Re: [PATCH v7 10/12] nvmem: s32g2_siul2: add NVMEM driver for SoC information

[Andrei Stefanescu <andrei.stefanescu@oss.nxp.com>]

Hi Krzysztof,

On 04/08/2025 10:26, Krzysztof Kozlowski wrote:
> On 04/08/2025 09:12, Andrei Stefanescu wrote:
>> Hi Krzysztof,
>>
>> Thank you for the quick response!
>> On 02/08/2025 11:32, Krzysztof Kozlowski wrote:
>>> On 02/08/2025 10:28, Krzysztof Kozlowski wrote:
>>>> On 01/08/2025 16:36, Andrei Stefanescu wrote:
>>>>> Apart from the proposed NVMEM driver, there is also an option of exporting
>>>>> a syscon regmap for the registers which provide information about the SoC.
>>>>>
>>>>> I have seen that typically NVMEM drivers export information read from fuses
>>>>> but I think having a NVMEM driver is nicer way to access the information
>>>>> instead of using a syscon regmap and manually extracting the needed bits. 
>>>>
>>>>
>>>> nvmem is not a syscon. Mixing these two means device is something
>>>> completely else.
>>
>> Yes, I don't want to mix them. The driver will either be a NVMEM driver or
>> a syscon. These registers are read-only. I suggested NVMEM because it's a
> 
> We do not talk about drivers here, but hardware.
> 
>> an abstraction layer which makes it easier for drivers which want to use
>> that information without knowing where to actually read it i.e. reg address,
>> bit mask.
> 
> Sorry, but no. You design it for drivers, that's not the way. Describe
> properly the hardware.

I don't think there's a clear way. These are read-only registers, I am not sure
if exporting them as a syscon regmap properly describes the hardware. Moreover,
I saw the following phrase in the NVMEM documentation [1]:

"NVMEM is the abbreviation for Non Volatile Memory layer. It is used to retrieve
configuration of SOC or Device specific data from non volatile memories like eeprom,
efuses and so on."

This suggests that NVMEM might be a fit. I incline more towards the NVVMEM driver
because it will also reduce code duplication (if multiple drivers need to read the
part number for example).

What do you think? Would you consider NVMEM the way to go in this case?

Best regards,
Andrei

[1] - https://www.kernel.org/doc/html/latest/driver-api/nvmem.html