On Mon, Oct 14 2024 at 14:02:37 -05:00:00, Rob Herring
<robh@xxxxxxxxxx> wrote:
On Mon, Oct 14, 2024 at 3:27 AM AngeloGioacchino Del Regno
<angelogioacchino.delregno@xxxxxxxxxxxxx> wrote:
Il 11/10/24 18:56, Rob Herring ha scritto:
> On Fri, Oct 11, 2024 at 03:03:46PM +0300, Yassine Oudjana wrote:
>> From: Yassine Oudjana <y.oudjana@xxxxxxxxxxxxxx>
>>
>> mediatek,pinctrl-mt6795.yaml has different node name patterns
which match
>> bindings of other MediaTek pin controllers, ref for
pinmux-node.yaml which
>> has a description of the pinmux property, as well as some
additional
>> descriptions for some pin configuration properties. Pull those
changes
>> into mediatek,mt6779-pinctrl.yaml and adjust the example DTS to
match in
>> preparation to combine the MT6795 document into it.
>>
>> Signed-off-by: Yassine Oudjana <y.oudjana@xxxxxxxxxxxxxx>
>> ---
>> .../pinctrl/mediatek,mt6779-pinctrl.yaml | 38
++++++++++++++-----
>> 1 file changed, 28 insertions(+), 10 deletions(-)
>>
>> diff --git
a/Documentation/devicetree/bindings/pinctrl/mediatek,mt6779-pinctrl.yaml
b/Documentation/devicetree/bindings/pinctrl/mediatek,mt6779-pinctrl.yaml
>> index 3bbc00df5548d..352a88d7b135e 100644
>> ---
a/Documentation/devicetree/bindings/pinctrl/mediatek,mt6779-pinctrl.yaml
>> +++
b/Documentation/devicetree/bindings/pinctrl/mediatek,mt6779-pinctrl.yaml
>> @@ -111,12 +111,12 @@ allOf:
>> - "#interrupt-cells"
>>
>> patternProperties:
>> - '-[0-9]*$':
>> + '-pins$':
>
> Worst case, this could be an ABI break. Best case, it's churn for
> mt6779. Is it worth unifying?
>
All those MediaTek pinctrl bindings are mostly the same, where only
the pin
definitions in the binding header does actually change.
I think that it's worth unifying them, not only to get rid of the
duplication
but mostly for consistency between all of those subnode names which
are wildly
differing for no real reason... and consistency is a long time
issue with
MediaTek bindings/dts in general (which is way way way better now,
but still)...
Besides - just for context and nothing else: the driver doesn't
care about
the names of the subnodes, anyway... so while this is technically
an ABI break
it's not really creating any functionality issue, and then,
actually, Yassine
is also modifying the devicetrees to comply with his consistency
changes, so,
in my own perspective, it's still acceptable.
Wait, I thought there were no users?
Right, When I said there were no users I was thinking of MT6779
strictly, but MT6797 is included in the bindings so it counts too.
mt6797.dtsi is currently the only place where these bindings are used.
We generally only consider node names ABI when/if something or someone
cares. Most of the time it doesn't matter. For the pinctrl nodes, it's
really just a question of churn renaming a lot of nodes.
Ultimately, it's up to you. I only care that the implications of the
changes are clear in the commit msg.
I'll mention MT6797 in the commit message.
