Question

Our scenario is as follow: When the TCP connection is established (we can check this in tcp_set_state() function), we can find out the egress net_device of this connection via sk->sk_dst_cache->dev. With this net_device we can distinguish our net_device with other net_device on the system.

This approach failed when the egress device is a sub-interface of a bridge, because sk->sk_dst_cache->dev always points to the bridge. Is it possible to get the real net_device (struct net_bridge_port) under the bridge if we are currently at Layer 4.

Any suggestions are appreciated. Thanks and Best Regards.

Answer 1

The struct net_bridge has a list of member ports, port_list. The list contains struct net_bridge_port elements, which in turn contain the struct net_dev.

You can get the struct net_bridge from the struct net_device of the bridge trivially:

struct net_bridge *br = netdev_priv(dev);

Of course, that still doesn’t tell you which of the member interfaces the packet came in on, but not knowing anything about why you’re trying to do this I can’t say if that’s a problem or not.