A phone meeting of the TIPC working group was held on October 25, 2007.

Attendees:

Al Stephens (Wind River) -- meeting chair; Ravi Rao (Nortel associate);
Jon Maloy (Ericsson); Dominic Herity (Red Plain); Elmer Horvath (Wind
River)

Minutes:

1. Meeting opening

The meeting was called to order at 11:03 AM (EDT).

2. TIPC support update

A number of bugs (mostly new) have been reported over the last month.

- Al reported that the kernel panic caused by TIPC mentioned on the
net-dev mailing list is a known issue relating to TIPC not supporting
non-contiguous sk_buffs; the fix in TIPC 1.7 needs to be submitted to
the kernel mainstream.
- Elmer discovered that very long fragmented messages may be discarded
prematurely at the receiving end because TIPC places a limit on how soon
the final fragment must arrive relative to the arrival of the first
fragment; Elmer has a fix that resets the limit each time a new fragment
arrives.
- Elmer also reported a couple of issues arising when TIPC switches from
single node mode to network mode; these would only affect users who run
TIPC applications before assigning TIPC's network address (i.e. its
<Z.C.N> value), so are unlikely to impact most Linux users.  Elmer will
pass on the fixes he developed to Al.
- Jon reported that the native API is not re-entrant, so that problems
can arise when two application threads try to send messages using the
same TIPC port at the same time.  After some discussion it was agreed
that it was sufficient to document this limitation, since adding new
locks to the native API would impact performance; applications that are
impacted by this issue always have the option of implementing the
required locking themselves.
- Jon also reported that Ericsson developers had discovered that TIPC is
missing a lock that can crash the system if it "discovers" the same node
on two different interfaces at the same time.  Jon will pass the fix
that was developed to Al and Elmer.

Al also reported that Wind River management is looking at getting a
designer to do the submission of TIPC 1.7 code into the mainstream
kernel (under Al's supervision, of course); hopefully this effort will
begin prior to the next TIPC WG meeting.  The patches will be based on
the code in Dave Miller's net-2.6 git repository (located at
git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6.git).

Ravi reported an issue that has been encountered when deleting a TIPC
socket.  Al requested that he provide details via the TIPC discussion
list, since the problem could not be described easily over the phone.

3. TIPC virtualization update

Ravi (and Randy) recently reported that their first cut at TIPC
virtualization is working.  They are now at work trying to partition
TIPC into "base" and "per stack" parts, to avoid unnecessary duplication
of parts of the TIPC code base that can be shared by multiple stacks.

Elmer asked whether it would make sense to use TIPC's network id value
as the stack identifier, rather than a separate protocol id value; Al
was concerned that this might raise backwards compatability issues (as
well as other issues), and requested that a written proposal be
distributed to the group so that it could be properly evaluated.

The question was also raised as to whether it should be possible for two
different TIPC stacks to communicate with each other.  Elmer and Jon
seemed to think this might be desirable, however Al objected that this
seemed to fly in the face of primary reason for creating multiple stacks
in the first place -- namely, to ensure that the two networks did not
impact one another.  It was agreed that the issue could be ignored for
the time being, as there was no requirement to provide inter-stack
communication.

4. Meeting close

The meeting adjourned at 11:52. The next meeting is scheduled for
Thursday, November 29 at 11:00 AM EST.

Please let me know of any errors or omissions.

Regards,
Al Stephens

[END]


-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
tipc-discussion mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/tipc-discussion

Reply via email to