From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S266380AbUG0P5T (ORCPT ); Tue, 27 Jul 2004 11:57:19 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S266381AbUG0P5T (ORCPT ); Tue, 27 Jul 2004 11:57:19 -0400 Received: from holomorphy.com ([207.189.100.168]:20100 "EHLO holomorphy.com") by vger.kernel.org with ESMTP id S266380AbUG0P5S (ORCPT ); Tue, 27 Jul 2004 11:57:18 -0400 Date: Tue, 27 Jul 2004 08:57:15 -0700 From: William Lee Irwin III To: linux@horizon.com Cc: linux-kernel@vger.kernel.org Subject: Re: 2.6.8-rc2 4K stack overflow Message-ID: <20040727155715.GD2334@holomorphy.com> Mail-Followup-To: William Lee Irwin III , linux@horizon.com, linux-kernel@vger.kernel.org References: <20040727074813.4596.qmail@science.horizon.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20040727074813.4596.qmail@science.horizon.com> User-Agent: Mutt/1.5.6+20040523i Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Tue, Jul 27, 2004 at 07:48:13AM -0000, linux@horizon.com wrote: > Just a crash report. Machine locked hard - no caps lock, no ping. > Machine mostly idle. Amanda network backup was running, but not daily cron. > Backtrace copied by hand. (I didn't copy the leading addresses.) > FWIW, only one partition (a RAID-0 non-critical data partition) is > mounted with ext2. > .config follows. Compiler is GCC 3.3.4 (Debian 3.3.4-3) Hardware is > Intel Celeron, 440BX motherboard. > Hopefully it helps someone. I don't have frame pointers enabled, so I > assume the confusing bits of the backtrace are clutter misidentified as > a return address. It looks like you took invalid opcode exceptions in addition to some rather suspicious oopsen in/around buffered_rmqueue(). The latter are likely stack gunk, the former OTOH I suspect may be real. Could you try to reproduce without CONFIG_REGPARM and turn on frame pointers? Might help a bit to cut down on ways backtraces might get misinterpreted. -- wli