Re: mac addresses of local interfaces do not obey setageing 0

On Tue, Feb 15, 2011 at 10:25:38AM -0800, Stephen Hemminger wrote:

On Wed, 9 Feb 2011 19:17:52 +0100
Veaceslav Falico [off-list ref] wrote:

quoted

Hello,

I have a host and a VM inside this host bridged. I've set ageing_time and
forward_delay to 0 and trying to capture all the traffic that goes through that
bridge from my VM, but it fails to capture the traffic that has dst ether
address the same as the hosts address (i.e. I can't capture the traffic to the
host).

From the code, I see that br->ageing_time doesn't really work with local mac
addresses - has_expired() function never says that a local interface mac address
is expired, because it verifies if fdb->is_static is set and returns right away.

Is this the desired behaviour? If so, is there a way to capture packets with
destination to a local interface from another interface?

I've also done a small patch and it seems to fix the situation, but I am not
sure if it's the right way to do it.

No.
Local addresses should never age.

The proper way to capture packet is to us AF_PACKET or tc actions.

If you really want to use a bridge for this kind of traffic capturing,
you can attach one side of a veth pair as third interface to the bridge,
clear off all the mac addresses (and have no higher-layer addresses on
the bridge either) and use the other side of the veth pair for your
host. Since the "far" side of the veth is not on the bridge, the bridge
will not treat its address any special.

(This is somewhat of a funny hack though. Use AF_PACKET really...)


-David