From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753085Ab2DQVRI (ORCPT ); Tue, 17 Apr 2012 17:17:08 -0400 Received: from ogre.sisk.pl ([217.79.144.158]:49133 "EHLO ogre.sisk.pl" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751542Ab2DQVRG (ORCPT ); Tue, 17 Apr 2012 17:17:06 -0400 From: "Rafael J. Wysocki" To: Linus Torvalds , Sven Joachim Subject: Re: kernel panic after suspend/resume (was: Linux 3.4-rc3) Date: Tue, 17 Apr 2012 23:21:40 +0200 User-Agent: KMail/1.13.6 (Linux/3.4.0-rc3+; KDE/4.6.0; x86_64; ; ) Cc: Ingo Molnar , Thomas Gleixner , Linux Kernel Mailing List References: <87pqb6fkji.fsf@turtle.gmx.de> In-Reply-To: MIME-Version: 1.0 Content-Type: Text/Plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <201204172321.41095.rjw@sisk.pl> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tuesday, April 17, 2012, Linus Torvalds wrote: > On Tue, Apr 17, 2012 at 8:24 AM, Sven Joachim wrote: > > > > With Linux 3.4-rc3, I'm experiencing crashes after resuming from > > suspend, not immediately but after a few minutes. This has happened > > three times so far, note that 3.4-rc2 worked fine. > > Hmm. Looks like "global_clock_event->event_handler" is NULL. Which > doesn't make any sense what-so-ever, but clearly it is. > > Added Ingo and Thomas to the cc, since that's a very x86 > timer-looking thing. And Rafael since it's about suspend/resume. I do > wonder if it's some odd memory corruption due to a wild pointer. Of > course, if it's somewhat repeatable, that's some *seriously* odd > corruption, though. So that sounds unlikely too - but that > global_clock_event thing looks odd. > > Oh: guys, one thing to look at is that "lapic_cal_handler" thing. > Weren't there some changes to timer calibration wrt SMP lately? Not in > -rc3, but we had some calibrate_delay() changes - skipping them on > other CPU's when the TSC was reliable, and irq disable things. > > Maybe the calibration at resume now does something different? > > Two questions: > > - if it is reasonably repeatable, can you try to bisect it? There's > just under 400 commits in between rc2 and rc3, and you don't really > need to do a full bisect, but if you do just four bisections, it > should narrow it down to just 25 commits or so. > > - how sure are you that rc2 is fine? I don't see anything suspicious > in this area since rc2, so I would ask you to really test it very well > to make sure it really was introduced after rc2. > > Thomas, Ingo, Rafael - any ideas? Well, commit fa4da365bc7772c kind of looks like it might be the source of this trouble. Sven, can you try to revert it, please? Rafael