On Wed, May 18, 2016 at 8:50 AM, Joel Stanley <joel@xxxxxxxxx> wrote:
> On Fri, May 13, 2016 at 2:50 AM, Rob Herring <robh@xxxxxxxxxx> wrote:
>> On Wed, May 11, 2016 at 8:01 PM, Benjamin Herrenschmidt
>> <benh@xxxxxxxxxxxxxxxxxxx> wrote:
>>> On Wed, 2016-05-11 at 09:33 -0500, Rob Herring wrote:
>>>> > +- interrupt-controller : Identifies the node as an interrupt controller
>>>> > +- #interrupt-cells : Specifies the number of cells needed to encode an
>>>> > + interrupt source. The value shall be 1.
>>>>
>>>> No need for level vs. edge flags?
>>>
>>> That's an open question. Most interrupts are fixed. A handful of GPIOs
>>> can be configured either way. For now I am relying on uboot setting up
>>> the right config for them and I read it back at boot time, but we could
>>> make it part of the binding I suppose.
>>
>> It is almost standard to just use 2 cells here even if reserved for
>> future use. Especially since the IP block seems to have registers to
>> control that.
>
> Yep, makes sense. I've added this to the bindings document.
>
> I was trying to use the second cell in our driver:
>
> static struct irq_domain_ops avic_dom_ops = {
> .map = avic_map,
> .xlate = irq_domain_xlate_twocell,
> };
>
> So we have irq_domain_xlate_twocell to parse the device tree and
> grabs the type property from the second cell.
>
> In avic_map we set the irq handler:
>
> if (vic->edge_sources[sidx] & sbit) {
> /* TODO: Check aginst type from dt and warn if not edge */
> irq_set_chip_and_handler(irq, &avic_chip, handle_edge_irq);
> } else {
> /* TODO: Check aginst type from dt and warn if not level */
> irq_set_chip_and_handler(irq, &avic_chip, handle_level_irq);
> }
>
> However, we don't have the type here, so we can't use it to check that
> we're setting the correct edge/level callback (or use it to set the
> register in the future if we want to use the device tree to override
> the bootloader settings).
>
> I had a look at some other drivers and they don't seem to use the dt
> property when mapping the irq. What am I missing here?
The type will be set by the irqdomain core when the mapping is created
and this should result in irq_set_type() being called on your irqchip.
The map function doesn't need to deal with type.
Rob
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
