QUERY · ISSUE

Cannot init USB devices in boot.py

openby mischifopened 2024-06-13updated 2025-12-13

bug

I tested this on a pico running MP 1.23 and the latest release of usb-device-cdc.

boot.py:

import usb.device
from usb.device.cdc import CDCInterface

IFACE_TWO = CDCInterface()
IFACE_TWO.init(timeout=0)
#usb.device.get().init(IFACE_TWO, builtin_driver=True)

main.py:

from time import sleep_ms

def main():
	while True:
		print(IFACE_TWO)
		sleep_ms(1000)

if __name__ == "__main__": main()

Running this as-is, IFACE_TWO does show up in main.py, but uncommenting the init function causes the process to error out. This discord post suggests you can call usb.device.get.init() in boot.py, is this no longer true?

8 comments

projectgus · 2024-06-19

causes the process to error out

Please give some more details about this means: what exactly do you do, and what error output do you see?

A soft reset (i.e. Ctrl-D in the REPL) will still disconnect the USB port each time, because the runtime USB interface objects have to all be freed and recreated.

A hard reset (i.e. plugging the rp2 in, or calling machine.reset()) should come up immediately with the two CDC ports.

The only difference between boot.py and main.py is that the first time the device boots from a hard reset, it runs boot.py before it initialises USB. If you put the code in main.py instead of boot.py then the rp2 would first try create a device using the default USB interface, then immediately switch to the custom USB interface. This can show up on the host as connect/disconnect/connect cycle, or as an error depending on the timing.

mischif · 2024-06-19

With the init line commented out in boot.py:

plug in pico
mpremote fs cp boot.py :
unplug/replug
mpremote run main.py

I see the print statements I expect:

9600/8N1 rts=False dtr=False
9600/8N1 rts=False dtr=False
[...]

With the init line uncommented in boot.py:

plug in pico
mpremote fs cp boot.py :
unplug/replug
mpremote run main.py

I do not see the print statements I expect:

Traceback (most recent call last):
  [...]
  File "/home/mischif/.local/lib/python3.10/site-packages/serial/serialposix.py", line 549, in in_waiting
    s = fcntl.ioctl(self.fd, TIOCINQ, TIOCM_zero_str)
OSError: [Errno 5] Input/output error

I get dropped back to the terminal, and connecting with mpremote a0 yields no output. Attempting to replace boot.py even after unplugging/replugging the pico also yields the above OSError; I have to drop to a shell and use os.unlink().

projectgus · 2024-06-19

Ah snap, mpremote run does a soft reset before it runs the provided file. Because boot.py has loaded the custom USB interface, it disconnects the USB port every time (then boot.py runs and adds it back, so you end up where you started.)

If you copy both boot.py and main.py to the filesystem then main.py works and you can use mpremote a0 to view the output, but you still can't use mpremote run any more.

I think the fix for this will be the planned work to have mpremote detect when the USB port is in a disconnect/reconnect cycle, and recover from it instead of erroring out.

mischif · 2024-07-10

Is there a timeline for when the planned work might be finished?

projectgus · 2024-07-11

Not at the moment, I'm afraid.

ziesemer · 2024-11-17

I assume this is impacting mpremote connect as well (also including a soft reset)? Running similar as the OP above, I'm finding that I can connect directly to REPL by connecting to its COM port - but not using mpremote connect.

ziesemer · 2025-12-13

Not sure if anything has changed since the last update here - though revisiting this a year later, now on MicroPython v1.27 - everything seems to be working, and I'm not sure I can find a way to make it not?

I'm seeing success with both (separately) MicroPico 4.3.4 in VS Code, and mpremote 1.27.0.

Curious if anyone else here still watching can help test / validate? Thanks!

ziesemer · 2025-12-13

Though for reasons I don't yet fully understand, having usb.device.get().init(...) included as part of boot.py interferes with running "MicroPico: Run current file on Pico" (micropico.run) from MicroPico in VS Code.

Disabling "Micropico: No Soft Reset On Run" (micropico.noSoftResetOnRun) seems to then resolve this.

Though then, and I suspect unrelated, I get into trouble if I try using Ctrl+C to quit a loop - same as reported in https://github.com/paulober/MicroPico/issues/294 - though pretty sure this is an issue with MicroPico and not micropython.

CANDIDATE · ISSUE

Writing to a second serial connection in a thread has errors

closedby mischifopened 2024-06-11updated 2024-09-23

Running on a Pi Pico with MicroPython 1.23.0 and whatever the latest version of usb-device-cdc is in mip; code to reproduce:

from utime import sleep_ms, time
from _thread import start_new_thread

import usb.device
from usb.device.cdc import CDCInterface

SERIAL_TWO = None


def update():
	while True:
		print("Update: {}".format(time()), file=SERIAL_TWO, end="\r\n")
		sleep_ms(1000)

def main():
	global SERIAL_TWO
	SERIAL_TWO = CDCInterface()
	SERIAL_TWO.init(timeout=0)
	usb.device.get().init(SERIAL_TWO, builtin_driver=True)

	while not SERIAL_TWO.is_open():
		sleep_ms(100)

	start_new_thread(update, ())

	while True:
		sleep_ms(1000)


if __name__ == "__main__": main()

If you connect to both ports with mpremote a0 and mpremote a1, you'll see this error sporadically:

Traceback (most recent call last):
  File "usb/device/core.py", line 316, in _xfer_cb
  File "usb/device/cdc.py", line 327, in _wr_cb
  File "usb/device/core.py", line 833, in finish_read
AssertionError:

But I did see this error once:

Unhandled exception in thread started by <function update at 0x2000aef0>
Traceback (most recent call last):
  File "<stdin>", line 12, in update
  File "usb/device/cdc.py", line 356, in write
  File "usb/device/cdc.py", line 322, in _wr_xfer
  File "usb/device/core.py", line 570, in submit_xfer
  File "usb/device/core.py", line 302, in _submit_xfer
RuntimeError: xfer_pending

4 comments

projectgus · 2024-06-11

Hi @mischif,

Thanks for the report, I can reproduce this here. The CDC class currently uses machine.disable_irq() and machine.enable_irq() to control concurrent access to the serial buffer, but this doesn't prevent a thread on CPU1 from accessing it at the same time a USB callback is running. EDIT: This is wrong, machine.disable_irq() locks a mutex.

Will look into an alternative locking mechanism for this.

projectgus · 2024-06-11

Root cause is not quite that actually, it's that the current USB transfer callback code marks the endpoint as free (no pending transfer) before it calls the transfer callback, so the callback code can call submit_xfer() if needed to submit a new transfer immediately.

However when the thread on the other CPU is submitting transfers, there's a race where it may see the endpoint has no pending transfer and submit a new transfer before the callback is done executing. Need to redesign the callbacks to handle this sequence.

mischif · 2024-07-10

Is there an update for this bug? I see one PR related to this was merged, but it's apparently only a partial fix.

projectgus · 2024-07-11

There is, I was going to return to these fixes this week but I got sick instead. I should have something soon.

Cannot init USB devices in boot.py

Writing to a second serial connection in a thread has errors

Keyboard