On Thursday 28 October 2004 08:43, Patrick Mochel wrote: > > > I'm not sure if PCI_D3cold should be included like > > this. pci_enable_wake at least ignores it... Actually that's not true. It's pci_set_power_state() that hides the distinction. If you pass 4 (D3cold) to pci_enable_wake, it sets that bit in config space so that devices that can wake from that state will be able to do that. > It's not a real state that devices can enter; IIRC it just describes the > state devices are in when the system is initially turned on. I don't know > why some drivers use it, but it doesn't seem right. (It's been a while > since I read the spec, though, so I could be wrong.) It's been a while since I read that part of the PCI PM spec too, but I remember that D3hot and D3cold are both distinct PCI PM states. Drivers can enter D3hot, but after resume it could have morphed into D3cold. - Dave
