Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

ipfw forward does not work

7 views
Skip to first unread message

Oleg Tarasov

unread,
Oct 2, 2006, 2:56:47 AM10/2/06
to
Hello,

I've got a machine running FreeBSD 6.0. This problem occured on 6.0-p0
and 6.0-p12.

Introduction
=============
I've got two internet connections from two different providers. One
is the main and second for failover. Both interfaces have attached
natd using divert function of ipfw. Here are interface parameters:

ng0: flags=88d1<UP,POINTOPOINT,RUNNING,NOARP,SIMPLEX,MULTICAST> mtu 1492
inet xxx.xxx.xxx.xxx --> XXX.XXX.XXX.XXX netmask 0xffffffff
ng8: flags=88d1<UP,POINTOPOINT,RUNNING,NOARP,SIMPLEX,MULTICAST> mtu 1492
inet yyy.yyy.yyy.yyy --> YYY.YYY.YYY.YYY netmask 0xffffffff

Here yyy.yyy.yyy.yyy is an IP address of main connection.

routing table looks like this:
-------------------------
default YYY.YYY.YYY.YYY UGS 0 21878 ng8
yyy.yyy.yyy.yyy lo0 UHS 0 51 lo0
xxx.xxx.xxx.xxx lo0 UHS 0 0 lo0
127.0.0.1 127.0.0.1 UH 0 3810 lo0
192.168.82 link#1 UC 0 0 rl0
192.168.82.253 00:30:4f:27:ae:85 UHLW 1 74 lo0
YYY.YYY.YYY.YYY yyy.yyy.yyy.yyy UH 3 0 ng8
XXX.XXX.XXX.XXX xxx.xxx.xxx.xxx UH 3 0 ng0
-------------------------

My kernel is compiled using following options:
-------------------------
options IPFIREWALL
options IPFIREWALL_VERBOSE
options IPFIREWALL_VERBOSE_LIMIT=300
options IPFIREWALL_DEFAULT_TO_ACCEPT
options IPFIREWALL_FORWARD
options IPDIVERT
options IPSTEALTH
options DUMMYNET
options HZ=1000
-------------------------

Both interfaces have real IPs and should simultaneously work supplying
DNS, mail and other services.

Usually this is implemented configuring ipfw fwd command for policy
routing so I've inserted two following lines into ipfw script:
-------------------------
fwd XXX.XXX.XXX.XXX ip from xxx.xxx.xxx.xxx to any out xmit ng8
fwd YYY.YYY.YYY.YYY ip from yyy.yyy.yyy.yyy to any out xmit ng0
-------------------------

This usually works and works on my second server. But for some reason
here I met strange behaviour. It just seems that fwd command does not
do anything at all.

When I ping xxx.xxx.xxx.xxx (which is failover one) icmp packets come
into ng0 but replies from xxx.xxx.xxx.xxx go through default route on
ng8. This should be normal if there were no fwd commands. But I see
counters on the rule increasing and logging these rules shows
following lines:
Oct 2 08:35:49 central kernel: ipfw: 20500 Forward to XXX.XXX.XXX.XXX
ICMP:0.0 xxx.xxx.xxx.xxx some.outer.ip.address out via ng8

but packets still go out through ng8 using default route.
There can be two reasons as I see. First is that fwd command does not
work for some reason and the second is that system routing table
considered that default route is preferrable over direct route to
router. The second near impossible so I wonder...

Please tell me if possible how to locate the possible reason of this
problem!

--
Best regards,
Oleg Tarasov mailto:subsc...@osk.com.ua

_______________________________________________
freebs...@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-ipfw
To unsubscribe, send any mail to "freebsd-ipfw...@freebsd.org"

Oleg Tarasov

unread,
Oct 2, 2006, 3:56:54 AM10/2/06
to
Hello,

[resolved]

Recompiling the kernel using IPFIREWALL_FORWARD_EXTENDED solved the
problem. I thought this one in 6.0-p12 is deprecated...

0 new messages