USB support

[japaric]

this PR builds on top of @jonas-schievink PR (#133)

I have managed to get a CDC ACM (using the usbd-serial crate) device enumerated and got some output out from it using the following program (snippet):

    loop {
        if !usb_dev.poll(&mut [&mut serial]) {
            continue;
        }

        if usb_dev.state() == UsbDeviceState::Configured {
            if once {
                serial.write(b"Hello, world!\n").unwrap();
                serial.flush().unwrap();

                once = false;
            }
        }
    }

$ cat /dev/ttyACM1
Hello, world!

(don't forget to send a newline or cat won't print anything)

However, I'm not getting reliable enumeration (*). I still need to dig deeper into that issue but would appreciate other people testing out this branch.

(*) though this may be mainly caused by how I'm terminating my embedded application: I'm halting the device without first disconnecting it from the USB bus; this seems to make the host USB controller very sad (it power cycles the USB bus a few times and never gets an answer from the device (because it's halted))

P.S. this branch depends on this Jonas' fork of usbd-device (which handles the nrf52 quirk of handling SET_ADDRESS requests entirely in hardware). Use the following patch before you try out this branch:

[patch.crates-io.usb-device]
git = "https://github.com/jonas-schievink/usb-device.git"
branch = "inhibit-setaddr-resp"

[Disasm]

We have a test repo with a few tests that you can port and run on your hardware: https://github.com/Disasm/stm32-usbd-tests

[Disasm]

For some reason, this driver doesn't work with the default EP0 size (8 bytes). Also a serial loopback example (based on https://github.com/Disasm/usb-otg-workspace/blob/master/example-f429zi-board/examples/serial.rs#L63-L85) fails to pass stress-testing.

[valpackett]

Rebasing the usb-device quirk onto the current (0.2.7) version, looks like rust-embedded-community/usb-device@7509c77 breaks device enumeration, I've had to revert that commit to make it work.

[valpackett]

Oh.

                    regs.tasks_ep0stall.write(|w| if stalled { w.tasks_ep0stall().set_bit() } else { w.tasks_ep0stall().clear_bit() });

[wez]

I went with:

  regs.tasks_ep0stall.write(|w| w.tasks_ep0stall().bit(stalled));

[wez]

@nrf-rs/nrf52 I've rebased this and plan to push this PR + my local fixup diff to close out this PR.

I have a few "Particle Xenon" boards with 52840's on them, as well as that weird MDK thing that I previously added a bsp for in this org, so I'm able to test this and confirmed that the device enumerates and responds:

[72570.386287] usb 3-3.4.4.2: New USB device strings: Mfr=0, Product=2, SerialNumber=0
[72570.386289] usb 3-3.4.4.2: Product: nRF52840 Serial Port Demo
[72570.472123] cdc_acm 3-3.4.4.2:1.0: ttyACM0: USB ACM device
[72570.472567] usbcore: registered new interface driver cdc_acm
[72570.472569] cdc_acm: USB Abstract Control Model driver for USB modems and ISDN adapters

$ sudo cat /dev/ttyACM0
Hello, world!
^C

[jonas-schievink]

Awesome! Happy to take a look.

[wez]

My fixup diff on top of this is minor; I'll just push it and if you spot something I'm around to act on it!

[jonas-schievink]

Please wait for CI to pass before pushing to master

[wez]

Too late... but I'm running those tests locally and fixing up now :)

[Disasm]

I'm going to run test_class as well as the serial benchmark soon

[Disasm]

Tested, the behavior is the same as mentioned in #144 (comment)