[RFC] Qi Bootmenu

Marc Andre Tanner openmoko at brain-dump.org
Wed Nov 18 13:10:29 CET 2009


On Thu, Oct 08, 2009 at 10:50:00PM +0200, Marc Andre Tanner wrote:
> Proof of Concept
> ================

Some updates on what I've done so far:

 - ecore-con is no longer needed (It can be disabled during ./configure,
   patch is in enlightenment svn)

 - uclibc can be built with a minimal set of locale data (patch is in
   upstream git)

 - I streamlined freetype's configuration by disabling everything which
   didn't seem necessary

 - integrated openmoko kernel build, the om-gta02-2.6.31 kernel branch 
   is used

 - added some init script which should set up usb networking and start
   dropbear

 - updated to latest FWL toolchain version, unfortunately the release
   tarballs seem to have a problem I therefore uploaded the toolchain
   which I' ve built and which seems to work here. For further 
   information see the my post to the FWL mailing list at:

    http://lists.impactlinux.com/pipermail/firmware-impactlinux.com/2009-November/000443.html

My current build scripts can be found at:

  git://repo.or.cz/qi-bootmenu-system.git

It would be nice if someone could try it out by following the instructions
in the README and reporting back the results.

I also started a bootmenu application which scans the available partitions
for bootable images.

  git://repo.or.cz/qi-bootmenu.git 

It doesn't yet include a GUI but Dave Ball started a prototype based on
evas which is basically option #2 from raster.

Things which I need to do: 

 - kexec doesn't have uImage support so either:

    (a) implement it
    (b) convince distros to also provide zImages 
    (c) strip the uImage header off at runtime with dd but this would 
        make it slower of course

 - there seems to be a segfault in the color conversation code from evas
   triggered by compiler optimizations (-O0 works everything else doesn't)
   and the elementary dialog application. I have yet to check if this
   also happens with the evas based GUI of qi-bootmenu.

 - kernel config needs to be trimed down to the absolute minimum, don't
   know whether it makes sense to look into gta02_micro_defconfig or start
   from scratch.

 - Qi's boot logic needs to be patched to load a custom kernel up on AUX
   press instead of skipping one.
 
 - actually build a decent GUI for qi-bootmenu
 
 - Try random things to make it fast(er)...

    - updating to 2.6.32 or backporting devtmpfs comes to mind because
      it would make the call to mdev -s unecessary.

    - ...

As always comments are appreciated.

Thanks,
Marc
 
-- 
 Marc Andre Tanner >< http://www.brain-dump.org/ >< GPG key: CF7D56C0



More information about the devel mailing list