hi folks,

I've checked in the MI changes to support read() and write() returning
the real error values from I/O errors.  I've also done the MD parts for
x86 and sparc (see my recent checkins to locore.s and trap.c on those
platforms).  as discussed before, I'm going to leave the rest of the
MD parts for other folks to do.  it's pretty straightforward on most
platforms, just store the return value from uvm_fault() into some register
in the trapframe, then have the onfault handler return that.

(Ben Harris is way ahead of the game, he's already updated arm26!)

-Chuck