[Gen 3] Fix USB CDC write triggering an assertion failure; migrates wiring/usbserial to be run under test runner, more usb tests by avtolstoy · Pull Request #2939 · particle-iot/device-os

avtolstoy · 2026-08-06T07:44:08Z

Problem

On Gen 3 a non-blocking Serial.write() racing a full TX buffer could SOS the device. hal_usb_cdc_pvt_send_data() SPARK_ASSERTed that app_fifo_write() succeeds, but a full FIFO returns NRF_ERROR_NO_MEM - so writing into a full buffer was an assertion failure instead of a short write. SerialUSBStream had the same problem one layer up: errors from the HAL propagated out of read()/peek()/write() instead of being reported as "0 bytes done". This is a 6.5.0 regression introduced in USB Serial/UART serial buffer read/write/peek + wait event #2935
wiring/usbserial was a manual test, so it never ran on HIL.
USB coverage beyond CDC data was basically nonexistent: control-request transport (chunking, concurrency, cancellation), descriptors/WCID, device modes and bus-reset robustness had no automated tests. user/tests/app/usb_ctrl_request predates the test runner and never runs.

Solution

hal_usb_cdc_pvt_send_data() returns SYSTEM_ERROR_NO_MEMORY on a full FIFO instead of asserting; SerialUSBStream treats it as a zero-length transfer.
Migrated wiring/usbserial under the test runner (20 tests). Host side drives the CDC tty via a new UsbCdcPort helper (user/tests/integration/test/usb_cdc.js): resolves the tty by USB serial number through sysfs, toggles DTR via open/close, creates TX backpressure by pausing the reader. Covers blocking/non-blocking TX stress, RX stress, isConnected() transitions, RX buffer fill, end()/begin() while the port is closed.
New wiring/usb suite (18 tests, gen3 + gen4) - everything USB that isn't CDC data:
- system control requests (CTRL_REQUEST_ECHO) across the allocation/chunking boundaries (1–10000 bytes: pooled vs async alloc at 64, host-side chunked SEND/RECV above 4096), 4 concurrent requests
- app custom request path end-to-end, including deferred completion so requests stay PENDING across multiple host CHECK polls and complete out of order
- raw vendor requests (bRequest 0x50: DEVICE_ID, SYSTEM_VERSION), raw INIT->RESET->CHECK cancellation, DIAGNOSTIC_INFO
- descriptor assertions: VID/PID vs @particle/device-constants, iSerialNumber == device ID, per-platform interface layout, MS OS string + WCID compat ID / extended properties
- listening mode enter/leave observed from both host and device; control requests keep working in listening mode
- host-initiated bus reset x5 with recovery and Serial.end()/begin() re-enumeration stress
Supersedes user/tests/app/usb_ctrl_request (GET_MODULE_INFO is already covered by the ota/* suites).
Test runner application: the request handler gained an e echo command with an optional delay - delayed requests are completed from testAppLoop(), same pattern as the old usb_ctrl_request app. Removed the dead RequestHandler::instance() singleton: the live handler is TestSuite::reqHandler_, and the stale singleton silently absorbed the loop hook (deferred requests queued on one object, drained on the other). Added test/platform_util.js - wiring specs can't require() node modules directly (module resolution uses the spec's realpath, which is outside integration/), so helpers with npm deps have to live in integration/test/.

Steps to Test

device-os-test --device-os-dir=. run <platform> wiring/usbserial wiring/usb

scott-brust

Works for me on msom

➜  device-os git:(fix/gen3-usb-cdc-migrate-usbserial-tests) ✗ device-os-test --device-os-dir=. run msom wiring/usbserial wiring/usb


  USB
    msom
      systemThread=enabled
        ✓ USB_00_SystemEchoShort (4.8s)
        ✓ USB_01_SystemEchoLarge (1s)
        ✓ USB_02_SystemEchoConcurrent
        ✓ USB_03_AppCustomEcho
        ✓ USB_04_AppCustomEchoDeferred (3.3s)
        ✓ USB_05_DiagnosticInfo
        ✓ USB_06_RawVendorRequests (1.1s)
        ✓ USB_07_ServiceRequestCancellation
        ✓ USB_08_DeviceDescriptor
        ✓ USB_09_InterfaceLayout
        ✓ USB_10_MsftOsStringDescriptor
        ✓ USB_11_WcidCompatIdDescriptor
        ✓ USB_12_EnterListeningMode (1.4s)
        ✓ USB_13_DeviceObservesListeningMode (1.3s)
        ✓ USB_14_DeviceObservesNormalMode
        ✓ USB_15_HostBusResetRecovery (16.9s)
          [2a] Hub-cycle disabled, device not found by portPath
          [2a] Hub-cycle disabled, device not found by portPath
          [2a] Hub-cycle disabled, device not found by portPath
          [2a] Hub-cycle disabled, device not found by portPath
Opened USB CDC port /dev/cu.usbmodem1102 for 0a10aced202194944a04b000
        ✓ USB_16_DeviceEndBeginStress (34.6s)
        ✓ USB_17_FinalCdcSanity

  USBSerial
    msom
      systemThread=enabled
Opened USB CDC port /dev/cu.usbmodem1102 for 0a10aced202194944a04b000
        ✓ USBSERIAL_00_RingBufferHelperIsSane (4.9s)
        ✓ USBSERIAL_01_SerialDoesNotDeadlockWhenInterruptsAreMasked
        ✓ USBSERIAL_02_ReadWrite
        ✓ USBSERIAL_03_ReadWriteVerifiesHostData
        ✓ USBSERIAL_04_isConnectedInitially
        ✓ USBSERIAL_05_ClosedPortWritesFailWithoutBlocking
        ✓ USBSERIAL_06_isConnectedDetectsReopenedPort
        ✓ USBSERIAL_07_EndBeginWhilePortIsClosed (4.3s)
        ✓ USBSERIAL_08_RxBufferSetup (1.1s)
        ✓ USBSERIAL_09_RxBufferFillsCompletelyFirst
        ✓ USBSERIAL_10_RxBufferFillsCompletelySecond
        ✓ USBSERIAL_11_RxBufferFillsCompletelyThird
        ✓ USBSERIAL_12_NonBlockingWriteStressSetup
        ✓ USBSERIAL_13_NonBlockingWriteStressHandlesBackpressure
        ✓ USBSERIAL_14_BlockingWriteStressWritesEveryBuffer (2.3s)
        ✓ USBSERIAL_15_NonBlockingCharacterWriteStressSetup
        ✓ USBSERIAL_16_NonBlockingCharacterWriteStressHandlesBackpressure (1.8s)
        ✓ USBSERIAL_17_BlockingCharacterWriteStressWritesEveryByte (16.3s)
        ✓ USBSERIAL_18_DeviceReceiveStressSetup
        ✓ USBSERIAL_19_DeviceReceiveStress (2.8s)


  38 passing (2m)

user/tests/app/usb_ctrl_request predates the test runner and never runs
Do we want to delete this app since its effectively been replaced?

scott-brust · 2026-08-26T16:28:36Z

 }

 int SerialUSBStream::skip(size_t size) {
    return read(nullptr, size);


I think there is an issue with this.
On gen3 read(nullptr, size) -> hal_usb_cdc_pvt_recv_data() -> app_fifo_read() with a null buffer will be interpreted as querying the length of the fifo and return NRF_SUCCESS with the size skipped set to the entire length of the fifo:

uint32_t app_fifo_read(app_fifo_t * p_fifo, uint8_t * p_byte_array, uint32_t * p_size) { ... const uint32_t byte_count = fifo_length(p_fifo); ... (*p_size) = byte_count; // Check if application has requested only the size. if (p_byte_array == NULL) { return NRF_SUCCESS; } ...

I think we need to have a dummy buffer to actually consume the bytes

skip() doesnt appear to be used right now but if someone does use it, I dont think it will work

Yup, looks like this is broken. Will fix and add tests, as it wasn't caught. Good catch.

There aren't any tests for SerialUSBStream right now, right? Its only used for USB tethering, which have their own tests on top of it I suppose.

yeah, but HAL_USB_USART_Peek_Buffer() and others are testable.

scott-brust · 2026-08-26T17:05:49Z

+    if (r == SYSTEM_ERROR_NO_MEMORY) {
+        return 0;


For gen4, SYSTEM_ERROR_NO_MEMORY means there is no data to peek?

Kinda weird, I admit. I'll take another look at peek/read in particular

scott-brust · 2026-08-26T17:07:28Z

+    uint32_t result = app_fifo_write(&m_usb_instance.tx_fifo, data, &bytesToWrite);
+    if (result == NRF_ERROR_NO_MEM) {
+        return SYSTEM_ERROR_NO_MEMORY;
+    }
+    SPARK_ASSERT(result == NRF_SUCCESS);


avtolstoy added 3 commits August 6, 2026 14:42

[test] migrate wiring/usbserial to be run under test runner

b8dff97

[gen3] usb: fix send buffer full buffer assertion failure

fde111d

minor

f212bcc

avtolstoy added this to the 6.5.1 milestone Aug 6, 2026

avtolstoy added the bug label Aug 6, 2026

avtolstoy added 2 commits August 19, 2026 02:32

[test] port more usb tests to be run under test runner

c0bfb0b

[test] require fixup

e05af52

avtolstoy changed the title ~~[Gen 3] Fix USB CDC write triggering an assertion failure; migrates wiring/usbserial to be run under test runner~~ [Gen 3] Fix USB CDC write triggering an assertion failure; migrates wiring/usbserial to be run under test runner, more usb tests Aug 25, 2026

minor

aefe559

avtolstoy requested review from scott-brust and technobly August 25, 2026 08:26

avtolstoy marked this pull request as ready for review August 25, 2026 08:26

scott-brust approved these changes Aug 26, 2026

View reviewed changes

Navigation Menu

Uh oh!

[Gen 3] Fix USB CDC write triggering an assertion failure; migrates wiring/usbserial to be run under test runner, more usb tests - #2939

[Gen 3] Fix USB CDC write triggering an assertion failure; migrates wiring/usbserial to be run under test runner, more usb tests#2939
avtolstoy wants to merge 6 commits into
developfrom
fix/gen3-usb-cdc-migrate-usbserial-tests

avtolstoy commented Aug 6, 2026 •

edited

Loading

Uh oh!

scott-brust left a comment

Uh oh!

scott-brust Aug 26, 2026

Uh oh!

avtolstoy Aug 26, 2026

Uh oh!

scott-brust Aug 26, 2026

Uh oh!

avtolstoy Aug 26, 2026

Uh oh!

scott-brust Aug 26, 2026

Uh oh!

avtolstoy Aug 26, 2026

Uh oh!

scott-brust Aug 26, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

avtolstoy commented Aug 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

Solution

Steps to Test

Uh oh!

scott-brust left a comment

Choose a reason for hiding this comment

Uh oh!

scott-brust Aug 26, 2026

Choose a reason for hiding this comment

Uh oh!

avtolstoy Aug 26, 2026

Choose a reason for hiding this comment

Uh oh!

scott-brust Aug 26, 2026

Choose a reason for hiding this comment

Uh oh!

avtolstoy Aug 26, 2026

Choose a reason for hiding this comment

Uh oh!

scott-brust Aug 26, 2026

Choose a reason for hiding this comment

Uh oh!

avtolstoy Aug 26, 2026

Choose a reason for hiding this comment

Uh oh!

scott-brust Aug 26, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

avtolstoy commented Aug 6, 2026 •

edited

Loading