[PATCH 12/14] ARM: tegra: tec: Add PCIe support

Stephen Warren swarren at wwwdotorg.org
Fri Jan 11 11:22:30 EST 2013


On 01/09/2013 01:43 PM, Thierry Reding wrote:
> Enable the first PCIe root port which is connected to an FPGA on the
> Tamonten Evaluation Carrier and add device nodes for each of the PCI
> endpoints available in the standard configuration.

> diff --git a/arch/arm/boot/dts/tegra20-tec.dts b/arch/arm/boot/dts/tegra20-tec.dts

> +	pcie-controller {
> +		vdd-supply = <&pci_vdd_reg>;
> +		pex-clk-supply = <&pci_clk_reg>;
> +		status = "okay";

Sorry this is also really picky. I'd prefer properties that exist in
/include/d files and are overidden here to appear first, followed by new
properties. In other words, move the status property to be first. I
believe/hope all the other (Tegra) .dts files follow this convention.

> +		pci at 1,0 {
> +			bus-range = <0x01 0x0a>;
> +			status = "okay";
> +
> +			pci at 0,0 {
> +				reg = <0x010000 0 0 0 0>;

Hmm. The unit address in that node name doesn't match the address in the
reg property, although I suppose there's nothing we can do about it
since those formats are both defined by the standard PCI binding?

What do the numbers "0,0" represent here; device/function? Is the same
true for the "0,0" in the child nodes?

> +				bus-range = <0x02 0x0a>;
> +
> +				compatible = "plda,pcie";

Are there DT binding documents for all these devices; plda,pcie,
ad,pcie, ad,pcie-test, etc.?

> +				pci at 4,0 {
> +					reg = <0x022000 0 0 0 0>;
> +					bus-range = <0x07 0x07>;
> +
> +					compatible = "ad,pcie";
> +					device_type = "pci";
> +
> +					#address-cells = <3>;
> +					#size-cells = <2>;
> +
> +					pci at 0,0 {
> +						compatible = "opencores,uart";
> +						reg = <0x070000 0 0 0 0>;
> +					};
> +				};

Do you need to include a node for the UART; I can see you need to for
the SPI/I2C controllers so you can instantiate the appropriate devices
on non-probe-able buses, but I think you can just let regular PCI device
probing find the UART, Ethernet MAC, etc., can't you?


More information about the devicetree-discuss mailing list