linux-kernel - Re: Commit 554c8aa8ecad causing severe performance degression with pcc-cpufreq

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]

Date:   Tue, 17 Jul 2018 17:29:21 +0200
From:   "Rafael J. Wysocki" <rafael@...nel.org>
To:     Andreas Herrmann <aherrmann@...e.com>
Cc:     "Rafael J. Wysocki" <rjw@...ysocki.net>,
        "Rafael J. Wysocki" <rafael.j.wysocki@...el.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Frederic Weisbecker <frederic@...nel.org>,
        Viresh Kumar <viresh.kumar@...aro.org>,
        Linux PM <linux-pm@...r.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: Commit 554c8aa8ecad causing severe performance degression with pcc-cpufreq

On Tue, Jul 17, 2018 at 4:03 PM, Andreas Herrmann <aherrmann@...e.com> wrote:
> On Tue, Jul 17, 2018 at 12:21:36PM +0200, Andreas Herrmann wrote:
>> On Tue, Jul 17, 2018 at 12:09:21PM +0200, Rafael J. Wysocki wrote:
>
>   ---8<---
>
>> > OK, the patch is below.
>> >
>> > First, I hope that if "Collaborative Power Control" is disabled, it will
>> > simply hide the PCCH object and so intel_pstate will still not load then.
>>
>> PCCH is hidden in that case.
>>
>> > The main question basically is what the OS is expected to do if
>> > "Dynamic Power Savings Mode" is set.  If we are *expected* to use
>> > the PCC interface then, intel_pstate may not work in that case, but
>> > I suspect that the PCC interface allows extra energy to be saved
>> > over what is possible without it.
>>
>> I'll test it and see what happens.
>
> I've tested it on top of v4.18-rc5-36-g30b06abfb92b. intel_pstate now
> loads instead of pcc-cpufreq and system looks stable.
>
> When disabling "Collaborative Power Control" no cpufreq driver is loaded
> (as expected).
>
> Performance (with kernbench) is as expected (always better than any
> brew of pcc-cpufreq + misc modifications to this driver + partial
> rollback of commit 554c8aa8ecad).
>
> If you like you can add either Tested-by or
> Reviewed-by: Andreas Herrmann <aherrmann@...e.com>
>
> I think this patch should be tagged for 4.17-stable.

OK, thank you for testing!