Re: [lwip-users] strange TCP ack

[Kieran Mansley]

On Wed, 2010-04-28 at 13:17 +0800, yueyue papa wrote:
> tcp_receive: pcb->rttest 400 rtseq 8655 ackno 9680
> tcp_receive: experienced rtt 3 ticks (1500 msec).
> tcp_receive: RTO 8 (4000 milliseconds

That's the bit that describes how the RTO is updated.  lwIP has counted
three slow timer ticks between sending the packet and getting an ACK for
it.  Using the TCP algorithm this results in an RTO value of 4
seconds.  

The only thing about this that looks odd is the large time between
sending the packet and getting an ACK: it's much longer than in the
first trace you gave.  That trace was taken without debug logging turned
on of course, so that could be part of it, but in the packet capture you
also saw high RTO values, so there must be something similar there.

Can you show us your code that calls tcp_tmr()?  I wonder if it is being
called more often than your setting for TCP_TMR_INTERVAL suggests.

Kieran