PA issue with pa_PATCH_OP_DELETE command in pdk_c6678_1_1_2_5

Hi, Aamir:

I have verified and confirmed that the fix is included at the PA LLD 1.2.3.3. I have also enhanced the unit test to cover this case and verified that it works. In the new test case, we remove 4 bytes after the UDP payload, and then 4 test bytes at the ned of packet.

Input packet:

static UInt8 natTPkt2[] = {
 0x00, 0x01, 0x02, 0x03, 0x04, 0xaa, 0x00, 0xe0,
 0xa6, 0x66, 0x57, 0x04, 0x08, 0x00, 0x45, 0x00,
 0x00, 0x62, 0x00, 0x00, 0x00, 0x00, 0x05, 0x11,
 0x16, 0x78, 0x9e, 0xda, 0x6d, 0x0b, 0xc8, 0xc9,
 0xca, 0x64, 0x80, 0x00, 0x11, 0x94, 0x00, 0x4e,
 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00,
 0x00, 0x00, 0x6c, 0x6d, 0x01, 0xc0, 0x01, 0xfa,
 0xce, 0x00, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79,
 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f, 0x80, 0x81,
 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89,
 0x8a, 0x8b, 0x8c, 0x8d, 0x8e, 0x8f, 0x90, 0x91,
 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99,
 0x9a, 0x9b, 0x9c, 0x9d, 0x9e, 0x9f, 0xa0, 0xa1,
 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7, 0xa8, 0xa9,
  0xba, 0xbe, 0xfa, 0xce
};

Output packet:

static UInt8 natTPkt2[] = {
 0x00, 0x01, 0x02, 0x03, 0x04, 0xaa, 0x00, 0xe0,
 0xa6, 0x66, 0x57, 0x04, 0x08, 0x00, 0x45, 0x00,
 0x00, 0x62, 0x00, 0x00, 0x00, 0x00, 0x05, 0x11,
 0x16, 0x78, 0x9e, 0xda, 0x6d, 0x0b, 0xc8, 0xc9,
 0xca, 0x64, 0x80, 0x00, 0x11, 0x94, 0x00, 0x4e,
 0x00, 0x00, 0x01, 0x00,
 0x00, 0x00, 0x6c, 0x6d, 0x01, 0xc0, 0x01, 0xfa,
 0xce, 0x00, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79,
 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f, 0x80, 0x81,
 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89,
 0x8a, 0x8b, 0x8c, 0x8d, 0x8e, 0x8f, 0x90, 0x91,
 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99,
 0x9a, 0x9b, 0x9c, 0x9d, 0x9e, 0x9f, 0xa0, 0xa1,
 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7, 0xa8, 0xa9
};

Test command:

static  paPatchInfo_t t5PatchCmd3 =
        {
            pa_PATCH_OP_DELETE, /* ctrlBitfield */ 
            4,                  /* nPatchBytes */
            0,                  /* totalPatchSize */
            0,                  /* offset */
            NULL                /* Pointer to the patch data */
        };
     
static paCmdInfo_t t5CmdSet3[] =
{
    /* Command: Delete Bytes */
    {
        pa_CMD_PATCH_DATA,
        {
            {
                0,                            /* ctrlBitfield */
                0,                            /* dest */
                0,                            /* pktType */
                0,                            /* flowId */ 
                0,                            /* queue */
                0,                            /* swInfo0 */
                0,                            /* swInfo1 */
                0                             /* multiRouteIndex */ 
            }
        }
    },
   
    /* Command 2: Remove Tail */
    {
        pa_CMD_REMOVE_TAIL,
        {
            {
                0,                            /* ctrlBitfield */
                0,                            /* dest */
                0,                            /* pktType */
                0,                            /* flowId */ 
                0,                            /* queue */
                0,                            /* swInfo0 */
                0,                            /* swInfo1 */
                0                             /* multiRouteIndex */ 
            }
        }
    }
};

Attached please find out the updated test code as part of PA unit test:

Did you change anything else in your application?

Best regards,

Eric

2671.test5pkts.h1524.test5.c

Eric,

The difference between your test case and my implementation is that I am removing the header first, but since it is custom LUT2, it removes the Ethernet and IP headers only leaving the UDP header. Then I am using the patch to remove the 8 byte UDP header.  I would suspect your test does not adequately capture this use case scenario and so I suspect that there is still a bug in the implementation that is resulting in removing the header and the last 8 bytes of the UDP payload.

This was working for me when I was using the older mcsdk with the older PA LLD and your fix.

Are you able to try something similar to my scenario.

Eric,

Further to my last post, I tried the PATCH delete with a non-custom LUT2 so when I request removal of header, it removes the ethernet, IP and UDP header. I also request PATCH delete 8 bytes and it removes 8 bytes from the start of the UDP payload and then 8 bytes at the end of the UDP payload. I was also removing the tail in this case.

When I try the same test, but without the removal of the tail, I see the PATCH only deletes the 8 bytes at the start of the UDP payload and not the 8 bytes at the end of the playload, so it definitely has something to do with the tail removing additional data.

Aamir

Aamir:

I can not reproduce the problem you reported at the latest pdk_C6678_1_1_2_6 which also uses PA LLD 1.2.3.3. It works whether or not I add pa_CMD_REMOVE_HEADER command. I know the root cause of the original additional data deletion problem and how the fix works and I have verified the fix does exist at PA LLD 1.2.3.3.

Is it possible that you can run the updated test code and see whether the problem occurs? You just need to set a breakpoint at function t5ReceiveNatTDataPkts().

  while ((hd = t5GetRxPkt (tf)) != NULL)  {
   
   /* Verify swInfo0 for packet match and packet ID number */
   Cppi_getSoftwareInfo (Cppi_DescType_HOST, (Cppi_Desc *)hd, (UInt8 **)&swInfo);
   
   if ((*swInfo & T5_CMD_SWINFO0_TYPE_MASK) != T5_CMD_SWINFO0_PKT_ID)  {
    System_printf ("%s (%s:%d): Found a packet in the receive packet queue but with incorrect swinfo0 Id: 0x%08x\n",
          tfName, __FILE__, __LINE__, *swInfo);
     testCommonRecycleLBDesc (tf, hd);
    return (-1);
   }
   
   chan = *swInfo & T5_CMD_SWINFO0_ID_MASK;
   <BP>         count++;
   if (chan <= T5_MAX_NAT_T_CHAN)
     actualPktCount[chan] += 1;

Check the receiving packet at the packet descriptor hd when chan = 11.

Best regards,

Eric

Eric,

Pinging you for your thoughts on a post about SRIo and PA interaction. Thanks, Aamir

http://e2e.ti.com/support/dsp/c6000_multi-core_dsps/f/639/p/271072/973135.aspx#973135