From mboxrd@z Thu Jan 1 00:00:00 1970 From: Prarit Bhargava Subject: Re: [PATCH] cpufreq, intel_pstate, set max_sysfs_pct and min_sysfs_pct on governor switch Date: Wed, 07 Oct 2015 10:10:33 -0400 Message-ID: <561527D9.3020009@redhat.com> References: <1444168147-17812-1-git-send-email-prarit@redhat.com> <1755198.JNkaHg87IV@vostro.rjw.lan> <1594304.lVcRDcB3yL@vostro.rjw.lan> <002701d100cc$98cb8c60$ca62a520$@net> <5614ED1C.8070709@redhat.com> <000001d10109$10d34630$3279d290$@net> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit Return-path: Received: from mx1.redhat.com ([209.132.183.28]:51265 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751648AbbJGOKf (ORCPT ); Wed, 7 Oct 2015 10:10:35 -0400 In-Reply-To: <000001d10109$10d34630$3279d290$@net> Sender: linux-pm-owner@vger.kernel.org List-Id: linux-pm@vger.kernel.org To: Doug Smythies Cc: 'Kristen Carlson Accardi' , linux-kernel@vger.kernel.org, 'Viresh Kumar' , linux-pm@vger.kernel.org, "'Rafael J. Wysocki'" On 10/07/2015 10:04 AM, Doug Smythies wrote: > > On 2015.10.07 03:00 Prarit Bhargava wrote: >> On 10/07/2015 02:51 AM, Doug Smythies wrote: >>> >>> And before patch I get, using primitives and not cpupower: >>> Executive Summary: Everything works fine (or at least as I thought it was supposed to). >>> >>> root@s15:/home/doug/temp# grep . /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor >>> /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor:powersave >>> ... >>> /sys/devices/system/cpu/cpu7/cpufreq/scaling_governor:powersave >>> root@s15:/home/doug/temp# grep . /sys/devices/system/cpu/intel_pstate/*_perf_* >>> /sys/devices/system/cpu/intel_pstate/max_perf_pct:100 >>> /sys/devices/system/cpu/intel_pstate/min_perf_pct:42 >>> root@s15:/home/doug/temp# echo 50 > /sys/devices/system/cpu/intel_pstate/min_perf_pct >>> root@s15:/home/doug/temp# echo 80 > /sys/devices/system/cpu/intel_pstate/max_perf_pct >>> root@s15:/home/doug/temp# grep . /sys/devices/system/cpu/intel_pstate/*_perf_* >>> /sys/devices/system/cpu/intel_pstate/max_perf_pct:80 >>> /sys/devices/system/cpu/intel_pstate/min_perf_pct:50 >>> root@s15:/home/doug/temp# for file in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor; do echo "performance" > $file; done >>> root@s15:/home/doug/temp# grep . /sys/devices/system/cpu/intel_pstate/*_perf_* >>> /sys/devices/system/cpu/intel_pstate/max_perf_pct:100 >>> /sys/devices/system/cpu/intel_pstate/min_perf_pct:100 >>> root@s15:/home/doug/temp# grep . /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor >>> /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor:performance >> >> The switch has to be from performance to powersave. Switch again and you'll see >> the problem. > > Yes, of course, and I did. > The first "powersave" listing was just to show the setup. > Continue reading my original reply, also added back below: > The "..." stuff is just where I deleted 6 of the 8 CPUs saying the same thing. > > root@s15:/home/doug/temp# grep . /sys/devices/system/cpu/intel_pstate/*_perf_* > /sys/devices/system/cpu/intel_pstate/max_perf_pct:100 > /sys/devices/system/cpu/intel_pstate/min_perf_pct:100 > root@s15:/home/doug/temp# grep . /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor:performance > ... > /sys/devices/system/cpu/cpu7/cpufreq/scaling_governor:performance > root@s15:/home/doug/temp# for file in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor; do echo "powersave" > $file; done > root@s15:/home/doug/temp# grep . /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor:powersave > ... > /sys/devices/system/cpu/cpu7/cpufreq/scaling_governor:powersave > root@s15:/home/doug/temp# grep . /sys/devices/system/cpu/intel_pstate/*_perf_* > /sys/devices/system/cpu/intel_pstate/max_perf_pct:80 > /sys/devices/system/cpu/intel_pstate/min_perf_pct:50 > >> I can also reproduce this without using 'cpupower'. > > Interesting. Then maybe it is a difference between legacy mode > and Hardware P state (HWP) mode. I do not have an HWP capable processor > to test. I have intel_pstate=no_hwp so that doesn't explain things. I will send you a printk debug patch shortly to figure out why your system doesn't see this problem. I have _6_ different variants of intel processors that see this same issue. What is your model #? P. > > I compiled cpupower, but get the same results as I get using primitives, > meaning things work properly before the patch and resort to defaults > after the patch. > >