From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753835AbZHIKDv (ORCPT ); Sun, 9 Aug 2009 06:03:51 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1753802AbZHIKDv (ORCPT ); Sun, 9 Aug 2009 06:03:51 -0400 Received: from bar.sig21.net ([80.81.252.164]:58796 "EHLO bar.sig21.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753788AbZHIKDu (ORCPT ); Sun, 9 Aug 2009 06:03:50 -0400 Date: Sun, 9 Aug 2009 12:03:48 +0200 From: Johannes Stezenbach To: x86@kernel.org Cc: linux-kernel@vger.kernel.org, "Rafael J. Wysocki" Subject: Re: 2.6.31-rc5 regression: x86 MCE malfunction on Thinkpad T42p Message-ID: <20090809100348.GA7186@sig21.net> References: <20090807170942.GB9177@sig21.net> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20090807170942.GB9177@sig21.net> User-Agent: Mutt/1.5.20 (2009-06-14) X-Spam-21-Score: -3.8 (---) X-Spam-21-Report: No, score=-3.8 required=5.0 tests=ALL_TRUSTED=-1.8,AWL=0.568,BAYES_00=-2.599 autolearn=no Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, Aug 07, 2009 at 07:09:42PM +0200, Johannes Stezenbach wrote: > > I'm currently running linux-2.6.31-rc5-246-g90bc1a6 on > an old Thinkpad T42p. During boot I get the following: ... > I guess I should try to boot with "lapic"? But I think > MCE worked without "lapic" in earlier kernels. On a 2.6.29.1 > kernel dmesg said: > > Local APIC disabled by BIOS -- you can enable it with "lapic" > ... > Intel machine check architecture supported. > Intel machine check reporting enabled on CPU#0. > > 2.6.29.1 doesn't log any MCE events, so I doubt this is a HW problem. I booted with "lapic", then the backtrace is gone but it still logs machine check events which I think are bogus since 2.6.29.1 does not log any. I also tried 2.6.30, no machine check messages in dmesg. But I noticed that mcelog complains about missing /dev/mcelog, it seems that /dev/mcelog support for 32bit kernels is new. However, I guess the old kernels should still have printed a messge to dmesg, right? "Uncorrected error" + "Processor context corrupt" sounds pretty serious, but the machine runs without problems. These are from 2.6.31-rc5 by running mcelog (dmesg just says "Machine check events logged"): HARDWARE ERROR. This is *NOT* a software problem! Please contact your hardware vendor MCE 0 CPU 0 BANK 1 TIME 1249725930 Sat Aug 8 12:05:30 2009 MCG status: MCi status: Error overflow Uncorrected error Error enabled Processor context corrupt MCA: MEMORY CONTROLLER AC_CHANNEL0_ERR Transaction: Address/Command error STATUS f2000000000000b0 MCGSTATUS 0 MCGCAP 5 APICID 0 SOCKETID 0 CPUID Vendor Intel Family 6 Model 13 HARDWARE ERROR. This is *NOT* a software problem! Please contact your hardware vendor MCE 1 CPU 0 BANK 1 TIME 1249728066 Sat Aug 8 12:41:06 2009 MCG status: MCi status: Error overflow Uncorrected error Error enabled Processor context corrupt MCA: Unknown Error 20 STATUS f200000000000020 MCGSTATUS 0 MCGCAP 5 APICID 0 SOCKETID 0 CPUID Vendor Intel Family 6 Model 13 HARDWARE ERROR. This is *NOT* a software problem! Please contact your hardware vendor MCE 2 CPU 0 BANK 1 TIME 1249747923 Sat Aug 8 18:12:03 2009 MCG status: MCi status: Error overflow Uncorrected error Error enabled Processor context corrupt MCA: Unknown Error 30 STATUS f200000000000030 MCGSTATUS 0 MCGCAP 5 APICID 0 SOCKETID 0 CPUID Vendor Intel Family 6 Model 13 HARDWARE ERROR. This is *NOT* a software problem! Please contact your hardware vendor MCE 3 CPU 0 BANK 1 TIME 1249765938 Sat Aug 8 23:12:18 2009 MCG status: MCi status: Error overflow Uncorrected error Error enabled Processor context corrupt MCA: Unknown Error 20 STATUS f200000000000020 MCGSTATUS 0 MCGCAP 5 APICID 0 SOCKETID 0 CPUID Vendor Intel Family 6 Model 13 Johannes