On Mon, Jun 15, 2020 at 12:42:50PM -0700, Florian Fainelli wrote:

> Or how about this: we slightly re-structure the interrupt handler such
> that all possible interrupt conditions are explicitly handled and
> terminate with a return IRQ_HANDLED, and those which are not, including
> in the case of a "spurious" (because the interrupt was triggered for
> another SPI controller instance), then we finish the function with
> IRQ_NONE. This would not impact the performance for the BCM2835/36/37
> which would still have a single controller/single interrupt line to handle.

That seems sensible - it's generally a good way to structure interrupt
handlers.  It's almost there already.