I just committed a (now months-old) change to ioasic interrupt
enabling into -Current.  (At least I *think* I did.  I'm behind
a firewall and it's hard to always be sure :-().

This change affects how the interrupt-enable mask on the motherboard
is re-set in the interrupt handler. This won't have any real
effect until the autoconfiguration code is also changed.

I've been using these changes on a 5k/240 since, oh, late January.
They work for me, but I don't have any other IOASIC machines to test
on.  It's also kind of hard to verify that this _actually_ works,
given that at least two users can't build working kernels from
-Current as of yesterday.  But if anyone using a 3min or Xine (5k/1xx
or 5k/xx) finds that their kernels die _really_ early, it's probably
my fault.

If I don't hear any complaints, I'll commit the corresponding changes
to autoconfiguration and interrupt enable over the weekend.

--Jonathan Stone