On Sun, Apr 16, 2000 at 01:33:57PM -0700, Eduardo Horvath wrote:
> If we're talking about registers then this should already be handled by
> bus_space_{read,write}*() routines and no htole()/htobe() should be
> needed.

Sure.

> 
> If we're talking about DMA memory then we should add an equivalent set of

Yes, that's it.

> bus_dma_{read,write}*() routines rather than use htole()/htobe().  

It may be nice to have a standartized way of doing it, for sure.

> 
> One *VERY IMPORTANT* thing to remember is that UltraSPARC machines can
> support both SBus (big-endian) and PCI (little-endian) on the same
> machine, and it is even possible to have the same device on both SBus and
> PCI so you cannot hardcode bitswapping for a particular device based on
> the architecture.

For 53c8xx devices it's not true; these are PCI-only devices,
but if the driver also supports the 53c720 then this theorically can
happen :)
The issue is more real for devices like the isp, I think (or maybe just
hme).

--
Manuel Bouyer <bouyer@antioche.eu.org>
--