Re: [RFC]: not possible to do policy routing for next hop on tunnels.

[Eric Dumazet <eric.dumazet@gmail.com>]

Le mercredi 01 septembre 2010 à 14:21 +0200, Anders Franzen a écrit :
> If I have multiple routing choices to reach a tunnel end point,
> I would like to mark different flows with iptables, after the
> postrouting of the primary route look up, this would put an fwmark on
> the skb. The route would in my case resolve to an ip6_tunnel.
> 
> In the tunnel a secondary route lookup is done to select next hop for
> the tunnel end point.
> 
> I would like to apply policy routing to the secondary lookup. 
> 
> This will not work, for two reasons:
> 
> 1. None of the tunnels (ipip, gre, ip6_tunnel) I looked at regards the
> fwmark at the skb, when performing the route lookup.
> 
> 2. ip6_tunnel is keeping a local dst cache, so it will never reroute as
> long as the current cache is valid.
> 
> 
> I wonder if there is a reason for not giving the fwmark at the tunnel
> route for tunnels in general?
> 
> And the local dst cache for ip6_tunnel, is it needed, is not the routing
> subsystem keeping some form of internal cache (ip route ls cache).

ip6_tunnel (or others) could be extended with a
IP6_TNL_F_USE_ORIG_FWMARK, and in this case not use/cache the route.

untested patch to get the idea :

diff --git a/include/linux/ip6_tunnel.h b/include/linux/ip6_tunnel.h
index acb9ad6..bf22b03 100644
--- a/include/linux/ip6_tunnel.h
+++ b/include/linux/ip6_tunnel.h
@@ -16,6 +16,8 @@
 #define IP6_TNL_F_MIP6_DEV 0x8
 /* copy DSCP from the outer packet */
 #define IP6_TNL_F_RCV_DSCP_COPY 0x10
+/* copy fwmark from inner packet */
+#define IP6_TNL_F_USE_ORIG_FWMARK 0x20
 
 struct ip6_tnl_parm {
 	char name[IFNAMSIZ];	/* name of tunnel device */
diff --git a/net/ipv6/ip6_tunnel.c b/net/ipv6/ip6_tunnel.c
index 0fd027f..e31a91f 100644
--- a/net/ipv6/ip6_tunnel.c
+++ b/net/ipv6/ip6_tunnel.c
@@ -858,7 +858,7 @@ static int ip6_tnl_xmit2(struct sk_buff *skb,
 	int err = -1;
 	int pkt_len;
 
-	if ((dst = ip6_tnl_dst_check(t)) != NULL)
+	if (!fl->mark && (dst = ip6_tnl_dst_check(t)) != NULL)
 		dst_hold(dst);
 	else {
 		dst = ip6_route_output(net, NULL, fl);
@@ -910,7 +910,7 @@ static int ip6_tnl_xmit2(struct sk_buff *skb,
 		skb = new_skb;
 	}
 	skb_dst_drop(skb);
-	skb_dst_set(skb, dst_clone(dst));
+	skb_dst_set(skb, fl->mark ? dst : dst_clone(dst));
 
 	skb->transport_header = skb->network_header;
 
@@ -940,7 +940,9 @@ static int ip6_tnl_xmit2(struct sk_buff *skb,
 		stats->tx_errors++;
 		stats->tx_aborted_errors++;
 	}
-	ip6_tnl_dst_store(t, dst);
+	if (!fl->mark)
+		ip6_tnl_dst_store(t, dst);
+		
 	return 0;
 tx_err_link_failure:
 	stats->tx_carrier_errors++;
@@ -976,6 +978,8 @@ ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
 		fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
 					  & IPV6_TCLASS_MASK;
+	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK))
+		fl.mark = skb->mark;
 
 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
 	if (err != 0) {
@@ -1026,7 +1030,8 @@ ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
 	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL))
 		fl.fl6_flowlabel |= (*(__be32 *) ipv6h & IPV6_FLOWLABEL_MASK);
-
+	if ((t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK))
+		fl.mark = skb->mark;
 	err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
 	if (err != 0) {
 		if (err == -EMSGSIZE)