From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754193AbbHEPaI (ORCPT ); Wed, 5 Aug 2015 11:30:08 -0400 Received: from bombadil.infradead.org ([198.137.202.9]:50578 "EHLO bombadil.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753465AbbHEPaE (ORCPT ); Wed, 5 Aug 2015 11:30:04 -0400 Date: Wed, 5 Aug 2015 17:29:53 +0200 From: Peter Zijlstra To: Uladzislau Rezki Cc: linux-kernel@vger.kernel.org, Ingo Molnar , Mike Galbraith , Dmitry Adamushko , Srivatsa Vaddagiri , Thomas Gleixner Subject: Re: [PATCH] sched: check pinned tasks before nohz balance, linux-4.2-rc5 Message-ID: <20150805152953.GB25159@twins.programming.kicks-ass.net> References: MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: 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 Wed, Aug 05, 2015 at 05:14:34PM +0200, Uladzislau Rezki wrote: > While investigating “load balancer” logic, I run into interesting > behavior related > to affinity and NO_HZ idle balance. Basically saying if there are two > tasks running > in certain rq and one of them is pinned/affined, then scheduler will > trigger load > balance on power collapsed core via soft irq. > > The corner case here is it may fail to unbalance because of affinity, as a > result core that has two tasks will interrupt any found idle one each its timer > tick, what can lead to wasting of power, especially in mobile area, what is > our case. > > Do you have any concern with provided patch? Yeah, I should go fix the set_cpus_allowed rework :/ lkml.kernel.org/r/20150515154333.712161952@infradead.org