arm64: dts: xilinx: Drop "label" property on dlg,slg7xl45106

[PATCH] arm64: dts: xilinx: Drop "label" property on dlg,slg7xl45106

Posted by Rob Herring (Arm) 1 day, 19 hours ago

The "label" property is not documented for the dlg,slg7xl45106. Nor is
it common to use for GPIO controllers. So drop it.

Signed-off-by: Rob Herring (Arm) <robh@kernel.org>
---
 arch/arm64/boot/dts/xilinx/zynqmp-sck-kd-g-revA.dtso | 1 -
 arch/arm64/boot/dts/xilinx/zynqmp-sck-kr-g-revA.dtso | 1 -
 arch/arm64/boot/dts/xilinx/zynqmp-sck-kr-g-revB.dtso | 1 -
 3 files changed, 3 deletions(-)

diff --git a/arch/arm64/boot/dts/xilinx/zynqmp-sck-kd-g-revA.dtso b/arch/arm64/boot/dts/xilinx/zynqmp-sck-kd-g-revA.dtso
index 02be5e1e8686..d0230da47068 100644
--- a/arch/arm64/boot/dts/xilinx/zynqmp-sck-kd-g-revA.dtso
+++ b/arch/arm64/boot/dts/xilinx/zynqmp-sck-kd-g-revA.dtso
@@ -75,7 +75,6 @@ u3: ina260@40 { /* u3 */
 	slg7xl45106: gpio@11 { /* u13 - reset logic */
 		compatible = "dlg,slg7xl45106";
 		reg = <0x11>;
-		label = "resetchip";
 		gpio-controller;
 		#gpio-cells = <2>;
 		gpio-line-names = "USB0_PHY_RESET_B", "",
diff --git a/arch/arm64/boot/dts/xilinx/zynqmp-sck-kr-g-revA.dtso b/arch/arm64/boot/dts/xilinx/zynqmp-sck-kr-g-revA.dtso
index b92dcb86e87e..5d1f031bbd89 100644
--- a/arch/arm64/boot/dts/xilinx/zynqmp-sck-kr-g-revA.dtso
+++ b/arch/arm64/boot/dts/xilinx/zynqmp-sck-kr-g-revA.dtso
@@ -103,7 +103,6 @@ u14: ina260@40 { /* u14 */
 	slg7xl45106: gpio@11 { /* u19 - reset logic */
 		compatible = "dlg,slg7xl45106";
 		reg = <0x11>;
-		label = "resetchip";
 		gpio-controller;
 		#gpio-cells = <2>;
 		gpio-line-names = "USB0_PHY_RESET_B", "USB1_PHY_RESET_B",
diff --git a/arch/arm64/boot/dts/xilinx/zynqmp-sck-kr-g-revB.dtso b/arch/arm64/boot/dts/xilinx/zynqmp-sck-kr-g-revB.dtso
index 99ad220d13d6..1d5c95bd6073 100644
--- a/arch/arm64/boot/dts/xilinx/zynqmp-sck-kr-g-revB.dtso
+++ b/arch/arm64/boot/dts/xilinx/zynqmp-sck-kr-g-revB.dtso
@@ -104,7 +104,6 @@ u14: ina260@40 { /* u14 */
 	slg7xl45106: gpio@11 { /* u19 - reset logic */
 		compatible = "dlg,slg7xl45106";
 		reg = <0x11>;
-		label = "resetchip";
 		gpio-controller;
 		#gpio-cells = <2>;
 		gpio-line-names = "USB0_PHY_RESET_B", "USB1_PHY_RESET_B",
-- 
2.51.0

Re: [PATCH] arm64: dts: xilinx: Drop "label" property on dlg,slg7xl45106

Posted by Michal Simek 1 day, 1 hour ago

On 12/16/25 18:59, Rob Herring (Arm) wrote:
> The "label" property is not documented for the dlg,slg7xl45106. Nor is
> it common to use for GPIO controllers. So drop it.
> 

Correct it is not documented but it is at least used in Linux also by
drivers/gpio/gpio-mmio.c:810:	err = device_property_read_string(dev, "label", 
&label);

which is also not documented in DT binding.

I don't have a problem with it because I can't see it used anywhere.

But there are other devices where label is the part of binding. And IMHO this 
should be more generic property.

Thanks,
Michal

Re: [PATCH] arm64: dts: xilinx: Drop "label" property on dlg,slg7xl45106

Posted by Rob Herring 17 hours ago

On Wed, Dec 17, 2025 at 6:39 AM Michal Simek <michal.simek@amd.com> wrote:
> On 12/16/25 18:59, Rob Herring (Arm) wrote:
> > The "label" property is not documented for the dlg,slg7xl45106. Nor is
> > it common to use for GPIO controllers. So drop it.
> >
>
> Correct it is not documented but it is at least used in Linux also by
> drivers/gpio/gpio-mmio.c:810:   err = device_property_read_string(dev, "label",
> &label);
>
> which is also not documented in DT binding.

It really should be for all GPIO providers or none of them. IMO, it
should be none of them as we already have "gpio-line-names" which is
essentially per line labels. I can't see why anyone should care about
a label for the controller.

> I don't have a problem with it because I can't see it used anywhere.
>
> But there are other devices where label is the part of binding. And IMHO this
> should be more generic property.

"label" is the generic property. Not sure what you are suggesting here.

Rob

Re: [PATCH] arm64: dts: xilinx: Drop "label" property on dlg,slg7xl45106

Posted by Krzysztof Kozlowski 1 day, 1 hour ago

On 17/12/2025 13:38, Michal Simek wrote:
> 
> 
> On 12/16/25 18:59, Rob Herring (Arm) wrote:
>> The "label" property is not documented for the dlg,slg7xl45106. Nor is
>> it common to use for GPIO controllers. So drop it.
>>
> 
> Correct it is not documented but it is at least used in Linux also by
> drivers/gpio/gpio-mmio.c:810:	err = device_property_read_string(dev, "label", 
> &label);
> 
> which is also not documented in DT binding.
> 
> I don't have a problem with it because I can't see it used anywhere.
> 
> But there are other devices where label is the part of binding. And IMHO this 
> should be more generic property.

No, please read the EXPLICIT comment:

"+	 * This property *must not* be used in device-tree sources, it's only"

It was added only for ACPI.

We have indeed the problem with documenting disallowed ABI for cases
where drivers use fwnode/device_property(). So far I insisted on
comments, but as we can see people can just ignore the comments and
still use something explicitly disallowed.

Rob,

Would it be okay to add for such cases "label: false" in the binding to
mark that this is not ABI for DT?

Best regards,
Krzysztof

Re: [PATCH] arm64: dts: xilinx: Drop "label" property on dlg,slg7xl45106

Posted by Krzysztof Kozlowski 1 day ago

On 17/12/2025 13:42, Krzysztof Kozlowski wrote:
> On 17/12/2025 13:38, Michal Simek wrote:
>>
>>
>> On 12/16/25 18:59, Rob Herring (Arm) wrote:
>>> The "label" property is not documented for the dlg,slg7xl45106. Nor is
>>> it common to use for GPIO controllers. So drop it.
>>>
>>
>> Correct it is not documented but it is at least used in Linux also by
>> drivers/gpio/gpio-mmio.c:810:	err = device_property_read_string(dev, "label", 
>> &label);
>>
>> which is also not documented in DT binding.
>>
>> I don't have a problem with it because I can't see it used anywhere.
>>
>> But there are other devices where label is the part of binding. And IMHO this 
>> should be more generic property.
> 
> No, please read the EXPLICIT comment:
> 
> "+	 * This property *must not* be used in device-tree sources, it's only"
> 
> It was added only for ACPI.

Uh, my bad, the comment is above "gpio-mmio,base", not "label", but I
have impression it was applicable to label as well. Maybe that's not the
case...

> 
> We have indeed the problem with documenting disallowed ABI for cases
> where drivers use fwnode/device_property(). So far I insisted on
> comments, but as we can see people can just ignore the comments and
> still use something explicitly disallowed.
> 
> Rob,
> 
> Would it be okay to add for such cases "label: false" in the binding to
> mark that this is not ABI for DT?
> 
> Best regards,
> Krzysztof


Best regards,
Krzysztof