http://lists.osdl.org/pipermail/virtualization/2005-November/000231.html This post to lkml highlights a number of flaws in kprobe, some of which may have a security context (currently unknown). Needs kernel team investigation.
I'm going to have to punt this to the IBM developer who did all of the kprobes work for RHEL4; there's too much stuff in the reference above (specifically the LDT-related stuff) that's well beyond my understanding. I'll pass the reference on to him, ananth.com, for his take.
Following the reference in comment #1, it's pretty obvious that kprobe development is ongoing. Prasanna Panchamukhi (prasanna.com) is working with the reporter Zachary Amsden (zach) on several upstream issues. In any case, they (IBM) are obviously aware of (i.e. participating in) the thread above, and this is Ananth's reply re: its affect on RHEL4: Regards bz# 173230, per Prasanna, Zach is consolidating all LDT access/verification code in kernel and is yet to post the final patch accounting for all the feedback given. We are OK in the current state wrt LDT checks for RHEL4-U3 - it can be relooked after Zach's LDT consolidation is complete.
With the large kprobes backport patches that made it into RHEL4 U4 addressed all of the issues mentioned in this bugzilla. LDT and several others mentiond here is no longer an issues in 2.6.9-34.12EL and beyond. This bugzilla can be closed.
unless somebody wants to point to a specific problem, i'm closing this. *** This bug has been marked as a duplicate of 180260 ***