System Software Teleconference Minutes - 2006-03-06

[Mitch Bradley]

System Software Teleconference - 2006-03-06

Mitch (scribe), Richard (moderator), Chris, Zephaniah, Jordan, Andres,
David, Victor, Vance, Chin, Marcelo, J5, Ronak, Nat (TamTam)


New Action Items:

   AI: Richard - update trac tickets after releasing WLAN_RESET firmware

Action Items Still Open:

   AI: dcbw and marcopg to investigate mesh/avahi not working.
        done, but we're seeing crashing reports
      - No new status this week


Status:

Mitch:
  * Suspend/resume working from firmware, requires DCONLOAD ECO
  * Gave Andres a recipe for the firmware side of uncached DMA
  * Lilian released first cut of WLAN driver with WIFI supplicant,
    so we can start testing

Richard
  * Chasing bugs
  * Will release new firmware with wireless reset fix (eth0 doesn't show 
up sometime)

Andres:
  * fixed last known major kernel bug - kernel now stable
  * working on performance - e.g. USB causes CPU slowdown
     Andres and Marcelo are implementing noncached DMA

J5:
  * Stable build
    + Using kernel 2.6.21.rc2
    - biggest problem is performance of mesh view layout (GUI)
    - a few small cleanups
    - new some country-specific stuff like AbiWord
  + Hope to mark a stable build by the end of the week
    - But we need to give the country reps something by Wednesday

Marcelo:
  * NetworkManager fixes to make mesh "just work"
  * New Libertas drivers are in the new build (multicast support and bug 
fixes)
  * iperf problem is fixed - bad flow control in driver
  * Wants to get link-local IP address stuff into the build - not dangerous

Chris Ball:
  * Owns release notes for new build but stalled waiting for new OS build

Battery management
  * Richard - problem with LiFe batteries when they get low.
     ? Are wakeup and trickle charging the same?
     + Quanta: Supposed to trickle charge until 5.3V
     - Richard left it for 48 hours and it didn't charge enough

  * EC microcode is getting better, but still not perfect
  * Supposedly the trickle charge works for battery recovery, but it
    may take a day or so to recover a battery - NOT Working!!!

Wireless wakeup:
  * Had a conference call between OLPC and Marvell, made a plan
    for how the wireless wakeup is supposed to work

  * Need information about timing requirements of EC_WAKEUP signal.

  * Missing infrastructure for processor wakeup via SCI interrupts -
    wireless, ebook mode sensors, etc.

* Finalized B3 mechanical changes
  - Keyboard lights are gone, among other things

* Get all the hardware working - what's left?
    - Possible issue with high speed SD, but not critical

* Cambridge is going to add serial connectors to 10 B2 systems for the
  use of people doing power management.  If you want to do power work
  and don't have a serial B2, email smithbone at gmail.com

* Tinderbox/power measurement - Richard will get onto it when he
  stops fighting other fires.

* Backup - Wad is working this.

TamTam:
   * Performance when run as user is sluggish
   + More investigation needed.  Suggestions offered.

Ronak:
   * Wireless performance measured at 2 mbps "open air"
   * Marvell still needs more clarification of CPU-triggered wakeup
   * Ronak will schedule another conference call
   * Marvell has also seen the "WLAN doesn't enumerate after reset" problem

Old action items:

AI: Blizzard to bug Marcelo to try latest TCP
Moot - iperf problem is solved

AI: Andres - look into what it would take to uncache USB descriptors.
In progress

AI: MLJ - get new schematics and .brd file from Quanta.
Done - 0110 is the latest

AI: Jordan and ajax to investigate xrandr segfault
Written up

AI: dcbw and marcopg to investigate mesh/avahi not working.
      done, but we're seeing crashing reports
    - No new status this week

AI: #601: cjb to check whether we have pango extents fix in
Done