[hercules-390] Translation exception 0005 in MVT boot.

[[e-mail] [hercules-390]]

My goal is to get OS/360 MVT running on an emulated 360/91.


Host system is x86_64 running Fedora 23.


I cloned hyperion: commit a87eaab20 Fri Nov 18 15:51:28 2016


Installed all needed and optional packages except libbz2.


Built and installed.


Following the instructions at Introduction to Generating and Running OS/360 on Hercules http://www.conmicro.com/hercos360/, performed through step 8, "Final MVT setup".


(The gen.cnf and mvt.cnf files were modified: TIMEZONE set to -0800, CPUMODEL set to 2091, SYSEPOCH set to 1960 -28)


On step 9 "IPLing MVT":


Start hyperion:
hercules -f mvt.cnf,
It appears to start normally


Start 3270 terminal
x3270 localhost:3270
The Hercules logo appears.


Do "ipl 350"


HHC01603I ipl 350
HHC00801I Processor CP00: Addressing exception code 0005 ilc 6
HHC02324I PSW=00000005C0000118 INST=D2FF4000FC7E MVC 0(256,4),3198(15) mo
HHC02326I R:00800000: Translation exception 0005
HHC02326I R:00000D00:K:06=00000000 00000000 00000000 00000000 ................
HHC02269I GR00=00000000 GR01=00000000 GR02=00000000 GR03=00000000
HHC02269I GR04=00800000 GR05=00000000 GR06=00000000 GR07=00000000
HHC02269I GR08=00000100 GR09=00FFFFFF GR10=0000011E GR11=00000000
HHC02269I GR12=00000000 GR13=00000000 GR14=00000000 GR15=40000082
HHC00801I Processor CP00: Addressing exception code 0005 ilc 2
HHC02324I PSW=0000000550000132 INST=0832 SSK 3,2 se
HHC02269I GR00=00000000 GR01=00000000 GR02=00800000 GR03=00000000
HHC02269I GR04=00000800 GR05=00000000 GR06=00000000 GR07=00000000
HHC02269I GR08=00000100 GR09=00FFFFFF GR10=0000013E GR11=00000000
HHC02269I GR12=00000000 GR13=00000000 GR14=00000000 GR15=40000082
HHC00107I Starting thread cckd_ra(), active=0, started=0, max=2
HHC00100I Thread id 00007f56ac74c700, prio 2147483647, name Read-ahead thread-1
HHC00107I Starting thread cckd_ra() from cckd_ra(), active=1, started=1, max=2
HHC00100I Thread id 00007f56ac64b700, prio 2147483647, name Read-ahead thread-2
HHC01315I 0:0350 CHAN: ccw 07000C08 40000006=>00000194 0000
HHC01312I 0:0350 CHAN: stat 0E00, count 0000
HHC01313I 0:0350 CHAN: sense 80000000 38931204 00000000 00000000 00000000 000000
HHC01314I 0:0350 CHAN: sense CMDREJ
HHC00809I Processor CP00: disabled wait state 00020000 80000005
herc =====>


According to the instructions, I should see:


IEA101A SPECIFY SYSTEM PARAMETERS FOR RELEASE 21.8F MVT


but the CPU has halted.


-- Charles

[[e-mail] [hercules-390]]

Post by ***@gmail.com [hercules-390]
CPUMODEL set to 2091

I don't know the solution to your problem,
but I would suggest you try putting the
CPUMODEL back to whatever it was
previously and see what effect that has.

BFN. Paul.

[[e-mail] [hercules-390]]

Changed back to original '0158'; no effect.

-- Charles

[[e-mail] [hercules-390]]

Post by ***@gmail.com [hercules-390]
Changed back to original '0158'; no effect.

Ok, another suggestion:

Try typing in "t+" to activate trace before
doing the IPL to see if it executes any
instructions successfully before failing.

And another suggestion: try using an
older version of Hercules such as 3.07
or 3.12 and see if you get the same result.

BFN. Paul.

[[e-mail] [hercules-390]]

588,291 instructions before dying.

I tried 3.02, 3.07 and 3.12 before trying hyperion; they all died with various instruction exceptions after the initial IPL much earlier in the OS/360 install process.


-- Charles

[Joe Monk [e-mail] [hercules-390]]

Your problem is here:

HHC01315I 0:0350 CHAN: ccw 07000C08 40000006=>00000194 0000


HHC01312I 0:0350 CHAN: stat 0E00, count 0000


HHC01313I 0:0350 CHAN: sense 80000000 38931204 00000000 00000000 00000000
000000

HHC01314I 0:0350 CHAN: sense CMDREJ


HHC00809I Processor CP00: disabled wait state 00020000 80000005

You are getting a CCW because you have an IO problem (the unit is rejecting
the channel command).

My advice: dump everything and rebuild from scratch. Something was not
right in your build.

Joe

Post by ***@gmail.com [hercules-390]
588,291 instructions before dying.
I tried 3.02, 3.07 and 3.12 before trying hyperion; they all died with
various instruction exceptions after the initial IPL much earlier in the
OS/360 install process.
-- Charles

[[e-mail] [hercules-390]]

The address exception/translation exception happens before the the channel error; I would think that the exception is the issue and the channel error a result of the earlier error.

-- Charles

[Mike Stramba [e-mail] [hercules-390]]

I get the same two addressing exceptions when running Kevin's
"turnkey" system, but the system IPL's and runs anyway.

I suspect the exceptions are "normal".

There are no mentions of them in Kevin's readme files.

Mike

Post by ***@gmail.com [hercules-390]
The address exception/translation exception happens before the the channel
error; I would think that the exception is the issue and the channel error a
result of the earlier error.
-- Charles

[Gerhard Postpischil [e-mail] [hercules-390]]

Post by ***@gmail.com [hercules-390]
The address exception/translation exception happens before the the
channel error; I would think that the exception is the issue and the
channel error a result of the earlier error.

Nucleus initialization tries all sorts of things that get errors.
Typical is an addressing exception while it tries to determine the
available memory size (and DOS goes into a loop with 16 MB). An
unexpected exceptoin would give you a wait state with an error code, not
continued running.

Gerhard Postpischil
Bradford, VT

[[e-mail] [hercules-390]]

Post by ***@gmail.com [hercules-390]
HHC01315I 0:0350 CHAN: ccw 07000C08 40000006=>00000194 0000
HHC01312I 0:0350 CHAN: stat 0E00, count 0000
HHC01313I 0:0350 CHAN: sense 80000000 38931204 00000000 00000000 00000000 000000
HHC01314I 0:0350 CHAN: sense CMDREJ
HHC00809I Processor CP00: disabled wait state 00020000 80000005
You are getting a CCW because you have an IO problem (the unit is rejecting the channel >command).
My advice: dump everything and rebuild from scratch. Something was not right in your build.
Joe

- - - old notes snipped - - -

If rebuilt the same way a few dozen times, expect the same results.

The error is from a seek to disk address cylinder X'194' Track zero
X'194' = decimal 404

S/360 disks like 2311 and 2314 were not anywhere near that size.
3330 model 1 had cylinders 0 through 403 for normal data and
alternate tracks starting at cylinder 404 but I don't think that
any S/360 operating system inspects alternate tracks on IPL?.
3330 double density had about twice as many cylinders.

I suspect that you have a device type mismatch?
What type of device is genned in MVT for address 350 ?
How is device 350 defined in the Hercules configuration file?
What device type is the 350 disk image?

Did you use dasdinit and specify the -a switch to get
alternate tracks?

[[e-mail] [hercules-390]]

Rebuilt hyperion '-O0'; now runs 8,078,793 instructions before:

Addressing exception code 0005 ilc 6



-- Charles

[Mike Stramba [e-mail] [hercules-390]]

For a "sanity check" on your Hercules / 360 build / Host OS, you might
try running Kevin
Leonard's "turnkey MVT" :

http://www.ibiblio.org/jmaynard/asp.zip

It runs fine for me on Hercules 3.12 on Win7, I haven't tried Hyperion

Mike

[[e-mail] [hercules-390]]

Running from asp.zip:

hercules -f conf/asp.conf


x3270 ***@localhost:3270


x3270 ***@localhost:3270


ipl 150


HHC01603I ipl 150
HHC00801I Processor CP00: Addressing exception code 0005 ilc 6
HHC02324I PSW=00000005C0000118 INST=D2FF4000FC7E MVC 0(256,4),3198(15) mo
HHC02326I R:00800000: Translation exception 0005
HHC02326I R:00000D00:K:06=00000000 00000000 00000000 00000000 ................
HHC02269I GR00=00000000 GR01=00000000 GR02=00000000 GR03=00000000
HHC02269I GR04=00800000 GR05=00000000 GR06=00000000 GR07=00000000
HHC02269I GR08=00000100 GR09=00FFFFFF GR10=0000011E GR11=00000000
HHC02269I GR12=00000000 GR13=00000000 GR14=00000000 GR15=40000082
HHC00801I Processor CP00: Addressing exception code 0005 ilc 2
HHC02324I PSW=0000000550000132 INST=0832 SSK 3,2 se
HHC02269I GR00=00000000 GR01=00000000 GR02=00800000 GR03=00000000
HHC02269I GR04=00000800 GR05=00000000 GR06=00000000 GR07=00000000
HHC02269I GR08=00000100 GR09=00FFFFFF GR10=0000013E GR11=00000000
HHC02269I GR12=00000000 GR13=00000000 GR14=00000000 GR15=40000082
HHC00107I Starting thread cckd_ra(), active=0, started=0, max=2
HHC00100I Thread id 00007f5109c27700, prio 2147483647, name Read-ahead thread-1
HHC00107I Starting thread cckd_ra() from cckd_ra(), active=1, started=1, max=2
HHC00100I Thread id 00007f5109b26700, prio 2147483647, name Read-ahead thread-2
herc =====>


-- Charles

[Joe Monk [e-mail] [hercules-390]]

Yep. Here's why it's not working:

The first five macros configure the OS itself. They describe an OS/360 MVT
system, *built to run on a 370/158 CPU*,

http://www.conmicro.com/hercos360/stage1.html

So, the sysgen is not right for a 360/91. The system is for a 370 machine.

Joe

Post by ***@gmail.com [hercules-390]
hercules -f conf/asp.conf
ipl 150
HHC01603I ipl 150
HHC00801I Processor CP00: Addressing exception code 0005 ilc 6
HHC02324I PSW=00000005C0000118 INST=D2FF4000FC7E MVC 0(256,4),3198(15)
mo
HHC02326I R:00800000: Translation exception 0005
HHC02326I R:00000D00:K:06=00000000 00000000 00000000 00000000
................
HHC02269I GR00=00000000 GR01=00000000 GR02=00000000 GR03=00000000
HHC02269I GR04=00800000 GR05=00000000 GR06=00000000 GR07=00000000
HHC02269I GR08=00000100 GR09=00FFFFFF GR10=0000011E GR11=00000000
HHC02269I GR12=00000000 GR13=00000000 GR14=00000000 GR15=40000082
HHC00801I Processor CP00: Addressing exception code 0005 ilc 2
HHC02324I PSW=0000000550000132 INST=0832 SSK 3,2
se
HHC02269I GR00=00000000 GR01=00000000 GR02=00800000 GR03=00000000
HHC02269I GR04=00000800 GR05=00000000 GR06=00000000 GR07=00000000
HHC02269I GR08=00000100 GR09=00FFFFFF GR10=0000013E GR11=00000000
HHC02269I GR12=00000000 GR13=00000000 GR14=00000000 GR15=40000082
HHC00107I Starting thread cckd_ra(), active=0, started=0, max=2
HHC00100I Thread id 00007f5109c27700, prio 2147483647, name Read-ahead thread-1
HHC00107I Starting thread cckd_ra() from cckd_ra(), active=1, started=1, max=2
HHC00100I Thread id 00007f5109b26700, prio 2147483647, name Read-ahead thread-2
herc =====>
-- Charles

[Tony Harminc [e-mail] [hercules-390]]

Post by ***@gmail.com [hercules-390]
HHC01603I ipl 150
HHC00801I Processor CP00: Addressing exception code 0005 ilc 6
HHC02324I PSW=00000005C0000118 INST=D2FF4000FC7E MVC 0(256,4),3198(15) mo
HHC02326I R:00800000: Translation exception 0005
HHC02326I R:00000D00:K:06=00000000 00000000 00000000 00000000 ................
HHC02269I GR00=00000000 GR01=00000000 GR02=00000000 GR03=00000000
HHC02269I GR04=00800000 GR05=00000000 GR06=00000000 GR07=00000000
HHC02269I GR08=00000100 GR09=00FFFFFF GR10=0000011E GR11=00000000
HHC02269I GR12=00000000 GR13=00000000 GR14=00000000 GR15=40000082

So this is an MVC copying data from location X'D00' to location
X'800000'. How much main storage do you have configured in Hercules
for this MVT system? Is it 8 MB by any chance? Why do you think this
is an error? Program checks are not necessarily wrong.

My only question is why Hercules calls this a "Translation exception".
It's a program check 5, which is an addressing exception, as it
correctly says elsewhere.

Post by ***@gmail.com [hercules-390]
HHC00801I Processor CP00: Addressing exception code 0005 ilc 2
HHC02324I PSW=0000000550000132 INST=0832 SSK 3,2 se
HHC02269I GR00=00000000 GR01=00000000 GR02=00800000 GR03=00000000
HHC02269I GR04=00000800 GR05=00000000 GR06=00000000 GR07=00000000
HHC02269I GR08=00000100 GR09=00FFFFFF GR10=0000013E GR11=00000000
HHC02269I GR12=00000000 GR13=00000000 GR14=00000000 GR15=40000082

This is trying to set the storage key of the same block at address
X'800000' to 0. It fails almost certainly for the same reason - you
don't have that much storage defined. I would expect MVT to continue
on after having discovered the amount of available storage, and indeed
it seems to.

So perhaps it *is* the I/O related problem that you next bump into
that is the real issue.

Tony H.

[[e-mail] [hercules-390]]

Okay, I managed to get the asp.zip image booted; it appears that the exception messages are expected, probably part of the memory sizing.

So I probably fumbled something in the OS/360 install process; I'll go over it (yet again), but for my immediate needs, the asp.zip image will suffice.


Thank you all,
-- Charles

[[e-mail] [hercules-390]]

Post by ***@gmail.com [hercules-390]
My goal is to get OS/360 MVT running on an emulated 360/91.
for my immediate needs, the asp.zip image will suffice.

Would you mind sharing what you are
doing? You seem to have some sort
of genuine production need for this
to work.

I can't imagine what you would want
to do on OS/370 that you can't do on
TK4- or MVS/380.

BFN. Paul.

[[e-mail] [hercules-390]]

Ahh.. NDA.

It's cool, and I will follow-up in a couple of months when the covers come off.


-- Charles

[[e-mail] [hercules-390]]

Post by ***@gmail.com [hercules-390]
Ahh.. NDA.
It's cool, and I will follow-up in a couple of
months when the covers come off.

Ok, I look forward to that. I can't even
imagine what feature MVT has that
TK4- and MVS/380 doesn't blow away.

Note that recently beta MVS/380 (with
beta Hercules/380) even supports
z/Arch 64-bit registers and instructions
like STGM. I'm not sure whether it supports
data ATB or not, because I don't know
how to compile a 64-bit version of
Windows Hercules/380 3.07 yet to find out.

BFN. Paul.

[[e-mail] [hercules-390]]

instructions like STGM.

STMG.

BFN. Paul.

[Joe Monk [e-mail] [hercules-390]]

"I can't even
imagine what feature MVT has that
TK4- and MVS/380 doesn't blow away."

Umm ... it launched apollo rockets to the moon on 360/75J processors?

Joe

Post by ***@yahoo.com.au [hercules-390]

Post by ***@gmail.com [hercules-390]
Ahh.. NDA.
It's cool, and I will follow-up in a couple of
months when the covers come off.

Ok, I look forward to that. I can't even
imagine what feature MVT has that
TK4- and MVS/380 doesn't blow away.
Note that recently beta MVS/380 (with
beta Hercules/380) even supports
z/Arch 64-bit registers and instructions
like STGM. I'm not sure whether it supports
data ATB or not, because I don't know
how to compile a 64-bit version of
Windows Hercules/380 3.07 yet to find out.
BFN. Paul.

[Gregg Levine [e-mail] [hercules-390]]

Hello!
Exactly. Outside of the hardware worn by the LEM and the CSM, the big
bird used the same logic blobs as the ground based systems. And it
makes sense because IBM built both, (The others were all built by hand
by a different outfit.)

And Paul? I arranged the problems that Yahoo was having. Dave? Please
stop giving money to that Yeti.....
-----
Gregg C Levine ***@gmail.com
"This signature fought the Time Wars, time and again."

Post by Joe Monk ***@gmail.com [hercules-390]
"I can't even
imagine what feature MVT has that
TK4- and MVS/380 doesn't blow away."
Umm ... it launched apollo rockets to the moon on 360/75J processors?
Joe

Post by ***@yahoo.com.au [hercules-390]

Post by ***@gmail.com [hercules-390]
Ahh.. NDA.
It's cool, and I will follow-up in a couple of
months when the covers come off.

Ok, I look forward to that. I can't even
imagine what feature MVT has that
TK4- and MVS/380 doesn't blow away.
Note that recently beta MVS/380 (with
beta Hercules/380) even supports
z/Arch 64-bit registers and instructions
like STGM. I'm not sure whether it supports
data ATB or not, because I don't know
how to compile a 64-bit version of
Windows Hercules/380 3.07 yet to find out.
BFN. Paul.

['Dave Wade' [e-mail] [hercules-390]]

-----Original Message-----
Sent: 24 November 2016 04:55
Subject: Re: [hercules-390] Re: Translation exception 0005 in MVT boot.
Hello!
Exactly. Outside of the hardware worn by the LEM and the CSM, the big bird
used the same logic blobs as the ground based systems. And it makes sense
because IBM built both, (The others were all built by hand by a different
outfit.)
And Paul? I arranged the problems that Yahoo was having. Dave? Please stop

How did you know?

giving money to that Yeti.....
-----
"This signature fought the Time Wars, time and again."

Post by Joe Monk ***@gmail.com [hercules-390]
"I can't even
imagine what feature MVT has that
TK4- and MVS/380 doesn't blow away."
Umm ... it launched apollo rockets to the moon on 360/75J processors?
Joe

Post by ***@gmail.com [hercules-390]
Ahh.. NDA.
It's cool, and I will follow-up in a couple of months when the covers come off.

Ok, I look forward to that. I can't even imagine what feature MVT has
that
TK4- and MVS/380 doesn't blow away.
Note that recently beta MVS/380 (with beta Hercules/380) even
supports z/Arch 64-bit registers and instructions like STGM. I'm not
sure whether it supports data ATB or not, because I don't know how to
compile a 64-bit version of Windows Hercules/380 3.07 yet to find
out.
BFN. Paul.

------------------------------------
------------------------------------
http://groups.yahoo.com/group/hercules-390
http://www.hercules-390.org
------------------------------------
Yahoo Groups Links

[[e-mail] [hercules-390]]

Post by ***@yahoo.com.au [hercules-390]
I'm not sure whether it supports
data ATB or not, because I don't know
how to compile a 64-bit version of
Windows Hercules/380 3.07 yet to find out.

I figured out how to do a 64-bit build,
but ATB writing is not working for
some reason so I need to debug that.

Also the 64-bit build crashes something
like 20% of the time so also needs
to be debugged.

BFN. Paul.

[[e-mail] [hercules-390]]

Post by ***@yahoo.com.au [hercules-390]
but ATB writing is not working for
some reason so I need to debug that.

ATB (writing above the 2 GiB bar) is
now working. :-) Windows binaries
will hopefully be in the hercules-os380
group tomorrow. There is some
technical problem with Yahoo at the
moment.

BFN. Paul.

[[e-mail] [hercules-390]]

Post by ***@yahoo.com.au [hercules-390]

Post by ***@gmail.com [hercules-390]
Ahh.. NDA.
It's cool, and I will follow-up in a couple of
months when the covers come off.

Ok, I look forward to that. I can't even
imagine what feature MVT has that
TK4- and MVS/380 doesn't blow away.
Note that recently beta MVS/380 (with
beta Hercules/380) even supports
z/Arch 64-bit registers and instructions
like STGM. I'm not sure whether it supports
data ATB or not, because I don't know
how to compile a 64-bit version of
Windows Hercules/380 3.07 yet to find out.
BFN. Paul.

What does you not knowing how to build
Hercules that cannot run on a 32-bit system
have to do with S/360 MVT /O area.

Since you run a 32-bit PC operating system,
I suggest that you use a Hercules that will
run on that.
If you ever switch to a 64-bit PC operating
system, then at that point considering rebuilding
Hercules to not be able to run on a 32-bit PC
operating system might be more reasonable.

[[e-mail] [hercules-390]]

Post by ***@yahoo.com [hercules-390]
Since you run a 32-bit PC operating system,
I suggest that you use a Hercules that will
run on that.

I run 64-bit Windows. I think that is
more standard for new PCs than
32-bit Windows.

BFN. Paul.

['Dave Wade' [e-mail] [hercules-390]]

-----Original Message-----
Sent: 23 November 2016 23:47
Subject: [hercules-390] Re: Translation exception 0005 in MVT boot.

Post by ***@gmail.com [hercules-390]
My goal is to get OS/360 MVT running on an emulated 360/91.

You can't use Hercules. Hercules only emulates S/370 onwards, and it does not emulate model specific features.

Dave
G4UGM

Post by ***@gmail.com [hercules-390]
for my immediate needs, the asp.zip image will suffice.

Would you mind sharing what you are
doing? You seem to have some sort
of genuine production need for this
to work.
I can't imagine what you would want
to do on OS/370 that you can't do on
TK4- or MVS/380.
BFN. Paul.
------------------------------------
------------------------------------
http://groups.yahoo.com/group/hercules-390
http://www.hercules-390.org
------------------------------------
Yahoo Groups Links

[Harold Grovesteen [e-mail] [hercules-390]]

Not True. Hercules S/370 mode supports basic-control mode used by
360-era operating systems.

Harold Grovesteen

Post by 'Dave Wade' ***@gmail.com [hercules-390]

-----Original Message-----
Sent: 23 November 2016 23:47
Subject: [hercules-390] Re: Translation exception 0005 in MVT boot.

Post by ***@gmail.com [hercules-390]
My goal is to get OS/360 MVT running on an emulated 360/91.

You can't use Hercules. Hercules only emulates S/370 onwards, and it does not emulate model specific features.
Dave
G4UGM

['Dave Wade' [e-mail] [hercules-390]]

-----Original Message-----
Sent: 25 November 2016 15:12
Subject: Re: [hercules-390] Re: Translation exception 0005 in MVT boot.
Not True. Hercules S/370 mode supports basic-control mode used by 360-era
operating systems.

I never said it wouldn't run on Hercules, but, he original target was VERY SPECIFIC it said

"..... running on an emulated 360/91"

Now the Hercules FAQ is very specific. Hercules does not emulate specific models and features, in archmode 370 it’s a generic S/370 not a S/360 and any specific S/360 only facilities will not be available.
(No ASCII mode, invalid decimal operations suppressed, not terminated)

Setting the model in the config to a S/360 model 91 does not alter the actual instructions available or the way the emulation works.

I only mentioned this because the M91 appears to be a very special machine, with a highly interleaved, pipelined, microcode architecture..
.. so yes sure it will run on Hercules, BUT Hercules does no implement any M91 specifics...

Harold Grovesteen

Dave

Post by 'Dave Wade' ***@gmail.com [hercules-390]

-----Original Message-----
Sent: 23 November 2016 23:47
Subject: [hercules-390] Re: Translation exception 0005 in MVT boot.

Post by ***@gmail.com [hercules-390]
My goal is to get OS/360 MVT running on an emulated 360/91.

You can't use Hercules. Hercules only emulates S/370 onwards, and it does

not emulate model specific features.

Post by 'Dave Wade' ***@gmail.com [hercules-390]
Dave
G4UGM

------------------------------------
------------------------------------
http://groups.yahoo.com/group/hercules-390
http://www.hercules-390.org
------------------------------------
Yahoo Groups Links

[Tony Harminc [e-mail] [hercules-390]]

Post by 'Dave Wade' ***@gmail.com [hercules-390]
Now the Hercules FAQ is very specific. Hercules does not emulate specific models and features, in archmode 370 it’s a generic S/370 not a S/360 and any specific S/360 only facilities will not be available.
(No ASCII mode, invalid decimal operations suppressed, not terminated)

No specification exception for unaligned operands of unprivileged
instructions...

Tony H.

[[e-mail] [hercules-390]]

Given the context that I am a CDC & Honeywell mainframe guy and IBM is new foray for me, where is a good place to start with understanding the architectural details of 360/370 architecture differences and model 91 specific differences?

-- Charles

['Dave Wade' [e-mail] [hercules-390]]

Charles,



Sadly, its getting difficult to find the original System/360 information. Until three or four years ago the relevant issues of the IBM System Journal were free available for download and so you could read the original source, but now they are behind a paywall.

I guess the original IBM manuals on Bitsavers would be a good start, beginning with the “Introduction To System 360” (not you can edit the URLS to go directly to BitSavers)




http://www.mirrorservice.org/sites/www.bitsavers.org/pdf/ibm/360/GC20-1667-1_intro360arch.pdf



The differences between each model is described in the “Functional Characteristics” manual for the specific model..



http://www.mirrorservice.org/sites/www.bitsavers.org/pdf/ibm/360/funcChar/



the differences between System/360 and System/370 are outlined in the initial System/370 Principles of Operation which is the first file in the list..



http://www.mirrorservice.org/sites/www.bitsavers.org/pdf/ibm/370/princOps/



there is also some information on the IBM web site..



http://www-03.ibm.com/ibm/history/ibm100/us/en/icons/system360/



https://www-03.ibm.com/ibm/history/exhibits/mainframe/mainframe_PR360.html



and of course there are Wikipedia articles, but they largely regurgitate a summary of the information in the above sources




https://en.wikipedia.org/wiki/IBM_System/360



so actually, perhaps a good starting place




Dave







From: hercules-***@yahoogroups.com [mailto:hercules-***@yahoogroups.com]
Sent: 27 November 2016 15:46
To: hercules-***@yahoogroups.com
Subject: Re: [hercules-390] Re: Translation exception 0005 in MVT boot.





Given the context that I am a CDC & Honeywell mainframe guy and IBM is new foray for me, where is a good place to start with understanding the architectural details of 360/370 architecture differences and model 91 specific differences?



-- Charles

[Tony Harminc [e-mail] [hercules-390]]

Post by ***@gmail.com [hercules-390]
Given the context that I am a CDC & Honeywell mainframe guy and IBM is new
foray for me, where is a good place to start with understanding the
architectural details of 360/370 architecture differences and model 91
specific differences?

One book to look at would be the very first edition of the S/370 Principles
of Operation GA22-7000-00. This is a very thin manual that is just an
update to the S/360 Principles of Operation GA22-6821. All subsequent
editions of the 370 POO were standalone books.

That first 370 POO points out in passing a number of restrictions in S/360
that were removed in S/370, so it may help you to work backwards from the
much more well known and easy to find info on (and assumptions about) S/370.

In general, the Principles of Operation from S/360 right up to the current
System z is a masterpiece of good documentation. One can pick the very rare
nit, but the succsessive authors over many years have kept up a very high
standard. In the 1970s IBM received an award from someone (ACM? IEEE? US
DOD?) for the S/370 POO, saying it was the only really completely
documented architecture on the market. The book is well known to be
produced from a common GML source base in two versions - one for the
public, and one for IBM. Evidently the really interesting stuff, including
design and implementation doc is kept only in the full version.

As for the model 91 - this was a rare beast that, while architecturaly
compatible, did have some differences in interrupt handling (so-called
imprecise interrupts) and out-of-order execution that (iirc) was visible to
channel programs, but not to CPU programs.

I guess the bottom line is that a S/370 machine or its emulation will
produce results compatible with those on a 360/91; rather, it's the /91
that can produce results incompatible with the more mainstream 360 and 370
models. Again, iirc, these differences are in the high performance floating
point area alone.

So it depends on which way you are going - if you want, say, to test
software behaviour under Hercules in preparation for running it on a real
/91, you won't be able to cover all the cases. If you want to, say, install
Hercules running on a Raspberry-Pi inside a mockup of a /91 (maybe someone
saved just an old front panel or the like), then your results will probably
be acceptable.

Of course we'd all love to hear what you're working on. There has been very
little talk of S/360 wrt Hercules over the years, except for some about the
quite different 360/67, which had virtual storage and 32-bit addressing.

Tony H..

[Charles Anthony [e-mail] [hercules-390]]

Post by Tony Harminc ***@gmail.com [hercules-390]

Post by Tony Harminc ***@gmail.com [hercules-390]
As for the model 91 - this was a rare beast that, while architecturaly

compatible, did have some differences in interrupt handling (so-called
imprecise interrupts) and out-of-order execution that (iirc) was visible to
channel programs, but not to CPU programs.

A non-CPU difference appears to be that the /91 predated 3270 (IIUC). The
/91 operator console was an I/O Selectric style. The ready-to-roll OS/360
packages appear to be 3270 oriented. I'm going to have to figure out the
intricacies of system configuration to match /91 era peripherals; possibly
rolling back to an earlier version of OS/360 if needed (and possible).

(Looking at Wikipedia, /91 was first released in 1967; there appear to have
been a slew of OS/360 releases that year; I would hazard a guess that one
of the releases was a /91 release. I might be asking for trouble with a OS
configured for the /91 if the interrupt structure is different and
unsupported by Hercules; but if possible a period correct release would be
preferable.)

I am more interested in /91 like behavior; I don't require emulation of the
OOO and related differences -- it is more about the user experience; it
looks and feels like a /91. So configured as a /91 (appropriate
peripherals, amount of memory), and if the OS or programs announce the
model number it should be 360/91 or 2091 as appropriate.

-- Charles

[Joe Monk [e-mail] [hercules-390]]

"The /91 operator console was an I/O Selectric style."

Nope. The Operator Console on a 360/91 was a 2250 CRT model 1. It's
detailed on page 6 of the Model 91 Functional Characteristics:

http://bitsavers.informatik.uni-stuttgart.de/pdf/ibm/360/funcChar/A22-6907-2_360-91_funcChar.pdf

"I'm going to have to figure out the intricacies of system configuration to
match /91 era peripherals; possibly rolling back to an earlier version of
OS/360 if needed (and possible)."

OS/360 raw distribution tapes and files you will find here:

http://www.ibiblio.org/jmaynard/

and instructions for installation to run on Hercules here:

http://www.conmicro.com/hercos360/index.html

Joe

Post by Charles Anthony ***@gmail.com [hercules-390]

Post by Tony Harminc ***@gmail.com [hercules-390]

Post by Tony Harminc ***@gmail.com [hercules-390]
As for the model 91 - this was a rare beast that, while architecturaly

compatible, did have some differences in interrupt handling (so-called
imprecise interrupts) and out-of-order execution that (iirc) was visible to
channel programs, but not to CPU programs.
A non-CPU difference appears to be that the /91 predated 3270 (IIUC). The

/91 operator console was an I/O Selectric style. The ready-to-roll OS/360
packages appear to be 3270 oriented. I'm going to have to figure out the
intricacies of system configuration to match /91 era peripherals; possibly
rolling back to an earlier version of OS/360 if needed (and possible).
(Looking at Wikipedia, /91 was first released in 1967; there appear to
have been a slew of OS/360 releases that year; I would hazard a guess that
one of the releases was a /91 release. I might be asking for trouble with a
OS configured for the /91 if the interrupt structure is different and
unsupported by Hercules; but if possible a period correct release would be
preferable.)
I am more interested in /91 like behavior; I don't require emulation of
the OOO and related differences -- it is more about the user experience; it
looks and feels like a /91. So configured as a /91 (appropriate
peripherals, amount of memory), and if the OS or programs announce the
model number it should be 360/91 or 2091 as appropriate.
-- Charles

[Joe Monk [e-mail] [hercules-390]]

BTW, I am a lover of all things related to NASA, including the model 91!

Joe

Post by Joe Monk ***@gmail.com [hercules-390]
"The /91 operator console was an I/O Selectric style."
Nope. The Operator Console on a 360/91 was a 2250 CRT model 1. It's
http://bitsavers.informatik.uni-stuttgart.de/pdf/ibm/360/
funcChar/A22-6907-2_360-91_funcChar.pdf
"I'm going to have to figure out the intricacies of system configuration
to match /91 era peripherals; possibly rolling back to an earlier version
of OS/360 if needed (and possible)."
http://www.ibiblio.org/jmaynard/
http://www.conmicro.com/hercos360/index.html
Joe
On Sun, Nov 27, 2016 at 2:20 PM, Charles Anthony

Post by Tony Harminc ***@gmail.com [hercules-390]

Post by Tony Harminc ***@gmail.com [hercules-390]
As for the model 91 - this was a rare beast that, while architecturaly

compatible, did have some differences in interrupt handling (so-called
imprecise interrupts) and out-of-order execution that (iirc) was visible to
channel programs, but not to CPU programs.
A non-CPU difference appears to be that the /91 predated 3270 (IIUC).

The /91 operator console was an I/O Selectric style. The ready-to-roll
OS/360 packages appear to be 3270 oriented. I'm going to have to figure out
the intricacies of system configuration to match /91 era peripherals;
possibly rolling back to an earlier version of OS/360 if needed (and
possible).
(Looking at Wikipedia, /91 was first released in 1967; there appear to
have been a slew of OS/360 releases that year; I would hazard a guess that
one of the releases was a /91 release. I might be asking for trouble with a
OS configured for the /91 if the interrupt structure is different and
unsupported by Hercules; but if possible a period correct release would be
preferable.)
I am more interested in /91 like behavior; I don't require emulation of
the OOO and related differences -- it is more about the user experience; it
looks and feels like a /91. So configured as a /91 (appropriate
peripherals, amount of memory), and if the OS or programs announce the
model number it should be 360/91 or 2091 as appropriate.
-- Charles

[Charles Anthony [e-mail] [hercules-390]]

Post by Joe Monk ***@gmail.com [hercules-390]
"The /91 operator console was an I/O Selectric style."
Nope. The Operator Console on a 360/91 was a 2250 CRT model 1. It's
Hmm. The LCM console clearly has a Selectric attached as part of the /91

console, as does the image in
http://www.subinet.es/un-poco-de-historia-el-ibm-36091/ which supposedly is
the Goddard machine.

-- Charles

['Dave Wade' [e-mail] [hercules-390]]

There are two consoles. There is a 2250 on the rhs of the picture 
 It was common practice to have a 2741 or other hard copy console so a printed log on the jobs run could be archived..







Dave



From: hercules-***@yahoogroups.com [mailto:hercules-***@yahoogroups.com]
Sent: 27 November 2016 23:29
To: hercules-***@yahoogroups.com
Subject: Re: [hercules-390] Re: Translation exception 0005 in MVT boot.












On Sun, Nov 27, 2016 at 1:53 PM, Joe Monk ***@gmail.com <mailto:***@gmail.com> [hercules-390] <hercules-***@yahoogroups.com <mailto:hercules-***@yahoogroups.com> > wrote:



"The /91 operator console was an I/O Selectric style."



Nope. The Operator Console on a 360/91 was a 2250 CRT model 1. It's detailed on page 6 of the Model 91 Functional Characteristics:



Hmm. The LCM console clearly has a Selectric attached as part of the /91 console, as does the image in http://www.subinet.es/un-poco-de-historia-el-ibm-36091/ which supposedly is the Goddard machine.



-- Charles

['Dave Wade' [e-mail] [hercules-390]]

But the picture here supposedly of the Goddard machine



http://www.wikiwand.com/en/IBM_System/360_Model_91



shows a 2250. I suspect you got a 2250 as standard and many sites ordered a 2741 for hard copy




Dave



From: hercules-***@yahoogroups.com [mailto:hercules-***@yahoogroups.com]
Sent: 27 November 2016 23:29
To: hercules-***@yahoogroups.com
Subject: Re: [hercules-390] Re: Translation exception 0005 in MVT boot.












On Sun, Nov 27, 2016 at 1:53 PM, Joe Monk ***@gmail.com <mailto:***@gmail.com> [hercules-390] <hercules-***@yahoogroups.com <mailto:hercules-***@yahoogroups.com> > wrote:



"The /91 operator console was an I/O Selectric style."



Nope. The Operator Console on a 360/91 was a 2250 CRT model 1. It's detailed on page 6 of the Model 91 Functional Characteristics:



Hmm. The LCM console clearly has a Selectric attached as part of the /91 console, as does the image in http://www.subinet.es/un-poco-de-historia-el-ibm-36091/ which supposedly is the Goddard machine.



-- Charles

['Dave Wade' [e-mail] [hercules-390]]

Sorry to ramble, if you look at the Photo on the console of the 360/91 at LCM on this page, you can see the 2250




http://www.wikiwand.com/en/IBM_System/360_Model_91



Dave



From: hercules-***@yahoogroups.com [mailto:hercules-***@yahoogroups.com]
Sent: 27 November 2016 23:29
To: hercules-***@yahoogroups.com
Subject: Re: [hercules-390] Re: Translation exception 0005 in MVT boot.












On Sun, Nov 27, 2016 at 1:53 PM, Joe Monk ***@gmail.com <mailto:***@gmail.com> [hercules-390] <hercules-***@yahoogroups.com <mailto:hercules-***@yahoogroups.com> > wrote:



"The /91 operator console was an I/O Selectric style."



Nope. The Operator Console on a 360/91 was a 2250 CRT model 1. It's detailed on page 6 of the Model 91 Functional Characteristics:



Hmm. The LCM console clearly has a Selectric attached as part of the /91 console, as does the image in http://www.subinet.es/un-poco-de-historia-el-ibm-36091/ which supposedly is the Goddard machine.



-- Charles

[Tony Harminc [e-mail] [hercules-390]]

Post by Joe Monk ***@gmail.com [hercules-390]
Hmm. The LCM console clearly has a Selectric attached as part of the /91
console, as does the image in http://www.subinet.es/un-
poco-de-historia-el-ibm-36091/ which supposedly is the Goddard machine.

And that's definitely a 1052, which was always a built-in Selectric based
printer/keyboard for almost all (maybe all) the S/360 models. It is a
channel-attached device with a very simple programming interface. One cute
aspect is that there is a Ring Bell CCW op code, and that bell on most 360s
was a 5-inch fire bell under the CPU covers, rather than the little
typewriter "ding" sort of bell you might expect from a Selectric device.

The 1052 should not be confused with either the 1050 or the 2740/2741,
which are all standalone terminal devices, and which all attach via TP
(sync or async lines) via a 270x or similar telecommunications controller.
The 1052 has its own control unit, and connects via bus & tag channel
cables.

Tony H.

[[e-mail] [hercules-390]]

that firebell was *loud*. I remember it going off on UIUCs 360/75 (we were reworking consoles, and got MVT to the point where it had no access to any working console, so it set off the bell). Scared the ^%$^ out of us.

definitely not "ding".

---In hercules-***@yahoogroups.com, <***@...> wrote :

And that's definitely a 1052, which was always a built-in Selectric based printer/keyboard for almost all (maybe all) the S/360 models. It is a channel-attached device with a very simple programming interface. One cute aspect is that there is a Ring Bell CCW op code, and that bell on most 360s was a 5-inch fire bell under the CPU covers, rather than the little typewriter "ding" sort of bell you might expect from a Selectric device.

[[e-mail] [hercules-390]]

scratch that.



https://groups.google.com/forum/#!msg/bit.listserv.ibm-main/nJL-FFH72pY/JqQGI9HMwZAJ https://groups.google.com/forum/#!msg/bit.listserv.ibm-main/nJL-FFH72pY/JqQGI9HMwZAJ


the ccw opcode was x'0B'.

[[e-mail] [hercules-390]]

any idea where to look up what the Ring Bell CCW op code was? I can't seem to find any docs on channel attached 1052 (Google and bitsavers came up dry)


The only thing I can think of is to look in the VM/370 R6 source; I recall that VM would put RING... RING... RING... on the console whenever the guest OS would run the bell....

---In hercules-***@yahoogroups.com, <***@...> wrote :

And that's definitely a 1052, which was always a built-in Selectric based printer/keyboard for almost all (maybe all) the S/360 models. It is a channel-attached device with a very simple programming interface. One cute aspect is that there is a Ring Bell CCW op code, and that bell on most 360s was a 5-inch fire bell under the CPU covers, rather than the little typewriter "ding" sort of bell you might expect from a Selectric device.

['Mark L. Gaubatz' [e-mail] [hercules-390]]

X'0B'. It's in the reference summaries as well as in the machine
materials on BitSavers.

Post by ***@sbcglobal.net [hercules-390]
any idea where to look up what the Ring Bell CCW op code was? I can't
seem to find any docs on channel attached 1052 (Google and bitsavers
came up dry)
.

[Gerhard Postpischil [e-mail] [hercules-390]]

Post by ***@sbcglobal.net [hercules-390]
any idea where to look up what the Ring Bell CCW op code was? I can't
seem to find any docs on channel attached 1052 (Google and bitsavers
came up dry)

It's on my 370 reference card as 0B. IIRC, it required a length of 1,
but I don't recall whether it accepted 0 as an address or required
something valid.


Gerhard Postpischil
Bradford, VT

[Ivan Warren [e-mail] [hercules-390]]

Post by Gerhard Postpischil ***@charter.net [hercules-390]

Post by ***@sbcglobal.net [hercules-390]
any idea where to look up what the Ring Bell CCW op code was? I can't
seem to find any docs on channel attached 1052 (Google and bitsavers
came up dry)

It's on my 370 reference card as 0B. IIRC, it required a length of 1,
but I don't recall whether it accepted 0 as an address or required
something valid.

Gerhard,

CCW opcode 0x0B on a 1052/3215 is an immediate control write operation
(that is Channel End is presented before the control unit asks for any
data). Therefore, the address portion of the CCW is ignored and the
length can be anything except 0 for format 0 CCW, and any value
including 0 for a format 1 CCW. If the value is not 0, SILI should also
be set in the CCW if the CCW is part of a CCW chain.

--Ivan



[Non-text portions of this message have been removed]

[Gerhard Postpischil [e-mail] [hercules-390]]

Post by Ivan Warren ***@vmfacility.fr [hercules-390]
CCW opcode 0x0B on a 1052/3215 is an immediate control write operation
(that is Channel End is presented before the control unit asks for any
data). Therefore, the address portion of the CCW is ignored and the
length can be anything except 0 for format 0 CCW, and any value
including 0 for a format 1 CCW. If the value is not 0, SILI should also
be set in the CCW if the CCW is part of a CCW chain.

Thanks, I had forgotten the details. Almost fifty years ago one of my
colleagues modified MFT to chain a bell CCW whenever a console message
had a routing and descriptor code of 1, and we added some WTOs when the
operators didn't respond to actions (tape mount, etc.) within 15
minutes. When we updated from a 360/40 to a 360/50, the system kept
crashing - IBM diagnostics didn't test the bell.

Gerhard Postpischil
Bradford, VT


------------------------------------

------------------------------------

Community email addresses:
Post message: hercules-***@yahoogroups.com
Subscribe: hercules-390-***@yahoogroups.com
Unsubscribe: hercules-390-***@yahoogroups.com
List owner: hercules-390-***@yahoogroups.com

Files and archives at:
http://groups.yahoo.com/group/hercules-390

Get the latest version of Hercules from:
http://www.hercules-390.org


------------------------------------

Yahoo Groups Links

<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/hercules-390/

<*> Your email settings:
Individual Email | Traditional

<*> To change settings online go to:
http://groups.yahoo.com/group/hercules-390/join
(Yahoo! ID required)

<*> To change settings via email:
hercules-390-***@yahoogroups.com
hercules-390-***@yahoogroups.com

<*> To unsubscribe from this group, send an email to:
hercules-390-***@yahoogroups.com

<*> Your use of Yahoo Groups is subject to:
https://info.yahoo.com/legal/us/yahoo/utos/terms/

['Mark L. Gaubatz' [e-mail] [hercules-390]]

Further information from SY22-2902-2 1052 Adapter and 2150 Console FETOP
(also available on BitSavers):

_Control Alarm_

The control alarm command is a control immediate. This command
activates an alarm for approximately two seconds. Channel end and
device end are transmitted during the initial selection status.
Control alarm commands should not be issued at intervals of less
than five seconds.

Sidebar: VM printing RING... RING... RING... plus the carriage return /
linefeed sequence took two seconds, effectively matched the real alarm
sounding time, for both the 1052 at 14.8 cps, and the 2741 at 14.1 cps.

Mark

Post by ***@sbcglobal.net [hercules-390]
any idea where to look up what the Ring Bell CCW op code was? I can't
seem to find any docs on channel attached 1052 (Google and bitsavers
came up dry)
The only thing I can think of is to look in the VM/370 R6 source; I
recall that VM would put RING... RING... RING... on the console
whenever the guest OS would run the bell....

['Dave Wade' [e-mail] [hercules-390]]

A non-CPU difference appears to be that the /91 predated 3270 (IIUC). The /91 operator console was an I/O Selectric style. The ready-to-roll OS/360 packages appear to be 3270 oriented. I'm going to have to figure out the intricacies of system configuration to match /91 era peripherals; possibly rolling back to an earlier version of OS/360 if needed (and possible).



The relevant screen as a 2260 (I think)



https://en.wikipedia.org/wiki/IBM_2260



but most sites did not have these. In any case TSO didn’t appear until 1971
.

.. we tend to use 3270 screens as there are 3270 emulators available for Windows and Linux, but there are no 2260 emulators.



(Looking at Wikipedia, /91 was first released in 1967; there appear to have been a slew of OS/360 releases that year; I would hazard a guess that one of the releases was a /91 release. I might be asking for trouble with a OS configured for the /91 if the interrupt structure is different and unsupported by Hercules; but if possible a period correct release would be preferable.)



I doubt it. The structure is similar, its just on a 91 you need extra checking to cater for the “imprecise interrupt”. I haven’t checked the config but as nothing emulates a 360/91 there won’t be any OS’s out there configured for a 91.



I am more interested in /91 like behavior; I don't require emulation of the OOO and related differences -- it is more about the user experience; it looks and feels like a /91. So configured as a /91 (appropriate peripherals, amount of memory), and if the OS or programs announce the model number it should be 360/91 or 2091 as appropriate.



I am not an OS specialist but I doubt that the model is visible anywhere, or if it is it would be generated in by hand.



I would suspect that the “user experience” didn’t usually exist. Users would typically write their Fortran code on Cards, put them in some in pigeon holes, they would get run, and the output and cards would be bundled together and you would pick them up. Now I know from some google searches that the 360/91 had some kind of graphics device. Perhaps a 2250 




<http://www.columbia.edu/cu/computinghistory/2250.html> http://www.columbia.edu/cu/computinghistory/2250.html



but there are no emulations for a mainframe attached 2250
.

(The SIMH IBM 1130 emulator does..)



Dave



From: hercules-***@yahoogroups.com [mailto:hercules-***@yahoogroups.com]
Sent: 27 November 2016 20:21
To: hercules-***@yahoogroups.com
Subject: Re: [hercules-390] Re: Translation exception 0005 in MVT boot.












On Sun, Nov 27, 2016 at 11:07 AM, Tony Harminc ***@gmail.com <mailto:***@gmail.com> [hercules-390] <hercules-***@yahoogroups.com <mailto:hercules-***@yahoogroups.com> > wrote:



On 27 November 2016 at 10:45, ***@gmail.com <mailto:***@gmail.com> wrote:



As for the model 91 - this was a rare beast that, while architecturaly compatible, did have some differences in interrupt handling (so-called imprecise interrupts) and out-of-order execution that (iirc) was visible to channel programs, but not to CPU programs.



A non-CPU difference appears to be that the /91 predated 3270 (IIUC). The /91 operator console was an I/O Selectric style. The ready-to-roll OS/360 packages appear to be 3270 oriented. I'm going to have to figure out the intricacies of system configuration to match /91 era peripherals; possibly rolling back to an earlier version of OS/360 if needed (and possible).



(Looking at Wikipedia, /91 was first released in 1967; there appear to have been a slew of OS/360 releases that year; I would hazard a guess that one of the releases was a /91 release. I might be asking for trouble with a OS configured for the /91 if the interrupt structure is different and unsupported by Hercules; but if possible a period correct release would be preferable.)



I am more interested in /91 like behavior; I don't require emulation of the OOO and related differences -- it is more about the user experience; it looks and feels like a /91. So configured as a /91 (appropriate peripherals, amount of memory), and if the OS or programs announce the model number it should be 360/91 or 2091 as appropriate.



-- Charles

[Mike Schwab [e-mail] [hercules-390]]

I think you could investigate the OS/MFT modified to run APL/360, it used
that era of consol.

Post by Charles Anthony ***@gmail.com [hercules-390]

Post by Tony Harminc ***@gmail.com [hercules-390]

Post by Tony Harminc ***@gmail.com [hercules-390]
As for the model 91 - this was a rare beast that, while architecturaly

compatible, did have some differences in interrupt handling (so-called
imprecise interrupts) and out-of-order execution that (iirc) was visible to
channel programs, but not to CPU programs.
A non-CPU difference appears to be that the /91 predated 3270 (IIUC). The

/91 operator console was an I/O Selectric style. The ready-to-roll OS/360
packages appear to be 3270 oriented. I'm going to have to figure out the
intricacies of system configuration to match /91 era peripherals; possibly
rolling back to an earlier version of OS/360 if needed (and possible).
(Looking at Wikipedia, /91 was first released in 1967; there appear to
have been a slew of OS/360 releases that year; I would hazard a guess that
one of the releases was a /91 release. I might be asking for trouble with a
OS configured for the /91 if the interrupt structure is different and
unsupported by Hercules; but if possible a period correct release would be
preferable.)
I am more interested in /91 like behavior; I don't require emulation of
the OOO and related differences -- it is more about the user experience; it
looks and feels like a /91. So configured as a /91 (appropriate
peripherals, amount of memory), and if the OS or programs announce the
model number it should be 360/91 or 2091 as appropriate.
-- Charles

[Tony Harminc [e-mail] [hercules-390]]

Post by Charles Anthony ***@gmail.com [hercules-390]
A non-CPU difference appears to be that the /91 predated 3270 (IIUC). The
/91 operator console was an I/O Selectric style. The ready-to-roll OS/360
packages appear to be 3270 oriented. I'm going to have to figure out the
intricacies of system configuration to match /91 era peripherals; possibly
rolling back to an earlier version of OS/360 if needed (and possible).

The Selectric-based console device (the 1052 and compatible 370-era
followon the dot-matrix 3215) continued to be supported in not only OS/360,
but in the subsequent OSs (MVS et al) for decades afterwords. It's only in
perhaps the last < ten years that support was dropped from z/OS. In fact
support continued long after any such hardware was being made or was even
attachable to a modern machine, because the 1052/3215 interface was a
logical one between VM and its guest OSs.

Hercules supports mapping a line-mode telnet connection to a 1052-protocol
device, so you should have no trouble running any version of OS/360 using
such a console.

(Looking at Wikipedia, /91 was first released in 1967; there appear to have

Post by Charles Anthony ***@gmail.com [hercules-390]
been a slew of OS/360 releases that year; I would hazard a guess that one
of the releases was a /91 release. I might be asking for trouble with a OS
configured for the /91 if the interrupt structure is different and
unsupported by Hercules; but if possible a period correct release would be
preferable.)

The interrupt *structure* is not different (unlike, say the 360/44), except
that a limited number of floating-point operations - when a program
interrupt results, which is not the normal course of events from the
application point of view - can set the instruction length field in the old
PSW to 0, indicating an imprecise interrupt. The 370/165 (and probably the
360/85 on which it was based) had a similar feature, though there were
fewer and/or different cases where this could occur. I'm not sure that any
OS support is required to take care of this, or if there is, it's trivial;
it's more a matter of the application programmer having to put more work
into figuring out what went wrong in his program.

I am more interested in /91 like behavior; I don't require emulation of the

Post by Charles Anthony ***@gmail.com [hercules-390]
OOO and related differences -- it is more about the user experience; it
looks and feels like a /91. So configured as a /91 (appropriate
peripherals, amount of memory), and if the OS or programs announce the
model number it should be 360/91 or 2091 as appropriate.

Well I think we're still not sure what "it" is here... But fair enough.

S/360 does not have the Store CPU ID (STIDP) instruction that was
introduced in S/370. So there is no architected way for the OS to discover
the machine model it's running on. (It is possible that various heuristics
could be used to do so, but in fact no IBM OS does.) So if there is a model
number in a message somewhere, it would have to be gen'd into the OS. and
that would mean chosen by you. BTW, the primary reason for telling the OS
what model it's running on is so that the appropriate machine check
routines are loaded and used when a machine check (interrupt) takes place.
Machine check recovery was much more basic in S/360 than in S/370, which
added a much better architected logging of information and ability to retry
without losing information (or at least knowing if information was lost or
corrupted). In S/360 little was done beyond logging the errors.

Tony H.

['Dave Wade' [e-mail] [hercules-390]]

Tony,

Did OS/360 ever support a 2250 console? Do you actually need a 2741 to run it?

Dave



From: hercules-***@yahoogroups.com [mailto:hercules-***@yahoogroups.com]
Sent: 28 November 2016 18:27
To: hercules-***@yahoogroups.com
Subject: Re: [hercules-390] Re: Translation exception 0005 in MVT boot.








On 27 November 2016 at 15:20, Charles Anthony ***@gmail.com <mailto:***@gmail.com> wrote:

A non-CPU difference appears to be that the /91 predated 3270 (IIUC). The /91 operator console was an I/O Selectric style. The ready-to-roll OS/360 packages appear to be 3270 oriented. I'm going to have to figure out the intricacies of system configuration to match /91 era peripherals; possibly rolling back to an earlier version of OS/360 if needed (and possible).



The Selectric-based console device (the 1052 and compatible 370-era followon the dot-matrix 3215) continued to be supported in not only OS/360, but in the subsequent OSs (MVS et al) for decades afterwords. It's only in perhaps the last < ten years that support was dropped from z/OS. In fact support continued long after any such hardware was being made or was even attachable to a modern machine, because the 1052/3215 interface was a logical one between VM and its guest OSs.

Hercules supports mapping a line-mode telnet connection to a 1052-protocol device, so you should have no trouble running any version of OS/360 using such a console.

(Looking at Wikipedia, /91 was first released in 1967; there appear to have been a slew of OS/360 releases that year; I would hazard a guess that one of the releases was a /91 release. I might be asking for trouble with a OS configured for the /91 if the interrupt structure is different and unsupported by Hercules; but if possible a period correct release would be preferable.)



The interrupt *structure* is not different (unlike, say the 360/44), except that a limited number of floating-point operations - when a program interrupt results, which is not the normal course of events from the application point of view - can set the instruction length field in the old PSW to 0, indicating an imprecise interrupt. The 370/165 (and probably the 360/85 on which it was based) had a similar feature, though there were fewer and/or different cases where this could occur. I'm not sure that any OS support is required to take care of this, or if there is, it's trivial; it's more a matter of the application programmer having to put more work into figuring out what went wrong in his program.

I am more interested in /91 like behavior; I don't require emulation of the OOO and related differences -- it is more about the user experience; it looks and feels like a /91. So configured as a /91 (appropriate peripherals, amount of memory), and if the OS or programs announce the model number it should be 360/91 or 2091 as appropriate.



Well I think we're still not sure what "it" is here... But fair enough.

S/360 does not have the Store CPU ID (STIDP) instruction that was introduced in S/370. So there is no architected way for the OS to discover the machine model it's running on. (It is possible that various heuristics could be used to do so, but in fact no IBM OS does.) So if there is a model number in a message somewhere, it would have to be gen'd into the OS. and that would mean chosen by you. BTW, the primary reason for telling the OS what model it's running on is so that the appropriate machine check routines are loaded and used when a machine check (interrupt) takes place. Machine check recovery was much more basic in S/360 than in S/370, which added a much better architected logging of information and ability to retry without losing information (or at least knowing if information was lost or corrupted). In S/360 little was done beyond logging the errors.

Tony H.

[Tony Harminc [e-mail] [hercules-390]]

Post by 'Dave Wade' ***@gmail.com [hercules-390]
Did OS/360 ever support a 2250 console?

Yes - I have actually used a 2250 as a console on a 370/165 (pre-DAT)
running MVT 21.7. I don't have great experience, but I remember it had
lighted PF keys and a light pen, both of which were supported by DIDOCS. I
also used it in graphics mode to run a chess program under MFT on a 360/44.

Post by 'Dave Wade' ***@gmail.com [hercules-390]
Do you actually need a 2741 to run it?

Not quite clear - to run the 2250, to run OS/360, or to operate a 360/91?
(Or something else?) In any case, I'm not sure that the 2741 is supported
as a console. It's an async serial terminal that runs at 134.5 bps, so
there would have to be a 270x or 370x controller to connect it to. I've
never seen one as a console, but it's certainly not impossible.

I don't believe any OS/360 version requires a printer/keyboard console.
Certainly we ran the 370/165 on MVT with the built in 3066 as the primary
console, a set of 2260s left over from an earlier 360/65 as
tape/disk/printer area consoles, and that 2250 that had been attached to a
360/44 used very briefly as a console as well as an application device.

Tony H.

['Dave Wade' [e-mail] [hercules-390]]

From: hercules-***@yahoogroups.com [mailto:hercules-***@yahoogroups.com]
Sent: 28 November 2016 23:00
To: hercules-***@yahoogroups.com
Subject: Re: [hercules-390] Re: Translation exception 0005 in MVT boot.








On 28 November 2016 at 17:44, 'Dave Wade' ***@gmail.com <mailto:***@gmail.com> wrote:

Did OS/360 ever support a 2250 console?



Yes - I have actually used a 2250 as a console on a 370/165 (pre-DAT) running MVT 21.7. I don't have great experience, but I remember it had lighted PF keys and a light pen, both of which were supported by DIDOCS. I also used it in graphics mode to run a chess program under MFT on a 360/44.



Do you actually need a 2741 to run it?



Not quite clear - to run the 2250, to run OS/360, or to operate a 360/91? (Or something else?) In any case, I'm not sure that the 2741 is supported as a console. It's an async serial terminal that runs at 134.5 bps, so there would have to be a 270x or 370x controller to connect it to. I've never seen one as a console, but it's certainly not impossible.

What I meant was did you need a printer keyboard console.

I don't believe any OS/360 version requires a printer/keyboard console. Certainly we ran the 370/165 on MVT with the built in 3066 as the primary console, a set of 2260s left over from an earlier 360/65 as tape/disk/printer area consoles, and that 2250 that had been attached to a 360/44 used very briefly as a console as well as an application device.

It sounds like multiple consoles were the norm for MVT on large systems. My only experience of 360 operations was the 360/67 at Newcastle which ran MTS and only had one keyboard/printer console.

Tony H.



Dave W

[[e-mail] [hercules-390]]

- - - In hercules-***@yahoogroups.com, <***@...> wrote:
- - - beginning snipped - - -

Post by Tony Harminc ***@gmail.com [hercules-390]
I don't believe any OS/360 version requires a printer/keyboard console.

- - - ending snipped - - -

We at one time had seven consoles including a 1052, 2741s, and 3277s.
We could do an output only 1403 console or a composite 2501 with a
1403 console.
We had an RJE/RES type workstation close by in our warehouse
which could enter console commands and receive responses.
Later we got rid of the 2741s and switch to an output 3210 when
we switched from a S360/65 to a S370/168 with a special
feature of not having DAT. When we upgraded the S370/168 to
have be normal and have DAT, we switched from MFT-II to OS/VS1.
Prior equipment included S360/40 and S360/50.

Back then, there was no SDSF and neither MFT-II nor OS/VS1 supported
TSO so to scan back through the console log, having a printer console
was almost a requirement. There were other ways but not as convenient.

[[e-mail] [hercules-390]]

The 91 does not have decimal instructions, so those are emulated by the OS. There should be a sysgen option to turn this on. Then turn them off on Hercules.

As others note, the 91 allows for imprecise interrupts, and the S0C0 abend code. I suspect Hercules won't generate those. There is a console switch on the 91 to turn off the ability to generate them.


Also, the floating point divide instructions return the rounded quotient, instead of the truncated quotient specified for the architecture. That would be an interesting addition to Hercules.

[Charles Anthony [e-mail] [hercules-390]]

Post by Tony Harminc ***@gmail.com [hercules-390]

Post by Charles Anthony ***@gmail.com [hercules-390]
I am more interested in /91 like behavior; I don't require emulation of
the OOO and related differences -- it is more about the user experience; it
looks and feels like a /91. So configured as a /91 (appropriate
peripherals, amount of memory), and if the OS or programs announce the
model number it should be 360/91 or 2091 as appropriate.

Well I think we're still not sure what "it" is here... But fair enough.
S/360 does not have the Store CPU ID (STIDP) instruction that was
introduced in S/370. So there is no architected way for the OS to discover
the machine model it's running on. (It is possible that various heuristics
could be used to do so, but in fact no IBM OS does.) So if there is a model
number in a message somewhere, it would have to be gen'd into the OS. and
that would mean chosen by you.

Hmm...

Seen on 3270 console:

IEA218I MOD=158, ALTSYS=350 ASSUMED S370

Poking around in os360s:

os360s/dlibs/Modgen2/OEAANIP.txt

.NODDR1 ANOP 19022
81550020
DC C' ASSUMED' 19022
81557020
.* THE FOLLOWING 4 LINES OF CODE WERE PUT IN FOR PROJ 21961 BY BH
81557521
AIF (('&MODID' EQ '0030') OR ('&MODID' EQ '0040')).MODTST1
81558021
AIF (('&MODID' EQ '0050') OR ('&MODID' EQ '0065')).MODTST1
81558521
AIF (('&MODID' EQ '0075') OR ('&MODID' EQ '0085')).MODTST1
81559021
AIF (('&MODID' EQ '0091') OR ('&MODID' EQ '0195')).MODTST1
81559521
DC C' S370' . CK 21108
81560521
AGO .MODTST2 CK 21108
81561221
.MODTST1 ANOP CK 21108
81561921
DC C' S360' . CK 21108
81562621

Wikipedia says the AIF is a conditional pseudo-op, so I am assuming that
this is a build-time thing. Looking thorough the os360mvt scripts, I find:

stage1.jcl: CENPROCS MODEL=158, 370/158 CPU

I will muck about with the "SYSTEM CONFIGURATION" parameters in stage1.jcl
to try to move the OS closer to the /91 configuration.

-- Charles



BTW, the primary reason for telling the OS what model it's running on is so

Post by Tony Harminc ***@gmail.com [hercules-390]
that the appropriate machine check routines are loaded and used when a
machine check (interrupt) takes place. Machine check recovery was much more
basic in S/360 than in S/370, which added a much better architected logging
of information and ability to retry without losing information (or at least
knowing if information was lost or corrupted). In S/360 little was done
beyond logging the errors.
Tony H.

[[e-mail] [hercules-390]]

Ah,yes - the manuals. They were a wonderful thing! I recall there were manual stands at all theee 360s in the computer room. Then there was a subset of themon a stand in the programming dept. And, us COBOL guys were blessed with a small number of programming-related manuals which we read cover-to-cover and referenced constantly.


On another note,later on when I was with Price Waterhouse working on a Federal water pollution monitoring system for the State, starting around 1971 or so we had to compatability test all our COBOL (and JCL, if I recall correctly) programs over at the Legislature's RCA Spectra/70 as a disaster recovery strayegy. The code was 100% compatable on both platforms - it was uncanny. Since this was before the Feds went after IBM for monopolistic practices, I don't know how RCA got away with this. Snow White and the Seven Dwarfs.

[Harold Grovesteen [e-mail] [hercules-390]]

Post by ***@gmail.com [hercules-390]
Given the context that I am a CDC & Honeywell mainframe guy and IBM is
new foray for me, where is a good place to start with understanding
the architectural details of 360/370 architecture differences and
model 91 specific differences?
-- Charles

Bitsavers has manuals related to nearly all 360 and 370 models,
including two on the 360-91:

A22-6907-2 and GA22-6907-3. Both of these are versions of the
Functional characteristics manual that usually detail the unique aspects
of the device being describe.

http://www.mirrorservice.org/sites/www.bitsavers.org/pdf/ibm/360/funcChar/

Harold Grovesteen

[Mike Stramba [e-mail] [hercules-390]]

Post by ***@gmail.com [hercules-390]
HHC00801I Processor CP00: Addressing exception code 0005 ilc 6
HHC02324I PSW=00000005 C0000118 INST=D2FF4000FC7E MVC 0(256,4),3198(15)

Interesting how there is no "HHCCP014I CPU0000: Specification
exception ...." message for the 'C0' in the
5th PSW byte.

I tried OSTAILOR NULL and OSTAILOR OS/390 and they both show the
Specification message.

Mike

Post by ***@gmail.com [hercules-390]
My goal is to get OS/360 MVT running on an emulated 360/91.
Host system is x86_64 running Fedora 23.
I cloned hyperion: commit a87eaab20 Fri Nov 18 15:51:28 2016
Installed all needed and optional packages except libbz2.
Built and installed.
Following the instructions at Introduction to Generating and Running
OS/360 on Hercules http://www.conmicro.com/hercos360/, performed through
step 8, "Final MVT setup".
(The gen.cnf and mvt.cnf files were modified: TIMEZONE set to -0800,
CPUMODEL set to 2091, SYSEPOCH set to 1960 -28)
hercules -f mvt.cnf,
It appears to start normally
Start 3270 terminal
x3270 localhost:3270
The Hercules logo appears.
Do "ipl 350"
HHC01603I ipl 350
HHC00801I Processor CP00: Addressing exception code 0005 ilc 6
HHC02324I PSW=00000005C0000118 INST=D2FF4000FC7E MVC 0(256,4),3198(15)
mo
HHC02326I R:00800000: Translation exception 0005
HHC02326I R:00000D00:K:06=00000000 00000000 00000000 00000000
................
HHC02269I GR00=00000000 GR01=00000000 GR02=00000000 GR03=00000000
HHC02269I GR04=00800000 GR05=00000000 GR06=00000000 GR07=00000000
HHC02269I GR08=00000100 GR09=00FFFFFF GR10=0000011E GR11=00000000
HHC02269I GR12=00000000 GR13=00000000 GR14=00000000 GR15=40000082
HHC00801I Processor CP00: Addressing exception code 0005 ilc 2
HHC02324I PSW=0000000550000132 INST=0832 SSK 3,2
se
HHC02269I GR00=00000000 GR01=00000000 GR02=00800000 GR03=00000000
HHC02269I GR04=00000800 GR05=00000000 GR06=00000000 GR07=00000000
HHC02269I GR08=00000100 GR09=00FFFFFF GR10=0000013E GR11=00000000
HHC02269I GR12=00000000 GR13=00000000 GR14=00000000 GR15=40000082
HHC00107I Starting thread cckd_ra(), active=0, started=0, max=2
HHC00100I Thread id 00007f56ac74c700, prio 2147483647, name Read-ahead thread-1
HHC00107I Starting thread cckd_ra() from cckd_ra(), active=1, started=1, max=2
HHC00100I Thread id 00007f56ac64b700, prio 2147483647, name Read-ahead thread-2
HHC01315I 0:0350 CHAN: ccw 07000C08 40000006=>00000194 0000
HHC01312I 0:0350 CHAN: stat 0E00, count 0000
HHC01313I 0:0350 CHAN: sense 80000000 38931204 00000000 00000000 00000000 000000
HHC01314I 0:0350 CHAN: sense CMDREJ
HHC00809I Processor CP00: disabled wait state 00020000 80000005
herc =====>
IEA101A SPECIFY SYSTEM PARAMETERS FOR RELEASE 21.8F MVT
but the CPU has halted.
-- Charles

[[e-mail] [hercules-390]]

Post by Mike Stramba ***@gmail.com [hercules-390]

HHC00801I Processor CP00: Addressing exception code 0005 ilc 6
HHC02324I PSW=00000005 C0000118 INST=D2FF4000FC7E MVC 0(256,4),3198(15)

Interesting how there is no "HHCCP014I CPU0000: Specification
exception ...." message for the 'C0' in the
5th PSW byte.
I tried OSTAILOR NULL and OSTAILOR OS/390 and they both show the
Specification message.
Mike

- - - old notes snipped - - -

In the BC mode PSW, the C0 means ilc 3 for the 6 byte MVC.
The 5 before the C0 means a PIC 5 which is to be expected.

If an EC mode PSW, it would have issues.

[[e-mail] [hercules-390]]

Post by Mike Stramba ***@gmail.com [hercules-390]
I tried OSTAILOR NULL and OSTAILOR OS/390 and they both show the
Specification message.

Not totally sure why you posted the above,
but if you're not already aware, there is a
"QUIET" option too, which seems to be
the opposite of "NULL":


if (strcasecmp (argv[1], "NULL") == 0)
{
sysblk.pgminttr = 0xFFFFFFFFFFFFFFFFULL;
return 0;
}
if (strcasecmp (argv[1], "QUIET") == 0)
{
sysblk.pgminttr = 0;
return 0;
}



BFN. Paul.

[[e-mail] [hercules-390]]

Okay, I switched to the instructions at TechInfo:OS:IBM Mainframes:OS/360 Installation - neurotica.com http://www.neurotica.com/wiki/TechInfo:OS:IBM_Mainframes:OS/360_Installation, much better results -- I was able to install and boot OS/360.

I had to make two changes from the page:


The lines


issue this command on the hercules console: m 350,vol=(sl,SCRTCH),use=private



should read "on the 3270 console window:"


When IPLing the installed system, the clock is invalid; I changed the response to "REPLY WITH SET PARAMETERS OR U" from


r 00,q=(,f)


to


r 00,'date=75.nnn,q=(,f)' <--- for "nnn" here, enter the three-digit day of year


I have not yet tackled the HASP and TSO installs....


Thank you for the help,


-- Charles