Re: [PATCH v2 5/6] arm: dts: amlogic: migrate pwms to new meson8 v2 binding

From: Krzysztof Kozlowski
Date: Wed Nov 22 2023 - 10:10:12 EST


On 22/11/2023 15:52, Jerome Brunet wrote:
>
> On Wed 22 Nov 2023 at 09:39, Krzysztof Kozlowski <krzysztof.kozlowski@xxxxxxxxxx> wrote:
>
>> On 17/11/2023 13:59, Jerome Brunet wrote:
>>> Update Amlogic based SoC PWMs to meson8-pwm-v2 compatible
>>
>> Why? Your commit msg must explain this. You break users of this DTS on
>> older kernels and also this makes it impossible to apply via different
>> branches in the same cycle. All this needs explanation and proper
>> justification. Your message tells here nothing, because "what" is quite
>> obvious.
>>
>
> I provided all the explanation possible through the different commits of
> this series. I can re-state here if it helps

DTS commits stand on their own and must not go via same branch as
driver, so how does driver commit msg help Git history?

>
>>>
>>> Signed-off-by: Jerome Brunet <jbrunet@xxxxxxxxxxxx>
>>> ---
>>> arch/arm/boot/dts/amlogic/meson.dtsi | 4 ++--
>>> arch/arm/boot/dts/amlogic/meson8.dtsi | 16 +++++++++++++---
>>> arch/arm/boot/dts/amlogic/meson8b-ec100.dts | 2 --
>>> arch/arm/boot/dts/amlogic/meson8b-mxq.dts | 2 --
>>> arch/arm/boot/dts/amlogic/meson8b-odroidc1.dts | 2 --
>>> arch/arm/boot/dts/amlogic/meson8b.dtsi | 16 +++++++++++++---
>>> 6 files changed, 28 insertions(+), 14 deletions(-)
>>>
>>> diff --git a/arch/arm/boot/dts/amlogic/meson.dtsi b/arch/arm/boot/dts/amlogic/meson.dtsi
>>> index 8e3860d5d916..80cc004ad5fe 100644
>>> --- a/arch/arm/boot/dts/amlogic/meson.dtsi
>>> +++ b/arch/arm/boot/dts/amlogic/meson.dtsi
>>> @@ -83,14 +83,14 @@ i2c_A: i2c@8500 {
>>> };
>>>
>>> pwm_ab: pwm@8550 {
>>> - compatible = "amlogic,meson-pwm";
>>> + compatible = "amlogic,meson8-pwm-v2";
>>
>> That's breaking users of this DTS (old kernel, out of tree, other
>> projects) for no real reasons without explanation.
>
> "amlogic,meson-pwm" will continue to match, meaning of bindings is unchanged

No, because new DTS does not have amlogic,meson-pwm, thus all existing
users see breakage.

>
> How do you propose to fix badly designed bindings then ?

Justify and introduce incompatible changes, breaking the ABI. Anyway
this is a requirement, because, as I said in other reply, you cannot
have compatible for software model!

>
> if we cant even introduce a new compatible to fix things up. It is supposed to
> stay and broken till the end of time ?

No, you cannot introduce new compatible for new OS.

Fix the bindings instead with proper justification. We did it many
times, what's the problem here?

Best regards,
Krzysztof