Dont hard-code GIC_IRQ_PHANDLE, read it from device tree instead

[axel-h]

Follow-up from #42 (comment)

Remove the hard coded GIC_IRQ_PHANDLE define and take the value from the device tree instead. Or add at least a sanity check that the hard-coded values machtes the device tree.

There have been cases, where the phande changes, and that causes strange errors then.
See also #27

[chrisguikema]

Can we define an INTERRUPT_CONTROLLER macro instead? That way you can read that node from the device tree, and get the phandle?

[kent-mcleod]

What about getting the phandle by looking up the interrupt parent of the primary boot core?

[chrisguikema]

Is that something that is consistent on a per-platform basis?

Here is the boot core node on a zynqmp device tree:

		cpu@0 {
			compatible = "arm,cortex-a53";
			device_type = "cpu";
			enable-method = "psci";
			operating-points-v2 = <0x01>;
			reg = <0x00>;
			cpu-idle-states = <0x02>;
			clocks = <0x03 0x0a>;
			phandle = <0x3d>;
		};

On a tx2 device tree, I see <interrupt-parent> right on top.

[kent-mcleod]

I see. I guess defining the path is good enough. It could also be used to setup the virtual gic without having to hardcode the addresses for each platform.

[chrisguikema]

Created a PR for that here: seL4/seL4_projects_libs#83

[axel-h]

Closing this, because #52 was merged and we use DTB paths now.