UART Debugging log on the same UART as Modbus

[Olgidos]

Is there an easy fix, so that I can send debug messages via the same UART interface from differen RTOS tasks like this:

        HAL_GPIO_WritePin(UART_Enable_GPIO_Port, UART_Enable_Pin, GPIO_PIN_SET);
        hstatus = HAL_UART_Transmit(gHuart, (uint8_t*) ptr, len, 100);
        HAL_GPIO_WritePin(UART_Enable_GPIO_Port, UART_Enable_Pin, GPIO_PIN_RESET);

it currently works until I get the first Modbus request, then both the debug log and modbus stops working, even so no task is hanging.

i also tried to use semaphores:

int _write(int fd, char *ptr, int len) {

    QueueHandle_t semaphoreHandle = (QueueHandle_t)ModbusH.ModBusSphrHandle;
    // Take the semaphore

    HAL_StatusTypeDef hstatus;
    if (fd == STDOUT_FILENO || fd == STDERR_FILENO) {

     if (xSemaphoreTake(semaphoreHandle, (TickType_t)100) == pdTRUE) {
        //prevent switching
         taskENTER_CRITICAL();
         __disable_irq();

        HAL_GPIO_WritePin(UART_Enable_GPIO_Port, UART_Enable_Pin, GPIO_PIN_SET);
        hstatus = HAL_UART_Transmit(gHuart, (uint8_t*) ptr, len, 100);
        HAL_GPIO_WritePin(UART_Enable_GPIO_Port, UART_Enable_Pin, GPIO_PIN_RESET);

        xSemaphoreGive(semaphoreHandle);
         taskEXIT_CRITICAL();
         __enable_irq();
        //till here
     }
        if (hstatus == HAL_OK)
            return len;
        else
            return EIO;
    }
        errno = EBADF;
        return -1;

}

[alejoseb]

I would not call a "fix" but rather a hack what you need here, and for sure it would break Modbus. This protocol is time sensitive and uses states to handle the telegrams according to a master-slave architecture not compatible with your modifications. Eventually it could work, but you need to first stop the serial port an reconfigure it for Modbus after every debug message.

Instead, I recommend you to explore overloading the Modbus protocol itself, you can for example have a specific memory area were you dump debug messages and then you retrieve that using the multi-register read function, but always maintaining the master-slave architecture.

[Olgidos]

Ok I see. This sounds quite error prone and I already expected that it would be unlikely to get it to work. Then I will try it this way.