RTOS: using unitiazed task_id to set the MAGIC_WORD for stack overflow check

[oresths]

In rt_CMSIS.c in svcThreadCreate(...) , rt_init_context(...) is called before a proper value is given to task_context->task_id (which is done one line after).

But rt_init_context(...) in rt_Task.c calls rt_init_stack(...) in HAL_CM.c with the following last two lines:
if (p_TCB->task_id != 0x01)
p_TCB->stack[0] = MAGIC_WORD;

That could cause a mismatch in some cases in rt_stk_check (...) in rt_System.c because
os_tsk.run->stack[0] != MAGIC_WORD

[adamgreen]

I have an uncommitted change on my machine from several months ago that attempts to address this issue. If I recall correctly, I found and fixed this issue because it was leading to a global variable being corrupted as the main task was getting set up. I never got to the bottom of all the issues I was hitting with my program at that time so I never got around to pushing up this fix. Here is my diff where I use the task_body instead of the task_id for this check. Maybe you can change the call order to fix it as well.

diff --git a/libraries/rtos/rtx/HAL_CM.c b/libraries/rtos/rtx/HAL_CM.c
index a5a2a9c..acabd65 100644
--- a/libraries/rtos/rtx/HAL_CM.c
+++ b/libraries/rtos/rtx/HAL_CM.c
@@ -52,6 +52,7 @@ BIT dbg_msg;

 /*--------------------------- rt_init_stack ---------------------------------*/

+extern int main (void);
 void rt_init_stack (P_TCB p_TCB, FUNCP task_body) {
   /* Prepare TCB and saved context for a first time start of a task. */
   U32 *stk,i,size;
@@ -88,11 +89,11 @@ void rt_init_stack (P_TCB p_TCB, FUNCP task_body) {
   p_TCB->ptask = task_body;

   /* Set a magic word for checking of stack overflow.
-   For the main thread (ID: 0x01) the stack is in a memory area shared with the
+   For the main thread the stack is in a memory area shared with the
    heap, therefore the last word of the stack is a moving target.
    We want to do stack/heap collision detection instead.
   */
-  if (p_TCB->task_id != 0x01)
+  if (task_body != (FUNCP)main)
       p_TCB->stack[0] = MAGIC_WORD;
 }

[oresths]

Thanks for answer. I just moved
OS_TID tsk = rt_get_TID ();
task_context->task_id = tsk;
in rt_CMSIS.c before rt_init_context(...).

Though it would be good to see a fix for everyone.

[0xc0170]

Hi,

did you fix it locally? I dont see any change in the rt_CMSIS.c file. Is this bug still present? If not, please close this, otherwise could you send a pull request with a fix for this issue?

Regards,
0xc0170

[oresths]

I haven’t submitted any fix. I just checked the current version in github and task_id is still used uninitialized.

Both mine and adamgreen’s solution seem to do the trick.
As I am not familiar with mbed coding practices should we change HAL_CM.c or rt_CMSIS.c ?

[0xc0170]

This should be now fixed via #827. Thanks Adam!