linux-kernel - Re: [PATCH] arm64: defconfig: Select schedutil as default cpufreq governor

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: <CAK8P3a1FinyyFr840-JkTcX2mL8QTtcnvYi3hpczs3Yx_ga-ww@mail.gmail.com>
Date:   Thu, 21 Dec 2017 16:32:32 +0100
From:   Arnd Bergmann <arnd@...db.de>
To:     Viresh Kumar <viresh.kumar@...aro.org>
Cc:     Catalin Marinas <catalin.marinas@....com>,
        arm-soc <arm@...nel.org>, Will Deacon <will.deacon@....com>,
        Vincent Guittot <vincent.guittot@...aro.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Linux ARM <linux-arm-kernel@...ts.infradead.org>
Subject: Re: [PATCH] arm64: defconfig: Select schedutil as default cpufreq governor

On Mon, Dec 18, 2017 at 5:29 AM, Viresh Kumar <viresh.kumar@...aro.org> wrote:
> On 15-12-17, 15:50, Catalin Marinas wrote:
>> On Thu, Nov 16, 2017 at 11:51:36AM +0530, Viresh Kumar wrote:
>> > Currently performance governor is getting selected by default, which is
>> > surely not a very good choice as its pretty much power hungry.
>> >
>> > Select schedutil instead.
>>
>> And why do we care about this in defconfig? People deploying their own
>> kernels in mobile may opt for this config, others may prefer the default
>> governor.
>>
>> Also it seems it would be the only architecture make this governor the
>> default, so NAK.
>
> This is a bit dangerous configuration IMHO.
>
> Other architectures have some *real* governor selected by default, like Ondemand
> or Conservative. Running your CPUs at max (because of the default performance
> governor in arm64 config) may end up burning some SoCs accidentally just because
> their thermal stuff doesn't kick in to cool SoC down properly.
>
> So, we should have one of ondemand, conservative and schedutil selected by
> default for arm64 as well IMO and schedutil is the one which every one is
> falling back to now a days, even android.

Maybe it's time to change the global 'default
CPU_FREQ_DEFAULT_GOV_PERFORMANCE' instead to avoid having to
do this for each architecture separately?

I think the general idea of using schedutil or ondemand instead of performance
makes sense, but it doesn't feel right to have to do this for every
single defconfig
that doesn't select a default.

      Arnd