From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755674AbYGZNWV (ORCPT ); Sat, 26 Jul 2008 09:22:21 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1751808AbYGZNWN (ORCPT ); Sat, 26 Jul 2008 09:22:13 -0400 Received: from mx2.mail.elte.hu ([157.181.151.9]:58096 "EHLO mx2.mail.elte.hu" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751600AbYGZNWN (ORCPT ); Sat, 26 Jul 2008 09:22:13 -0400 Date: Sat, 26 Jul 2008 15:21:18 +0200 From: Ingo Molnar To: Steven Rostedt Cc: Peter Zijlstra , Thomas Gleixner , Andrew Morton , linux-kernel@vger.kernel.org, Linus Torvalds , Steven Rostedt Subject: Re: [PATCH 1/2] ftrace: single CPU tracers use CPU clock Message-ID: <20080726132118.GA30903@elte.hu> References: <20080725220040.911370182@goodmis.org> <20080725220334.781928198@goodmis.org> <20080726124058.GD20713@elte.hu> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.5.18 (2008-05-17) 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 * Steven Rostedt wrote: > > this is not a good idea. We want to fix cpu_clock(), not work around > > any deficiencies it might have. > > cpu_clock currently is "sched_clock" when the > CONFIG_HAVE_UNSTABLE_SCHED_CLOCK is not set. Which is done via > configuration, and I noticed is set on my boxes with a stable TSC?? define 'stable TSC' ;-) > Perhaps we need to make cpu_clock change dynamically when an unstable > sched_clock is detected. > > Even in this case, forcing the tracers that are single CPU to use a > clock source that modifies itself to try to look stable across CPUs > still seems wrong to me. The goal of looking stable across CPUs will > always be at odds with the irqsoff tracer that does not care about > other CPUS but cares tremendously about accurate latencies. other tracers care too - for example to have the right chronology of trace events. The scheduler cares too. What kind of worst-case cross-CPU effects have you observed? Ingo