CCS/TMS320C6748: NIMU receive function issue

Tool/software: Code Composer Studio

Hello,

I am using LCDKC6748 and processor_sdk_rtos_omapl138_6_01_00_08.

I run "NIMU_emacExample_lcdkOMAPL138C674xBiosExampleProject", and sending ICMP echo requests continuously from a remote PC.

There is no problem most of the time. However, very rarely errors occur.

When the error occurred, the received data in the memory has been corrupted.
It is occurring in the function "EmacRxPkt()" in the "pdk_omapl138_1_0_10\packages\ti\transport\ndk\nimu\src\v6\nimu_eth.c".

static void
EmacRxPkt
(
    uint32_t              port_num,
    EMAC_PKT_DESC_T*    p_desc
)
{
    uint32_t              dataOffset, protocol;
    uint8_t*              pBuffer;

    (void)port_num;

    PBM_setValidLen((PBM_Handle) p_desc->AppPrivate, p_desc->ValidLen);
    PBM_setIFRx((PBM_Handle) p_desc->AppPrivate, pPDI[NIMU_PORT_NUM_USED]->hEther );

    /* Handle raw frames separately, i.e. check the
    * Ethernet Protocol type in this packet and see
    * if its a well known ether type. If so, this is normal
    * IP stream, enqueue this is in usual Rx queue and let the
    * stack know that a packet has arrived for it. However, if
    * the ethernet type in the packet is not a well known one,
    * this could be a custom raw ethernet packet, enqueue it
    * separately in the Raw Rx queue and notiofy stack. The Raw
    * Ethernet packets when being handed over are given
    * preferential treatment and are serviced before the normal
    * IP traffic. Hence the 2 queues.
    */
    dataOffset = PBM_getDataOffset((PBM_Handle) p_desc->AppPrivate);
    pBuffer = (Uint8 *)Convert_CoreLocal2GlobalAddr((uint32_t)(PBM_getDataBuffer((PBM_Handle) p_desc->AppPrivate))) + dataOffset;

    /* Extract the ethernet type from the packet. */
    protocol = ( pBuffer[12] << 8) | pBuffer[13] ;
    protocol = (protocol & 0xFFFFu);

    /* Is it a standard ethernet type? */
    if (protocol != ETHERTYPE_IP && protocol != ETHERTYPE_IPv6 && protocol != ETHERTYPE_VLAN
        && protocol != ETHERTYPE_PPPOECTL && protocol != ETHERTYPE_PPPOEDATA )
    {
        /* This is a raw packet, enqueue in Raw Rx Queue */
        PBMQ_enq( &pPDI[NIMU_PORT_NUM_USED]->PBMQ_rawrx, (PBM_Handle) p_desc->AppPrivate );
    }
    else
    {            /* This is a normal IP packet. Enqueue in Rx Queue */
        PBMQ_enq( &pPDI[NIMU_PORT_NUM_USED]->PBMQ_rx, (PBM_Handle) p_desc->AppPrivate );
    }

    /* Notify NDK stack of pending Rx Ethernet packet */
    STKEVENT_signal( pPDI[NIMU_PORT_NUM_USED]->hEvent, STKEVENT_ETHERNET, 1 );

    /* Clean the cache for external addesses */
    if( ((uint32_t)(PBM_getDataBuffer(p_desc->AppPrivate)) & EMAC_EXTMEM) || (((uint32_t)(PBM_getDataBuffer(p_desc->AppPrivate)) & EMAC_MSMCMEM) == EMAC_MSMCMEM) )
    {
        CacheP_wbInv( (void *)(PBM_getDataBuffer(p_desc->AppPrivate)), PBM_getBufferLen(p_desc->AppPrivate) );
    }
}

"CacheP_wbInv()" overwrites the received data in the external memory with data in the cache.
If the latest received data is not in the cache, the latest received data in the external memory is corrupted.

Is there any problem if I use "CacheP_Inv()" instead of "CacheP_wbInv()" ?

Thank you,