20090430 systemtap meeting notes

[Stan Cox <scox at redhat dot com>]

Here are the notes I took from the meeting.
-Stan

The systemtap 0.9.7 release will become RHEL 5.4 candidate.

The systemtap roadmap is being redrawn soon.  Let Frank know what your
interests are.

The current dwarfless kprobe.function, which uses kprobe symbol
resolution, doesn't conflict with the dwarf
symbol table mechanism and is perhaps good enough for the task at hand.
Is plain old symbol table info available for user space?

Folks are encouraged to try Tim Moore's graphing widget.

Prasad will be at OLS in July.  Folks are trying to organize an adjunct
tracing session. There will be a Systemtap paper presented by IBM.

Ananth wonders if anyone has looked at ksplice?
It may necessitate generating debug info for functions that are in the
kernel although Josh's on the fly debug info may make this less critical.
How will stap probe a ksplice instrument?  Source is a requirement
for case splice to work.  Ksplice replaces a function
by linking a new version of the function into the kernel and placing a
jump instruction at the start of the obsolete function, to direct
execution to the replacement code. 

Masami mentions that the first version of the insn decoder has normal
kernel opcodes but not floating point or coprocessor opcodes.  Still
needed is a build time test to insure that the decoder can
exhaustively determine the length of every insn in the kernel.  This
will be used by kprobe etc.