> 
> Michael-John Turner <mj@turner.org.za> wrote:
> > > At this point, my best guess is that there is a serious problem in the MP
> > > NetBSD 3.1 code on sparc, at least for the 50 MHz SuperSPARC CPUs i
> use.
> > 
> > Hmm... I have a pair of machines, each with a pair of SM61s and they've
> > both been rock solid with 3.0 and 3.1. Both have uptimes of ~26 days and
> > are fairly busy serving DNS, DHCP, ntp, Kerberos, etc. They're both running
> > a custom kernel, but whilst upgrading them and before building a custom
> > kernel they ran GENERIC.MP.
> > 
> > I have a pair of SM50s in my parts box and can bring up another SS20 to see
> > if I can reproduce the problem. Before I got the SM61s, I did have one of
> > the machines mentioned above running 3.0 stably with an MP kernel and a
> > pair of SM50s.

I replied::
>  I'd appreicate the help.  It could be a problem in my build environment,
> as well -- the 3.1 installation i'm running was built locally (cross
> compiled on a pentium4 running NetBSD 3.1).

 OK, found the time to do some hardware renoberation, and some testing.  I
got my production machine (vishnu) working multiprocessor with a different
CPU arrangement -- a dual-50MHz board (2 X 390Z55) instead of two single
50 MHz boards (2 X 390Z50). This required swapping the motherboard, so in
essence, vishnu is a "new" machine, with the "old" hard disks, HME network
card, and some of the RAM.

 So the current setup i have:

Two SPARC-20s, they appear to have the same ROM revision, but the 390Z55
will only work in one of them (vishnu).  When i install the dual CPU
390Z55 in the other SPARC-20, it only sees one of the CPUs.

vishnu:
SPARCstation 20 MP (2 X 390Z55), No Keyboard
ROM Rev. 2.22, 512 MB memory installed, Serial #7779815.

kali:
SPARCstation 20 MP (2 X 390Z50), No Keyboard
ROM Rev. 2.22, 384 MB memory installed, Serial #7448106.


Unpredictable and bad behavior occurs on NetBSD 3.1 on the dual 390Z50
machine (kali), but the 2 X 390Z55s (vishnu) seem to be OK.

 The 2 X 390Z50s worked OK with NetBSD 2.0.2_STABLE, so it seems like
something that changed between the revisions.

The current error during boot is:

cpu0: booting secondary processors: cpu1
warning: no /dev/console
init: Creating mfs /dev (410 blocks, 1024 inodes)
uid 0, pid 8, command sh, on /dev: file system full
init: Unable to run MAKEDEV: Invalid argument


 I may rebuild the OS for kali in case i botched something there.  I'll
also try kali with just one CPU.

 After that I'll open a PR unless anyone has any thoughts otherwise.

 -johan