On Sat, 22 Dec 2007, Joerg Sonnenberger wrote:
>> This is clearly wrong:
>
> You can just read the registers on other chipsets as well, I don't think
> there's any hard in writing them back as well.

You misunderstand. You are reading four PCI conf registers into a single 
pcireg_t, and restoring garbage in the resume handler. This will never 
work.

Jared