Op 3/01/2012 18:42, Richard Weinberger schreef:

Am 03.01.2012 17:15, schrieb Stephen Hemminger:

quoted

On Tue,  3 Jan 2012 14:26:04 +0100
Richard Weinberger[off-list ref]  wrote:

quoted

If net.bridge.bridge-nf-call-iptables or net.bridge.bridge-nf-call-ip6tables
are set to zero xt_physdev has no effect because skb->nf_bridge has not been set up.

Signed-off-by: Richard Weinberger<richard@nod.at>

I am not sure if this is a valid configuration. The setting of sysctl is saying
"don't do iptables on bridge (since I won't be using it)" and then you are later
doing iptables and expecting the settings as if the iptables setup was being
done.

I don't think so.

Also rules like this one are broken:
iptables -A INPUT -i bridge0 -m physdev --physdev-in eth0 -j ...

No firewalling is done on the bridge, xt_physdev is only using some meta
information.

At least a big fat warning would be nice that xt_physdev does not work
if bridge-nf-call-iptables=0.
It took me some time to figure out why my firewall rule set gone nuts on
RHEL6...

The documentation is probably not explicit enough, but I would keep the 
behavior as it is now. Setting bridge-nf-call-iptables to 0 makes 
iptables behave as if bridge-netfilter was not enabled at compilation.
Anyway, your patch is almost certainly flawed since the fact that 
skb->nf_bridge can be NULL is used as part of the logic in 
br_netfilter.c: it indicates that bridge-nf-call-iptables was 0 when the 
packet was first processed by bridge-netfilter and should therefore not 
be given to iptables in any other netfilter hook.

cheers,
Bart


-- 
Bart De Schuymer
www.artinalgorithms.be