perry@sbcs.UUCP (Perry Kivolowitz) (08/19/83)
We at Stony Brook recently installed 4.1Bsd on the first of a bunch of
750's. We would like to take this opportunity to share a little of what
we learned in the hopes that we can save someone (anyone!) some grief.
First: The version of 4.1 we installed has a revision date of Sept 1,
1981.
Second: The configuration.
Vax 750.
2Mbytes Memory.
Floating Point Board (although this shouldn't enter into the
following discussion).
Uda50 Disk Controller.
DZ 11's.
TU-80 - (man, oh man, if ever there was a tape drive intended
for PATIENT people.)
RL11 - with an rl02.
RA80 -
RA81 .
________________________________________________________________________
TROUBLE GETTING BERKELEY BOOTSTRAP TO WORK.
------------------------------------------------------------------------
The TU-80 will work from a TS-11 bootstrap (and driver). We immediately
ran into problems in getting the Berkeley bootstrap to work. The first
problem turned out to be a schiz'ed cpu board which would run the VMS
16 user simulator (UETP) but failed its basic instruction set
diagnostics (you figure it out). The second problem is that DEC normally
sets the UDA-50 device address to 0772150. The Berkeley bstrap expects
this to be 0777550.
Solution: Have DEC install the UDA-50 at 777550 (or do it yourself).
________________________________________________________________________
NO BOOTSTRAP FOR THE UDA-50.
------------------------------------------------------------------------
We currently face this problem by hoping we don't have to reboot. One
day, some kind soul will send us one (hint!) We didn't receive any
UDA-50 programming spec's so writting the bstrap ourselves is out of the
question.
________________________________________________________________________
NO DRIVER FOR THE RL02
------------------------------------------------------------------------
Who cares. (But if someone happens to have one...)
________________________________________________________________________
NO SUPPORT (READILY) AVAILABLE FOR MULTIPLE DRIVES ON THE UDA-50
------------------------------------------------------------------------
The UDA-50 driver we received had knowledge of the RA-80 only.
Moreover, it dealt with only a SINGLE RA-80. Luckily, the hooks were
there to add multiple drive support as well as adding a second level to
the partition size data structure.
If anyone wants this - send mail. BUT - it really wasn't much trouble to
do (about twenty minutes).
------------------------------------------------------------------------
YES!!!! JUST COMMENT OUT N_DH and DH11 FROM SYMBOLS.RAW
------------------------------------------------------------------------
Nuff said.
------------------------------------------------------------------------
TU-80
------------------------------------------------------------------------
The TU-80 is great fun. Get one and hope for lots of rainy days. You can
cheer and take bets on when it'll find the beggining of tape mark. Also,
hire a grunt to make believe he's an automatic on-line feature.
________________________________________________________________________
IT RUNS NOW - BUT...
------------------------------------------------------------------------
o If you ever get into a situation where Unix wants to do a core
image save to disk (after a crash, for instance) there is no
UDA-50 DUMP routine (as noted in a previous letter to this news
group).
o SIT DOWN - Take a deep breath:
With a 1K blocking granularity (Cluster setting 2 in param.h)
block i/o on the UDA-50 takes place at a blinding 20 to 30Kbytes
per second. (You did not read that wrong - I DID say 20 to 30 K
bytes per second).
We took measurements which showed that we could achieve:
(ALL ON RAW DEVICE)
453Kb/sec with 64Kb transfers.
192Kb/sec with 4Kb transfers.
and 29Kb with 512 byte transfers.
What can be done? Well, the file system data structure allows
one more increment in the blocking factor - to 2048 byte units).
Adopting this will help a tad but cost in fragmentation.
We expect that the problem is due to having to pay the piper
each time we grab the Unibus. We believe that an entire rewrite
of the uba code is on order as the driver was probably written
before the 750 was on the scene (ie - Hey we have a Massbus -
Why fiddle with that ol'Unithingy).
We are working on this (but we don't want to).
------------------------------------------------------------------------
SUMMARY
------------------------------------------------------------------------
We are firm in our belief that a Vax is a great personal
workstation.
________________________________________________________________________
ETC
------------------------------------------------------------------------
Noone that we could find at Berkeley knew how to regen a boot
tape (ie the stand alone system's tape directory structure).
Any clues?
------------------------------------------------------------------------
Hope this helped someone somewhere.
Perry S Kivolowitz
Rick J Spanbauer
SUNY/Stony Brook
allegra!sbcs!perry
philabs!sbcs!perry
(516) 246-7146
so there!