Hello everyone,

In order to solve this issue, I have opened a PR here:
https://github.com/apache/nuttx/pull/16281

This allows multiple GPIO interrupt types to be selected once, since the
RP2040 supports that. Now one GPIO pin can trigger up to 4 interrupt events.

I would appreciate your review!
Matteo

On Tue, Mar 18, 2025 at 4:36 PM Kevin Witteveen <kevinwit1...@gmail.com>
wrote:

> Hey Matteo,
>
> It appears the RP2040 can support falling and rising interrupts at once
> according to the RP2040 datasheet section 2.19.5.2 "Enable a GPIO
> interrupt", which provides an example of how to do this.
> Ofcourse this example is not NuttX, but can help for implementation.
> https://datasheets.raspberrypi.com/rp2040/rp2040-datasheet.pdf
>
> Maybe this can help you find a solution.
>
> Best regards,
>
> Kevin
>
>
> Op di 18 mrt 2025 om 21:25 schreef Alan C. Assis <acas...@gmail.com>:
>
> > Hi Matteo,
> >
> > I think RP2040 and new RP2350 are good MCUs, but Raspberry Pi Foundation
> > did some terribles mistakes, like using a serial port control that
> doesn't
> > have indication of transmission done (useful for RS485 support) and GPIO
> > INT with both edges support.
> >
> > When I created the ultrasound sensor driver (drivers/sensors/hc_sr04.c) I
> > already considered that some "cheap" MCU will not have support for both
> > edge detection, so initially I detect the raising signal and when that
> > interrupt happens I change the GPIO to detect falling edge. Look at
> > hcsr04_start_measuring() and hcsr04_int_handler().
> >
> > BR,
> >
> > Alan
> >
> > On Tue, Mar 18, 2025 at 5:07 PM Matteo Golin <matteo.go...@gmail.com>
> > wrote:
> >
> > > Thank you everyone for the suggestions.
> > >
> > > I don't have enough GPIO pins to dedicate two per switch unfortunately,
> > nor
> > > do I have a suitable hardware method for handling debouncing. I suppose
> > > this is taken care of in the button driver framework via software, but
> > not
> > > for the bare GPIO framework. I don't believe the button framework will
> > work
> > > for this application because the switches are toggle switches, not
> > buttons.
> > >
> > > I like Marco's solution of switching the interrupt type on the toggle,
> > but
> > > I am fearful of the race conditions you mentioned, Nathan. I suppose I
> > > could take the interrupt type toggling approach but read the
> interrupted
> > > pin in software with a delay to ensure the correct value, which would
> > > remedy the debouncing issue? I believe this can be done from
> application
> > > code at least. That still leaves the issue of the interrupt type not
> > being
> > > toggled in time for the next edge, which could miss an input. I was
> > hoping
> > > there would be a method to attach both a rising and falling edge
> > interrupt
> > > handler to the GPIO pin simultaneously but it appears the current
> RP2040
> > > support doesn't allow that, although the chip itself does. Maybe that
> > would
> > > be a feature worth adding?
> > >
> > > Thanks again for the suggestions, I'll have to do some thinking but
> maybe
> > > the polling approach is the safest bet.
> > >
> > > Matteo
> > >
> > > On Tue, Mar 18, 2025 at 8:23 AM Nathan Hartman <
> hartman.nat...@gmail.com
> > >
> > > wrote:
> > >
> > > > On Mon, Mar 17, 2025 at 5:18 PM Matteo Golin <matteo.go...@gmail.com
> >
> > > > wrote:
> > > >
> > > > > Hello everyone,
> > > > >
> > > > > I have an application wherein I am using a W5500-EVB Pico as the
> MCU
> > > for
> > > > a
> > > > > network controlled system. I need to connect
> > > > > several switch inputs into this MCU. The switches are normally held
> > > high
> > > > > via the RP2040's internal pullups, and pulled
> > > > > low when flipped.
> > > >
> > > >
> > > >
> > > > By switch inputs, do you mean physical switches, like toggle
> switches?
> > If
> > > > so, then I hope you're handling switch debouncing somehow, either by
> > > > software (multiple consecutive identical readings, spaced apart by
> some
> > > > constant time interval, must occur before reacting to the change in
> > > switch
> > > > state) or by hardware (at the very least a RC network with a schmitt
> > > > trigger). Otherwise, if you interrupt on rising and falling edges,
> > you'll
> > > > likely see spurious interrupts and unwanted state changes.
> > > >
> > > >
> > > > The problem I'm encountering is that it appears that the current
> RP2040
> > > > > support only allows for one type of interrupt
> > > > > per pin: either rising edge or falling edge. I need to trigger on
> > both
> > > > > rising and falling edge to accurately track the
> > > > > state of the switch. I'm wondering if anyone else has had a similar
> > > issue
> > > > > or has suggestions about how I can resolve
> > > > > this.
> > > >
> > > >
> > > >
> > > > Assuming that switch bounce is either handled or is not an issue (see
> > > > above), you could either use 2 IO pins as suggested by William or
> > > > reconfigure the interrupt edge as suggested by Marco.
> > > >
> > > > If you use Marco's idea, you must handle two issues that will cause
> > > > problems if left unhandled:
> > > >
> > > > 1) You must determine the correct initial edge to configure, and
> > > >
> > > > 2) You must handle race conditions between changes in the hardware IO
> > > state
> > > > and reconfiguration of the interrupt edge, i.e., a scenario like:
> > > >
> > > > a) rising edge interrupt occurs
> > > > b) some small amount of time passes
> > > > c) interrupt handler runs
> > > > d) in interrupt handler, you reconfigure to interrupt on falling edge
> > > >
> > > > but somewhere between a and d, falling edge already occurred, so
> > falling
> > > > edge interrupt will be missed.
> > > >
> > > > That's not the only scenario that could go wrong.,You have to ensure
> > that
> > > > all possible race conditions are handled in a reliable way or this
> will
> > > be
> > > > a repeated source of headaches.
> > > >
> > > > For reasons like this, I prefer a timer-based polling and software
> > > > debouncing for switches, but YMMV.
> > > >
> > > > Hope this helps,
> > > > Nathan
> > > >
> > >
> >
>

Reply via email to