On Mon, Nov 05, 2001 at 01:51:42AM -0800, David S. Miller wrote:
> ..causing all sorts of havoc when you try to enslave an encrypted ethernet
> tunnel device (CIPE) to a bridge. Attached patch should fix it.
>
> + skb->protocol = 0;
>
> Why "0"? That is about as good as initializing it random garbage.
> It just means now instead of going through ptype[random_value]
> in net/core/dev.c receive packet processing, it will go through
> ptype[0].
>
> Please explain how this change "fixes" somthing. If skb->protocol
> doesn't matter, you should be able to leave it alone. If it _DOES_
> matter, you should initialize it to a proper value not "0".
OK. Spanning tree packets have 0x4242 in their ethertype field,
so I assume this will do.
cheers,
Lennert
--- linux-2.4.13-ac1/net/bridge/br_stp_bpdu.c.orig Thu Nov 1 18:43:01 2001
+++ linux-2.4.13-ac1/net/bridge/br_stp_bpdu.c Thu Nov 1 18:44:38 2001
@@ -42,6 +42,7 @@
}
skb->dev = dev;
+ skb->protocol = 0x4242;
skb->mac.raw = skb_put(skb, size);
memcpy(skb->mac.raw, bridge_ula, ETH_ALEN);
memcpy(skb->mac.raw+ETH_ALEN, dev->dev_addr, ETH_ALEN);
_______________________________________________
Bridge mailing list
[EMAIL PROTECTED]
http://www.math.leidenuniv.nl/mailman/listinfo/bridge
