[PATCH] arm: dts: add initial support for TBS2910 Matrix ARM mini PC

[Sebastian Hesselbarth]

On 21.10.2014 16:17, Sascha Hauer wrote:
> On Tue, Oct 21, 2014 at 02:03:51PM +0200, Sebastian Hesselbarth wrote:
>> On 21.10.2014 13:57, Sören Moch wrote:
>>>>> +/ {
>>>>> +    model = "TBS2910 Matrix ARM mini PC";
>>>>> +    compatible = "tbs,imx6q-tbs2910", "fsl,imx6q";
>>>>
>>>> [...]
>>>> Also, board compatible should only be "tbs,tbs2910".
>>>
>>> Similar boards use e.g.
>>>      compatible = "fsl,imx6q-sabresd", "fsl,imx6q";
>>>      compatible = "udoo,imx6q-udoo", "fsl,imx6q";
>>>      compatible = "wand,imx6q-wandboard", "fsl,imx6q";
>>>
>>> So is it really better to use only "tbs,tbs2910" here?
>>
>> Hmm, having the SoC name again in the board compatible seems
>> odd to me, e.g. we have "google,chromecast" without the SoC
>> name in it.
>
> No. My vanilla 3.17 kernel has this:
>
> arch/arm/boot/dts/berlin2cd-google-chromecast.dts:17:   compatible = "google,chromecast", "marvell,berlin2cd", "marvell,berlin";

Right, I didn't question that there should be "fsl,imx6q" at the end
of the array of compatibles. But there is no need for another "imx6q"
in the board name, e.g. "tbs,tbs2910" instead of "tbs,imx6q-tbs2910".

>> Anyway, it is just a compatible and if it is common for imx-
>> based boards or Shawn/Sascha like to have it this way you
>> should stick with that, of course.
>
> It has to be that way, because the kernel matches to "marvell,berlin",
> not the board name. Otherwise we would have to compile in all board
> compatible strings into the kernel or the kernel would refuse to start.

Yup. No doubt, it was just about imx6q in the board's compatible.

Sebastian