chopps@merit.edu said:
> So while I was fixing my i386's pnpbios I noticed that we have a lot
> of options and information available with pnpbios that we aren't
> currnetly using. 

Yes. Using the ressource allocation data even if no pnpbios
driver is attached would help a lot. But there is quite some
work necessary to the bus_space stuff to allow other drivers
to take over a ressource originally allocated to pnpbios.
These are the limitations of the one-pass autoconfiguration.

> As 1 example we don't know all the different ways to disable and enable
> Cardbus controller's pcic compat (i.e., legacy isa mode) and so
> sometimes autconf just loses (finding both).  But through pnpbios you
> can just say `disable this' and the bios DTRT.

Well, my testing base is a bit limited (consists of one Thinkpad
only), but I've not yet seen a PnP node describing a pcic.
Cardbus controllers shouldn't be handled by the PnP BIOS anyway
because they are logically PCI devices.
(What I've seen is that a PnP node exists for a soundblaster/wss
compatible audio device which is physically implemented within
a Neomagic PCI audio chip afaik. The legacy ressources shouldn't be
described by the PCI device then.)

> So what I'm wondering is what is the proper way to take advantage of the
> information and control ability presented to the kernel though the
> bios/firmware in a nice and MI way?

Don't know... These mixes between clean buses as PCI and legacy
devices are nasty, and I'm not very optimistic that it can be dealt
with an an mi way.

best regards
Matthias