From mboxrd@z Thu Jan 1 00:00:00 1970 From: "F. P. Beekhof" Subject: Re: Machine check exception Date: Wed, 27 Jul 2011 23:30:08 +0200 Message-ID: <4E308360.8080302@gmail.com> References: <4E2FF0EC.2020704@gmail.com> <20110727130319.GA26923@gere.osrc.amd.com> <4E302F6C.3@gmail.com> <20110727170309.GB26923@gere.osrc.amd.com> <4E307AF3.40109@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from mail-ww0-f44.google.com ([74.125.82.44]:54588 "EHLO mail-ww0-f44.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755262Ab1G0VaM (ORCPT ); Wed, 27 Jul 2011 17:30:12 -0400 Received: by wwe5 with SMTP id 5so1835265wwe.1 for ; Wed, 27 Jul 2011 14:30:11 -0700 (PDT) In-Reply-To: <4E307AF3.40109@gmail.com> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: Borislav Petkov Cc: Jeff Garzik , Mikael Pettersson , linux-ide@vger.kernel.org Note: after a suspend/resume cycle, the register value is back at 8, so I have to run the commands again to set it to 100008 # rdmsr -x 0xc001001f 100008 (suspend / resume) # rdmsr -x 0xc001001f 8 On 07/27/2011 10:54 PM, F. P. Beekhof wrote: > On 07/27/2011 07:03 PM, Borislav Petkov wrote: >> On Wed, Jul 27, 2011 at 05:31:56PM +0200, F. P. Beekhof wrote: >>> $ sudo rdmsr 0xc001001f >>> 8 >>> >>> Is there anything else I can do ? >> >> Ok, I'd like you to try something out: >> >> Boot into runlevel 1: you need this because you get the MCE before >> you've done starting apache. For that, add a "1" to your kernel command >> line and boot. When you get the prompt, type in your root pwd and do >> >> wrmsr 0xc001001f $(( $(rdmsr -u 0xc001001f) | (1<< 20) )) >> >> Then do >> >> rdmsr -x 0xc001001f >> >> to verify that the write has succeeded. It should say >> >> 0x00100008. >> >> then exit the prompt to continue to runlevel 2 to see whether this >> setting fixes your MCE issue. >> >> This is for now, >> thanks. >> > > Ok, writing the register worked. > > Now we just need to wait, these crashes occur at random moments. > Sometimes there is 5 minutes between two crashes, sometimes a few days... > > I'll post an update as soon as I have more information. > > Thanks for helping out! >