Re: macvtap performs IP defragmentation, causing MTU problems for virtual machines

[Henrik Lindström]

Had to change "return 0" to "return vif" but other than that your changes
seem to work, even with macvlan defragmentation removed.

I tested it by sending 8K fragmented multicast packets, with 5 macvlans on
the receiving side. I consistently received 6 copies of the packet (1 from the
real interface and 1 per macvlan). While doing this i had my VM running with
a macvtap, and it was receiving fragmented packets as expected.

Here are the changes i was testing with, first time sending a diff over mail
so hope it works :-)

diff --git a/drivers/net/macvlan.c b/drivers/net/macvlan.c
index 02bd201bc7e5..5f638433cef9 100644
--- a/drivers/net/macvlan.c
+++ b/drivers/net/macvlan.c
@@ -462,10 +462,6 @@ static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
 	if (is_multicast_ether_addr(eth->h_dest)) {
 		unsigned int hash;
 
-		skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN);
-		if (!skb)
-			return RX_HANDLER_CONSUMED;
-		*pskb = skb;
 		eth = eth_hdr(skb);
 		if (macvlan_forward_source(skb, port, eth->h_source)) {
 			kfree_skb(skb);
diff --git a/net/ipv4/ip_fragment.c b/net/ipv4/ip_fragment.c
index a4941f53b523..30b822dfa222 100644
--- a/net/ipv4/ip_fragment.c
+++ b/net/ipv4/ip_fragment.c
@@ -479,11 +479,29 @@ static int ip_frag_reasm(struct ipq *qp, struct sk_buff *skb,
 	return err;
 }
 
+static int ip_defrag_vif(const struct sk_buff *skb, const struct net_device *dev)
+{
+	int vif = l3mdev_master_ifindex_rcu(dev);
+
+	if (vif)
+		return vif;
+
+	/* some folks insist that receiving a fragmented mcast dgram on n devices shall
+	 * result in n defragmented packets.
+	 */
+	if (skb->pkt_type == PACKET_BROADCAST || skb->pkt_type == PACKET_MULTICAST) {
+		if (dev)
+			vif = dev->ifindex;
+	}
+
+	return vif;
+}
+
 /* Process an incoming IP datagram fragment. */
 int ip_defrag(struct net *net, struct sk_buff *skb, u32 user)
 {
 	struct net_device *dev = skb->dev ? : skb_dst(skb)->dev;
-	int vif = l3mdev_master_ifindex_rcu(dev);
+	int vif = ip_defrag_vif(skb, dev);
 	struct ipq *qp;
 
 	__IP_INC_STATS(net, IPSTATS_MIB_REASMREQDS);