From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758806AbZJGIYw (ORCPT ); Wed, 7 Oct 2009 04:24:52 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1758793AbZJGIYv (ORCPT ); Wed, 7 Oct 2009 04:24:51 -0400 Received: from mail-px0-f179.google.com ([209.85.216.179]:45423 "EHLO mail-px0-f179.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1758791AbZJGIYu convert rfc822-to-8bit (ORCPT ); Wed, 7 Oct 2009 04:24:50 -0400 MIME-Version: 1.0 In-Reply-To: <1254903263.30157.69.camel@eenurkka-desktop> References: <1254832609.31336.290.camel@eenurkka-desktop> <1254893230.30157.28.camel@eenurkka-desktop> <1254901778.30157.62.camel@eenurkka-desktop> <1254903263.30157.69.camel@eenurkka-desktop> Date: Wed, 7 Oct 2009 01:24:13 -0700 Message-ID: Subject: Re: [BISECTED] "conservative" cpufreq governor broken From: Steven Noonan To: ext-eero.nurkkala@nokia.com Cc: "linux-kernel@vger.kernel.org" , Thomas Gleixner , Rik van Riel , Venkatesh Pallipadi Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8BIT Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Oct 7, 2009 at 1:14 AM, Eero Nurkkala wrote: > On Wed, 2009-10-07 at 10:05 +0200, ext Steven Noonan wrote: >> > Those look good. >> > >> > Well, might as well then go for: >> > /drivers/cpufreq/cpufreq_conservative.c >> > dbs_check_cpu() -> >> > load = 100 * (wall_time - idle_time) / wall_time; <- What is your load? >> >> That's probably the problem... >> >> [   40.632277] cpufreq load = 100 * (66667 - 3310) / 66667 = 95 >> [   40.698947] cpufreq load = 100 * (66661 - 3238) / 66661 = 95 >> [   73.965425] cpufreq load = 100 * (66667 - 12820) / 66667 = 80 >> [   74.032095] cpufreq load = 100 * (66661 - 1124) / 66661 = 98 >> [  107.298571] cpufreq load = 100 * (66666 - 13092) / 66666 = 80 >> [  107.365301] cpufreq load = 100 * (66722 - 3317) / 66722 = 95 >> [  140.631717] cpufreq load = 100 * (66666 - 3311) / 66666 = 95 >> [  140.698387] cpufreq load = 100 * (66662 - 3237) / 66662 = 95 >> >> idle_time is wrong. >> >> > Let assume load is sane, look for (in dbs_check_cpu()) >> >        if (load < (dbs_tuners_ins.down_threshold - 10)) { >> > >> > whether it is taken ever...if not, what is your >> > (dbs_tuners_ins.down_threshold - 10) ? >> > >> > - Eero >> > >> > > > <------------Sidenote------------------------------> > Vankatesh, I think the following is not smp friendly: >        cpufreq_conservative.c: >        /* Get Absolute Load */ >        for_each_cpu(j, policy->cpus) { >        ... >        load = 100 * (wall_time - idle_time) / wall_time; > >        now the load is only considered from the last cpu in line? >        in ondemand, load goes to max_load_freq...so max is taken >        from all cpus out there. (well, doesn't seem like the case with >        conservative). Of course I may be wrong, and things are like >        supposed. > > > Steven, how do the cpu loads look like without the patch? They're sane: [ 40.019381] cpufreq load = 100 * (66666 - 66337) / 66666 = 0 [ 40.019396] cpufreq load = 100 * (66666 - 66299) / 66666 = 0 [ 73.352580] cpufreq load = 100 * (66717 - 66349) / 66717 = 0 [ 73.352595] cpufreq load = 100 * (66634 - 63848) / 66634 = 4