[ns]bug in full-tcp: sequence number after duplicated ACKs(2)

[Hyewon Jun <hyewon@cc.gatech.edu>]

To be precise, I got the previous result from FullTcp/Sack.
I also printed the reason of retransmission of packets, and the weird
packets have been retransmitted for reason 4: REASON_SACK.

After switching "FullTcp/Sack" into "FullTcp", I didn't get the weird
sequence number.

Thus, FullTcp/Sack must have some problem.
is it incomplete?

Thank you!

--hyewon



On Mon, 24 Dec 2001, Hyewon Jun wrote:

> hi,
> 
> I found a weird behavior and I would like to know what's happening.
> I have a simple scenario:
> 
> There is one tcp connection between a server and a client to use ftp
> application. In the middle of transmission, background traffic is added to
> cause congestion to the link. After receiving 3 duplicated ACKs, the TCP
> agent cut cwnd_ to half. So far so good.
> 
> but after that, TCP agent generates TCP packets having much smaller
> sequence number. Not lost packet. (MSS=536)
> 
> seqno_ in TCP packet generated by the same TCP agent: 
>    377345->15009->15545->increasing
> t_seqno_ in TCP agent side: 
>    378417 -> 378417 -> 378417 -> same
> ack_ in TCP agent side :
>    361801 - >361801 ->361801-> same
> 
> After receiving 35 duplicated ACKs, the tCP agent reduces cwnd_ into 1.
> and the sequence number of tCP packets starts from -1:
>    (25729)->(361801)-> -1->535->1071->1607->increasing
> t_seqno_ in TCP agent side:
>    (378417)->(378417)-> 362337 -> 362337 ->362337 ->362337-> ...
>    -> same until receiving ack(362873: ACK for 361801)
> 
> 
> Topology:
> 
> backtraffic -------------|
>                          |
> client------------------router--------------server    
> 
> 
> Does anyone have similar experience?
> If you have any idea what's causing this, would you please reply me?
> Is it a bug in fulltcp?
> I am using :ns-allinone-2.1b8a.
> 
> Thank you!
> and Merry Christmas!
> 					
> 			-Hyewon Jun (hyewon@cc.gatech.edu)
> 
>