linux-kernel - Re: [PATCH 1/3] cpufreq: Add a callback to update the min_freq

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: <CAJZ5v0iKOQkAUuZaHf1Zcm5sO6xD-dYkeTg8nyC3EuMmY0qDqQ@mail.gmail.com>
Date: Mon, 7 Oct 2024 17:46:18 +0200
From: "Rafael J. Wysocki" <rafael@...nel.org>
To: Dhananjay Ugwekar <Dhananjay.Ugwekar@....com>
Cc: gautham.shenoy@....com, mario.limonciello@....com, perry.yuan@....com, 
	ray.huang@....com, viresh.kumar@...aro.org, linux-pm@...r.kernel.org, 
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/3] cpufreq: Add a callback to update the min_freq_req
 from drivers

Hi,

On Mon, Oct 7, 2024 at 6:40 AM Dhananjay Ugwekar
<Dhananjay.Ugwekar@....com> wrote:
>
> Hello Rafael,
>
> On 10/4/2024 11:47 PM, Rafael J. Wysocki wrote:
> > On Thu, Oct 3, 2024 at 10:44 AM Dhananjay Ugwekar
> > <Dhananjay.Ugwekar@....com> wrote:
> >>
> >> Currently, there is no proper way to update the initial lower frequency
> >> limit from cpufreq drivers.
> >
> > Why do you want to do it?
>
> We want to set the initial lower frequency limit at a more efficient level
> (lowest_nonlinear_freq) than the lowest frequency, which helps save power in
> some idle scenarios, and also improves benchmark results in some scenarios.
> At the same time, we want to allow the user to set the lower limit back to
> the inefficient lowest frequency.

So you want the default value of scaling_min_freq to be greater than
the total floor.

I have to say that I'm not particularly fond of this approach because
it is adding a new meaning to scaling_min_freq: Setting it below the
default would not cause the driver to use inefficient frequencies
which user space may not be aware of.  Moreover, it would tell the
driver how far it could go with that.

IMV it would be bettwr to have a separate interface for this kind of tuning.