From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754685AbYDHNya (ORCPT ); Tue, 8 Apr 2008 09:54:30 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1751902AbYDHNyX (ORCPT ); Tue, 8 Apr 2008 09:54:23 -0400 Received: from mx3.mail.elte.hu ([157.181.1.138]:54890 "EHLO mx3.mail.elte.hu" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751872AbYDHNyX (ORCPT ); Tue, 8 Apr 2008 09:54:23 -0400 Date: Tue, 8 Apr 2008 15:54:00 +0200 From: Ingo Molnar To: Frans Pop Cc: linux-kernel@vger.kernel.org, Dhaval Giani , "Rafael J. Wysocki" , Srivatsa Vaddagiri , Peter Zijlstra Subject: Re: [rc8] Regression: key repeats and skips in music during compilation Message-ID: <20080408135400.GA18493@elte.hu> References: <200804041222.35025.elendil@planet.nl> <200804071403.59500.elendil@planet.nl> <20080407123234.GA8521@elte.hu> <200804081137.20063.elendil@planet.nl> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <200804081137.20063.elendil@planet.nl> User-Agent: Mutt/1.5.17 (2007-11-01) X-ELTE-VirusStatus: clean X-ELTE-SpamScore: -1.5 X-ELTE-SpamLevel: X-ELTE-SpamCheck: no X-ELTE-SpamVersion: ELTE 2.0 X-ELTE-SpamCheck-Details: score=-1.5 required=5.9 tests=BAYES_00 autolearn=no SpamAssassin version=3.2.3 -1.5 BAYES_00 BODY: Bayesian spam probability is 0 to 1% [score: 0.0000] Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org * Frans Pop wrote: > > and a more laborous way would be to trace such skipping: > > http://people.redhat.com/mingo/sched-devel.git/howto-trace-latencies.txt > > Tried this too, but the problem is that the skips don't happen while > tracing is enabled. Apparently the system overhead is so high that it > changes the scheduling behavior. Attached one trace that was made > immediately after a skip happened. Maybe it helps. one way to reduce tracing overhead would be: echo nostacktrace > /sys/kernel/debug/tracing/iter_ctrl echo nosched-tree > /sys/kernel/debug/tracing/iter_ctrl and making sure you only have CONFIG_CONTEXT_SWITCH_TRACER=y enabled, not any of the other tracers. but ... if there's a narrow wakeup race, even lightweight tracing could make it go away. Ingo