[OLPC-devel] BIOS Meeting minutes - 2006-08-09

[Mitch Bradley]

BIOS meeting minutes. 2006-08-09

Present: Ray, Richard, Jim, Mitch (scribe)

Loading FLASH:

Quanta has monitor servers that will control image servers

Recommendation for B-test - use USB key to load FLASH

Jim suggests loading Linux from the USB stick, then to mount the diagnostics
from the server.  The goal is to load the NAND FLASH once, not twice.

Ray suggests using a USB stick to load a diag image into NAND FLASH.  
The diag image
would run for some time without a LAN connected, then finally connect to 
the LAN
and load the final image from the server, using a driver in the diag 
image.  The
diagnostic could run for 8 to 20 hours.

Jim suggests including the diagnostics into the full software load, then 
delete
it later.

Jim wants to stress the machine close to its thermal limits for 
functional test, but
not necessary for manufacturing test.

Loading NAND FLASH twice might not be a problem if the diag image is small.

In the SPI FLASH, the manufacturing info will be in the EC section.

Physical connection between EC and 1-wire battery gauge is okay.  
Schedule for finishing
EC stuff is end of August.  Ray to ask hardware team when can we get 
battery samples,
so we can write and test battery driver.

Whither to use ACPI interface for battery info?  - Will put battery info 
in EC RAM
so Linux can read it more or less directly.

Question of whether to define a new Linux abstraction for accessing 
battery info?
Dave Woodhouse has been tasked to see if we can lead the way.

Mechanism for reporting switch status such as cover-closed switch?  ACPI?
Mono/color key can be reported as a keyboard scan code.
Answer: Report lid switch as an ACPI event, mono/color as a keyboard event.

Suspend/Resume stuff needs to be done at high priority.

Richard has made progress on SPI programming.  About to hold breath and pull
the trigger - tomorrow night or Friday.  Richard has 2 boards.  The intent
is to write LinuxBIOS at offset 64K, skipping the EC part.

Discussion of how to write protect the SPI chip.  Ray has defined a 
software interface
to control the SPIWP line by doing accesses via I/O ports 68 and 6c 
(keyboard
ports).  But we also need a "human lock" - the user must hold down a key or
something, to block BIOS-rewriting worms.

Discussion of what to include in LinuxBIOS for B-test?  Graphical 
screens or not?

Dave Woodhouse and Dave Zeuthen both have images that will boot from
NAND FLASH.  The installation has not been automated - must convert
ext3 image to JFFS2.  Jim will try to get instructions into the wiki 
tomorrow.