• State Accepted Answer
  • Replies 3 replies
  • Subscribers 23 subscribers
  • Views 1463 views
  • Users 0 members are here
Options
Related

Semihosting output is garbled

paperchalice

I create a STM32CubeMX based STM32F407VET project, and try to use printf in project:

int main(void)
{

  /* USER CODE BEGIN 1 */

  /* USER CODE END 1 */

  /* MCU Configuration--------------------------------------------------------*/

  /* Reset of all peripherals, Initializes the Flash interface and the Systick. */
  HAL_Init();

  /* USER CODE BEGIN Init */

  /* USER CODE END Init */

  /* Configure the system clock */
  SystemClock_Config();

  /* USER CODE BEGIN SysInit */

  /* USER CODE END SysInit */

  /* Initialize all configured peripherals */
  MX_GPIO_Init();
  /* USER CODE BEGIN 2 */
  printf("Hello semihosting!\n");
  /* USER CODE END 2 */

  /* Infinite loop */
  /* USER CODE BEGIN WHILE */
  while (1)
  {
    /* USER CODE END WHILE */
    printf("Hello semihosting!\n");
    /* USER CODE BEGIN 3 */
  }
  /* USER CODE END 3 */
}
But the output is garbled...
I have the following solution configuration:
  # List different hardware targets that are used to deploy the solution.
  target-types:
    - type: STM32F407VETx
      device: STM32F407VETx
      target-set:
        - set:
          images:
            - project-context: Project.Debug
          debugger:
            name: ST-Link@pyOCD
            clock: 4000000
            protocol: swd
            telnet:
              - mode: monitor
The output captured by serial monitor in hex is
20 11 03 00 08 a9 2a 00 08 79 2a 00 08 a5 2a 00
08 39 1f
Same project using STM32CubeIDE vscode is OK, with correct rdimon configuration.
How to let it output correctly?
  • +1 Arm Employee Badge

    I tested this on an STM32F769I-DISCO project. I started from the CubeMX basic Solution template, generated the code with STM32CubeMX without any extra configuration, selected Telnet/Serial Monitor in Mange Solution and added a simple printf() for semihosting output.

    The output appeared when I connected the VS Code Serial Monitor to the pyOCD STDIO/Telnet TCP port, as shown in the screenshot. So I would suggest checking the Serial Monitor settings first.

    Note: the port number 4444 is reserved in Keil Studio for semihosting, and it can be checked in the generated *.cbuild-run.yml.

    One extra thing I had to fix was the CubeMX-generated MDK startup file. The startup file still had a non-microlib __user_initial_stackheap block. With AC6 linker scripts, the heap and stack are already provided by ARM_LIB_HEAP and ARM_LIB_STACK, so this old block can conflict with the linker-script setup. In my case, removing the non-microlib __user_initial_stackheap block from startup_stm32f769xx.s fixed standard printf(), preventing from HardFault. So it may be worth checking that block as well.

  • 0 in reply to Jen-Tse Huang

    Thanks! The problem is the User Stack and Heap initialization part in startup_stm32f407xx.s, it can output correctly if I remove this part.

  • +1 Arm Employee Badge

    Glad to hear that it works on your side as well.

    Quick addition: if you'd rather not touch the startup file at all, you can fix it from the linker script side instead. Keep the startup's STACK/HEAP areas and __user_initial_stackheap as they are, and just tell the default scatter file to place them instead of reserving its own:

    // replace ARM_LIB_HEAP / ARM_LIB_STACK with:
STARTUP_HEAP  AlignExpr(+0, 8) { *(HEAP)  }
STARTUP_STACK AlignExpr(+0, 8) { *(STACK) }

    Either way, pick one place (startup code or linker script) to manage them and keep it consistent.