On 28/05/2024 15:06, Andreas Kemnade wrote:
> On Tue, 28 May 2024 13:25:29 +0200
> Krzysztof Kozlowski <krzk@xxxxxxxxxx> wrote:
>
>> On 28/05/2024 13:16, Andreas Kemnade wrote:
>>> On Tue, 28 May 2024 12:04:22 +0200
>>> Krzysztof Kozlowski <krzk@xxxxxxxxxx> wrote:
>>>
>>>> On 28/05/2024 08:57, Andreas Kemnade wrote:
>>>>> Convert the regulator bindings to yaml files. To allow only the regulator
>>>>> compatible corresponding to the toplevel mfd compatible, split the file
>>>>> into one per device.
>>>>>
>>>>> To not need to allow any subnode name, specify clearly node names
>>>>> for all the regulators.
>>>>>
>>>>> Drop one twl5030 compatible due to no documentation on mfd side and no
>>>>> users of the twl5030.
>>>>>
>>>>> Signed-off-by: Andreas Kemnade <andreas@xxxxxxxxxxxx>
>>>>> ---
>>>>> Reason for being RFC:
>>>>> the integration into ti,twl.yaml seems not to work as expected
>>>>> make dt_binding_check crashes without any clear error message
>>>>> if used on the ti,twl.yaml
>>>>>
>>>>> .../devicetree/bindings/mfd/ti,twl.yaml | 4 +-
>>>>> .../regulator/ti,twl4030-regulator.yaml | 402 ++++++++++++++++++
>>>>> .../regulator/ti,twl6030-regulator.yaml | 292 +++++++++++++
>>>>> .../regulator/ti,twl6032-regulator.yaml | 238 +++++++++++
>>>>> .../bindings/regulator/twl-regulator.txt | 80 ----
>>>>> 5 files changed, 935 insertions(+), 81 deletions(-)
>>>>> create mode 100644 Documentation/devicetree/bindings/regulator/ti,twl4030-regulator.yaml
>>>>> create mode 100644 Documentation/devicetree/bindings/regulator/ti,twl6030-regulator.yaml
>>>>> create mode 100644 Documentation/devicetree/bindings/regulator/ti,twl6032-regulator.yaml
>>>>> delete mode 100644 Documentation/devicetree/bindings/regulator/twl-regulator.txt
>>>>>
>>>>> diff --git a/Documentation/devicetree/bindings/mfd/ti,twl.yaml b/Documentation/devicetree/bindings/mfd/ti,twl.yaml
>>>>> index c2357fecb56cc..4ced6e471d338 100644
>>>>> --- a/Documentation/devicetree/bindings/mfd/ti,twl.yaml
>>>>> +++ b/Documentation/devicetree/bindings/mfd/ti,twl.yaml
>>>>> @@ -50,7 +50,7 @@ allOf:
>>>>> properties:
>>>>> compatible:
>>>>> const: ti,twl4030-wdt
>>>>> -
>>>>> + $ref: /schemas/regulator/ti,twl4030-regulator.yaml
>>>>
>>>> That's not needed, just like othehr refs below.
>>>>
>>> but how to prevent error messages like this:
>>>
>>> arch/arm/boot/dts/ti/omap/omap2430-sdp.dtb: twl@48: Unevaluated properties are not allowed ('gpio', 'keypad', 'pwm', 'pwmled', 'regulator-vaux1', 'regulator-vaux2', 'regulator-vaux3', 'regulator-vaux4', 'regulator-vdac', 'regulator-vdd1', 'regulator-vintana1', 'regulator-vintana2', 'regulator-vintdig', 'regulator-vio', 'regulator-vmmc1', 'regulator-vmmc2', 'regulator-vpll1', 'regulator-vpll2', 'regulator-vsim', 'regulator-vusb1v5', 'regulator-vusb1v8', 'regulator-vusb3v1
>>>
>>> esp. the regulator parts without adding stuff to ti,twl.yaml?
>>
>> Eh? That's a watchdog, not regulator. Why do you add ref to regulator?
>>
> hmm, wrongly indented? At what level doet it belong? But as the regualor.yaml stuff can
> be shortened, maybe just add it directly to ti,twl.yaml to avoid that trouble.
I don't follow. The diff here and in other two places suggest you add
twl-regulator reference to wdt/gpio/whatnot nodes, not to regulators.
>
>> ...
>>
>>>>> +
>>>>> + regulator-vaux2:
>>>>> + type: object
>>>>> + $ref: regulator.yaml#
>>>>> + unevaluatedProperties: false
>>>>> + properties:
>>>>> + compatible:
>>>>> + const: "ti,twl4030-vaux2"
>>>>> +
>>>>> + regulator-initial-mode:
>>>>> + items:
>>>>> + - items:
>>>>> + enum:
>>>>> + - 0x08 # Sleep mode, the nominal output voltage is maintained
>>>>> + # with low power consumption with low load current capability
>>>>> + - 0x0e # Active mode, the regulator can deliver its nominal output
>>>>> + # voltage with full-load current capability
>>>>
>>>> These entries are the same. Just use patternProperties and enum for
>>>> compatible.
>>>>
>>> hmm, if I am using that, how do I prevent e.g. constructions like this to be
>>> valid?
>>>
>>> regulator-vaux2 {
>>> compatible = "ti,twl4030-vaux1";
>>> };
>>>
>>
>> Why would node name matter if you have compatible? The entire point of
>> compatibles is to not to rely on node names.
>>
> Hmm, even if we rely on them, it should somehow match what is inside that node
> usually. We have @xx and reg=<xx>; e.g. So relax the stuff to allowing
I don't follow what reg has anything to do with it. There are no reg
properties in regulator nodes.
> any regulator-.* as node name independently of the contents?
As I said: patternProperties+enum
BTW, the example in MFD (so main node) is heavily incomplete. This
should be full, complete, passing dt_binding_check example.
>
> And since that all is then shorter, maybe add stuff just directly to ti,twl.yaml?
Sure.
Best regards,
Krzysztof
