On 14 Jul, Rafal Boni punched keys in this particular order:
> Next time you see this happen, please also post a traceback (via the `tr'
> command).  What's happening is that you're jumping through a NULL pointer
> somewhere, but I'm not sure where.  

First boot I tried ended up in a TLB miss in kvm_mkdb again. Here's the
trace (very short...?):

db> tr
PC 0x0: not in kernel space
0+0 (1,9fdfd,3f12d9dd,883200b0) ra 0 sz 0
User-level pid 82
db>

Sometimes it gets beyond kvm_mkdb and has a TLB miss in sh, and
sometimes it runs for a while; at least a few hours if I don't _do_
anything with the system - haven't tested longer than that, yet (I
usually _do_ something after a while).

> (more specific than the 1.6U), but if it happens in 1.6.1, that's much more
> of a concern :-/

Unfortunately, this is the case.
I just double-checked by downloading a fresh 1.6.1 kernel, as I wasn't
entirely sure that I haven't been experimenting with different kernels.
It does have this problem.


PS. How do I ascertain that this is not a hardware problem? I couldn't
find any memory testing software in the packages, and other tests are
welcome too. The built-in SGI test program unfortunately requires
software on the system disk (not necessarily a harddisk in the system
itself, AFAICS).

PS 2. I will probably sell this system in a few weeks, when it's
replacement arrives (an R10000 Indigo2, _with_ harddisk and CD-ROM), so
after that I won't be able to test NetBSD sgimips any further (until the
R10000 is supported, of course).
I am willing to keep it around for a while, though, for testing
purposes.

-- 

Alban Hertroys                              http://solfertje.student.utwente.nl
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
              The difference between your legs and your bicycle is
              that when they are in bad shape it is unwise to give
              your bicycle more training.