[Pm-utils] Set performance governor instead of userspace before
suspend
Holger Macht
hmacht at suse.de
Wed Oct 18 15:24:41 PDT 2006
On Tue 10. Oct - 20:04:59, Peter Jones wrote:
> On Fri, 2006-10-06 at 18:09 +0200, Holger Macht wrote:
> > Hi,
> >
> > I'm just looking over the 94cpufreq hook which sets the userspace governor
> > immediately before suspend. I think this was needed because of dead locks
> > of the ondemand governor in the past, right?
>
> Yes -- in fact, I don't think it's a problem any more, so we can
> probably just remove it soon, or just disable it in distro's packaging,
> but leave it in CVS. Maybe we should add some auto* stuff to disable
> individual scripts' installation by default.
Something else comes to my mind when thinking about the
distro-hook-directories I have proposed in another mail. We could add a
--with-hooks=(suse|fedora|gentoo) so that distros just can configure their
package this way and run make install and everythings they want gets
installed.
>
> > Wouldn't it be better to set the performance governor then? Because if
> > frequency is low and you set the userspace governor, frequency will keep
> > to be low if you have no userspace daemon caring about.
> >
> > Anyway, setting the performance governor would be a good idea in any case
> > because of compression and stuff we will have in the near future.
>
> Doesn't matter -- the machine isn't going to be _running_. And when it
> is running again, we're going to go back to the same governor that we
> had before. So basically, we're talking about ~2 seconds of runtime
> that this will effect.
>
> Since that's the case, I'd rather just leave it as "userspace", if only
> for the reason that it doesn't _do_ anything. Less stuff being done
> means less chances of having to hack on this again when a bug is
> introduced.
Tim mentioned in another mail that the userspace module doesn't need to be
available at all. So what speaks agains using 'performance', which is
available in all cases?
Regards,
Holger
More information about the Pm-utils
mailing list