Behaviour of instructions not supported on z390

15 views
Skip to first unread message

Abe Kornelis

unread,
Feb 19, 2022, 11:26:48 AM2/19/22
to z390
Dear all,

Now that V1.8.0 is out, the development team are beginning work on the next set of changes.

For me, that means I'm looking into a set of changes to make "unsupported instructions" behave consistently.
For example, most privileged instructions are not supported on z390.

Making these instructions behave consistently means we need to define what the desired
behaviour should be. Currently, we some of these instructions cause a S0C1 abend,
some are implemented as a no-op, and some are partially implemented.

Ideally, we would fully implement these instructions in a meaningful way, but that
appears out of reach for the moment. Therefore I would suggest the following:
- in problem mode, issue a privileged-operation exception or a special-operation exception,
  depending on the instruction's  definition in the PoP.
- in supervisor mode (which we do not support yet!)
  issue a special-operation exception - unless/until we have a better implementation.

It's a rather arbitrary choice, so please do feel free to come up with a better suggestion!

Kind regards,
Abe
===

d...@higgins.net

unread,
Feb 19, 2022, 12:08:56 PM2/19/22
to z390

Abe, all

 

I think we should abort all supervisor instructions and currently unsupported instructions.  The preferred method to perform supervisory functions with z390 is via svc calls which in tern get matted into java functions for the 3 different supported host environments: Windows, Linux, and Mac OS.

 

--
You received this message because you are subscribed to the Google Groups "z390" group.
To unsubscribe from this group and stop receiving emails from it, send an email to z390+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/z390/2562d7d6-ec33-d812-d98c-a5a2818f6ec9%40bixoft.nl.

Abe Kornelis

unread,
Feb 20, 2022, 6:44:19 AM2/20/22
to z3...@googlegroups.com

Don,

fair enough - does that mean you concur with my proposal below?
That would be pretty close to the set of changes that John and I are currently working on.

Kind regards,
Abe
===

Op 19/02/2022 om 18:08 schreef d...@higgins.net:

d...@higgins.net

unread,
Feb 20, 2022, 9:40:59 AM2/20/22
to a...@bixoft.nl, z3...@googlegroups.com

Abe, John, all

 

How about 2 abend codes:

  1. S0C2 for privileged opcode not supported
  2. S0C1 non-privaledged opcode not supported yet

 

Today I’ve started on issue 300 to add MGRK support.  It currently causes S0C1.

Reply all
Reply to author
Forward
0 new messages