hawk.ro / stories / NetBSD on MicroVAX

NetBSD on MicroVAX

"Of course it runs NetBSD!"

The machines

After getting neofetch to run on 386 and 486 based computers, I wanted to try and use an even older machine. I have two older VAX computers in my collection. One is a heavily upgraded MicroVAX 3600 (that I use to print my calendar), the other a MicroVAX II. There is a significant speed difference between the two, and the MicroVAX II doesn't have a working HDD, so I thought I'd better start with the 3600. As result of its upgrades, it has 2 QBUS SCSI controllers. It also has 3 internal HDDs on one controller, while the other has an external standard SCSI connector. By the way, none of the controllers from the MicroVAX 3600 would fit the MicroVAX II, one for electrical reasons, the other for mechanical ones. Long story.

I don't want to erase the internal drives of the MicroVAX 3600 but I have a huge (in size; around 300MB capacity) external SCSI HDD, also from DIGITAL. That one can easily be imaged to a file using a modern computer (a crude but fast form of backup&restore). That leaves me with a free HDD on which to install NetBSD.

The huge SCSI HDD. The 3.5" IDE drive on top of it is there for scale.


I know that this VAX is capable of booting from SCSI CD-ROM but it requires a particular type of drive that I don't have at hand. Another option - that doesn't seem very difficult - is netbooting. Yes, these old computers were capable of netbooting all by themselves. Of course, it's not PXE, but a proprietary protocol by DEC, called MOP (Maintenance Operations Protocol). This protocol would also allow for things like remote console over Ethernet, but I'm only interested in booting. mopd is not included in Slackware, nor is it available at SlackBuilds but there are sources at https://github.com/qu1j0t3/mopd.

Next hop: A small documentation error that took a while to untangle. Following the instructions at http://www.NetBSD.org/docs/network/netboot/mop.html:

First, you need to get a copy of the bootloader. This can be found in the NetBSD distribution as: installation/netboot/boot
I get the file, copy it to /tftpboot/mop/ with the right name but after loading it, the VAX HALTs. After multiple experiments (thinking that maybe something is wrong with my setup/VAX/whatever), I found out that in NetBSD 1.5.3 distribution, the file is called boot.mop. That one works. Also, while reading about mopd I learned out about mopcopy(1) - a NetBSD utility that performs this exact conversion, from ELF or a.out executable to MOP image. So now I can use the proper boot file for each distribution (this will prove handy later).

Next, NetBSD boot expects to read the kernel from NFS, not from TFTP. Edit /etc/exports, add a directory for that purpose (/export/micro), copy install.ram.gz there (as netbsd.vax since that's what it was trying to read), then start nfsd (on Slackware /etc/rc.d/rc.nfsd needs to be chmod-ed +x). Once everything is in place NetBSD boots and not much later sysinst displays the first screen.

First install attempt

Spoilery title, I know

English, install NetBSD to a hard disk, continue, I looked carefully to make sure I will install to the right HDD (ra3, fourth one), Full installation, Set sizes of NetBSD partitions, 283MB for root partition, 32MB swap. Continue. Progress bar (recommended). Install from FTP, configure network, server (on local network) and let's proceed. Everything runs fine (at around 300KB/s transfer rate) until.. it no longer does. System seems to be hanged but it responds to ping (so kernel and network are still running). The external HDD LED remains lit (that seems to indicate a SCSI error). One more attempt offers a similar result, however it hangs at a different point in transfer.

Thinking a newer BSD might eventually help, I go for 7, but first attempt (using boot.mop from 1.5.3) quickly gets into a loop (I switched from the VT-220 terminal to my regular machine running screen in order to log what's happening):

>> NetBSD/vax boot [Jan  6 2002 22:13:30] <<
>> Press any key to abort autoboot 0
Trying BOOTP
Using IP address:
myip: vaxnetbsd.local (
root addr= path=/export/client/root
3733560stray interrupt: vector 0x18, ipl 31
stray interrupt: vector 0x18, ipl 31
stray interrupt: vector 0x18, ipl 31
stray interrupt: vector 0x18, ipl 31
...ad infinitum

The same happens on SIMH. I fire up a NetBSD machine, mopcopy VAX boot and install.ram.gz from 8.0 (why not?), copy the result to my Linux machine and try again. This time it loads the kernel but then panics right after network card detection:
Copyright (c) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005,
    2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017,
    2018 The NetBSD Foundation, Inc.  All rights reserved.
Copyright (c) 1982, 1986, 1989, 1991, 1993
    The Regents of the University of California.  All rights reserved.

NetBSD 8.0 (INSTALL) #0: Tue Jul 17 14:59:51 UTC 2018
MicroVAX 3500/3600
total memory = 32708 KB
avail memory = 27328 KB
mainbus0 (root)
cpu0 at mainbus0: KA650, CVAX microcode rev 2 Firmware rev 83
lance at mainbus0 not configured
uba0 at mainbus0: Q22
mtc0 at uba0 csr 174500 vec 774 ipl 17
mscpbus0 at mtc0: version 2 model 14
mscpbus0: DMA burst size set to 4
uda0 at uba0 csr 172150 vec 770 ipl 17
mscpbus1 at uda0: version 5 model 13
mscpbus1: DMA burst size set to 4
qt !Turbo
qe0 at uba0 csr 174440 vec 764 ipl 17: delqa, hardware address 08:00:2b:0a:eb:3d
machine check 10
      vap 80 istate1 6c istate2 dbe0fc80 pc ffc07403 psl 80009c45
dmaser=0x88<QBNXM,LOSTERR> qbear=0x0 dmaear=0x0
 parity error:        cacheing disabled
memory err!
panic: mchk

        PC = 8000C25B
This doesn't happen on SIMH. I don't think there is an actual defect with this machine (it runs fine and I tested it with both network and disk activity under VMS just to make sure) but its configuration might be too unusual for NetBSD. I suspect the disk issue to be with that secondary controller. These SCSI controllers are not like the PC ones, i.e. they are transparent as far as the OS is concerned. The OS doesn't know it's not using a typical MSCP (Mass Storage Control Protocol) unit. All the SCSI processing is done in the controller's firmware, however, the MSCP it implements might be subtly different from what NetBSD expects. After all, we're talking about a time when NetBSD was more than 10 years into the future. However, the two controllers are different, one is Emulex UC08, the other is CMD Technology CQD-223, the CQD-223 is the one I've tried. But for the moment I have another idea.

MicroVAX II to the rescue!

Why not try diskless NetBSD? Since I already know NFS works, how hard could it be? (Yes, I know it's not a good question to ask rhetorically). I started light, with the same NetBSD 3, but instead of following instructions carefully, I skipped some steps (crucially I set rc.configured=YES in /etc/rc.conf and forgot to MAKEDEV). And I also forgot to create dev/console beforehand. Oops! A "mknod /export/uvaxii/root/dev/console c 0 0" on the server and a reset (of the MicroVAX) solved this last one and after a while I got the login prompt. The only trouble was that, despite answering vt220 (or even vt100) at login, it didn't like my terminal and vi would not work. However, entering tset vt100 seems to solve the issue. Good enough for now.

Having the real thing up and running, I start another instance of NetBSD 3 on SIMH/VAX to compile bash (the emulated one, without NFS, is much faster). Transfer the binary to the exported filesystem and check that it's running, then...

First run finished after almost 13 minutes
The entire setup. The cable on the left connects the MicroVAX II to the VT-220 (I only have one VT-220)
... it mostly works. It seems to be stopping (crashing?) early, I assume it expects something that's missing. However, it crashes just the same on SIMH so there's that.

While I'm rather pleased with the above, I really would like a full run with all the information. I try to mount /proc (and /kern) thinking that maybe it would help, but no. In fact /proc doesn't look to have anything in it besides process information.

Going for newer

Next attempt is NetBSD 7.1.1. Quite a huge leap, I know, but I have it available on a PC machine and I want to try cross-compiling a custom kernel - and in the process I discover that it's surprisingly easy to cross-compile NetBSD. I configure the kernel for this machine (I should have tried the same for the MicroVAX 3600 -- maybe later) and start the build. In the meantime I start the diskless install for 7.1.1, meaning:

# mkdir -p /export/uvaxii/root/dev
# cd /export/uvaxii/
# mkdir usr
# mkdir home
# cd root/
# for i in ~hawk/netbsd7/sets/*.tgz ; do tar --numeric-owner -xvpzf $i ; done
# mv usr/* ../usr/
# mknod dev/console c 0 0
# mkdir kern
# mkdir home
# mkdir swap
# dd if=/dev/zero of=../swap bs=4k count=4k
# chmod 0700 ../swap
# cat - >etc/ifconfig.qe0
inet netmask broadcast
# cat - >etc/fstab
hgw:/export/uvaxii/swap   none  swap  sw,nfsmntpt=/swap
hgw:/export/uvaxii/root   /     nfs   rw 0 0
hgw:/export/uvaxii/usr    /usr  nfs   rw 0 0
hgw:/export/uvaxii/home   /home nfs   rw 0 0

I don't know if etc/ifconfig.qe0 is really needed but it can't hurt. Then add to etc/rc.conf:

# from NetBSD instructions
# more defaults that I want to override

and to etc/hosts:    hgw.local hgw   uvaxii.local uvaxii

On the server side, add to /etc/exports:

# for diskless MicroVAX II

In fact usr and home should not need "no_root_squash" except that I need to install a binary package (bash) and after that, add a regular user and configure its home directory. Next is /etc/dhcpd.conf on the server:

host uvaxii.local {
	hardware ethernet 08:00:2b:06:17:c2;
	always-reply-rfc1048 true;
	always-broadcast true;
	option root-path "/export/uvaxii/root";

Restart nfsd and dhcpd. I then configure SIMH to use the same hardware address as the MicroVAX II in order to finish the installation quicker (MAKEDEV, install bash) then boot it. Oops, wrong kernel (I really meant it when compiling just for the MicroVAX II, SIMH emulating a Microvax 3900 halts). Try again with generic kernel. This time... no

>>>b xqa0


>> NetBSD/vax boot [1.12 (Tue Jul 17 14:59:51 UTC 2018)] <<
>> Press any key to abort autoboot 4
Press '?' for help
> boot netbsd.generic
Trying BOOTP
Using IP address:
myip: uvaxii.local (
root addr= path=/export/uvaxii/root
3101284+172988 [230096+220228]=0x38d948
Copyright (c) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005,
    2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017
    The NetBSD Foundation, Inc.  All rights reserved.
Copyright (c) 1982, 1986, 1989, 1991, 1993
    The Regents of the University of California.  All rights reserved.

NetBSD 7.1.1 (GENERIC.201712222334Z)
MicroVAX 3800/3900
total memory = 65468 KB
avail memory = 59216 KB
mainbus0 (root)
cpu0 at mainbus0: KA655, CVAX microcode rev 6 Firmware rev 83
lance at mainbus0 not configured
uba0 at mainbus0: Q22
dz1 at uba0 csr 160100 vec 304 ipl 17
mtc0 at uba0 csr 174500 vec 774 ipl 17
mscpbus0 at mtc0: version 5 model 3
mscpbus0: DMA burst size set to 4
uda0 at uba0 csr 172150 vec 770 ipl 17
mscpbus1 at uda0: version 3 model 3
mscpbus1: DMA burst size set to 4
qt0 at uba0 csr 174440 vec 764 ipl 17
qt0: delqa-plus in Turbo mode, hardware address 08:00:2b:06:17:c2
mt0 at mscpbus0 drive 0: TK50
mt1 at mscpbus0 drive 1: TK50
mt2 at mscpbus0 drive 2: TK50
mt3 at mscpbus0 drive 3: TK50
ra0 at mscpbus1 drive 0: RA92
ra1 at mscpbus1 drive 1: RA92
ra2 at mscpbus1 drive 2: RA92
ra3 at mscpbus1 drive 3: RA92
ra0: size 2940951 sectors
ra1label: 70fc
: no disk label: size 2940951 sectors
ra2: attempt to bring on line failed:  unit offline (not mounted) (code 3, subcode 1)
ra3: attempt to bring on line failed:  unit offline (not mounted) (code 3, subcode 1)
boot device: qt0
root on qt0
nfs_boot: trying DHCP/BOOTP
nfs_boot: timeout...
nfs_boot: timeout...
nfs_boot: timeout...
nfs_boot: trying RARP (and RPC/bootparam)
revarp failed, error=51
Supported file systems: union umap tmpfs ptyfs procfs overlay null nfs mfs lfs kernfs ffs
fdesc cd9660
no file system for qt0
cannot mount root, error = 79
root device (default qt0): 

Right, SIMH is emulating a different network card and that gets a different device name. Maybe copying etc/ifconfig.qe0 to etc/ifconfig.qt0 would help? No. tcpdump reveals that the DHCP server responds but for some reason NetBSD doesn't get the answer. After much head-scratching and experiments I notice that the real MicroVAX 3600 (with a regular DELQA card, qe0) works (to be investigated, why doesn't it panic?). Fortunately, SIMH supports setting the network card as DELQA (set xq mac=08-00-2b-06-17-c2 type=delqa in vax.ini) and this time it boots.

boot device: qe0
root on qe0
nfs_boot: trying DHCP/BOOTP
nfs_boot: DHCP next-server:
nfs_boot: my_name=uvaxii.local
nfs_boot: my_domain=local
nfs_boot: my_addr=
nfs_boot: my_mask=
nfs_boot: gateway=
root on
root file system type: nfs
/etc/rc.conf is not configured.  Multiuser boot aborted.
Enter pathname of shell or RETURN for /bin/sh:
We recommend that you create a non-root account and use su(1) for root access.
uvaxii# mount /usr
uvaxii# cd /dev
uvaxii# /bin/sh MAKEDEV all
uvaxii# cd /
uvaxii# swapctl -A
swapctl: adding hgw:/export/uvaxii/swap as swap device at priority 0

MAKEDEV took a while even on the faster SIMH. exit in order to get a new sh that would ask for a terminal, edit /etc/rc.conf and set rc.configured=YES then proceed with multiuser. Download bash binary package from pkgsrc and copy it to the exported filesystem, then

pkg_add bash-4.3.039.tgz
This is when I realized the problem with root_squash for usr. Make the appropriate changes to exports, restart and finally install bash.
useradd -m -s /usr/pkg/bin/bash hawk
Then I make a mistake. I add a password to user hawk. The reason is that I want to be able to telnet into this machine (by enabling telnet in /etc/intetd.conf, with -a off for "insecure" access). However, telnet doesn't work. Everything else works though - on SIMH. A test run of logging in as hawk followed by running neofetch succeeds. Time to move on to real hardware.

NetBSD 7.1.1 on MicroVAX II with 16MB of RAM and no HDD

Some minutes after power-on I get a login prompt. Enter hawk, wait, wait some more, a password prompt appears, enter the password and then... nothing. Oops. login timed out after 300 seconds. Maybe I didn't press ENTER? Try again. Same. I assume it has something to do with newer password encryption (that is taking too long on a 5MHz CPU). Luckily root is passwordless. Indeed, login as root works, su - hawk, ./neofetch

Around 20 minutes later
Now it shows memory but that causes the flag to scroll

neofetch done, I'm going back to NetBSD 3 for a little slacking. Not in the Slackware sense, true, but nevertheless...

Before the Elder Scrolls, before almost any other dungeon delving game, there was .. rogue

Published 2019-05-25 by Mihai Gaitos - contacthawk.ro