[ECOS] TCP/IP Fast Retransmission fails rarely

Gary Schneider schneider@hi-techniques.com
Thu Mar 31 07:12:00 GMT 2011

Hi and thanks in advance from a forum newb. I searched the FAQs and archives
before writing.
We have a PowerPC 440 IBM-reference-design card running PPC405 ecos 2
w/Berkeley stack that has been working well for several years. It streams
digitized data (6.4MB/s, ~ 4000 packets/sec) over Gb TCP/IP and unmanaged
switch to a Windows DAQ application. Up to four cards stream to one PC on
the same Gb port. Almost all of the time, the card successfully handles
Ethernet collisions with a Fast Retransmit after 3 Dup Acks. On very rare
occasions (many minutes to many hours) this fails. Wireshark shows ecos
ignores the Dup Acks and keeps sending new data until the PC Receive Window
is filled, resulting in tens to hundreds of unanswered Dup Acks. It finally
retransmits the requested packet after almost exactly 1.0 second of
inactivity. Ecos returns a very polite error message to the application
“error sending data reply to host – Try again later”. This seems to happen
under two conditions:

1. If two different Dup Ack retransmission requests occur too quickly (under
a millisecond)
2. If a Fast Retransmission packet is lost again and a second Dup Ack series
is sent for the same packet
Is this familiar to anyone? One obvious answer is try the newer ecos version
3 but this will take enough time and open enough technical risk I’d
appreciate any insight.

Thanks and regards,
Gary Schneider
Engineering Manager
Hi-Techniques Inc.

Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

More information about the Ecos-discuss mailing list