTP id E278743D54; Tue, 20 Jul 2004 15:38:01 +0000 (GMT) (envelope-from oberman@es.net) Received: from ptavv.es.net ([198.128.4.29]) by postal3.es.net (Postal Node 3) with ESMTP (SSL) id IBA74465; Tue, 20 Jul 2004 08:38:01 -0700 Received: from ptavv (localhost [127.0.0.1]) by ptavv.es.net (Tachyon Server) with ESMTP id 4CA0A5D09; Tue, 20 Jul 2004 08:38:00 -0700 (PDT) To: Brian Fundakowski Feldman In-reply-to: Your message of "Tue, 20 Jul 2004 11:20:44 EDT." <20040720152044.GB1009@green.homeunix.org> Date: Tue, 20 Jul 2004 08:38:00 -0700 From: "Kevin Oberman" Message-Id: <20040720153800.4CA0A5D09@ptavv.es.net> cc: freebsd-current@freebsd.org cc: freebsd-gnome@freebsd.org Subject: Re: [amd64] System lockups still continuing X-BeenThere: freebsd-current@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list List-Id: Discussions about the use of FreeBSD-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 20 Jul 2004 15:38:02 -0000 > Date: Tue, 20 Jul 2004 11:20:44 -0400 > From: Brian Fundakowski Feldman > > On Tue, Jul 20, 2004 at 05:32:20AM -0700, Kevin Oberman wrote: > > > Date: Mon, 19 Jul 2004 16:09:19 -0500 (CDT) > > > From: "Conrad J. Sabatier" > > > > > > > > > On 19-Jul-2004 Kevin Oberman wrote: > > > >> > > > >> On 18-Jul-2004 Conrad J. Sabatier wrote: > > > >> > I had posted several weeks ago about constant hard lockups of my > > > >> > amd64 box. At first, I attributed the problem to GNOME, then to > > > >> > my kernel config, but now it appears to me that the problem > > > >> > really lies with Mozilla/Firefox. > > > >> > > > >> Well, another theory bites the dust. I tried running GNOME > > > >> yesterday, never touched Mozilla or Firefox, and the system locked > > > >> up while doing some file management chores in Nautilus. > > > >> > > > >> It must be something low-level in GNOME, some shared library, > > > >> probably, but I have yet to track it down. > > > > > > > > I have lost track of the full thread and this mail list is not > > > > tracked by google :-(, so this may have been ruled out. Are you sure > > > > that it's not the preemption problem that is biting you? This applies > > > > to CURRENT systems only, so may not be applicable to what you are > > > > seeing, the symptoms are a match. > > > > > > Hmm, you may be right. Is there any workaround for this? I've tried > > > both SCHED_ULE and SCHED_4BSD, both producing the same results. > > > > > > What's strange is that this only seems to occur when either running a > > > full-blown GNOME desktop or using some GNOME-related app under another > > > environment. > > > > > > I'm currently working under KDE, just for comparison, and no such > > > problems have occurred, unless I try using, say, Mozilla or Firefox. > > > > > > I'll Cc: this to current and see what they think. > > > > The problem is not specific to ULE or 4BSD. A fix was checked in last > > night, so cvsup and rebuild might just fix it. Early reports are that > > the patch resolves the problem, but the author is not sure that this is > > a complete fix. > > It is not a complete fix in i386/SMP, but seems to make it harder to > cause the lock-up. I am not able to cause the lock-up at all simply > by disabling PREEMPTION in src/sys/param.h. Brian, Thanks for the information. In my case it's i386/UP. I am currently building a new kernel and will try it out soon. When I had the problem, it showed up reliably in a buildworld, so I will fire off one as soon as the new kernel is up. I tried disabling PREEMPTION, but the system was barely usable whenever it was running anything that was CPU intensive (such as compiling a port), so I backed out the changes instead. While this works, it is something I want to get away from ASAP. -- R. Kevin Oberman, Network Engineer Energy Sciences Network (ESnet) Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab) E-mail: oberman@es.net Phone: +1 510 486-8634