> Date: Tue, 22 Feb 1994 10:25:49 -0800
> From: Adam Glass <glass@sun-lamp.cs.berkeley.edu>
> 
> Yeah...i tried a scheme like this using timeout(), and the prom
> maygetchar() function.  it just didn't work, though i should probably
> take another look at it.

The maygetchar() function works fine with the NMI left active.
I think the PROM console would be easiest to get working.
Also, I have found it really handy to be able to stop the kernel
in its tracks and poke around with the PROM. (i.e. see below)

> I actually have a system alive enough to boot, exec /sbin/init which
> opens /dev/console which (effectively) opens /dev/prom, requesting
> single-user mode, execs /bin/sh, prints a shell prompt, and blocks on
> read() afterwards.  I'll commit all of the changes necessary to get
> that far as soon as I fix either the prom console driver or the zs
> driver I ported such that /dev/console input/output actually works.

The biggest problem I am encountering right now is that the kernel
takes some vm_fault traps and then somehow jumps immediately to
the PROM reset which clears the screen so I can't tell what's up.
I was able to see some of what happened by hitting L1-A at just the
right instant (before it does the reset).  I can get a stack trace
at this point (using the PROM "g4" function) but I have not yet done
the tedious conversion from hex to symbol+offset or source line.

I would find it very helpful to have any of the kernel bug fixes
you've done so far, even if the drivers are not yet in place.
(If you want to wait until it "works" I can understand.)

Thanks,
Gordon

------------------------------------------------------------------------------