On Mon 18 Aug 1997, Matthias Drochner wrote:
> Excerpts from netbsd: 18-Aug-97 Re: bus.h style question "Gordon W.
> Ross"@mc.com (880)
> 
> > Not only ISA bridges on big-endian processors present this problem.
> > Look at the Sun "ie" board (Intel Ethernet on VME).  It is a 16-bit
> > VME board with its VME interface hard-wired to do byte swaps.
> 
> I have to deal with a similar problem here: The Dolphin PCI-SCI
> adapter. The chip was designed for SBUS first, and it needs all data
> byte-swapped.
> This is exotic however, and not enough reason to bloat the bus.h
> interface for imho.
> I thought about options to the bus space mapping functions at some
> time. It's hard to find a sufficient abstraction for all the swapping
> possibilities.

This is exactly why I decided to accept the fact to tack an endian label
to a bus; it should work for most of the hardware and it provides a
reasonably sane interface for most (MI) driver writers. If you drop this..
uhh "certainty", nearly every thought about an MI driver turns into a
headache. 

Leo.