Re: [B.A.T.M.A.N.] [RFC 5/6] batman-adv: Generalize snooping mechanism in order to suport NDP too

Hi Mihail,

On Mon, Jul 08, 2013 at 03:12:44AM +0300, mihail.costea2005@gmail.com wrote:

From: Mihail Costea mihail.costea90@gmail.com /**

• • batadv_dat_snoop_outgoing_pkt_reply - snoop the ARP reply and fill the DHT

• • batadv_dat_snoop_outgoing_pkt_reply - snoop the ARP reply / NA and fill
• • the DHT
  • @bat_priv: the bat priv with all the soft interface information
  • @skb: packet to check
  */

void batadv_dat_snoop_outgoing_pkt_reply(struct batadv_priv *bat_priv, struct sk_buff *skb) {

--- snip ---

• /* not a solicited advertisement (see snooping mechanism) */
• if (ip_dst) {

• batadv_dat_entry_add(bat_priv, ip_dst, dat_pair_type.data_type,
  

• 		     hw_dst, vid);
  

• batadv_dat_send_data(bat_priv, skb, ip_dst,
  

• 		     dat_pair_type.data_type,
  

• 		     BATADV_P_DAT_DHT_PUT);
  

• }

}

I'm currently wondering about the unsolicited neighbor advertisement case. So far, your concept/patchset does not seem to apply any optimizations for those packets, does it? They are still flooded unconditionally, right?

What do you think, would it make sense to prevent unsolicited neighbor advertisements from being forwarded into the mesh network? Or would that create certain issues?

Cheers, Linus