> I was reading the manual for an explanation of CPU lock: > CPU lock -- provides MVS-recognized (valid) disablement for I/O and > external interrupts. > > The manual gives a further explanation: > MVS does not guarantee preservation of the interrupt status of > programs that explicitly disable for I/O and external interrupts > through the STNSM instruction.
> What does 'does not guarantee' mean? The hardware architecture provides a simple mechanism for masking, enabling and disabling I/O and external interrupts. And if you're a supervisor state program and you know what you're doing those will do just what you expect. The operating system exposes a bunch of services on top of those mechanisms that are supposed to make them palatable for mere mortals to use. But as Walt said, there's a lot of stuff whirling around below the surface, so it is unreasonable to expect non-OS programmers to understand all of the environmental conditions and get it right with any sort of consistency. So IBM (has always) documented the CPU lock as the "official" mechanism for obtaining disablement. And by the way, the same is true for all of the other spin locks. That's what a "spin" lock implies - legal disablement, as well as serial access to whatever the specific lock implies. The cpu lock simply provides serial access to the cpu that holds it. Cute huh? Now with that said, no matter how you got disabled, once you -are- disabled for external and I/O interrupts, the only other kind of interrupts that can happen are the non-maskable ones - program check, machine check etc. The interrupted program's state is always stored by the hardware and (IMO) it would be astonishing for a disabled program to find itself redispatched in an enabled state. For the OS to do so, it would in effect be playing chicken with itself in many cases. I suspect RTM would simply abend the work that was considered to be in some state of sin. And I don't have any doubt that such cases might well exist in nature, for example in spin recovery or ACR. Bottom line, play by the rules and there's a lot less chance of getting hurt. And even then, it is not any place for the faint hearted. 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
