On Monday 18 November 2013, Gerhard Sittig wrote: > [ trimmed Cc: list for DT ] > [ is the bindings/gpio/8xxx_gpio.txt document incomplete? ] > > On Mon, Nov 18, 2013 at 12:38 +0100, Arnd Bergmann wrote: > > > > See Documentation/devicetree/bindings/gpio/8xxx_gpio.txt for an > > example of a device whose interrupt line is connected to a gpio > > controller. The key here is to set the "interrupt-parent" to > > the gpio node and have the irq specifier define an interrupt > > local to that node. > > I was wondering whether the gpio-controller nodes lack both the > 'interrupt-controller' boolean flag as well as the > '#interrupt-cells = <2>' property. The former may be optional, > I'm not certain there. But the latter should be a hard > requirement without which the 'funkyfpga' references cannot get > parsed I'm afraid. You are absolutely right, I picked an example that was actually wrong. > Am I missing something? Did I get the names right? Shall I send > a patch? Is the code working without those specs since it > assumes knowledge that was not specified in the device tree? And > would this be against the idea that the binding should be > "complete" and independent of who is interpreting the data? Or > is the binding document just incomplete? I think the problem is that this particular controller is never used in "interrupts" properties, so nobody noticed the mistake. It probably still works for gpio references and drivers using gpio_to_irq on those numbers. It would be nice if you can send a patch to add those as optional properties to the binding and the example. We can't really make them mandatory properties now because that would make all existing dts files with this controller invalid. Arnd -- 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