I'm having a strange problem with IP Masq on a 2.0.36 kernel.  I have
these rules:

/sbin/ipfwadm -F -p deny
/sbin/ipfwadm -F -a accept -m -S 192.168.100.50/32 -D 0.0.0.0/0 -W eth1
/sbin/ipfwadm -F -a accept -m -S 192.168.100.55/32 -D 0.0.0.0/0 -W eth1
/sbin/ipfwadm -F -a accept -m -S 192.168.100.89/32 -D 0.0.0.0/0 -W eth1
/sbin/ipfwadm -F -a accept -m -S 192.168.100.230/32 -D 0.0.0.0/0 -W eth1
/sbin/ipfwadm -M -s 7200 10 60
/sbin/ipfwadm -F -a accept -m -S 192.168.100.0/24 -D 207.xxx.xxx.xxx/32

The local network is on eth0, and eth1 connects to the Internet.  I
also have Squid running (port 3128).  Everything works fine, until one
of the four machines above tries to use Squid.  Those packets
shouldn't be masqueraded, since they aren't going out over eth1, but
they show up in 'ipfwadm -M -l':

tcp  00:41.61 sys89.here.com  localhost          1906 (61105) -> 3128

As soon as that happens, all outgoing tcp on the machine locks up; I
can't even telnet to localhost!  Ping still works fine.  It looks like
the kernel suddenly decides to masquerade all tcp traffic, whether it's
going out eth1 or not.

Any ideas what could be causing this, or rules that could fix it.
I've never had this problem on a 2.0.30 system I'm using, so if all
else fails I'll downgrade the kernel to that.  The only other
difference is that the problem machine has two Ethernet cards, while
the working system uses ppp for the gateway connection.


Thanks,
Aaron
-- 
Aaron Baugher - [EMAIL PROTECTED] - Quincy, IL, USA
Extreme Systems Consulting - http://haruchai.rnet.com/esc/
CGI, Perl, Java, and Linux/Unix Administration


_______________________________________________
Masq maillist  -  [EMAIL PROTECTED]
http://tiffany.indyramp.com/mailman/listinfo/masq
Admin requests can be handled by web (above) or [EMAIL PROTECTED]

Reply via email to