Re: [PATCH v2 05/11] regulator: dt-bindings: mediatek: Add MT6366 PMIC

[Krzysztof Kozlowski]

On 23/08/2023 06:20, Chen-Yu Tsai wrote:
> On Wed, Aug 23, 2023 at 3:40 AM Krzysztof Kozlowski
> <krzysztof.kozlowski@xxxxxxxxxx> wrote:
>>
>> On 22/08/2023 21:39, Krzysztof Kozlowski wrote:
>>> On 22/08/2023 10:45, Chen-Yu Tsai wrote:
>>>> From: Zhiyong Tao <zhiyong.tao@xxxxxxxxxxxx>
>>>>
>>>> The MediaTek MT6366 PMIC is similar to the MT6358 PMIC. It is designed
>>>> to be paired with the MediaTek MT8186 SoC. It has 9 buck regulators and
>>>> 29 LDO regulators, not counting ones that feed internally and basically
>>>> have no controls. The regulators are named after their intended usage
>>>> for the SoC and system design, thus not named generically as ldoX or
>>>> dcdcX, but as vcn33 or vgpu.
>>>>
>>>> Add a binding document describing all the regulators and their supplies.
>>>>
>>>> Signed-off-by: Zhiyong Tao <zhiyong.tao@xxxxxxxxxxxx>
>>>> [wens@xxxxxxxxxxxx: major rework and added commit message]
>>>> Signed-off-by: Chen-Yu Tsai <wenst@xxxxxxxxxxxx>
>>>> ---
>>>> Changes since v1:
>>>> - Replaced underscores in supply names to hyphens
>>>> - Merged with MT6358 regulator binding
>>>> - Added MT6358 fallback compatible to MT6366 regulator
>>>>
>>>> Changes since Zhiyong's last version (v4) [1]:
>>>> - simplified regulator names
>>>> - added descriptions to regulators
>>>> - removed bogus regulators (*_sshub)
>>>> - merged vcn33-wifi and vcn33-bt as vcn33
>>>> - added missing regulators (vm18, vmddr, vsram-core)
>>>> - cut down examples to a handful of cases and made them complete
>>>> - expanded commit message a lot
>>>>
>>>> [1] https://lore.kernel.org/linux-arm-kernel/20220823123745.14061-1-zhiyong.tao@xxxxxxxxxxxx/
>>>>  .../regulator/mediatek,mt6358-regulator.yaml  | 227 +++++++++++++-----
>>>>  1 file changed, 168 insertions(+), 59 deletions(-)
>>>>
>>>> diff --git a/Documentation/devicetree/bindings/regulator/mediatek,mt6358-regulator.yaml b/Documentation/devicetree/bindings/regulator/mediatek,mt6358-regulator.yaml
>>>> index 82328fe17680..b350181f33ff 100644
>>>> --- a/Documentation/devicetree/bindings/regulator/mediatek,mt6358-regulator.yaml
>>>> +++ b/Documentation/devicetree/bindings/regulator/mediatek,mt6358-regulator.yaml
>>>> @@ -16,14 +16,18 @@ description: |
>>>>
>>>>  properties:
>>>>    compatible:
>>>> -    const: mediatek,mt6358-regulator
>>>> +    oneOf:
>>>> +      - const: mediatek,mt6358-regulator
>>>> +      - items:
>>>> +          - const: mediatek,mt6366-regulator
>>>> +          - const: mediatek,mt6358-regulator
>>>>
>>>>    vsys-ldo1-supply:
>>>>      description: Supply for LDOs vfe28, vxo22, vcn28, vaux18, vaud28, vsim1, vusb, vbif28
>>>>    vsys-ldo2-supply:
>>>> -    description: Supply for LDOs vldo28, vio28, vmc, vmch, vsim2
>>>> +    description: Supply for LDOs vldo28 (MT6358 only), vio28, vmc, vmch, vsim2
>>>>    vsys-ldo3-supply:
>>>> -    description: Supply for LDOs vcn33, vcama1, vcama2, vemc, vibr
>>>> +    description: Supply for LDOs vcn33, vcama[12] (MT6358 only), vemc, vibr
>>>>    vsys-vcore-supply:
>>>>      description: Supply for buck regulator vcore
>>>>    vsys-vdram1-supply:
>>>> @@ -43,75 +47,138 @@ properties:
>>>>    vsys-vs2-supply:
>>>>      description: Supply for buck regulator vs2
>>>>    vs1-ldo1-supply:
>>>> -    description: Supply for LDOs vrf18, vefuse, vcn18, vcamio, vio18
>>>> +    description: Supply for LDOs vrf18, vefuse, vcn18, vcamio (MT6358 only), vio18
>>>>    vs2-ldo1-supply:
>>>> -    description: Supply for LDOs vdram2
>>>> +    description: Supply for LDOs vdram2, vmddr (MT6366 only)
>>>>    vs2-ldo2-supply:
>>>>      description: Supply for LDOs vrf12, va12
>>>>    vs2-ldo3-supply:
>>>> -    description: Supply for LDOs vsram-gpu, vsram-others, vsram-proc11, vsram-proc12
>>>> -  vs2-ldo4-supply:
>>>> -    description: Supply for LDO vcamd
>>>> -
>>>> -patternProperties:
>>>> -  "^buck_v(core|dram1|gpu|modem|pa|proc1[12]|s[12])$":
>>>> -    description: Buck regulators
>>>> -    type: object
>>>> -    $ref: regulator.yaml#
>>>> -    unevaluatedProperties: false
>>>> -
>>>> -  "^ldo_v(a|rf)12":
>>>> -    description: LDOs with fixed 1.2V output and 0~100/10mV tuning
>>>> -    type: object
>>>> -    $ref: regulator.yaml#
>>>> -    unevaluatedProperties: false
>>>> -
>>>> -  "^ldo_v((aux|cn|io|rf)18|camio)":
>>>> -    description: LDOs with fixed 1.8V output and 0~100/10mV tuning
>>>> -    type: object
>>>> -    $ref: regulator.yaml#
>>>> -    unevaluatedProperties: false
>>>> -
>>>> -  "^ldo_vxo22":
>>>> -    description: LDOs with fixed 2.2V output and 0~100/10mV tuning
>>>> -    type: object
>>>> -    $ref: regulator.yaml#
>>>> -    unevaluatedProperties: false
>>>> -
>>>> -  "^ldo_v(aud|bif|cn|fe|io)28":
>>>> -    description: LDOs with fixed 2.8V output and 0~100/10mV tuning
>>>> -    type: object
>>>> -    $ref: regulator.yaml#
>>>> -    unevaluatedProperties: false
>>>> -
>>>> -  "^ldo_vusb":
>>>> -    description: LDOs with fixed 3.0V output and 0~100/10mV tuning
>>>> -    type: object
>>>> -    $ref: regulator.yaml#
>>>> -    unevaluatedProperties: false
>>>> -
>>>> -  "^ldo_vsram_(gpu|others|proc1[12])$":
>>>> -    description: LDOs with variable output
>>>> -    type: object
>>>> -    $ref: regulator.yaml#
>>>> -    unevaluatedProperties: false
>>>> -
>>>> -  "^ldo_v(cama[12]|camd|cn33|dram2|efuse|emc|ibr|ldo28|mc|mch|sim[12])$":
>>>> -    description: LDOs with variable output and 0~100/10mV tuning
>>>> -    type: object
>>>> -    $ref: regulator.yaml#
>>>> -    unevaluatedProperties: false
>>>
>>> I don't understand. You just added it and it is already wrong? Please,
>>> do not add code which is clearly incorrect.
>>
>> Sent too early - anyway properties cannot be defined in allOf:. That's
>> not the place for them and there is no single reason for it. From which
>> regulator binding you got this example?
> 
> None. It was simply a way I figured out when I was reading up on JSON
> schema syntax. I wanted to split the definitions cleanly, since they
> are very different. And with "unevaluatedProperties: false" in the base
> schema it did seem to work, successfully evaluating existing device trees
> and producing errors when extra properties were added, or if types didn't
> match up.

If they are very different, this should not have been one binding. There
is little benefit of that.

> 
> Now that you mention it, I suppose the preferred way to write it is to
> have all the properties in the base schema, then negate the ones that
> don't belong in the allOf: section? It just seems really repetitive given
> the child node names for the chip variants are completely different. OOTH
> I guess it would produce better error messages.


For regular cases yes, but not if devices differ so much.

Best regards,
Krzysztof