In message <20040419.184219.83793818.he@uninett.no>, you write: 

-> Hi,
-> 
-> a small update -- another way to trigger the problem:
-> 
-> I took the system up single-user, created a few users and
-> configured rc.conf to start sshd, so that I could log in
-> remotely.  That worked, but as soon as I typed "su<RET>", the
-> machine paniced:

I don't have time to look at this in any level of detail before the
weekend, but this sounds very much like the kernel stack overflows we
were seeing before 1.6(?) due to interrupt nesting getting too deep.
Since the interrupt stuff has had a lot of whacking done to it (and
since I have been too busy with the Real World to follow most of the
changes, never mind actively hack on port-sgimips), I could be out in
left field.

However, I did recently (week or two ago) install a new kernel on the
Challenge and it seemed fine, but that may have been before the change
that caused the breakage.  In any case, I'll try and at least see if I
can reproduce it there or on my Indigo2 and maybe try and put in a bit
of time debugging if I can make it happen.

BTW, what was the problem with haltwo vs. Challenge S?  I seem to recall
booting a GENERIC_xxx (whatever it's called today :-) kernel on it without
problems when I did the last upgrade (as I said, in the last couple of
weeks).

--rafal

----
Rafal Boni                                                     rafal@pobox.com
  We are all worms.  But I do believe I am a glowworm.  -- Winston Churchill