From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758224AbYDNMwQ (ORCPT ); Mon, 14 Apr 2008 08:52:16 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752822AbYDNMwA (ORCPT ); Mon, 14 Apr 2008 08:52:00 -0400 Received: from www.tglx.de ([62.245.132.106]:55731 "EHLO www.tglx.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752600AbYDNMv7 (ORCPT ); Mon, 14 Apr 2008 08:51:59 -0400 Date: Mon, 14 Apr 2008 14:51:50 +0200 (CEST) From: Thomas Gleixner To: Chr cc: "Rafael J. Wysocki" , Linux Kernel Mailing List Subject: Re: [Bug #10369] The never ending BEEEEP/__smp_call_function_mask with 2.6.25-rc7 In-Reply-To: <200804132237.10401.chunkeey@web.de> Message-ID: References: <200804132237.10401.chunkeey@web.de> User-Agent: Alpine 1.10 (LFD 962 2008-03-14) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Sun, 13 Apr 2008, Chr wrote: > On Sunday 13 April 2008 22:02:08 Thomas Gleixner wrote: > > > > The problem still persists and I have no clue at all what's the root > > cause. The symptom is a stuck local apic timer interrupt as we have > > seen it on those AMD X2 systems with C1E enabled. This box does not > > advertise C1E, but the behaviour is the same. > > > > Christian, it would be great if you could bisect the git tree so we > > get an idea which patch is exposing this problem. My feeling is that > > the problem itself is there with 2.6.24 as well but not > > exposed. That's just a gut feeling and I could be wrong as usual. > > Ok! The biggest problem however is: I can only run bisects on weekends. :-( > And even then: it takes up to 2 hours until the timer gets stuck... and the > system freezes... Oh well, that's a bad idea then. > But, on the other hand... does ANYBODY know a way how I can see if the > apictimer drifts/(getting slower and slower)? Well, it does not drift. It's stopped by some magic, but I do not understand why we do not see that behaviour in 2.6.24. The code logic in this area is really unchanged. Can you disable CPUIDLE for a test please ? Thanks, tglx