Re: [RFC] New system device API

[Pavel Machek <pavel@suse.cz>]

Hi!

> > Well, can you be a little more concrete? I do not see any description
> > about what is system device and what is not.
> > 
> > Keyboard controller is very deeply integrated into the system. If it
> > is not system device, what is it?
> 
> I apologize that the description of system devices is not in the driver 
> model documentation. From the linux.conf.au paper:
> 
> System-level devices are devices that are integral to the routine
> operation of the system. This includes devices such as processors,
> interrupt controllers, and system timers. System devices do not follow
> normal read/write semantics. Because of this, they are not typically
> regarded as I/O devices, and are not represented in any standard
> way. 

What about mtrr's? They seem like system-level devices to me. Still
its usefull to have kmalloc in its suspend routine, which moves it to
SAVE_STATE phase.

Decision on which level to put it is up to programmer, and it seems
wrong to hardcode it into architecture. It may be more convient to do
save stating at place where you still can kmalloc...
								Pavel
-- 
When do you have a heart between your knees?
[Johanka's followup: and *two* hearts?]