Hi Linus,

Le 24/10/2016 16:39, Linus Walleij a écrit :

On Mon, Oct 24, 2016 at 6:51 AM, Andrey Smirnov
[off-list ref] wrote:

quoted

It seem strange to me that the driver uses "handle_edge_irq", given
how none of the individual interrupts seem to require any ACKing,
since it is all handled in sx150x_irq_thread_fn(), line 533. More so,
I had trouble finding who/where sets .irq_ack() callback, which AFAIU
is mandatory for handle_edge_irq().

Yes that looks strange.

Neil have you tested IRQs with this code?

To be frank, I took the IRQ code from the GPIO driver verbatim, and only tested
a simple use case on beagle bone black.
The interrupt code is very complex and sincerely I was not able to understand
clearly how it worked.

The IRQ_TYPE_EDGE_BOTH did work for me since the BBB irq controller dealt with it.

I will dig in the datasheet and run more uses cases next week, but since the
code hasn't changed since the GPIO driver, so this is a nice to have !

If there is trouble, please follow up with a fix for the edge handler.
Maybe it should just be handle_simple_irq().

Yours,
Linus Walleij

Andrey, can we talk over irc somehow ? I'm idling as narmstrong on freenode.

Neil