On Thu, 29 Mar 2012 11:00:15 -0500, Tom Marchant wrote: >On Thu, 29 Mar 2012 15:21:41 GMT, MD Johnson wrote: > >>Does anyone know what I could look for to detect when >>an SVC contains code to place the caller into and >>authorized state (key 0). > >You could run a GTF trace and examine all SVC calls and >returns. Key 0 is not an adequate test, though. You'd >have to test for supervisor state or any system key. >Even that is not sufficient, because you can't tell from >that whether they turned on JSCBAUTH. And then >there are PC routines.... > >If the SVC (or PC) passes control to an address supplied >by the caller, allowing the caller's code to run as part of >the SVC, you might be out of luck. > If the SVC (or PC) passes control with escalated privilege to an address supplied by the caller, isn't it ipso facto unsafe?
But overall this sounds related to attempts to prove program correctness mathematically; a theoretical impossibility outside finite-state systems, and impractical for many systems with a large finite number of states. -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
