[lwip-devel] TCP ACK - Bug or Acceptable Behavior

[jhinkle]

lwip running on an ESP - downloading large file from AWS server -- process is
bogged down and communication is going through many retransmits with the
server in flow control delaying messages up to 2 minutes.

lwip (ESP) acks a msg and still has several messages in its queue to
process. (WNDO is 4 messages)

The server retransmits msg based on receiving an ack but this time includes
a FIN (server giving up) - note - the original msg is in the ESP queue

The ESP acks the next in its queue (which does not have a FIN associated
with it) and then sometime later receives the msg with the FIN and ignores
it because it's already been processed and ACKed.

Sever is gone because it sent a FIN.  ESP is still waiting for data because
it ignored the FIN on a retransmitted packet that it had already processed
and ACK'd.

Is this behavior acceptable or a bug?

I can understand lwip not wanting to test for flags (FIN) on messages it has
already ACK'd.

Thanks for any comments.






--
Sent from: http://lwip.100.n7.nabble.com/lwip-devel-f11621.html