X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=datapath%2Ftunnel.c;h=899d1cdcb8b4fa037fbc92bffc8148aeb993aede;hb=d7436f763874179e086ff1f86bb2fe72596973a0;hp=3f25c9b4ad3fcc28efa4dd5dd02e49a3b19d0741;hpb=d1eb60ccff0c2dbef2300cdfc5fead3c2d394beb;p=sliver-openvswitch.git

diff --git a/datapath/tunnel.c b/datapath/tunnel.c
index 3f25c9b4a..899d1cdcb 100644
--- a/datapath/tunnel.c
+++ b/datapath/tunnel.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2010 Nicira Networks.
+ * Copyright (c) 2010, 2011 Nicira Networks.
  * Distributed under the terms of the GNU GPL version 2.
  *
  * Significant portions of this file may be copied from parts of the Linux
@@ -15,6 +15,7 @@
 #include <linux/jhash.h>
 #include <linux/kernel.h>
 #include <linux/version.h>
+#include <linux/workqueue.h>
 
 #include <net/dsfield.h>
 #include <net/dst.h>
@@ -28,24 +29,60 @@
 #include <net/xfrm.h>
 
 #include "actions.h"
+#include "checksum.h"
 #include "datapath.h"
 #include "table.h"
 #include "tunnel.h"
+#include "vlan.h"
 #include "vport.h"
 #include "vport-generic.h"
+#include "vport-internal_dev.h"
 
-/* Protected by RCU. */
-static struct tbl *port_table;
+#ifdef NEED_CACHE_TIMEOUT
+/*
+ * On kernels where we can't quickly detect changes in the rest of the system
+ * we use an expiration time to invalidate the cache.  A shorter expiration
+ * reduces the length of time that we may potentially blackhole packets while
+ * a longer time increases performance by reducing the frequency that the
+ * cache needs to be rebuilt.  A variety of factors may cause the cache to be
+ * invalidated before the expiration time but this is the maximum.  The time
+ * is expressed in jiffies.
+ */
+#define MAX_CACHE_EXP HZ
+#endif
+
+/*
+ * Interval to check for and remove caches that are no longer valid.  Caches
+ * are checked for validity before they are used for packet encapsulation and
+ * old caches are removed at that time.  However, if no packets are sent through
+ * the tunnel then the cache will never be destroyed.  Since it holds
+ * references to a number of system objects, the cache will continue to use
+ * system resources by not allowing those objects to be destroyed.  The cache
+ * cleaner is periodically run to free invalid caches.  It does not
+ * significantly affect system performance.  A lower interval will release
+ * resources faster but will itself consume resources by requiring more frequent
+ * checks.  A longer interval may result in messages being printed to the kernel
+ * message buffer about unreleased resources.  The interval is expressed in
+ * jiffies.
+ */
+#define CACHE_CLEANER_INTERVAL (5 * HZ)
+
+#define CACHE_DATA_ALIGN 16
+
+static struct tbl __rcu *port_table __read_mostly;
+
+static void cache_cleaner(struct work_struct *work);
+static DECLARE_DELAYED_WORK(cache_cleaner_wq, cache_cleaner);
 
 /*
  * These are just used as an optimization: they don't require any kind of
  * synchronization because we could have just as easily read the value before
  * the port change happened.
  */
-static unsigned int key_local_remote_ports;
-static unsigned int key_remote_ports;
-static unsigned int local_remote_ports;
-static unsigned int remote_ports;
+static unsigned int key_local_remote_ports __read_mostly;
+static unsigned int key_remote_ports __read_mostly;
+static unsigned int local_remote_ports __read_mostly;
+static unsigned int remote_ports __read_mostly;
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36)
 #define rt_dst(rt) (rt->dst)
@@ -63,49 +100,86 @@ static inline struct tnl_vport *tnl_vport_table_cast(const struct tbl_node *node
 	return container_of(node, struct tnl_vport, tbl_node);
 }
 
-/* RCU callback. */
-static void free_config(struct rcu_head *rcu)
+/* This is analogous to rtnl_dereference for the tunnel cache.  It checks that
+ * cache_lock is held, so it is only for update side code.
+ */
+static inline struct tnl_cache *cache_dereference(struct tnl_vport *tnl_vport)
+{
+	return rcu_dereference_protected(tnl_vport->cache,
+					 lockdep_is_held(&tnl_vport->cache_lock));
+}
+
+static inline void schedule_cache_cleaner(void)
+{
+	schedule_delayed_work(&cache_cleaner_wq, CACHE_CLEANER_INTERVAL);
+}
+
+static void free_cache(struct tnl_cache *cache)
+{
+	if (!cache)
+		return;
+
+	flow_put(cache->flow);
+	ip_rt_put(cache->rt);
+	kfree(cache);
+}
+
+static void free_config_rcu(struct rcu_head *rcu)
 {
 	struct tnl_mutable_config *c = container_of(rcu, struct tnl_mutable_config, rcu);
 	kfree(c);
 }
 
+static void free_cache_rcu(struct rcu_head *rcu)
+{
+	struct tnl_cache *c = container_of(rcu, struct tnl_cache, rcu);
+	free_cache(c);
+}
+
 static void assign_config_rcu(struct vport *vport,
 			      struct tnl_mutable_config *new_config)
 {
 	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
 	struct tnl_mutable_config *old_config;
 
-	old_config = rcu_dereference(tnl_vport->mutable);
+	old_config = rtnl_dereference(tnl_vport->mutable);
 	rcu_assign_pointer(tnl_vport->mutable, new_config);
-	call_rcu(&old_config->rcu, free_config);
+	call_rcu(&old_config->rcu, free_config_rcu);
+}
+
+static void assign_cache_rcu(struct vport *vport, struct tnl_cache *new_cache)
+{
+	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
+	struct tnl_cache *old_cache;
+
+	old_cache = cache_dereference(tnl_vport);
+	rcu_assign_pointer(tnl_vport->cache, new_cache);
+
+	if (old_cache)
+		call_rcu(&old_cache->rcu, free_cache_rcu);
 }
 
 static unsigned int *find_port_pool(const struct tnl_mutable_config *mutable)
 {
-	if (mutable->port_config.flags & TNL_F_IN_KEY_MATCH) {
-		if (mutable->port_config.saddr)
+	if (mutable->flags & TNL_F_IN_KEY_MATCH) {
+		if (mutable->saddr)
 			return &local_remote_ports;
 		else
 			return &remote_ports;
 	} else {
-		if (mutable->port_config.saddr)
+		if (mutable->saddr)
 			return &key_local_remote_ports;
 		else
 			return &key_remote_ports;
 	}
 }
 
-enum lookup_key {
-	LOOKUP_TUNNEL_TYPE	= 0,
-	LOOKUP_SADDR		= 1,
-	LOOKUP_DADDR		= 2,
-	LOOKUP_KEY		= 3,
-};
-
 struct port_lookup_key {
-	u32 vals[4];			/* Contains enum lookup_key keys. */
 	const struct tnl_mutable_config *mutable;
+	__be64 key;
+	u32 tunnel_type;
+	__be32 saddr;
+	__be32 daddr;
 };
 
 /*
@@ -117,56 +191,114 @@ static int port_cmp(const struct tbl_node *node, void *target)
 	const struct tnl_vport *tnl_vport = tnl_vport_table_cast(node);
 	struct port_lookup_key *lookup = target;
 
-	lookup->mutable = rcu_dereference(tnl_vport->mutable);
+	lookup->mutable = rcu_dereference_rtnl(tnl_vport->mutable);
 
-	return (lookup->mutable->tunnel_type == lookup->vals[LOOKUP_TUNNEL_TYPE]) &&
-	       lookup->mutable->port_config.daddr == lookup->vals[LOOKUP_DADDR] &&
-	       lookup->mutable->port_config.in_key == lookup->vals[LOOKUP_KEY] &&
-	       lookup->mutable->port_config.saddr == lookup->vals[LOOKUP_SADDR];
+	return (lookup->mutable->tunnel_type == lookup->tunnel_type &&
+		lookup->mutable->daddr == lookup->daddr &&
+		lookup->mutable->in_key == lookup->key &&
+		lookup->mutable->saddr == lookup->saddr);
+}
+
+static u32 port_hash(struct port_lookup_key *k)
+{
+	u32 x = jhash_3words((__force u32)k->saddr, (__force u32)k->daddr,
+			     k->tunnel_type, 0);
+	return jhash_2words((__force u64)k->key >> 32, (__force u32)k->key, x);
 }
 
-static u32 port_hash(struct port_lookup_key *lookup)
+static u32 mutable_hash(const struct tnl_mutable_config *mutable)
 {
-	return jhash2(lookup->vals, ARRAY_SIZE(lookup->vals), 0);
+	struct port_lookup_key lookup;
+
+	lookup.saddr = mutable->saddr;
+	lookup.daddr = mutable->daddr;
+	lookup.key = mutable->in_key;
+	lookup.tunnel_type = mutable->tunnel_type;
+
+	return port_hash(&lookup);
+}
+
+static void check_table_empty(void)
+{
+	struct tbl *old_table = rtnl_dereference(port_table);
+
+	if (tbl_count(old_table) == 0) {
+		cancel_delayed_work_sync(&cache_cleaner_wq);
+		rcu_assign_pointer(port_table, NULL);
+		tbl_deferred_destroy(old_table, NULL);
+	}
 }
 
 static int add_port(struct vport *vport)
 {
+	struct tbl *cur_table = rtnl_dereference(port_table);
 	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
-	struct port_lookup_key lookup;
 	int err;
 
 	if (!port_table) {
 		struct tbl *new_table;
 
-		new_table = tbl_create(0);
+		new_table = tbl_create(TBL_MIN_BUCKETS);
 		if (!new_table)
 			return -ENOMEM;
 
 		rcu_assign_pointer(port_table, new_table);
+		schedule_cache_cleaner();
 
-	} else if (tbl_count(port_table) > tbl_n_buckets(port_table)) {
-		struct tbl *old_table = port_table;
+	} else if (tbl_count(cur_table) > tbl_n_buckets(cur_table)) {
 		struct tbl *new_table;
 
-		new_table = tbl_expand(old_table);
+		new_table = tbl_expand(cur_table);
 		if (IS_ERR(new_table))
 			return PTR_ERR(new_table);
 
 		rcu_assign_pointer(port_table, new_table);
-		tbl_deferred_destroy(old_table, NULL);
+		tbl_deferred_destroy(cur_table, NULL);
+	}
+
+	err = tbl_insert(rtnl_dereference(port_table), &tnl_vport->tbl_node,
+			 mutable_hash(rtnl_dereference(tnl_vport->mutable)));
+	if (err) {
+		check_table_empty();
+		return err;
 	}
 
-	lookup.vals[LOOKUP_SADDR] = tnl_vport->mutable->port_config.saddr;
-	lookup.vals[LOOKUP_DADDR] = tnl_vport->mutable->port_config.daddr;
-	lookup.vals[LOOKUP_KEY] = tnl_vport->mutable->port_config.in_key;
-	lookup.vals[LOOKUP_TUNNEL_TYPE] = tnl_vport->mutable->tunnel_type;
+	(*find_port_pool(rtnl_dereference(tnl_vport->mutable)))++;
 
-	err = tbl_insert(port_table, &tnl_vport->tbl_node, port_hash(&lookup));
+	return 0;
+}
+
+static int move_port(struct vport *vport, struct tnl_mutable_config *new_mutable)
+{
+	int err;
+	struct tbl *cur_table = rtnl_dereference(port_table);
+	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
+	u32 hash;
+
+	hash = mutable_hash(new_mutable);
+	if (hash == tnl_vport->tbl_node.hash)
+		goto table_updated;
+
+	/*
+	 * Ideally we should make this move atomic to avoid having gaps in
+	 * finding tunnels or the possibility of failure.  However, if we do
+	 * find a tunnel it will always be consistent.
+	 */
+	err = tbl_remove(cur_table, &tnl_vport->tbl_node);
 	if (err)
 		return err;
 
-	(*find_port_pool(tnl_vport->mutable))++;
+	err = tbl_insert(cur_table, &tnl_vport->tbl_node, hash);
+	if (err) {
+		(*find_port_pool(rtnl_dereference(tnl_vport->mutable)))--;
+		check_table_empty();
+		return err;
+	}
+
+table_updated:
+	(*find_port_pool(rtnl_dereference(tnl_vport->mutable)))--;
+	assign_config_rcu(vport, new_mutable);
+	(*find_port_pool(rtnl_dereference(tnl_vport->mutable)))++;
 
 	return 0;
 }
@@ -176,32 +308,33 @@ static int del_port(struct vport *vport)
 	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
 	int err;
 
-	err = tbl_remove(port_table, &tnl_vport->tbl_node);
+	err = tbl_remove(rtnl_dereference(port_table), &tnl_vport->tbl_node);
 	if (err)
 		return err;
 
-	(*find_port_pool(tnl_vport->mutable))--;
+	check_table_empty();
+	(*find_port_pool(rtnl_dereference(tnl_vport->mutable)))--;
 
 	return 0;
 }
 
-struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be32 key,
+struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be64 key,
 			    int tunnel_type,
 			    const struct tnl_mutable_config **mutable)
 {
 	struct port_lookup_key lookup;
-	struct tbl *table = rcu_dereference(port_table);
+	struct tbl *table = rcu_dereference_rtnl(port_table);
 	struct tbl_node *tbl_node;
 
-	if (!table)
+	if (unlikely(!table))
 		return NULL;
 
-	lookup.vals[LOOKUP_SADDR] = saddr;
-	lookup.vals[LOOKUP_DADDR] = daddr;
+	lookup.saddr = saddr;
+	lookup.daddr = daddr;
 
 	if (tunnel_type & TNL_T_KEY_EXACT) {
-		lookup.vals[LOOKUP_KEY] = key;
-		lookup.vals[LOOKUP_TUNNEL_TYPE] = tunnel_type & ~TNL_T_KEY_MATCH;
+		lookup.key = key;
+		lookup.tunnel_type = tunnel_type & ~TNL_T_KEY_MATCH;
 
 		if (key_local_remote_ports) {
 			tbl_node = tbl_lookup(table, &lookup, port_hash(&lookup), port_cmp);
@@ -210,19 +343,19 @@ struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be32 key,
 		}
 
 		if (key_remote_ports) {
-			lookup.vals[LOOKUP_SADDR] = 0;
+			lookup.saddr = 0;
 
 			tbl_node = tbl_lookup(table, &lookup, port_hash(&lookup), port_cmp);
 			if (tbl_node)
 				goto found;
 
-			lookup.vals[LOOKUP_SADDR] = saddr;
+			lookup.saddr = saddr;
 		}
 	}
 
 	if (tunnel_type & TNL_T_KEY_MATCH) {
-		lookup.vals[LOOKUP_KEY] = 0;
-		lookup.vals[LOOKUP_TUNNEL_TYPE] = tunnel_type & ~TNL_T_KEY_EXACT;
+		lookup.key = 0;
+		lookup.tunnel_type = tunnel_type & ~TNL_T_KEY_EXACT;
 
 		if (local_remote_ports) {
 			tbl_node = tbl_lookup(table, &lookup, port_hash(&lookup), port_cmp);
@@ -231,7 +364,7 @@ struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be32 key,
 		}
 
 		if (remote_ports) {
-			lookup.vals[LOOKUP_SADDR] = 0;
+			lookup.saddr = 0;
 
 			tbl_node = tbl_lookup(table, &lookup, port_hash(&lookup), port_cmp);
 			if (tbl_node)
@@ -246,6 +379,79 @@ found:
 	return tnl_vport_to_vport(tnl_vport_table_cast(tbl_node));
 }
 
+static void ecn_decapsulate(struct sk_buff *skb, u8 tos)
+{
+	if (unlikely(INET_ECN_is_ce(tos))) {
+		__be16 protocol = skb->protocol;
+
+		skb_set_network_header(skb, ETH_HLEN);
+
+		if (protocol == htons(ETH_P_8021Q)) {
+			if (unlikely(!pskb_may_pull(skb, VLAN_ETH_HLEN)))
+				return;
+
+			protocol = vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
+			skb_set_network_header(skb, VLAN_ETH_HLEN);
+		}
+
+		if (protocol == htons(ETH_P_IP)) {
+			if (unlikely(!pskb_may_pull(skb, skb_network_offset(skb)
+			    + sizeof(struct iphdr))))
+				return;
+
+			IP_ECN_set_ce(ip_hdr(skb));
+		}
+#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
+		else if (protocol == htons(ETH_P_IPV6)) {
+			if (unlikely(!pskb_may_pull(skb, skb_network_offset(skb)
+			    + sizeof(struct ipv6hdr))))
+				return;
+
+			IP6_ECN_set_ce(ipv6_hdr(skb));
+		}
+#endif
+	}
+}
+
+/**
+ *	tnl_rcv - ingress point for generic tunnel code
+ *
+ * @vport: port this packet was received on
+ * @skb: received packet
+ * @tos: ToS from encapsulating IP packet, used to copy ECN bits
+ *
+ * Must be called with rcu_read_lock.
+ *
+ * Packets received by this function are in the following state:
+ * - skb->data points to the inner Ethernet header.
+ * - The inner Ethernet header is in the linear data area.
+ * - skb->csum does not include the inner Ethernet header.
+ * - The layer pointers are undefined.
+ */
+void tnl_rcv(struct vport *vport, struct sk_buff *skb, u8 tos)
+{
+	struct ethhdr *eh;
+
+	skb_reset_mac_header(skb);
+	eh = eth_hdr(skb);
+
+	if (likely(ntohs(eh->h_proto) >= 1536))
+		skb->protocol = eh->h_proto;
+	else
+		skb->protocol = htons(ETH_P_802_2);
+
+	skb_dst_drop(skb);
+	nf_reset(skb);
+	skb_clear_rxhash(skb);
+	secpath_reset(skb);
+
+	ecn_decapsulate(skb, tos);
+	compute_ip_summed(skb, false);
+	vlan_set_tci(skb, 0);
+
+	vport_receive(vport, skb);
+}
+
 static bool check_ipv4_address(__be32 addr)
 {
 	if (ipv4_is_multicast(addr) || ipv4_is_lbcast(addr)
@@ -413,7 +619,7 @@ static void ipv6_build_icmp(struct sk_buff *skb, struct sk_buff *nskb,
 #endif /* IPv6 */
 
 bool tnl_frag_needed(struct vport *vport, const struct tnl_mutable_config *mutable,
-		     struct sk_buff *skb, unsigned int mtu, __be32 flow_key)
+		     struct sk_buff *skb, unsigned int mtu, __be64 flow_key)
 {
 	unsigned int eth_hdr_len = ETH_HLEN;
 	unsigned int total_length = 0, header_length = 0, payload_length;
@@ -466,7 +672,6 @@ bool tnl_frag_needed(struct vport *vport, const struct tnl_mutable_config *mutab
 	}
 #endif
 
-	total_length = min(total_length, mutable->mtu);
 	payload_length = total_length - header_length;
 
 	nskb = dev_alloc_skb(NET_IP_ALIGN + eth_hdr_len + header_length +
@@ -486,7 +691,8 @@ bool tnl_frag_needed(struct vport *vport, const struct tnl_mutable_config *mutab
 
 		vh->h_vlan_TCI = vlan_eth_hdr(skb)->h_vlan_TCI;
 		vh->h_vlan_encapsulated_proto = skb->protocol;
-	}
+	} else
+		vlan_set_tci(nskb, vlan_get_tci(skb));
 	skb_reset_mac_header(nskb);
 
 	/* Protocol */
@@ -504,7 +710,7 @@ bool tnl_frag_needed(struct vport *vport, const struct tnl_mutable_config *mutab
 	 * not symmetric then PMTUD needs to be disabled since we won't have
 	 * any way of synthesizing packets.
 	 */
-	if ((mutable->port_config.flags & (TNL_F_IN_KEY_MATCH | TNL_F_OUT_KEY_ACTION)) ==
+	if ((mutable->flags & (TNL_F_IN_KEY_MATCH | TNL_F_OUT_KEY_ACTION)) ==
 	    (TNL_F_IN_KEY_MATCH | TNL_F_OUT_KEY_ACTION))
 		OVS_CB(nskb)->tun_id = flow_key;
 
@@ -514,195 +720,435 @@ bool tnl_frag_needed(struct vport *vport, const struct tnl_mutable_config *mutab
 	return true;
 }
 
-static struct sk_buff *check_headroom(struct sk_buff *skb, int headroom)
+static bool check_mtu(struct sk_buff *skb,
+		      struct vport *vport,
+		      const struct tnl_mutable_config *mutable,
+		      const struct rtable *rt, __be16 *frag_offp)
 {
-	if (skb_headroom(skb) < headroom || skb_header_cloned(skb)) {
-		struct sk_buff *nskb = skb_realloc_headroom(skb, headroom + 16);
-		if (unlikely(!nskb)) {
-			kfree_skb(skb);
-			return ERR_PTR(-ENOMEM);
-		}
+	bool pmtud = mutable->flags & TNL_F_PMTUD;
+	__be16 frag_off = 0;
+	int mtu = 0;
+	unsigned int packet_length = skb->len - ETH_HLEN;
 
-		set_skb_csum_bits(skb, nskb);
+	/* Allow for one level of tagging in the packet length. */
+	if (!vlan_tx_tag_present(skb) &&
+	    eth_hdr(skb)->h_proto == htons(ETH_P_8021Q))
+		packet_length -= VLAN_HLEN;
 
-		if (skb->sk)
-			skb_set_owner_w(nskb, skb->sk);
+	if (pmtud) {
+		int vlan_header = 0;
 
-		dev_kfree_skb(skb);
-		return nskb;
+		frag_off = htons(IP_DF);
+
+		/* The tag needs to go in packet regardless of where it
+		 * currently is, so subtract it from the MTU.
+		 */
+		if (vlan_tx_tag_present(skb) ||
+		    eth_hdr(skb)->h_proto == htons(ETH_P_8021Q))
+			vlan_header = VLAN_HLEN;
+
+		mtu = dst_mtu(&rt_dst(rt))
+			- ETH_HLEN
+			- mutable->tunnel_hlen
+			- vlan_header;
 	}
 
-	return skb;
-}
+	if (skb->protocol == htons(ETH_P_IP)) {
+		struct iphdr *iph = ip_hdr(skb);
 
-static inline u8 ecn_encapsulate(u8 tos, struct sk_buff *skb)
-{
-	u8 inner;
+		frag_off |= iph->frag_off & htons(IP_DF);
 
-	if (skb->protocol == htons(ETH_P_IP))
-		inner = ((struct iphdr *)skb_network_header(skb))->tos;
+		if (pmtud && iph->frag_off & htons(IP_DF)) {
+			mtu = max(mtu, IP_MIN_MTU);
+
+			if (packet_length > mtu &&
+			    tnl_frag_needed(vport, mutable, skb, mtu,
+					    OVS_CB(skb)->tun_id))
+				return false;
+		}
+	}
 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
-	else if (skb->protocol == htons(ETH_P_IPV6))
-		inner = ipv6_get_dsfield((struct ipv6hdr *)skb_network_header(skb));
+	else if (skb->protocol == htons(ETH_P_IPV6)) {
+		/* IPv6 requires PMTUD if the packet is above the minimum MTU. */
+		if (packet_length > IPV6_MIN_MTU)
+			frag_off = htons(IP_DF);
+
+		if (pmtud) {
+			mtu = max(mtu, IPV6_MIN_MTU);
+
+			if (packet_length > mtu &&
+			    tnl_frag_needed(vport, mutable, skb, mtu,
+					    OVS_CB(skb)->tun_id))
+				return false;
+		}
+	}
 #endif
-	else
-		inner = 0;
 
-	return INET_ECN_encapsulate(tos, inner);
+	*frag_offp = frag_off;
+	return true;
 }
 
-static inline void ecn_decapsulate(struct sk_buff *skb)
+static void create_tunnel_header(const struct vport *vport,
+				 const struct tnl_mutable_config *mutable,
+				 const struct rtable *rt, void *header)
 {
-	u8 tos = ip_hdr(skb)->tos;
+	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
+	struct iphdr *iph = header;
+
+	iph->version	= 4;
+	iph->ihl	= sizeof(struct iphdr) >> 2;
+	iph->frag_off	= htons(IP_DF);
+	iph->protocol	= tnl_vport->tnl_ops->ipproto;
+	iph->tos	= mutable->tos;
+	iph->daddr	= rt->rt_dst;
+	iph->saddr	= rt->rt_src;
+	iph->ttl	= mutable->ttl;
+	if (!iph->ttl)
+		iph->ttl = ip4_dst_hoplimit(&rt_dst(rt));
+
+	tnl_vport->tnl_ops->build_header(vport, mutable, iph + 1);
+}
 
-	if (INET_ECN_is_ce(tos)) {
-		__be16 protocol = skb->protocol;
-		unsigned int nw_header = skb_network_header(skb) - skb->data;
+static inline void *get_cached_header(const struct tnl_cache *cache)
+{
+	return (void *)cache + ALIGN(sizeof(struct tnl_cache), CACHE_DATA_ALIGN);
+}
 
-		if (skb->protocol == htons(ETH_P_8021Q)) {
-			if (unlikely(!pskb_may_pull(skb, VLAN_ETH_HLEN)))
-				return;
+static inline bool check_cache_valid(const struct tnl_cache *cache,
+				     const struct tnl_mutable_config *mutable)
+{
+	return cache &&
+#ifdef NEED_CACHE_TIMEOUT
+		time_before(jiffies, cache->expiration) &&
+#endif
+#ifdef HAVE_RT_GENID
+		atomic_read(&init_net.ipv4.rt_genid) == cache->rt->rt_genid &&
+#endif
+#ifdef HAVE_HH_SEQ
+		rt_dst(cache->rt).hh->hh_lock.sequence == cache->hh_seq &&
+#endif
+		mutable->seq == cache->mutable_seq &&
+		(!is_internal_dev(rt_dst(cache->rt).dev) ||
+		(cache->flow && !cache->flow->dead));
+}
 
-			protocol = vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
-			nw_header += VLAN_HLEN;
-		}
+static int cache_cleaner_cb(struct tbl_node *tbl_node, void *aux)
+{
+	struct tnl_vport *tnl_vport = tnl_vport_table_cast(tbl_node);
+	const struct tnl_mutable_config *mutable = rcu_dereference(tnl_vport->mutable);
+	const struct tnl_cache *cache = rcu_dereference(tnl_vport->cache);
 
-		if (protocol == htons(ETH_P_IP)) {
-			if (unlikely(!pskb_may_pull(skb, nw_header
-			    + sizeof(struct iphdr))))
-				return;
+	if (cache && !check_cache_valid(cache, mutable) &&
+	    spin_trylock_bh(&tnl_vport->cache_lock)) {
+		assign_cache_rcu(tnl_vport_to_vport(tnl_vport), NULL);
+		spin_unlock_bh(&tnl_vport->cache_lock);
+	}
 
-			IP_ECN_set_ce((struct iphdr *)(nw_header + skb->data));
-		}
-#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
-		else if (protocol == htons(ETH_P_IPV6)) {
-			if (unlikely(!pskb_may_pull(skb, nw_header
-			    + sizeof(struct ipv6hdr))))
-				return;
+	return 0;
+}
 
-			IP6_ECN_set_ce((struct ipv6hdr *)(nw_header
-							  + skb->data));
-		}
+static void cache_cleaner(struct work_struct *work)
+{
+	schedule_cache_cleaner();
+
+	rcu_read_lock();
+	tbl_foreach(rcu_dereference(port_table), cache_cleaner_cb, NULL);
+	rcu_read_unlock();
+}
+
+static inline void create_eth_hdr(struct tnl_cache *cache,
+				  const struct rtable *rt)
+{
+	void *cache_data = get_cached_header(cache);
+	int hh_len = rt_dst(rt).hh->hh_len;
+	int hh_off = HH_DATA_ALIGN(rt_dst(rt).hh->hh_len) - hh_len;
+
+#ifdef HAVE_HH_SEQ
+	unsigned hh_seq;
+
+	do {
+		hh_seq = read_seqbegin(&rt_dst(rt).hh->hh_lock);
+		memcpy(cache_data, (void *)rt_dst(rt).hh->hh_data + hh_off, hh_len);
+	} while (read_seqretry(&rt_dst(rt).hh->hh_lock, hh_seq));
+
+	cache->hh_seq = hh_seq;
+#else
+	read_lock_bh(&rt_dst(rt).hh->hh_lock);
+	memcpy(cache_data, (void *)rt_dst(rt).hh->hh_data + hh_off, hh_len);
+	read_unlock_bh(&rt_dst(rt).hh->hh_lock);
 #endif
-	}
 }
 
-static struct sk_buff *handle_gso(struct sk_buff *skb)
+static struct tnl_cache *build_cache(struct vport *vport,
+				     const struct tnl_mutable_config *mutable,
+				     struct rtable *rt)
 {
-	if (skb_is_gso(skb)) {
-		struct sk_buff *nskb = skb_gso_segment(skb, 0);
+	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
+	struct tnl_cache *cache;
+	void *cache_data;
+	int cache_len;
 
-		dev_kfree_skb(skb);
-		return nskb;
+	if (!(mutable->flags & TNL_F_HDR_CACHE))
+		return NULL;
+
+	/*
+	 * If there is no entry in the ARP cache or if this device does not
+	 * support hard header caching just fall back to the IP stack.
+	 */
+	if (!rt_dst(rt).hh)
+		return NULL;
+
+	/*
+	 * If lock is contended fall back to directly building the header.
+	 * We're not going to help performance by sitting here spinning.
+	 */
+	if (!spin_trylock_bh(&tnl_vport->cache_lock))
+		return NULL;
+
+	cache = cache_dereference(tnl_vport);
+	if (check_cache_valid(cache, mutable))
+		goto unlock;
+	else
+		cache = NULL;
+
+	cache_len = rt_dst(rt).hh->hh_len + mutable->tunnel_hlen;
+
+	cache = kzalloc(ALIGN(sizeof(struct tnl_cache), CACHE_DATA_ALIGN) +
+			cache_len, GFP_ATOMIC);
+	if (!cache)
+		goto unlock;
+
+	cache->len = cache_len;
+
+	create_eth_hdr(cache, rt);
+	cache_data = get_cached_header(cache) + rt_dst(rt).hh->hh_len;
+
+	create_tunnel_header(vport, mutable, rt, cache_data);
+
+	cache->mutable_seq = mutable->seq;
+	cache->rt = rt;
+#ifdef NEED_CACHE_TIMEOUT
+	cache->expiration = jiffies + tnl_vport->cache_exp_interval;
+#endif
+
+	if (is_internal_dev(rt_dst(rt).dev)) {
+		struct sw_flow_key flow_key;
+		struct tbl_node *flow_node;
+		struct vport *dst_vport;
+		struct sk_buff *skb;
+		bool is_frag;
+		int err;
+
+		dst_vport = internal_dev_get_vport(rt_dst(rt).dev);
+		if (!dst_vport)
+			goto done;
+
+		skb = alloc_skb(cache->len, GFP_ATOMIC);
+		if (!skb)
+			goto done;
+
+		__skb_put(skb, cache->len);
+		memcpy(skb->data, get_cached_header(cache), cache->len);
+
+		err = flow_extract(skb, dst_vport->port_no, &flow_key, &is_frag);
+
+		kfree_skb(skb);
+		if (err || is_frag)
+			goto done;
+
+		flow_node = tbl_lookup(rcu_dereference(dst_vport->dp->table),
+				       &flow_key, flow_hash(&flow_key),
+				       flow_cmp);
+		if (flow_node) {
+			struct sw_flow *flow = flow_cast(flow_node);
+
+			cache->flow = flow;
+			flow_hold(flow);
+		}
 	}
 
-	return skb;
+done:
+	assign_cache_rcu(vport, cache);
+
+unlock:
+	spin_unlock_bh(&tnl_vport->cache_lock);
+
+	return cache;
 }
 
-static int handle_csum_offload(struct sk_buff *skb)
+static struct rtable *find_route(struct vport *vport,
+				 const struct tnl_mutable_config *mutable,
+				 u8 tos, struct tnl_cache **cache)
 {
-	if (skb->ip_summed == CHECKSUM_PARTIAL)
-		return skb_checksum_help(skb);
-	else {
-		skb->ip_summed = CHECKSUM_NONE;
-		return 0;
+	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
+	struct tnl_cache *cur_cache = rcu_dereference(tnl_vport->cache);
+
+	*cache = NULL;
+	tos = RT_TOS(tos);
+
+	if (likely(tos == mutable->tos && check_cache_valid(cur_cache, mutable))) {
+		*cache = cur_cache;
+		return cur_cache->rt;
+	} else {
+		struct rtable *rt;
+#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,39)
+		struct flowi fl = { .nl_u = { .ip4_u =
+					      { .daddr = mutable->daddr,
+						.saddr = mutable->saddr,
+						.tos = tos } },
+				    .proto = tnl_vport->tnl_ops->ipproto };
+
+		if (unlikely(ip_route_output_key(&init_net, &rt, &fl)))
+			return NULL;
+#else
+		struct flowi4 fl = { .daddr = mutable->daddr,
+				     .saddr = mutable->saddr,
+				     .flowi4_tos = tos,
+				     .flowi4_proto = tnl_vport->tnl_ops->ipproto };
+
+		rt = ip_route_output_key(&init_net, &fl);
+		if (IS_ERR(rt))
+			return NULL;
+#endif
+
+		if (likely(tos == mutable->tos))
+			*cache = build_cache(vport, mutable, rt);
+
+		return rt;
 	}
 }
 
-/* Called with rcu_read_lock. */
-void tnl_rcv(struct vport *vport, struct sk_buff *skb)
+static struct sk_buff *check_headroom(struct sk_buff *skb, int headroom)
 {
-	skb->pkt_type = PACKET_HOST;
-	skb->protocol = eth_type_trans(skb, skb->dev);
+	if (skb_headroom(skb) < headroom || skb_header_cloned(skb)) {
+		struct sk_buff *nskb = skb_realloc_headroom(skb, headroom + 16);
+		if (unlikely(!nskb)) {
+			kfree_skb(skb);
+			return ERR_PTR(-ENOMEM);
+		}
 
-	skb_dst_drop(skb);
-	nf_reset(skb);
-	secpath_reset(skb);
-	skb_reset_network_header(skb);
+		set_skb_csum_bits(skb, nskb);
+
+		if (skb->sk)
+			skb_set_owner_w(nskb, skb->sk);
 
-	ecn_decapsulate(skb);
+		kfree_skb(skb);
+		return nskb;
+	}
 
-	skb_push(skb, ETH_HLEN);
-	compute_ip_summed(skb, false);
+	return skb;
+}
 
-	vport_receive(vport, skb);
+static inline bool need_linearize(const struct sk_buff *skb)
+{
+	int i;
+
+	if (unlikely(skb_shinfo(skb)->frag_list))
+		return true;
+
+	/*
+	 * Generally speaking we should linearize if there are paged frags.
+	 * However, if all of the refcounts are 1 we know nobody else can
+	 * change them from underneath us and we can skip the linearization.
+	 */
+	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++)
+		if (unlikely(page_count(skb_shinfo(skb)->frags[i].page) > 1))
+			return true;
+
+	return false;
 }
 
-static int build_packet(struct vport *vport, const struct tnl_mutable_config *mutable,
-			struct iphdr *iph, struct rtable *rt, int max_headroom,
-			int mtu, struct sk_buff *skb)
+static struct sk_buff *handle_offloads(struct sk_buff *skb,
+				       const struct tnl_mutable_config *mutable,
+				       const struct rtable *rt)
 {
-	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
+	int min_headroom;
 	int err;
-	struct iphdr *new_iph;
-	int orig_len = skb->len;
-	__be16 frag_off = iph->frag_off;
 
-	skb = check_headroom(skb, max_headroom);
-	if (unlikely(IS_ERR(skb)))
-		goto error;
+	forward_ip_summed(skb);
 
-	err = handle_csum_offload(skb);
+	err = vswitch_skb_checksum_setup(skb);
 	if (unlikely(err))
 		goto error_free;
 
-	if (skb->protocol == htons(ETH_P_IP)) {
-		struct iphdr *old_iph = ip_hdr(skb);
-
-		if ((old_iph->frag_off & htons(IP_DF)) &&
-		    mtu < ntohs(old_iph->tot_len)) {
-			if (tnl_frag_needed(vport, mutable, skb, mtu, OVS_CB(skb)->tun_id))
-				goto error_free;
-		}
+	min_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len
+			+ mutable->tunnel_hlen
+			+ (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0);
 
+	skb = check_headroom(skb, min_headroom);
+	if (IS_ERR(skb)) {
+		err = PTR_ERR(skb);
+		goto error;
 	}
-#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
-	else if (skb->protocol == htons(ETH_P_IPV6)) {
-		unsigned int packet_length = skb->len - ETH_HLEN
-			- (eth_hdr(skb)->h_proto == htons(ETH_P_8021Q) ? VLAN_HLEN : 0);
 
-		/* IPv6 requires PMTUD if the packet is above the minimum MTU. */
-		if (packet_length > IPV6_MIN_MTU)
-			frag_off = htons(IP_DF);
+	if (skb_is_gso(skb)) {
+		struct sk_buff *nskb;
 
-		if (mtu < packet_length) {
-			if (tnl_frag_needed(vport, mutable, skb, mtu, OVS_CB(skb)->tun_id))
+		nskb = skb_gso_segment(skb, 0);
+		kfree_skb(skb);
+		if (IS_ERR(nskb)) {
+			err = PTR_ERR(nskb);
+			goto error;
+		}
+
+		skb = nskb;
+	} else if (skb->ip_summed == CHECKSUM_PARTIAL) {
+		/* Pages aren't locked and could change at any time.
+		 * If this happens after we compute the checksum, the
+		 * checksum will be wrong.  We linearize now to avoid
+		 * this problem.
+		 */
+		if (unlikely(need_linearize(skb))) {
+			err = __skb_linearize(skb);
+			if (unlikely(err))
 				goto error_free;
 		}
-	}
-#endif
 
-	new_iph = (struct iphdr *)skb_push(skb, mutable->tunnel_hlen);
-	skb_reset_network_header(skb);
-	skb_set_transport_header(skb, sizeof(struct iphdr));
+		err = skb_checksum_help(skb);
+		if (unlikely(err))
+			goto error_free;
+	} else if (skb->ip_summed == CHECKSUM_COMPLETE)
+		skb->ip_summed = CHECKSUM_NONE;
 
-	memcpy(new_iph, iph, sizeof(struct iphdr));
-	new_iph->frag_off = frag_off;
-	ip_select_ident(new_iph, &rt_dst(rt), NULL);
+	return skb;
 
-	tnl_vport->tnl_ops->build_header(skb, vport, mutable);
+error_free:
+	kfree_skb(skb);
+error:
+	return ERR_PTR(err);
+}
 
-	/* Allow our local IP stack to fragment the outer packet even if the
-	 * DF bit is set as a last resort. */
-	skb->local_df = 1;
+static int send_frags(struct sk_buff *skb,
+		      const struct tnl_mutable_config *mutable)
+{
+	int sent_len;
 
-	memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
-	IPCB(skb)->flags = 0;
+	sent_len = 0;
+	while (skb) {
+		struct sk_buff *next = skb->next;
+		int frag_len = skb->len - mutable->tunnel_hlen;
+		int err;
 
-	err = ip_local_out(skb);
-	if (likely(net_xmit_eval(err) == 0))
-		return orig_len;
-	else {
-		vport_record_error(vport, VPORT_E_TX_ERROR);
-		return 0;
+		skb->next = NULL;
+		memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
+
+		err = ip_local_out(skb);
+		skb = next;
+		if (unlikely(net_xmit_eval(err)))
+			goto free_frags;
+		sent_len += frag_len;
 	}
 
-error_free:
-	kfree_skb(skb);
-error:
-	vport_record_error(vport, VPORT_E_TX_DROPPED);
+	return sent_len;
 
-	return 0;
+free_frags:
+	/*
+	 * There's no point in continuing to send fragments once one has been
+	 * dropped so just free the rest.  This may help improve the congestion
+	 * that caused the first packet to be dropped.
+	 */
+	tnl_free_linked_skbs(skb);
+	return sent_len;
 }
 
 int tnl_send(struct vport *vport, struct sk_buff *skb)
@@ -710,15 +1156,19 @@ int tnl_send(struct vport *vport, struct sk_buff *skb)
 	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
 	const struct tnl_mutable_config *mutable = rcu_dereference(tnl_vport->mutable);
 
-	struct iphdr *old_iph;
-	int orig_len;
-	struct iphdr iph;
+	enum vport_err_type err = VPORT_E_TX_ERROR;
 	struct rtable *rt;
-	int max_headroom;
-	int mtu;
+	struct dst_entry *unattached_dst = NULL;
+	struct tnl_cache *cache;
+	int sent_len = 0;
+	__be16 frag_off = 0;
+	u8 ttl;
+	u8 inner_tos;
+	u8 tos;
 
 	/* Validate the protocol headers before we try to use them. */
-	if (skb->protocol == htons(ETH_P_8021Q)) {
+	if (skb->protocol == htons(ETH_P_8021Q) &&
+	    !vlan_tx_tag_present(skb)) {
 		if (unlikely(!pskb_may_pull(skb, VLAN_ETH_HLEN)))
 			goto error_free;
 
@@ -727,196 +1177,252 @@ int tnl_send(struct vport *vport, struct sk_buff *skb)
 	}
 
 	if (skb->protocol == htons(ETH_P_IP)) {
-		if (unlikely(!pskb_may_pull(skb, skb_network_header(skb)
-		    + sizeof(struct iphdr) - skb->data)))
+		if (unlikely(!pskb_may_pull(skb, skb_network_offset(skb)
+		    + sizeof(struct iphdr))))
 			skb->protocol = 0;
 	}
 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 	else if (skb->protocol == htons(ETH_P_IPV6)) {
-		if (unlikely(!pskb_may_pull(skb, skb_network_header(skb)
-		    + sizeof(struct ipv6hdr) - skb->data)))
+		if (unlikely(!pskb_may_pull(skb, skb_network_offset(skb)
+		    + sizeof(struct ipv6hdr))))
 			skb->protocol = 0;
 	}
 #endif
-	old_iph = ip_hdr(skb);
 
-	iph.tos = mutable->port_config.tos;
-	if (mutable->port_config.flags & TNL_F_TOS_INHERIT) {
-		if (skb->protocol == htons(ETH_P_IP))
-			iph.tos = old_iph->tos;
-#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
-		else if (skb->protocol == htons(ETH_P_IPV6))
-			iph.tos = ipv6_get_dsfield(ipv6_hdr(skb));
-#endif
-	}
-	iph.tos = ecn_encapsulate(iph.tos, skb);
-
-	{
-		struct flowi fl = { .nl_u = { .ip4_u =
-					      { .daddr = mutable->port_config.daddr,
-						.saddr = mutable->port_config.saddr,
-						.tos = RT_TOS(iph.tos) } },
-				    .proto = tnl_vport->tnl_ops->ipproto };
-
-		if (unlikely(ip_route_output_key(&init_net, &rt, &fl)))
-			goto error_free;
-	}
-
-	iph.ttl = mutable->port_config.ttl;
-	if (mutable->port_config.flags & TNL_F_TTL_INHERIT) {
-		if (skb->protocol == htons(ETH_P_IP))
-			iph.ttl = old_iph->ttl;
+	/* ToS */
+	if (skb->protocol == htons(ETH_P_IP))
+		inner_tos = ip_hdr(skb)->tos;
 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
-		else if (skb->protocol == htons(ETH_P_IPV6))
-			iph.ttl = ipv6_hdr(skb)->hop_limit;
+	else if (skb->protocol == htons(ETH_P_IPV6))
+		inner_tos = ipv6_get_dsfield(ipv6_hdr(skb));
 #endif
-	}
-	if (!iph.ttl)
-		iph.ttl = dst_metric(&rt_dst(rt), RTAX_HOPLIMIT);
+	else
+		inner_tos = 0;
 
-	iph.frag_off = (mutable->port_config.flags & TNL_F_PMTUD) ? htons(IP_DF) : 0;
-	if (iph.frag_off)
-		mtu = dst_mtu(&rt_dst(rt))
-			- ETH_HLEN
-			- mutable->tunnel_hlen
-			- (eth_hdr(skb)->h_proto == htons(ETH_P_8021Q) ? VLAN_HLEN : 0);
+	if (mutable->flags & TNL_F_TOS_INHERIT)
+		tos = inner_tos;
 	else
-		mtu = mutable->mtu;
+		tos = mutable->tos;
 
-	if (skb->protocol == htons(ETH_P_IP)) {
-		iph.frag_off |= old_iph->frag_off & htons(IP_DF);
-		mtu = max(mtu, IP_MIN_MTU);
-	}
-#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
-	else if (skb->protocol == htons(ETH_P_IPV6))
-		mtu = max(mtu, IPV6_MIN_MTU);
-#endif
+	tos = INET_ECN_encapsulate(tos, inner_tos);
 
-	iph.version = 4;
-	iph.ihl = sizeof(struct iphdr) >> 2;
-	iph.protocol = tnl_vport->tnl_ops->ipproto;
-	iph.daddr = rt->rt_dst;
-	iph.saddr = rt->rt_src;
+	/* Route lookup */
+	rt = find_route(vport, mutable, tos, &cache);
+	if (unlikely(!rt))
+		goto error_free;
+	if (unlikely(!cache))
+		unattached_dst = &rt_dst(rt);
 
+	/* Reset SKB */
 	nf_reset(skb);
 	secpath_reset(skb);
 	skb_dst_drop(skb);
-	skb_dst_set(skb, &rt_dst(rt));
+	skb_clear_rxhash(skb);
 
-	/*
-	 * If we are doing GSO on a pskb it is better to make sure that the
-	 * headroom is correct now.  We will only have to copy the portion in
-	 * the linear data area and GSO will preserve headroom when it creates
-	 * the segments.  This is particularly beneficial on Xen where we get
-	 * lots of GSO pskbs.  Conversely, we delay copying if it is just to
-	 * get our own writable clone because GSO may do the copy for us.
-	 */
-	max_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len
-			+ mutable->tunnel_hlen;
+	/* Offloading */
+	skb = handle_offloads(skb, mutable, rt);
+	if (IS_ERR(skb))
+		goto error;
 
-	if (skb_headroom(skb) < max_headroom) {
-		skb = check_headroom(skb, max_headroom);
-		if (unlikely(IS_ERR(skb))) {
-			vport_record_error(vport, VPORT_E_TX_DROPPED);
-			goto error;
-		}
+	/* MTU */
+	if (unlikely(!check_mtu(skb, vport, mutable, rt, &frag_off))) {
+		err = VPORT_E_TX_DROPPED;
+		goto error_free;
 	}
 
-	forward_ip_summed(skb);
+	/*
+	 * If we are over the MTU, allow the IP stack to handle fragmentation.
+	 * Fragmentation is a slow path anyways.
+	 */
+	if (unlikely(skb->len + mutable->tunnel_hlen > dst_mtu(&rt_dst(rt)) &&
+		     cache)) {
+		unattached_dst = &rt_dst(rt);
+		dst_hold(unattached_dst);
+		cache = NULL;
+	}
 
-	if (unlikely(vswitch_skb_checksum_setup(skb)))
-		goto error_free;
+	/* TTL */
+	ttl = mutable->ttl;
+	if (!ttl)
+		ttl = ip4_dst_hoplimit(&rt_dst(rt));
 
-	skb = handle_gso(skb);
-	if (unlikely(IS_ERR(skb))) {
-		vport_record_error(vport, VPORT_E_TX_DROPPED);
-		goto error;
+	if (mutable->flags & TNL_F_TTL_INHERIT) {
+		if (skb->protocol == htons(ETH_P_IP))
+			ttl = ip_hdr(skb)->ttl;
+#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
+		else if (skb->protocol == htons(ETH_P_IPV6))
+			ttl = ipv6_hdr(skb)->hop_limit;
+#endif
 	}
 
-	/*
-	 * Process GSO segments.  Try to do any work for the entire packet that
-	 * doesn't involve actually writing to it before this point.
-	 */
-	orig_len = 0;
-	do {
+	while (skb) {
+		struct iphdr *iph;
 		struct sk_buff *next_skb = skb->next;
 		skb->next = NULL;
 
-		orig_len += build_packet(vport, mutable, &iph, rt, max_headroom, mtu, skb);
-
+		if (unlikely(vlan_deaccel_tag(skb)))
+			goto next;
+
+		if (likely(cache)) {
+			skb_push(skb, cache->len);
+			memcpy(skb->data, get_cached_header(cache), cache->len);
+			skb_reset_mac_header(skb);
+			skb_set_network_header(skb, rt_dst(rt).hh->hh_len);
+
+		} else {
+			skb_push(skb, mutable->tunnel_hlen);
+			create_tunnel_header(vport, mutable, rt, skb->data);
+			skb_reset_network_header(skb);
+
+			if (next_skb)
+				skb_dst_set(skb, dst_clone(unattached_dst));
+			else {
+				skb_dst_set(skb, unattached_dst);
+				unattached_dst = NULL;
+			}
+		}
+		skb_set_transport_header(skb, skb_network_offset(skb) + sizeof(struct iphdr));
+
+		iph = ip_hdr(skb);
+		iph->tos = tos;
+		iph->ttl = ttl;
+		iph->frag_off = frag_off;
+		ip_select_ident(iph, &rt_dst(rt), NULL);
+
+		skb = tnl_vport->tnl_ops->update_header(vport, mutable, &rt_dst(rt), skb);
+		if (unlikely(!skb))
+			goto next;
+
+		if (likely(cache)) {
+			int orig_len = skb->len - cache->len;
+			struct vport *cache_vport = internal_dev_get_vport(rt_dst(rt).dev);
+
+			skb->protocol = htons(ETH_P_IP);
+			iph = ip_hdr(skb);
+			iph->tot_len = htons(skb->len - skb_network_offset(skb));
+			ip_send_check(iph);
+
+			if (cache_vport) {
+				OVS_CB(skb)->flow = cache->flow;
+				compute_ip_summed(skb, true);
+				vport_receive(cache_vport, skb);
+				sent_len += orig_len;
+			} else {
+				int xmit_err;
+
+				skb->dev = rt_dst(rt).dev;
+				xmit_err = dev_queue_xmit(skb);
+
+				if (likely(net_xmit_eval(xmit_err) == 0))
+					sent_len += orig_len;
+			}
+		} else
+			sent_len += send_frags(skb, mutable);
+
+next:
 		skb = next_skb;
-	} while (skb);
+	}
 
-	return orig_len;
+	if (unlikely(sent_len == 0))
+		vport_record_error(vport, VPORT_E_TX_DROPPED);
+
+	goto out;
 
 error_free:
-	kfree_skb(skb);
-	vport_record_error(vport, VPORT_E_TX_ERROR);
+	tnl_free_linked_skbs(skb);
 error:
-	return 0;
+	vport_record_error(vport, err);
+out:
+	dst_release(unattached_dst);
+	return sent_len;
 }
 
-int tnl_init(void)
-{
-	return 0;
-}
-
-void tnl_exit(void)
-{
-	tbl_destroy(port_table, NULL);
-	port_table = NULL;
-}
+static const struct nla_policy tnl_policy[ODP_TUNNEL_ATTR_MAX + 1] = {
+	[ODP_TUNNEL_ATTR_FLAGS]    = { .type = NLA_U32 },
+	[ODP_TUNNEL_ATTR_DST_IPV4] = { .type = NLA_U32 },
+	[ODP_TUNNEL_ATTR_SRC_IPV4] = { .type = NLA_U32 },
+	[ODP_TUNNEL_ATTR_OUT_KEY]  = { .type = NLA_U64 },
+	[ODP_TUNNEL_ATTR_IN_KEY]   = { .type = NLA_U64 },
+	[ODP_TUNNEL_ATTR_TOS]      = { .type = NLA_U8 },
+	[ODP_TUNNEL_ATTR_TTL]      = { .type = NLA_U8 },
+};
 
-static int set_config(const void __user *uconfig, const struct tnl_ops *tnl_ops,
-		      const struct vport *cur_vport,
-		      struct tnl_mutable_config *mutable)
+/* Sets ODP_TUNNEL_ATTR_* fields in 'mutable', which must initially be zeroed. */
+static int tnl_set_config(struct nlattr *options, const struct tnl_ops *tnl_ops,
+			  const struct vport *cur_vport,
+			  struct tnl_mutable_config *mutable)
 {
 	const struct vport *old_vport;
 	const struct tnl_mutable_config *old_mutable;
+	struct nlattr *a[ODP_TUNNEL_ATTR_MAX + 1];
+	int err;
 
-	if (copy_from_user(&mutable->port_config, uconfig, sizeof(struct tnl_port_config)))
-		return -EFAULT;
-
-	mutable->tunnel_hlen = tnl_ops->hdr_len(&mutable->port_config);
-	if (mutable->tunnel_hlen < 0)
-		return mutable->tunnel_hlen;
+	if (!options)
+		return -EINVAL;
 
-	mutable->tunnel_hlen += sizeof(struct iphdr);
+	err = nla_parse_nested(a, ODP_TUNNEL_ATTR_MAX, options, tnl_policy);
+	if (err)
+		return err;
 
-	if (mutable->port_config.daddr == 0)
+	if (!a[ODP_TUNNEL_ATTR_FLAGS] || !a[ODP_TUNNEL_ATTR_DST_IPV4])
 		return -EINVAL;
 
+	mutable->flags = nla_get_u32(a[ODP_TUNNEL_ATTR_FLAGS]) & TNL_F_PUBLIC;
+
+	if (a[ODP_TUNNEL_ATTR_SRC_IPV4])
+		mutable->saddr = nla_get_be32(a[ODP_TUNNEL_ATTR_SRC_IPV4]);
+	mutable->daddr = nla_get_be32(a[ODP_TUNNEL_ATTR_DST_IPV4]);
+
+	if (a[ODP_TUNNEL_ATTR_TOS]) {
+		mutable->tos = nla_get_u8(a[ODP_TUNNEL_ATTR_TOS]);
+		if (mutable->tos != RT_TOS(mutable->tos))
+			return -EINVAL;
+	}
+
+	if (a[ODP_TUNNEL_ATTR_TTL])
+		mutable->ttl = nla_get_u8(a[ODP_TUNNEL_ATTR_TTL]);
+
 	mutable->tunnel_type = tnl_ops->tunnel_type;
-	if (mutable->port_config.flags & TNL_F_IN_KEY_MATCH) {
+	if (!a[ODP_TUNNEL_ATTR_IN_KEY]) {
 		mutable->tunnel_type |= TNL_T_KEY_MATCH;
-		mutable->port_config.in_key = 0;
-	} else
+		mutable->flags |= TNL_F_IN_KEY_MATCH;
+	} else {
 		mutable->tunnel_type |= TNL_T_KEY_EXACT;
+		mutable->in_key = nla_get_be64(a[ODP_TUNNEL_ATTR_IN_KEY]);
+	}
+
+	if (!a[ODP_TUNNEL_ATTR_OUT_KEY])
+		mutable->flags |= TNL_F_OUT_KEY_ACTION;
+	else
+		mutable->out_key = nla_get_be64(a[ODP_TUNNEL_ATTR_OUT_KEY]);
 
-	old_vport = tnl_find_port(mutable->port_config.saddr,
-				  mutable->port_config.daddr,
-				  mutable->port_config.in_key,
-				  mutable->tunnel_type,
+	mutable->tunnel_hlen = tnl_ops->hdr_len(mutable);
+	if (mutable->tunnel_hlen < 0)
+		return mutable->tunnel_hlen;
+
+	mutable->tunnel_hlen += sizeof(struct iphdr);
+
+	old_vport = tnl_find_port(mutable->saddr, mutable->daddr,
+				  mutable->in_key, mutable->tunnel_type,
 				  &old_mutable);
 
 	if (old_vport && old_vport != cur_vport)
 		return -EEXIST;
 
-	if (mutable->port_config.flags & TNL_F_OUT_KEY_ACTION)
-		mutable->port_config.out_key = 0;
-
 	return 0;
 }
 
-struct vport *tnl_create(const char *name, const void __user *config,
+struct vport *tnl_create(const struct vport_parms *parms,
 			 const struct vport_ops *vport_ops,
 			 const struct tnl_ops *tnl_ops)
 {
 	struct vport *vport;
 	struct tnl_vport *tnl_vport;
+	struct tnl_mutable_config *mutable;
+	int initial_frag_id;
 	int err;
 
-	vport = vport_alloc(sizeof(struct tnl_vport), vport_ops);
+	vport = vport_alloc(sizeof(struct tnl_vport), vport_ops, parms);
 	if (IS_ERR(vport)) {
 		err = PTR_ERR(vport);
 		goto error;
@@ -924,22 +1430,33 @@ struct vport *tnl_create(const char *name, const void __user *config,
 
 	tnl_vport = tnl_vport_priv(vport);
 
-	strcpy(tnl_vport->name, name);
+	strcpy(tnl_vport->name, parms->name);
 	tnl_vport->tnl_ops = tnl_ops;
 
-	tnl_vport->mutable = kmalloc(sizeof(struct tnl_mutable_config), GFP_KERNEL);
-	if (!tnl_vport->mutable) {
+	mutable = kzalloc(sizeof(struct tnl_mutable_config), GFP_KERNEL);
+	if (!mutable) {
 		err = -ENOMEM;
 		goto error_free_vport;
 	}
 
-	vport_gen_rand_ether_addr(tnl_vport->mutable->eth_addr);
-	tnl_vport->mutable->mtu = ETH_DATA_LEN;
+	vport_gen_rand_ether_addr(mutable->eth_addr);
+
+	get_random_bytes(&initial_frag_id, sizeof(int));
+	atomic_set(&tnl_vport->frag_id, initial_frag_id);
 
-	err = set_config(config, tnl_ops, NULL, tnl_vport->mutable);
+	err = tnl_set_config(parms->options, tnl_ops, NULL, mutable);
 	if (err)
 		goto error_free_mutable;
 
+	spin_lock_init(&tnl_vport->cache_lock);
+
+#ifdef NEED_CACHE_TIMEOUT
+	tnl_vport->cache_exp_interval = MAX_CACHE_EXP -
+				       (net_random() % (MAX_CACHE_EXP / 2));
+#endif
+
+	rcu_assign_pointer(tnl_vport->mutable, mutable);
+
 	err = add_port(vport);
 	if (err)
 		goto error_free_mutable;
@@ -947,59 +1464,39 @@ struct vport *tnl_create(const char *name, const void __user *config,
 	return vport;
 
 error_free_mutable:
-	kfree(tnl_vport->mutable);
+	kfree(mutable);
 error_free_vport:
 	vport_free(vport);
 error:
 	return ERR_PTR(err);
 }
 
-int tnl_modify(struct vport *vport, const void __user *config)
+int tnl_set_options(struct vport *vport, struct nlattr *options)
 {
 	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
+	const struct tnl_mutable_config *old_mutable;
 	struct tnl_mutable_config *mutable;
 	int err;
-	bool update_hash = false;
 
-	mutable = kmemdup(tnl_vport->mutable, sizeof(struct tnl_mutable_config), GFP_KERNEL);
+	mutable = kzalloc(sizeof(struct tnl_mutable_config), GFP_KERNEL);
 	if (!mutable) {
 		err = -ENOMEM;
 		goto error;
 	}
 
-	err = set_config(config, tnl_vport->tnl_ops, vport, mutable);
+	/* Copy fields whose values should be retained. */
+	old_mutable = rtnl_dereference(tnl_vport->mutable);
+	mutable->seq = old_mutable->seq + 1;
+	memcpy(mutable->eth_addr, old_mutable->eth_addr, ETH_ALEN);
+
+	/* Parse the others configured by userspace. */
+	err = tnl_set_config(options, tnl_vport->tnl_ops, vport, mutable);
 	if (err)
 		goto error_free;
 
-	/*
-	 * Only remove the port from the hash table if something that would
-	 * affect the lookup has changed.
-	 */
-	if (tnl_vport->mutable->port_config.saddr != mutable->port_config.saddr ||
-	    tnl_vport->mutable->port_config.daddr != mutable->port_config.daddr ||
-	    tnl_vport->mutable->port_config.in_key != mutable->port_config.in_key ||
-	    (tnl_vport->mutable->port_config.flags & TNL_F_IN_KEY_MATCH) !=
-	    (mutable->port_config.flags & TNL_F_IN_KEY_MATCH))
-		update_hash = true;
-
-
-	/*
-	 * This update is not atomic but the lookup uses the config, which
-	 * serves as an inherent double check.
-	 */
-	if (update_hash) {
-		err = del_port(vport);
-		if (err)
-			goto error_free;
-	}
-
-	assign_config_rcu(vport, mutable);
-
-	if (update_hash) {
-		err = add_port(vport);
-		if (err)
-			goto error_free;
-	}
+	err = move_port(vport, mutable);
+	if (err)
+		goto error_free;
 
 	return 0;
 
@@ -1009,42 +1506,54 @@ error:
 	return err;
 }
 
-static void free_port(struct rcu_head *rcu)
+int tnl_get_options(const struct vport *vport, struct sk_buff *skb)
 {
-	struct tnl_vport *tnl_vport = container_of(rcu, struct tnl_vport, rcu);
+	const struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
+	const struct tnl_mutable_config *mutable = rcu_dereference_rtnl(tnl_vport->mutable);
+
+	NLA_PUT_U32(skb, ODP_TUNNEL_ATTR_FLAGS, mutable->flags & TNL_F_PUBLIC);
+	NLA_PUT_BE32(skb, ODP_TUNNEL_ATTR_DST_IPV4, mutable->daddr);
+
+	if (!(mutable->flags & TNL_F_IN_KEY_MATCH))
+		NLA_PUT_BE64(skb, ODP_TUNNEL_ATTR_IN_KEY, mutable->in_key);
+	if (!(mutable->flags & TNL_F_OUT_KEY_ACTION))
+		NLA_PUT_BE64(skb, ODP_TUNNEL_ATTR_OUT_KEY, mutable->out_key);
+	if (mutable->saddr)
+		NLA_PUT_BE32(skb, ODP_TUNNEL_ATTR_SRC_IPV4, mutable->saddr);
+	if (mutable->tos)
+		NLA_PUT_U8(skb, ODP_TUNNEL_ATTR_TOS, mutable->tos);
+	if (mutable->ttl)
+		NLA_PUT_U8(skb, ODP_TUNNEL_ATTR_TTL, mutable->ttl);
 
-	kfree(tnl_vport->mutable);
-	vport_free(tnl_vport_to_vport(tnl_vport));
+	return 0;
+
+nla_put_failure:
+	return -EMSGSIZE;
 }
 
-int tnl_destroy(struct vport *vport)
+static void free_port_rcu(struct rcu_head *rcu)
 {
-	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
-	const struct tnl_mutable_config *old_mutable;
-
-	if (vport == tnl_find_port(tnl_vport->mutable->port_config.saddr,
-	    tnl_vport->mutable->port_config.daddr,
-	    tnl_vport->mutable->port_config.in_key,
-	    tnl_vport->mutable->tunnel_type,
-	    &old_mutable))
-		del_port(vport);
+	struct tnl_vport *tnl_vport = container_of(rcu,
+						   struct tnl_vport, rcu);
 
-	call_rcu(&tnl_vport->rcu, free_port);
-
-	return 0;
+	free_cache((struct tnl_cache __force *)tnl_vport->cache);
+	kfree((struct tnl_mutable __force *)tnl_vport->mutable);
+	vport_free(tnl_vport_to_vport(tnl_vport));
 }
 
-int tnl_set_mtu(struct vport *vport, int mtu)
+int tnl_destroy(struct vport *vport)
 {
 	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
-	struct tnl_mutable_config *mutable;
+	const struct tnl_mutable_config *mutable, *old_mutable;
 
-	mutable = kmemdup(tnl_vport->mutable, sizeof(struct tnl_mutable_config), GFP_KERNEL);
-	if (!mutable)
-		return -ENOMEM;
+	mutable = rtnl_dereference(tnl_vport->mutable);
 
-	mutable->mtu = mtu;
-	assign_config_rcu(vport, mutable);
+	if (vport == tnl_find_port(mutable->saddr, mutable->daddr,
+				   mutable->in_key, mutable->tunnel_type,
+				   &old_mutable))
+		del_port(vport);
+
+	call_rcu(&tnl_vport->rcu, free_port_rcu);
 
 	return 0;
 }
@@ -1054,7 +1563,8 @@ int tnl_set_addr(struct vport *vport, const unsigned char *addr)
 	struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
 	struct tnl_mutable_config *mutable;
 
-	mutable = kmemdup(tnl_vport->mutable, sizeof(struct tnl_mutable_config), GFP_KERNEL);
+	mutable = kmemdup(rtnl_dereference(tnl_vport->mutable),
+			  sizeof(struct tnl_mutable_config), GFP_KERNEL);
 	if (!mutable)
 		return -ENOMEM;
 
@@ -1064,7 +1574,6 @@ int tnl_set_addr(struct vport *vport, const unsigned char *addr)
 	return 0;
 }
 
-
 const char *tnl_get_name(const struct vport *vport)
 {
 	const struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
@@ -1074,11 +1583,14 @@ const char *tnl_get_name(const struct vport *vport)
 const unsigned char *tnl_get_addr(const struct vport *vport)
 {
 	const struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
-	return rcu_dereference(tnl_vport->mutable)->eth_addr;
+	return rcu_dereference_rtnl(tnl_vport->mutable)->eth_addr;
 }
 
-int tnl_get_mtu(const struct vport *vport)
+void tnl_free_linked_skbs(struct sk_buff *skb)
 {
-	const struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
-	return rcu_dereference(tnl_vport->mutable)->mtu;
+	while (skb) {
+		struct sk_buff *next = skb->next;
+		kfree_skb(skb);
+		skb = next;
+	}
 }