Hi, Arnd

On Wed, 2018-02-07 at 09:31 +1100, Benjamin Herrenschmidt wrote:

On Tue, 2018-02-06 at 13:42 +0800, Ryder Lee wrote:

quoted

Thanks for explanation.

So I guess the better way to achieve my aim - one IRQ per slot that is
connected to all INTx and get propagated through the bridges (and for
those root ports own interrupts (PME ..)} is to add interrupt-map
properties in both parent and root port nodes.

Something like this: https://patchwork.kernel.org/patch/9970923// ,right?

Yup.

Cheers,
Ben.

Do you have any thoughts on the original approach? If you are okay with
it, I will resend the DT patch.

Thanks.