1999-07-07 23:17:55

by Matti Aarnio

[permalink] [raw]
Subject: Re: IBM Mainframe Support

On Wed, Jul 07, 1999 at 06:23:36PM -0400, Jim Nance wrote:
> On Wed, Jul 07, 1999 at 01:45:10AM -0500, Linas Vepstas wrote:
> > Hi,
> >
> > I have patches to the Linux kernel that add support for the IBM
> > mainframe ESA/390 series of computers (and clones e.g. Hitachi
>
> This is great! I glanced at the web page, but its not clear to me if the
> kernel will run in a VM, of if you need the entire machine. I sort of
> got the idea that it would not run in a VM.

Add some extra privilege bits to your VM account, and you
will be able to run guest operating systems just fine.

It is so long since I have ran IX/370 at an 3033 complex, that
I don't remember details anymore :-( -- but my library contains
bound copy of IX/370 manuals :-) ( Compacts them from about two
meters of binders to half a meter of books. )
( I collect odd computer manuals for possible reference use.. )

A problem at IX/370 time (and still, I think) is that IBM
mainframes don't have asynchronous serial ports, nor any
equivalent interfaces. Mega-IO-throughput, and async serial
interfaces don't really mix... Back then a Series/1 computer
was hooked into a channel attachment, and it acted as serial
port multiplexor processor. ( IBM mainframes are like any
other computer, just that all interfaces they have are alike
SCSI ... Attachment to network/terminals/disks/whatnot goes
via attachment units. )

Installation guide has these notes for the IX370 virtual machine:

"IX/370 Entry in the VM/SP Directory"

USER IX370 password 4M 16M G # Start with 4 MB memory, max 16 MB
IPL CMS # Do IPL to CMS
ACCOUNT nnnnnn
OPTION ECMODE BMX SVCACCL # For running guest operating-systems
# in VM/SP, SVCACCL is for use of
# "VM/SP Handshaking" facility (optional)
CONSOLE ...
IUCV ALLOW # allows incoming IUCV messages
OPTION MAXCONN nn # how many concurrent IUCV sessions

+++ DASD defines, SPOOL defines, etc +++

There is then an EXEC3 script for a set of CP commands to be
executed at the virtual machine boot, then doing another
IPL from a disk where the real IX/370 boot loader image is
resident.

At IBM mainframes there is *no* BIOS, *NOTHING*.
Also, there are no initial filesystems from which to pull
anything, there are just device interface addresses to which
you speak with archaic IO commands.

You tell the VM/CP program at first to attach certain parts
of physical disks to your virtual machine, and activate them.
Then you do Initial Program Load (IPL) from any of loadable
devices (including your virtual card reader), and the program
you want to use comes up.

Because you usually want to do something more convenient, than
IPL from hard/virtualized devices, VM/SP users usually have
(had) their IPL as CMS - Conversational Monitor(ing) System.
If you compare CMS with anything more widely known, it was
back then "CP/M Done Right". (Filenames were 8+8, no stupid
8+3 things. But no directories either!)


I am afraid running Linux at "Big Iron" will be a curiousity
issue... ... but i you get it up, I would love to have a guest
account :)

> Jim

/Matti Aarnio <[email protected]> -- these days using silicon instead
of iron :)