linux-kernel - Re: [PATCH] thermal: power allocator: Add control for non-power actor devices

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] [day] [month] [year] [list]

Message-ID: <2550a12b-aed6-dd89-0260-69241d04d39b@arm.com>
Date:   Mon, 18 Jan 2021 17:15:38 +0000
From:   Lukasz Luba <lukasz.luba@....com>
To:     Daniel Lezcano <daniel.lezcano@...aro.org>
Cc:     linux-kernel@...r.kernel.org, linux-pm@...r.kernel.org,
        amitk@...nel.org, rui.zhang@...el.com
Subject: Re: [PATCH] thermal: power allocator: Add control for non-power actor
 devices



On 1/18/21 4:07 PM, Daniel Lezcano wrote:
> On 05/01/2021 20:01, Lukasz Luba wrote:
>> The cooling devices which are used in IPA should provide power mapping
>> functions. The callback functions are used for power estimation and state
>> setting. When these functions are missing IPA ignores such cooling devices
>> and does not limit their performance. It could happen that the platform
>> configuration is missing these functions in example when the Energy Model
>> was not setup properly (missing DT entry 'dynamic-power-coefficient').
>>
>> The patch adds basic control over these devices' performance. It
>> manages to throttle them to stay safe and not overheat. It also adds a
>> warning during the binding phase, so it can be captured during testing.
>>
>> The patch covers also a corner case when all of the cooling devices are
>> non-power actors.
> 
> In my opinion this is a user space problem. If a device does not have
> power information, then it should use the step-wise governor instead.
> 
> It is not the power allocator to overcome a wrong or unsupported setup.
> 
> Usually, the default governor is the step-wise and the userspace sets
> the power allocator policy.

True, but currently there is even no warning to the user, that some
device (quite big, like GPU) is not controlled. This could be dangerous.

> 
> A solution can be to fail to change the policy or bind if the associated
> cooling devices are not all power actors.

I tend to agree, we can simply 'fail' during the binding and print
some warning.

The code in that case would be much simpler. Let me send a v2 then.
Thank you for your comments.

Regards,
Lukasz