Re: [PATCH v1 2/6] dt-bindings: mediatek,mt6779-keypad: use unevaluatedProperties

[Mattijs Korpershoek <mkorpershoek@baylibre.com>]

On Thu, Jul 21, 2022 at 11:16, Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> wrote:

> On 21/07/2022 11:06, Mattijs Korpershoek wrote:
>> On Wed, Jul 20, 2022 at 19:14, Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> wrote:
>> 
>>> On 20/07/2022 16:48, Mattijs Korpershoek wrote:
>>>> writing-bindings.rst states:
>>>>> - If schema includes other schema (e.g. /schemas/i2c/i2c-controller.yaml) use
>>>>>   "unevaluatedProperties:false". In other cases, usually use
>>>>>   "additionalProperties:false".
>>>>
>>>> mt6779-keypad includes matrix-keymap.yaml so replace additionalProperties:false
>>>> by unevaluatedProperties:false.
>>>
>>> This is not sufficient explanation. You now allow all properties from
>>> matrix-keymap.yaml, which might be desired or might be not (e.g. they
>>> are not valid for this device). Please investigate it and mention the
>>> outcome.
>> 
>> Hi Krzysztof,
>> 
>> Thank you for your prompt review.
>> 
>> In mt6779_keypad_pdrv_probe(), we call
>> * matrix_keypad_parse_properties() which requires keypad,num-rows and keypad,num-cols.
>> * matrix_keypad_build_keymap() which uses linux,keymap
>> 
>> Therefore, all properties from matrix-keymap.yaml are
>> required by the mt6779-keypad 
> Better to mention the device, not driver.

I mixed up driver versus device (hardware). Sorry about that.

For successful key detection, the hardware (called MediaTek keypad) 
requires that we program rows/columns via the KP_SEL register.
So num-rows and num-cols are valid properties for this device.

The MediaTek keypad has a set of bits representing keys, from KEY0 to KEY77. 
These keys are organized in a 8x8 hardware matrix.
Therefore, linux,keymap is also a valid property for this device.
>
>> 
>> In v2, I will add the above justification and also add all 3 properties
>> in the "required" list.
>> 
>> Initially, I did not do this because from a dts/code perspective it seemed
>> interesting to split out SoC specific keyboard node vs board specific key configuration:
>> * [PATCH v1 5/6] arm64: dts: mediatek: mt8183: add keyboard node # SoC specific
>> * [PATCH v1 6/6] arm64: dts: mediatek: mt8183-pumpkin: add keypad support # board specific
>> 
>> What would be the recommend approach for above?
>> I see at least 2:
>> * "move the whole keyboard node into the board file (mt8183-pumpkin.dts)" even if it generates
>>   duplication between boards using the same SoC.
>> * "add a "dummy keymap,row,cols" properties in the soc node which can be overriden in board file.
>>   For example, use rows and cols = 0 which would have the driver early exit.
>> 
> SoC DTSI should have only SoC properties. The keyboard module is part of
> SoC. The keys and how it is wired to them - not.

Indeed. So the split I send in v1 is "valid", from a device(hardware)
point of view.
In that case i'll not make the properties from matrix-keymap.yaml
*required* in v2.

Thanks again for your feedback.

Mattijs

>
> Best regards,
> Krzysztof