On Fri, 5 Oct 2007, Joerg Sonnenberger wrote:
> Jared wants to do similiar things for the audio devices. In general, the
> decision about when to power down or not should be left for higher
> layers to decide. A device is certainly not the right place.

My plan for audio is to have MI audio inform the PNP policy of device 
activity, and the pnp dev class policy for audio devices handles power 
states. No device driver support required (apart from standard pnp_power 
API), and minimal support required in MI audio. dev class policy for audio 
devices gradually steps down the power level (D0 -> D1 -> D2 -> D3, as 
supported) on idle timeouts, and restored back to D0 on the next 
pnp_active() invokation. IOW, the longer the device is idle, the deeper 
it will sleep; as a result, the longer a device is idle, the longer it 
will take to wake back up.

Cheers,
Jared