From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1761772AbXF0Qaj (ORCPT ); Wed, 27 Jun 2007 12:30:39 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1753192AbXF0Qac (ORCPT ); Wed, 27 Jun 2007 12:30:32 -0400 Received: from e5.ny.us.ibm.com ([32.97.182.145]:47950 "EHLO e5.ny.us.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752209AbXF0Qab (ORCPT ); Wed, 27 Jun 2007 12:30:31 -0400 Date: Wed, 27 Jun 2007 09:30:25 -0700 From: "Paul E. McKenney" To: Ingo Molnar Cc: Andrew Morton , linux-kernel@vger.kernel.org, josh@freedesktop.org, dipankar@in.ibm.com Subject: Re: [PATCH] Force rcutorture tasks to spread over CPUs Message-ID: <20070627163025.GB8604@linux.vnet.ibm.com> Reply-To: paulmck@linux.vnet.ibm.com References: <20070613042804.GA6132@linux.vnet.ibm.com> <20070626161317.767d5ec0.akpm@linux-foundation.org> <20070627063055.GB3186@elte.hu> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20070627063055.GB3186@elte.hu> User-Agent: Mutt/1.5.13 (2006-08-11) Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Jun 27, 2007 at 08:30:55AM +0200, Ingo Molnar wrote: > > * Andrew Morton wrote: > > > On Tue, 12 Jun 2007 21:28:04 -0700 > > "Paul E. McKenney" wrote: > > > > > + while (!startwriters) > > > + barrier(); /* Force scheduler to spread over CPUs. */ > > > > one wonders whether a cpu_relax() would be a bit nicer here. That > > implicitly does a barrier(). > > > > This patch doesn't make much sense for non-SMP builds? > > i think this patch should be unnecessary because we found the real SMP > balancing bug in the upstream scheduler causing this rcu problem, see: > > commit 92c4ca5c3a5e180e9762438db235f41d192cb955 > Author: Christoph Lameter > Date: Sat Jun 23 17:16:33 2007 -0700 > > sched: fix next_interval determination in idle_balance() Ingo is correct -- applying the above patch caused the scheduler to correctly balance the rcutorture tasks, so that my patch to rcutorture is no longer needed. Which is a very good thing! ;-) Thanx, Paul