From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757251AbZCRV5m (ORCPT ); Wed, 18 Mar 2009 17:57:42 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1760229AbZCRVvP (ORCPT ); Wed, 18 Mar 2009 17:51:15 -0400 Received: from hrndva-omtalb.mail.rr.com ([71.74.56.124]:50907 "EHLO hrndva-omtalb.mail.rr.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1760190AbZCRVvO (ORCPT ); Wed, 18 Mar 2009 17:51:14 -0400 Date: Wed, 18 Mar 2009 17:51:11 -0400 (EDT) From: Steven Rostedt X-X-Sender: rostedt@gandalf.stny.rr.com To: LKML cc: Ingo Molnar , Peter Zijlstra , Rusty Russell Subject: Re: [BUG] circular lock dependency in tip In-Reply-To: Message-ID: References: User-Agent: Alpine 2.00 (DEB 1167 2008-08-23) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, 18 Mar 2009, Steven Rostedt wrote: > > > Seems that the issue is that we have: > > static void idle_balance(int this_cpu, struct rq *this_rq) > { > struct sched_domain *sd; > int pulled_task = 0; > unsigned long next_balance = jiffies + HZ; > cpumask_var_t tmpmask; > > if (!alloc_cpumask_var(&tmpmask, GFP_ATOMIC)) > return; Is this really a good idea? I mean, if the system is very active, it can easily fail to acquire memory, you just need to have cache filled. Remember, this is GFP_ATOMIC, if the memory is not available, then we do not try to reclaim. With the system busy, this may be exactly the time we need to do the idle_balance. I'm thinking a blind conversion of cpumask_var_t is not the best way to handle changes. This is as bad as fixing compiler warnings without looking into the real cause. -- Steve