Enable FDCAN_IT_ARB_PROTOCOL_ERROR

[MFrie.7]

Hello
I use CAN FD and have a functioning communication with 3 participants. Now I want to activate the interrupt that notifies me when the device has lost the arbitration contest.

I forward each of the four NVIC interrupts in stm32h7xx_hal_fdcan:

void FDCAN2_IT0_IRQHandler(void)
{
HAL_FDCAN_IRQHandler(&hfdcan2);
}
When initializing CAN, I activate the interrupt:
retVal = HAL_FDCAN_ActivateNotification(m_Fdcan2Handle, FDCAN_IT_ARB_PROTOCOL_ERROR, 0);

and set the CallbackHandle beforehand:
retVal = HAL_FDCAN_RegisterTxEventFifoCallback(m_Fdcan2Handle, MyCallback);
SYSM_ASSERT_STOP(retVal == HAL_OK, retVal);
retVal = HAL_FDCAN_RegisterErrorStatusCallback(m_Fdcan2Handle, MyCallback);
SYSM_ASSERT_STOP(retVal == HAL_OK, retVal);
retVal = HAL_FDCAN_RegisterTxBufferAbortCallback(m_Fdcan2Handle, MyCallback);
SYSM_ASSERT_STOP(retVal == HAL_OK, retVal);
retVal = HAL_FDCAN_RegisterTxEventFifoCallback(m_Fdcan2Handle, MyCallback);
SYSM_ASSERT_STOP(retVal == HAL_OK, retVal);

Despite arbitration collisions, I do not end up in my callback “MyCallback”.

What have I overlooked?

Best regards

[mƎALLEm]

Hello,

First, please use </> button to share your code. See this link. (I've edited your post).

Second, according to the description of the PEA bit it might be not intended to signal a lost of arbitration but to signal a protocol error that occurred during the arbitration phase:

I may ask internally and get back to you as soon as I have an answer. Internal ticket 206406 has been submitted (not accessible by the community users).

[Saket_Om]

Hello @MFrie.7 

For better debugging, please ensure that the return value retVal is equal to HAL_OK when calling the HAL functions.

 retVal = HAL_FDCAN_ActivateNotification(m_Fdcan2Handle, FDCAN_IT_ARB_PROTOCOL_ERROR, 0);
 if (retVal != HAL_OK)
 {
 // Error handler
 }

[MFrie.7]

Hello,

Thank you for your reply and opening an internal ticket.

In the reference manual I see an other possible solution for this problem, but I can't get it to work:

I check TXBTO and TXBFC for a lost of the arbitration contest like this:

            if (TXBFC & ~TXBTO)               
            {
                ...

The problem with this is, that I now count to much lost of arbitration contest. Perhaps I need to trigger TXBRP from 1 to 0? I have not found a way (Interrupt) for that.

[mƎALLEm]

I don't think this is an efficient way as it's not exclusively to signal the loss of the arbitration but also an issue in transmission. 

[MFrie.7 · Accepted solution]

Hallo, as a workaround I made it work with the FiFo size of 1 and the FiFoEmpty-Interrupt:

    void CCan::txFifoEmptyCallback(FDCAN_HandleTypeDef *hfdcan)
    {

        uint32_t TXBRP = hfdcan->Instance->TXBRP;
        uint32_t TXBFC = hfdcan->Instance->TXBCF;
        uint32_t TXBTO = hfdcan->Instance->TXBTO;

        /* Arbitration loss or frame transmission disturbed? */
        if ((TXBFC & ~TXBTO) != 0 && !TXBRP)
        {
            /* Check if there was an error in the data or arbitration phase.
             * Lost arbitration is not rated as error (See FDCAN_PSR.LEC). */
            bool dataPhaseError = hfdcan->Instance->IR & FDCAN_FLAG_DATA_PROTOCOL_ERROR;
            bool dataArbError = hfdcan->Instance->IR & FDCAN_FLAG_ARB_PROTOCOL_ERROR;

            if(dataPhaseError == false && dataArbError == false )
            {
                /* Retransmit the Message */
                HAL_StatusTypeDef retVal;
                if(hfdcan->Instance == FDCAN1)
                {
                    retVal = HAL_FDCAN_AddMessageToTxFifoQ(hfdcan, &latestCanMessage[0].txHeader, latestCanMessage[0].dataWithSerial);
                    SYSM_ASSERT_NO_STOP(retVal == HAL_OK, hfdcan->ErrorCode);
                }
                else if(hfdcan->Instance == FDCAN2)
                {
                    retVal = HAL_FDCAN_AddMessageToTxFifoQ(hfdcan, &latestCanMessage[1].txHeader, latestCanMessage[1].dataWithSerial);
                    SYSM_ASSERT_NO_STOP(retVal == HAL_OK, hfdcan->ErrorCode);
                }
                else
                {
                    SYSM_ASSERT_STOP(0, hfdcan->Instance);
                }
            }
        }
    }

There are still some cases, in which an error in the arbitration-phase is evaluated as arbitration loss (See FDCAN_PSR.LEC), but since it will only occure rarely, its fine for my SW.

Best regards

Manuel