* Jan Weitzel [off-list ref] [110715 02:42]:

Am Freitag, den 15.07.2011, 00:50 -0700 schrieb Tony Lindgren:

quoted

Well we can add it even before device tree support if it makes
sense from code coalescing point of view. In this case it would
mean creating board-panda-common.c or similar so the code can
be shared amongst the panda variants.

It seems that some GPIO pins are different and there are some
difference in devices connected, but big parts of the code can be
shared.

Adding the pending boards without combining code would be adding code
that will be going away for most part with the device tree support.

And most likely the beagle and panda based boards will be the first
ones to work with device tree. So anything we can do to have common
board init code will also help with this effort.

Regards,

Tony

pcm049 and panda board have some more different devices.
I am working on patches to add NAND support and using tlv320aic3x audio
codec which need regulators in platformcode. I need a hack to controll
usb otg by gpio (controlling a external power circuit). The patches are
not mainline ready by now. 

OK, let's figure out how we can add the basic support first then.

If using machine_is_xxxx() in board-omap4panda.c then also is good way,
I could provide our board as patches on top of board-omap4panda.c.
When will pandaboard use device tree to boot?

Yes, but it would better just to use separate .init_machine entries to
initialize thing. Gpio entries can be initialized along the lines of the
recent beagle commit 5fe8b4c19dc24e3bb873daf9e96a2439a83bbd79 that adds
support for beagl xm board. Except you already have a separate machine_id,
so no runtime detection is needed.

That avoids adding the machine_is entries all over the place that tend
to cause problems when some other panda variant is added as it requires
patching in multiple places.

Regards,

Tony