Re: [PATCH] dt-bindings: marvell: Update Armada 37xx platform bindings

[Pali Rohár]

On Tuesday 09 August 2022 18:46:04 Krzysztof Kozlowski wrote:
> On 09/08/2022 16:13, Pali Rohár wrote:
> > On Tuesday 09 August 2022 08:58:50 Krzysztof Kozlowski wrote:
> >> On 08/08/2022 23:23, Pali Rohár wrote:
> >>> PING?
> >>>
> >>> On Wednesday 13 July 2022 22:01:23 Pali Rohár wrote:
> >>>> Distinguish between Armada 3700 family, Armada 3710 SoC and Armada 3720 SoC.
> >>>> Armada 3720 DB is name of the board with Armada 3720 SoC, so correctly
> >>>> indicate SoC in example.
> >>>>
> >>>> Signed-off-by: Pali Rohár <pali@xxxxxxxxxx>
> >>>> ---
> >>>>  .../devicetree/bindings/arm/marvell/armada-37xx.txt        | 7 ++++++-
> >>>>  1 file changed, 6 insertions(+), 1 deletion(-)
> >>>>
> >>>> diff --git a/Documentation/devicetree/bindings/arm/marvell/armada-37xx.txt b/Documentation/devicetree/bindings/arm/marvell/armada-37xx.txt
> >>>> index f6d6642d81c0..d2ca008de266 100644
> >>>> --- a/Documentation/devicetree/bindings/arm/marvell/armada-37xx.txt
> >>>> +++ b/Documentation/devicetree/bindings/arm/marvell/armada-37xx.txt
> >>>> @@ -4,6 +4,11 @@ Marvell Armada 37xx Platforms Device Tree Bindings
> >>>>  Boards using a SoC of the Marvell Armada 37xx family must carry the
> >>>>  following root node property:
> >>>>  
> >>>> + - compatible: must contain "marvell,armada3700"
> >>>> +
> >>>> +In addition, boards using the Marvell Armada 3710 SoC shall have the
> >>>> +following property before the previous one:
> >>>> +
> >>
> >> The change is an ABI break, which looks reasonable, but still platform
> >> maintainer should comment on it. Especially on the aspect why the
> >> marvell,armada3710 fallback was chosen at the first place.
> > 
> > I do not think this is ABI break but rather incorrect documentation and
> > bug in some board dts files.
> 
> This documentation is the ABI, especially if actually implemented in the
> DTS files, therefore by all definitions of DT ABI this is an ABI break.

Ok.

> What you are saying about "incorrect documentation and bug in some DTS
> files" is a good reason to break the ABI, but it is still a break.
> 
> Best regards,
> Krzysztof

I think it is now clear what we both mean, just we used different
terminology for the same thing.