Re: [PATCH] dt-bindings: mfd: Add Renesas R2A11302FT PMIC

[Krzysztof Kozlowski <krzk@kernel.org>]

On 10/10/2025 09:58, Wolfram Sang wrote:
> Hi Krzysztof,
> 
> thanks for the review!
> 
> On Fri, Oct 10, 2025 at 04:37:14AM +0200, Krzysztof Kozlowski wrote:
>> On 09/10/2025 20:12, Wolfram Sang wrote:
>>> This PMIC is referenced in upstream DTs for the Renesas Lager and
>>> Koelsch boards. Sadly, we don't have documentation for more complete
>>> bindings, but due to the spi-cpol/cpha it also doesn't belong to trivial
>>
>> That's not a problem.
> 
> It is not? I found this mail[1] saying that enabling spi-cpol et al.
> will disable sanity checks for other trivial devices.
> 
> Adding this PMIC to trivial devices would be my fav solution, actually.

I meant, that not being complete is not yet problem.

> 
> [1] https://www.spinics.net/lists//devicetree/msg731369.html
> 
>>> Change since RFC [1]:
>>
>> Please start using b4. This is v2. Your wrong numbering makes any
>> comparisons broken. Try yourself with b4.
> 
> I checked the link [1] I gave last mail and the mail said "[RFC PATCH]".
> Looks correct to me, am I overlooking something? I am open to switching
> to b4, but I need a time slot for that.

Heh? I said your numbering is wrong. RFC was v1, this is v2.

Your broken numbering breaks tools. Try yourself:

b4 diff '<20251009181916.2431-2-wsa+renesas@sang-engineering.com>'

Works? Not. Should work? Obviously yes.

> 
>>> * dropped "unevaluatedProperties" because we know this binding is not
>>>   complete currently
>>
>> You cannot. Bindings must be complete (or complete "enough"), otherwise
>> I see no point in accepting them.
> 
> I can see that.
> 
>> It is like you sent driver code which does not build because you did not
>> write half of it. Why would we want it?
> 
> I would agree if it does not build. But it does build. So, the proper
> comparison for me is a limited driver which probes and gives access to
> the device. Features can be implemented incrementally. This bindings
> works/builds with the few information we have. It even removes a
> warning.
> 
> My reasoning is that DT describes HW. And on the Renesas Lager board,
> this PMIC *is* connected to the SPI controller. I am open to removing
> this node because we don't have a driver. That would mean, though,
> otherwise rare to find information is lost.


If you have it in DTS, then you need the binding. Just write it as much
as complete as possible, mention incompleteness in binding description
and provide reasons of incompleteness in commit msg.


Best regards,
Krzysztof