From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1030586AbVIPEz2 (ORCPT ); Fri, 16 Sep 2005 00:55:28 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1030595AbVIPEz2 (ORCPT ); Fri, 16 Sep 2005 00:55:28 -0400 Received: from dsl027-180-168.sfo1.dsl.speakeasy.net ([216.27.180.168]:54195 "EHLO sunset.davemloft.net") by vger.kernel.org with ESMTP id S1030586AbVIPEz2 (ORCPT ); Fri, 16 Sep 2005 00:55:28 -0400 Date: Thu, 15 Sep 2005 21:55:27 -0700 (PDT) Message-Id: <20050915.215527.42819293.davem@davemloft.net> To: rdunlap@xenotime.net Cc: linux-kernel@vger.kernel.org Subject: Re: soft lockup disease (2.6.14-rc1, x86_64) From: "David S. Miller" In-Reply-To: References: X-Mailer: Mew version 4.2.53 on Emacs 21.4 / Mule 5.0 (SAKAKI) Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org From: "Randy.Dunlap" Date: Thu, 15 Sep 2005 15:20:03 -0700 (PDT) > (somewhat like http://bugzilla.kernel.org/show_bug.cgi?id=5159 ) > but not IO/Storage related AFAICT, and not xseries. > > It always includes ext3 in the backtrace (from what I have noticed). > [The serial console output appears a bit garbled.] > > Are there patches for this or is it an outstanding issue? I've seen similar triggers on sparc64 from a reporter, but it was NFSD and slab poisioning in the backtrace in that case. I think the common denominator is the presence of very busy kernel daemons unable to schedule out to let other tasks (and in particular the per-cpu softlockup daemon) onto the cpu. Looking at KNFSD specifically, I don't see anywhere that it tries to yield the cpu if it's been working for too long. I guess that's exactly what the softlockup thing was meant to catch :-)