Subject: Re: TK50 SCSI controller [was: VS2000 PS quirk?]
To: Aaron J. Grier <agrier@poofy.goof.com>
From: Tom Ivar Helbekkmo <tih+mail@Hamartun.Priv.NO>
List: port-vax
Date: 06/09/1998 22:52:01
"Aaron J. Grier" <agrier@poofy.goof.com> writes:

> I think it's only 4 AB/CD slots (or so it says on the cardcage.)

Oh -- I was so convinced it was 5, but OK, if it's marked with the
number of such slots, then that should be right.  Maybe not all BA123
systems have the same backplane, and I picked up the 5 from another
configuration?  Anyway:

> I actually have a quad-height drive controller, which is a RQDX2.  (oops)
> so I changed to this:
> 
> 1:    KA630
> 2:    MS630
> 3: M9047 empty
> 4:    M3104
> 5: TQK50 empty
> 6: DEQNA empty
> 7:    RQXD2

This won't do.  The reason the number of AB/CD slots before the AB/AB
ones is significant is "grant continuity".  Bus priority on the Qbus
is created by having the CPU, all the way up there at the top, send a
signal on a particular bus line, saying that the bus is available for
use.  Each module passes this signal along to the next (and the M9047
grant continuity cards do exactly this), until some module actually
wants the bus, at which time it doesn't pass the signal on, instead
sending a response back saying it's got the bus.

Now, this bus grant signal follows the AB slots.  Its path, assuming
the first four slots are AB/CD, goes like this:

  ---------   ---------   AB/CD
      v
  ---------   ---------   AB/CD
      v
  ---------   ---------   AB/CD
      v
  ---------   ---------   AB/CD
      v
  --------- > ---------   AB/AB
                  v
  --------- < ---------   AB/AB
      v
  --------- > ---------   AB/AB
                  v
  --------- < ---------   AB/AB

I'll stop there.  You see the criss-cross pattern.  Always straight
down the AB/CD slots, then meandering through the AB/AB ones, starting
at the left.  This path _has_ to be kept intact, all the way up to the
last module, which is to say that something has to be inserted into
all the slot halves that it goes through.

Note that there are other types of backplanes, for instance ones with
all AB/CD slots or all AB/AB slots, but we're talking BA[1]23 here.

> FWIW, the original configuration of the box (written on the label of the
> cardcage cover) has:
> 
> 1:    KA630
> 2:    MS630
> 3: M9047
> 4:    DHV11
> 5: TQK50 M9047
> 6:    RQDX2

Refering to the above chart, this is good.  The first four slots have
quad height modules, with the exception of slot 3, where a grant
continuity board has been placed to forward the bus grant lines, thus
keeping the position free for another memory board.  Then, the grant
passes to the TQK50, and through another continuity board to the old
RQDX2 disk controller.

What you've added to this configuration is a DEQNA.  The natural place
for it (for practical reasons and to sort out the priority in an
acceptable way) is in the half of slot 5 where the bus grant card is:

1:    KA630
2:    MS630
3: M9047
4:    DHV11
5: TQK50 DEQNA
6:    RQDX2

-tih
-- 
Popularity is the hallmark of mediocrity.  --Niles Crane, "Frasier"