interrupts-extended only allowed with different parents? [Was: Re: [PATCH 3/4] ARM: dts: imx28-tx28: fix interrupt flags and use interrupts-extended property]

[Rob Herring]

On Mon, Oct 16, 2017 at 4:03 AM, Uwe Kleine-König
<u.kleine-koenig at pengutronix.de> wrote:
> On Mon, Oct 16, 2017 at 10:56:32AM +0200, Lothar Waßmann wrote:
>> Hi,
>>
>> On Mon, 16 Oct 2017 09:17:26 +0200 Uwe Kleine-König wrote:
>> > Hello,
>> >
>> > On Wed, Oct 11, 2017 at 01:05:38PM +0200, Lothar Waßmann wrote:
>> > > diff --git a/arch/arm/boot/dts/imx28-tx28.dts b/arch/arm/boot/dts/imx28-tx28.dts
>> > > index 211e67d..3c852f7 100644
>> > > --- a/arch/arm/boot/dts/imx28-tx28.dts
>> > > +++ b/arch/arm/boot/dts/imx28-tx28.dts
>> > > @@ -328,8 +328,7 @@
>> > >           reg = <0x20>;
>> > >           pinctrl-names = "default";
>> > >           pinctrl-0 = <&tx28_pca9554_pins>;
>> > > -         interrupt-parent = <&gpio3>;
>> > > -         interrupts = <28 0>;
>> > > +         interrupts-extended = <&gpio3 28 IRQ_TYPE_NONE>;
>> > >           gpio-controller;
>> > >           #gpio-cells = <2>;
>> > >           interrupt-controller;
>> >
>> > While interrupts-extended looks nice,
>> > Documentation/devicetree/bindings/interrupt-controller/interrupts.txt
>> > has:
>> >
>> >     "interrupts-extended" should only be used when a device has
>> >     multiple interrupt parents.
>> >
>> > If this is still true, this patch is wrong.
>> >
>> Thanks for the hint. It really helps to read the documentation
>> sometimes, rahter than relying on existing code only...
>>
>> A quick check shows, that more than 100 of the 130 uses of
>> interrupts-extended are wrong. :(
>
> That's why I honestly consider that these documentation bits are stale.
> I adapted the Subject to maybe catch the attention of the devicetree
> guys.

The documentation is correct as that is recommended practice IMO.

I wouldn't go fixing the 100 cases found either.

> (BTW: The current wording is likely imprecise. I'd expect that it really
> should mean "Use interrupt-parent + interrupts if possible", but the
> following still fulfills the documented condition:

"should" is pretty standard to mean recommended vs. "shall" or "must"
meaning required.

Rob