On Wednesday 31 May 2006 1:33 am, Rafał Bilski wrote: > Please consider using pci_choose_state() for translation > from message.event to pci_power_t.
I think this is wrong, and that pci_choose_state() is so broken in its current form that it's not worth considering. You should fix that pci call to be worth worth calling first! The **ONLY** virtue of a pci standard "choose" call is to be an escape to platform code -- notably ACPI. Otherwise the current approach of simple driver heuristics (like "D3, or D0/legacy") still works just fine. But acpi_pci_choose_state() always fails, and it doesn't actually return a PCI_Dx state anyway... so in terms of its current goals, that routine is pointless since it only applies the same "D3, or D0/legacy" logic found in pretty much every Linux driver already. Then hmm, there might be platform hooks needed to enter those power states, even for "legacy pci pm" devices; maybe there should be one routine handling _both_ tasks. Plus setting up the wakeup event handling... Once the PCI routine actually does something interesting, it'd be the right time to try switching USB over to use that code. I expect some systems would then be breaking for a while. - Dave _______________________________________________ linux-usb-devel@lists.sourceforge.net To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
