🚀 go-pugleaf

In article <1992Dec27.195328.125@vms.huji.ac.il> allon@nesher.fh.huji.ac.il (Allon herman) writes:
>I am trying to install a non IBM scsi disk on my RT (running AOS 4.3).
>I am also aware that there has been recently a discussion regarding this topic
>in this newsgroup. I will be most thankful to the person who can tell me if
>and where can I find old articles that were posted on this newsgroup.

The IBM RT SCSI controller only supports differential SCSI disks (a different
electrical spec than most workstation-class disks); I also wonder how
flexible the driver is in handling any drive other than the Athens 400mb
drive (a humongous, noisy and power-hungry 8" or 12" Winchester which
appears like two 200mb devices to the software.)

Christer Bernerus at Sweden's Chalmers University developed a device driver
for the Adaptec 1540a/1542a which supports disks and tapes, though I've only
had experience with disks (a friend seemed to run into some problems with
using both a disk and a 150mb tape, but that may be pilot error--I don't know.)
I've had two SCSI disks running on it flawlessly for over 3 years now.
I'm not sure if the 1542a is still sold--it's a fast AT-bus SCSI DMA
controller card.

I don't know if Christer is still around: his email address used to be
bernerus@utc.chalmers.se
-- 
Steve Dyer
dyer@ursa-major.spdcc.com aka {ima,harvard,rayssd,linus,m2c}!spdcc!dyer

In article <1992Dec27.195328.125@vms.huji.ac.il>, allon@nesher.fh.huji.ac.il (Allon herman) writes:
> I am trying to install a non IBM scsi disk on my RT (running AOS 4.3).
> I am also aware that there has been recently a discussion regarding this topic
> in this newsgroup. I will be most thankful to the person who can tell me if
> and where can I find old articles that were posted on this newsgroup.

Those discussions on SCSI disks were most likely by myself.  
However.. I was using AIX and non-IBM SCSI disks.  Can be done, just have
to trick AIX and VRM into configing the disks.
I did not get into gory details on how I tricked AIX and VRM.
For AOS however.. I don't have it, so I don't know what would be
needed.

Here part of the article I posted on differential SCSI cut up to contain
some of the relavant info on RT scsi.

Subject: Re: What is differential SCSI?
Message-ID: <1645@airgun.wg.waii.com>
Date: 17 Nov 92 22:46:41 GMT
References: <1992Nov11.104513.1412@cs.hw.ac.uk> <BxKw13.AFs.2@cs.cmu.edu> <1992Nov16.095729.23423@cs.hw.ac.uk>

> |> You can run NON-IBM SCSI disks on the IBM SCSI adapter.  As long as
> |> they are either native differential, or run thru a Rancho diff-to-single
> |> converter box.  
 
> Is this convertor a two way thing?

Yes the converter is BI-DIRECTIONAL...  It HAS to be... 
There are several companies that manufacture the DIFF <--> SINGLE converters,
in addition to Rancho Technology, but that vendor makes the ones that I have
personally used.

The box is more complex than looks on the surface, as EACH pair of
differential signals and each single ended lead is BI-DIRECTIONAL.  It is
truely a PARALLEL BUS.  That is why it is called a SCSI BUS.  :-)
The converter must arbritate and decide which end is "speaking" and drive
the leads for the other side, and "listen" to the talking device.  Remember
in a SCSI bus, even though there is a "initatior" and a "target".  The 
target devices are intelligent.  They can be given a command by the host,
disconnect from the bus, perform the command, arbitrate for the bus, and
re-connect asyncronously and deliver status information.  Independently
of operations pending or ongoing on other devices on the same bus.
Some target devices can themselves become initiators, allowing a 
disk drive to back itself up to a tape drive.. At least in theory...
I don't think that anybody has MADE one that does that... :-)
You can also have TWO hosts attached to the same bus, sharing a
common device.  HOWEVER.. Both host adapters would need to 
co-ordinate access to the device, and beware of hidden terminators on
most HOST interface boards.  Specs indicate that terminators should only
be on EACH end of the bus.  Also the host adapter in the other computer
must be changed to a different unique SCSI address  (range from 0 thru 7),
than either the attached device or other host computer.

The IBM RT/PC SCSI board from IBM is designed to DIFFERENTIAL SCSI specs.
There are two wires for each signal lead, when one goes hi, the other goes
low, in a push-pull arangement.
Differential SCSI on shielded cable will allow up to 25 Meters total cable
length.  Single ended cable length total bus length is 6 meters.
Both bus types allow for up to 8 devices (initatiors OR targets) on one bus,
provided that the total cable length is not exceded.

Most PC and workstation vendors sell SINGLE ended SCSI devices.  Single-ended
devices have only ONE lead is used per signal line.  The interface is
slightly simpler electronicly is the primary reason, and most devices are
usually "under the covers" of the host machine anyway, and don't need the
extra cable length or noise immunity that differential scsi provides.

SO you could end up with this:

 +----+
 |    |              +--------+                   +------+   +------+ 
 | RT |  DIFF SCSI   |        |   single-ended    |      |   |      |
 |    |--------------+  CONV  | ------------------+ DEV1 +---+ DEV2 +-- T
 |    |  up to       |        |   total length    |      |   |      |
 |    |  25 meters   +--------+   6 meters to     +------+   +------+ 
 +----+                           the terminator.

T = Terminator 

NOTE:  The converter has no concept of IN or OUT.  It can have a 
single ended host, and differential devices just as easily.  

You can also  do this:

   Single                                                         single
   ended  <-- *N1 --> CONV <--- differential -- *N2 ---> CONV <-- ended --T
   host                                                           device
                                                                  *N3
*N1 - There could be single ended device(s) here..  6 Meters..
*N2 - And differential device(s) attached here...  25 Meters
*N3 - and of course single ended device(s) here..   6 Meters.
T   - Terminator.  The ranco converters have built in terminators for
      the differential AND the single-ended buses.
As long as there are no more that 8 total scsi controller devices.

The EXTREME example above actually has THREE buses.  Two single ended,
and one differential!
For a total cable length of 37 meters!  Mind you, this is pushing all
devices to their maximum specs, and there could be timing problems.
Especially if you had devices attached at all the points indicated. 
AND if especially fast devices are nvolved.
The converter box DOES introduce some delay when switching bus directions. 
Devices attached at the N1 and N2 spots on the buses shown in the last
example, MUST have no internal termination installed.. Usually a small resistor DIP or SIP pack, sometimes termination be enabled/disabled by jumpers.

Later,
markw
-- 
M. Whetzel 
Home: markw@antimatr.hou.tx.us        | IBM RT/125 running AIX 2.2.1 
UUCP ..!uhnix1!lobster!antimatr!markw 

In article <1992Dec27.195328.125@vms.huji.ac.il>, allon@nesher.fh.huji.ac.il (Allon herman) writes:
> I am trying to install a non IBM scsi disk on my RT (running AOS 4.3).
> I am also aware that there has been recently a discussion regarding this topic
> in this newsgroup. I will be most thankful to the person who can tell me if
> and where can I find old articles that were posted on this newsgroup.
> 
> 			Thanks,
> 				Allon Herman.
> 
> E-Mail: allon@batata.fh.huji.ac.il


The fact that it is a not an IBM SCSI disk shoulod have no effect.  SCSI is a
standard.  If it is a SCSI drive and it does not work with a SCSI controller, one
of the devices does not meet the SCSI standard.

Mark


-- 
Mark 
______________________________________________________________________________
mark@uk.ac.ox.physchem

mark@physchem.ox.ac.uk (Mark) writes:


>In article <1992Dec27.195328.125@vms.huji.ac.il>, allon@nesher.fh.huji.ac.il (Allon herman) writes:
>> I am trying to install a non IBM scsi disk on my RT (running AOS 4.3).
>> I am also aware that there has been recently a discussion regarding this topic
>> in this newsgroup. I will be most thankful to the person who can tell me if
>> and where can I find old articles that were posted on this newsgroup.
>> 
>> 			Thanks,
>> 				Allon Herman.
>> 
>> E-Mail: allon@batata.fh.huji.ac.il
>
>
>The fact that it is a not an IBM SCSI disk shoulod have no effect.  SCSI is a
>standard.  If it is a SCSI drive and it does not work with a SCSI controller, one
>of the devices does not meet the SCSI standard.

Let's not forget the difference between Differential SCSI, as implemented on
the RT and several minicomputers, and Single-Ended SCSI, on the PC, Mac, RS/6000
and other workstations.  They are not electrically compatible; the only difference
which I understand between them is that Differential SCSI allows longer distances
(the RT SCSI disks were in racks external to the cabinet).

I've heard that there are 3rd-party converters (black boxes?) which allow these
to be intermixed, but have no experience with them.
--

Bob Shair                          shair@vnet.ibm.com
Scientific Computing Specialist    SHAIR@UIUCVMD (bitnet)
IBM Champaign

In article <C0E5wM.943@news.cso.uiuc.edu>, shair@barolo.cso.uiuc.edu (Bob Shair) writes:
> mark@physchem.ox.ac.uk (Mark) writes:
> >In article <1992Dec27.195328.125@vms.huji.ac.il>, allon@nesher.fh.huji.ac.il (Allon herman) writes:
> >> I am trying to install a non IBM scsi disk on my RT (running AOS 4.3).
> >> I am also aware that there has been recently a discussion regarding this
> >> topic in this newsgroup. I will be most thankful to the person who can
> >> tell me if and where can I find old articles that were posted on this
> >> newsgroup.
> >
> >The fact that it is a not an IBM SCSI disk shoulod have no effect.  SCSI is a
> >standard.  If it is a SCSI drive and it does not work with a SCSI
> >controller, one of the devices does not meet the SCSI standard.

Not true.. As Bob Shair said below, ANSI standards do allow for two electricly 
incompatible interface styles.  Also, different adapters have different
maximum data transfer rates, and not all adapters supported async and 
syncronous transfer modes.  Most modern controllers do now, but that
was not always the case.  The RT adapter supported both transfer modes
and up to 3.5M Bytes per second.  I believe that this was improved on
later versions of the IBM SCSI adapter.
 
> Let's not forget the difference between Differential SCSI, as implemented on
> the RT and several minicomputers, and Single-Ended SCSI, on the PC, Mac,
> RS/6000 and other workstations.  They are not electrically compatible;
> the only difference which I understand between them is that Differential
> SCSI allows longer distances (the RT SCSI disks were in racks external
> to the cabinet).
 
Single-ended SCSI uses only one signal wire for each control signal.
Differential-SCSI uses two wires in a push-pull driving pair for each
control signal.  This allows better noise immunity, and the longer driving
distance from the computer.. up to 25 Meters.

> I've heard that there are 3rd-party converters (black boxes?) which allow
> these to be intermixed, but have no experience with them.

True.  The boxes and internally mounted converter boards from
Rancho Technology work quite well for this task.  There are also at least
two other vendors making single<->differential converters that I know about.

I have used non-IBM disks on an RT running AIX 2.2.1.
I have run HP 9" disks, Maxtor 5.25" disks, and Fujitsu 5.25" SCSI disks on
an RT running AIX, as well as other SCSI disks from several manufacturers.

I also have running today at my office, HP 88780A 9-track tape drives,
Exabyte 8MM tape drives, and Fujitsu 3480 compatible tape drives on several
of our RT's running AIX 2.2.1. We had to write our own driver for the tape
drive support.

AIX and VRM can be tricked into running OEM disks, without too much trouble.
While it is awkard to get the OEM SCSI disks initially running, once you do,
they work fine, just stay away from drives bigger than 1 Gigabyte.  
VRM just gets completely lost and wraps accesses over 1 Gig back to
cylinder 0, destroying the minidisk partition table.  I believe this to
be a bug in the SCSI driver software, not a hardware problem.

People need to understand IBM's methods when the IBM RT was developed.
You buy a IBM computer, you get IBM devices to attach to it.  This was true
of several major makers of computer equipment at that time.

IBM only supported for attachment by the SCSI interface (at least for AIX,
not sure about AOS), the IBM 9332 external disk drive on the RT. 
The disks came in either 200 or 400 MByte capacity drives. We had an RT 125
on evaluation once, and it had one of the external table top 9332 drives on it.
They were monster drives with 11 inch in diameter platters, and were
available in standalone table top cabinets, or rack mounted. 
The internal drive electronics for those drives had special extended command
sets that supported IBM diagnostics, and also the ability for the RT to
download new microcode to the drive.  The drives came preformatted with
VRM partition table information already written to the drives.

The drives can be only used for DATA filesystems, /, /usr, /tmp and /u
filesystems were not allowed to be placed on the SCSI disks.
Initial IPL support did not understand the SCSI controllers.  The
"varyon" command brought in the proper support for the SCSI disks. into
the running AIX kernel.

Later,
markw
-- 
M. Whetzel 
Home: markw@antimatr.hou.tx.us        | IBM RT/125 running AIX 2.2.1 
UUCP ..!uhnix1!lobster!antimatr!markw