Re: [lwip-devel] [lwip-users] TCP Previous segment lost - lwip down to crawl

[Pow, Christopher (SWCOE)]

I sent this email directly to Simon but I think I should also probably ask here.  Help, of course, would be wildly appreciated!  Thanks!

 

Hi Simon.  I have exactly the same problem as mentioned in this thread.  Applying your patch has not fixed it.  I’ve attached a PCAP file.  Please let me know if I can supply anything else helpful.  I’m running lwip-1.4.0 and so far I’ve applied patches 32417 [I was getting the problem described in that patch, and the patch seems to have fixed it] and 34517 [because of the slow-down issue].

 

Scenario:

 

I have 9 TCP connections [ports 4660-4667 and port 1030] between a lwIP embedded target and a Windows host PC.  The lwIP embedded target sends a chunk of data on each connection [hundreds of bytes 80 times a second].  The Windows host PC only has telnet opened on the TCP ports to see the streaming data from the embedded target.

 

It seems that there may be a dropped packet that causes the death-spiral to occur, around packet 3345.

 

The packet trace is the entire scenario.  I started the servers, and within seconds *all* 9 ports have slowed to a ~1Hz crawl, and I can see on my lwIP embedded target that the ISR, RX and TX counters have slowed also.

 

Christopher S. Pow ___________________ Senior Engineer
Honeywell _____________________________  SWCOE-RTOS
2600 Ridgway Pkwy. ______  o:(612)951-5852 c:(612)910-1271
Minneapolis, MN 55413 __  address@hidden

 

 

lwip-1.4.0+34517+32417.zip
Description: lwip-1.4.0+34517+32417.zip