The problem turned out to be a bug in the Sun
diagnostic boot PROMs; i.e., the system would fail
when the diag/norm switch was in diag mode.
This seems to be a feature of Sun 4/280 CPUs,
at least those with boot PROM 3/3.0; the system
will not pass the diag self tests with more than
48 megabytes of RAM (at least, 56 and 64 megs will
fail self test :-(.
I had initially misconfigured the Dataram 32 meg board
that I was trying to add, and had placed the CPU in
diag mode to get more info about the failure. Once
properly configured, I continued to test against diag
mode (not norm) because there was an obvious failure.
It didn't occur to me to try normal mode until today,
when we put my Dataram board into another 4/280 which
had a working Dataram board in it. It failed, of course,
in diag mode; but then so did the formerly working Dataram
board we had swapped mine for. Bingo! Try both boards
in normal mode and they work. Argh! I guess that implies
that you can not use diag for anything useful if you have
more than 48 megs in your system (since it will always
crash in the ECC tests).
So now I'm up and happily running with 56 megs: one
32 meg Dataram board, and 3 Sun 8 meg boards
Thanks to all that helped:
rct@fid.Morgan.COM (Robert Terzi)
zjat02@trc.amoco.com (Jon A. Tankersley)
Mike Raffety <miker@sbcoc.com>
tester@cmcl2.NYU.EDU (thx1138@nyu.edu)
And a special thanks to
rlg@ida.org (Randy garrett)
who took the time to answer several queries of mine
and
wargo@cs (Dave Wargo)
who let me cram my board in his machine.
and
Allen Hirschorn and Kathy at Dataram customer
support (good folks)
Mark Wallen
UCSD, cogsci
mwallen@ucsd.edu
The original problem:
-----------
From: wallen@cogsci.UCSD.EDU (Mark R. Wallen)
Message-Id: <9112160513.AA04927@cogsci.UCSD.EDU>
Date: 15 December 1991 2113-PST (Sunday)
To: sun-managers@eecs.nwu.edu
Subject: adding memory to a sun4/280
Hi
I'm trying to add a 32meg Dataram board to a
sun4/280, which has/had 4 8meg sun memory boards
(for a total of 32megs). I believe that the sun4/280
can only have 4 memory boards (boards numbered 0-3,
which you configure via jumpers on each memory board).
Sooooo, I pull out one of the sun 8meg boards (board 0)
and replace it with the Dataram board. This should give
me 56megs total. However, the system won't boot; it dies
in self test looking for board number 4 (a fifth memory
board!).
The system works with the Datram board and 2 sun 8meggers
(for a total of 48 megs). Adding the third sun 8meg
causes the system to want to look for board #4 (a
fifth memory board) during the ECC selftests; this produces
a bus error and causes a loop.
I've received a hot spare from Dataram which gives the
same results. I've swapped sun 8meg boards and slots as
the 3rd sun board, so I don't think it's a particular sun
memory board or slot. And Sun came in and swapped the
CPU board as a test, with exactly the same results.
So I'm stumped. The Dataram folks say that there is
a new prom out, version 3.03, which is needed to run
more than 56 megs. Actually there is a small debate
about whether you need the new proms to run 56 and
above megs or 64 and above megs. Both my CPU and the
spare Sun brought in are version 3.0, and the Sun
FE says that 3.0 is the newest version of the proms
and has never heard of 3.03.
I've also pulled out the nonstandard peripherals,
thinking that a device register might conflict with
one of the memory board registers. Same lack of joy.
I suspect something simple/obvious in my setup, soooooo
I'm askin'. Has anyone a similar setup? Had any problems
making it work? Is there a version 3.03 of the
sun4/280 boot prom? Is there a Santa Claus?
Thanks in advance for any pointers
Mark Wallen
Cognitive Science, UCSD
mwallen@ucsd.edu
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:21 CDT