25 Dec
2015
25 Dec
'15
5:55 p.m.
I've recently been poking about with various bits of emulation with
hardware interfaces... Dave's MFM emulator; various SCSI-to-USB or
SCSI-to-SDcard devices; my Setasi RP12 Massbus disk emulator; the
Sigma Seven Lisa widget/ProFile emulator etc.
What about IBM channel-attached DASD?
There are various CPUs lying around in private collections and museums
- System/360s; System/370s; System/3 Model 15s; all used
channel-attached DASD: and working reliable disks are much rarer than
the damn CPUs!
Questions:
1. There are or were various 3rd party companies producing rather
obscure emulated DASD replacement subsystems - Virtualblue and Bustech
are two names that come to mind. Has anyone looked into the
possibility of using them to emulate older devices that would be
usable on the above vintage CPUs?
2. To those with hardware design experience: how big a task do you
reckon it would be to do this as a home-brew with modern hardware -
exactly as Dave did with his MFM emulator? Is it feasible? Do the
entire thing in software - Pi or Arduino or FPGA - with appropriate
driver electronics to drive a channel interface?
I've BCC'd some experts with experience - Rich Alderson; William
Donzelli; Henk Stegeman - in the hope that they'll be able to
contribute.
What do people reckon would be the best target for emulation? 3340
springs to mind initially... would that work on machines as old as
System/360s? It's about the *only* option for 5415 DASD...
(To digress briefly - a modern reimplementation of something like the
Setasi Massbus disk emulator would also be very useful; Rich - weren't
LCM working on something like that?)
Mike
http://www.corestore.org
'No greater love hath a man than he lay down his life for his brother.
Not for millions, not for glory, not for fame.
For one person, in the dark, where no one will ever know or see.'
25 Dec
25 Dec
7:12 p.m.
On 25 December 2015 at 20:55, Mike Ross <tmfdmike at gmail.com> wrote:
What do people reckon would be the best target for
emulation? 3340
springs to mind initially... would that work on machines as old as
System/360s? It's about the *only* option for 5415 DASD...
The 3340 is a pretty "small" disk at least for mainframe DASD.
More useful would be the 3330-1 and 3330-11. You can gen OS/360 21.8F
and MVS 3.8J to 3330 (assuming of course you have a 370 of some sort
knocking around).
Unless I'm completely wrong, most of the various control units
presented a programming interface that was relatively similar/the same
amongst the various kinds of DASD (looking at the 2844/2314 and
3830/3330 for example). Naturally the best way to double check that
would be to look at sufficient reference manuals, though if you felt
less inclined you could try browsing the source of the Hercules
emulator for how it does DASD emulation.
Now for the topic of System/3 DASD; firstly the control "speaks" a
completely different set of commands compared to the channel attached
DASD of the "proper" mainframe world. The 5444 disk is it's own thing
and not related to any of the channel attached DASD. The 5445 is
basically a single 2314 unit; and the 3340/3344 are exactly what they
say they are. Images of the 5445 and 334x drives would be
interchangable with the 2314 and 334x (since the format the System/3
uses on those disks is compatible with mainframe systems).
As a kind of dumb thought, for storing the images of the
Count-Key-Data DASD perhaps abuse the Hercules CKD/CCKD code into the
hardware emulator? (Means you can move your emulated disk packs to and
from a Hercules instance if you had reason to do so.)
Sorry for blathering on, I'm not an expert and I haven't a clue how
easy or difficult it would be to implement any of these ideas.
Cheers,
Christian
--
Christian M. Gauger-Cosgrove
STCKON08DS0
Contact information available upon request.
8:16 p.m.
I
On 12/25/15 5:55 PM, Mike Ross wrote:
What about IBM channel-attached DASD?
There are various CPUs lying around in private collections and museums
- System/360s; System/370s; System/3 Model 15s; all used
channel-attached DASD: and working reliable disks are much rarer than
the damn CPUs!
It's typical of most of the vintage gear. Folks save the CPUs
and ditch
the peripherals.
Questions:
2. To those with hardware design experience: how big a task do you
reckon it would be to do this as a home-brew with modern hardware -
exactly as Dave did with his MFM emulator? Is it feasible? Do the
entire thing in software - Pi or Arduino or FPGA - with appropriate
driver electronics to drive a channel interface?
I've looked at this briefly.
I would expect that there will have to be
more than just
software. Talking with David re: MFM emulator, that was at about the
limit of what
could be done with S/W. Anything else would require at least *some* H/W.
What do people reckon would be the best target for
emulation? 3340
springs to mind initially... would that work on machines as old as
System/360s? It's about the *only* option for 5415 DASD...
From my limited
experience, I would think that any of the CCK drives
would be
reasonable targets. I think once you get one (3330?) the others should
follow fairly
easily. I think once you've emulated the controller, supporting the
different drives
should be fairly straightforward.
(To digress briefly - a modern reimplementation of something like the
Setasi Massbus disk emulator would also be very useful; Rich - weren't
LCM working on something like that?)
Again, I've looked at it. I think the biggest problem is that there
isn't a spec per-se
on Massbus. A lot of reverse engineering will be required to make it
work properly
in all cases. It'll be a lot of reading (and fully understanding) the
drive schematics and
then trying to generalize it.
TTFN - Guy
9:32 p.m.
On Dec 26, 2015 5:16 PM, "Guy Sotomayor" <ggs at shiresoft.com> wrote:
isn't a spec per-se
I
On 12/25/15 5:55 PM, Mike Ross wrote:
(To digress
briefly - a modern reimplementation of something like the
Setasi Massbus disk emulator would also be very useful; Rich - weren't
LCM working on something like that?)
Again, I've looked at it. I think the biggest problem is that there
on Massbus. A lot of reverse engineering will be
required to make it
work properly
in all cases. It'll be a lot of reading (and
fully understanding) the
drive schematics and
then trying to generalize it.
TTFN - Guy
Another approach would be reverse-engineering the Setasi itself. That's a
known production-quality Massbus emulation, based on a 486 PC with one
custom card doing the Massbus interface... A card driven by a Xilinx FPGA
which could possibly be 'read out'...?
Crying shame time... Back in the mid 2000s I was in contact with John
Jones... ex DEC hardware guru and the guy behind Setasi. I agreed to buy
all the remaining bits of Setasi from him... Several units, doc, software -
and his development PC with all source and design files.
Then I had a bout with cancer, and by the time I picked up the threads
again he had given up on me and chucked it all in the dumpster :(
Mike
10:14 p.m.
On Fri, Dec 25, 2015 at 10:32 PM, Mike Ross <tmfdmike at gmail.com> wrote:
A card driven by a Xilinx FPGA
which could possibly be 'read out'...?
Could be "read out" in that the same bistream could be programmed into
another Xilinx FPGA of the same part number.
But if you're thinking that reading the FPGA bitstream is somehow
useful for reverse-engineering, no. The bitstream format is complex
and undocumented.
26 Dec
26 Dec
1:54 a.m.
-----Original Message-----
From: cctalk [mailto:cctalk-bounces at classiccmp.org] On Behalf Of Mike Ross
Sent: 26 December 2015 01:55
To: General Discussion: On-Topic and Off-Topic Posts
<cctalk at classiccmp.org>
Subject: IBM channel-attached DASD emulation
I've recently been poking about with various bits of emulation with hardware
interfaces... Dave's MFM emulator; various SCSI-to-USB or SCSI-to-SDcard
devices; my Setasi RP12 Massbus disk emulator; the Sigma Seven Lisa
widget/ProFile emulator etc.
What about IBM channel-attached DASD?
The Channel Protocols and Electrical Characteristics of the channel are reasonably well
documented. E.g.
http://bitsavers.trailing-edge.com/pdf/ibm/370/channel/GA22-6974-4_360_370_…
but IBM DASD are complex in that the physical block size can differ from track to track,
and can be changed by the user.
There are various CPUs lying around in private collections and museums
- System/360s; System/370s; System/3 Model 15s; all used channel-attached
DASD: and working reliable disks are much rarer than the damn CPUs!
The ICL2960 at TNMOC at Bletchley used similar disks. It now has emulated disk attached.
I don't think there are any working S/360's or even any that are complete enough
to be restored.
The same goes for real S/370. There might be 30xx or 43xx boxes but those are not 370/s..
Questions:
1. There are or were various 3rd party companies producing rather obscure
emulated DASD replacement subsystems - Virtualblue and Bustech are two
names that come to mind. Has anyone looked into the possibility of using
them to emulate older devices that would be usable on the above vintage
CPUs?
These seem rarer than the original DASD and less well understood. Probably as hard to keep
running as the originals.
2. To those with hardware design experience: how big a task do you reckon it
would be to do this as a home-brew with modern hardware - exactly as Dave
did with his MFM emulator? Is it feasible? Do the entire thing in software - Pi
or Arduino or FPGA - with appropriate driver electronics to drive a channel
interface?
The later IBM channel runs at 4Mbytes/second. Not sure software could keep up.
I've BCC'd some experts with experience - Rich Alderson; William Donzelli;
Henk Stegeman - in the hope that they'll be able to contribute.
What do people reckon would be the best target for emulation? 3340 springs
to mind initially... would that work on machines as old as System/360s? It's
about the *only* option for 5415 DASD...
I think the hard bit is getting the channel protocol working, once that?s running I think
emulating many models of DASD would be possible.
As someone else mentioned Hercules already has code that allows disk files to contain DASD
images in various sector sizes.
It may be possible to use this to talk to a real channel. Implementing some of the channel
protocol in FPGA might be necessary.
But I think finding a real S/360 would be harder...
..... but some one has an FPGA implementation...
http://www.ljw.me.uk/ibm360/
(To digress briefly - a modern reimplementation of
something like the Setasi
Massbus disk emulator would also be very useful; Rich - weren't LCM
working on something like that?)
Mike
http://www.corestore.org
'No greater love hath a man than he lay down his life for his brother.
Not for millions, not for glory, not for fame.
For one person, in the dark, where no one will ever know or see.'
Dave
G4UGM
11:14 a.m.
On Dec 26, 2015, at 4:54 AM, Dave Wade <dave.g4ugm
at gmail.com> wrote:
...
but IBM DASD are complex in that the physical block size can differ from track to track,
and can be changed by the user.
From sector to sector, actually. And sectors can have
key fields along with the data -- the device will find a sector by key match on a track.
paul
27 Dec
27 Dec
12:15 p.m.
Actually the DASD themselves are easy to emulate when fully buffered and
error free as IBM does today in its mainframes.
What is really hard to emulate is the Storage Control Unit and the
appropriate set of CCWs associated with a specific Storage Control Unit and
attached DASD. If I were doing this I might start with the 2314 DASF; it
might run on any Selector or Block Multiplexor channel and since the tracks
would be fully buffered in memory there really is no performance issue since
there would be no seek time or latency time. For the command set to emulate
see, Initial CKD implementation
<https://en.wikipedia.org/wiki/Count_key_data#Initial_CKD_implementation>
on Wikipedia or the 2314 reference manual. If there was more than one
device on the channel u might have some serious hogging issues
The 29MB/spindle or 232 MB/facility might be a limitation on some of the
newer systems, but it is possible to modify IBM's Oss to change the
characteristic of a DASD to almost anything by changing the device
characteristics in a table of such - that's how the PDMs attached double
capacity 2314s and 3350s; they redefined an unused device, e.g. 2311, to
have the characteristics of these non-IBM supported devices. In this case
all one might have to do is redefine the 2314 characteristics
The full step which would be really really interesting would be to build an
emulator that could operate as a Selector/2314/2318r, BMux/3380-2/3350
and/or (Bmux and Escon) /3990/3390. All together IBM implemented 83 CCWs
for its CKD DASD, most but not all of which would have to executed by such
an emuiator.
Tom
-----Original Message-----
From: Paul Koning [mailto:paulkoning at comcast.net]
Sent: Saturday, December 26, 2015 11:15 AM
To: General Discussion: On-Topic and Off-Topic Posts
Subject: Re: IBM channel-attached DASD emulation
On Dec 26, 2015, at 4:54 AM, Dave Wade <
<mailto:dave.g4ugm at gmail.com>
dave.g4ugm at gmail.com> wrote:
...
but IBM DASD are complex in that the physical block
size can differ from
track to track, and can be changed by the user.
From sector to sector, actually. And sectors can have
key fields along with
the data -- the device will find a sector by key match on a
track.
paul
26 Dec
26 Dec
8:21 a.m.
On 12/25/15 5:55 PM, Mike Ross wrote:
I've recently been poking about with various bits
of emulation with
hardware interfaces... Dave's MFM emulator; various SCSI-to-USB or
SCSI-to-SDcard devices; my Setasi RP12 Massbus disk emulator; the
Sigma Seven Lisa widget/ProFile emulator etc.
What about IBM channel-attached DASD?
FWIW, I picked this up in October
http://www.ebay.com/itm/230304621233
It is a PC-based device that lets you exercise channel peripherals off-line
simulating the CPU.
9:32 a.m.
FWIW, I picked this up in October
http://www.ebay.com/itm/230304621233
It is a PC-based device that lets you exercise channel peripherals off-line
simulating the CPU.
Does it work?
Can I borrow it?
For a long, long time?
--
Will
9:36 a.m.
On 12/26/15 9:32 AM, William Donzelli wrote:
Sure. I got it cheap and I owe you a favor.
FWIW, I picked
this up in October
http://www.ebay.com/itm/230304621233
It is a PC-based device that lets you exercise channel peripherals off-line
simulating the CPU.
Does it work?
Can I borrow it?
For a long, long time?
--
Will
9:45 a.m.
Does it work?
Can I borrow it?
For a long, long time?
--
Will
Sure. I got it cheap and I owe you a favor.
3066
days inactive
3067
days old
11 comments
9 participants
participants (9)
-
aek＠bitsavers.org -
captainkirk359＠gmail.com -
dave.g4ugm＠gmail.com -
ggs＠shiresoft.com -
paulkoning＠comcast.net -
spacewar＠gmail.com -
t.gardner＠computer.org -
tmfdmike＠gmail.com -
wdonzelli＠gmail.com