lkml.org 
[lkml]   [2014]   [Mar]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH 0/2] Add exit_prepare callback to the cpufreq_driver interface.
From
Hi,

It was a long weekend in India due to some holidays and so couldn't reply.

On Fri, Mar 14, 2014 at 11:59 PM, Dirk Brandewie
<dirk.brandewie@gmail.com> wrote:
> On 03/14/2014 10:07 AM, Viresh Kumar wrote:
> Suspend and hotplug are two very different things and if we start
> crossing those wires bad things are going to happen IMHO.
>
> In "normal" operation using the suspend path to do this work could
> work in principal but doesn't handle the case where the user does
> echo 0 | sudo tee /sys/devices/system/cpu/cpuX/online
>
> Trying force hotplug and suspend into a common mechanism would
> lead to a bunch of special case code or a significant rework of the
> core code IMHO.

What you said is correct, we shouldn't do it. But what I am asking for
is a bit different. The stuff we are doing in core on system suspend
isn't actually related to suspend but only CPU online/offline.

There are platforms which want to set CPUs to a particular frequency
before they are taken out by disable_nonboot_cpus. And then there
are platforms which want to do similar thing when CPUs are taken
down with help of sysfs files. But there is a common baseline there:
Set CPUs to a particular P-state before they are taken down.

And so I wanted to keep a common solution for both these requirements.

> This is guaranteed by the hardware. Each core has its own MSR for P state
> request. Any coordination that is required between cores to select the
> package P state is handled by the hardware.

I see.. Let me send some patches which I have in my mind and then we can
decide which set looks more reasonable :)


\
 
 \ /
  Last update: 2014-03-18 07:03    [W:0.102 / U:0.252 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site