From: Bharadwaj Raju <bharadwaj.raju@machinesoul.in>

Add (and require) the dvdd-supply property for awinic,aw88261 in
the awinic,aw88395.yaml binding.

Signed-off-by: Bharadwaj Raju <bharadwaj.raju@machinesoul.in>
---
 Documentation/devicetree/bindings/sound/awinic,aw88395.yaml | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/Documentation/devicetree/bindings/sound/awinic,aw88395.yaml b/Documentation/devicetree/bindings/sound/awinic,aw88395.yaml
index bb92d6ca3144..994d68c074a9 100644
--- a/Documentation/devicetree/bindings/sound/awinic,aw88395.yaml
+++ b/Documentation/devicetree/bindings/sound/awinic,aw88395.yaml
@@ -33,6 +33,8 @@ properties:
   reset-gpios:
     maxItems: 1
 
+  dvdd-supply: true
+
   awinic,audio-channel:
     description:
       It is used to distinguish multiple PA devices, so that different
@@ -65,6 +67,17 @@ allOf:
     then:
       properties:
         reset-gpios: false
+  - if:
+      properties:
+        compatible:
+          contains:
+            const: awinic,aw88261
+    then:
+      required:
+        - dvdd-supply
+    else:
+      properties:
+        dvdd-supply: false
 
 unevaluatedProperties: false
 

-- 
2.43.0

On 11/01/2026 11:10, Bharadwaj Raju via B4 Relay wrote:
> From: Bharadwaj Raju <bharadwaj.raju@machinesoul.in>
> 
> Add (and require) the dvdd-supply property for awinic,aw88261 in

Why? Was it missing? Why require it, which is an ABI break (or is not?)?
Why is ABI break allowed or what is its impact?

> the awinic,aw88395.yaml binding.



Best regards,
Krzysztof

On Sun, 11 Jan 2026, at 5:00 PM, Krzysztof Kozlowski wrote:
> Why? Was it missing? Why require it, which is an ABI break (or is not?)?
> Why is ABI break allowed or what is its impact?

Right now there are no users of aw88261 in the kernel device tree sources.
This patch is part of an effort to mainline the FairPhone 5, for which we 
will add the first use of this compatible, and there we need to specify 
dvdd-supply for this chip's power supply.

Since there are no present users, I thought it was OK to add a new required 
property. If not, I can make it optional.

On 11/01/2026 13:48, Bharadwaj Raju wrote:
> On Sun, 11 Jan 2026, at 5:00 PM, Krzysztof Kozlowski wrote:
>> Why? Was it missing? Why require it, which is an ABI break (or is not?)?
>> Why is ABI break allowed or what is its impact?
> 
> Right now there are no users of aw88261 in the kernel device tree sources.
> This patch is part of an effort to mainline the FairPhone 5, for which we 
> will add the first use of this compatible, and there we need to specify 
> dvdd-supply for this chip's power supply.

You mean the bindings were incomplete?

> 
> Since there are no present users, I thought it was OK to add a new required 
> property. If not, I can make it optional.

Nothing like that was explained in commit msg. Also your explanation
above does not consider out of tree users of this ABI. That's fine in
general, but needs reason why you are doing this.

Best regards,
Krzysztof

On Sun, 11 Jan 2026, at 9:40 PM, Krzysztof Kozlowski wrote:
> On 11/01/2026 13:48, Bharadwaj Raju wrote:
>> On Sun, 11 Jan 2026, at 5:00 PM, Krzysztof Kozlowski wrote:
>>> Why? Was it missing? Why require it, which is an ABI break (or is not?)?
>>> Why is ABI break allowed or what is its impact?
>> 
>> Right now there are no users of aw88261 in the kernel device tree sources.
>> This patch is part of an effort to mainline the FairPhone 5, for which we 
>> will add the first use of this compatible, and there we need to specify 
>> dvdd-supply for this chip's power supply.
>
> You mean the bindings were incomplete?

Yes, the chip needs DVDD to power on.

>> 
>> Since there are no present users, I thought it was OK to add a new required 
>> property. If not, I can make it optional.
>
> Nothing like that was explained in commit msg. Also your explanation
> above does not consider out of tree users of this ABI. That's fine in
> general, but needs reason why you are doing this.

After looking at other codec drivers, they generally make it optional 
in the binding but error out from probe if it can't be enabled.

I'll resend the series with it made optional.

Regards,
Bharadwaj

On Mon, Jan 12, 2026 at 09:57:46PM +0530, Bharadwaj Raju wrote:
> On Sun, 11 Jan 2026, at 9:40 PM, Krzysztof Kozlowski wrote:

> > You mean the bindings were incomplete?

> Yes, the chip needs DVDD to power on.

> After looking at other codec drivers, they generally make it optional 
> in the binding but error out from probe if it can't be enabled.

> I'll resend the series with it made optional.

No, if the chip needs the supply it shouldn't be optional in the
bindings - it is an ABI break, but in practice only for validation since
the kernel will assume some supply is there even if not described.