Subject: Re: SCSI on Q-bus
To: None <port-vax@NetBSD.ORG>
From: John Wilson <wilson@dbit.dbit.com>
List: port-vax
Date: 02/02/1998 11:53:30
>From: "emanuel stiebler" <emu@ecubics.com>
>what is a "realistic" price for a Q-bus SCSI controller supperted by NetBSD?
About $500-700 is what I've seen from used dealers.
>What makes this scsi so expensive ?
>Is there a patent on this MSCP, TMSCP protocols ?
Yes, and DEC charges $100 per board for licenses (and if the board does both
MSCP and TMSCP I'll bet that's $200). The patent runs out in about 3 more
years for what that's worth.
>I'm thinking about developing one for the q-bus, should be cheaper than
>.., and i could use my disks here.
>(its not my first one, i build such stuff on vme-bus before)
Sounds like a cool project. Just so you know, the SCSI side is the easiest
(since you can use your favorite SCSI chip to do all the work), being a
Q-bus bus master is a bit more complicated, and writing the firmware to do
MSCP/TMSCP is quite a project. The "cleansed for external use" MSCP doc is
apparently no longer sold by DEC, and there was never an external TMSCP doc.
I've been working on reverse-engineering TMSCP lately, from the various OS
drivers and from doing all kinds of testing on a TQK50, I've been thinking
of writing a web page documenting what I've found so that other people won't
have to go through the same pain all over again! It's mostly what you would
think (read/write/compare do one record at a time, OP.WTM writes a tape mark),
the "reposition" command is a little tricky since it has so many options,
and it seems just about every kind of return status triggers a "serious
exception" -- I found that RSTS/E reads an 80-byte label with a 78-byte buffer
and things don't work if it doesn't get a serious exception for that. It
sure doesn't seem like a serious error to me, the mismatched block sizes are
already known from the "read" response packet. Also, all the I/O commands
fill the LBN field of the response with the current tape position, which is
0 for BOT and increments by 1 for every record or tape mark.
Most of the nastiness of (T)MSCP is in the underlying protocol, the commands
themselves are pretty straightforward.
John Wilson
D Bit