STM32H743的FDCAN导致CAN发送线程卡死的处理方法

芯片型号STM32H743IIT6，测试时发现如果外面没有连接CAN设备，程序调用CAN发送时会一直等待发送反馈，导致相关线程挂起。
分析发现是卡在can.c文件的168行_can_int_tx函数：rt_completion_wait(&(tx_tosnd->completion), RT_WAITING_FOREVER);

rt_inline int _can_int_tx(struct rt_can_device *can, const struct rt_can_msg *data, int msgs)
{
int size;
struct rt_can_tx_fifo *tx_fifo;
RT_ASSERT(can != RT_NULL);
size = msgs;
tx_fifo = (struct rt_can_tx_fifo *) can->can_tx;
RT_ASSERT(tx_fifo != RT_NULL);
while (msgs)
{
rt_base_t level;
rt_uint32_t no;
rt_uint32_t result;
struct rt_can_sndbxinx_list tx_tosnd = RT_NULL;
rt_sem_take(&(tx_fifo->sem), RT_WAITING_FOREVER);
level = rt_hw_interrupt_disable();
tx_tosnd = rt_list_entry(tx_fifo->freelist.next, struct rt_can_sndbxinx_list, list);
RT_ASSERT(tx_tosnd != RT_NULL);
rt_list_remove(&tx_tosnd->list);
rt_hw_interrupt_enable(level);
no = ((rt_uint32_t)tx_tosnd - (rt_uint32_t)tx_fifo->buffer) / sizeof(struct rt_can_sndbxinx_list);
tx_tosnd->result = RT_CAN_SND_RESULT_WAIT;
if (can->ops->sendmsg(can, data, no) != RT_EOK)
{
/ send failed. */
level = rt_hw_interrupt_disable();
rt_list_insert_after(&tx_fifo->freelist, &tx_tosnd->list);
rt_hw_interrupt_enable(level);
rt_sem_release(&(tx_fifo->sem));
continue;
}
can->status.sndchange = 1;
rt_completion_wait(&(tx_tosnd->completion), RT_WAITING_FOREVER);
说明一直在等待完成信号，而发送完成信号的地方在can.c文件的900行rt_hw_can_isr函数：rt_completion_done(&(tx_fifo->buffer[no].completion));

case RT_CAN_EVENT_TX_DONE:
case RT_CAN_EVENT_TX_FAIL:
{
    struct rt_can_tx_fifo *tx_fifo;
    rt_uint32_t no;
    no = event >> 8;
    tx_fifo = (struct rt_can_tx_fifo *) can->can_tx;
    RT_ASSERT(tx_fifo != RT_NULL);
    if ((event & 0xff) == RT_CAN_EVENT_TX_DONE)
    {
        tx_fifo->buffer[no].result = RT_CAN_SND_RESULT_OK;
    }
    else
    {
        tx_fifo->buffer[no].result = RT_CAN_SND_RESULT_ERR;
    }
    rt_completion_done(&(tx_fifo->buffer[no].completion));
    break;
}

然后想到如果can总线没有其他设备，CAN发送报文应该属于出错的情况，查看drv_fdcan.c文件中关于几种中断的处理，发现故障后的回调函数里没有调用rt_hw_can_isr。
于是参考HAL_FDCAN_TxBufferCompleteCallback函数的处理方式，对HAL_FDCAN_ErrorCallback进行了如下处理。

void HAL_FDCAN_ErrorCallback(FDCAN_HandleTypeDef *hfdcan)
{
rt_uint32_t tmp_u32Errcount;
rt_uint32_t tmp_u32status;
uint32_t ret = HAL_FDCAN_GetError(hfdcan);
if(hfdcan->Instance == FDCAN1)
{
#ifdef BSP_USING_FDCAN1
//can1
if( (ret & FDCAN_IT_ARB_PROTOCOL_ERROR) &&
(hfdcan->Instance->CCCR & FDCAN_CCCR_INIT_Msk))
{
//hfdcan->Instance->CCCR |= FDCAN_CCCR_CCE_Msk;
hfdcan->Instance->CCCR &= ~FDCAN_CCCR_INIT_Msk;
st_DrvCan1.device.status.errcode = 0xff;
}
else
{
tmp_u32Errcount = st_DrvCan1.fdcanHandle.Instance->ECR;
tmp_u32status = st_DrvCan1.fdcanHandle.Instance->PSR;
st_DrvCan1.device.status.rcverrcnt = (tmp_u32Errcount>>8)&0x000000ff;
st_DrvCan1.device.status.snderrcnt = (tmp_u32Errcount)&0x000000ff;
st_DrvCan1.device.status.lasterrtype = tmp_u32status&0x000000007;
rt_hw_can_isr(&st_DrvCan1.device, RT_CAN_EVENT_TX_FAIL);
}
#endif /BSP_USING_FDCAN1/
}
else
{
#ifdef BSP_USING_FDCAN2
if( (ret & FDCAN_IT_ARB_PROTOCOL_ERROR) &&
(hfdcan->Instance->CCCR & FDCAN_CCCR_INIT_Msk))
{
//hfdcan->Instance->CCCR |= FDCAN_CCCR_CCE_Msk;
hfdcan->Instance->CCCR &= ~FDCAN_CCCR_INIT_Msk;
st_DrvCan2.device.status.errcode = 0xff;
}
else
{
//can2
tmp_u32Errcount = st_DrvCan2.fdcanHandle.Instance->ECR;
tmp_u32status = st_DrvCan2.fdcanHandle.Instance->PSR;
st_DrvCan2.device.status.rcverrcnt = (tmp_u32Errcount>>8)&0x000000ff;
st_DrvCan2.device.status.snderrcnt = (tmp_u32Errcount)&0x000000ff;
st_DrvCan2.device.status.lasterrtype = tmp_u32status&0x000000007;
rt_hw_can_isr(&st_DrvCan2.device, RT_CAN_EVENT_TX_FAIL);
}
#endif /BSP_USING_FDCAN2/
}
}
经过测试发现即使CAN总线上没有别的设备，调用发送函数也不会一直等待，而是返回发送失败。

原作者：小白小白小白白