Re: [PATCH 1/1] net/macb: add DT support

On Fri, Dec 02, 2011 at 04:30:36PM +0100, Nicolas Ferre wrote:

On 11/20/2011 06:11 PM, Jamie Iles :

quoted

On Sun, Nov 20, 2011 at 05:47:40PM +0100, Jean-Christophe PLAGNIOL-VILLARD wrote:

quoted

On 15:58 Fri 18 Nov     , Jamie Iles wrote:

[...]

quoted

quoted

quoted

quoted

+		compatible = "atmel,macb";

This should be "cdns,macb" as it isn't Atmel specific.  I believe cdns
is the correct stock ticker symbol for Cadence.

here I put "atmel,macb" on purpose to specify the difference of the IP between
the soc, in fact it should have been atmel-at91,macb

Well if we really can't detect the difference from the revision register
then we should have "cdns,macb" *and* "atmel,at91-macb" at least then
where platforms could claim compatibility as:

compatible = "atmel,at91-macb", "cdns,macb";

re-thinking about this I propose that we go for the following
compatible string for macb:

- compatible: Should be "cdns,<chip>-macb"

And as the first SoC that have embedded an emacb that is compatible
with current 10/100 AT91 usage is AVR32 at32ap7000... We may end up
with "cdns,at32ap7000-macb" compatible string. The first ones with
different synthesis parameters where at91sam9260/3 so I may also
add:
"cdns,at91sam9260-macb".
Then you will have to add the first SoC that uses the gigabit
version of the macb...
What do you think about that?

Sure, that works for me, though I guess this is a much more general 
thing than this one binding, but that does make sense to me.  I think 
that keeping a general "cdns,macb" _too_ still makes sense though as 
lots of it may well be detectable and it will probably be difficult for 
one SoC vendor to know whether their IP instantiation really is the same 
as another vendors...  Either way I don't have a strong opinion on that.

BTW, "cdns" seems not included in the vendor-prefixes.txt file yet...

No, that one is missing.  If you want to add it then feel free, if not 
I'll add it to my list of patches to do!

Jamie