Re: [PATCH v2] dt-bindings: trivial-devices: add onnn,adt7462

[Guenter Roeck <linux@roeck-us.net>]

On 9/23/24 21:17, Chanh Nguyen wrote:
> On 24/09/2024 04:23, Conor Dooley wrote:
>> On Mon, Sep 23, 2024 at 09:38:00AM +0000, Chanh Nguyen wrote:
>>> The adt7462 supports monitoring and controlling up to
>>> four PWM Fan drive outputs and eight TACH inputs measures.
>>> The adt7462 supports reading a single on chip temperature
>>> sensor and three remote temperature sensors. There are up
>>> to 13 voltage monitoring inputs.
>>>
>>> Add device tree bindings for the adt7462 device.
>>>
>>> Signed-off-by: Chanh Nguyen <chanh@os.amperecomputing.com>
>>> ---
>>> Change in v2:
>>>     - Add onnn,adt7462 to the list of trivial devices       [Guenter]
>>
>> Is this really a trivial device? If it monitors and controls fans, how
>> come those do not need to be represented in the devicetree? How is it
>> possible to tell the difference between monitoring 1 and 4 fans without
>> the extra detail?
>>
> 
> Hi Conor, Thank you for your comments!
> 
> The chip is old. The driver was added back in 2008.
> 
> Really, this is such an old chip that it would make more sense to just leave its driver alone unless there is a problem with it; this is viewpoint from Guenter.
> 
> I'm using the driver and the device tree with only the "compatible" and "reg" properties; now it's being good for me without any extra detail.
> 
> Guenter, Rob, Krzysztof, and I discussed making the decision to add this device to the list of trivial devices. You can get more information at thread https://lore.kernel.org/lkml/20240918220553.GA2216504-robh@kernel.org/T/ (Because the commit title changed between v1 and v2, it's so hard for everyone to find it. Sorry! I missed mentioning the link to pacth v1).
> 
> Guenter, who supported the driver development before, he suggested me add this device to the list of trivial devices.
> 

Historically it was ok to add an entry into trivial devices and extending
it later if/when needed. That was still widely done at least until very
recently. Maybe that changed recently. If so, sorry for bringing up the idea.
I did not envision that this might be a problem.

Can you live with no devicetree entry at all for the chip ? That is of
course less than perfect, but it seems better than trying to design a
devicetree description for the chip only to never implement it.

Thanks,
Guenter