[PATCH] Please test: new ACPI release (20041105) import

[Warner Losh]

> I do suspect that this is a PCI power issue and not an ACPI issue, but
> I'd like to know if it shows up on more models. It is most obvious when
> playing a stream as the buffers keep emptying and the sound stops after
> a short time. Music is clearly too fast and high pitched, but at 10%
> overspeed, some may not notice. (Does Dr. Dre sound much different 10%
> faster?)

I've flipped the sense of the PCI power sysctl in current.  Once
that's sanner, we should investigate setting the power states on
resume.  Not sure what to do about suspend, since I think that acpi
has different notions of what D level to put a device in for different
S levels and it would likely have to take care of that after giving
the device a chance to save state.

Warner