probably shouldn't bind to PCI devices based only on ssvid/sspid
There was a problem on certain dell machines where their ssvid/sspid is an old amr(7D) device ID. We then attempt to bind based on that, and try to bind amr to basically every device on the mainboard (the cpus, etc...). We hang. It took danmcd a long painful time to debug this. It sucked.
It's almost certainly insane to be generating the properties such that we're willing to bind based only on ssvid/sspid. Rather than a more-specific ID including ssvid and sspid.
We should (very carefully) stop doing it. Unfortunately, x86 is a mess, and fixing this incautiously could have who-knows-what effect. Anyone attempting it should be very careful.