From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1161222AbWF0RKj (ORCPT ); Tue, 27 Jun 2006 13:10:39 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1161224AbWF0RKj (ORCPT ); Tue, 27 Jun 2006 13:10:39 -0400 Received: from scrub.xs4all.nl ([194.109.195.176]:13960 "EHLO scrub.xs4all.nl") by vger.kernel.org with ESMTP id S1161222AbWF0RKh (ORCPT ); Tue, 27 Jun 2006 13:10:37 -0400 Date: Tue, 27 Jun 2006 19:10:32 +0200 (CEST) From: Roman Zippel X-X-Sender: roman@scrub.home To: Valdis.Kletnieks@vt.edu cc: Andrew Morton , linux-kernel@vger.kernel.org Subject: Re: 2.6.17-mm2 hrtimer code wedges at boot? In-Reply-To: <200606271643.k5RGh9ZQ004498@turing-police.cc.vt.edu> Message-ID: References: <20060624061914.202fbfb5.akpm@osdl.org> <200606262141.k5QLf7wi004164@turing-police.cc.vt.edu> <200606271643.k5RGh9ZQ004498@turing-police.cc.vt.edu> 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 Hi, On Tue, 27 Jun 2006, Valdis.Kletnieks@vt.edu wrote: > On Tue, 27 Jun 2006 12:16:15 +0200, Roman Zippel said: > > > Could you please try the patch below? > > tv_nsec can shortly become negative, but its absolute value will always be > > smaller then the current nsec offset. > > > Index: linux-2.6-mm/kernel/timer.c > > =================================================================== > > --- linux-2.6-mm.orig/kernel/timer.c 2006-06-27 11:59:19.000000000 +0200 > > +++ linux-2.6-mm/kernel/timer.c 2006-06-27 12:10:28.000000000 +0200 > > @@ -1129,7 +1129,7 @@ static void update_wall_time(void) > > clocksource_adjust(clock, offset); > > > > /* store full nanoseconds into xtime */ > > - xtime.tv_nsec = clock->xtime_nsec >> clock->shift; > > + xtime.tv_nsec = (s64)clock->xtime_nsec >> clock->shift; > > clock->xtime_nsec -= (s64)xtime.tv_nsec << clock->shift; > > > > /* check to see if there is a new clocksource to use */ > > Sorry Roman... This may indeed be a legitimate bugfix, but it doesn't > fix the problem I'm seeing. I've been doing the mm-bisect polka for a bit, > and have it narrowed down to this set of patches: I'm afraid the problem is somehow related, in the longer boot log one can see the edi counting down, so I think it's likely in timespec_add_ns(). What's weird is that you can trigger it this easily... bye, Roman