[EV3] Implement Pybricks USB protocol #361

ArcaneNibble · 2025-07-30T22:11:21Z

This fully implements the rest of the Pybricks USB protocol, allowing pybricksdev to run.

coveralls · 2025-07-30T22:21:40Z

coverage: 56.982% (+0.02%) from 56.967%
when pulling 0198f7b on ArcaneNibble:usb
into ccafa60 on pybricks:master.

github-actions · 2025-07-30T22:24:39Z

Download the artifacts for this pull request:

ArcaneNibble · 2025-07-30T23:16:53Z

With this in place (and possibly some fixes in pybricksdev? I was noticing some strange issues) we might be able to call this an early alpha release?

dlech · 2025-07-31T03:25:14Z

and possibly some fixes in pybricksdev?

Can you elaborate? (Or send a PR)

dlech

Not the most thorough review and I won't be able to test until sometime next week, but looks pretty good to me.

lib/pbio/drv/usb/usb_ev3.c

lib/pbio/include/pbio/protocol.h

lib/pbio/drv/usb/usb_ev3.c

dlech · 2025-07-31T03:39:39Z

lib/pbio/drv/usb/usb_ev3.c

+    static uint32_t prev_status_flags = ~0;
+    static uint32_t new_status_flags = 0;
+
+    (void)ep1_tx_stdout_buf;


What is this for?

This silences a build failure due to warnings-treated-as-errors in the intermediate commit. This is because the variable is never used until stdout is implemented.

dlech · 2025-07-31T03:40:48Z

lib/pbio/drv/usb/usb_ev3.c

+
+        // Handle timeouts
+        is_transmitting = usb_tx_response_is_not_ready || usb_tx_status_is_not_ready || usb_tx_stdout_is_not_ready;
+        if (was_transmitting && is_transmitting) {


Perhaps can use pbio_oneshot()?

The reason I am not is because the code needs to handle both the rising edge case and the "continuously active" case.

dlech · 2025-07-31T03:43:45Z

lib/pbio/drv/usb/usb_ev3.c

+        // Handle timeouts
+        is_transmitting = usb_tx_response_is_not_ready || usb_tx_status_is_not_ready || usb_tx_stdout_is_not_ready;
+        if (was_transmitting && is_transmitting) {
+            if (pbio_os_timer_is_expired(&tx_timeout_timer)) {


Are we checking for timeout before setting the timer?

No, the timer gets set in the else if later, which this code has to pass through before both flags can become true.

lib/pbio/drv/usb/usb_ev3.c

laurensvalk · 2025-07-31T07:22:45Z

Amazing!

and possibly some fixes in pybricksdev? I was noticing some strange issues

At a glance, Pybricksdev seems to still always start slot 0 rather than start the selected slot. So code loaded to other slots does not appear to run. It does run if you start them with the button.

With this in place (...) we might be able to call this an early alpha release?

Let's wait a while. Anyone who is able to run it in the current state is probably also able to grab the nightly release.

I'd like to turn off some of the developer hacks (the hard reset; sensor 1 disabled, repl on 0) and get the official devices enabled (except I2C).

With the rapid progress here, I think I'll put aside some of the internal BLE refactoring and return focus to EV3 sensors and UX.

I was also occasionally getting this:

  File "/home/laurens/git-pybricks/pybricksdev/pybricksdev/connections/pybricks.py", line 612, in run
    await self.start_user_program()
  File "/home/laurens/git-pybricks/pybricksdev/pybricksdev/connections/pybricks.py", line 503, in start_user_program
    await self.write_gatt_char(
  File "/home/laurens/git-pybricks/pybricksdev/pybricksdev/connections/pybricks.py", line 947, in write_gatt_char
    await self._send_message(bytes([PybricksUsbOutEpMessageType.COMMAND]) + data)
  File "/home/laurens/git-pybricks/pybricksdev/pybricksdev/connections/pybricks.py", line 933, in _send_message
    raise RuntimeError(f"Write failed: {reply[0]}")

ArcaneNibble · 2025-08-01T23:14:01Z

Forgot to reply to this, but

pybricksdev seems to still always start slot 0

seems to describe the bulk of the strangeness I was experiencing. The other issues included:

download doesn't seem to work (OSError: Could not find hub.)
run doesn't seem to work if the hub is in REPL mode (RuntimeError: Write failed: 129)

I have cleaned up most of the review comments though.

laurensvalk · 2025-08-04T09:43:04Z

Forgot to reply to this, but

pybricksdev seems to still always start slot 0

seems to describe the bulk of the strangeness I was experiencing.

This is happening because pybricksdev is not getting the status update until after starting the program. See pybricks/support#2306 So we could fix this in Pybricksdev, but it also sounds like the EV3 firmware is doing something slightly different than SPIKE?

laurensvalk · 2025-08-05T12:49:49Z

Would you say we should merge it and further improve later? This has worked well for the SPI flash/adc code, so I'm happy to go this way here too.

It would be pretty useful to have #361, #363, #364 merged to facilitate other developments.

ArcaneNibble · 2025-08-05T13:30:04Z

Yes, I was hoping to merge these first and then make fixes as needed. I was waiting for @dlech to get back, unless you want to go ahead and click the button.

These transfers allow reading "characteristics" which would otherwise be exchanged over BLE GATT. They contain information about the firmware version and hub capabilities. Also move the bRequest values into pbio/protocol.h so they can be shared

In order to be able to detect disconnection, we cannot force device mode in the PHY. Forcing device mode overrides the analog comparators used to detect VBUS. Not forcing device mode should be fine because device mode is the default (enabling host mode requires setting DEVCTL.SESSION explicitly, and the ID pin is explicitly disconnected on the EV3 PCB.)

This handles Pybricks commands, status flag reporting, and timeouts. stdout is not yet handled.

This replaces the old stopgap stdout logic with stdio over USB

Without this timer, if the hub state doesn't change and the user program doesn't output anything, the hub won't be able to detect that the host software has closed (because it never sends anything which could time out). Keeping this state reasonably up-to-date is important for the hub auto-power-off functionality.

laurensvalk · 2025-08-05T14:24:14Z

Done, thanks! Yes, let's build on this in follow up commits as needed.

ArcaneNibble force-pushed the usb branch from 7d081d1 to 1b02ac6 Compare July 30, 2025 22:15

dlech reviewed Jul 31, 2025

View reviewed changes

laurensvalk mentioned this pull request Jul 31, 2025

[EV3] Various issues to dig into pybricks/support#2301

Closed

ArcaneNibble force-pushed the usb branch from 1b02ac6 to efa93ee Compare August 1, 2025 10:15

ArcaneNibble added 7 commits August 5, 2025 15:52

pbio/drv/usb/usb_ev3.c: Handle multiple TX completions in ISR

35f7835

pbio/drv/usb/usb_ev3.c: Implement command and status handling

492892e

This handles Pybricks commands, status flag reporting, and timeouts. stdout is not yet handled.

pbio/drv/usb/usb_ev3.c: Correctly report host connection status

c85cf2a

pbio/drv/usb/usb_ev3.c: Implement stdout handling

e6ead10

This replaces the old stopgap stdout logic with stdio over USB

laurensvalk force-pushed the usb branch from efa93ee to 0198f7b Compare August 5, 2025 13:53

laurensvalk merged commit 0198f7b into pybricks:master Aug 5, 2025
17 checks passed

Uh oh!

[EV3] Implement Pybricks USB protocol #361

[EV3] Implement Pybricks USB protocol #361

Uh oh!

Conversation

ArcaneNibble commented Jul 30, 2025

Uh oh!

coveralls commented Jul 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Jul 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ArcaneNibble commented Jul 30, 2025

Uh oh!

dlech commented Jul 31, 2025

Uh oh!

dlech left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

dlech Jul 31, 2025

Choose a reason for hiding this comment

Uh oh!

ArcaneNibble Aug 1, 2025

Choose a reason for hiding this comment

Uh oh!

dlech Jul 31, 2025

Choose a reason for hiding this comment

Uh oh!

ArcaneNibble Aug 1, 2025

Choose a reason for hiding this comment

Uh oh!

dlech Jul 31, 2025

Choose a reason for hiding this comment

Uh oh!

ArcaneNibble Aug 1, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

laurensvalk commented Jul 31, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ArcaneNibble commented Aug 1, 2025

Uh oh!

laurensvalk commented Aug 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

laurensvalk commented Aug 5, 2025

Uh oh!

ArcaneNibble commented Aug 5, 2025

Uh oh!

Uh oh!

laurensvalk commented Aug 5, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

coveralls commented Jul 30, 2025 •

edited

Loading

github-actions bot commented Jul 30, 2025 •

edited

Loading

laurensvalk commented Jul 31, 2025 •

edited

Loading

laurensvalk commented Aug 4, 2025 •

edited

Loading