[Arm-netbook] CT-PC89E kernel 2.6.24.7 source investigation

[Luke Kenneth Casson Leighton]

On Fri, Mar 5, 2010 at 4:44 PM, Frans Pop <elendil at planet.nl> wrote:
> On Thursday 04 March 2010, Luke Kenneth Casson Leighton wrote:
>> 1) frans, i worked out a better (easier) way, which is very very
>> simple: move or remove datang-epc.tar.gz from the firmware flashing SD
>> card, _but_ place a replacement kernel on it, zImage_dt.  press-hold
>> both mouse-buttons; u-boot and then the (new) kernel is copied to the
>> NAND flash, and then the search for datang-epc.tar.gz fails, and
>> everything stops.  voila.  reboot.  new kernel runs.  repeat.
>
> Hmm, but earlier you said very convincingly that partial firmware upgrades
> were not possible... That was exactly what I was thinking of. :-)

 ho hum, well it was actually a mistake on my part that i found out
that it worked: instead of naming the tarball datang-epc.tar.gz, i had
accidentally just copied rootfs.small.tgz onto the SDcard.  it was 3
or 4 successful kernel-compile cycles before i noticed the mistake,
and the fact that the kernels _were_ being successfully flashed onto
the device :)


>> 2) i noticed the tell-tale sign "S3C_LCD got initialized: 180mhz" and
>> started looking on google for this.  up came the SmartQ5 MID device.
>> downloaded the source code which was a 2.6.24.7 kernel aaaand bingo,
>> the subdirectory drivers/video/s3c/ is missing from 2.6.33 stable.
>> so, that's what we need.
>
> I assume you mean:
> http://gitorious.org/mer-smartq/mer-smartq-kernel
> And in particular commit 9b25305 "Apply samsung kernel patch".

 i ended up at http://zenvoid.org/software/qi-smartq/

> I've added that repo as a remote branch to my local git checkout of the
> upstream kernel and taken a look at it. It's your typical completely
> clueless commit from hardware vendors.

 hurrah! :)

> It's an aggregated patch including
> fixes and new hardware support and even two binary files. But probably
> they were given it that way by Samsung.
>
> But it does indeed look as if ChiTech have used that same patch as their
> basis (only on top of .24.2, not .24.7; the included changelog suggests
> it's intended for .24.2).
>
> AFAICT the video support added there in that video/s3c directory is an
> extension to the s3c-fb support already present in the standard .24
> and .33 kernels (and which I had enabled).

 ok.  the video/s3c bit is the LCD initialisation sequences, without
which you're hosed.

 i.e. these:

+       A drivers/video/s3c/s3cfb_lts222qv.c
+       A drivers/video/s3c/s3cfb_ltv350qv.c
+       A drivers/video/s3c/s3cfb_lte480wv.c

 it's one of _these_ that is so important to have, as it sets the
resolution (1024x600), and does the proper power-up / register setting
sequence (if needed)

> I've uploaded the patch (without the binary blobs) and a diffstat here:
> http://people.debian.org/~fjp/tmp/CT-PC89E/samsung_2.6.24.2.patch.gz

 excellent, thank you.

> I also had a look around and found the following possibly useful links:
> http://www.mail-archive.com/linux-samsung-soc@vger.kernel.org/index.html
> http://git.fluff.org/gitweb?p=bjdooks/linux.git;a=summary
>
> Ben Dooks seems to be the upstream coordinator for Samsung stuff, but
> possibly more intersting, Mark Brown (DD!) is a frequent contributor.
>
> How about asking either of them for suggestions?

 good call.