IBM Mainframe Discussion List <[email protected]> wrote on 09/21/2006 03:05:36 AM:
> On Wed, 20 Sep 2006 17:40:22 -0400 Jim Mulder <[EMAIL PROTECTED]> wrote: > > :>IBM Mainframe Discussion List <[email protected]> wrote on 09/20/2006 > :>11:21:40 AM: > > :>> On Wed, 20 Sep 2006 08:05:18 -0500 Andy Wood <[EMAIL PROTECTED]> > :>wrote: > > :>> :>On Wed, 20 Sep 2006 12:16:12 +0300, Binyamin Dissen > :>> :><[EMAIL PROTECTED]> wrote: > > :>> :>>It appears that all task related storage, including loaded modules, > :>are > :>> :>>released after the initialization routine ends despite the fact that > :>the > :>> :>TCB > :>> :>>remains around. > > :>> :>Does ATTR=(...,KEEPRGN) on the ASCRE help? If so, perhaps you need > :>that - > :>> :>in more ways than one. > > :>> That took care of the problem. > > :>> I wonder about the side effects that the PMR describes, but I can now go > :>on. > > :> Usually, the initiator calls VSM Get Region processing, which frees > :>all region (a.k.a Low Private) subpools (0-132, 250-252), and sets up > :>the region limits requested by the REGION parameter or IEFUSI. That is > :>what > :>freed the storage for your LOADed modules. KEEPRGN causes this processing > :>to > :>be bypassed, as well as bypassing the TRANSWAP for the PPT NOSWAP > :>attribute. > :>(I don't know why the TRANSWAP is bypassed - that may have been > :>unintentional. > :>But it means that NOSWAP in the PPT is effectively ignored when KEEPRGN is > :>specified). > > :> Since the region subpools are not freed, any region subpools used > :>during the initialization routine will continue to be key 0, when they > :>would usually be key 8 or the PPT key. So, for example, if you LOAD > :>a non-reentrant module during the initialization routine, subpool 251 > :>will be created as a key 0 subpool. Then later when your jobstep task > :>gets attached, since the region was not freed, subpool 251 will still > :>exist as a key 0 subpool. Subsequent LOADs of non-reentrant modules > :>into subpool 251 will be into key 0 storage (and fetch protected), > :>so you won't be able to execute them unless you are running key 0. > :>Also, the REGION/IEFUSI limits will not be honored. > > Just imagine how much easier this would be if this was in the manual. > > I guess an RCF would be in order, but all it could state is > > "Please document undocumented options" > > (which would be valid for pretty much any manual). > > Now that I know what is going on I can program a better solution. > KEEPRGN was not in the original ASCRE in SP3.1.0. It was added in SP4.1.0 because it was needed for some system address space. It may have been intentionally left undocumented due to the somewhat unintuitive side effects discussed above. If we wanted to document it, we would need to document these side effects, and figure out if there are other side effects, and decide if ignoring PPT NOSWAP should be a documented side effect, or a bug which should fixed. So there would be more development cost to this than just adding 2 lines of text to the current manual. Of course, you can also argue that the current manual is deficient for not explaining about the free region processing that occurs if the initialization routine returns to the system. If you choose to submit an RCF, I would suggest that you include the text from these IBM-MAIN posts. Jim Mulder z/OS System Test IBM Corp. Poughkeepsie, NY ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
