[PATCH v2 0/3] arm64: dts: rockchip: Fix vdec register blocks order on RK3576/RK3588

[PATCH v2 0/3] arm64: dts: rockchip: Fix vdec register blocks order on RK3576/RK3588

[Cristian Ciocaltea]

When building device trees for the RK3576 based boards, DTC shows the
following complaint:

  rk3576.dtsi:1282.30-1304.5: Warning (simple_bus_reg): /soc/video-codec@27b00000: simple-bus unit address format error, expected "27b00100"

The first patch updates 'reg-names' property in rockchip,vdec.yaml to
allow providing the register blocks in the expected address-based order
for RK3576 & RK3588 video decoder, while the following two patches
correct 'reg' and 'reg-names' accordingly for the impacted vdec DT nodes.

Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
---
Changes in v2:
- Added patch for updating rockchip,vdec.yaml binding
- Added patch for updating RK3588 vdec nodes
- Link to v1: https://lore.kernel.org/r/20260223-vdec-reg-order-rk3576-v1-1-560976566bd3@collabora.com

---
Cristian Ciocaltea (3):
      media: dt-bindings: rockchip,vdec: Correct reg-names order for RK35{76,88}
      arm64: dts: rockchip: Fix vdec register blocks order on RK3576
      arm64: dts: rockchip: Fix vdec register blocks order on RK3588

 Documentation/devicetree/bindings/media/rockchip,vdec.yaml | 13 ++++++-------
 arch/arm64/boot/dts/rockchip/rk3576.dtsi                   |  6 +++---
 arch/arm64/boot/dts/rockchip/rk3588-base.dtsi              | 12 ++++++------
 3 files changed, 15 insertions(+), 16 deletions(-)
---
base-commit: 6de23f81a5e08be8fbf5e8d7e9febc72a5b5f27f
change-id: 20260223-vdec-reg-order-rk3576-cc2ec6e05e98

[PATCH v2 1/3] media: dt-bindings: rockchip,vdec: Correct reg-names order for RK35{76,88}

[Cristian Ciocaltea]

Update 'reg-names' property to allow providing the register blocks in
the expected address-based order for RK3576 & RK3588 video decoder and,
consequently, fix some DTC warnings.

Additionally, drop the 'reg' description items as the order is not fixed
anymore, while the information they offer is not very relevant anyway.

Fixes: c6ffb7e1fb90 ("media: dt-bindings: rockchip: Document RK3588 Video Decoder bindings")
Fixes: a5c4a6526476 ("media: dt-bindings: rockchip: Add RK3576 Video Decoder bindings")
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
---
 Documentation/devicetree/bindings/media/rockchip,vdec.yaml | 13 ++++++-------
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/Documentation/devicetree/bindings/media/rockchip,vdec.yaml b/Documentation/devicetree/bindings/media/rockchip,vdec.yaml
index 809fda45b3bd..2d3164a2882b 100644
--- a/Documentation/devicetree/bindings/media/rockchip,vdec.yaml
+++ b/Documentation/devicetree/bindings/media/rockchip,vdec.yaml
@@ -28,16 +28,15 @@ properties:
 
   reg:
     minItems: 1
-    items:
-      - description: The function configuration registers base
-      - description: The link table configuration registers base
-      - description: The cache configuration registers base
+    maxItems: 3
 
   reg-names:
-    items:
+    oneOf:
       - const: function
-      - const: link
-      - const: cache
+      - items:
+          - const: link
+          - const: function
+          - const: cache
 
   interrupts:
     maxItems: 1

-- 
2.52.0

[PATCH v2 2/3] arm64: dts: rockchip: Fix vdec register blocks order on RK3576

[Cristian Ciocaltea]

When building device trees for the RK3576 based boards, DTC shows the
following complaint:

  rk3576.dtsi:1282.30-1304.5: Warning (simple_bus_reg): /soc/video-codec@27b00000: simple-bus unit address format error, expected "27b00100"

Provide the register blocks in the expected address-based order.

Fixes: da0de806d8b4 ("arm64: dts: rockchip: Add the vdpu383 Video Decoder on rk3576")
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
---
 arch/arm64/boot/dts/rockchip/rk3576.dtsi | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/arch/arm64/boot/dts/rockchip/rk3576.dtsi b/arch/arm64/boot/dts/rockchip/rk3576.dtsi
index 49ccdf12ef7e..45eb0d053a6f 100644
--- a/arch/arm64/boot/dts/rockchip/rk3576.dtsi
+++ b/arch/arm64/boot/dts/rockchip/rk3576.dtsi
@@ -1281,10 +1281,10 @@ gpu: gpu@27800000 {
 
 		vdec: video-codec@27b00000 {
 			compatible = "rockchip,rk3576-vdec";
-			reg = <0x0 0x27b00100 0x0 0x500>,
-			      <0x0 0x27b00000 0x0 0x100>,
+			reg = <0x0 0x27b00000 0x0 0x100>,
+			      <0x0 0x27b00100 0x0 0x500>,
 			      <0x0 0x27b00600 0x0 0x100>;
-			reg-names = "function", "link", "cache";
+			reg-names = "link", "function", "cache";
 			interrupts = <GIC_SPI 308 IRQ_TYPE_LEVEL_HIGH>;
 			clocks = <&cru ACLK_RKVDEC_ROOT>, <&cru HCLK_RKVDEC>,
 				 <&cru ACLK_RKVDEC_ROOT_BAK>, <&cru CLK_RKVDEC_CORE>,

-- 
2.52.0

[PATCH v2 3/3] arm64: dts: rockchip: Fix vdec register blocks order on RK3588

[Cristian Ciocaltea]

Provide the register blocks in the expected address-based order for
vdec0 & vdec1 nodes.  This aligns with a similar fix for RK3576 where
DTC complained about the bus address format.

Fixes: f61731bd6062 ("arm64: dts: rockchip: Add the vdpu381 Video Decoders on RK3588")
Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
---
 arch/arm64/boot/dts/rockchip/rk3588-base.dtsi | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/arch/arm64/boot/dts/rockchip/rk3588-base.dtsi b/arch/arm64/boot/dts/rockchip/rk3588-base.dtsi
index 7fe9593d8c19..b95129f85aba 100644
--- a/arch/arm64/boot/dts/rockchip/rk3588-base.dtsi
+++ b/arch/arm64/boot/dts/rockchip/rk3588-base.dtsi
@@ -1355,10 +1355,10 @@ vepu121_3_mmu: iommu@fdbac800 {
 
 	vdec0: video-codec@fdc38000 {
 		compatible = "rockchip,rk3588-vdec";
-		reg = <0x0 0xfdc38100 0x0 0x500>,
-		      <0x0 0xfdc38000 0x0 0x100>,
+		reg = <0x0 0xfdc38000 0x0 0x100>,
+		      <0x0 0xfdc38100 0x0 0x500>,
 		      <0x0 0xfdc38600 0x0 0x100>;
-		reg-names = "function", "link", "cache";
+		reg-names = "link", "function", "cache";
 		interrupts = <GIC_SPI 95 IRQ_TYPE_LEVEL_HIGH 0>;
 		clocks = <&cru ACLK_RKVDEC0>, <&cru HCLK_RKVDEC0>, <&cru CLK_RKVDEC0_CA>,
 			 <&cru CLK_RKVDEC0_CORE>, <&cru CLK_RKVDEC0_HEVC_CA>;
@@ -1387,10 +1387,10 @@ vdec0_mmu: iommu@fdc38700 {
 
 	vdec1: video-codec@fdc40000 {
 		compatible = "rockchip,rk3588-vdec";
-		reg = <0x0 0xfdc40100 0x0 0x500>,
-		      <0x0 0xfdc40000 0x0 0x100>,
+		reg = <0x0 0xfdc40000 0x0 0x100>,
+		      <0x0 0xfdc40100 0x0 0x500>,
 		      <0x0 0xfdc40600 0x0 0x100>;
-		reg-names = "function", "link", "cache";
+		reg-names = "link", "function", "cache";
 		interrupts = <GIC_SPI 97 IRQ_TYPE_LEVEL_HIGH 0>;
 		clocks = <&cru ACLK_RKVDEC1>, <&cru HCLK_RKVDEC1>, <&cru CLK_RKVDEC1_CA>,
 			 <&cru CLK_RKVDEC1_CORE>, <&cru CLK_RKVDEC1_HEVC_CA>;

-- 
2.52.0

Re: [PATCH v2 1/3] media: dt-bindings: rockchip,vdec: Correct reg-names order for RK35{76,88}

[Heiko Stübner]

Am Montag, 23. Februar 2026, 20:49:49 Mitteleuropäische Normalzeit schrieb Cristian Ciocaltea:
> Update 'reg-names' property to allow providing the register blocks in
> the expected address-based order for RK3576 & RK3588 video decoder and,
> consequently, fix some DTC warnings.

nit: I think this could use an additional explanation, describing that the
individual components got moved around inside the vdec io block
at least between rk3588 and rk3576.


> Additionally, drop the 'reg' description items as the order is not fixed
> anymore, while the information they offer is not very relevant anyway.
> 
> Fixes: c6ffb7e1fb90 ("media: dt-bindings: rockchip: Document RK3588 Video Decoder bindings")
> Fixes: a5c4a6526476 ("media: dt-bindings: rockchip: Add RK3576 Video Decoder bindings")
> Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
> ---
>  Documentation/devicetree/bindings/media/rockchip,vdec.yaml | 13 ++++++-------
>  1 file changed, 6 insertions(+), 7 deletions(-)
> 
> diff --git a/Documentation/devicetree/bindings/media/rockchip,vdec.yaml b/Documentation/devicetree/bindings/media/rockchip,vdec.yaml
> index 809fda45b3bd..2d3164a2882b 100644
> --- a/Documentation/devicetree/bindings/media/rockchip,vdec.yaml
> +++ b/Documentation/devicetree/bindings/media/rockchip,vdec.yaml
> @@ -28,16 +28,15 @@ properties:
>  
>    reg:
>      minItems: 1
> -    items:
> -      - description: The function configuration registers base
> -      - description: The link table configuration registers base
> -      - description: The cache configuration registers base
> +    maxItems: 3
>  
>    reg-names:
> -    items:
> +    oneOf:
>        - const: function
> -      - const: link
> -      - const: cache
> +      - items:
> +          - const: link
> +          - const: function
> +          - const: cache
>  
>    interrupts:
>      maxItems: 1
> 
> 

Re: [PATCH v2 1/3] media: dt-bindings: rockchip,vdec: Correct reg-names order for RK35{76,88}

[Krzysztof Kozlowski]

On Mon, Feb 23, 2026 at 09:49:49PM +0200, Cristian Ciocaltea wrote:
> Update 'reg-names' property to allow providing the register blocks in
> the expected address-based order for RK3576 & RK3588 video decoder and,

What? The register blocks SHOULD NOT be provided in address-based order.
There was never such coding style, never ever we gave such hint and if
you found Qcom discussions you would see me and Rob objecting to it.

> consequently, fix some DTC warnings.
> 
> Additionally, drop the 'reg' description items as the order is not fixed
> anymore, while the information they offer is not very relevant anyway.
> 
> Fixes: c6ffb7e1fb90 ("media: dt-bindings: rockchip: Document RK3588 Video Decoder bindings")
> Fixes: a5c4a6526476 ("media: dt-bindings: rockchip: Add RK3576 Video Decoder bindings")

No, nothing to be fixed here. Describe the bug, because missing order
to address-based is for sure not a bug.

> Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
> ---
>  Documentation/devicetree/bindings/media/rockchip,vdec.yaml | 13 ++++++-------
>  1 file changed, 6 insertions(+), 7 deletions(-)
> 
> diff --git a/Documentation/devicetree/bindings/media/rockchip,vdec.yaml b/Documentation/devicetree/bindings/media/rockchip,vdec.yaml
> index 809fda45b3bd..2d3164a2882b 100644
> --- a/Documentation/devicetree/bindings/media/rockchip,vdec.yaml
> +++ b/Documentation/devicetree/bindings/media/rockchip,vdec.yaml
> @@ -28,16 +28,15 @@ properties:
>  
>    reg:
>      minItems: 1
> -    items:
> -      - description: The function configuration registers base
> -      - description: The link table configuration registers base
> -      - description: The cache configuration registers base
> +    maxItems: 3
>  
>    reg-names:
> -    items:
> +    oneOf:
>        - const: function
> -      - const: link
> -      - const: cache
> +      - items:
> +          - const: link
> +          - const: function
> +          - const: cache

No, ABI break without point.

>  
>    interrupts:
>      maxItems: 1
> 
> -- 
> 2.52.0
> 

Re: [PATCH v2 3/3] arm64: dts: rockchip: Fix vdec register blocks order on RK3588

[Krzysztof Kozlowski]

On Mon, Feb 23, 2026 at 09:49:51PM +0200, Cristian Ciocaltea wrote:
> Provide the register blocks in the expected address-based order for

No, there is NO such convention. Don't make up rules.

> vdec0 & vdec1 nodes.  This aligns with a similar fix for RK3576 where
> DTC complained about the bus address format.
> 
> Fixes: f61731bd6062 ("arm64: dts: rockchip: Add the vdpu381 Video Decoders on RK3588")

Not true, otherwise explain user-visible bug.

> Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
> ---
>  arch/arm64/boot/dts/rockchip/rk3588-base.dtsi | 12 ++++++------
>  1 file changed, 6 insertions(+), 6 deletions(-)
> 
> diff --git a/arch/arm64/boot/dts/rockchip/rk3588-base.dtsi b/arch/arm64/boot/dts/rockchip/rk3588-base.dtsi
> index 7fe9593d8c19..b95129f85aba 100644
> --- a/arch/arm64/boot/dts/rockchip/rk3588-base.dtsi
> +++ b/arch/arm64/boot/dts/rockchip/rk3588-base.dtsi
> @@ -1355,10 +1355,10 @@ vepu121_3_mmu: iommu@fdbac800 {
>  
>  	vdec0: video-codec@fdc38000 {
>  		compatible = "rockchip,rk3588-vdec";
> -		reg = <0x0 0xfdc38100 0x0 0x500>,
> -		      <0x0 0xfdc38000 0x0 0x100>,
> +		reg = <0x0 0xfdc38000 0x0 0x100>,
> +		      <0x0 0xfdc38100 0x0 0x500>,
>  		      <0x0 0xfdc38600 0x0 0x100>;

NAK, impacts DTS users without any reason.

Best regards,
Krzysztof

Re: [PATCH v2 2/3] arm64: dts: rockchip: Fix vdec register blocks order on RK3576

[Krzysztof Kozlowski]

On Mon, Feb 23, 2026 at 09:49:50PM +0200, Cristian Ciocaltea wrote:
> When building device trees for the RK3576 based boards, DTC shows the
> following complaint:
> 
>   rk3576.dtsi:1282.30-1304.5: Warning (simple_bus_reg): /soc/video-codec@27b00000: simple-bus unit address format error, expected "27b00100"

So you need to fix the unit address. This is what the warning tells you.

> 
> Provide the register blocks in the expected address-based order.
> 
> Fixes: da0de806d8b4 ("arm64: dts: rockchip: Add the vdpu383 Video Decoder on rk3576")
> Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
> ---
>  arch/arm64/boot/dts/rockchip/rk3576.dtsi | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/arm64/boot/dts/rockchip/rk3576.dtsi b/arch/arm64/boot/dts/rockchip/rk3576.dtsi
> index 49ccdf12ef7e..45eb0d053a6f 100644
> --- a/arch/arm64/boot/dts/rockchip/rk3576.dtsi
> +++ b/arch/arm64/boot/dts/rockchip/rk3576.dtsi
> @@ -1281,10 +1281,10 @@ gpu: gpu@27800000 {
>  
>  		vdec: video-codec@27b00000 {
>  			compatible = "rockchip,rk3576-vdec";
> -			reg = <0x0 0x27b00100 0x0 0x500>,
> -			      <0x0 0x27b00000 0x0 0x100>,
> +			reg = <0x0 0x27b00000 0x0 0x100>,
> +			      <0x0 0x27b00100 0x0 0x500>,
>  			      <0x0 0x27b00600 0x0 0x100>;

The main block, so probably the lowest address as in unit address,
should be used, but this ship has sailed. You shipped this DTS, because
the order of items is FIXED. Your binding change is clearly incorrect
and ABI break without explanation and without reason, so you cannot use
such as an argument here.

You cannot change it without impact, so just fix the unit address.

Best regards,
Krzysztof

Re: [PATCH v2 1/3] media: dt-bindings: rockchip,vdec: Correct reg-names order for RK35{76,88}

[Nicolas Dufresne]

[-- Attachment #1: Type: text/plain, Size: 2824 bytes --]

Hi,

Le mardi 24 février 2026 à 08:17 +0100, Krzysztof Kozlowski a écrit :
> On Mon, Feb 23, 2026 at 09:49:49PM +0200, Cristian Ciocaltea wrote:
> > Update 'reg-names' property to allow providing the register blocks in
> > the expected address-based order for RK3576 & RK3588 video decoder and,
> 
> What? The register blocks SHOULD NOT be provided in address-based order.
> There was never such coding style, never ever we gave such hint and if
> you found Qcom discussions you would see me and Rob objecting to it.

Would be nice to make a suggestion to how to fix this warning differently.

rk3576.dtsi:1282.30-1304.5: Warning (simple_bus_reg): /soc/video-codec@27b00000: simple-bus unit address format error, expected "27b00100"


> 
> > consequently, fix some DTC warnings.
> > 
> > Additionally, drop the 'reg' description items as the order is not fixed
> > anymore, while the information they offer is not very relevant anyway.
> > 
> > Fixes: c6ffb7e1fb90 ("media: dt-bindings: rockchip: Document RK3588 Video Decoder bindings")
> > Fixes: a5c4a6526476 ("media: dt-bindings: rockchip: Add RK3576 Video Decoder bindings")
> 
> No, nothing to be fixed here. Describe the bug, because missing order
> to address-based is for sure not a bug.

The warning should be part of this this commit message, since the cover letter
is not going into git.

> 
> > Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
> > ---
> >  Documentation/devicetree/bindings/media/rockchip,vdec.yaml | 13 ++++++-------
> >  1 file changed, 6 insertions(+), 7 deletions(-)
> > 
> > diff --git a/Documentation/devicetree/bindings/media/rockchip,vdec.yaml b/Documentation/devicetree/bindings/media/rockchip,vdec.yaml
> > index 809fda45b3bd..2d3164a2882b 100644
> > --- a/Documentation/devicetree/bindings/media/rockchip,vdec.yaml
> > +++ b/Documentation/devicetree/bindings/media/rockchip,vdec.yaml
> > @@ -28,16 +28,15 @@ properties:
> >  
> >    reg:
> >      minItems: 1
> > -    items:
> > -      - description: The function configuration registers base
> > -      - description: The link table configuration registers base
> > -      - description: The cache configuration registers base
> > +    maxItems: 3
> >  
> >    reg-names:
> > -    items:
> > +    oneOf:
> >        - const: function
> > -      - const: link
> > -      - const: cache
> > +      - items:
> > +          - const: link
> > +          - const: function
> > +          - const: cache
> 
> No, ABI break without point.

Its not released yet, otherwise both order would need to be allowed.

Nicolas

> 
> >  
> >    interrupts:
> >      maxItems: 1
> > 
> > -- 
> > 2.52.0
> > 

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

Re: [PATCH v2 2/3] arm64: dts: rockchip: Fix vdec register blocks order on RK3576

[Nicolas Dufresne]

[-- Attachment #1: Type: text/plain, Size: 2193 bytes --]

Hi,

Le mardi 24 février 2026 à 08:22 +0100, Krzysztof Kozlowski a écrit :
> On Mon, Feb 23, 2026 at 09:49:50PM +0200, Cristian Ciocaltea wrote:
> > When building device trees for the RK3576 based boards, DTC shows the
> > following complaint:
> > 
> >   rk3576.dtsi:1282.30-1304.5: Warning (simple_bus_reg): /soc/video-codec@27b00000: simple-bus unit address format error, expected "27b00100"
> 
> So you need to fix the unit address. This is what the warning tells you.

The unit address is where the register range starts. Picking a random point
inside the range is just bad idea. For anyone that uses spec to develop these
drivers and device tree, its just plain difficult and error prone.

> 
> > 
> > Provide the register blocks in the expected address-based order.
> > 
> > Fixes: da0de806d8b4 ("arm64: dts: rockchip: Add the vdpu383 Video Decoder on rk3576")
> > Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
> > ---
> >  arch/arm64/boot/dts/rockchip/rk3576.dtsi | 6 +++---
> >  1 file changed, 3 insertions(+), 3 deletions(-)
> > 
> > diff --git a/arch/arm64/boot/dts/rockchip/rk3576.dtsi b/arch/arm64/boot/dts/rockchip/rk3576.dtsi
> > index 49ccdf12ef7e..45eb0d053a6f 100644
> > --- a/arch/arm64/boot/dts/rockchip/rk3576.dtsi
> > +++ b/arch/arm64/boot/dts/rockchip/rk3576.dtsi
> > @@ -1281,10 +1281,10 @@ gpu: gpu@27800000 {
> >  
> >  		vdec: video-codec@27b00000 {
> >  			compatible = "rockchip,rk3576-vdec";
> > -			reg = <0x0 0x27b00100 0x0 0x500>,
> > -			      <0x0 0x27b00000 0x0 0x100>,
> > +			reg = <0x0 0x27b00000 0x0 0x100>,
> > +			      <0x0 0x27b00100 0x0 0x500>,
> >  			      <0x0 0x27b00600 0x0 0x100>;
> 
> The main block, so probably the lowest address as in unit address,
> should be used, but this ship has sailed. You shipped this DTS, because
> the order of items is FIXED. Your binding change is clearly incorrect
> and ABI break without explanation and without reason, so you cannot use
> such as an argument here.

Its not released yet.


> You cannot change it without impact, so just fix the unit address.

I completely disagree with that.

Nicolas


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

Re: [PATCH v2 1/3] media: dt-bindings: rockchip,vdec: Correct reg-names order for RK35{76,88}

[Krzysztof Kozlowski]

On 24/02/2026 14:14, Nicolas Dufresne wrote:
> Hi,
> 
> Le mardi 24 février 2026 à 08:17 +0100, Krzysztof Kozlowski a écrit :
>> On Mon, Feb 23, 2026 at 09:49:49PM +0200, Cristian Ciocaltea wrote:
>>> Update 'reg-names' property to allow providing the register blocks in
>>> the expected address-based order for RK3576 & RK3588 video decoder and,
>>
>> What? The register blocks SHOULD NOT be provided in address-based order.
>> There was never such coding style, never ever we gave such hint and if
>> you found Qcom discussions you would see me and Rob objecting to it.
> 
> Would be nice to make a suggestion to how to fix this warning differently.
> 
> rk3576.dtsi:1282.30-1304.5: Warning (simple_bus_reg): /soc/video-codec@27b00000: simple-bus unit address format error, expected "27b00100"

I provided the suggestion in the commit mentioning the issue. Nothing
here suggests any warning, so why would I suggest some steps?

> 
> 
>>
>>> consequently, fix some DTC warnings.
>>>
>>> Additionally, drop the 'reg' description items as the order is not fixed
>>> anymore, while the information they offer is not very relevant anyway.
>>>
>>> Fixes: c6ffb7e1fb90 ("media: dt-bindings: rockchip: Document RK3588 Video Decoder bindings")
>>> Fixes: a5c4a6526476 ("media: dt-bindings: rockchip: Add RK3576 Video Decoder bindings")
>>
>> No, nothing to be fixed here. Describe the bug, because missing order
>> to address-based is for sure not a bug.
> 
> The warning should be part of this this commit message, since the cover letter
> is not going into git.
> 
>>
>>> Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
>>> ---
>>>  Documentation/devicetree/bindings/media/rockchip,vdec.yaml | 13 ++++++-------
>>>  1 file changed, 6 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/Documentation/devicetree/bindings/media/rockchip,vdec.yaml b/Documentation/devicetree/bindings/media/rockchip,vdec.yaml
>>> index 809fda45b3bd..2d3164a2882b 100644
>>> --- a/Documentation/devicetree/bindings/media/rockchip,vdec.yaml
>>> +++ b/Documentation/devicetree/bindings/media/rockchip,vdec.yaml
>>> @@ -28,16 +28,15 @@ properties:
>>>  
>>>    reg:
>>>      minItems: 1
>>> -    items:
>>> -      - description: The function configuration registers base
>>> -      - description: The link table configuration registers base
>>> -      - description: The cache configuration registers base
>>> +    maxItems: 3
>>>  
>>>    reg-names:
>>> -    items:
>>> +    oneOf:
>>>        - const: function
>>> -      - const: link
>>> -      - const: cache
>>> +      - items:
>>> +          - const: link
>>> +          - const: function
>>> +          - const: cache
>>
>> No, ABI break without point.
> 
> Its not released yet, otherwise both order would need to be allowed.

Nothing in commit msg explains that and we expressed such expectation
multiple times. Otherwise you get above comment.

Best regards,
Krzysztof

Re: [PATCH v2 1/3] media: dt-bindings: rockchip,vdec: Correct reg-names order for RK35{76,88}

[Krzysztof Kozlowski]

On 24/02/2026 14:14, Nicolas Dufresne wrote:
>>>  
>>>    reg-names:
>>> -    items:
>>> +    oneOf:
>>>        - const: function
>>> -      - const: link
>>> -      - const: cache
>>> +      - items:
>>> +          - const: link
>>> +          - const: function
>>> +          - const: cache
>>
>> No, ABI break without point.
> 
> Its not released yet, otherwise both order would need to be allowed.
> 
Also, not true.

git describe --contains  c6ffb7e1fb90
v6.17-rc1~90^2~3

So clearly released.

Best regards,
Krzysztof

Re: [PATCH v2 2/3] arm64: dts: rockchip: Fix vdec register blocks order on RK3576

[Krzysztof Kozlowski]

On 24/02/2026 14:17, Nicolas Dufresne wrote:
> Hi,
> 
> Le mardi 24 février 2026 à 08:22 +0100, Krzysztof Kozlowski a écrit :
>> On Mon, Feb 23, 2026 at 09:49:50PM +0200, Cristian Ciocaltea wrote:
>>> When building device trees for the RK3576 based boards, DTC shows the
>>> following complaint:
>>>
>>>   rk3576.dtsi:1282.30-1304.5: Warning (simple_bus_reg): /soc/video-codec@27b00000: simple-bus unit address format error, expected "27b00100"
>>
>> So you need to fix the unit address. This is what the warning tells you.
> 
> The unit address is where the register range starts. Picking a random point
> inside the range is just bad idea.

Of course and that's why you run tools BEFORE you send patches to tell
you that.

> For anyone that uses spec to develop these
> drivers and device tree, its just plain difficult and error prone.
> 
>>
>>>
>>> Provide the register blocks in the expected address-based order.
>>>
>>> Fixes: da0de806d8b4 ("arm64: dts: rockchip: Add the vdpu383 Video Decoder on rk3576")
>>> Signed-off-by: Cristian Ciocaltea <cristian.ciocaltea@collabora.com>
>>> ---
>>>  arch/arm64/boot/dts/rockchip/rk3576.dtsi | 6 +++---
>>>  1 file changed, 3 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/arch/arm64/boot/dts/rockchip/rk3576.dtsi b/arch/arm64/boot/dts/rockchip/rk3576.dtsi
>>> index 49ccdf12ef7e..45eb0d053a6f 100644
>>> --- a/arch/arm64/boot/dts/rockchip/rk3576.dtsi
>>> +++ b/arch/arm64/boot/dts/rockchip/rk3576.dtsi
>>> @@ -1281,10 +1281,10 @@ gpu: gpu@27800000 {
>>>  
>>>  		vdec: video-codec@27b00000 {
>>>  			compatible = "rockchip,rk3576-vdec";
>>> -			reg = <0x0 0x27b00100 0x0 0x500>,
>>> -			      <0x0 0x27b00000 0x0 0x100>,
>>> +			reg = <0x0 0x27b00000 0x0 0x100>,
>>> +			      <0x0 0x27b00100 0x0 0x500>,
>>>  			      <0x0 0x27b00600 0x0 0x100>;
>>
>> The main block, so probably the lowest address as in unit address,
>> should be used, but this ship has sailed. You shipped this DTS, because
>> the order of items is FIXED. Your binding change is clearly incorrect
>> and ABI break without explanation and without reason, so you cannot use
>> such as an argument here.
> 
> Its not released yet.

Nothing in commit msg explained that and your previous comment about
binding was false. Binding WAS released.

I am not going to check every claim you make whether it is true or not.

Best regards,
Krzysztof

Re: [PATCH v2 1/3] media: dt-bindings: rockchip,vdec: Correct reg-names order for RK35{76,88}

[Nicolas Dufresne]

[-- Attachment #1: Type: text/plain, Size: 938 bytes --]

Le mardi 24 février 2026 à 14:18 +0100, Krzysztof Kozlowski a écrit :
> On 24/02/2026 14:14, Nicolas Dufresne wrote:
> > > >  
> > > >    reg-names:
> > > > -    items:
> > > > +    oneOf:
> > > >        - const: function
> > > > -      - const: link
> > > > -      - const: cache
> > > > +      - items:
> > > > +          - const: link
> > > > +          - const: function
> > > > +          - const: cache
> > > 
> > > No, ABI break without point.
> > 
> > Its not released yet, otherwise both order would need to be allowed.
> > 
> Also, not true.
> 
> git describe --contains  c6ffb7e1fb90
> v6.17-rc1~90^2~3

Ack, I thought this was part of 7.0, it would have been logical to be part of
7.0-rc1 I must say. Whatever happen, backward compat will unfortunatly be
needed.

Nicolas

> 
> So clearly released.
> 
> Best regards,
> Krzysztof

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 228 bytes --]

Re: [PATCH v2 1/3] media: dt-bindings: rockchip,vdec: Correct reg-names order for RK35{76,88}

[Krzysztof Kozlowski]

On 24/02/2026 14:48, Nicolas Dufresne wrote:
> Le mardi 24 février 2026 à 14:18 +0100, Krzysztof Kozlowski a écrit :
>> On 24/02/2026 14:14, Nicolas Dufresne wrote:
>>>>>  
>>>>>    reg-names:
>>>>> -    items:
>>>>> +    oneOf:
>>>>>        - const: function
>>>>> -      - const: link
>>>>> -      - const: cache
>>>>> +      - items:
>>>>> +          - const: link
>>>>> +          - const: function
>>>>> +          - const: cache
>>>>
>>>> No, ABI break without point.
>>>
>>> Its not released yet, otherwise both order would need to be allowed.
>>>
>> Also, not true.
>>
>> git describe --contains  c6ffb7e1fb90
>> v6.17-rc1~90^2~3
> 
> Ack, I thought this was part of 7.0, it would have been logical to be part of
> 7.0-rc1 I must say. Whatever happen, backward compat will unfortunatly be
> needed.


I was here checking the binding, so the ABI. DTS maybe was in v7.0-rc1
indeed. Anyway, the commit msg should describe the impact on users of
this ABI, at least known users.

Best regards,
Krzysztof

Re: [PATCH v2 1/3] media: dt-bindings: rockchip,vdec: Correct reg-names order for RK35{76,88}

[Cristian Ciocaltea]

On 2/24/26 3:54 PM, Krzysztof Kozlowski wrote:
> On 24/02/2026 14:48, Nicolas Dufresne wrote:
>> Le mardi 24 février 2026 à 14:18 +0100, Krzysztof Kozlowski a écrit :
>>> On 24/02/2026 14:14, Nicolas Dufresne wrote:
>>>>>>  
>>>>>>    reg-names:
>>>>>> -    items:
>>>>>> +    oneOf:
>>>>>>        - const: function
>>>>>> -      - const: link
>>>>>> -      - const: cache
>>>>>> +      - items:
>>>>>> +          - const: link
>>>>>> +          - const: function
>>>>>> +          - const: cache
>>>>>
>>>>> No, ABI break without point.
>>>>
>>>> Its not released yet, otherwise both order would need to be allowed.
>>>>
>>> Also, not true.
>>>
>>> git describe --contains  c6ffb7e1fb90
>>> v6.17-rc1~90^2~3
>>
>> Ack, I thought this was part of 7.0, it would have been logical to be part of
>> 7.0-rc1 I must say. Whatever happen, backward compat will unfortunatly be
>> needed.
> 
> 
> I was here checking the binding, so the ABI. DTS maybe was in v7.0-rc1
> indeed. Anyway, the commit msg should describe the impact on users of
> this ABI, at least known users.

My bad, I only checked the DTS changes which landed in v7.0-rc1, and wrongly
assumed the binding also landed about the same time.  I think it's unlikely
there are any (external) users of this binding, but obviously we cannot be 100%
sure.

Would it be acceptable to extend the binding so that it allows both the current
and the new proposed order to coexist?

Thanks,
Cristian

Re: [PATCH v2 1/3] media: dt-bindings: rockchip,vdec: Correct reg-names order for RK35{76,88}

[Krzysztof Kozlowski]

On 24/02/2026 15:07, Cristian Ciocaltea wrote:
> On 2/24/26 3:54 PM, Krzysztof Kozlowski wrote:
>> On 24/02/2026 14:48, Nicolas Dufresne wrote:
>>> Le mardi 24 février 2026 à 14:18 +0100, Krzysztof Kozlowski a écrit :
>>>> On 24/02/2026 14:14, Nicolas Dufresne wrote:
>>>>>>>  
>>>>>>>    reg-names:
>>>>>>> -    items:
>>>>>>> +    oneOf:
>>>>>>>        - const: function
>>>>>>> -      - const: link
>>>>>>> -      - const: cache
>>>>>>> +      - items:
>>>>>>> +          - const: link
>>>>>>> +          - const: function
>>>>>>> +          - const: cache
>>>>>>
>>>>>> No, ABI break without point.
>>>>>
>>>>> Its not released yet, otherwise both order would need to be allowed.
>>>>>
>>>> Also, not true.
>>>>
>>>> git describe --contains  c6ffb7e1fb90
>>>> v6.17-rc1~90^2~3
>>>
>>> Ack, I thought this was part of 7.0, it would have been logical to be part of
>>> 7.0-rc1 I must say. Whatever happen, backward compat will unfortunatly be
>>> needed.
>>
>>
>> I was here checking the binding, so the ABI. DTS maybe was in v7.0-rc1
>> indeed. Anyway, the commit msg should describe the impact on users of
>> this ABI, at least known users.
> 
> My bad, I only checked the DTS changes which landed in v7.0-rc1, and wrongly
> assumed the binding also landed about the same time.  I think it's unlikely
> there are any (external) users of this binding, but obviously we cannot be 100%
> sure.
> 
> Would it be acceptable to extend the binding so that it allows both the current
> and the new proposed order to coexist?

Depends also on the drivers and that's your "users analysis" should
cover. If there are no known released users of relatively new ABI, it is
fine to propose the change with such explanation and reason - poor
hardware description because convention expects the main device's
address to be used as first "reg" entry.

You can have also oneOf with older list "deprecated: true", if want to
keep any users unaffected.


Best regards,
Krzysztof