Re: Power Management Framework architectural design

[Martin Husemann <martin%duskware.de@localhost>]

On Tue, Apr 15, 2008 at 07:29:23AM -0700, Jason Thorpe wrote:
> The problem is that power domains don't necessarily fit into that  
> strict hierarchical model; one "device" (the main board or whatever)  
> might have multiple power domains with control logic of those domains  
> not in the ancestry of the devices whose power is being managed.

But if the domain decides to power down, the device still needs to
power down it's children first (no matter in what domains those actually
are). This is not directly an issue in the case at hand due to the very
flat device tree.

So we have a (potential empty) set of domains each device is part of, and
events (like "I am idle") that the device communicates to pmf, which could
do the bookeeping and if all members of the domain meet some (domain specific)
condition do the legwork for a domain status change and save device state,
before the domain driver pulls the trigger.

Martin