Using rppal with embedded_hal_async (async support)

[IhsenBouallegue]

I am trying to get lora_phy working with rppal. But I am getting the following errors

It seems like lora_phy is using embedded_hal_async next to embedded_hal. As far as I understood, only embedded_hal is implemented for rppal? is there a relatively simple way to add embedded_hal_async support, or is it pretty much a rewrite?

Here is also a reproduction repo. If you want to check my setup
https://github.com/IhsenBouallegue/lora-phy-embedded-linux

[golemparts]

Hi @IhsenBouallegue. As you noted, embedded_hal_async isn't currently supported. From a cursory glance, supporting the async version should be relatively straightforward. However, since rppal doesn't support async natively, someone would have to write a wrapper to provide the needed functionality using the existing sync functions.

The best short-term solution would be to use a crate that supports embedded_hal_async on linux as I'm not sure when/if rppal will get support for embedded_hal_async. However, if anyone is up for adding support, I'd be happy to accept a PR.

[IhsenBouallegue]

Hey @golemparts thanks for the reply! I would also be happy to open a PR, but I am still lost on how to do this (and learning rust along the way).
Would perhaps using something like this be possible? https://github.com/embassy-rs/embassy/blob/main/embassy-embedded-hal/src/adapter/blocking_async.rs
I couldn't get it working yet but wanted to get more opinions and tips first.

[golemparts]

Thanks for the link. While that would technically work, it's not the approach I would go for, since it just turns the async implementation into a potentially blocking one, which could cause issues. rppal's native functionality can already be used in a (mostly) non-blocking way, although I haven't checked if that's possible for all peripherals and operations supported by embedded-hal-async. Some additional internal support might be needed.

Once embedded-hal 1.0.0 releases it'll be time for some clean-up anyway. If noone has had a chance to add support for async before then, I'll add it myself, since it does seem useful to have available for drivers that require it.

[IhsenBouallegue]

Awesome I will be looking forward to that and hopefully I can help out. I am writing my bachelor thesis in this more or less and I would to get to use the async functions!

[IhsenBouallegue]

@golemparts with the release of embedded-hal 1.0.0 is there a possibility to add proper async with embedded-hal-async now?
I gave it a try myself. While I managed to make it "work" I am pretty sure this is an ugly workaround

impl embedded_hal_async::digital::Wait for InputPin {
    async fn wait_for_low(&mut self) -> core::result::Result<(), Self::Error> {
        let (sender, receiver) = oneshot::channel();
        let sender_mutex = Arc::new(std::sync::Mutex::new(Some(sender)));
        let _ = self.set_async_interrupt(super::Trigger::FallingEdge, move |_| {
            if let Some(sender) = sender_mutex.lock().unwrap().take() {
                sender.send(()).unwrap();
            }
        });
        Ok(receiver.await.unwrap())
    }
    ...

for spi I made a wrapper that fakes async but a proper implementation would benefit the library a lot. Sadly I couldn't manage that 😢

impl<T, E> embedded_hal_async::spi::SpiBus<u8> for BlockingAsync<T>
where
    E: embedded_hal::spi::Error + 'static,
    T: blocking::spi::Transfer<u8, Error = E> + blocking::spi::Write<u8, Error = E>,
{
    async fn flush(&mut self) -> Result<(), Self::Error> {
        Ok(())
    }

    async fn write(&mut self, data: &[u8]) -> Result<(), Self::Error> {
        self.wrapped.write(data)?;
        Ok(())
    }

    async fn read(&mut self, data: &mut [u8]) -> Result<(), Self::Error> {
        self.wrapped.transfer(data)?;
        Ok(())
    }

    async fn transfer(&mut self, read: &mut [u8], write: &[u8]) -> Result<(), Self::Error> {
        for i in 0..core::cmp::min(read.len(), write.len()) {
            read[i] = write[i].clone();
        }
        self.wrapped.transfer(read)?;
        Ok(())
    }

    async fn transfer_in_place(&mut self, data: &mut [u8]) -> Result<(), Self::Error> {
        self.wrapped.transfer(data)?;
        Ok(())
    }
}

[golemparts]

with the release of embedded-hal 1.0.0 is there a possibility to add proper async with embedded-hal-async now?

That's the plan, yes 😁 . I'm glad to see you figured out a workaround for yourself for now, assuming you're using that in your own projects. I haven't had a chance to dive into the async traits yet myself, other than a cursory glance when you first opened this issue. I'll be publishing the latest rppal with embedded-hal 1.0.0 support this week. I can't give you an exact schedule for when to expect async support, but it's on the list!

[IhsenBouallegue]

@golemparts Glad to see embedded-hal 1.0.0 coming to rppal! Awesome work 🥳
Sadly my workaround is just a workaround and things are starting to act weird now, and I suspect it's because of my implementation :'( I am using this as part of my thesis, so I am a bit constrained on time. I am up for implementing this in the library if I get some guidance.
Let me know if I can be of any help! 😄

[golemparts]

@IhsenBouallegue unfortunately I'm pretty swamped with another project at the moment, so I wouldn't be able to offer much guidance. As you are still new to Rust and are constrained on time, I would recommend checking if there are any other embedded-hal-async implementations available. A generic linux implementation should generally work for the Pi. It just won't be as performant as rppal's Pi-specific implementation. Alternatively, perhaps another LoRa PHY crate offers similar functionality without the need for embedded-hal-async. Another option could be to implement the driver yourself using rppal or embedded-hal.

[coolreader18]

Would it be reasonable to use something like tokio or async-io for this? It feels like a lot of unnecessary work to reimplement an event loop when async_io::Async::new(Interrupt::new()).poll_readable() does the same thing, ready-made.