> Initially, neither program was authorized. I was using the ARCRPEXT > exit, which, I believe, is not authorized(AC=0),
That's a non-sequitur. The exit routine (typically) is NOT linked AC(1) because that is only relevant for job step programs. The exit routine must be loaded from an APF-authorized library, but that is primarily to ensure installation control over the software that gets control in exit points. The exit routine's "authorization" is dictated by what ever environmental state that exists when the exit is entered. Most system exits are entered in supervisor state (and key zero) and are therefore able to do anything they want. I have NO knowledge of HSM, but I would bet large that all of its exits are entered in supervisor state. If so, you have been solving a non-problem, lo these many years. > and letting Control-O > intercept the WTOR message and format the modify command.. ARCRPEXT > runs in the HSM address space Then it almost certainly is entered in sup state... but at a minimum it would be considered an authorized environment, either by JSCBAUTH being set as a result of the AC(1) job step program, or by running in a system key. > and some requests were getting through to > the local HSM, which is a badness. I changed to using a front end to > the IGX00024 exit, so I am now running authorized in the local address > space. Which address space is "the local address space"? The issuer of HRECALL? > This has some rather intriguing side effects. The first is that > I can now stack modify commands. The ARCRPEXT appears to be single > threaded. The second is that, now that I am actually authorized, I may > as well issue the modify myself, instead of having Control-O intercept it. Or you could just bag the whole idea and implement a reliable signaling mechanism between HSM and your STC. > PS. I have read your second e-mail about Pause, Release and Transfer, I > will look into their availability in 1.4(my current environment) and, if > not, play with them in 1.7(my new sandbox). Thanks for the tip. Pause/Release services first became available in (I think) OS/390 2.5, if not earlier. They are available on every currently supported release. They are also a far superior mechanism for signaling between address spaces than dinking around with WTORs and (unserialized!!) ORE chains. You may still need to provide a mechanism for queuing requests from multiple callers unless (as you surmised) the exit is naturally serialized by HSM. CC ---------------------------------------------------------------------- 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
