Re: [lwip-users] Patch for bug 23119

[Kieran Mansley]

On Tue, 2009-03-10 at 14:22 -0700, Ken Smith wrote:
> I'm using lwip-1.3.0 and I believe I have encountered the problem
> discussed here.
> 
> http://savannah.nongnu.org/bugs/?23119
> 
> I am using SO_RCVTIMEO (set to 30 seconds) and occasionally, my
> application gets into a loop where lwip with TCP_DEBUG enabled just
> keeps displaying:
> 
> tcp_slowtmr: processing active pcb
> tcp_slowtmr: max SYN retries reached
> tcp_pcb_purge
> tcp_pcb_purge: data left on ->unacked

Sounds like you have a different problem to the bug you mention: that
one was just to do with whether timeouts were classed as fatal or not.
I therefore think that the patch won't help.

The output above suggests that lwIP has tried to open a connection to
another host but has received no reply after a lot of attempts, and so
is giving up.  When removing the connection it noticed that there is
some data on the unacked queue, which is a little odd for a connection
that has not yet been opened, but it should cope with that, free the
unacked queue, and remove the connection.

Can you explain more about how your application uses the sockets API?
Maybe with a code sample to show what it's doing when things go wrong.
A packet capture can often be useful too.

Thanks

Kieran