Message-ID: <fca941a4-7c24-48dd-b36a-2f9b5c44575c@kernel.org>
Date: Sun, 5 Jan 2025 11:39:53 +0100
From: Krzysztof Kozlowski <krzk@...nel.org>
To: Ivaylo Ivanov <ivo.ivanov.ivanov1@...il.com>
Cc: Rob Herring <robh@...nel.org>, Conor Dooley <conor+dt@...nel.org>,
 Alim Akhtar <alim.akhtar@...sung.com>,
 Sam Protsenko <semen.protsenko@...aro.org>, devicetree@...r.kernel.org,
 linux-arm-kernel@...ts.infradead.org, linux-samsung-soc@...r.kernel.org,
 linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 2/3] dt-bindings: soc: samsung: usi: add USIv1 and
 samsung,exynos8895-usi

On 05/01/2025 10:51, Ivaylo Ivanov wrote:
> On 1/5/25 11:18, Krzysztof Kozlowski wrote:
>> On Sat, Jan 04, 2025 at 06:29:14PM +0200, Ivaylo Ivanov wrote:
>>>  
>>>    reg:
>>>      maxItems: 1
>>> @@ -64,7 +75,6 @@ properties:
>>>  
>>>    samsung,mode:
>>>      $ref: /schemas/types.yaml#/definitions/uint32
>>> -    enum: [0, 1, 2, 3]
>> Widest constraints stay here, so minimum/maximum.
> 
> Why?

Because that's the coding style and that's how you define the field,
considering you might miss a variant in multiple if:then: .


> If we are going to add new enum values specific for each USI version,
> isn't it better to selectively constrain them in the binding?

You are supposed to constrained them.

Again: widest constrains always stay in top level property. This applies
to all bindings, all fields. Repeated multiple times, so here is
standard example:

https://elixir.bootlin.com/linux/v6.11-rc6/source/Documentation/devicetree/bindings/ufs/qcom,ufs.yaml#L127


> 
>>
>>>      description:
>>>        Selects USI function (which serial protocol to use). Refer to
>>>        <include/dt-bindings/soc/samsung,exynos-usi.h> for valid USI mode values.
>>> @@ -101,18 +111,42 @@ required:
>>>    - samsung,sysreg
>>>    - samsung,mode
>>>  
>>> +allOf:
>>> +  - if:
>>> +      properties:
>>> +        compatible:
>>> +          contains:
>>> +            enum:
>>> +              - samsung,exynos850-usi
>>> +    then:
>>> +      properties:
>>> +        reg:
>>> +          maxItems: 1
>>> +
>>> +        samsung,mode:
>>> +          enum: [0, 1, 2, 3]
>>> +
>>> +      required:
>>> +        - reg
>>> +
>>> +    else:
>>> +      properties:
>>> +        reg: false
>>> +        samsung,clkreq-on: false
>>> +
>>> +        samsung,mode:
>>> +          enum: [4, 5, 6, 7, 8, 9, 10]
>> Is it really true? Previously for example GS101 had values 0-3, now you
>> claim has values 4-10, so an ABI change without explanation.
> 
> How is it unexplained? Citing the commit message:
> "Add constants for choosing USIv1 configuration mode in device tree.
> Those are further used in the USI driver to figure out which value to
> write into SW_CONF register."
> 

I don't see reference here about GS101 and others.

> USIv1 and v2 write different values to the SW_CONF register. For example:
> 
> #define USI_V1_SW_CONF_UART        0x8
> #define USI_V2_SW_CONF_UART    BIT(0)
> 
> ..
>  [USI_V2_UART] =    { .name = "uart", .val = USI_V2_SW_CONF_UART },
>  [USI_V1_UART] =    { .name = "uart", .val = USI_V1_SW_CONF_UART },
> ..
> 
> Hence the decision to have separate constants for different USI versions,
> which in my opinion is cleaner than meshing the enums together and
> choosing what to use with IFs in the driver code.

This does not answer at all why GS101 receives now different values than
before.

Explain why you are changing ABI.

> 
>>
>>> +
>>>  if:
>> Missing allOf:
>>
>>>    properties:
>>>      compatible:
>>>        contains:
>>>          enum:
>>>            - samsung,exynos850-usi
>>> +          - samsung,exynos8895-usi
>> Effect is not readable and not correct. You have two if:then:else.
>> Usually it is easier to just have separate if: for each group of
>> variants and define/constrain complete for such group within its if:.
>>
>>>  
>>>  then:
>>>    properties:
>>> -    reg:
>>> -      maxItems: 1
>>> -
>>>      clocks:
>>>        items:
>>>          - description: Bus (APB) clock
>>> @@ -122,16 +156,13 @@ then:
>>>        maxItems: 2
>>>  
>>>    required:
>>> -    - reg
>>>      - clocks
>>>      - clock-names
>>>  
>>>  else:
>>>    properties:
>>> -    reg: false
>>>      clocks: false
>>>      clock-names: false
>>> -    samsung,clkreq-on: false
>>>  
>>>  additionalProperties: false
>>>  
>>> diff --git a/include/dt-bindings/soc/samsung,exynos-usi.h b/include/dt-bindings/soc/samsung,exynos-usi.h
>>> index a01af169d..4c077c9a8 100644
>>> --- a/include/dt-bindings/soc/samsung,exynos-usi.h
>>> +++ b/include/dt-bindings/soc/samsung,exynos-usi.h
>>> @@ -13,5 +13,12 @@
>>>  #define USI_V2_UART		1
>>>  #define USI_V2_SPI		2
>>>  #define USI_V2_I2C		3
>>> +#define USI_V1_NONE		4
>> Drop, it is already there.
>>
>>> +#define USI_V1_I2C0		5
>>> +#define USI_V1_I2C1		6
>>> +#define USI_V1_I2C0_1		7
>>> +#define USI_V1_SPI		8
>> Drop
>>
>>> +#define USI_V1_UART		9
>> Drop
> 
> How so? These bring different configuration values. Could you specify how
> exactly you want me to implement these in the driver?

Heh, so the binding was made poorly for the driver and driver was
developed in a matching way, so now this became an argument. Binding and
drivers are independent, so whatever argument was in the driver does not
matter really.

What I don't understand is downstream for USIv1 and USIv2 has it correct
- proper string values without mentioning any version. So, surprisingly
proper downstream binding, really rare case, was converted to something
tied to current driver implementation.

You have only one sort of property - the mode how you configure the USI
engine. The mode can be: I2C, SPI, I2C0, I2C1 for special cases with two
I2C etc.

The mode is not "USI_V1_I2C" because it is redundant. USI V1 cannot be
USI V2. You cannot tell USI to be v1 or v2. It is either v1 or v2. Only
one of these, thus encoding this information in the binding is meaningless.

I even mentioned this in original binding review:
"so please drop everywhere v2 (bindings, symbols, Kconfig,
functions) except the compatible."
Well, then I missed to check on that, so now this mess has to be fixed.

Best regards,
Krzysztof

Powered by blists - more mailing lists