On Wed, Sep 20, 2017 at 3:04 PM, Timur Tabi [off-list ref] wrote:

On 9/20/17 6:43 AM, Linus Walleij wrote:

quoted

Doesn't that mean we need something like irq_valid_mask but rather
gpio_valid_mask that just block all usage of certain GPIOs?

That raises a lot of questions.  In the meantime, my current patches for
4.14 work fine.

Do we replace irq_valid_mask with gpio_valid_mask?  That would break drivers
where the GPIO is valid but the interrupt is not.  If we keep both, what
happens if gpio_valid_mask is false but irq_valid_mask is true?  And then we
would need to audit all gpio drivers to see which ones should be updated for
the new infrastructure.

I guess gpio_valid_mask would take precedence over irq_valid_mask.
I.e if the GPIO is not valid then the IRQ is per definition not valid either.

Since it is a new thing, we can simply define a semantic like that
and document it.

Yours,
Linus Walleij