Re: [PATCH v2] ASoC: samsung: Mark unused Odroid compatibles as deprecated

[Krzysztof Kozlowski]

On Mon, Mar 19, 2018 at 4:14 PM, Sylwester Nawrocki
<s.nawrocki@xxxxxxxxxxx> wrote:
> On 03/19/2018 11:56 AM, Krzysztof Kozlowski wrote:
>> On Mon, Mar 19, 2018 at 11:29 AM, Sylwester Nawrocki
>> <s.nawrocki@xxxxxxxxxxx> wrote:
>>> On 03/18/2018 04:35 PM, Krzysztof Kozlowski wrote:
>>>> Compatible for XU4 audio is not being used.  Instead the board uses the
>>>> same compatible as XU3.  The devices are now just compatible so they
>>>> should use the same value.  Mark "hardkernel,odroid-xu4-audio" as being
>>>> deprecated so in this future could be removed to limit useless
>>>> properties.
>>>
>>> It doesn't feel right to obsolete the "hardkernel,odroid-xu4-audio"
>>> compatible, there is significant difference between XU3 and XU4 - there
>>> is no audio CODEC on XU4, this board only supports audio over HDMI interface.
>>> XU4 could be compatible with XU3, but not the other way around.
>>> It just happens we have other DT properties that help to handle such HW
>>> design difference.
>>
>> The compatible does not describe physical differences. It does not
>> mean that devices are the same. In this case they are just coming from
>> the same family and they operate the same, from the bindings
>> perspective.
>
> From the ePAPR 'compatible' string definition you cited, the compatible
> string is supposed to indicate programming model of a device, for the purpose
> of matching a driver.

Yes, you're correct, it refers to programming model. Although later
you will find second explanation (chapter 4): "The compatible property
of a device node describes the specific binding (or bindings) to which
the node complies.".

> I thought the programming model refers to the driver's
> SW interfaces used to control the hardware, rather than only to a particular
> DT binding design. And XU4 is not compatible with XU3 from device programming
> perspective.

The programming models of XU4 and XU3 audio components, to which we
refer now and which are implemented/used, are the same. I mean not the
same in general, but how we use them. The used subset of each is the
same. Therefore the binding and the driver do not distinguish any
differences (like codec).

>> The XU4 binding is not being used. Adding a compatible which is not
>> used in the moment of adding is a proof that this compatible is not
>> needed. It is just a duplicate. There is no point of adding
>> duplicates.
>
> I disagree it is just an unnecessary duplicate, I think dts for XU4 could
> fixed instead of dropping that compatible from the binding.

You know, there is nothing to fix - changing the compatible to XU4
will not change anything. The executed code will be exactly the
same...

>>> Moreover, only XU4 is still in production and should be in few more years [1],
>>> others are obsoleted now.
>>
>> It is not a problem. Whether device is manufactured or not, does not
>> reflect what bindings we are using. Deprecated XU4 compatible does not
>> mean that XU4 itself is deprecated. Just this compatible should not be
>> used for new DTS.
>
>>> So I think we should keep at least these 2 compatible strings:
>>>
>>> - "hardkernel,odroid-xu3-audio" - for boards with audio CODEC,
>>> - "hardkernel,odroid-xu4-audio" - for boards without audio CODEC,
>>>    supporting only HDMI interface.
>>
>> Yeah, and then we inflate this list into X, X2, U3, HC1 and all others
>> which are the same. And then we should add XU3-lite (it is different
>> device). This goes to some nonsense. Compatible is not for each device
>> but for family even though there are differences between specific
>> devices.
>
> You are not listening, I refer only to major audio subsystem differences.
> It would have been:
>
>  - "hardkernel,odroid-xu3-audio" for: U2, U3, X, X2, XU, XU3, XU3-Lite
>  - "hardkernel,odroid-xu4-audio" for: XU4
>
> But if you insist on only one compatible I'm not going to argue further,
> you will be responsible for this. :)

Ah, I read too fast and missed that point. Actually that is nice
consensus in this case.

Best regards,
Krzysztof