Marks Weblog

In trying to make modifications to some of the supplied operating code I was finding it quite frustrating going through the source code trying to guess which FMID I needed to apply module changes to. In many cases I was guessing through a range of FMIDs documented as being installed on the Turnkey3 documentation pages until I found one that didn’t report a mismatch.

One issue with that approach was it identified some of the FMIDs had PTFs installed that were not documented in the installed PTF documentation page on the Turnkey3 site. At that point I decided to read the SMP4 manual.

And joy, SMP provides a function that will tell you exactly what executable load module a program I am enhancing (or breaking) will be linked into along with the FMID.

A batch job to do that is

//MARKSMPL JOB  (SMP),'LIST SMP4 INFO',CLASS=A,MSGCLASS=A,REGION=4096K
//HMASMP  EXEC PGM=HMASMP,PARM='DATE=U',REGION=5120K,TIME=1440
//SYSUT1   DD  UNIT=WORK,SPACE=(1700,(600,100))
//SYSUT2   DD  UNIT=WORK,SPACE=(1700,(600,100))
//SYSUT3   DD  UNIT=WORK,SPACE=(1700,(600,100))
//SYSUT4   DD  UNIT=WORK,SPACE=(80,(2,2))
//SYSPRINT DD  SYSOUT=*
//ASMPRINT DD  SYSOUT=*
//CMPPRINT DD  SYSOUT=*
//COPPRINT DD  SYSOUT=*
//LKDPRINT DD  SYSOUT=*
//E37PRINT DD  SYSOUT=*
//UPDPRINT DD  SYSOUT=*
//ZAPPRINT DD  SYSOUT=*
//*************************** SMP DATASETS *********************
//SMPOUT   DD  SYSOUT=*
//SMPLOG   DD  DUMMY
//SMPTLIB  DD  UNIT=3350,VOL=SER=SMP004,DISP=OLD
//SMPACDS  DD  DISP=SHR,DSN=SYS1.SMPACDS
//SMPACRQ  DD  DISP=SHR,DSN=SYS1.SMPACRQ
//SMPCDS   DD  DISP=SHR,DSN=SYS1.SMPCDS
//SMPCRQ   DD  DISP=SHR,DSN=SYS1.SMPCRQ
//SMPMTS   DD  DISP=SHR,DSN=SYS1.SMPMTS
//SMPPTS   DD  DISP=SHR,DSN=SYS1.SMPPTS
//SMPSTS   DD  DISP=SHR,DSN=SYS1.SMPSTS
//SMPSCDS  DD  DISP=SHR,DSN=SYS1.SMPSCDS
//SMPWRK1  DD  UNIT=WORK,SPACE=(CYL,(5,10,84)),DCB=(BLKSIZE=3120,
//             LRECL=80)
//SMPWRK2  DD  UNIT=WORK,SPACE=(CYL,(5,10,84)),DCB=(BLKSIZE=3120,
//             LRECL=80)
//SMPWRK3  DD  UNIT=WORK,SPACE=(CYL,(5,10,84)),DCB=(BLKSIZE=3120,
//             LRECL=80)
//SMPWRK4  DD  UNIT=WORK,SPACE=(CYL,(1,10,84)),DCB=(BLKSIZE=3120,
//             LRECL=80)
//SMPWRK5  DD  UNIT=WORK,SPACE=(CYL,(30,10,250))
//************************ DO SOME LISTING **********************
//SMPPTFIN DD   DUMMY
//SMPCNTL  DD   *
 LIST CDS MOD(IEFAB421) XREF .
 LIST CDS MOD(IEFAB434) XREF .
/*
//

Which provides the information I needed to determine exactly what my changed code will affect when it is installed as a usermod.

     DATE 14.351  TIME 09:41:55      HMASMP LVL 04.48 MESSAGES AND CONTROL STATEMENTS                        PAGE 0001

HMA4240    HMASMP EXEC PARM = 'DATE=U'
 LIST CDS MOD(IEFAB421) XREF .
     DATE 14.351  TIME 09:41:55      HMASMP LVL 04.48 MESSAGES AND CONTROL STATEMENTS                        PAGE 0002

SMPCDS  MODULE ENTRIES


  NAME

IEFAB421  LASTUPD         = JCLIN    TYPE=UPD
          LIBRARIES       = DISTLIB=AOSB3
          FMID            = FBB1221
          RMID            = UZ58427
          LMOD            = IEFW21SD
          SYSMOD HISTORY  = SYSMOD   TYPE       DATE   MCS       ---------- STATUS ----------
                            EBB1102  FUNCTION  00.000  MOD                 ACC            RGN
                            FBB1221  FUNCTION  00.000  MOD                 ACC            RGN
                            UZ58427  PTF       74.164  MOD       APP       ACC

     DATE 14.351  TIME 09:41:58      HMASMP LVL 04.48 MESSAGES AND CONTROL STATEMENTS                        PAGE 0003

HMA2050    LIST PROCESSING COMPLETED - HIGHEST RETURN CODE IS 00


 LIST CDS MOD(IEFAB434) XREF .
     DATE 14.351  TIME 09:41:58      HMASMP LVL 04.48 MESSAGES AND CONTROL STATEMENTS                        PAGE 0004

SMPCDS  MODULE ENTRIES


  NAME

IEFAB434  LASTUPD         = JCLIN    TYPE=UPD
          LIBRARIES       = DISTLIB=AOSB3
          FMID            = EBB1102
          RMID            = UZ59776
          LMOD            = IEFW21SD
          SYSMOD HISTORY  = SYSMOD   TYPE       DATE   MCS       ---------- STATUS ----------
                            EBB1102  FUNCTION  00.000  MOD                 ACC            RGN

     DATE 14.351  TIME 09:42:02      HMASMP LVL 04.48 MESSAGES AND CONTROL STATEMENTS                        PAGE 0005

HMA2050    LIST PROCESSING COMPLETED - HIGHEST RETURN CODE IS 00

It will also provide a full summary of everything belonging to a function if you use “LIST CDS SYSMOD FUNCTION .” But thats a lot of output so not listed here.

The SMP4 manual at the mvs3.8j level is available at bitsavers.org, but as I refer to it a lot it is also available here as GC28-0673-6_SMP_SysPgmrGde_Sep80OCR.pdf.

Of interest to mvs3.8j hobbiests, probably not unless they want to run it on an OpenStack cloud; which they would probably need to be running at home as rackspace does not allow custom images installed.

Ok, to be truthful it is actually a Fedora 20 cloud-init enabled image. With hercules being the application that is installed on the image.

When booted into an instance it will automatically start hercules running my customised copy of Turnkey3.

As long as you associate a floating ip for the instance and open port 3270/tcp to the instance all you have to do is boot the instance, wait a while, and connect a 3270 emulator to the ip-address assigned on port 3270 for a TSO session.

But if you don’t want to fire up a kvm VM and would prefer a cloud fire-up/tear-down environment it is possible.

But I have added the image as an option in the hercules section of my website covering ‘how to obtain a working copy’ for those that may not want a live-DVD/live-USB solution.

Ok, real reason, I just wanted to do it.
And it’s tested and working; a new mvs38j system can be deployed and running in minutes as a standard OpenStack instance.

It’s available from my website page explaining the mvs3.8j OpenStack image if anyone is interested.

My intent is to create new console display commands. As MVS3.8J predates the ability to define new commands in the mpflist (actually it predates the mpflist the only option is to modify the appropriate source files for the operating system.

Found the source that needs to be modified

Making progress in reading through the MVS3.8J source files; so far I am fairly sure

• The console commands are all routed through IEE0403D (in MVSSRC.SYM101.F12) which then calls the appropriate module
• The module that manages the display commands is IEE3503D (in MVSSRC.SYM101.F13), somewhere to start from
• The sysgen macro SGIEF441 is apparently used to assemble those and all the many interwined dependant modules into lpalib

Complications identified

• I do not particularly want to do a new sysgen (well I do, but thats to gen new devices for another issue) so knowning the relevant sysgen macro is SGIEF441 is just commented here for others who may want to look at doing this properly
• IEE3503D is an alias to IGC0003D, and apparently linked into that program rather than being a standalone module (have not found where the source for IGC0003D lives yet), the SYS1.LPALIB entry for IEE3503D is an alias to IGC0003D
• also IEE0003D (the SVC34 handler) (which is not in sys1.lpalib,sys1.svclib or sys1.linklib) … is apparently also an alias to IGC0003D also (from somewhere, only comments in various source modules indicate that however), so IGC0003D is a bad thing to break
• all these interconnecting modules branch to each other with the “R R15,V(module)” then “BR R15” method, so they are either all linked in together or absolutely must be in memory (need to read another manual, or play and find out)
• did I mention all these are sysgen’ed, and all (apart from IEE0003D I cannot find in a load library) live in the primary LPA area

This has the potential to be as messy as when I was fiddling about replacing SVC modules; but I learnt a little from that :-). Whether it can be applied here I guess I will find out.

Current implementation plan

The easiest way forward I can see is for me (as I don’t want to do a new sysgen just for this, and I want to back out easily when (when not if) I break it is to

1. find a way to identify all the modules linked into IGC0003D
2. customise IEE3503D source code to link to my custom command handler before it branches to the program that spits out the “invalid command” console message (returning safely back if it is not one I handle to continue to the invalid command), store in a custom load library
3. relink IGC0003D with all the modules I found linked into it (todo) but replacing the IEE3503D procedure in IGC0003D with mine and store the new IGC0003D in a custom load library
4. adjust my MLPA list member (well create a new one) to include the new module and IPL with the CLPA and MLPA options to bring it in
5. undoubtably followed by an IPL with CLPA and my old mlpa member to get back to a working system :-) and months of debugging

Or, depending on whether I find BR R15,V(progname) requires the progname to be in the same linked module or just in memory I might try removing the alias to IGC0003D and MLPA’ing in my copy. I expect that to break things but is certainly the fastest way to break things :-).

However, where did I put my doco on using MLPA… I think in mvs3.8J modules MLPA’ed in must also live in sys1.lpalib or sys1.svclib anyway so I may have to replace the existing program (after backing up the system of course).

Current possible diversions

My current method of searching for source using existing utilities is one large job file that kicks off three jobs (due to mvs3.8j limit of 254 jobsteps I need three jobs) to perform a ‘pdsscan’ to locate members containing text strings I am looking for; one file scanned per job step and lots of output to wade through from a text prrinter file (impossible to view the output via Q due to so many jobsteps).

I need a program that takes a datatset prefix (ie: mvssrc or sys1), searches the catalogs to find all the datasets, reads the vtoc entry for the datasets to see if they are either dsorg=po+recfm=fb/f or dsorg=ps+recfm=fb/f (ie: ignore load libraries and vsam files etc), search for matching text in the files (different method depending on po or ps dataset of course), and therefore creates one output report to look at containing only datasets and members that matched. One report is easier to look through that 200+ saying no match found.

I find myself needing this program more and more often as I start trying to find out what connects where in all the OS source files. You can expect to find one in my website download section in a few weeks.

Current activites

It is a lovely sunny day outside. But as Teodora would not be the slightest bit interested in going on a picnic with me no matter how nice the weather I will stick to staying inside and coding; at least assemblers still talk to me.

• cloned a copy of my test mvs3.8j system to break
• extracted IEE3503D, want to see how many unresolved external modules a linkedit will throw out as a starting point
• started work on the file scanning programi; I thing I’m going to need it
• and trying to work out what live rodent the damn cat has just let lose in the house this time, scream

In the MVS3.8J section; if you don’t play with that move along.

The Oops was the number macro library I posted on the website in the mvs38j section yesterday was buggy; the main issue was there was a character 2 where there should have been a character 0, in assembler when addressing memory that’s a big oops. It works now :-).

Now the main post, “where is the damn command code”.

I see another few months (or years) of frustration ahead poking through the mvs3.8j source code for me.

In z/OS I believe you can use the MPF list to define new console commands; shame MPF doesn’t exist for mvs3.8j.

All I want to do is add a new display command available at the console, ie : D xxx.

So all I have to do is find the OS source code routine that handles console commands and insert/zap the new command so it runs the command instead of saying invalid display command.

All I have to do… I could search all the source PDS files to see what members contain the string “D” but that will not get me anywhere; I assume they all will.

I guess todo next weekend will be to find what module issues the “command invalid” error message, and over the next few months try and identify every module that calls that one.

Just because it irritates me.
The command I wanted to emulate was simply “D SMF”. As a temporary workaround I am using “S DSMF” and have also discovered that CMD1 can return the same info with #CS. But why settle for workarounds.

I may never find the OS code I need to modify; but it will keep me amused (or frustrated) for a few months anyway; until I find it or give up.