From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755674Ab0A1M4G (ORCPT ); Thu, 28 Jan 2010 07:56:06 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1754125Ab0A1M4E (ORCPT ); Thu, 28 Jan 2010 07:56:04 -0500 Received: from ozlabs.org ([203.10.76.45]:53014 "EHLO ozlabs.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752353Ab0A1M4B (ORCPT ); Thu, 28 Jan 2010 07:56:01 -0500 Date: Thu, 28 Jan 2010 23:52:02 +1100 From: Anton Blanchard To: Peter Zijlstra , Andrew Morton Cc: Balbir Singh , mingo@redhat.com, hpa@zytor.com, linux-kernel@vger.kernel.org, tglx@linutronix.de, mingo@elte.hu, linux-tip-commits@vger.kernel.org Subject: [PATCH 2/2] sched: cpuacct: Use bigger percpu counter batch values for stats counters Message-ID: <20100128125202.GG2996@kryten> References: <20100118044142.GS12666@kryten> <661de9471001270534ve6f7bf0nb58ef58f9a6ff10e@mail.gmail.com> <20100127132212.cd43cd94.akpm@linux-foundation.org> <1264628600.4283.2080.camel@laptop> <20100128124854.GF2996@kryten> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20100128124854.GF2996@kryten> User-Agent: Mutt/1.5.20 (2009-06-14) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org When CONFIG_VIRT_CPU_ACCOUNTING and CONFIG_CGROUP_CPUACCT are enabled we can call cpuacct_update_stats with values much larger than percpu_counter_batch. This means the call to percpu_counter_add will always add to the global count which is protected by a spinlock and we end up with a global spinlock in the scheduler. Based on an idea by KOSAKI Motohiro, this patch scales the batch value by cputime_one_jiffy such that we have the same batch limit as we would if CONFIG_VIRT_CPU_ACCOUNTING was disabled. His patch did this once at boot but that initialisation happened too early on PowerPC (before time_init) and it was never updated at runtime as a result of a hotplug cpu add/remove. This patch instead scales percpu_counter_batch by cputime_one_jiffy at runtime, which keeps the batch correct even after cpu hotplug operations. We cap it at INT_MAX in case of overflow. For architectures that do not support CONFIG_VIRT_CPU_ACCOUNTING, cputime_one_jiffy is the constant 1 and gcc is smart enough to optimise min(s32 percpu_counter_batch, INT_MAX) to just percpu_counter_batch at least on x86 and PowerPC. So there is no need to add an #ifdef. On a 64 thread PowerPC box with CONFIG_VIRT_CPU_ACCOUNTING and CONFIG_CGROUP_CPUACCT enabled, a context switch microbenchmark is 234x faster and almost matches a CONFIG_CGROUP_CPUACCT disabled kernel: CONFIG_CGROUP_CPUACCT disabled: 16906698 ctx switches/sec CONFIG_CGROUP_CPUACCT enabled: 61720 ctx switches/sec CONFIG_CGROUP_CPUACCT + patch: 16663217 ctx switches/sec Tested with: wget http://ozlabs.org/~anton/junkcode/context_switch.c make context_switch for i in `seq 0 63`; do taskset -c $i ./context_switch & done vmstat 1 Signed-off-by: Anton Blanchard --- v2: Added a comment and fixed the UP build. Index: linux-cpumask/kernel/sched.c =================================================================== --- linux-cpumask.orig/kernel/sched.c 2010-01-22 18:23:43.377212514 +1100 +++ linux-cpumask/kernel/sched.c 2010-01-28 23:24:02.677233753 +1100 @@ -10885,12 +10885,30 @@ static void cpuacct_charge(struct task_s } /* + * When CONFIG_VIRT_CPU_ACCOUNTING is enabled one jiffy can be very large + * in cputime_t units. As a result, cpuacct_update_stats calls + * percpu_counter_add with values large enough to always overflow the + * per cpu batch limit causing bad SMP scalability. + * + * To fix this we scale percpu_counter_batch by cputime_one_jiffy so we + * batch the same amount of time with CONFIG_VIRT_CPU_ACCOUNTING disabled + * and enabled. We cap it at INT_MAX which is the largest allowed batch value. + */ +#ifdef CONFIG_SMP +#define CPUACCT_BATCH \ + min_t(long, percpu_counter_batch * cputime_one_jiffy, INT_MAX) +#else +#define CPUACCT_BATCH 0 +#endif + +/* * Charge the system/user time to the task's accounting group. */ static void cpuacct_update_stats(struct task_struct *tsk, enum cpuacct_stat_index idx, cputime_t val) { struct cpuacct *ca; + int batch = CPUACCT_BATCH; if (unlikely(!cpuacct_subsys.active)) return; @@ -10899,7 +10917,7 @@ static void cpuacct_update_stats(struct ca = task_ca(tsk); do { - percpu_counter_add(&ca->cpustat[idx], val); + __percpu_counter_add(&ca->cpustat[idx], val, batch); ca = ca->parent; } while (ca); rcu_read_unlock();