[manet] FYI: Bug in ns-2 implementation of MAC802.11 still exists in 2.27

[Evgeny Osipov <Evgeny.Osipov at unibas.ch>]

May be most of you are aware of this but nevertheless...

This bug concerns the following part of code in mac-802_11.cc (taken from
version 2.27, and present in 2.26).

In void Mac802_11::recvACK(Packet *p)

        /*
         * The successful reception of this ACK packet implies
         * that our DATA transmission was successful.  Hence,
         * we can reset the Short/Long Retry Count and the CW.
         *
         * need to check the size of the packet we sent that's being
         * ACK'd, not the size of the ACK packet.
         */
        if((u_int32_t) HDR_CMN(pktTx_)->size() <= macmib_.getRTSThreshold())
                ssrc_ = 0;
        else
                slrc_ = 0;
        rst_cw();
        Packet::free(pktTx_); pktTx_ = 0;

        /*
         * Backoff before sending again.
         */


On reception of MAC level ACK the condition in "if" is always false, resulting
in not reseting the Short Retry Counter (ssrc_). With every retransmission of
an RTS ssrc_ keeps on increasing even if previous packet was transmitted
successfuly. After 7 retransmissions  (for different packets actually) an
unecessary data packet drop occurs with the reason
DROP_MAC_RETRY_COUNT_EXCEEDED.  For MANET experiments this will result in lower
packet delivery ratio in experiments with large number of nodes/connections.

This bug was reported in December 2003 (see
http://mailman.isi.edu/pipermail/ns-users/2003-December/038074.html), but still
exists in version 2.27.

One can fix this problem by modifying "if" as follows.

        if((u_int32_t) HDR_CMN(pktTx_)->size() <= macmib_.getRTSThreshold())
                //report on erroneous ACK
        else
         {
           ssrc_ = 0;
           slrc_ = 0;
         }
PS: The problem is really observable with even three nodes in a row where two
TCP connections are initiated from the central node to other two nodes. In such
a lightwieght scenario MAC should not drop data packets at all.

PPS: another useful fix: 
http://www.tkt.cs.tut.fi/research/daci/ra_protocols_ns2fix.html

--
Evgeny Osipov, Tekn.Lic.

E-mail: Evgeny.Osipov at unibas.ch
Tel. +46(70)738 42 50

----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.


_______________________________________________
manet mailing list
manet at ietf.org
https://www1.ietf.org/mailman/listinfo/manet