ipvlan: remove excessive packet scrubbing

IPvlan currently scrubs packets at every location where packets may be
crossing namespace boundary. Though this is desirable, currently IPvlan
does it more than necessary. e.g. packets that are going to take
dev_forward_skb() path will get scrubbed so no point in scrubbing them
before forwarding. Another side-effect of scrubbing is that pkt-type gets
set to PACKET_HOST which overrides what was already been set by the
earlier path making erroneous delivery of the packets.

Also scrubbing packets just before calling dev_queue_xmit() has detrimental
effects since packets lose skb->sk and because of that miss prio updates,
incorrect socket back-pressure and would even break TSQ.

Fixes: b93dd49 ('ipvlan: Scrub skb before crossing the namespace boundary')
Signed-off-by: Mahesh Bandewar <maheshb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ipvlan/ipvlan_core.c

@@ -315,13 +315,13 @@ static int ipvlan_rcv_frame(struct ipvl_addr *addr, struct sk_buff **pskb,
 
 		*pskb = skb;
 	}
-	ipvlan_skb_crossing_ns(skb, dev);
 
 	if (local) {
 		skb->pkt_type = PACKET_HOST;
 		if (dev_forward_skb(ipvlan->dev, skb) == NET_RX_SUCCESS)
 			success = true;
 	} else {
+		skb->dev = dev;
 		ret = RX_HANDLER_ANOTHER;
 		success = true;
 	}
@@ -586,7 +586,7 @@ static int ipvlan_xmit_mode_l2(struct sk_buff *skb, struct net_device *dev)
 		return NET_XMIT_SUCCESS;
 	}
 
-	ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
+	skb->dev = ipvlan->phy_dev;
 	return dev_queue_xmit(skb);
 }