Alan,

Using LOADMOD with ORIGIN is just an old habit to simplify debugging of
a relocatable CMS module. Easier to calc addresses/offsets of entry
points from a loadmap and reuse PER TRACE commands and address ranges
multiple times.

The Operation Exception occurs just running this relocatable CMS program
in whatever address space to gets loaded into. It runs fine sometimes,
and other times it fails repeatedly. Segments of the module are not
getting loaded into storage which is very strange. PER STORE INTO the
missing address range of code, followed by the LOADMOD cmd never
triggers either.

This is a test environment with multiple VM LPARs and releases. I just
noticed that this LPAR, which is running V4 R4.0, is taking daily PRG004
CP abends as well. There is a larger issue that is making the system
unstable.
 
This system is at a remote location and I have limited access to it, so
I am going to let those folks who own this test system deal with the
issue. I suspect that there is a DASD sharing/Page space/Cache issue of
some kind.

Regards,

Mike     


 
-----Original Message-----
From: The IBM z/VM Operating System [mailto:[EMAIL PROTECTED] On
Behalf Of Alan Altmark
Sent: Thursday, November 27, 2008 9:58 AM
To: [email protected]
Subject: Re: CMS module doesn't get loaded completely on CMS level 20

On Wednesday, 11/26/2008 at 09:12 EST, Mike Rydberg
<[EMAIL PROTECTED]> 
wrote:
> I have a CMS module that when loaded on CMS level 20 is missing a
large 
chunk 
> of code for some reason.  Executable code is missing beyond offset
4D4B0 
and 
> the program gets an operation exception branching to zeros.
> 
> When loaded on CMS level 22 it runs fine. It also runs fine on earlier

versions 
> of CMS as far back as CMS level 15. The program hasn?t been recompiled

in 
> several years.
> 
> To debug the problem I used the loadmod and progmap commands. I then 
displayed 
> the routines entry point at offset 4D4B0  (6D4B0?base 20000)
> 
> CMS level 22 has executable code at this offset, CMS 20 does not, at 
least not 
> today anyway. It has run on CMS 20 in the past and the problem seems
to 
be 
> intermittent.
> 
> Anyone know of a bug of this type on VM 4.4.0 CMS 20? 

I've never heard of it and there's nothing relevant I can see in the 
problem tracking system.  Mike D. may know something.  (ORIGIN has been
on 
LOADMOD since VM/ESA 2.0.)

If you do a "STORMAP 20000-70000" prior to your LOADMOD, does it show
all 
of the range is unallocated?  If you just let the program load wherever
it 
wants (no ORIGIN), is the code at origin+4D4B0 there?

Also try this:
IPL CMS PARM NOSPROF INSTSEG NO
ACCESS (NOPROF
ACCESS the disk with the MODULE
LOADMOD ..
DISPLAY ..

This ensures that there isn't a rogue nucleus extension or other program

running about.

Alan Altmark
z/VM Development
IBM Endicott

Reply via email to