[lwip-users] LWIP 2.0.2 / Xilinx Zynq: TCP hang data transmission after packets loss events

[Shemesh mimran, Nir [BWIIL]]

Hi LWIP,

 

I hope this is the relevant list to ask about an issue I’m working on for a while.

 

My setup as follow:

 

1. Unit A: Embedded unit: Xilinx board (Zynq 7020) running FreeRTOS +LWIP 2.0.2
2. Unit B: PC client running WIN10

 

Connectivity between the units is fiber – and I added sniffer in the middle for debugging of the issue.

 

Unit A transmit in TCP (LWIP 2.0.2) data at around 1.7[Mbps] to unit B.

 

The problem: after hour of system running the throughput is dropped to zero (except of TCP keep alive) and return back after around 6 seconds.

 

After reviewing the wireshark in the middle (connected via sniffer) I found a 3 dup acks event that cause fast retransmission reply,

 

From that moment the Application traffic stopped for 6 seconds and only TCP keep alive was active.

 

After checking the location of the packet loss I can tell for sure it was in the PC side (runnin Iperf UDP test client /server locally shows some packet loss due to the load on the PC).

 

I attached two wiresharks files of two events.

 

Now my question are:

 

1. Why there is a drop in the traffic from Unit A ?

 

2. Why Fast recovery is not working as it should after fast retransmission event ?

 

3. Is there a way to fix it by timers/windows size configuration ?

 

4. Is there a known issue about it ? – I looked on the change log since 2.0.2 and havent found something special.

 

Any help here will be much appreciated.

 

Kind regards,

 

Nir Shemesh

System Engineer
Biosense Webster (Israel) Ltd.

4 Hatnufa Street

P.O. Box 275

Yokneam 2066717 Israel

 

T: +972-(0)4-813-6114

M: +972-(054)-5-887726

nshemesh@its.jnj.com

http://www.biosensewebster.com

Visit our  LinkedIn  Page

 

 

Event_1.pcapng.gz
Description: Event_1.pcapng.gz

Event_2.pcapng.gz
Description: Event_2.pcapng.gz