tunnels: Optimize tx path

We currently dirty a cache line to update tunnel device stats
(tx_packets/tx_bytes). We better use the txq->tx_bytes/tx_packets
counters that already are present in cpu cache, in the cache
line shared with txq->_xmit_lock

This patch extends IPTUNNEL_XMIT() macro to use txq pointer
provided by the caller.

Also &tunnel->dev->stats can be replaced by &dev->stats
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

include/net/ipip.h

@@ -42,9 +42,9 @@ struct ip_tunnel_prl_entry
 	ip_select_ident(iph, &rt->u.dst, NULL);				\
 									\
 	err = ip_local_out(skb);					\
-	if (net_xmit_eval(err) == 0) {					\
-		stats->tx_bytes += pkt_len;				\
-		stats->tx_packets++;					\
+	if (likely(net_xmit_eval(err) == 0)) {				\
+		txq->tx_bytes += pkt_len;				\
+		txq->tx_packets++;					\
 	} else {							\
 		stats->tx_errors++;					\
 		stats->tx_aborted_errors++;				\

net/ipv4/ip_gre.c

@@ -662,7 +662,8 @@ static int ipgre_rcv(struct sk_buff *skb)
 static netdev_tx_t ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 {
 	struct ip_tunnel *tunnel = netdev_priv(dev);
-	struct net_device_stats *stats = &tunnel->dev->stats;
+	struct net_device_stats *stats = &dev->stats;
+	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
 	struct iphdr  *old_iph = ip_hdr(skb);
 	struct iphdr  *tiph;
 	u8     tos;
@@ -810,7 +811,7 @@ static netdev_tx_t ipgre_tunnel_xmit(struct sk_buff *skb, struct net_device *dev
 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 		if (!new_skb) {
 			ip_rt_put(rt);
-			stats->tx_dropped++;
+			txq->tx_dropped++;
 			dev_kfree_skb(skb);
 			return NETDEV_TX_OK;
 		}

net/ipv4/ipip.c

@@ -390,7 +390,8 @@ static int ipip_rcv(struct sk_buff *skb)
 static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 {
 	struct ip_tunnel *tunnel = netdev_priv(dev);
-	struct net_device_stats *stats = &tunnel->dev->stats;
+	struct net_device_stats *stats = &dev->stats;
+	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
 	struct iphdr  *tiph = &tunnel->parms.iph;
 	u8     tos = tunnel->parms.iph.tos;
 	__be16 df = tiph->frag_off;
@@ -478,7 +479,7 @@ static netdev_tx_t ipip_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 		if (!new_skb) {
 			ip_rt_put(rt);
-			stats->tx_dropped++;
+			txq->tx_dropped++;
 			dev_kfree_skb(skb);
 			return NETDEV_TX_OK;
 		}

net/ipv6/sit.c

@@ -555,7 +555,8 @@ static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 				     struct net_device *dev)
 {
 	struct ip_tunnel *tunnel = netdev_priv(dev);
-	struct net_device_stats *stats = &tunnel->dev->stats;
+	struct net_device_stats *stats = &dev->stats;
+	struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
 	struct iphdr  *tiph = &tunnel->parms.iph;
 	struct ipv6hdr *iph6 = ipv6_hdr(skb);
 	u8     tos = tunnel->parms.iph.tos;
@@ -688,7 +689,7 @@ static netdev_tx_t ipip6_tunnel_xmit(struct sk_buff *skb,
 		struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 		if (!new_skb) {
 			ip_rt_put(rt);
-			stats->tx_dropped++;
+			txq->tx_dropped++;
 			dev_kfree_skb(skb);
 			return NETDEV_TX_OK;
 		}