From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752029AbYDWGvv (ORCPT ); Wed, 23 Apr 2008 02:51:51 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1750792AbYDWGvo (ORCPT ); Wed, 23 Apr 2008 02:51:44 -0400 Received: from pentafluge.infradead.org ([213.146.154.40]:42343 "EHLO pentafluge.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750783AbYDWGvn (ORCPT ); Wed, 23 Apr 2008 02:51:43 -0400 Subject: Re: Second soft lockup regression in yesterday's sched.git merge From: Peter Zijlstra To: David Miller Cc: mingo@elte.hu, linux-kernel@vger.kernel.org, tglx@linutronix.de, guichaz@yahoo.fr In-Reply-To: <20080422.233935.37053549.davem@davemloft.net> References: <20080422.233935.37053549.davem@davemloft.net> Content-Type: text/plain Date: Wed, 23 Apr 2008 08:51:39 +0200 Message-Id: <1208933499.7115.324.camel@twins> Mime-Version: 1.0 X-Mailer: Evolution 2.22.0 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, 2008-04-22 at 23:39 -0700, David Miller wrote: > Two changets after the buggy cpu_clock() changes, we have: > > commit 15934a37324f32e0fda633dc7984a671ea81cd75 > Author: Guillaume Chazarain > Date: Sat Apr 19 19:44:57 2008 +0200 > > sched: fix rq->clock overflows detection with CONFIG_NO_HZ > > When using CONFIG_NO_HZ, rq->tick_timestamp is not updated every TICK_NSEC. > We check that the number of skipped ticks matches the clock jump seen in > __update_rq_clock(). > > Signed-off-by: Guillaume Chazarain > Signed-off-by: Ingo Molnar > > Which also causes softlockup warnings on my Niagara systems. > > Note this is with the cpu_clock() change reverted, and the > usual test case kernel build after a delay: > > sleep 10m; time make -j64 >build.log 2>&1; make -j64 image The effect of this one is that it should properly account idle time in rq->clock when waking from nohz. This 'extra' idle time would then propagate through cpu_clock() into the softlockup code. Could it be we touch the soft watchdog before we correct all these idle times?