[lwip-users] Unusual ARP behavior

[Roger Cover]

Greetings List,

I have encountered an unusual sequence of ARP requests from the lwIP library. I 
am using the raw API in a Power-PC system with NO_SYS=1. In the attached pcap 
file, IP 192.168.0.1 is a Windows 7-64 bit computer and IP 192.168.0.175 is my 
instrument. The post in packet 4 initiates a data set collection. The post in 
packet 15 initiates a data set transfer. The transfer does not show in the 
capture as I had UDP packets filtered out. The transfer would normally take 
about 6.5 seconds, mostly waiting for the collected data to be ready to send.

The first thing I see is that the ARP table entry timed out during the UDP 
transfer, causing the stack to issue an ARP request at packet 21. The next 3 
packets are duplicate ARP requests, as is packet 29.

The host computer software times-out after receiving no UDP packets for 1/2 
second and attempts to restart the UDP transfer at packet 31, even though the 
UDP transfer should be continuing until about time 7.0. There is no reply from 
the instrument, so a retransmission occurs at packet 32. Between packet 32 and 
packet 33 (from my other data) the instrument completes a watchdog timer reset 
(roughly time 10.145).

I am confused about a couple of things here. Why does the stack send 4 
duplicate ARP requests in 2 microseconds? Why does it send another ARP request, 
after those have been answered, 239 microseconds later? Why are no UDP packets 
passed out of the stack after these 5 ARP requests have been answered? Is there 
somewhere in the stack that it could reasonably remain for longer than the 1.3 
seconds my watchdog timer requires to reset my instrument? I look forward to 
your reply.

Regards,
Roger W. Cover
Spectral Instruments, Inc.
420 N. Bonita Ave.
Tucson, AZ 85745
Voice: 520-884-8821 ext. 144
FAX: 520-884-8803

IrregularEvent.pcap
Description: IrregularEvent.pcap