Re: [RFC PATCH v2 1/3] PM / Core: suspend_again callback for device PM.

["Rafael J. Wysocki" <rjw@sisk.pl>]

On Tuesday, April 26, 2011, Pavel Machek wrote:
> Hi!
> 
> > > > What is the root problem that this is trying to solve?
> > > 
> > > It is trying to fix machines that need to run periodic kernel tasks
> > > even when user asked them to sleep. Zaurus needs to periodicaly wake
> > > up to be able to charge battery in s2ram state, for example.
> > 
> > Well, if you wake up periodically, it's not S2RAM any more, I'd say.
> 
> It certainly looks a lot like s2ram to the user.
> 
> Now, your PC wakes up periodically in s2ram, too; but it is keyboard
> controller that wakes up, not CPU, so you do not know.

I seriously doubt it's a keyboard controller in contemporary machines.

Yes, there are parts of them that don't actually sleep, but that's not
under OS control.

> On Zaurus, you do not know, either; nothing visible happens and
> userspace is shielded from these details.

If you wake up device drivers, it _is_ visible.  Maybe not to user
space, but that's a semantic difference.

> Should user really have to know if battery charger is implemented on
> keyboard controller or on main cpu before he selects which interface
> to use?

Please don't confuse user space with the user.  You can resume all the
way to user space without turning the backlight on and the user will not
notice, right?

Rafael