On Thu, Jun 04, 2020 at 01:21:12PM +0200, Lukas Wunner wrote:
> On Thu, Jun 04, 2020 at 12:13:25PM +0100, Mark Brown wrote:

> > Regardless of what's going on with the interrupts the compatible string
> > should reflect the IP version so unless for some reason someone taped
> > out two different versions of the IP it seems odd that the compatible
> > strings would vary within a given SoC.

> Hm.  I guess it may be possible to search the DT for other devices
> sharing the same interrupt line and thereby determine whether
> IRQF_SHARED is necessary.  The helper to perform this search could
> live in drivers/of/irq.c as I imagine it might be useful in general.

That's another option, yeah - it'd be DT specific but it seems neater
than a property and much more tractable than trying to dance around
doing this in genirq (where we'd end up with callbacks when the second
device registers or something else horrible).