From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S933883AbdBQMH6 (ORCPT ); Fri, 17 Feb 2017 07:07:58 -0500 Received: from mail-wr0-f182.google.com ([209.85.128.182]:35839 "EHLO mail-wr0-f182.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932674AbdBQMHf (ORCPT ); Fri, 17 Feb 2017 07:07:35 -0500 From: Matt Fleming To: Peter Zijlstra , Ingo Molnar , Frederic Weisbecker Cc: linux-kernel@vger.kernel.org, Matt Fleming , Mike Galbraith , Morten Rasmussen , stable@vger.kernel.org, Vincent Guittot Subject: [PATCH v2 1/2] sched/loadavg: Avoid loadavg spikes caused by delayed NO_HZ accounting Date: Fri, 17 Feb 2017 12:07:30 +0000 Message-Id: <20170217120731.11868-2-matt@codeblueprint.co.uk> X-Mailer: git-send-email 2.10.0 In-Reply-To: <20170217120731.11868-1-matt@codeblueprint.co.uk> References: <20170217120731.11868-1-matt@codeblueprint.co.uk> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org If we crossed a sample window while in NO_HZ we will add LOAD_FREQ to the pending sample window time on exit, setting the next update not one window into the future, but two. This situation on exiting NO_HZ is described by: this_rq->calc_load_update < jiffies < calc_load_update In this scenario, what we should be doing is: this_rq->calc_load_update = calc_load_update [ next window ] But what we actually do is: this_rq->calc_load_update = calc_load_update + LOAD_FREQ [ next+1 window ] This has the effect of delaying load average updates for potentially up to ~9seconds. This can result in huge spikes in the load average values due to per-cpu uninterruptible task counts being out of sync when accumulated across all CPUs. It's safe to update the per-cpu active count if we wake between sample windows because any load that we left in 'calc_load_idle' will have been zero'd when the idle load was folded in calc_global_load(). This issue is easy to reproduce before, commit 9d89c257dfb9 ("sched/fair: Rewrite runnable load and utilization average tracking") just by forking short-lived process pipelines built from ps(1) and grep(1) in a loop. I'm unable to reproduce the spikes after that commit, but the bug still seems to be present from code review. Fixes: commit 5167e8d ("sched/nohz: Rewrite and fix load-avg computation -- again") Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Morten Rasmussen Cc: Vincent Guittot Cc: Frederic Weisbecker Cc: # v3.5+ Signed-off-by: Matt Fleming --- kernel/sched/loadavg.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) Changes in v2: - Folded in Peter's suggestion for how to fix this. - Tried to clairfy the changelog based on feedback from Peter and Frederic diff --git a/kernel/sched/loadavg.c b/kernel/sched/loadavg.c index a2d6eb71f06b..ec91fcc09bfe 100644 --- a/kernel/sched/loadavg.c +++ b/kernel/sched/loadavg.c @@ -201,8 +201,9 @@ void calc_load_exit_idle(void) struct rq *this_rq = this_rq(); /* - * If we're still before the sample window, we're done. + * If we're still before the pending sample window, we're done. */ + this_rq->calc_load_update = calc_load_update; if (time_before(jiffies, this_rq->calc_load_update)) return; @@ -211,7 +212,6 @@ void calc_load_exit_idle(void) * accounted through the nohz accounting, so skip the entire deal and * sync up for the next window. */ - this_rq->calc_load_update = calc_load_update; if (time_before(jiffies, this_rq->calc_load_update + 10)) this_rq->calc_load_update += LOAD_FREQ; } -- 2.10.0