IPFW: shape and status in CURRENT

[O. Hartmann]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Am Thu, 29 Sep 2016 21:02:16 +0200
"O. Hartmann" <ohartman at zedat.fu-berlin.de> schrieb:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA256
> 
> 
> Since a couple of months now, I use IPFW on several projects. I use IPFW again after a
> long term hiatus since ~ 2003. Before I used pf. The reasons are mannyfold and one
> reason is very dogmatic - it is the FreeBSD's native firewall and several performance
> diagrams shown in the net tells me a significant performance benefit in case being
> setup optimal over pf. pf in FreeBSD lacks behind the OpenBSD's development.
> 
> Since last year I try to setup IPFW only on all of our systems. So I do also at home
> and at some places, where we have to use NAT via PPPoE/modem. And here the struggle
> begins. While most setups of a firewall on a router/gateway with several NICs directly
> attached to the internet with on interface, the outbound interface, the same starts to
> be a horrible story when it comes to NAT.
> 
> The handbook offers some simple examples, but in most cases, I see the supposed to be
> outdated external natd daemon still in favour over in-kernel NAT! This is also the
> case with the manpage for ipfw(8). I miss a more recent example of setting up NAT with
> in-kernel NAT and the caveats of one-pass and none-one-pass and some hints how the IP
> packet's header gets rewritten when being translated by NAT and reinjected into the
> pipeline. For me, as a non-source-code-expert-and-simple-system's administrator, it is
> sometimes hard to understand how IPFW works. And the problems reported do tell me that
> I'm not alone.
> 
> The handbook has some examples. One of them contains a traversal of 37/TCP, timeserver.
> It is a long time since I saw this kind of setup, most time synchronisation methods use
> NTP and 123/UDP. The example also seems a bit outdated.
> 
> Manpage firewall(7) lacks also of an modern in-kernel NAT example - it still referes to
> the natd. Also, there is a kind of anti-spoof rule shown that leaves the impression that
> this page is quite antique. Doesn't IPFW has a antispoof rule, or even "verrepath" as
> the manpage ipfw(8) states?
> 
> Somehow I miss some more detailed explanations what happens with check-state, since this
> causes much trouble, even in combination with NAT.
> 
> Well, as said, I'm no expert, maybe I'm simply too blunt to understand, but again, it
> seems I'm not alone. People switched to pf and even Apple moved from ipfw to pf. That
> leaves the question here: what is the status of the development of IPFW in FreeBSD? is
> it maintained-only or is there development going on? Are there plans for refurbished,
> more up to time man pages and examples?
> 
> Thanks in advance and for your patience reading my bad English.
> 
> Oliver

[...]

Looking at firewall(7) and trying to simply fowllow the example, one will discover that
rules

     add 01500 deny all from not 10.0.1.0/24 in via fxp1
     add 01500 deny all from not 10.0.2.0/24 in via fxp2
     add 01501 deny all from 10.0.1.0/24 in via fxp0
     add 01501 deny all from 10.0.2.0/24 in via fxp0

will produce a "missing to" error in recent IPFW (it is the case in my installation of
CURRENT, hope it does not differ from yours)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2

iQEcBAEBCAAGBQJX7pyyAAoJEOgBcD7A/5N8I44IAK10SPfEeSH+JGGOBeExMLUt
HSP6lZHtzvprl6t2HaJzqOhl/Xu2jAKbBBjcdLcCZ5kP6qGsKvR1fySuz/0l42dO
rtnQMqwuuDEIsvA8LScTkPkDK7VtCUpD8xJEQ4jnCfhtk4qoGAZAV9+JlNdr3l1d
LNee2rAi9jhO5JcI3JRLLQIIOla+YMQKcMOwzLHGYfxuTFbp2qyQNSQpIFc3HilH
tX3owVeJw9yNmTrC3VYOt1NwdknrolEHOG3KeWdpEtNu7UA+31dkcom+qy1JTpuN
/KvkOGykVaPoUEO2CYjO2cFBB7bm3rk8UXoEr4doCVex05dl97ElP0W168dC1fk=
=BSxQ
-----END PGP SIGNATURE-----