[Simh] VAX 780 emulation - How to install Netbsd

[Mark Pizzolato - Info Comm]

On Sunday, December 28, 2014 at 9:02 AM, Anders Magnusson wrote:
> Mark Pizzolato - Info Comm skrev den 2014-12-28 03:02:
> > As it turns out, there were issues with booting and basic installing of
> NetBSD on the older VAX systems.  Much of NetBSD evolved after the
> relatively low cost MicroVAX systems were very common.  It seems that
> although some folks with older VAX hardware (VAX750, VAX780, VAX730
> and VAX8600) may have been running NetBSD, they didn't install it from CD
> media, but merely upgraded existing older BSD operating system installs.
> I may be able to give some explanations here that can be helpful (as I
> was the one that actually ported NetBSD to VAX).
> Modulo some memory errors, this was >20 years ago.
>
> It was originally ported to 750 /w RP06, since that was what I had
> available.  I wrote three ways to load /boot (required both for
> installation and booting the system)
> - Fetch from the TU58 (to also get the needed microcode update). Older
> boot ROMs only accepted microcode patching if booted from TU58 by some
> reason.
> - Fetch from TU77 (key in a program in hex on the console).
> - Fetch from RP0x (using boot ROMs).
> 
> The boot program could _not_ be loaded from VMB, neither from RA disks
> (missing boot ROMs, was added by DEC to some machines).
> 
> Then we got hold of a 780, so I ported over NetBSD there.  Could only be
> installed from TU77 via hex load code typing.
> But;  subsequent loading of the system required boot to be loaded from
> the RX01 floppy, so it had to be copied there (using arff).
> boot could not be loaded via VMB, it had to be put in memory from the
> loader script.
> Note that the NetBSD 1.2 VAX release were put together on a 11/780. It
> took 11 days to compile the system and 2 to gzip the distribution files :-)
> 
> The 8600 port was done similar to the 780.
> 
> The MicroVAX port was written by Rick Macklem, and since these machines
> have some sort of sub-VMB in ROM the first parts of VMB-support were
> added.
> 
> Around 2000 the boot procedure was finally altered so that VMB could be
> used as the boot program.
> Also other stuff like memory sizes are read from the RPB setup by VMB
> these days.
> But; when this was done virtually no big VAXen were still running, so
> the support for them may have bit-rotten.  This is what Johnny always
> encounters when he tries to power up his 8600 ;-)
> 
> So; bottom line:  If installing NetBSD on a non-MicroVAX machine, it is
> a safe bet to put the boot program on console media :-)

Good information.

Meanwhile, VMB's BootBlock support didn't require the booted program to have any knowledge of or take any information from whatever efforts VMB may have taken prior to loading the sector 0 contents and starting execution at offset 2.  This loading of the sector 0 contents and execution start at offset 2 was what the 750's console ROM was doing when booting from any particular device (RP, TU58, etc.).  This convention was initially defined when the 750 was first implemented.   On the 750 all operating systems booted with a BootBlock boot.  The original UCB 4.2BSD work done on VAX systems leveraged this paradigm as well.  Hence, simh's current use of VMB as the initial load program merely performs what the 750's ROM was doing anyway and thus makes a best effort to reflect original hardware behavior.  Anyone wanting to load an operating system via a different method can certainly use the LOAD command to directly load a file into memory and the start its execution directly.

- Mark