Re: [PATCH v1 1/2] dt-bindings: i2c: exynos5: Add samsung,exynos8895-hsi2c compatible

[Ivaylo Ivanov <ivo.ivanov.ivanov1@gmail.com>]

On 12/18/24 11:22, Krzysztof Kozlowski wrote:
> On 17/12/2024 11:04, Ivaylo Ivanov wrote:
>> On 12/17/24 11:43, Krzysztof Kozlowski wrote:
>>> On 17/12/2024 10:31, Ivaylo Ivanov wrote:
>>>> On 12/17/24 11:26, Krzysztof Kozlowski wrote:
>>>>> On 17/12/2024 10:08, Ivaylo Ivanov wrote:
>>>>>>>>>>        - items:
>>>>>>>>>>            - enum:
>>>>>>>>>> @@ -94,9 +95,28 @@ allOf:
>>>>>>>>>>          - clock-names
>>>>>>>>>>  
>>>>>>>>>>      else:
>>>>>>>>>> -      properties:
>>>>>>>>>> -        clocks:
>>>>>>>>>> -          maxItems: 1
>>>>>>>>>> +      if:
>>>>>>>>>> +        properties:
>>>>>>>>>> +          compatible:
>>>>>>>>>> +            contains:
>>>>>>>>>> +              enum:
>>>>>>>>>> +                - samsung,exynos8895-hsi2c
>>>>>>>>>> +
>>>>>>>>>> +      then:
>>>>>>>>>> +        properties:
>>>>>>>>>> +          clocks:
>>>>>>>>> Missing minItems
>>>>>>>>>
>>>>>>>>>> +            maxItems: 2
>>>>>>>>>> +
>>>>>>>>>> +          clock-names:
>>>>>>>>> Ditto
>>>>>>>>>
>>>>>>>>>> +            maxItems: 2
>>>>>>>>>> +
>>>>>>>>>> +        required:
>>>>>>>>>> +          - clock-names
>>>>>>>>> I don't understand why do you need second, same branch in if, basically
>>>>>>>> Because, as I stated in the commit message, we have HSI2C controllers
>>>>>>>> both implemented in USIv1 blocks and outside. These that are a part of
>>>>>>> On Exynos8895? Where? With the same compatible?
>>>>>> hsi2c_0 which has a clock from BUSC and hsi2c_1 to hsi2c_4 which use clocks
>>>>>> from PERIC1 (CLK_GOUT_PERIC1_HSI2C_CAM{0,1,2,3}_IPCLK). Why would
>>>>>> they need a different compatible though? It's functionally the same i2c design
>>>>>> as the one implemented in USIv1 blocks.
>>>>> If one block is part of USI and other not, they might not be the same
>>>>> I2C blocks, even if interface is similar. If they were the same or even
>>>>> functionally the same, they would have the same clock inputs. However
>>>> I see, so in such case I should make samsung,exynos8895-hsi2c-nonusi or
>>>> something like that?
>>>>
>>>>> user manual also suggests that there is only one clock, not two (for
>>>>> both cases), so they could be functionally equivalent but then number of
>>>>> clocks looks incorrect.
>>>> That'd be weird. Both according to downstream and upstream clk driver,
>>>> for the USI-implemented i2cs we have a pclk and an sclk_usi.
>>> Something is not precise here, as usually with Samsung clock topology.
>>>
>>> First, the non-USI instances have the IPCLK as well, e.g. things like
>>> PERIC1_UID_HSI2C_CAM1_IPCLKPORT_iPCLK
>>>
>>> USI have BLK_PERIC0_UID_USI03_IPCLKPORT_i_SCLK_USI, but that's USI
>>> clock, not HSI2C in USI. Datasheet mentions this is UART and SPI special
>>> clock, but not I2C.
>> That's weird. Don't we need the clock enabled in order for the
>> USIv1's HSI2C to work?
> The clock goes to USI, so it is enabled, no?

Yes, and as Markuss said:
"USI PCLK is used for the internal AMBA APB bus clock and the IPCLK
signal is used for the peripheral controller blocks (i2c/spi/uart)."

So perhaps referencing the USI PCLK in the hsi2c driver for USIv2, as
well as USIv1, is a wrong approach and should be dropped/fixed?

Best regards,
Ivo

>
> Best regards,
> Krzysztof