Problem of repeated A-MPDU retransmissions

[Vahid]

I am using A-MPDU aggregation on the DL of a BSS. Every once in a while the AP needs to retransmit a frame (Retry=1). Things work fine as long as the MPDUs are transmitted individually but once they are aggregated with each other or with other new packets in an A-MPDU, then the retransmission pattern continues indefinitely. That is, all the previously retransmitted MPDUs and the new ones will get Retry=1.

I haven't found the source of the problem yet, but after digging around into the code I came across a few things that I didn't understand.

In MacLow::AggregateToAmpdu (), why is qosPolicy set to zero (i.e., Normal ACK) when we are doing A-MPDU aggregation? Shouldn't this be qosPolicy=3? Here's the piece of code I am referring to. I have marked that line with >>>> :
----------------------------------------------------
  if (hdr.IsQosData() || hdr.IsBlockAck()|| hdr.IsBlockAckReq())
    {
      Time tstamp;
      uint8_t tid = GetTid (packet, hdr);
      Ptr<PerStaWifiMacQueue> queue;
      AcIndex ac = QosUtilsMapTidToAc (tid);
      //since a blockack agreement always preceeds mpdu aggregation there should always exist blockAck listener
      std::map<AcIndex, MacLowBlockAckEventListener*>::const_iterator listenerIt= m_edcaListeners.find(ac);
      NS_ASSERT (listenerIt != m_edcaListeners.end ());

      queue = listenerIt->second->GetQueue()->GetObject<PerStaWifiMacQueue>();
     
      if (!hdr.GetAddr1 ().IsBroadcast () && m_mpduAggregator!= 0)
        {
          //Have to make sure that their exist a block Ack agreement before sending an AMPDU (BlockAck Manager)
          if (listenerIt->second->GetBlockAckAgreementExists (hdr.GetAddr1(), tid))
            {
              /* here is performed mpdu aggregation */
              /* MSDU aggregation happened in edca if the user asked for it so m_currentPacket may contains a normal packet or a A-MSDU*/
              Ptr<Packet> currentAggregatedPacket = Create<Packet> ();
              peekedHdr = hdr;
              uint16_t startingSequenceNumber = 0;
              uint16_t currentSequenceNumber = 0;
     >>>>     uint8_t qosPolicy = 0;
              uint16_t blockAckSize = 0;
-------------------------------------
I have checked ans this line is always executed!

Also, in BlockAckManager::GetNextPacket (), why is the else statement always taken? Shouldn't we use BLOCK_ACK?

--------------------------------------------------
           if (!agreement->second.first.IsHtSupported ()//sva why is there a ! (not) here? shouldn't it be removed?
               && (ExistsAgreementInState (recipient, tid, OriginatorBlockAckAgreement::ESTABLISHED)
                  || SwitchToBlockAckIfNeeded (recipient, tid, hdr.GetSequenceNumber ())))
             {
                hdr.SetQosAckPolicy (WifiMacHeader::BLOCK_ACK);
             }
           else
            {
               /* From section 9.10.3 in IEEE802.11e standard:
                * In order to improve efficiency, originators using the Block Ack facility
                * may send MPDU frames with the Ack Policy subfield in QoS control frames
                * set to Normal Ack if only a few MPDUs are available for transmission.[...]
                * When there are sufficient number of MPDUs, the originator may switch back to
                * the use of Block Ack.
                */
               hdr.SetQosAckPolicy (WifiMacHeader::NORMAL_ACK);
               AgreementsI i = m_agreements.find (std::make_pair (recipient, tid));
               i->second.second.erase (*it);
             }
--------------------------------------------------
I have checked and all three function calls in the if statement return true. In particular, why should we resort to NORMAL_ACK if
agreement->second.first.IsHtSupported () == true?

I would appreciate any hint.

Thanks,

[Tommaso Pecorella]

Hi,

I'd suggest to check Bugzilla for open bugs and enhancements, we have a few pending.

If you suspect a bug, I'd also suggest to ping the Wi-Fi module maintainer (but Sébastien reads the forum too).

Cheers,

T.

[Sebastien Deronne]

Please refer to the standard, normal ACK is set as policy in MPDUs belonging to an A-MPDU, so this sounds perfectly ok in the code.

A-MPDU is a HT/VHT feature, while 802.11e-style "frame aggregation" makes uses of BAR/BACK exchange, which requires the Block Ack policy.