linux-kernel - Re: [PATCH] sched/fair: check for idle core

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <alpine.DEB.2.22.394.2010211519390.8475@hadrien>
Date:   Wed, 21 Oct 2020 15:43:06 +0200 (CEST)
From:   Julia Lawall <julia.lawall@...ia.fr>
To:     Peter Zijlstra <peterz@...radead.org>
cc:     Julia Lawall <julia.lawall@...ia.fr>, Mel Gorman <mgorman@...e.de>,
        Ingo Molnar <mingo@...hat.com>,
        kernel-janitors@...r.kernel.org,
        Juri Lelli <juri.lelli@...hat.com>,
        Vincent Guittot <vincent.guittot@...aro.org>,
        Dietmar Eggemann <dietmar.eggemann@....com>,
        Steven Rostedt <rostedt@...dmis.org>,
        Ben Segall <bsegall@...gle.com>,
        Daniel Bristot de Oliveira <bristot@...hat.com>,
        linux-kernel@...r.kernel.org,
        Valentin Schneider <valentin.schneider@....com>,
        Gilles Muller <Gilles.Muller@...ia.fr>,
        "Rafael J. Wysocki" <rjw@...ysocki.net>, viresh.kumar@...aro.org,
        srinivas.pandruvada@...ux.intel.com
Subject: Re: [PATCH] sched/fair: check for idle core



On Wed, 21 Oct 2020, Peter Zijlstra wrote:

> On Wed, Oct 21, 2020 at 02:42:20PM +0200, Julia Lawall wrote:
> >
> >
> > On Wed, 21 Oct 2020, Peter Zijlstra wrote:
> >
> > > On Wed, Oct 21, 2020 at 01:56:55PM +0200, Julia Lawall wrote:
> > > > Prior to 5.8, my machine was using intel_pstate and had few background
> > > > tasks.  Thus the problem wasn't visible in practice.  Starting with 5.8
> > > > the kernel decided that intel_cpufreq would be more appropriate, which
> > > > introduced kworkers every 0.004 seconds on all cores.
> > >
> > > That still doesn't make any sense. Are you running the legacy on-demand
> > > thing or something?
> > >
> > > Rafael, Srinivas, Viresh, how come it defaults to that?
> >
> > The relevant commits are 33aa46f252c7, and 39a188b88332 that fixes a small
> > bug.  I have a Intel(R) Xeon(R) CPU E7-8870 v4 @ 2.10GHz that does not
> > have the HWP feature, even though the cores seemed to be able to change
> > their frequencies at the hardware level.
>
> That just makes intel_pstate not prefer active mode. With the clear
> intent that it should then go use schedutil, but somehow it looks like
> you landed on ondemand, which is absolutely atrocious.
>
> What's:
>
>   $ for i in /sys/devices/system/cpu/cpu0/cpufreq/scaling_*; do echo -n $i ": "; cat $i; done
>
> say, for you? And if you do:

/sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors :
ondemand performance schedutil
/sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq : 1197706
/sys/devices/system/cpu/cpu0/cpufreq/scaling_driver : intel_cpufreq
/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor : ondemand
/sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq : 3000000
/sys/devices/system/cpu/cpu0/cpufreq/scaling_min_freq : 1200000
/sys/devices/system/cpu/cpu0/cpufreq/scaling_setspeed : <unsupported>

>
>   $ for i in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor ; do echo schedutil > $i; done
>
> Are the kworkers gone?

Yes, they are gone.  The performance seems to be more uniform and better
with the kworkers, though (see attached traces).  This is a very short
benchmark, though.

julia
Download attachment "cg2.pdf" of type "application/pdf" (274165 bytes)