checking stack usage

mr-m-from-k wrote on Friday, March 01, 2019:

Hello,

I struggle with stack checks because increasing the size in TaskCreateStatic did not give a corresponding result when using usStackHighWaterMark. I work on a Cortex-M7, StackType_t is set to uint32_t.
Now, when I increase the stack size passed into TaskCreateStatic by one the result I get in uxTaskGetSystemState[n].usStackHighWaterMark increases by 2. Increasing in TaskCreateStatic by 2 also increases usStackHighWaterMark by 2, 3-> 4, 4->4 and so on. I always check at the same system state, so the real stack usage at that point is the same (I suppose).
FreeRTOS Tutorial says usStackHighWaterMark gives the remaining stack in bytes. I also looked at prvTaskCheckFreeStackSpace. Somehow I can’t make sense of it.

Any help is welcome - thanks

Martin

rtel wrote on Friday, March 01, 2019:

Stack sizes are specified in words, whereas stack high water mark is reported in bytes. Does that help?

mr-m-from-k wrote on Friday, March 01, 2019:

Hello Richard,

yes, that’s also what I have learned from the docu.
But then I would expect when increasing the stack size value passed into TaskCreate(Static) by 1 that I get 4 more out with usStackHighWaterMark and 2 -> 8 and so on. But that is not what I found.

I checked again prvTaskCheckFreeStackSpace that is called to fill usStackHighWaterMark and it has this line:
ulCount /= ( uint32_t ) sizeof( StackType_t );
which makes me think the output is in units of StackType_t.
Please kindly check that again and confirm, as I just did a quick text search. My FreeRTOS version is V10.1.1.

kind regards

Martin

richarddamon wrote on Saturday, March 02, 2019:

One other thing to note is that on some processors (like the Arm Cortex series) the stack needs to be aligned more strictly than a stack ‘word’ by the ABI, so there is a ‘round up’ aspect in the stack size. Thus the size of the stack will be rounded up to the next multiple of that stack alignment.