>From: wrstuden@loki.stanford.edu (Bill Studenmund)
>
>> I agree that a portable way to do the IO without the extra copy would
>> be a win in general.  However a bus-spaceified version of uiomove()
>> doesn't look trivial at first glance, because of copyin() and
>> copyout() which are MD.
>
>But aren't all the bus-space routines M.D. anyway? It'd basically
>be munging a bus space read with the kernel side of copyin() and
>copyout().

The bus-space routines are MD, but the interface they present is
theoretically MI.  My point was that the bcopy()s in uiomove could be
replaced with bus_space_move_multi_1()s in a bus-spaced version, but
there's no similar generic way to replace the calls to copyin() and
copyout().

If you feel you can put together something functional, I'm certainly
not going to stop you.  I know my limits and this is well beyond the
boundary.

Cheers,
entropy

--
entropy -- it's not just a good idea, it's the second law.