Minutes of 7/7/05 meeting

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

Minutes of SystemTap Conference Call, July 7, 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)
	Maneesh Soni (maneesh@in.ibm.com -- Bangalore)
Red Hat:
	Elena Zannoni (ezannoni@redhat.com -- Boston)
	Will Cohen (wcohen@redhat.com -- Raleigh)
	Martin Hunt (hunt@redhat.com -- Seattle)
	Frank Eigler (fche@redhat.com -- Toronto)
	Graydon Hoare (graydon@redhat.com)
Intel:
	Brad Chen (brad.chen@intel.com -- Santa Clara)
	Charles Spirakis (charles.spirakis@intel.com)
Hitachi:
	Satoshi Oshima

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

21. Rusty, Will: Look into adding kprobes tests to LTP and the
Red Hat suite, respectively.

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.

32. Martin/Tom: Benchmark relayfs performance vs. netlink in a
kprobes/SystemTap context.  [Update: in progress]

34. Vara/Satoshi: Discuss possibility of LKST demo at BOF.  [Update:
No plans to do an LKST demo at the OLS BOF.  Satoshi's team is
planning a different approach that involves a kprobes enhancement.
Satoshi has discussed it with Ananth, and will post the idea to the
mailing list next week.]

34a. Satoshi: Post description of LKST team's idea for a new approach.

35. Prasanna: Contact Andrew M to request that all approved kprobes
patches get pushed to Linus.  [DONE.  All patches to date have been
pushed to Linus.]

36. Jim: Create and submit Documentation/kprobes.txt writeup.
[Update: Doc by Prasanna & Jim posted to SystemTap list 6/28/05]

>From 6/30/05:
37. Vara: Publish C Tapsets Rev 3 after Jim's review.  [Update:
Jim's review is done.  Vara posted "What is a tapset?" proposal
7/7/05, which led to a discussion of C tapsets as well.]

>From 7/7/05:
38. Kprobes developers: Update status document
(htdocs/kprobes-status.html in CVS).

39. Ananth: Summarize the status of the ppc64 runtime port for Martin.

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

The Red Hat kernel team has accepted all kprobes patches into RHEL4 U2.
The exception is some ia64 patches, which Ananth has received from
Intel and will test and submit very soon.  Also, Ingo Molnar has
incorporated relayfs into RHEL4 U2.

Especially in view of the above, the kprobes status page on the web
site needs to be updated.

We reviewed ARs.  See above for updates to ARs 23, 26, 30, 34, 34a,
35, and 37.

Ananth's port of the runtime to ppc64 still has a problem or two.
Ananth will summarize for Martin.

Graydon (and later Frank) reported on the status of the translator
system.  They hope to have an "end-to-end" demo for OLS.  It should
include integration with the runtime, but probably won't include
access to target variables in the probed function.

Prasanna has resurrected watchpoint (data-breakpoint) support for i386.
Should he go ahead and post it to LKML?  Yes.  Jim says the x86_64
port should be straightforward.  Ananth says that ppc64 can do
watchpoints, but has only 1 watchpoint register.

Martin reported a transport issue in the runtime, which he will discuss
with Tom.  Martin may also look into packaging with RPM.

Now that the RHEL4 U2 kernel includes the desired kprobes patches,
Will will retest with the new kernel.

Will has incorporated Rusty's exhaustive test of probes on the ia64
instruction set.  Ananth will port this test to ppc64, but Will needs
to adjust the test to handle architectures such as ppc64 where probes
on certain instruction types are intentionally rejected.

Will has looked at the gcc and Tinderbox build/test systems.
No decision yet, but leaning toward staying with the existing
infrastructure.

There's always room in the test suite for more tests.  New tests
should be posted to the SystemTap mailing list.

Brad is looking for more feedback on his validator.

Charles is experimenting with asynchronous probes.  Dtrace has both
an OS timer (which fires on a random CPU) and a profiler (on each
CPU).  On Linux, Charles is not seeing what he expected (in terms
of the IP?).  See his 7/6/05 post.  It was suggested that he check
his sample program in under src/runtime/probes.  It was suggested
that he look at the KSTK_EIP macro and, from SGI's KDB, kdba_bp.c
(or maybe kdba_bt.c?).

Please review Vara's "What is a tapset?" document, posted 7/7/05.

Ananth has been working on the kprobes backport to RHEL4 U2, and
on the ppc64 runtime port.

Prasanna has posted patches to LKML that aim to prevent the insertion
of probes on kprobes code, by moving all such code to a .kprobes.text
section.

Tom is preparing a post for LKML that he hopes will result in the
inclusion of relayfs in Linus's kernel.

Tom is also working on making the netlink/relayfs choice in the runtime
transparent to upper layers.  Frank feels that this is unnecessary,
as he envisions that the same process that registers the probes will
also handle the output from the probes, and so will control whether
netlink or relayfs is used.

Who's going to OLS?
IBM: Vara, Ananth, Maneesh
Red Hat: Elena, Will, Frank, Andrew (?), several Red Hat kernel folks
Intel: Brad, Charles, Rusty, Anil
Hitachi: Satoshi and 5-6 others from the LKST team

There will be no meeting on 7/21/05, which is during OLS.