🚀 go-pugleaf

In article <2766@milton.acs.washington.edu>, seymour@milton.acs.washington.edu (Richard Seymour) writes:
> A few months ago, i had a diskless VAXstation 3200 connected
> via an Able Qniverter booting vms v4.7 from an 11/750's Unibus
> cage with a UDA50 controlled RA81.
> Everything worked fine.
> 
> Last weekend, i tried it again, with vms v5.3-1 and a new model 3200
> card (fixed CQBIC chip, new console ROM).
> 
> It would not fully boot.  The initial "VMS v5.3" message appeared, 
> the screen cleared and everything got very quiet.

I think what is going on here, is the following:

As of VMS Version 5.0 Digital provided device driver writers the ability to
use the extended Qbus mapping registers available on MicroVAX II & MicroVAX
III qbus machines.  At that point in time, none of the DEC supplied drivers
used these routines themselves, but third party device driver writers could
use them.

As of VMS Version 5.2, Digital began using these extended Qbus mapping
registers in their drivers.  The driver that actually uses the extended
mapping registers now for your disk I/O is PUDRIVER which is loaded from
SYS$SYSROOT:[SYS$LDR] during the boot process, and only switched to part way
through the boot process. The ROM based driver which are used in the earlier
stages of the boot still only use mapping register which map the first 256K
of Qbus space.

In your configuration, the Qbus to Unibus converter only maps the first 256K
bytes of the Qbus into the Unibus since the Unibus is only 18bits that's all
you can hope for.  Since the VMS driver being loaded during the boot process
(PUDRIVER) doesn't know of this limitation in your hardware (as far as it
knows since it is on a MicroVAX III machine it has a 22bit Qbus, at it's
disposal), things get lost once it trys to start using bus addresses beyond
the initial 256K.

Other sites will also observe similar problems, with a VMS V5.2 or V5.3
upgrade, when using some third party disk controllers, which have not been
properly configured when they were initially installed.  Some of the
controllers in question include: Emulex QD01, QD21, QD32, QD33, DM01, DM02,
etc.  Proper configuration in this case, is to enable the 22bit option on
the controller and to install their 22bit addressing chip which was
initially shipped with almost all of these controllers. 

--
Mark Pizzolato - INFO COMM Computer Consulting, Redwood City, Ca
PHONE:	(415)369-9366	
UUCP:	mark@infopiz.UUCP or decwrl!infopiz!mark or uunet!lupine!infopiz!mark