Minutes of SystemTap Conference Call, Oct 6, 2005

[Kevin Stafford <kevinrs at us dot ibm dot com>]

Minutes of SystemTap Conference Call, Oct 6, 2005

Participants:
IBM:
       Ananth Mavinakayanahalli (mananth@in.ibm.com -- Boston)
       Jim Keniston (jkenisto@us.ibm.com -- Beaverton)
       Kevin Stafford (kevinrs@us.ibm.com -- Beaverton)
       Vara Prasad (varap@us.ibm.com -- Beaverton)
Red Hat:
       Will Cohen (wcohen@redhat.com -- Raleigh)
Intel:
       Josh Stone (joshua.i.stone@intel.com -- Santa Clara)
       Anil Keshavamurthy (anil.s.keshavamurthy@intel.com -- Hillsboro)
       Charles Spirakis (charles.spirakis@intel.com)
Hitachi:
       Satoshi Oshima (soshima@redhat.com)

Actions Required (ARs):
>From 5/19/05:

22. Rusty: Look into x86_64 limitation re: temporary disarming of
   reentrant probes. [DONE: posted to ML 6/24]

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 10/6/05: Will be done now that Intel has Jay's email
   address: jkt@redhat.com]

>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.]

>From 7/28/05:
41. Anil: Post a summary of known djprobes implementation issues.
   [Update 10/6/05: Defer at least until djprobes discussion settles
   down a bit.]

>From 8/11/05:
44. Tom: Automate relayfs tests and post them to mailing list.
   [Update: In progress]

>From 8/18/05:

53. Anil: Test RCU-based SMP scalability feature on ia64.
   [Update: continue testing on 32 way machine and post to ML]

>From 8/25/05:

55. Elena: Investigate approaches to making the full SystemTap bug
   list available to SystemTap developers outside Red Hat.

>From 9/8/05:

57. Martin: Runtime build issues fix or unlink documentation from 
Runtime webpage.

>From 9/22/05:

73. Prasanna: post user space probes paper to ML
   [Update: Will post to ML after 10/11-14/05 conference]

74. Ananth: post kprobes RCU patches to LKML, contingent upon resolution of
   BZ# 1345 and possibly others.
   [Update: Bug shows up with gcc 3.2.3-24, but not recent versions.
   Will post RCU patches to LKML after a final code review.]

76. Hien: post dumpstack routine that solves kallsyms_lookup PPC issue.

77. Frank: post prototype to do symbol lookup in the translator. 
(alternative
   to #76)

78. Vara: Correspond with Elena about the next face-to-face meeting.
   [Update: In progress]

>From 9/29/05:

81. Jim: Follow up with owners of AR's 22,30,41,44.  [DONE 10/5/05]

82. Josh: Follow up with owner of AR 26.  [DONE 10/6/05]

83. Kevin: Post SystemTap documentation, mirror examples on website.

>From 10/6/05

84. Jim: Propose face-to-face meeting agenda, post document.

85. Kevin: Run kprobes modules on RHEL4U2 x86 machine to test kprobes RCU
   patches.
   [DONE: Posted to ML 10/6-7]

86. Kevin: Compile list of known SystemTap package requirements to post to
   website.

87. Anil: Post feedback to Frank's "Next Steps" proposal.

The next face-to-face meeting will probably be in November
in Beaverton, OR.  Duration and agenda are TBD.  Jim will draft
an agenda for review.  We'll probably use the opportunity to give
SystemTap more visibility among kernel developers at IBM Beaverton
and Intel Hillsboro.

Examples and how-to documentation need to be augmented, and what
we do have needs to be more visible.  Kevin has posted the stap(1),
stapex(5), stapfuncs(5), & stapprobes(5) man pages to the web site.

Some potential users would like to use SystemTap on distros other
than RHEL4 U2 and Fedora Core 4.  Kevin will put together a list
of kernel and package requirements and post to the web.

The only known bug associated with the RCU (SMP scalability) patches
appears to be dependent on an old version (or range of versions) of
i386 gcc.  Comparison of kprobes-related .s files failed to reveal
a bug.  Ananth will post RCU to LKML, since it works with recent
gcc versions.

There was some discussion of djprobes.  Anil and/or Hitachi will look at
corresponding code in kerninst.  Anil suggested using a subset of
the djprobes approach to eliminate single-stepping (and the associated
trap).

--
Kevin Stafford
DES 2 | MS 2M3
Beaverton - OR
Linux Technology Center
IBM Systems & Technology
Phone: 1-503-578-3039
Email: kevinrs@us.ibm.com