Re: [PATCH 1/2] MIPS: dts: correct gpio-keys names and properties

[Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>]

On 25/06/2022 22:15, Paul Cercueil wrote:
> Hi Krzysztof,
> 
> Le sam., juin 25 2022 at 21:58:08 +0200, Krzysztof Kozlowski 
> <krzysztof.kozlowski@linaro.org> a écrit :
>> On 24/06/2022 20:40, Paul Cercueil wrote:
>>>  Hi Krzysztof,
>>>
>>>  Le ven., juin 24 2022 at 19:07:39 +0200, Krzysztof Kozlowski
>>>  <krzysztof.kozlowski@linaro.org> a écrit :
>>>>  gpio-keys children do not use unit addresses.
>>>>
>>>>  Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
>>>>
>>>>  ---
>>>>
>>>>  See:
>>>>  
>>>> https://lore.kernel.org/all/20220616005224.18391-1-krzysztof.kozlowski@linaro.org/
>>>>  ---
>>>>   arch/mips/boot/dts/img/pistachio_marduk.dts   |  4 +--
>>>>   arch/mips/boot/dts/ingenic/gcw0.dts           | 31
>>>>  +++++++++----------
>>>>   arch/mips/boot/dts/ingenic/rs90.dts           | 18 +++++------
>>>>   arch/mips/boot/dts/pic32/pic32mzda_sk.dts     |  9 ++----
>>>>   .../boot/dts/qca/ar9132_tl_wr1043nd_v1.dts    |  6 ++--
>>>>   arch/mips/boot/dts/qca/ar9331_dpt_module.dts  |  4 +--
>>>>   .../mips/boot/dts/qca/ar9331_dragino_ms14.dts |  6 ++--
>>>>   arch/mips/boot/dts/qca/ar9331_omega.dts       |  4 +--
>>>>   .../qca/ar9331_openembed_som9331_board.dts    |  4 +--
>>>>   arch/mips/boot/dts/qca/ar9331_tl_mr3020.dts   |  8 ++---
>>>>   10 files changed, 37 insertions(+), 57 deletions(-)
>>>>
>>>>  diff --git a/arch/mips/boot/dts/img/pistachio_marduk.dts
>>>>  b/arch/mips/boot/dts/img/pistachio_marduk.dts
>>>>  index a8708783f04b..a8da2f992b1a 100644
>>>>  --- a/arch/mips/boot/dts/img/pistachio_marduk.dts
>>>>  +++ b/arch/mips/boot/dts/img/pistachio_marduk.dts
>>>>  @@ -59,12 +59,12 @@ led-1 {
>>>>
>>>>   	keys {
>>>>   		compatible = "gpio-keys";
>>>>  -		button@1 {
>>>>  +		button-1 {
>>>>   			label = "Button 1";
>>>>   			linux,code = <0x101>; /* BTN_1 */
>>>>   			gpios = <&gpio3 6 GPIO_ACTIVE_LOW>;
>>>>   		};
>>>>  -		button@2 {
>>>>  +		button-2 {
>>>>   			label = "Button 2";
>>>>   			linux,code = <0x102>; /* BTN_2 */
>>>>   			gpios = <&gpio2 14 GPIO_ACTIVE_LOW>;
>>>>  diff --git a/arch/mips/boot/dts/ingenic/gcw0.dts
>>>>  b/arch/mips/boot/dts/ingenic/gcw0.dts
>>>>  index 4abb0318416c..5d33f26fd28c 100644
>>>>  --- a/arch/mips/boot/dts/ingenic/gcw0.dts
>>>>  +++ b/arch/mips/boot/dts/ingenic/gcw0.dts
>>>>  @@ -130,89 +130,86 @@ backlight: backlight {
>>>>
>>>>   	gpio-keys {
>>>>   		compatible = "gpio-keys";
>>>>  -		#address-cells = <1>;
>>>>  -		#size-cells = <0>;
>>>
>>>  Are you sure you can remove these?
>>
>> Yes, from DT spec point of view, DT bindings and Linux implementation.
>> However this particular change was not tested, except building.
>>
>>>
>>>  Looking at paragraph 2.3.5 of the DT spec, I would think they have 
>>> to
>>>  stay (although with #address-cells = <0>).
>>
>> The paragraph 2.3.5 says nothing about regular properties (which can 
>> be
>> also child nodes). It says about children of a bus, right? It's not
>> related here, it's not a bus.
> 
> I quote:
> "A DTSpec-compliant boot program shall supply #address-cells and 
> #size-cells on all nodes that have children."

And paragraph 2.2.3 says:
"A unit address may be omitted if the full path to the node is unambiguous."

You have address/size cells for nodes with children having unit
addresses. If they don't unit addresses, you don't add address/size
cells (with some exceptions).

The paragraph 2.3.5 mentions "child device nodes" and these properties
are not devices, although I agree that DT spec here is actually confusing.

> 
> The gpio-keys node has children nodes, therefore it should have 
> #address-cells and #size-cells, there's no room for interpretation here.
> 
>> Second, why exactly this one gpio-keys node is different than all 
>> other
>> gpio-keys everywhere and than bindings? Why this one has to be
>> incompatible/wrong according to bindings (which do not allow
>> address-cells and nodes with unit addresses)?
> 
> Nothing is different. I'm just stating that your proposed fix is 
> invalid if we want to enforce compliance with the DT spec.

In such case, we rather enforce the compliance with the bindings.

Best regards,
Krzysztof