Re: [PATCH v3 03/12] media: dt-bindings: Add binding doc for i.MX8QXP and i.MX8QM ISI

From: Frank Li
Date: Thu Mar 27 2025 - 16:22:22 EST


On Thu, Mar 27, 2025 at 08:44:25PM +0200, Laurent Pinchart wrote:
> Hi Frank,
>
> Thank you for the patch.
>
> On Mon, Feb 10, 2025 at 03:59:22PM -0500, Frank Li wrote:
> > Add binding documentation for i.MX8QXP and i.MX8QM ISI. The clock-names,
> > power-domains, and ports differ significantly from the existing
> > nxp,imx8-isi.yaml. Create a new file to avoid complex if-else branches.
>
> Mixed feelings about having different bindings files for what is
> essentially the same IP, but I won't object.

Rob suggest split it at v1.

https://lore.kernel.org/all/20250203221659.GA130749-robh@xxxxxxxxxx/

"I think this addition is borderline whether it should be its own schema
doc. The if/then schemas are larger than the main part. The ports are
not even the same."

>
> > Add new file to MAINTAINERS.
> >
> > Signed-off-by: Frank Li <Frank.Li@xxxxxxx>
> > ---
> > change from v2 to v3
> > - none
> > change from v1 to v2
> > - create new file for 8qm and 8qxp accroding rob's suggestion.
> > ---
> > .../devicetree/bindings/media/fsl,imx8qm-isi.yaml | 117 +++++++++++++++++++++
> > .../devicetree/bindings/media/fsl,imx8qxp-isi.yaml | 103 ++++++++++++++++++
> > MAINTAINERS | 1 +
> > 3 files changed, 221 insertions(+)
> >
> > diff --git a/Documentation/devicetree/bindings/media/fsl,imx8qm-isi.yaml b/Documentation/devicetree/bindings/media/fsl,imx8qm-isi.yaml
> > new file mode 100644
> > index 0000000000000..61c551673e2a4
> > --- /dev/null
> > +++ b/Documentation/devicetree/bindings/media/fsl,imx8qm-isi.yaml
> > @@ -0,0 +1,117 @@
> > +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> > +%YAML 1.2
> > +---
> > +$id: http://devicetree.org/schemas/media/fsl,imx8qm-isi.yaml#
> > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > +
> > +title: i.MX8QM Image Sensing Interface
> > +
> > +maintainers:
> > + - Frank Li <Frank.Li@xxxxxxx>
> > +
> > +description:
> > + The Image Sensing Interface (ISI) combines image processing pipelines with
> > + DMA engines to process and capture frames originating from a variety of
> > + sources. The inputs to the ISI go through Pixel Link interfaces, and their
> > + number and nature is SoC-dependent. They cover both capture interfaces (MIPI
> > + CSI-2 RX, HDMI RX, ...) and display engine outputs for writeback support.
> > +
> > +properties:
> > + compatible:
> > + enum:
> > + - fsl,imx8qm-isi
> > +
> > + reg:
> > + maxItems: 1
> > +
> > + clocks:
> > + maxItems: 8
> > +
> > + clock-names:
> > + items:
> > + - const: per0
> > + - const: per1
> > + - const: per2
> > + - const: per3
> > + - const: per4
> > + - const: per5
> > + - const: per6
> > + - const: per7
> > +
> > + interrupts:
> > + maxItems: 8
> > +
> > + power-domains:
> > + maxItems: 8
>
> This surprises me. The reference manual does list 8 clocks and
> interrupts, but only 6 channels in the ISI section (for instance in
> 15.6.1.1.4). Which one is wrong ?

Support 6 input, 8 output.

"The crossbar is a 6 input 8 output multiplexer where each output port can
be configured to connect to any of the 6 inputs."

8 irq and clocks is for output dmac.

>
> > +
> > + ports:
> > + $ref: /schemas/graph.yaml#/properties/ports
> > + properties:
> > + port@2:
> > + $ref: /schemas/graph.yaml#/properties/port
> > + description: MIPI CSI-2 RX 0
> > + port@3:
> > + $ref: /schemas/graph.yaml#/properties/port
> > + description: MIPI CSI-2 RX 1
> > + port@4:
> > + $ref: /schemas/graph.yaml#/properties/port
> > + description: HDMI RX
>
> Figure 15-59 in the reference manual list MIPI CSI-2 RX 0 and RX 1 as
> connected to inputs 0 and 1 respectively.

Reference document should be wrong, I reference another internal document

0: display control 0
1: display control 1
2: csi2 rx0
3: csi2 rx1
4: hdmi rx

>
> > +
> > +required:
> > + - compatible
> > + - reg
> > + - clocks
> > + - clock-names
> > + - interrupts
> > + - power-domains
> > + - ports
> > +
> > +additionalProperties: false
> > +
> > +examples:
> > + - |
> > + #include <dt-bindings/interrupt-controller/arm-gic.h>
> > + #include <dt-bindings/interrupt-controller/irq.h>
> > + #include <dt-bindings/clock/imx8-clock.h>
> > + #include <dt-bindings/clock/imx8-lpcg.h>
> > + #include <dt-bindings/firmware/imx/rsrc.h>
> > +
> > + image-controller@58100000 {
> > + compatible = "fsl,imx8qm-isi";
> > + reg = <0x58100000 0x90000>;
>
> The memory map in the reference manual lists the "Pixel DMA" region as
> ending at 0x5817ffff. Shouldn't the length of the region be 0x80000 ?

Yes, it should be 0x80000.

>
> > + interrupts = <GIC_SPI 297 IRQ_TYPE_LEVEL_HIGH>,
> > + <GIC_SPI 298 IRQ_TYPE_LEVEL_HIGH>,
> > + <GIC_SPI 299 IRQ_TYPE_LEVEL_HIGH>,
> > + <GIC_SPI 300 IRQ_TYPE_LEVEL_HIGH>,
> > + <GIC_SPI 301 IRQ_TYPE_LEVEL_HIGH>,
> > + <GIC_SPI 302 IRQ_TYPE_LEVEL_HIGH>,
> > + <GIC_SPI 303 IRQ_TYPE_LEVEL_HIGH>,
> > + <GIC_SPI 304 IRQ_TYPE_LEVEL_HIGH>;
> > + clocks = <&pdma0_lpcg IMX_LPCG_CLK_0>,
> > + <&pdma1_lpcg IMX_LPCG_CLK_0>,
> > + <&pdma2_lpcg IMX_LPCG_CLK_0>,
> > + <&pdma3_lpcg IMX_LPCG_CLK_0>,
> > + <&pdma4_lpcg IMX_LPCG_CLK_0>,
> > + <&pdma5_lpcg IMX_LPCG_CLK_0>,
> > + <&pdma6_lpcg IMX_LPCG_CLK_0>,
> > + <&pdma7_lpcg IMX_LPCG_CLK_0>;
> > + clock-names = "per0", "per1", "per2", "per3",
> > + "per4", "per5", "per6", "per7";
> > + power-domains = <&pd IMX_SC_R_ISI_CH0>, <&pd IMX_SC_R_ISI_CH1>,
> > + <&pd IMX_SC_R_ISI_CH2>, <&pd IMX_SC_R_ISI_CH3>,
> > + <&pd IMX_SC_R_ISI_CH4>, <&pd IMX_SC_R_ISI_CH5>,
> > + <&pd IMX_SC_R_ISI_CH6>, <&pd IMX_SC_R_ISI_CH7>;
> > +
> > + ports {
> > + #address-cells = <1>;
> > + #size-cells = <0>;
> > +
> > + port@2 {
> > + reg = <2>;
> > + endpoint {
> > + remote-endpoint = <&mipi_csi0_out>;
> > + };
> > + };
> > + };
> > + };
> > +...
> > diff --git a/Documentation/devicetree/bindings/media/fsl,imx8qxp-isi.yaml b/Documentation/devicetree/bindings/media/fsl,imx8qxp-isi.yaml
> > new file mode 100644
> > index 0000000000000..818fea0e4679f
> > --- /dev/null
> > +++ b/Documentation/devicetree/bindings/media/fsl,imx8qxp-isi.yaml
> > @@ -0,0 +1,103 @@
> > +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> > +%YAML 1.2
> > +---
> > +$id: http://devicetree.org/schemas/media/fsl,imx8qxp-isi.yaml#
> > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > +
> > +title: i.MX8QXP Image Sensing Interface
> > +
> > +maintainers:
> > + - Frank Li <Frank.Li@xxxxxxx>
> > +
> > +description:
> > + The Image Sensing Interface (ISI) combines image processing pipelines with
> > + DMA engines to process and capture frames originating from a variety of
> > + sources. The inputs to the ISI go through Pixel Link interfaces, and their
> > + number and nature is SoC-dependent. They cover both capture interfaces (MIPI
> > + CSI-2 RX, HDMI RX, ...) and display engine outputs for writeback support.
> > +
> > +properties:
> > + compatible:
> > + enum:
> > + - fsl,imx8qxp-isi
> > +
> > + reg:
> > + maxItems: 1
> > +
> > + clocks:
> > + maxItems: 5
> > +
> > + clock-names:
> > + items:
> > + - const: per0
> > + - const: per4
> > + - const: per5
> > + - const: per6
> > + - const: per7
> > +
> > + interrupts:
> > + maxItems: 5
> > +
> > + power-domains:
> > + maxItems: 5
>
> Here you have 5 channels, while the reference manual lists 8 interrupts
> and 6 channels in the ISI documentation.

QXP should only have 5 irqs and clocks. QM have 8, see above reply.

>
> > +
> > + ports:
> > + $ref: /schemas/graph.yaml#/properties/ports
> > + properties:
> > + port@2:
> > + $ref: /schemas/graph.yaml#/properties/port
> > + description: MIPI CSI-2 RX 0
> > + port@6:
> > + $ref: /schemas/graph.yaml#/properties/port
> > + description: CSI-2 Parallel RX
>
> Table 15-6 in the reference manual lists the parallel port as input 4.

Reference manual is wrong.

>
> > +
> > +required:
> > + - compatible
> > + - reg
> > + - clocks
> > + - clock-names
> > + - interrupts
> > + - power-domains
> > + - ports
> > +
> > +additionalProperties: false
> > +
> > +examples:
> > + - |
> > + #include <dt-bindings/interrupt-controller/arm-gic.h>
> > + #include <dt-bindings/interrupt-controller/irq.h>
> > + #include <dt-bindings/clock/imx8-clock.h>
> > + #include <dt-bindings/clock/imx8-lpcg.h>
> > + #include <dt-bindings/firmware/imx/rsrc.h>
> > +
> > + image-controller@58100000 {
> > + compatible = "fsl,imx8qxp-isi";
> > + reg = <0x58100000 0x90000>;
>
> Same comment here about the registers range.
>
> > + interrupts = <GIC_SPI 297 IRQ_TYPE_LEVEL_HIGH>,
> > + <GIC_SPI 301 IRQ_TYPE_LEVEL_HIGH>,
> > + <GIC_SPI 302 IRQ_TYPE_LEVEL_HIGH>,
> > + <GIC_SPI 303 IRQ_TYPE_LEVEL_HIGH>,
> > + <GIC_SPI 304 IRQ_TYPE_LEVEL_HIGH>;
> > + clocks = <&pdma0_lpcg IMX_LPCG_CLK_0>,
> > + <&pdma4_lpcg IMX_LPCG_CLK_0>,
> > + <&pdma5_lpcg IMX_LPCG_CLK_0>,
> > + <&pdma6_lpcg IMX_LPCG_CLK_0>,
> > + <&pdma7_lpcg IMX_LPCG_CLK_0>;
> > + clock-names = "per0", "per4", "per5", "per6", "per7";
> > + power-domains = <&pd IMX_SC_R_ISI_CH0>, <&pd IMX_SC_R_ISI_CH4>,
> > + <&pd IMX_SC_R_ISI_CH5>, <&pd IMX_SC_R_ISI_CH6>,
> > + <&pd IMX_SC_R_ISI_CH7>;
> > +
> > + ports {
> > + #address-cells = <1>;
> > + #size-cells = <0>;
> > +
> > + port@2 {
> > + reg = <2>;
> > + endpoint {
> > + remote-endpoint = <&mipi_csi0_out>;
> > + };
> > + };
> > + };
> > + };
> > +...
> > diff --git a/MAINTAINERS b/MAINTAINERS
> > index 40d1b7ec30fde..f243257ef7653 100644
> > --- a/MAINTAINERS
> > +++ b/MAINTAINERS
> > @@ -17000,6 +17000,7 @@ NXP i.MX 8M ISI DRIVER
> > M: Laurent Pinchart <laurent.pinchart@xxxxxxxxxxxxxxxx>
> > L: linux-media@xxxxxxxxxxxxxxx
> > S: Maintained
> > +F: Documentation/devicetree/bindings/media/fsl,imx8*-isi.yaml
> > F: Documentation/devicetree/bindings/media/nxp,imx8-isi.yaml
>
> Should nxp,imx8-isi.yaml be renamed to fsl,imx8-isi.yaml ?

Suppose yes, it should match one of compatible string name. This patch
have not touch nxp,imx8-isi.yaml. we may rename it later

Frank
>
> > F: drivers/media/platform/nxp/imx8-isi/
> >
>
> --
> Regards,
>
> Laurent Pinchart