> There is a "status" byte read at the beginning of the interrupt
> handler. Bit 5 of it tells whether the pending data are
> from the mouse port. This all looks OK in the manual and in
> the code.

Hmm...  It looks to me like it's more of a 'there is data in the mouse
buffer'.  When I boot with the keyboard switched away from the computer,
I get alternating:
	0x34
	0x35
When I boot with the keyboard switched to the computer, I get
alternating:
	0x14
	0x15

So the rest of the status is the same:
	self-test OK, Keylock status free, and a toggle on the keyboard
		output buffer

In the non-working case, there is a solid
"mouse output buffer holds mouse data" condition.

> Yes, these are different namespaces.

That's what I figured--just wanted to be sure.

OK.  I added a quick hack:
	if (stat & 0x01) slot = PCKBC_KBD_SLOT;

after
	slot = (t->t_haveaux && (stat & 0x20)) ?
                    PCKBC_AUX_SLOT : PCKBC_KBD_SLOT;

and it's working.  Not only is it working, but the "bad" mouse
buffer bit (garbage from a noisy switch?) is gone after the first
pass through the interrupt routine and the stat register toggles
between 0x14 and 0x15 again.

-allen