> so that we can just drop in new guts for the mutex primitives when
> the kernel becomes multi-threaded.

Hmm.  I don't think things are quite that simple..  We'll probably end
up with several different flavors of lock (e.g., spinlock
vs. sleeplock; mutexes vs reader/writer locks if we *really* care
about MP scalability).

Unless you turn "interrupts" into "threads", you also can't just
blindly attempt to take spinlocks at interrupt level, as the
interrupted process might already hold the spinlock and will be in no
position to release it until after the interrupt completes..

						- Bill