4.3-QJ0 installation problem (was: no subject)

[Michael Sokolov]

(I'm Cc'ing this to the PUPS list because the original message was, but this
 discussion belongs on the Quasijarus list. Please don't Cc follow-ups to PUPS,
 instead everyone who is interested in this discussion please send:

 subscribe quasijarus

 to Majordomo at minnie.cs.adfa.edu.au)

Wilko Bulte <wilko at yedi.iaf.nl> wrote:

> I've been trying to get 4.3BSD to run on my newly acquired MicroVAXII.
> I followed the Ultrix route described in the docs in the pups tree.
>
> I get as far as:
>
> [...]
>
> 4.3 BSD Quasijarus UNIX #0: Fri Dec 25 14:22:17 EST 1998
>     msokolov at polygon:/usr/src/sys/GENERIC
>
> [...]
>
> ra0 at uda0 slave 0: MICROP  , size = 1303998 sectors

OK, good, I see you've labeled your system disk.

> ra1 at uda0 slave 1trap type 6, code = 2, pc = 80031b1c
> panic: Arithmetic fault

OK, trap 6 code 2 is integer divide by zero on the VAX. My obvious guess is
that the disk has a garbage label on it and when the kernel tries to interpret
it, it divides by zero and blows up. A garbage label is something worse than no
label at all, because the label structure has a magic at the beginning, and
trust me, the kernel does check it and it does not blow up with a divide by
zero when block 0 is all zeros.

Wilko, what exactly do you have in the label block of disk 1? If you've been
following my installation instructions to the letter, that disk would be your
Ultrix disk. My installation instructions call for labeling the BSD disk, but
not the Ultrix disk. In fact, putting a BSD label on an Ultrix bootable disk
would render it unbootable, as Ultrix has boot code where BSD has the label.
This means that normally when someone follows my Ultrix-based installation
procedure, BSD will simply view the Ultrix disk as unlabeled and make it one
big partition a. You obviously have something else in there.

> Exactly the same thing happens when I use 4.3reno instead of the Quasijarus
> kit.

Well, this at least means that this is not yet another one of my own bugs, so
that's the good news. :-) But sure, the kernel could do with a few more label
sanity checks so that it prints a nice error message instead of blowing up.
I'll look into it.

--
Michael Sokolov
Special Agent
International Free Computing Task Force

Harhan Computer Operation Facility

ARPA Internet SMTP mail: msokolov at baryon.trailing-edge.com

Received: (from major at localhost)
	by minnie.cs.adfa.edu.au (8.9.3/8.9.3) id NAA47798
	for pups-liszt; Sat, 28 Aug 1999 13:58:50 +1000 (EST)