I pretty much agree, except there is not any "extreme" overhead in doing so. 
And there is not any issue with catching instructions
executed out of the data section via an EX instruction.


As a counter example, our production systems run with embedded debugger code, 
which does *not* affect execution speed at all.




-Paul

-----Original Message-----
From: Gibney, Dave [mailto:[email protected]]
Sent: Wednesday, July 14, 2010 02:16 PM
To: [email protected]
Subject: Re: z Linux assembler relative or friend or foe?

The original assertion was that under normal zLinux processing, certainproblem 
state instructions (specifically EPSW) were not allowed to beexecuted.As there 
seems to be no (without incurring the massive overhead ofeffectively 
interpreting/single stepping the execution (and even thatseems bypassable using 
the example give via the EX instruction)) nativeway to do this with zArch, this 
long discussion is occurring.The conclusion still must be that under routine 
execution of zLinux userprograms, all problem state instructions execute 
according to PoP. Yes,incurring the overhead of various tracing or screening 
during the loadof the executable code (or even scanning loaded code for 
dynamicallycreated instructions) could, at the expense of extreme overhead, 
preventEPSW or MVC or whatever, but IMO, no production system would bepractical 
with such a burden of overhead for such small, if any gain.Dave 
GibneyInformation Technology ServicesWashington State University

Reply via email to