On Mon, Aug 17, 2020 at 08:21:58PM +0200, Bartosz Golaszewski wrote:

On Fri, Aug 14, 2020 at 5:05 AM Kent Gibson [off-list ref] wrote:

quoted

Add support for setting debounce on a line via the GPIO uAPI.
Where debounce is not supported by hardware, a software debounce is
provided.

Signed-off-by: Kent Gibson <warthog618@gmail.com>
---

[snip]

quoted

+       debounce_period = READ_ONCE(desc->debounce_period);
+       if (debounce_period) {
+               info->attrs[num_attrs].id = GPIO_V2_LINE_ATTR_ID_DEBOUNCE;
+               info->attrs[num_attrs].debounce_period = debounce_period;
+               num_attrs++;
+       }
+       info->num_attrs = num_attrs;

AFAICT this (reading it in gpio_desc_to_lineinfo) is the only reason
to store the debounce period in struct gpio_desc. I'm wondering if we
can avoid extending this struct only for such uncommon case and store
it elsewhere. In all other cases where you read or write to it - you
have access to the underlying edge detector. Would the single-line
struct line I suggested elsewhere be a good place? On the other hand
I'm not sure how to get it having only the desc. I need to think about
it more.

Yeah, it is stored there so it can be returned by lineinfo_get() for the
GPIO_V2_GET_LINEINFO_IOCTL and GPIO_V2_GET_LINEINFO_WATCH_IOCTL.
And the same applies to any future config fields.
I would also like to not pollute the desc, or anything else in gpiolib,
but wasn't sure where else to put it.

I'm open to suggestions.

Cheers,
Kent.