Re: [PATCH 1/1] ARM: dt: add header to define tegra20 clocks

[Stephen Warren <swarren-3lzwWm7+Weoh9ZMKESR00Q@public.gmane.org>]

On 02/18/2013 10:31 PM, Shawn Guo wrote:
> On Thu, Feb 14, 2013 at 10:54:28AM -0700, Stephen Warren wrote:
>> On 02/13/2013 11:38 PM, Hiroshi Doyu wrote:
>>> To replace magic number in "clocks = <&tegra_car 28>;"
>>
>> I like the concept here; I was thinking about doing this today, but you
>> beat me to it:-) Feel free to create the Tegra30 header too, and modify
>> all the *.dts* files.
>>
>> To address other comments in this thread: Yes, I think that we will want
>> to modify the clock driver to include this header to avoid
>> duplication/errors (that will require adjusting Linux's include path),
>> and also remove the list of IDs from the binding document; it can just
>> refer the the new header by name and cause the header to *be* part of
>> the binding document.
> 
> I like it, since doing so will help us hijack kernel to have dts stay
> in the tree rather than going to a separate repository :)

Well, at least the header files would need to stay in the kernel. The
.txt binding documentation wouldn't have to though.

I guess this is a good argument for not putting these header files into
Documentation/devicetree/bindings, and adding that to the include path,
but rather putting the headers somewhere else.

> Seriously, is maintaining dts in a separate repository still a plan?
> If yes, we will have duplication problem someday anyway.
> 
> For reason of it, I vote for having dts stay in the kernel tree.
> 
> Shawn
>