Matthias Scheler <tron@zhadum.de> writes:

> On Sat, Apr 09, 2005 at 02:22:02PM -0400, Nathan J. Williams wrote:
> > > The above configuration line should be removed otherwise.
> > Why?
> 
> Probably because it worked fine on my PC except on i386 and 486 systems
> I had ages ago.
> 
> > I'd almost say the opposite. It's not like the pcib driver or the
> > isa-at-pcib attachment do anything useful.
> 
> I don't know about that. But that attachment is the one which every
> new PC build the last 10 years uses.

Only because it's there, and the isa at mainbus code only runs if ISA
hasn't already been found. All those machines would work equally well
with isa0 at mainbus?. Look at sys/arch/amd64/pci/pcib.c
vs. sys/arch/amd64/amd64/mainbus.c (about line 245).

In the days when machines had real ISA buses hanging off of a
southbridge chip, one could argue that it was nice to have the
attachment hierarchy reflect the physical hierarchy. But now that all
the remaining "isa" devices are integrated in one super-io chip or
another, it's a lot less clear that that's true.

What I think would be genuine forward progress would be to use ACPI to
attach the remaining legacy devices and punt the ISA infrastructure
entirely.

(looking at the amd64 GENERIC config: does "wdc at isa" *ever* happen
on an amd64 machine?)

        - Nathan