Re: [lwip-users] 1.3.0 Converting issues - chained pbuf, parts sent twice

[Kieran Mansley]

On Wed, 2008-07-30 at 11:48 -0700, Alan L wrote:
> I'm running an AT91SAM7X256 port, working to update from 1.1.0.
> 
> 1.3.0 is working for me, but not yet stable.
> Below is the debug output and attached is the wireshark log of my recent
> problem.
> 
> In the debug output, tcp_write is called 4 times to serve the web page.  The
> 3rd and 4th writes (37 and 407 bytes) result in tcp_enqueue chaining the
> segments to a 446 byte length (1.1.0 sent them separately).  The problem is
> the wireshark log shows the 446 byte packet was sent, but the ~407 bytes
> that were chained are sent again, but without any tcp header - just raw data
> (see wireshark No. 16, duplicate data sent in No. 12).
> 
> Maybe related, but for another post...
> 1. After serving the web page, sys_timeout soon locks infinitely, stuck in
> the for loop with t->time = 0

Sounds like you've got a locking problem, most likely two or more
threads active in the lwIP core at one time, and a list is getting
corrupted.  Unfortunately this means it's probably a problem in your
port rather than in lwIP itself, but I hope that this helps you track it
down.

Kieran