Re: Solving suspend-level confusion

On Friday 06 August 2004 13:04, Pavel Machek wrote:

quoted

These look to me like "wrong device-level suspend state" cases.

Actually, suspend-to-disk has to suspend all devices *twices*. Once it
wants them in "D0 but DMA/interrupts stopped", and once in "D3cold but
I do not really care power is going to be cut anyway". I do not think
this can be expressed with PCI states.

How are those different from "PCI_D1" then later "PCI_D3hot"?
I'd understood that loss of VAUX was always possible, so robust
drivers always had to handle resume  from PCI_D3cold.