Re: [RFC] New system device API

[Patrick Mochel]

> > Well, you need to suspend devices used to write the image, too, so you
> > have state to return to after resume. You only do not want disks to
> > spin down. Perhaps disk can just special-case it ("If I am going to
> > swsusp, I need to save state, but do not really need to spin down").
> 
> Mmm. Sounds ugly though. Would it be fair to say we want to S5 some
> devices and S3 others? Perhaps that sort of terminology might be
> helpful.

Those are not even valid states for devices. Device states are commonly 
D0-D3, though they are not represented the same way for all kinds of 
devices. 

Pavel is right, and we should be able to do that generically, though in a 
slightly different manner. We should have all devices save state, write 
the image, then power them down (including spinning down disks). 

	-pat