linux-kernel - Re: [PATCH 4/4] cpuidle/menu: add per cpu pm_qos_resume

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:   Fri, 23 Sep 2016 03:36:30 +0200
From:   "Rafael J. Wysocki" <rafael.j.wysocki@...el.com>
To:     Vincent Guittot <vincent.guittot@...aro.org>
Cc:     "Rafael J. Wysocki" <rjw@...ysocki.net>,
        Alex Shi <alex.shi@...aro.org>,
        open list <linux-kernel@...r.kernel.org>,
        "linux-pm@...r.kernel.org" <linux-pm@...r.kernel.org>,
        Ulf Hansson <ulf.hansson@...aro.org>,
        Daniel Lezcano <daniel.lezcano@...aro.org>,
        Rasmus Villemoes <linux@...musvillemoes.dk>,
        Arjan van de Ven <arjan@...ux.intel.com>,
        Rik van Riel <riel@...hat.com>
Subject: Re: [PATCH 4/4] cpuidle/menu: add per cpu pm_qos_resume_latency
 consideration

On 9/14/2016 10:28 AM, Vincent Guittot wrote:
> Hi Rafael,
>
> On 14 September 2016 at 00:10, Rafael J. Wysocki <rjw@...ysocki.net> wrote:
>> On Tuesday, September 13, 2016 10:02:49 PM Alex Shi wrote:
>>> Hi Daniel & Rafael,
>>>
>>> Any comments on this patch?
>> I actually am not sure about the whole series.
>>
>> I know your motivation, but honestly the changes here may not be the best way
>> to achieve what you need.
>>
>> You may think that the changes are trivial, but in fact they are not.  There
>> are side effects and I'm not sure about the resulting user space interface
>> at all.
> This patchset has got 2 parts:
> - one part is about taking into account per-device resume latency
> constraint when selecting the idle state of a CPU. This value can
> already be set by kernel (even if it's probably not done yet) but this
> constraint is never taken into account
> - the other part is about exposing the resume latency to userspace.
> This part might raise more discussion but I see one example that could
> take advantage of this. When you have several clusters of CPUs and you
> want to dedicate some CPUs  to latency sensitive activity and prevent
> deep sleep  state on these CPUs but you want to let the other CPUs
> using all C-state

The first very basic question about this I have is whether or not the 
device PM QoS mechanism is suitable for the task at hand at all.

It certainly hasn't been invented with it in mind.

Thanks,
Rafael