FIN_WAIT_2

[Charles Swiger]

On Mar 21, 2005, at 1:52 PM, Robert Gogolok wrote:
> I've looked at the "Closing a Connection" chapter from the RFC and 
> tried to understand it. The state diagram above shows that from the 
> FINWAIT-2 state there is only one possible way to reach TIME WAIT.

That's right.

> So FreeBSD must be using another extension of the RFC-793, when it's 
> sending ACK messages in the FINWAIT-2 state?

If the TCP connection is in FIN_WAIT_2, FreeBSD may send out ACKs 
periodicly, trying to nudge the other side to send a FIN to finish 
closing the connection.

That may be controlled by the keepalive sysctl, but we're starting to 
go beyond my specific knowledge.  One of the true FreeBSD network 
wizards like Andre Oppermann might be able to provide more information.

> Wow, I'm confused at this point, I have a linux box here which was the 
> previous webserver, and I can't remember seeing ACK's hitting the 
> firewall logs as it is now with the FreeBSD webserver.

Hmm, it's hard to say.  Having a complete tcpdump of a TCP connection 
handy would help, as would making sure that your firewall rules aren't 
doing something to interrupt or block the end of the connection.  Are 
you getting any responses back from the clients, or have they 
disappeared?

-- 
-Chuck