From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754853AbdDMOyo (ORCPT ); Thu, 13 Apr 2017 10:54:44 -0400 Received: from mail.kernel.org ([198.145.29.136]:43004 "EHLO mail.kernel.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754825AbdDMOyk (ORCPT ); Thu, 13 Apr 2017 10:54:40 -0400 Date: Thu, 13 Apr 2017 10:54:35 -0400 From: Steven Rostedt To: Peter Zijlstra Cc: LKML , Ingo Molnar , Andrew Morton , "Paul E. McKenney" , Thomas Gleixner Subject: Re: [PATCH v2] sched: Have do_idle() call __schedule() without enabling preemption Message-ID: <20170413105435.4aa3d012@gandalf.local.home> In-Reply-To: <20170413143807.gukvroojodipwuvg@hirez.programming.kicks-ass.net> References: <20170412142744.6ff07236@gandalf.local.home> <20170413084453.ad4falb7kc53viri@hirez.programming.kicks-ass.net> <20170413094840.347b9829@gandalf.local.home> <20170413143807.gukvroojodipwuvg@hirez.programming.kicks-ass.net> X-Mailer: Claws Mail 3.14.0 (GTK+ 2.24.31; x86_64-pc-linux-gnu) MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, 13 Apr 2017 16:38:07 +0200 Peter Zijlstra wrote: > On Thu, Apr 13, 2017 at 09:48:40AM -0400, Steven Rostedt wrote: > > On Thu, 13 Apr 2017 10:44:53 +0200 > > Peter Zijlstra wrote: > > > > > On Wed, Apr 12, 2017 at 02:27:44PM -0400, Steven Rostedt wrote: > > > > + * schedule_idle() is similar to schedule_preempt_disable() except > > > > + * that it never enables preemption. > > > > > > That's not right. The primary distinction is that it doesn't call > > > sched_submit_work(). > > > > That has nothing to do with fixing synchronize_rcu_tasks(), which is > > the entire point of my patch, thus it is *not* the primary distinction. > > It is. The only reason you _can_ avoid that preemption window is by not > doing sched_submit_work(). We are talking apples and oranges here. It's a distinction of the generic call to schedule_preempt_disable() and only one instance of the callers can actually run that. But it is not the idle case, in which case it is a nop. But we are now bike shedding the issue. > > The moment you need to call that, the preemption window comes back. I'll update my change log and comment and repost. Thanks, -- Steve