From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754302AbaDOLO1 (ORCPT ); Tue, 15 Apr 2014 07:14:27 -0400 Received: from merlin.infradead.org ([205.233.59.134]:54537 "EHLO merlin.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753925AbaDOLOZ (ORCPT ); Tue, 15 Apr 2014 07:14:25 -0400 Date: Tue, 15 Apr 2014 13:14:10 +0200 From: Peter Zijlstra To: Denys Vlasenko Cc: linux-kernel@vger.kernel.org, Frederic Weisbecker , Hidetoshi Seto , Fernando Luis Vazquez Cao , Tetsuo Handa , Thomas Gleixner , Ingo Molnar , Andrew Morton , Arjan van de Ven , Oleg Nesterov Subject: Re: [PATCH 2/2] nohz: Synchronize sleep time stats with memory barriers Message-ID: <20140415111410.GN13658@twins.programming.kicks-ass.net> References: <1396469872-14877-1-git-send-email-dvlasenk@redhat.com> <1396469872-14877-2-git-send-email-dvlasenk@redhat.com> <20140415105138.GO11096@twins.programming.kicks-ass.net> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20140415105138.GO11096@twins.programming.kicks-ass.net> User-Agent: Mutt/1.5.21 (2012-12-30) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, Apr 15, 2014 at 12:51:38PM +0200, Peter Zijlstra wrote: > That said; I thing the fundamental flaw in the entire thing is > accounting the entire nohz idle period as one type, since the period is > basically unbounded. So basically anything with the current per-cpu nr_iowait is bound to yield crack induced results. Like you said in your other email, we need to somehow intersect actual idle time with the presence of iowait tasks. Now that's a global problem and is unlikely to scale well. Which is of course why we have this problem to begin with, if it were easy it'd be done right to begin with (optimistic world view here).