[pkg-fso-maint] d-i installation on embedded devices (was Re: MTD Installation)

Wed Mar 25 12:00:48 UTC 2009

Hi there!

On Wed, 25 Mar 2009 11:45:09 +0100, Per Andersson wrote:
> (Cc:ing debian-boot.)

Done, but please keep the pkg-fso-maint@ mailing list cc:ed as well,
TIA.  Doing so, there is no need to cc: me.  I also cc:ed the
debian-embedded@ mailing list, people from there should be interested as
well.

I changed the title to something more appropriate.

> 2009/3/25 Joachim Breitner <nomeata at debian.org>:
>>> Maybe some other thread discusses this already,
>>> but why don't you use d-i for installation?
>>
>> Simplicity. We wanted to get started as fast as possible. But given the
>> ever increasing complexity of install.sh, this should be the goal, yet.
>>
>>> Could porting d-i to GTA02 (and GTA01 but I don't
>>> have one of those) be worth doing as a GSoC project?

For anyone working on/for the pkg-fso, thanks to Arne Anka [1] I can
borrow/offer a GTA01 (if needed the full development kit, i.e. with a
debug board as well).  Since I was in contact with people interested to
sell their GTA01, I can also find more than one device, if needed.  This
is *only* for Debian-related work, in any field (e.g. X11, ALSA, GPS,
etc.) and not only for d-i.

>> Quite possible – please raise that idea on the pkg-fso-maint list. Some
>> others have expressed interested in trying to port d-i to the FreeRunner
>> (Steffen Moeller, IIRC). He’s subscribed to the list.

I am interested as well, and this is one of the first item in my pkg-fso
ToDo list, it seems that other stuff (kernel, bug fixing and so on) is
keeping me away from it.

>> Do you know any d-i knowledgeable person that has a FreeRunner and would
>> be able to mentor this project?
>
> @debian-boot: Anyone up for this?

NB, I have no knowledge at all WRT how the GSoC work, but I am not sure
we can still change the different projects we submitted.

In any case, I think that the MTD project should still have an higher
priority, because:

- its use is wider (e.g. for NAS devices)

- it involves more people (not only the pkg-fso crew), thus it is
  probably easier to have advices

- ATM no official Debian kernel can be used on the FreeRunner and I
  really do not have ETA about when one will be available (Openmoko
  has now switched to 2.6.29 and the last time I checked for IIRC 2.6.26
  they had something like more than 100 patches)

> I have a FreeRunner myself.
>
> Without having done any research at all yet, I think
> that similar infrastructure to how Debian Installer is
> booted and installed on NAS devices could be used.
> The bootstrapping part seems very similar to the
> GSoC project I did last year, since the FreeRunner
> uses U-Boot as well. (Of course it would have to be
> made to work with the Qi bootloader as well which
> widens the scope.)

Disclaimer: everything below is my personal opinion.

I think we clearly need to have a plan, divided into two phases:

1) installation into the microSD

   a. booting d-i on the FR, which is an easy task given that we must
      use the pkg-fso kernel package [2].

   b. starting the graphical installer with a virtual keyboard, since
      there is no physical keyboard, thus the textual installer cannot
      be used.

      After my talk at last FOSDEM'09 [3], I talked with Jeremy Bobbio about
      that: from a quick view he had at d-i, this should not be so
      difficult either, given that d-i already uses the framebuffer.

   c. installing everything on the microSD, which means that d-i should
      not use the microSD at all, thus it should probably be run from a
      ramdisk.

   d. deafening some tasks for tasksel and then installing everything.

   e. modifying the boot loader.

      On this point, I am a strong opponent of Qi [4] (the now default
      bootloader used by Openmoko), because it lacks at least one very
      important feature U-Boot has: a menu.  This is why I think that
      any effort should firstly target U-Boot and then Qi, which AFAIK
      is ATM available only for the Openmoko.

2) installation into the MTD

   Given that everything above will be already done, we will have only
   two showstopper:

   a. MTD support in d-i.

      This is also why I consider implementing MTD in d-i of an higher
      priority than adapting d-i to the FR.

   b. disk space.

      The FR (GTA02) has 256MB of flash memory, while the 1973 (GTA01)
      only 64MB [5].  In both case, there is no way we can fit a normal
      Debian system on it, especially considering that we should target
      the FreeSmartphone.Org initiative [6], which ATM relies on Python.

      The solution in this case is EmDebian.

Given the above, if it was not clear yet, I have a clear vision of what
everything should move.  Not that this vision is the correct one, but I
think it makes sense.  Obviously, I am available to discuss it: I am not
looking for glory, on the contrary I would like that we progress by
small and well-defined steps :-)

Thx, bye,
Gismo / Luca

Footnotes: 
[1] http://lists.linuxtogo.org/pipermail/smartphones-userland/2009-February/001173.html
[2] http://git.debian.org/?p=pkg-fso/linux-2.6-openmoko.git;a=summary
[3] http://www.fosdem.org/2009/schedule/events/debian_openmoko
    http://people.debian.org/~gismo/talks/20090207_FOSDEM09_DebianOpenmokoFreeRunner/
[4] http://wiki.openmoko.org/wiki/Qi
[5] http://wiki.openmoko.org/wiki/Neo1973:_GTA01Bv4_versus_GTA02_comparison
[6] http://www.freesmartphone.org/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 314 bytes
Desc: not available
Url : http://lists.alioth.debian.org/pipermail/pkg-fso-maint/attachments/20090325/6e76798a/attachment.pgp 