[Bug kprobes/2184] New: jprobe_return() requirement should be eliminated

["anil dot s dot keshavamurthy at intel dot com" <sourceware-bugzilla@sourceware.org>]

In the current design their is a requirement that the 
instrumented jprobe_handler must call jprobe_return()
at the end. The jprobe_return()is nothing but 
executes break/int3 instruction so the control 
is passed back to kprobes infrastructure.

This restriction can be eliminated, if the normal return
from jprobe_handler returns to some well known kernel function
and from where we execute a break/int3. I.e before passing
the control to instrumented jprobe_handler() the retun address
is modified to go to well known function inside the kernel and
from here we execute break/int3 their by eliminating the need for
jprobe_return() from the instrumented jprobe_handler.

This technique is implemented in IA64 and seems to be working
very well and can be done for other architecture. In IA64,
jprobe_return() is defined as
static inline void jprobe_return(void) {}.

-- 
           Summary: jprobe_return() requirement should be eliminated
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: kprobes
        AssignedTo: systemtap at sources dot redhat dot com
        ReportedBy: anil dot s dot keshavamurthy at intel dot com


http://sourceware.org/bugzilla/show_bug.cgi?id=2184

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.