Minutes of 8/4/05 meeting

[Jim Keniston <jkenisto at us dot ibm dot com>]

Minutes of SystemTap Conference Call, August 4, 2005

Participants:
IBM:
	Larry Kessler (lkessler@us.ibm.com -- Beaverton)
	Jim Keniston (jkenisto@us.ibm.com -- Beaverton)
	Hien Nguyen (nguyhien@us.ibm.com -- Beaverton)
	Kevin Stafford (kevinrs@us.ibm.com - Beaverton)
	Tom Zanussi (trz@us.ibm.com -- Chicago)
	Ananth Mavinakayanahalli (amavin@redhat.com, ananth@in.ibm.com --
Boston)
	Prasanna Panchamukhi (prasanna@in.ibm.com -- Bangalore)
Red Hat:
	Elena Zannoni (ezannoni@redhat.com -- Boston)
	Martin Hunt (hunt@redhat.com -- Seattle)
	Frank Eigler (fche@redhat.com -- Toronto)
	Graydon Hoare (graydon@redhat.com)
Intel:
	Charles Spirakis (charles.spirakis@intel.com)
	Anil KeshavaMurthy (anil.s.keshavamurthy@intel.com -- Hillsboro)
	Chris Elford
	Josh Stone
Hitachi:
	Satoshi Oshima

Actions Required (ARs):
>From 5/19/05:
19. Vara: Collect sample kprobes modules and post them to the
SystemTap web site.

22. Rusty: Look into x86_64 limitation re: temporary disarming of
reentrant probes.

23. Jim, Martin: Ensure that the runtime's stack-trace feature works
appropriately when there are return probes in the call stack.
[Update: Martin is working on this.  Jim will prototype an improvement
whereby the saved IP is updated before the handler is called.]

>From 6/9/05:
26. Rusty/Brad: Forward email address of Intel test person to Jay
Turner.  [Update: As of 7/7/05, Brad hasn't done this.  Did Rusty?]

>From 6/23/05:
30. Rusty: Review Will's rendition of Rusty's test that exercises
probes on a wide variety of instruction types.  [Update: As of 7/7/05,
Will still needs to hear back from Rusty.]

31. Anil: Publish description of problem and solution regarding ia64
probes set on slot 1 of a bundle. [DONE: See 7/11/05 posting]

36. Jim, Prasanna: Create and submit Documentation/kprobes.txt writeup.
[DONE: Accepted into -mm kernel 8/3/05]

>From 7/7/05:
39. Ananth: Summarize the status of the ppc64 runtime port for Martin.

40. Ananth, Will: Port the probe-all-instruction-types test to ppc64.

>From 7/28/05:
41. Anil: Post a summary of known djprobes implementation issues.
[Update: This discussion is still very active.]

>From 8/4/05:
42. Frank/Graydon: Post Graydon's summary of the state of the
translator.  [DONE 8/4/05]

43. Prasanna: Post work on user-space probes.

Frank reported on the translator.  Within limitations, all phases of
the translator system are operational, and the translator system
can be used to create and install instrumentation expressed as .stp
scripts.  Current limitations have been summarized by Graydon and
will be [have been] posted.  Folks are invited to start coding and
testing .stp scripts.  SystemTap will be included in RHEL U2 as
planned; near-term availability was discussed.

Martin reports that the kernel-user transport system is working well.
No relayfs fixes were needed.  However, Martin is investigating another
problem that looks like it might be related to unregistering kprobes
under heavy load.

Frank asked for more stress-testing of kprobes.  Will has added some
tests, and Ananth is adding some instruction-specific tests (e.g.,
cli/sti and pushf/popf on i386).  Prasanna is working on adding
kprobes tests to IBM's ABAT test framework.

Ananth has posted an update to his SMP scalability patches.
This update uses RCU instead of read locks.  He has tested it on an
8-way i386 system.  No solid benchmark numbers yet, but building a
probed kernel with -j8 is noticeably faster than when the old (Big
Kprobe Lock) implementation.

Prasanna asked for more review of his watchpoint and
debug-register-allocation patches.

Ananth is consulting with IBM performance people who are interested
in kprobes and SystemTap.  [Jim and Kevin are engaged in similar
activities.]

Satoshi reported on djprobes.  A number of issues have been raised,
and the djprobes authors are addressing them.  They plan to publish
an improved version of their documentation.  There was some discussion
about the safety check that must be completed before the transition from
a kprobe to a djprobe.  Frank reiterated his support for pursuing this
technology and working it into SystemTap whenever djprobes is deemed
mature.

What about porting djprobes to other architectures?  There certainly
could be issues -- e.g., CPU rendezvous on ppc64 is rather hit-or-miss.
The consensus was to make sure djprobes is mature on i386 before
porting it to other architectures.

Prasanna talked about user-space probes.  A simple patch for this is
checked into CVS.  [Where?]  Architecture-specific issues appear to be
minimal.  Probes on shared libraries are supported.  Prasanna will
post this work soon.