run resume code only for S1-S4 states

[Jung-uk Kim]

On Monday 27 April 2009 03:20 pm, Fabian Keil wrote:
> Andriy Gapon <avg at freebsd.org> wrote:
> > on 25/04/2009 11:21 Fabian Keil said the following:
> > > Sure. It turns out that the problem is unrelated to your patch.
> > > I can reproduce it with an unpatched kernel too, by once
> > > pressing the power button before the second core is started.
> > >
> > > I probably did the same a few days ago, and forgot about it.
> > > Sorry.
> >
> > thank you very much for the testing and the insight, this is very
> > useful and interesting.
> > I think that it might be that 'init' process in pre-natal state
> > loses a signal sent to it.
> >
> > I decided to follow Nate's advice and exempt S5 from timeout
> > policy (after all it is possible to execute shutdown(8) multiple
> > times and concurrently with any other sleep request). With
> > previous version of the patch once shutdown_nice() failed once it
> > was impossible to enter any sleep state ever. shutdown_nice
> > failure is quite exotic event, but as you have proven it is not
> > impossible.
> >
> > So the new patch is attached.
>
> Thanks. The patch works and pressing the power button early on
> boot before it actually has any effect no longer prevents the
> power button from working later on.

Can you try r191699 acpi.c and tell me if there is any regression?

Thanks!

Jung-uk Kim