From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758180AbaD3KSs (ORCPT ); Wed, 30 Apr 2014 06:18:48 -0400 Received: from hqemgate16.nvidia.com ([216.228.121.65]:11080 "EHLO hqemgate16.nvidia.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751330AbaD3KSq (ORCPT ); Wed, 30 Apr 2014 06:18:46 -0400 X-PGP-Universal: processed; by hqnvupgp07.nvidia.com on Wed, 30 Apr 2014 03:09:23 -0700 From: Peter De Schrijver To: Peter De Schrijver CC: Sai Gurrappadi , Thomas Gleixner , Subject: [PATCH] tick: Don't clear idle and iowait sums on CPU down Date: Wed, 30 Apr 2014 13:18:34 +0300 Message-ID: <1398853114-27544-1-git-send-email-pdeschrijver@nvidia.com> X-Mailer: git-send-email 1.7.7.rc0.72.g4b5ea.dirty X-NVConfidentiality: public MIME-Version: 1.0 Content-Type: text/plain Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org From: Sai Gurrappadi NOHZ related per-cpu data is cleared on CPU down. This was introduced by 4b0c0f294 "tick: Cleanup NOHZ per cpu data on cpu down" which breaks /proc/stats because the idle and iowait sums are now non-monotonic across a CPU down/up cycle. Fix this by not clearing the idle_sleeptime and iowait_sleeptime fields on CPU down. Signed-off-by: Sai Gurrappadi Signed-off-by: Peter De Schrijver --- kernel/time/tick-sched.c | 13 ++++++++++++- 1 files changed, 12 insertions(+), 1 deletions(-) diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c index c12c006..2cd27ec 100644 --- a/kernel/time/tick-sched.c +++ b/kernel/time/tick-sched.c @@ -1164,6 +1164,17 @@ void tick_setup_sched_timer(void) #endif /* HIGH_RES_TIMERS */ #if defined CONFIG_NO_HZ_COMMON || defined CONFIG_HIGH_RES_TIMERS + +static inline void clear_tick_sched(struct tick_sched *ts) +{ + ktime_t idle_sleeptime = ts->idle_sleeptime; + ktime_t iowait_sleeptime = ts->iowait_sleeptime; + + memset(ts, 0, sizeof(*ts)); + ts->idle_sleeptime = idle_sleeptime; + ts->iowait_sleeptime = iowait_sleeptime; +} + void tick_cancel_sched_timer(int cpu) { struct tick_sched *ts = &per_cpu(tick_cpu_sched, cpu); @@ -1173,7 +1184,7 @@ void tick_cancel_sched_timer(int cpu) hrtimer_cancel(&ts->sched_timer); # endif - memset(ts, 0, sizeof(*ts)); + clear_tick_sched(ts); } #endif -- 1.7.7.rc0.72.g4b5ea.dirty