Re: [PATCH v16 1/3] dt-bindings: i2c: aspeed: support for AST2600-i2cv2

[Krzysztof Kozlowski <krzk@kernel.org>]

On 27/02/2025 09:19, Ryan Chen wrote:
>>
>>
>>> aspeed,enable-byte:
>>> Force i2c controller use byte mode transfer. the byte mode transfer
>>> will send i2c data each byte by byte, inlcude address read/write.
>>
>> Isn't this standard FIFO mode?
> Yes, it is.
>>
>> Why anyone would need to enable byte mode for given board?
> By default, it is buffer-mode, for performance, I don't want user enable byte-mode, it will increase CPU utilize.
> But someone want to be force enable byte-mode, so I add properties. 
> https://patchwork.ozlabs.org/project/linux-aspeed/patch/20241007035235.2254138-3-ryan_chen@aspeedtech.com/


I don't see the reason why this would be a board property.

I understood need for DMA because it is shared and only some of the
controllers can use it. But why choice between buffer and FIFO depending
on hardware?


>>
>>>
>>>>
>>>>> +      may require a DTS to manually allocate which controller can
>>>>> + use
>>>> DMA mode.
>>>>> +      The "aspeed,enable-dma" property allows control of this.
>>>>> +
>>>>> +      In cases where one the hardware design results in a specific
>>>>> +      controller handling a larger amount of data, a DTS would likely
>>>>> +      enable DMA mode for that one controller.
>>>>> +
>>>>> +  aspeed,enable-byte:
>>>>> +    type: boolean
>>>>> +    description: |
>>>>> +      I2C bus enable byte mode transfer.
>>>>
>>>> No, either this is expressed as lack of dma mode property or if you
>>>> have three modes, then rather some enum (aspeed,transfer-mode ?)
>>>
>>> Thanks suggestion, I will using an enum property like aspeed,transfer-mode
>> instead.
>>>>
>>>>
>>>>
>>>>> +
>>>>> +  aspeed,global-regs:
>>>>> +    $ref: /schemas/types.yaml#/definitions/phandle
>>>>> +    description: The phandle of i2c global register node.
>>>>
>>>> For what? Same question as usual: do not repeat property name, but
>>>> say what is this used for and what exactly it points to.
>>>>
>>>> s/i2c/I2C/ but then what is "I2C global register node"? This is how
>>>> you call your device in datasheet?
>>>>
>>> I do descript in cover, should add into the yaml file ?
>>
>>
>> Again, cover letter does not matter. Your hardware must be explained here.
> Will add into commit. 
>>
>>>
>>> aspeed,global-regs:
>>> This global register is needed, global register is setting for new
>>> clock divide control, and new register set control.
>>
>> So this means you implement clock controller via syscon?
> No, it is just mode switch. It also explain in cover. I will add it in commit. 
> The legacy register layout is mix controller/target register control together. The following is add more detail description about new register layout. And new feature set add for register.
>>
>>>
>>>>
>>>>> +
>>>>>  required:
>>>>>    - reg
>>>>>    - compatible
>>>>>    - clocks
>>>>>    - resets
>>>>>
>>>>> +allOf:
>>>>> +  - $ref: /schemas/i2c/i2c-controller.yaml#
>>>>> +  - if:
>>>>> +      properties:
>>>>> +        compatible:
>>>>> +          contains:
>>>>> +            const: aspeed,ast2600-i2cv2
>>>>
>>>> NAK, undocumented compatible.
>>>
>>> Sorry, I should add what kind of document about this compatible?
>>
>> You cannot add new compatibles without documenting them. Documentation
>> is in the form of DT schema and each compatible must be listed (in some
>> way) in compatible property description.
> 
> Sorry, do you mean, I add following in yaml or commit message?

You need to list this in compatibles first.

> 
> This series add AST2600 i2cv2 new register set driver. The i2cv2 driver is new register set that have new clock divider option for more flexiable generation. And also have separate i2c controller and target register set for control, patch #2 is i2c controller driver only, patch #3 is add i2c target mode driver.

All this describes driver, not hardware.

> 
> The legacy register layout is mix controller/target register control together. The following is add more detail description about new register layout. And new feature set add for register.
> 
> -Add new clock divider option for more flexible and accurate clock rate generation -Add tCKHighMin timing to guarantee SCL high pulse width.
> -Add support dual pool buffer mode, split 32 bytes pool buffer of each device into 2 x 16 bytes for Tx and Rx individually.
> -Increase DMA buffer size to 4096 bytes and support byte alignment.
> -Re-define the base address of BUS1 ~ BUS16 and Pool buffer.
> -Re-define registers for separating controller and target mode control.
> -Support 4 individual DMA buffers for controller Tx and Rx, target Tx and Rx.

Does it mean hardware changed on AST2600? Or these are different devices
than aspeed,ast2600-i2c-bus? If this is not a different device, how one
SoC can have two different flavors of same device in the same instance?




Best regards,
Krzysztof