From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S933104Ab3ECH4d (ORCPT ); Fri, 3 May 2013 03:56:33 -0400 Received: from mga02.intel.com ([134.134.136.20]:40020 "EHLO mga02.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932934Ab3ECH4c (ORCPT ); Fri, 3 May 2013 03:56:32 -0400 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="4.87,602,1363158000"; d="scan'208";a="307361323" Message-ID: <51836D74.2030409@intel.com> Date: Fri, 03 May 2013 15:55:32 +0800 From: Alex Shi User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130329 Thunderbird/17.0.5 MIME-Version: 1.0 To: Peter Zijlstra CC: mingo@redhat.com, tglx@linutronix.de, akpm@linux-foundation.org, arjan@linux.intel.com, bp@alien8.de, pjt@google.com, namhyung@kernel.org, efault@gmx.de, morten.rasmussen@arm.com, vincent.guittot@linaro.org, gregkh@linuxfoundation.org, preeti@linux.vnet.ibm.com, viresh.kumar@linaro.org, linux-kernel@vger.kernel.org, len.brown@intel.com, rafael.j.wysocki@intel.com, jkosina@suse.cz, clark.williams@gmail.com, tony.luck@intel.com, keescook@chromium.org, mgorman@suse.de, riel@redhat.com Subject: Re: [PATCH v4 0/6] sched: use runnable load based balance References: <1367040344-14485-1-git-send-email-alex.shi@intel.com> <20130501121421.GA19602@dyad.programming.kicks-ass.net> <5181B579.9000808@intel.com> <20130502103508.GA13196@dyad.programming.kicks-ass.net> In-Reply-To: <20130502103508.GA13196@dyad.programming.kicks-ass.net> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org > That should probably look like: > > preempt_disable(); > raw_spin_unlock_irq(); > preempt_enable_no_resched(); > schedule(); > > Otherwise you might find a performance regression on PREEMPT=y kernels. Yes, right! Thanks a lot for reminder. The following patch will fix it. > > OK, so what I was asking after is if you changed the scheduler after PJTs > patches landed to deal with this bulk wakeup. Also while aim7 might no longer > trigger the bad pattern what is to say nothing ever will? In particular > anything using pthread_cond_broadcast() is known to be suspect of bulk wakeups. Just find a benchmark named as pthread_cond_broadcast. http://kristiannielsen.livejournal.com/13577.html. will play with it. :) > > Anyway, I'll go try and make sense of some of the actual patches.. :-) > --- >>From 4c9b4b8a9b92bcbe6934637fd33c617e73dbda97 Mon Sep 17 00:00:00 2001 From: Alex Shi Date: Fri, 3 May 2013 14:51:25 +0800 Subject: [PATCH 8/8] rwsem: small optimizing rwsem_down_failed_common Peter Zijlstra suggest adding a preempt_enable_no_resched() to prevent a unnecessary scheduler in raw_spin_unlock. And we also can pack 2 raw_spin_lock to save one. So has this patch. Thanks Peter! Signed-off-by: Alex Shi --- lib/rwsem.c | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-) diff --git a/lib/rwsem.c b/lib/rwsem.c index ad5e0df..9aacf81 100644 --- a/lib/rwsem.c +++ b/lib/rwsem.c @@ -212,23 +212,25 @@ rwsem_down_failed_common(struct rw_semaphore *sem, adjustment == -RWSEM_ACTIVE_WRITE_BIAS) sem = __rwsem_do_wake(sem, RWSEM_WAKE_READ_OWNED); - raw_spin_unlock_irq(&sem->wait_lock); - /* wait to be given the lock */ for (;;) { - if (!waiter.task) + if (!waiter.task) { + raw_spin_unlock_irq(&sem->wait_lock); break; + } - raw_spin_lock_irq(&sem->wait_lock); - /* Try to get the writer sem, may steal from the head writer: */ + /* Try to get the writer sem, may steal from the head writer */ if (flags == RWSEM_WAITING_FOR_WRITE) if (try_get_writer_sem(sem, &waiter)) { raw_spin_unlock_irq(&sem->wait_lock); return sem; } + preempt_disable(); raw_spin_unlock_irq(&sem->wait_lock); + preempt_enable_no_resched(); schedule(); set_task_state(tsk, TASK_UNINTERRUPTIBLE); + raw_spin_lock_irq(&sem->wait_lock); } tsk->state = TASK_RUNNING; -- 1.7.12