Inspired by Chuck's House of VAX, and eager to relive those halcyon days
of my youth in which we happy few Waterloo undergrads shared a
VAX-11/780 running 4.1 BSD with a load average around 100, I bought a
VAXstation 4000/60 off eBay last summer.  It arrived damaged, with about
fifteen gold leads mashed together on two of the QFPs.  It languished in
the garage.  (Later I bought two more 4000/60s, unnecessarily, it turned
out.)

[I apologize if any of the following recollections are in fact
mis-recollections.]


1. FIRST INSTALL TRAVAILS

I recently installed NetBSD 1.5.2 for the first time, on one of the'60s.
Here's what worked for me.

I built a MMJ-DB9 cable (but still get lots of CRPT errors unless I plug
it in after the machine starts self-tests).  I flipped up S3 and got to
the console and experimented with ? and SHOW and SET and so forth.  I
attached an AUI-10baseT transceiver and flipped the network connect
switch to the AUI (10base5) side, away from the 10base2 coax side.  I
tried network boot using with B ESA0.  It worked ESA0 ... RETRY ... .

From a RedHat Linux 7.2 box, I fetched and built and fixed (loop.c) and
installed mopd.  It served up the MOPBOOT.SYS image (linked to
lowercase-mac-address.SYS in /tftpboot/mop).  Then I configured
/etc/bootptab and ran /usr/sbin/bootpd to serve BOOTP to the VAX.  Then
I configured and started NFS.  I kept having trouble with NFS mount
failing, but in the end it worked, although I don't recall how I fixed
my auth problem.)

At this point, I was booting esa0, loading MOPBOOT.SYS, loading
netbsd.ram, and running sysinst.  I had downloaded binary/sets/*.tgz to
a PC serving FTP, and attempted to install these sets by FTP.  Sysinst
failed in network configuration.  My mistake was in answering all the
prompts, gateway, hostname, etc.  When I would do that, I would get some
peculiar error about unable to contact the gateway (IIRC).  Curiously
enough, my network *was* running, so I was able to drop down to /bin/sh,
mount /dev/sd0d on /mnt, and fetch the binary sets via FTP.  Then I
reran sysisnt but alas, it insisted on repartitioning my sd0 drive.
Fortunately I had a second disk in this '60, so I did the same thing,
but to sd1d.  Then I reran sysinst and took the binary sets images from
sd1d.  Then I booted NetBSD from my hard disk!  Wonderful!

In the VAX console I SET FBOOT 1 and SET HALT 2 (or was it 3?).  This
seems to be all that is needed for the machine to quickly boot to NetBSD
automatically on power up.

I studied /etc/rc.d/, especially network, and /etc/rc.conf and found to
set a static IP address for a machine with an le0 ethernet I should set
up /etc/ifconfig.le0; for DHCP I should instead (of ifconfig.le0) add
these lines to /etc/rc.conf:
hostname="<whatever>"
dhclient=YES
dhclient_flags="le0"

Then I tried to install NetBSD on the second 4000/60, which had only one
disk.  The sysinst network configuration problem, and its "ftp (from
/bin/sh) the sets to a second hard disk" workaround would not work
there.  After some experimentation, I found that if I entered nothing
("") to most of the network configuration prompts, except for host IP
address, then network configuration just worked, and I was able to fetch
the binary sets with ftp, from inside sysinst.  Much nicer.

Then, inspired, I used a very fine knife and carefully straightened all
the mashed QFP pins on the first, broken 60.  It booted fine and NetBSD
installed easily.


2. 4000/90 TRAVAILS

I also purchased a 4000/90 on eBay.  It just arrived, in bad shape, with
the graphics daughtercard partially sticking out the back of the
machine, with its two-row fine pitch motherboard connector broken and
header pins bent and shorted.  A "loose screw" rattling around under the
motherboard was revealed to be a capacitor that had broken off the
graphics card.  Again, I carefully straigtened the pins, and reinserted
the graphics card, and booted.  Unfortunately, this card exhibits
errors.  Fortunately the presence or absence of the card seems to have
no effect on NetBSD.

As soon as I got to the console >>> prompt, things seemed a little
peculiar  When I would issue a command, even ? or SHOW CONFIG, the text
did not burst forth at 9600 baud, but rather in fits and starts -- blast
out three or four lines of text; pause for 100 ms or so; then another
couple of lines; and so forth.  As I recall, at the time, removing the
questionable graphics card did not seem to help; but now the problem is
gone and the '90 console writes text just as fast as the others do.

This time, I installed NetBSD on an old 4 GB Seagate drive.  (BTW I net
booted off EZA0, not ESA0 as on the '60s).  This worked beautifully.

Then I found I could not auto boot on power up because of an error, SYS
0512 or some such, which discussions on this list identify as a
corrupted NV RAM problem.  I also find that whatever I set HALT to, in
the range 1-3, it gets reset to 0 (on next power cycle?).  Oh well.

This old hard drive was noisy, so I tried a newer SCSI-3 68-pin SCSI
Compaq 10K RPM drive, OEM'd from Seagate, I think.  I used a 68-50 pin
adapter; the drive was already in SE mode.  This drive seemed fine
during show config.  Using netbsd.ram, during autoconfig, I get some
interesting messages:

  probe(asc0:1:0): max sync rate 6.25MB/s
  probe(asc0:1:0): unrecognized MESSAGE EXTENDED; sending REJECT
  probe(asc0:1:0): max sync rate 20.83MB/s

Is this benign?  I assume the drive is offering commands or information
that NetBSD will not use.

After booting netbsd.ram, and fully installing the system using sysinst,
I was disappointed that no matter what I tried, I could not then boot
rka100 (or whatever) from the VAX console.  If stopped, the drive
doesn't even spin up.  I surmise that the drive speaks a newer richer
SCSI command protocol, too rich for the old VAX ROMs' taste.
Fortunately, after booting NetBSD via another disk, I have had no
problems mounting and using the new drive under NetBSD.

This new 10K RPM drive is FAST.  time cat /usr/bin/* >/dev/null takes 9
s using the old drive and about 6 s using the new one.


3. VLC INSTALL "CAKEWALK"

No problems at all!  Last week, a friend went to the RE-PC (computer
surplus) in Tukwilla, WA, and bought each of us two VAXstation VLCs at
$30 each.  (I think the Q1 starting price is $40.)  These installed
NetBSD beautifully.  My friend is running OpenBSD on his two.  There are
about 5-6 more, by the way, with keyboards, not that they are of any
use.

I also picked up 6 AUI-10baseT transceivers at Boeing Surplus (Renton,
WA) for $5 each.  BTW, Boeing Surplus had some pretty hefty SGI Indigo^2
IMPACTs and IMPACT 10000s's for $50-$150, and even a couple of (very
substantial) Onyx Infinite Reality's (?) for $200.


4. THANK YOU

I second the recent sentiments of Matthew Sell -- I am very grateful for
this fantastic, free, high quality product.  It is quite apparent how
much effort and craft and love have been poured into it.  Thank you.  I
have already learned a lot, had a lot of fun, have been rereading my old
4.3 BSD text, and now (thanks Chuck) certain family members now know
that while Kestrels may be push-overs, you should often let sleeping
Ice-monsters be.

I design and build new processors in FPGAs (see www.fpgacpu.org) and I
am getting excited about porting NetBSD to a new FPGA CPU.  (First job
is porting GCC, glibc, binutils, and GDB, but I have some big short-cuts
in mind there.)


5. QUESTIONS

1. Re: 4000/90 corrupted NV RAM (SYS 0512): Has anyone fixed theirs, and
if so, how?  An earlier message indicated a MOP program might fix it.

2. Why can't my VAXen boot off my newer 10K RPM drive?  Is there
anything I can do except netboot my kernel off an NFS share, or boot
locally from an older SCSI disk?  Can I boot (fetch netbsd) off one disk
but use another for my entire file system, including the root partition?

3. For automatic boot on power up on my other (non-90) machines, should
I set HALT to 2 or 3?

4. I have 1 4000/90, 3 4000/60s, and 2 VLCs.  I think I want 1, 1, and
2.  I'm willing to donate one of the unwanted 4000/60s to the NetBSD/vax
effort -- do you folks need a new build or test machine...?

5. I am looking for more RAM for my 4000/90.  Can I use 4000/60 memory?
What other systems' memory is compatible with 4000/90?

6. What is the definitive story with the MMJ-DB9 cable CRPT bit errors?
I even tried an H8571-J (see
http://www.partner.compaq.com:9003/public/cheat_sheets/cables/padapters.
html) on my PC's DB-9 serial port, and I still see these errors.  I also
see them on a cable with just TXD+ TXD- RXD- RXD+ wired up.  I find the
only thing that definitely works is to disconnect the cable, power on
the machine, wait until it is well through self-test, then connect the
cable.

7. With any of the graphics displays (VLC, 60, 90) I think -- is it
possible to access and write the frame buffer RAM?  I have searched
google for specs on these cards to no avail.  Are there VAX VMS driver
sources available that would reveal how to address the frame buffer
memory and/or control regs?

8. Does anyone have a source for the DEC adapter that converts 3 BNC
with sync on green into a standard VGA connector?

9. Has anyone else ever seen the VAXstation 4000/90 console print output
in "fits and starts"?  Any idea what the cause might be?


Thanks again everyone.

Jan Gray
Gray Research LLC