From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S966085AbXHIPHb (ORCPT ); Thu, 9 Aug 2007 11:07:31 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1765067AbXHIPHW (ORCPT ); Thu, 9 Aug 2007 11:07:22 -0400 Received: from cantor.suse.de ([195.135.220.2]:40819 "EHLO mx1.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1760191AbXHIPHU (ORCPT ); Thu, 9 Aug 2007 11:07:20 -0400 To: Phil Moors Cc: linux-kernel@vger.kernel.org Subject: Re: EDAC k8 MC1: unknown syndrome References: <20070809141333.GA10188@speakeasy.net> From: Andi Kleen Date: 09 Aug 2007 18:01:38 +0200 In-Reply-To: <20070809141333.GA10188@speakeasy.net> Message-ID: User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.3 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Phil Moors writes: > > But is the error really reporting a DIMM problem or a race condition in > the kernel code handling ECC? First if you don't use a mainline kernel you should at least mention it in the report. It could well be a race -- the k8 DAC code is always racing against the builtin machine check handler which reads the same registers. That is why it was never accepted for mainline and shouldn't be used. However these races should normally only cause lost events, not bogus data, so it's probably something the CPU really reported and EDAC doesn't understand. The builtin machine check code will report all the same information through mcelog (on x86-64) or syslog (on i386) -Andi