[OLPC-devel] Minutes of 7/11 software teleconference.

[Jim Gettys]

OLPC Software Call 07/11/2006  Minutes by Chris Blizzard, edited by Jim
Gettys.

* EXA driver status from Jordan Crouse
- new EXA driver is working pretty well
- Adam Jackson ajax is packaging it
- ACTION: get it packaged into our images
   - it's in review, almost done
   - will want people to start using it ASAP to get time on it.
- status of Xv in this driver is unknown, though the XAA 
	driver has known problems:
   - problems with offset scaling?
- need to do some testing with codecs, etc
- theora? for introductory video: might not be fast enough
- 16 or 24 bits for depth of the driver?
   - 16 for memory?
   - use 4444 for frame buffer organization instead of 565
   - GX lacks bit depth translation support
   - interesting question: how fast could the Geode do 
     this transformation in software?  probably not fast enough.

* status from david woodhouse
- jffs2 optimizations left - but nothing major
- everything else is upstream  in Linus' tree (mtd, jffs2 changes)
- playing with the smbus driver
- olpc development kernel git tree has been set up
   - pulls from kernel + libertas + geode tree
- analog changes for ALSA control also already upstream.
  - need to do analog in properly in the framework.
- discussion of generic user space API for power management
   - want rough consensus before proceeding: discussions will be at OLS
     next week.
   - could just emulate an existing user space api (e.g. APM/ACPI)
- discussion of new way to hook up to the flash driver
   - improved flash controller: a variant of an IDE chip that can 
     talk to the flash quickly
   - could be months, could be shorter - not sure yet
   - discussions of various ways we could interface with the flash chip
- discussion of DCON
   - mark is meeting with various partners about the DCON
   - should know the schedule in the next 2 days or so
   - smbus, frame buffer integration a little later
   - have smbus interface to talk to the registers in debug and 
      then script to set registers
   - need interrupt handler for geode <-> dcon handoff of video
   - discussion of battery and a temp sensor in the battery
     - if we put a temp sensor in the battery we can determine when 
we've charged without damaging the batteries - maybe get a lot more 
cycles but still doing testing
   - might be an eprom on the memory for the size?
     - not sure there is any way to tell how much memory size 
       is in the machine

* discussion of quanta manufacturer information
   - 52 + 54 smbus BIOS addresses
   - date of manufacture and serial number
   - the bios util needs to be able to preserve the serial number and 
manufacture date in the flash (see mail from ray.tseng subject 'OLPC 
"bootloader"')
   - can we preserve?  need to know if the non-volatile settings 
     are 128k or 4k
   - linuxbios does need to support it - need to tell ron et al
     - same with keyboard
     - the EC flash needs to be non-volatile as well?  
       (same with the serial/date?)
     - see this stuff on most machines: "lshal| grep -i smbios"

* discussion of kexec() issues
   - kernel panic in the nand driver with a very strange, very 
     high address?
   - dave woodhouse will look at the email and follow up

* quanta documents
   - linuxbios checklist incl smbios question
   - discussion of acpi - to do or not to do
     - not using it in the conventional way
     - mjg has apparently suspended and resumed it from a PS/2 keyboard
   - looking for help as to how to do maufacturer tests
   - stress test program does exist
   - need manufacturer testing, not heat and stress testing
   - jg will send those out

* can we do chainload if linux as bootloader is corrupted?
   - jg will check with ron about that

* flash loader needs work, clearly
   - preserve smbios info if nothing else

* VGA - no VESA - always fbdev.  This avoids the VSA VESA emulation
  IP problems as well.
* driver support - audio driver is in better shape
   - we know how to turn off the audio port from the EC
   - are they in the driver?  need to set both bits
     - AD1888 - maintenance register that must be set/cleared that
       disables the high pass filter on the input side
     - GPIO bit through the EC that control the capacitive coupling
        into  the AD1888 - ray got some code that lets you set any 
        GPIO register
       - Quanta has asked the manufacturer of the EC to remove the
         "confidential" stamps on the EC docs.
     - jaya suggested a mixer flag in alsa 
      (already upstream in alsa now).
     - jg suggests when the device closes turn off those bits
        automatically?
   - need to get test-a schematics to jordan (done)

* UI reflash
- right now a text based UI for determining what to do
- part of the kernel boot
- has to be available on the console, not just serial device
- need to hold down magic keys to get the option
- images for boot is good - use compressed bitmap for size: gzip
  does a good job on synthetic images.  Would avoid I18N problem
  entirely, and be "kid friendly"
- need to look at this carefully; may be easier than dealing with N 
  translations, where N becomes huge.

-- 
Jim Gettys
One Laptop Per Child

-------------- next part --------------
A non-text attachment was scrubbed...
Name: Image_103_.jpg
Type: image/jpeg
Size: 9241 bytes
Desc: not available
URL: <http://lists.laptop.org/pipermail/devel/attachments/20060711/98cee4a8/attachment.jpg>