Re: [RFC] [PATCH] Device Tree on ARM platform

[Peter Korsgaard <jacmet@sunsite.dk>]

>>>>> "Robert" == Robert Schwebel <r.schwebel@pengutronix.de> writes:

Hi,

 Robert> - The whole concept is based on the assumption that bindings
 Robert> are defined *once*, then never to be changed again. As this
 Robert> is not true (check MPC5200 to find out what I mean), oftree
 Robert> wreckage is *the* main cause of new kernels not working on
 Robert> old bootloaders any more. Is there a solution of this
 Robert> problem? I have not seen a good idea how to avoid the
 Robert> constant change in definitions.

Just bundle the .dtb with the kernel and they'll always be in sync. I
know this isn't really in the spirit of OF, but currently imho the
only realistic solution.

I did a (nacked) patch to do this with the multi-image support in
U-Boot some time ago:

http://patchwork.ozlabs.org/patch/589/

 Robert> - The oftree layering is fundamentally broken. We already
 Robert> have a sane abstraction for arbitrary hardware in the kernel:
 Robert> platform devices.  Why not instanciate platform devices from
 Robert> a generic oftree core?

 Robert> - Platform data makes it possible to store function
 Robert> pointers. There is no equivalent to this concept in
 Robert> oftree-land.

Yeah.

-- 
Bye, Peter Korsgaard