Apparently, ms file shares fall back on IPX & netbui when IP isn't available. The linux bridge code was bridging all ethernet traffic, and of course iptables only sees IP. So, I added rules to ebtables to block everything but ARP and IP, and that seems to have stopped the ms file sharing tunneling through. The extra commands on the end of my bridge script are:
/sbin/ebtables --policy DROP /sbin/ebtables --append FORWARD -p ARP -j ACCEPT /sbin/ebtables --append FORWARD -p IP -j ACCEPT
Sorry for the noise.
M
Matthew Pocock wrote:
Hi,
I'm having trouble setting fire wall rules on a bridge. I'm running rh9, and
have patched the kernel with bridge-nf-0.0.10-against-2.4.20.diff and
ebtables-v2.0.003_vs_2.4.20.diff as well as installing the utilities brctl and
ebtables.
Using iptables (editing my uni2lan & lan2uni chains), I seem to be able to
control what goes through the firewall well enough to convince ping, http,
nmap and traceroute (with and without the -I option) that protocols & ports
are reachable or not.
However, windows file sharing seems go go right through. If I pull either
cable out of the bridge, then windows file sharing across the bridge stops (no
supprise there), but if the cable is in, file sharing is always functional,
even if I set all chains to deny everything (by uncommenting the debugging
rules below).
Down at the bottom of the netfilter/iptables docs, it says I should be using
the physdev module to match bridge ports, but it seems that this module is not
present on my system. Other docs say that it is only needed for kernels 2.5.44
and greater...
iptables-restore v1.2.7a: Couldn't load match
`physdev':/lib/iptables/libipt_physdev.so: cannot open shared object file: No
such file or directory
tcpdump (run seperately on eth0 and eth1) seems to be showing exactly what you
would expect given the firewall rules & doesn't show the packets transmitting
information between the two PCs, but the two PCs still are doing file sharing.
I'm feeling bewildered.
I use this script to set up the bridge:
/usr/sbin/brctl addbr br0 /usr/sbin/brctl stp br0 off /usr/sbin/brctl addif br0 eth0 /usr/sbin/brctl addif br0 eth1 /sbin/ifconfig eth0 down /sbin/ifconfig eth1 down /sbin/ifconfig eth0 0.0.0.0 up /sbin/ifconfig eth1 0.0.0.0 up /sbin/ifconfig br0 128.240.227.17 echo "1" > /proc/sys/net/ipv4/ip_forward /sbin/route add default gw 128.240.227.251
My iptables config looks like this:
##############################################################################
#
# iptables config file written by Matthew Pocock (matthew.pocock@xxxxxxxxx)
#
## main chains *filter :INPUT ACCEPT [0:0] :FORWARD DROP [0:0] :OUTPUT ACCEPT [0:0] :lan2uni - [0:0] :uni2lan - [0:0]
############################################################################## # ## debugging rules #-A FORWARD -j DROP #-A INPUT -j DROP #-A OUTPUT -j DROP
############################################################################## # ## drop all invalid packets, iregardless -A FORWARD -m state --state INVALID -j DROP
## split traffic depending upon direction -A FORWARD -i eth0 -o eth1 -j uni2lan -A FORWARD -i eth1 -o eth0 -j lan2uni
############################################################################## # ## let everything out - is this a good plan? ## we should realy be a bit more careful here, but hey-ho -A lan2uni -j ACCEPT
############################################################################## # ## let only specific things in
## stuff we've seen before should get through -A uni2lan -m state --state RELATED,ESTABLISHED -j ACCEPT
## accept some new connections for 'nice' protocols
## we know they are new, as we've dropped all invalid things earlier, and just
## now we let related & established through.
## ping, ssh -A uni2lan -p icmp -j ACCEPT -A uni2lan -p tcp --dport 22 -j ACCEPT
## drop everything else comming in -A uni2lan -j REJECT
COMMIT
Thanks,
Matthew
