On 01/02/2023 15:03, Mark Brown wrote:
On Wed, Feb 01, 2023 at 02:13:46PM +0100, Krzysztof Kozlowski wrote:
On 01/02/2023 14:10, Krzysztof Kozlowski wrote:
Because the binding:
- Was never tested,
- Was never sent to Devicetree maintainers,
- Is entirely broken and wrong, so it would have to be almost rewritten from scratch,
- It does not match the driver, IOW, the binding is fake.
I understand that in general we tend to fix, not just to revert. But the poor quality of this binding and the next patch, which was suppose to fix it, plus complete lack of testing, means I do not believe the author will send correct binding.
More over, fixing binding might require dropping incorrect properties, thus changing the driver. I am not willing to do that, I doubt that anyone has the time for it.
It is an absolutely trivial binding as is, it is utterly disproportionate to delete both the binding and the driver to fix whatever it is that the issues you're seeing are (I can't really tell TBH). Undocumented properties are a separate thing but again a revert is obviously disproportionate here, glancing at the driver the code is all well enough separated and can have default values. Looking again I did miss the sysclk selection which should be dropped, clocks should use the clock bindings.
It's the job of submitter to work on it.
It's also not the end of the world if we have a driver that isn't perfect.
Please, try to keep things constructive.
I tried. I started writing patch to fix few things in this binding and then noticed that it is entirely empty and documents nothing.
The trouble is that soon you will send it to Linus and then it becomes the ABI even though no one ever approved or reviewed the actual ABI.
Best regards, Krzysztof