ESP32can_basic example hangs after disconnect of TX-line

[TLS1000]

Hi, I just tried the esp32_can basic example and it worked right-away; thanks!

BUT: when I disconnect the tx line temporarily communication doesn't resume, even after resetting via the 'EN' button on the doitV1 board doesn help. I can only get it back running after a new power-up

I connected two ESP32 doitV1 boards together via the gpio2(TX) and gpio4(rx) and loaded the example sketch in each one

[TLS1000]

update: it also hangs after disconnection of one the the two the CAN lines, which is real-life event.

[allex1978]

Yes. it hangs after third send attempt to disconnected line. I looked line via logic analazer and it endlessly repeats sending the last packet to the line with a very short intervals.

[jerryyhl]

Hi, I'm encountering this same issue, did anyone figure out a fix for this?

[rocrail]

I worked around this by replacing portMAX_DELAY with 500 in CAN.c This will return after max. 500ms with zero if it could not be send.

int CAN_write_frame(const CAN_frame_t *p_frame) {
  if (sem_tx_complete == NULL) {
    return -1;
  }
 
  // Write the frame to the controller
  CAN_write_frame_phy(p_frame);
 
  // wait for the frame tx to complete
  int rc = xSemaphoreTake(sem_tx_complete, 500);
  return rc;
}

But it would be better if the original library would not block if the send message is not acknowledge...

[Satnet]

Doing this is a good solution and the proper way I would say

xSemaphoreTake(sem_tx_complete, 10 * portTICK_PERIOD_MS);

Itt uses tick in the low end code and adds 10 ms. So it's real time 10ms. Sending a regular frame takes about 6 - 7ms so you could make it less if you want to. Works fine! You'll find this inside the CAN.c file inside the function with name CAN_write_frame(const CAN_frame_t *p_frame)

[katealhola]

Best solution would be having int CAN_write_frame(const CAN_frame_t *p_frame, const tTickType_t xTicksToWait)

Current implementation where CAN_write_frame can hang infinite time is about blocker issue in real time system. Even having separate task just for sending and then having task hang infinite time is blocker issue. I have not find any other fix than patching this library because sem_tx_complete is static and there is no way from external code to check does function hang or no.

[reshma26994]

using same library rx starts after that tx stops till restart and tx stops after rx starts. please give me replay as soon as possible