[OLPC-devel] Thoughts on S3 resume.

[Chris Ball]

>> On Mon, 07 Aug 2006 21:54:24 -0400, Jim Gettys <jg at laptop.org> said: 

   > The measurement that is interesting is when the kernel is executing
   > off of the internal flash, and nothing plugged into USB.  I'd like
   > to have two numbers: with and without the USB stack loaded (I know,
   > it will be painful to get the second number until the PS/2
   > interface is easy to use).

I think I have a setup able to measure the second number (using a
ramdisk), but the results are still poor so far.  I:

   * put busybox in the developer image initrd
   * told the initrd to run a busybox shell instead of switchrooting
   * wrote a simple program (timestamp) to output the result of 
     gettimeofday(2), and put it in the initrd.
   * booted my initrd from usb, with a serial console, then unplugged
     the usb stick and hub.
   * rmmod until no modules loaded (which included usb_storage,
     ehci-hcd, uhci-hcd, scsi_mod, sd_mod, ext3, jbd)
   * set an alarm, and echo -n "mem" > /sys/power/state && timestamp
 
The resulting serial console log file is attached, with the timestamp
output as the last line.  It still takes 19 seconds from the board
drawing full power again to my userspace code running; this time we at
least have kernel BUG()s to go on, though.

Anyone have advice on how to proceed, especially with regards to the
BUG()s?  I plan to build AMD_IDE (amd74xx.c) as a module rather than
compiled in, so that I can unload that too, and add CONFIG_PRINTK_TIME
in case it shows that a particular driver's resume path is responsible.
I'll be able to do those tomorrow.

- Chris.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: olpc-s3-log
URL: <http://lists.laptop.org/pipermail/devel/attachments/20060810/51950a68/attachment.ksh>
-------------- next part --------------
-- 
Chris Ball   <cjb at mrao.cam.ac.uk>    <http://blog.printf.net/>