In some email I received from James R Grinter, sie wrote:
> (Yes! I really am still running a SunOS4.1.4 system.)
> 
> I've been running 3.4.19 for some time, but have made a couple of 
> recent aborted attempts to get the latest version (3.4.31) running.
> 
> The symptoms are that connections present in the state table aren't 
> matched by returning packets for those connections. As an experiment, I 
> removed all my 'flags S...' sections from TCP 'keep state' rules, and 
> observed that multiple state entries are created for an outgoing 
> connection's traffic, and returning packets are mostly permitted (the 
> closing FIN seems not to be). NAT'd connections don't function (the 
> returning packet doesn't match the NAT state and get mapped.)
> 
> I've done a bit of regression, with a simple set of test rules just to 
> avoid any unnecessary complexity/errors in my usual set, and this 
> breakage seems to have appeared with 3.4.23 - unfortunately a version 
> that came with a large number of changes.

Can you compose a regression test similar to those in the tests directory
which highlights the problem ?

Use tcpdump (-s 1500 -x) to capture packets and generate the hex input
required.

Darren

Reply via email to