Hi Andrew,

On Wed, 15 Sept 2021 at 00:59, Andrew Lunn [off-list ref] wrote:

How are the GPIOs mapped to the interrupts? Is it a simple 1:1?

Unfortunately, no.
I wanted to add the GPIO controller part of this to this same series
but there are some patches in flight for that so it would have been
messy.
You can see that here though:
https://github.com/linux-chenxing/linux/commit/88345dc470bf07d36aa1ddab09551ed33a1cfb22

They've really made a mess of this. Their whole GPIO thing is a mess
with no clear logic between the pin names and the register locations
etc.
This IRQ part is no exception. IRQ 0 from this thing isn't for the pin
called GPIO0 or anything sane like that.

The GPIO core has some support for the GPIO drivers to be also
interrupt controllers. So if this interrupt control is dedicated to
GPIO, you would be better to make it part of the GPIO driver.

I don't think so. One reason is the non-linear mapping stuff. A second
reason is this GPIO interrupt controller might handle GPIO interrupts
for multiple GPIO controller blocks.
Finally, in newer chips they've replaced one of the GPIO blocks with a
new IP which will need it's own driver. That GPIO controller still
seems to use this same IRQ block to handle it's interrupts.
So if this code is wrapped into the GPIO driver itself it would end up
duplicated in two GPIO drivers.

Cheers,

Daniel