On Thu, 22 Nov 2007, Matthias Drochner wrote:
> No, they are made compatible of course, but the logical structures
> within current chipsets are different. There is no address aliasing
> anymore. And the devices don't go away if you disable the PCI-ISA
> bridge. (Actually, the bridge's io/mem space enable bits are read only
> on my box.) So having these devices attached to acpi (or whatever
> virtual mainbus) reflects reality better than having them on isa.

We are still speaking to the compatible interface, and will continue to do 
so. The compatible interface is an ISA interface -- why not treat it like 
one?

>> Like it or not, isa(4) is not going away any time soon on x86.
>
> There are just some standard PC parts which depend on isa
> (timer, intr controller etc) which could be as well hardwired
> into common x86 code because they are part of what we know
> as PC architecture. ISA also means indirect configuration, and
> incomplete decoding, and that is dead code within most kernels
> already now.

I meant that our ISA support code (bus_* etc) is not going away, and that 
the ACPI glue for these devices uses it anyway.

Cheers,
Jared