CTCE links fail using TSAF under VM

[jeff-snyder]

CTCE links fail using TSAF under VM

[Peter-J-Jansen]

Hi Jeff,

Interesting, I never tried VM's TSAF, but spent quite some effort before I got SSI to work with CTCE's. Am I correct to assume that even the newer z/VM's also still support TSAF? I would certainly like to make TSAF work as well, and actually already started investigating the logs you sent. The first pair, JS01 <-> JS04, already show that the TSAF startup does not work, which in the past for me was always a big stumbling block, as about every CTC link CCW program solves the startup problem differently.

Solving this TSAF - CTCE problem will take some time, also because I'll be offline soon until the second half of March. So please bear with me.

Cheers,

Peter

[jeff-snyder]

Thanks for looking at this, Peter.

I'll write up a doc about how I have TSAF set up between two z/VMs, hopefully that will be enough to get you started.

One thing to be aware of, the TSAF on SP5 is incompatible with the TSAF on the other versions, so don't compare those logs (JS02 and JS10) hoping to find a fix for ESA and z/VM. When I run SP5 as a guest of VM/ESA, the VCTC links come up, but the SP5 side immediately drops with a message:

VCTCA between VM/ESA and VM/SP 5

VM/ESA side:

add link 818                                                                    
21:35:31 ATSLLM724I Link 0818 added                                             
21:35:36 ATSLMN707I Link 0818 came up

VM/SP 5 side:

add link 700                                                                    
ATSLLM724I Link 700 added                                                       
ATSNHR602E Incompatible release or service level detected on lin                
ATSLLM713I Link 700 deleted

I don't know how much their CTC drivers differ, but given that SP5 is the only 370 mode OS in the mix, it could be a lot.

I'm also working to get a VM trace of the VCTC when TSAF comes up successfully with JS01 is running as a guest of JS04. I should have more on that soon, I just need to do some research first. I've done a VM trace before, but it's been a while and I didn't really know what I was doing then...

Don't worry about the timing. I've got plenty to do until this gets going. Just make sure and let me know if there's anything I can do to help it along.

Jeff

[jeff-snyder]

Hey Peter,

Here's version 0.1 of the document on running TSAF over a VCTCA. I have a section in there for running it over a CTCE, but I'll flesh that out when we get further into this process and I can get past the first step without it failing.

FYI, this is rough and rambles a bit, but I'm hoping you'll be able to follow it and recreate what I've done. I figured it was more important to get it to you sooner and work on polishing it later.

Jeff

• TSAF between VMs.pdf

[jeff-snyder]

Hi @Peter-J-Jansen

I'm researching the VM I/O TRACE options and looking at what's avaiable with an eye toward getting it as close to what Hercules provides so it'll be easier to compare and contrast.

z/VM provides the ability to trace the following I/O "types":

SIO, SIOF, SSCH, RIO, RSCH

Use TRACE SIO, SIOF, SSCH, RIO, or RSCH to trace the SIO (START I/O), SIOF (START I/O FAST RELEASE), SSCH (START SUBCHANNEL), RIO (RESUME I/O), or RSCH (RESUME SUBCHANNEL) instructions for all input/output (I/O) devices or for a range of devices. Tracing is limited to instructions that do not produce an operation or privileged operation exception.

I was guessing that the SSCH would be most comparable to what you already have, but maybe it would be SIO? What do you think?

From what I've read so far, I believe I can trace multiple types. So, if you want SSCH and RSCH for example, that should be doable too. Or if you're like me, maybe you want to trace them all in the hopes you'll recognize something?  :)

Jeff

[jeff-snyder]

Hi @Peter-J-Jansen

After a bit more reading in the manuals, I found the "trace i/o ccw" command that looks like a good match for what we need. I created the attached trace and log files for TSAFVM on both the host and guest VMs.

Just a reminder, JS16 is the host VM. VCTC 807 in JS16's TSAFVM couples to VCTC 816 on the JS07 (guest) VM.

In the below attached zip, I have included the following files:

• JS07.TSAFVM.log.txt       (the guest TSAFVM session log)
• JS07.TSAFVM.trace.txt    (the trace of 816 on the guest)
• JS16.TSAFVM.log.txt       (the host TSAFVM session log)
• JS16.TSAFVM.trace.txt    (the trace of 807 on the host)

Jeff

• VCTC logs and traces.zip

[Peter-J-Jansen]

Interesting Jeff! The traces show how easy the TSAF startup CCW's apparently are using VCTC's: No OPeration (NOP), Set Extended Mode (SEM), REaD, Sense Adapter State (SAS) on one side only, and then just REaD's and WRiTes. So what's my CTCE code doing wrong?

I'll try to set up TSAF between two z/VM's, as per usual to first reproduce the problem, so I can start debugging for real.

But, as I said, I'm going to be offline for almost 2 weeks, for a little winter vacation over here. So please bear with me.

Cheers,

Peter

[jeff-snyder]

Great, I'm glad it's useful. Let me know if I can gather any more information or if I can help you recreate the environment for testing.

Don't worry about the downtime, it'll take me a day just to reorganize my To Do list and see which of the many things I have listed that I want to play with while you're out!

Jeff