Aymeric Vincent <xmimic@free.fr> writes:

> As stated in the PR, preventing the second pcib from attaching the ISA
> bus made the trick and he could boot properly.
> 
> However, I believe it would be nice to have a isa_match_hook() called
> by isamatch() which could prevent attachment as soon as one isa bus
> has been seen on the i386 arch.
> 
> I don't feel confident enough to do the change myself, and such a
> change would mean adding a isa_match_hook() to every architecture that
> has ISA busses.
> 
> Obviously such a restriction (enforcing only one isa bus) can not go
> into the MI isa code since I don't see any conceptual reason to
> prevent multiple isa busses from coexisting in address space or
> something.

This has been discussed almost to death. While there's no conceptual
reason you couldn't have multiple ISA busses, there's no such hardware
that anyone knows about, and we certainly don't have support code for
any such thing.

I think the best solution is simply to change all instances of "isa*
at ???" to "isa0 at ???" in the config files, which is a nice minimal
change that prevents multiple busses attaching.

That said, there *are* new devices on things like docking stations,
and we need some mechanisim (ACPI?) to find them, or re-probe the
(one!) ISA bus.

        - Nathan