On Sun, 6 May 2012 15:26:33 -0700, Charles Mills wrote:
>
>I have z/OS tasks of all stripes driving a z/OS system exit. I have a finite
>supply of a resource that each task needs in the exit. Sometimes the
>resources are exhausted and will not be replenished for a while. What if we
>come through the exit and are out of the resources? It would be nice if all
>of the tasks in such a situation could wait until the resource was
>replenished, without each task having to do a GETMAIN or CPOOL GET, chaining
>it to the previous (not a trivial thing in a totally asynchronous multi-CPU
>situation), WAITing, and then unchaining the block and freeing it. It might
>be nice if there were a single control block on which all of the tasks could
>(effectively) "wait."
>
The conventional solution to this problem is the semaphore:
http://en.wikipedia.org/wiki/Semaphore_%28programming%29
I don't know that classic z/OS provides semaphores; others will surely
know better than I. In a z/OS UNIX (USS) shell script I've done something
using a POSIX FIFO.
Initially, an executive writes one token to the FIFO for each unit
resource to be used. Then it enters a loop reading tokens from
the FIFO. If no token is available, it blocks. When it successfully
reads a token, it forks a child to perform a chore. When a child
completes a chore, it returns a token to the FIFO.
I suspect my situation differs from yours in that the executive
manages the queue of chores; the forked children are not autonomous
as you describe your tasks.
-- gil
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to lists...@bama.ua.edu with the message: INFO IBM-MAIN
