(Michael L. Hitch) <hitch@lightning.oscs.montana.edu. writes:


  The following patch seens to fix it for me (with an sfb console):

Looks good to me.  (see, trying something on the hardware *did*
help!). 

The other non-interrupt-driven console getchar routines also do
spltty() or splhigh(), presumably so the getchar routine *can*.  poll
without the bottom-half driver getting in the way.

I think we should commit this and build an updated diskimage and ECOFF
kernel when that's done.
The only question in my mind is whether to use:

	* spltty()

	* splhigh() (so other devices don't interrupt the ``polled''
	     console input)

	* Something elseconformant with Charles'  ``splhigh does not
	  block hardware tty interrupts'' SPL semantics.

	* Either of the previous two, but which also   leaves
	  hardware-reset interrupts enabled, so the reset
	  button on the back works.