[cgl_discussion] PoC concall minutes, 3/6

Rod Van Meter Rod.VanMeter at nokia.com
Thu Mar 6 10:36:33 PST 2003

Intel: Julie, Rusty
Nokia: rdv
OSDL: Randy, Mark
HP: Khalid
IBM: Venkat

Rusty is worried about moving /proc/net stuff to sysfs.  Worried about
single-value-per-file migration from a report-like structure.

Rusty asking about language for hi-res timers -- is it in
requirements?  If not, declare victory & move on.

Julie thinks there are still POSIX bugs; will report to LKML &
cgl_discussion.  More like, post status to cgl_d.

TIPC: Rusty thinks it has a wide enough audience that the goal is to
push it into mainline kernel.  rdv not sure, maybe distros is better.
Mika thinks aim is 2.7, but it's a long road.  Some parts, such as
name server, may need to be moved outside kernel.  Current approach is
to send patches to Mika and Jon.  Rusty thinking about bkbits tree for

Rusty points out that in panic.c the panic notification handlers are
not called.  He thinks maybe it should be a kernel option to call

If you want to test TIPC, become a lab associate and you can get
access to the resource pool of PCs at OSDL to use.

Please look at the list below and help me fill it out -- what are our


Mika wants a list of projects to be pushed into kernel.  I've added some
others for mainline:

pushing into 2.5:

* IPv6:
  - most core IPv6 ND, autoconf, bugfixes, etc. have gone in,
    including privacy extensions.  Some in 2.4.20, more in 2.4.21.
  - IPSec accepted by Dave Miller todayish, pushing to Linus
    probably 2.5.65
  - MIPv6 pending; may or may not make it?  MN & CN probably will, HA
    may not
  - SNMP (Randy/Yixiong) should make 2.5, but worried about changes to
  - timers (POSIX interface to clocks & timers is in)
  - possible bugs (see below)

waiting for 2.7:

* IPv6:
  - maybe MIP, esp. HA
  - hi-res timers


* kernel debugger
* kernel dumps


Non-kernel, therefore not subject to this logic:

(These items may have kernel dependencies, but will not be shipped as
part of the kernel.)

Well under way:

* OpenPOSIX test suite
* Hot device identity (status?)
  (MontaVista one may have had kernel change, Nokia was user-only)

More problematic:

* TAHI cooperation/IPv6 testing
* TIPC testing
* application loading (done, but not accepted by glibc?)

* IPv6:
  - glibc -- e.g., getifaddrs() needed by DHCPv6 (hosted by OSDL)
    ported from BSD, but haven't been pushed to glibc maintainers yet


Red flag items:

* boot cycle detection
* live SW upgrade
* resource monitoring
* OpenHPI still waiting on SAF/legal stuff; resources expected to be
  available (probably all non-kernel)

More information about the cgl_discussion mailing list