On Thursday 28 October 2004 12:32, Pavel Machek wrote: > Hi! > > > > I guess it would be handy. It is incompatible change but right thing > > > to do. > > > > Not clear to me why it should be needed though. We know > > that in ACPI terms the system is going back to the G0 state; > > what else is there to say? > > Unfreeze during suspend: enable queues, almost nothing to do. How can we know it's safe to re-enable all queues? If they're empty, and nothing fills them ... it's obviously OK. Likewise if they're never re-enabled. But it's not clear to me that re-enabling all I/O queues should be safe. If there's a pending write, it could be completed both at that point ... then again later on resume. Is that always going to be safe, for all devices? Probably not. > Unfreeze during resume: reinitialize the hardware, then enable queues. Don't always re-init; S4 may have had it preserve key state. Like how the device requested the wakeup from S4, or the state of downstream links. - Dave
