Re: [PATCH 5/6] arm64: dts: imx8mm: add PCIe support

[Rob Herring <robh+dt@kernel.org>]

On Fri, Jul 23, 2021 at 2:50 PM Tim Harvey <tharvey@gateworks.com> wrote:
>
> Add PCIe node for PCIe support.
>
> Signed-off-by: Tim Harvey <tharvey@gateworks.com>
> ---
>  arch/arm64/boot/dts/freescale/imx8mm.dtsi | 36 +++++++++++++++++++++++
>  1 file changed, 36 insertions(+)
>
> diff --git a/arch/arm64/boot/dts/freescale/imx8mm.dtsi b/arch/arm64/boot/dts/freescale/imx8mm.dtsi
> index 3bec6b8d52a0..45017f50a11b 100644
> --- a/arch/arm64/boot/dts/freescale/imx8mm.dtsi
> +++ b/arch/arm64/boot/dts/freescale/imx8mm.dtsi
> @@ -1134,6 +1134,10 @@
>                                 reg = <0x32e50200 0x200>;
>                         };
>
> +                       pcie_phy: pcie-phy@32f00000 {
> +                                 compatible = "fsl,imx7d-pcie-phy";
> +                                 reg = <0x32f00000 0x10000>;

The phy really has 64KB worth of registers? This wastes virtual space
too, but I guess that's 'free' on 64-bit.

> +                       };
>                 };
>
>                 dma_apbh: dma-controller@33000000 {
> @@ -1233,5 +1237,37 @@
>                         reg = <0x3d800000 0x400000>;
>                         interrupts = <GIC_SPI 98 IRQ_TYPE_LEVEL_HIGH>;
>                 };
> +
> +               pcie0: pcie@33800000 {
> +                       compatible = "fsl,imx8mm-pcie";
> +                       reg = <0x33800000 0x400000>,
> +                             <0x1ff00000 0x80000>;
> +                       reg-names = "dbi", "config";

I don't think the DBI space ever has 4MB of registers. And IIRC, only
4KB is used for config space unless ECAM is used.

> +                       #address-cells = <3>;
> +                       #size-cells = <2>;
> +                       device_type = "pci";
> +                       bus-range = <0x00 0xff>;
> +                       ranges = <0x81000000 0 0x00000000 0x1ff80000 0 0x00010000 /* downstream I/O 64KB */
> +                                0x82000000 0 0x18000000 0x18000000 0 0x07f00000>; /* non-prefetchable memory */
> +                       num-lanes = <1>;
> +                       num-viewport = <4>;

This is deprecated and ignored. The driver has gotten smarter and detects this.

> +                       interrupts = <GIC_SPI 122 IRQ_TYPE_LEVEL_HIGH>;
> +                       interrupt-names = "msi";
> +                       #interrupt-cells = <1>;
> +                       interrupt-map-mask = <0 0 0 0x7>;
> +                       interrupt-map = <0 0 0 1 &gic GIC_SPI 125 IRQ_TYPE_LEVEL_HIGH>,
> +                                       <0 0 0 2 &gic GIC_SPI 124 IRQ_TYPE_LEVEL_HIGH>,
> +                                       <0 0 0 3 &gic GIC_SPI 123 IRQ_TYPE_LEVEL_HIGH>,
> +                                       <0 0 0 4 &gic GIC_SPI 122 IRQ_TYPE_LEVEL_HIGH>;
> +                       fsl,max-link-speed = <2>;

There's a standard property for this.

> +                       power-domains = <&pgc_pcie>;
> +                       resets = <&src IMX8MQ_RESET_PCIEPHY>,
> +                                <&src IMX8MQ_RESET_PCIE_CTRL_APPS_EN>,
> +                                <&src IMX8MQ_RESET_PCIE_CTRL_APPS_CLK_REQ>,
> +                                <&src IMX8MQ_RESET_PCIE_CTRL_APPS_TURNOFF>;
> +                       reset-names = "pciephy", "apps", "clkreq", "turnoff";

The phy reset belongs in the phy node.

> +                       fsl,imx7d-pcie-phy = <&pcie_phy>;

Didn't we deprecate this? Either way, use the phy binding.

> +                       status = "disabled";
> +               };
>         };
>  };
> --
> 2.17.1
>