Re: [PATCH] powerpc/mpc85xx:Add initial device tree support of T104x

[Valentin Longchamp <valentin.longchamp@keymile.com>]

On 09/13/2013 04:53 PM, Kumar Gala wrote:
> On Sep 13, 2013, at 4:14 AM, Valentin Longchamp wrote:
>> On 09/11/2013 08:58 AM, Prabhakar Kushwaha wrote:
>>> +
>>> +&pci0 {
>>> +	compatible = "fsl,t1042-pcie", "fsl,qoriq-pcie-v2.4", "fsl,qoriq-pcie";
>>> +	device_type = "pci";
>>> +	#size-cells = <2>;
>>> +	#address-cells = <3>;
>>> +	bus-range = <0x0 0xff>;
>>> +	interrupts = <20 2 0 0>;
>>> +	fsl,iommu-parent = <&pamu0>;
>>> +	pcie@0 {
>>> +		reg = <0 0 0 0 0>;
>>> +		#interrupt-cells = <1>;
>>> +		#size-cells = <2>;
>>> +		#address-cells = <3>;
>>> +		device_type = "pci";
>>> +		interrupts = <20 2 0 0>;
>>> +		interrupt-map-mask = <0xf800 0 0 7>;
>>> +		interrupt-map = <
>>> +			/* IDSEL 0x0 */
>>> +			0000 0 0 1 &mpic 40 1 0 0
>>> +			0000 0 0 2 &mpic 1 1 0 0
>>> +			0000 0 0 3 &mpic 2 1 0 0
>>> +			0000 0 0 4 &mpic 3 1 0 0
>>> +			>;
>>> +	};
>>> +};
>>> +
>>> +&pci1 {
>>> +	compatible = "fsl,t1042-pcie", "fsl,qoriq-pcie-v2.4", "fsl,qoriq-pcie";
>>> +	device_type = "pci";
>>> +	#size-cells = <2>;
>>> +	#address-cells = <3>;
>>> +	bus-range = <0 0xff>;
>>> +	interrupts = <21 2 0 0>;
>>> +	fsl,iommu-parent = <&pamu0>;
>>> +	pcie@0 {
>>> +		reg = <0 0 0 0 0>;
>>> +		#interrupt-cells = <1>;
>>> +		#size-cells = <2>;
>>> +		#address-cells = <3>;
>>> +		device_type = "pci";
>>> +		interrupts = <21 2 0 0>;
>>> +		interrupt-map-mask = <0xf800 0 0 7>;
>>> +		interrupt-map = <
>>> +			/* IDSEL 0x0 */
>>> +			0000 0 0 1 &mpic 41 1 0 0
>>> +			0000 0 0 2 &mpic 5 1 0 0
>>> +			0000 0 0 3 &mpic 6 1 0 0
>>> +			0000 0 0 4 &mpic 7 1 0 0
>>> +			>;
>>> +	};
>>> +};
>>> +
>>> +&pci2 {
>>> +	compatible = "fsl,t1042-pcie", "fsl,qoriq-pcie-v2.4", "fsl,qoriq-pcie";
>>> +	device_type = "pci";
>>> +	#size-cells = <2>;
>>> +	#address-cells = <3>;
>>> +	bus-range = <0x0 0xff>;
>>> +	interrupts = <22 2 0 0>;
>>> +	fsl,iommu-parent = <&pamu0>;
>>> +	pcie@0 {
>>> +		reg = <0 0 0 0 0>;
>>> +		#interrupt-cells = <1>;
>>> +		#size-cells = <2>;
>>> +		#address-cells = <3>;
>>> +		device_type = "pci";
>>> +		interrupts = <22 2 0 0>;
>>> +		interrupt-map-mask = <0xf800 0 0 7>;
>>> +		interrupt-map = <
>>> +			/* IDSEL 0x0 */
>>> +			0000 0 0 1 &mpic 42 1 0 0
>>> +			0000 0 0 2 &mpic 9 1 0 0
>>> +			0000 0 0 3 &mpic 10 1 0 0
>>> +			0000 0 0 4 &mpic 11 1 0 0
>>> +			>;
>>> +	};
>>> +};
>>> +
>>> +&pci3 {
>>> +	compatible = "fsl,t1042-pcie", "fsl,qoriq-pcie-v2.4", "fsl,qoriq-pcie";
>>> +	device_type = "pci";
>>> +	#size-cells = <2>;
>>> +	#address-cells = <3>;
>>> +	bus-range = <0x0 0xff>;
>>> +	interrupts = <23 2 0 0>;
>>> +	fsl,iommu-parent = <&pamu0>;
>>> +	pcie@0 {
>>> +		reg = <0 0 0 0 0>;
>>> +		#interrupt-cells = <1>;
>>> +		#size-cells = <2>;
>>> +		#address-cells = <3>;
>>> +		device_type = "pci";
>>> +		interrupts = <23 2 0 0>;
>>> +		interrupt-map-mask = <0xf800 0 0 7>;
>>> +		interrupt-map = <
>>> +			/* IDSEL 0x0 */
>>> +			0000 0 0 1 &mpic 43 1 0 0
>>> +			0000 0 0 2 &mpic 0 1 0 0
>>> +			0000 0 0 3 &mpic 4 1 0 0
>>> +			0000 0 0 4 &mpic 8 1 0 0
>>> +			>;
>>> +	};
>>> +};
>>> +
>>
>> The above 4 nodes have the consequence that it will then be mandatory that a
>> board support .dts file that would like to inlcude the SOC-NAMEsi-post.dtsi
>> defines the pci0, pci1, pci2, pci3 aliases.
>>
>> Now it is possible that a board does not implement pci1 for instance. So its
>> .dts file would ideally not define a node for it, and thus not define the
>> respective alias. However, this triggers this dtc compile error (which is correct):
>>
>>> [chlongv1@chber1-10533x linux-km]$ make kmp204x.dtb
>>>  DTC     arch/powerpc/boot/kmp204x.dtb
>>> Error: arch/powerpc/boot/dts/fsl/p2041si-post.dtsi:98.2-3 label or path, 'pci1', not found
>>> FATAL ERROR: Syntax error parsing input tree
>>> make[1]: *** [arch/powerpc/boot/kmp204x.dtb] Error 1
>>> make: *** [kmp204x.dtb] Error 2
>>
>> The solution I have found is to define a "dummy" disabled node so that I can
>> define the alias, but I am not really happy about this:
>>
>>> 	pci1: pcie@ffe201000 {
>>> 		status = "disabled";
>>> 	};
>>
>> I am here missing something obvious or shouldn't it be possible that such .dtsi
>> files allow not to define unused/unnecessary nodes ?
> 
> Isn't this correct, that you are disabling the PCIe1 interface on the SoC for your board?
> 

Yes it is correct. So this confirms that a board .dts file must disable all not
used interfaces/peripherals with such a "disabled" node.

I just wanted to be sure that this was the correct way of doing things and that
I had not missed a "better" way (where you don't need to define such "disabled"
nodes in .dts files).

Valentin