You are further illustrating my point.  Your 
astonishment was due to your inexperience. 

  The use of REFR for storage error recovery was only
in predecessors of MVS, and that was before my time
at IBM. I only know about that because of Shmuel's 
posts about it. 

   Although it was before my time, I would guess that the 
key 0 protection for RENT programs from APF authorized 
libraries was added only because it was required for 
system integrity, and limited to APF authorized libraries in
order to avoid migration impediments that would have 
resulted from breakage to existing self-modifying RENT
 programs in non-APF authorized libraries.  That would 
have been consistent with MVS's emphasis on compatibility,
and thus is not  astonishing to me. 

  Your "by happenstance it is refreshed and fails" scenario is
unlikely to ever occur, since refreshing is currently done 
only for LPA modules, which are DAT-protected.  z/OS will
not be implementing refreshing for storage machine checks.
The probability of a storage machine check on current machines
is so low that, if anything, we would be simplifying or removing
storage error machine check processing in z/OS, not  enhancing it.
In fact, while doing some work on System Trace in z/OS 2.5, I 
discovered that exploitation of 1MB frames for trace buffers in 
z/OS 1.10 had introduced a bug where a storage error 
machine check recovery for trace structures could result in
an infinite loop.  Based on discussion with hardware engineers
about the projections for storage machine checks on the machines
supported by z/OS 2.5, I remediated that by simplifying the 
recovery to just discard and rebuild the whole trace structure
for a processor instead of trying to fix the old logic for 
removing a single 4K frame from the processor's trace 
structure. 

Jim Mulder z/OS Diagnosis, Design, Development, Test  IBM Corp. 
Poughkeepsie NY


"IBM Mainframe Discussion List" <[email protected]> wrote on 
09/01/2021 01:58:03 PM:

> From: "Paul Gilmartin" <[email protected]>
> To: [email protected]
> Date: 09/01/2021 02:40 PM
> Subject: Re: RENT binder option
> Sent by: "IBM Mainframe Discussion List" <[email protected]>
> 
> On Wed, 1 Sep 2021 13:25:03 -0400, Jim Mulder wrote:
> >    ...
> >  As to the Astonishment in POLA, I would
> >suggest that astonishment is relative to the 
> >experience of the beholder, ...
> >
> The oldest memories are the sharpest.  I remember vividly my
> astonishment as an MVS/370 novice when I discovered that
> RER programs were not protected against modifying themselves
> despite ready availability of hardware facilities.  (Then I  could
> grasp the PoOps; no longer.)
> 
> I was even further astonished to learn that a program need to
> be Authorized in order to not modify itself.
> 
> I imagine a program incorrectly marked REFR which depends
> on self-modification.  It works for years until by happenstance
> it is refreshed  and fails.  Of course: "User Error.  WAD."
> 
> --gil



----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to