On 08/02/18 15:08, Rasmus Villemoes wrote:
> On 2018-02-05 07:07, Rob Herring wrote:
>>> +Example:
>>> +   scfg: scfg@1570000 {
>>> +           compatible = "fsl,ls1021a-scfg", "syscon";
>>> +           ...
>>> +           extirq: interrupt-controller {
>>> +                   compatible = "fsl,ls1021a-extirq";
>>> +                   #interrupt-cells = <3>;
>>> +                   interrupt-controller;
>>> +                   interrupt-parent = <&gic>;
>>> +                   offset = <0x1ac>;
>>
>> Use reg here instead (with a length).
> 
> Hm, ok, but what does the length buy us? Should the driver just ignore
> it, or should it check that it is 4 and bail out if not?
> 
>>> +                   interrupts = <163 164 165 167 168 169>;
>>
>> These don't look like GIC interrupt cells. Building this with current 
>> dtc will have errors.
> 
> Indeed, they are not. They simply record which interrupt lines on the
> GIC the external interrupt lines IRQ0...IRQ5 map to (the arm64 socs
> apparently have 12 such lines, but I don't know what they map to). I
> originally had that mapping in the driver, but I was asked to move it to
> DT. Is the problem the use of the name "interrupts" for this property?
> I'm happy to use something else (parent-interrupts, interrupt-mapping,
> ...) I find it very hard to figure out which property names have
> magic/reserved meanings.
> 
> I don't see any warnings/errors from dtc in the 4.14 tree I'm working
> on. Does it require an even newer dtc than that?

Most interrupt controllers use a private property, potentially with a
range (see the recent example of the Qualcomm PDC [1]).

Thanks,

        M.

[1] https://patchwork.kernel.org/patch/10208037/
-- 
Jazz is not dead. It just smells funny...

Reply via email to