vApplicationMallocFailedHook when using queues

Kernel
1

Hello!

I have got STM32F429, CubeIDE and FreeRTOS
I use UART, Ethernet(netconn API), SDIO
I

#define configMINIMAL_STACK_SIZE                (4096)
#define configTOTAL_HEAP_SIZE                   65536
#define TCPIP_THREAD_STACKSIZE 4096
#define DEFAULT_THREAD_STACKSIZE 4096
#define TCP_MSS 4096
#define DEFAULT_TCP_RECVMBOX_SIZE       2048
#define INTERFACE_THREAD_STACK_SIZE (4096)

I use appropriate queues to each peripheral
I have got the following queue structure

typedef struct MesNode_
{  
  char* message;
  size_t size;
} MesNode;

I have got the following buffers

#define BUFFER_SIZE 4096
#define Middle_BUFFER_SIZE 1024
char uartBuffer[BUFFER_SIZE];
char TCPBuffer[BUFFER_SIZE];

Im tryng to use queues with circular with the following circular buffer

MesNode make_msg_circ(const char* buffer, const char* begin, const char* end)
{
  MesNode mNode;
  mNode.size = (end > begin) ? end - begin: BUFFER_SIZE - (begin - end);
  mNode.message = pvPortMalloc(msgNode.size + 1);
  if (end > begin) {
    memcpy(msgNode.message, begin, msgNode.size);
  }
  else {
    size_t tailSize = buffer + BUFFER_SIZE - begin;
    size_t headSize = end - buffer;
    memcpy(msgNode.message, begin, tailSize);
    memcpy(msgNode.message + tailSize, buffer, headSize);
  }
  mNode.message[msgNode.size] = '\0';
  return mNode;
}

I applicate buffer in the following function of data processing. What function get last adress of pointer in buffer

typedef void (*process_fa)(MesNode* mNode);

char* data_process(osThreadId_t thread, char* beginBuffer)
{
  DMA_HandleTypeDef* dmaRx;
  char* uartBuffer;
  process_fa process_func;
  find_circ_fa find_func;
  uint16_t dmaRxv;

  if (thread == uartHandle) {
    dmaRx = dmaRxPpm;
    uartBuffer = uartPpmBuffer;
    process_func = data_process_ppm;
    find_func = find_circ_end_lsn;
  }
 
  else if (thread == TCPTxHandle) {
	  dmaRx = dmaRxPpm;
	  if(!tcpbuf)
	  {
		  uartBuffer = uartPpmBuffer;
		  process_func = data_process_ttcp;
		  find_func = find_circ_end_lsn;
	  }
	  else
	  {
		  uartBuffer = TCPBuffer;
		  dmaRxv = 	rx_size;
		  process_func = data_process_ftcp2;
		  find_func = find_circ_end_lsn;//find_circ_end_line;
	  }
	  
  }

  else if (thread == TCPProcHandle) {
	  	  dmaRxv = lenc;//dmaTCP.length;
	      uartBuffer = TCPBuffer;

	      process_func = data_process_ftcp;
	      find_func = find_circ_end_line;


  	  }

  else {
    return beginBuffer;
  }

  vPortEnterCritical();
  size_t dmaRxSize;
  if(tcpbuf)
	  dmaRxSize = BUFFER_SIZE - dmaRxv;
  else
	  dmaRxSize = BUFFER_SIZE - dmaRx->Instance->NDTR;
  char* endRxDmaBuffer = uartBuffer + dmaRxSize;
  size_t lenRxData = (endRxDmaBuffer >= beginBuffer) ? endRxDmaBuffer - beginBuffer:
      dmaRxSize + uartBuffer + BUFFER_SIZE - beginBuffer;
  vPortExitCritical();

  char* startPosition = beginBuffer;
  char* endPosition;
  char* startFind = beginBuffer;
  size_t lenTail = lenRxData;

  while (1) {
    endPosition = find_func(uartBuffer, startFind, lenTail);
    if (endPosition) {
      MesNode mNode = make_msg_circ(uartBuffer, startPosition, endPosition);
      lenTail -= mNode.size;
      process_func(&mNode);
      startPosition = endPosition;
      startFind = endPosition;
      continue;
    }
    break;
  }
  return startFind;

}

I suppose what portmalloc problem cause definition of mNode.size but still found the reason
Can anybody help to clarify this moment?

I operate the 1K data and tranfer it

2

So what’s your actual problem ? Is vApplicationMallocFailedHook invoked and you don’t know why ?
Did you verify that the size calculation is always right and you have sufficient heap available ?
Is the heap exhausted later on perhaps due to a memory leak somewhere else ?
Which heap implementation do you use ?

BTW are you sure this configuration is right ?

With the default/standard MTU size 1500 the MSS is 1460 (MTU - IP/TCP header size).

3

Thank you for your replay!
I use heap4 and increased TCP_MSS due to dividied meassage during transfer
Yes I stopped after about 5 min of transfer in debug I stopped in`vApplicationMallocFailedHook. I see what msg.size not stable in debug but cant find it in code

Also maybe heap leak cause of netconn API using

Pls help!

4

Not sure what you mean with divided message. It’s TCP and this is a stream protocol i.e. there is no such thing as a message or datagram known from UDP.
A receiver needs to take care to assemble an application layer message correctly according to your application protocol (sending messages).
Regarding the memory leak I afraid no one can help you … it’s up to you to find the reason/bug in your code where you forgot to free up buffers or messages after use.
Good luck !