From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753316Ab3BFLf1 (ORCPT ); Wed, 6 Feb 2013 06:35:27 -0500 Received: from mail-ie0-f175.google.com ([209.85.223.175]:37903 "EHLO mail-ie0-f175.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751032Ab3BFLfZ (ORCPT ); Wed, 6 Feb 2013 06:35:25 -0500 MIME-Version: 1.0 In-Reply-To: <20130205164520.GA4276@mtj.dyndns.org> References: <1359657696-2767-1-git-send-email-laijs@cn.fujitsu.com> <1359657696-2767-4-git-send-email-laijs@cn.fujitsu.com> <20130204212808.GC27963@mtj.dyndns.org> <51111E9A.4000504@gmail.com> <20130205164520.GA4276@mtj.dyndns.org> Date: Wed, 6 Feb 2013 19:35:24 +0800 Message-ID: Subject: Re: [PATCH 03/13] workqueue: don't set work cwq until we queued it on pool From: Lai Jiangshan To: Tejun Heo Cc: Lai Jiangshan , linux-kernel@vger.kernel.org Content-Type: text/plain; charset=ISO-8859-1 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Feb 6, 2013 at 12:45 AM, Tejun Heo wrote: > Hey, Lai. > > On Tue, Feb 05, 2013 at 11:00:42PM +0800, Lai Jiangshan wrote: >> The @cpu can be WORK_CPU_UNBOUND, and the current code will delay >> to determine on which cpu should this work be queued until timeout, >> I didn't want to change this behavior when I wrote the patch. > > Hmmm? The current code determines the CPU on queue_delayed_work(). > > /* > * If we cannot get the last pool from @work directly, > * select the last CPU such that it avoids unnecessarily > * triggering non-reentrancy check in __queue_work(). > */ > lcpu = cpu; > if (pool) > lcpu = pool->cpu; > if (lcpu == WORK_CPU_UNBOUND) > lcpu = raw_smp_processor_id(); > > @lcpu is never WORK_CPU_UNBOUND for bound workqueues and for unbound > workqueues we of course have cwq for WORK_CPU_UNBOUND. Are you sure? @lcpu is only used when it tries to set cwq to work data without breaking nrt logic. void delayed_work_timer_fn(unsigned long __data) { struct delayed_work *dwork = (struct delayed_work *)__data; struct cpu_workqueue_struct *cwq = get_work_cwq(&dwork->work); /* should have been called from irqsafe timer with irq already off */ __queue_work(dwork->cpu, cwq->wq, &dwork->work); } > >> PS, again, @cpu and @wq can be saved/encoded in dwork.work.entry if needed. > > If possible, I'd much prefer to keep it inside delayed_work() proper. > > Thanks. > > -- > tejun