On Thu, 2013-11-21 at 06:24 -0800, Eric Dumazet wrote:

On Thu, 2013-11-21 at 14:08 +0100, Johannes Berg wrote:

quoted

On Thu, 2013-11-21 at 14:05 +0100, Johannes Berg wrote:

quoted

@@ -498,7 +500,8 @@ struct sk_buff {
 	 * headers if needed
 	 */
 	__u8			encapsulation:1;
-	/* 7/9 bit hole (depending on ndisc_nodetype presence) */
+	__u8			drop_unicast:1;

The obvious question is here, and for IPv4/IPv6 - should the wireless
stack be responsible for doing this instead?

I don't really like the idea of reserving a bit for this in sk_buff,
and propagate it in every cloning ...

Someone should replace __copy_skb_header() by a single memset(),
because copying all these bits one by one is not really clever.

And then, adding a test in fast path (ip_rcv_finish()) is really not
nice.

Yeah, that was a concern too. I'll do it entirely in the wireless stack
instead I guess. At least it'll be hidden away inside the if that
already does the group key check etc.

Thanks.

johannes