• State Accepted Answer
  • +1 person also asked this people also asked this
  • Locked Locked
  • Replies 6 replies
  • Subscribers 352 subscribers
  • Views 9822 views
  • Users 0 members are here
Options
Related
How was your experience today?
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

4 Bytes repeating when using memcpy

Krishna prasad

Hi ,

We are using Cortex-M4(STM32F429IIH6) processor for our IOT project. We are seeing a data corruption when memcpy used, the corruption pattern was identified by feeding a Fixed pattern of size 512 Bytes from Src to Dest using memcpy, and it is not happening in all the iterations.

If you see the image attached below that Data "6 N 7 O" is replaced with 2 Bytes("Highlighted with orange color" 4L).  

Can any one tells that what can be the reason for this wired behavior? We are in a critical phase of our project.

 

Regards,

Krishna Prasad

Parents
  • 0 in reply to Joseph Yiu
    Hi Joseph Yiu,
    Thank you for the valuable suggestion.
    1) We kept the process memcpy inside Critical section.
    The block of code is added below.
    /******************Critical section Handling********************/
    #define F076_ENTER_CRITICAL_P(a) do{ \
    (*a)=__get_PRIMASK();\
    __disable_irq();\
    __DSB();\
    __ISB(); }while(0)

    #define F076_EXIT_CRITICAL_P(a) __set_PRIMASK(*a)
    /*****************Critical section Handling*********************/

    Function fctnTestCase(){
    intr_status_t criticalEntryIrqStat;
    F076_ENTER_CRITICAL_P(&criticalEntryIrqStat);
    memcpy(g_BufferForParsing.buffer, pMsgFromHost->msgBegin, pMsgFromHost->msgSize);
    g_BufferForParsing.size = pMsgFromHost->msgSize;
    F076_EXIT_CRITICAL_P(&criticalEntryIrqStat);

    }

    So the Pre-emption will be restricted here, but issue still observed.

    2) Ensured double word stack alignment at function boundaries.
    In icf file saw that alignment is 8 Bytes, and checked under Registers\System Control Block\CCR\STACKALIGN also set to 1.
    define symbol __ICFEDIT_size_cstack__ = 0x4000;
    ...
    define block CSTACK with alignment = 8, size = __ICFEDIT_size_cstack__ { };
    ...

    Regards,
    Krishna
Reply
  • 0 in reply to Joseph Yiu
    Hi Joseph Yiu,
    Thank you for the valuable suggestion.
    1) We kept the process memcpy inside Critical section.
    The block of code is added below.
    /******************Critical section Handling********************/
    #define F076_ENTER_CRITICAL_P(a) do{ \
    (*a)=__get_PRIMASK();\
    __disable_irq();\
    __DSB();\
    __ISB(); }while(0)

    #define F076_EXIT_CRITICAL_P(a) __set_PRIMASK(*a)
    /*****************Critical section Handling*********************/

    Function fctnTestCase(){
    intr_status_t criticalEntryIrqStat;
    F076_ENTER_CRITICAL_P(&criticalEntryIrqStat);
    memcpy(g_BufferForParsing.buffer, pMsgFromHost->msgBegin, pMsgFromHost->msgSize);
    g_BufferForParsing.size = pMsgFromHost->msgSize;
    F076_EXIT_CRITICAL_P(&criticalEntryIrqStat);

    }

    So the Pre-emption will be restricted here, but issue still observed.

    2) Ensured double word stack alignment at function boundaries.
    In icf file saw that alignment is 8 Bytes, and checked under Registers\System Control Block\CCR\STACKALIGN also set to 1.
    define symbol __ICFEDIT_size_cstack__ = 0x4000;
    ...
    define block CSTACK with alignment = 8, size = __ICFEDIT_size_cstack__ { };
    ...

    Regards,
    Krishna
Children
  • 0 Arm Employee Badge in reply to Krishna prasad
    Just by visual inspection I don't see anything stands out.
    Other areas to check:
    - any chance of stack overflow?
    - any possibility that pMsgFromHost got changed by other hardware while memcpy is on going?
  • 0 in reply to Krishna prasad
    1. check address alignment
    printf("%p\n", g_BufferForParsing.buffer);
    printf("%p\n", pMsgFromHost->msgBegin);

    2. try to replace memcpy with the following code to see any difference.

    volatile unsigned int * dst = (volatile unsigned int*) g_BufferForParsing.buffer;
    volatile unsigned int * src = (volatile unsigned int*) pMsgFromHost->msgBegin;
    int sz = pMsgFromHost->msgSize / 4;
    for (int i = 0; i < sz; i++) {
    dst[i] = src[i];
    }