diff --git a/include/net/ip6_tunnel.h b/include/net/ip6_tunnel.h
index b8529aa1dae7..65c2a9397b3c 100644
--- a/include/net/ip6_tunnel.h
+++ b/include/net/ip6_tunnel.h
@@ -32,6 +32,12 @@ struct __ip6_tnl_parm {
 	__be32			o_key;
 };
 
+struct ip6_tnl_dst {
+	seqlock_t lock;
+	struct dst_entry __rcu *dst;
+	u32 cookie;
+};
+
 /* IPv6 tunnel */
 struct ip6_tnl {
 	struct ip6_tnl __rcu *next;	/* next tunnel in list */
@@ -39,8 +45,7 @@ struct ip6_tnl {
 	struct net *net;	/* netns for packet i/o */
 	struct __ip6_tnl_parm parms;	/* tunnel configuration parameters */
 	struct flowi fl;	/* flowi template for xmit */
-	struct dst_entry *dst_cache;    /* cached dst */
-	u32 dst_cookie;
+	struct ip6_tnl_dst __percpu *dst_cache;	/* cached dst */
 
 	int err_count;
 	unsigned long err_time;
@@ -60,9 +65,11 @@ struct ipv6_tlv_tnl_enc_lim {
 	__u8 encap_limit;	/* tunnel encapsulation limit   */
 } __packed;
 
-struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t);
+struct dst_entry *ip6_tnl_dst_get(struct ip6_tnl *t);
+int ip6_tnl_dst_init(struct ip6_tnl *t);
+void ip6_tnl_dst_destroy(struct ip6_tnl *t);
 void ip6_tnl_dst_reset(struct ip6_tnl *t);
-void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst);
+void ip6_tnl_dst_set(struct ip6_tnl *t, struct dst_entry *dst);
 int ip6_tnl_rcv_ctl(struct ip6_tnl *t, const struct in6_addr *laddr,
 		const struct in6_addr *raddr);
 int ip6_tnl_xmit_ctl(struct ip6_tnl *t, const struct in6_addr *laddr,
diff --git a/net/ipv6/addrconf.c b/net/ipv6/addrconf.c
index 030fefdc9aed..900113376d4e 100644
--- a/net/ipv6/addrconf.c
+++ b/net/ipv6/addrconf.c
@@ -5127,13 +5127,12 @@ static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
 
 			rt = addrconf_get_prefix_route(&ifp->peer_addr, 128,
 						       ifp->idev->dev, 0, 0);
-			if (rt && ip6_del_rt(rt))
-				dst_free(&rt->dst);
+			if (rt)
+				ip6_del_rt(rt);
 		}
 		dst_hold(&ifp->rt->dst);
 
-		if (ip6_del_rt(ifp->rt))
-			dst_free(&ifp->rt->dst);
+		ip6_del_rt(ifp->rt);
 
 		rt_genid_bump_ipv6(net);
 		break;
diff --git a/net/ipv6/ip6_fib.c b/net/ipv6/ip6_fib.c
index 418d9823692b..8a9ec01f4d01 100644
--- a/net/ipv6/ip6_fib.c
+++ b/net/ipv6/ip6_fib.c
@@ -155,6 +155,11 @@ static void node_free(struct fib6_node *fn)
 	kmem_cache_free(fib6_node_kmem, fn);
 }
 
+static void rt6_rcu_free(struct rt6_info *rt)
+{
+	call_rcu(&rt->dst.rcu_head, dst_rcu_free);
+}
+
 static void rt6_free_pcpu(struct rt6_info *non_pcpu_rt)
 {
 	int cpu;
@@ -169,7 +174,7 @@ static void rt6_free_pcpu(struct rt6_info *non_pcpu_rt)
 		ppcpu_rt = per_cpu_ptr(non_pcpu_rt->rt6i_pcpu, cpu);
 		pcpu_rt = *ppcpu_rt;
 		if (pcpu_rt) {
-			dst_free(&pcpu_rt->dst);
+			rt6_rcu_free(pcpu_rt);
 			*ppcpu_rt = NULL;
 		}
 	}
@@ -181,7 +186,7 @@ static void rt6_release(struct rt6_info *rt)
 {
 	if (atomic_dec_and_test(&rt->rt6i_ref)) {
 		rt6_free_pcpu(rt);
-		dst_free(&rt->dst);
+		rt6_rcu_free(rt);
 	}
 }
 
@@ -933,6 +938,10 @@ int fib6_add(struct fib6_node *root, struct rt6_info *rt,
 	int replace_required = 0;
 	int sernum = fib6_new_sernum(info->nl_net);
 
+	if (WARN_ON_ONCE((rt->dst.flags & DST_NOCACHE) &&
+			 !atomic_read(&rt->dst.__refcnt)))
+		return -EINVAL;
+
 	if (info->nlh) {
 		if (!(info->nlh->nlmsg_flags & NLM_F_CREATE))
 			allow_create = 0;
@@ -1025,6 +1034,7 @@ int fib6_add(struct fib6_node *root, struct rt6_info *rt,
 		fib6_start_gc(info->nl_net, rt);
 		if (!(rt->rt6i_flags & RTF_CACHE))
 			fib6_prune_clones(info->nl_net, pn);
+		rt->dst.flags &= ~DST_NOCACHE;
 	}
 
 out:
@@ -1049,7 +1059,8 @@ out:
 			atomic_inc(&pn->leaf->rt6i_ref);
 		}
 #endif
-		dst_free(&rt->dst);
+		if (!(rt->dst.flags & DST_NOCACHE))
+			dst_free(&rt->dst);
 	}
 	return err;
 
@@ -1060,7 +1071,8 @@ out:
 st_failure:
 	if (fn && !(fn->fn_flags & (RTN_RTINFO|RTN_ROOT)))
 		fib6_repair_tree(info->nl_net, fn);
-	dst_free(&rt->dst);
+	if (!(rt->dst.flags & DST_NOCACHE))
+		dst_free(&rt->dst);
 	return err;
 #endif
 }
diff --git a/net/ipv6/ip6_gre.c b/net/ipv6/ip6_gre.c
index 4038c694ec03..646512488c28 100644
--- a/net/ipv6/ip6_gre.c
+++ b/net/ipv6/ip6_gre.c
@@ -634,20 +634,20 @@ static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
 	}
 
 	if (!fl6->flowi6_mark)
-		dst = ip6_tnl_dst_check(tunnel);
+		dst = ip6_tnl_dst_get(tunnel);
 
 	if (!dst) {
-		ndst = ip6_route_output(net, NULL, fl6);
+		dst = ip6_route_output(net, NULL, fl6);
 
-		if (ndst->error)
+		if (dst->error)
 			goto tx_err_link_failure;
-		ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
-		if (IS_ERR(ndst)) {
-			err = PTR_ERR(ndst);
-			ndst = NULL;
+		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
+		if (IS_ERR(dst)) {
+			err = PTR_ERR(dst);
+			dst = NULL;
 			goto tx_err_link_failure;
 		}
-		dst = ndst;
+		ndst = dst;
 	}
 
 	tdev = dst->dev;
@@ -702,12 +702,9 @@ static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
 		skb = new_skb;
 	}
 
-	if (fl6->flowi6_mark) {
-		skb_dst_set(skb, dst);
-		ndst = NULL;
-	} else {
-		skb_dst_set_noref(skb, dst);
-	}
+	if (!fl6->flowi6_mark && ndst)
+		ip6_tnl_dst_set(tunnel, ndst);
+	skb_dst_set(skb, dst);
 
 	proto = NEXTHDR_GRE;
 	if (encap_limit >= 0) {
@@ -762,14 +759,12 @@ static netdev_tx_t ip6gre_xmit2(struct sk_buff *skb,
 	skb_set_inner_protocol(skb, protocol);
 
 	ip6tunnel_xmit(NULL, skb, dev);
-	if (ndst)
-		ip6_tnl_dst_store(tunnel, ndst);
 	return 0;
 tx_err_link_failure:
 	stats->tx_carrier_errors++;
 	dst_link_failure(skb);
 tx_err_dst_release:
-	dst_release(ndst);
+	dst_release(dst);
 	return err;
 }
 
@@ -1223,6 +1218,9 @@ static const struct net_device_ops ip6gre_netdev_ops = {
 
 static void ip6gre_dev_free(struct net_device *dev)
 {
+	struct ip6_tnl *t = netdev_priv(dev);
+
+	ip6_tnl_dst_destroy(t);
 	free_percpu(dev->tstats);
 	free_netdev(dev);
 }
@@ -1245,9 +1243,10 @@ static void ip6gre_tunnel_setup(struct net_device *dev)
 	netif_keep_dst(dev);
 }
 
-static int ip6gre_tunnel_init(struct net_device *dev)
+static int ip6gre_tunnel_init_common(struct net_device *dev)
 {
 	struct ip6_tnl *tunnel;
+	int ret;
 
 	tunnel = netdev_priv(dev);
 
@@ -1255,16 +1254,37 @@ static int ip6gre_tunnel_init(struct net_device *dev)
 	tunnel->net = dev_net(dev);
 	strcpy(tunnel->parms.name, dev->name);
 
+	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
+	if (!dev->tstats)
+		return -ENOMEM;
+
+	ret = ip6_tnl_dst_init(tunnel);
+	if (ret) {
+		free_percpu(dev->tstats);
+		dev->tstats = NULL;
+		return ret;
+	}
+
+	return 0;
+}
+
+static int ip6gre_tunnel_init(struct net_device *dev)
+{
+	struct ip6_tnl *tunnel;
+	int ret;
+
+	ret = ip6gre_tunnel_init_common(dev);
+	if (ret)
+		return ret;
+
+	tunnel = netdev_priv(dev);
+
 	memcpy(dev->dev_addr, &tunnel->parms.laddr, sizeof(struct in6_addr));
 	memcpy(dev->broadcast, &tunnel->parms.raddr, sizeof(struct in6_addr));
 
 	if (ipv6_addr_any(&tunnel->parms.raddr))
 		dev->header_ops = &ip6gre_header_ops;
 
-	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
-	if (!dev->tstats)
-		return -ENOMEM;
-
 	return 0;
 }
 
@@ -1460,19 +1480,16 @@ static void ip6gre_netlink_parms(struct nlattr *data[],
 static int ip6gre_tap_init(struct net_device *dev)
 {
 	struct ip6_tnl *tunnel;
+	int ret;
+
+	ret = ip6gre_tunnel_init_common(dev);
+	if (ret)
+		return ret;
 
 	tunnel = netdev_priv(dev);
 
-	tunnel->dev = dev;
-	tunnel->net = dev_net(dev);
-	strcpy(tunnel->parms.name, dev->name);
-
 	ip6gre_tnl_link_config(tunnel, 1);
 
-	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
-	if (!dev->tstats)
-		return -ENOMEM;
-
 	return 0;
 }
 
diff --git a/net/ipv6/ip6_tunnel.c b/net/ipv6/ip6_tunnel.c
index b0ab420612bc..983f0d20f96d 100644
--- a/net/ipv6/ip6_tunnel.c
+++ b/net/ipv6/ip6_tunnel.c
@@ -126,36 +126,92 @@ static struct net_device_stats *ip6_get_stats(struct net_device *dev)
  * Locking : hash tables are protected by RCU and RTNL
  */
 
-struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
+static void ip6_tnl_per_cpu_dst_set(struct ip6_tnl_dst *idst,
+				    struct dst_entry *dst)
 {
-	struct dst_entry *dst = t->dst_cache;
-
-	if (dst && dst->obsolete &&
-	    !dst->ops->check(dst, t->dst_cookie)) {
-		t->dst_cache = NULL;
-		dst_release(dst);
-		return NULL;
+	write_seqlock_bh(&idst->lock);
+	dst_release(rcu_dereference_protected(
+			    idst->dst,
+			    lockdep_is_held(&idst->lock.lock)));
+	if (dst) {
+		dst_hold(dst);
+		idst->cookie = rt6_get_cookie((struct rt6_info *)dst);
+	} else {
+		idst->cookie = 0;
 	}
+	rcu_assign_pointer(idst->dst, dst);
+	write_sequnlock_bh(&idst->lock);
+}
 
+struct dst_entry *ip6_tnl_dst_get(struct ip6_tnl *t)
+{
+	struct ip6_tnl_dst *idst;
+	struct dst_entry *dst;
+	unsigned int seq;
+	u32 cookie;
+
+	idst = raw_cpu_ptr(t->dst_cache);
+
+	rcu_read_lock();
+	do {
+		seq = read_seqbegin(&idst->lock);
+		dst = rcu_dereference(idst->dst);
+		cookie = idst->cookie;
+	} while (read_seqretry(&idst->lock, seq));
+
+	if (dst && !atomic_inc_not_zero(&dst->__refcnt))
+		dst = NULL;
+	rcu_read_unlock();
+
+	if (dst && dst->obsolete && !dst->ops->check(dst, cookie)) {
+		ip6_tnl_per_cpu_dst_set(idst, NULL);
+		dst_release(dst);
+		dst = NULL;
+	}
 	return dst;
 }
-EXPORT_SYMBOL_GPL(ip6_tnl_dst_check);
+EXPORT_SYMBOL_GPL(ip6_tnl_dst_get);
 
 void ip6_tnl_dst_reset(struct ip6_tnl *t)
 {
-	dst_release(t->dst_cache);
-	t->dst_cache = NULL;
+	int i;
+
+	for_each_possible_cpu(i)
+		ip6_tnl_per_cpu_dst_set(raw_cpu_ptr(t->dst_cache), NULL);
 }
 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset);
 
-void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
+void ip6_tnl_dst_set(struct ip6_tnl *t, struct dst_entry *dst)
 {
-	struct rt6_info *rt = (struct rt6_info *) dst;
-	t->dst_cookie = rt6_get_cookie(rt);
-	dst_release(t->dst_cache);
-	t->dst_cache = dst;
+	ip6_tnl_per_cpu_dst_set(raw_cpu_ptr(t->dst_cache), dst);
+
 }
-EXPORT_SYMBOL_GPL(ip6_tnl_dst_store);
+EXPORT_SYMBOL_GPL(ip6_tnl_dst_set);
+
+void ip6_tnl_dst_destroy(struct ip6_tnl *t)
+{
+	if (!t->dst_cache)
+		return;
+
+	ip6_tnl_dst_reset(t);
+	free_percpu(t->dst_cache);
+}
+EXPORT_SYMBOL_GPL(ip6_tnl_dst_destroy);
+
+int ip6_tnl_dst_init(struct ip6_tnl *t)
+{
+	int i;
+
+	t->dst_cache = alloc_percpu(struct ip6_tnl_dst);
+	if (!t->dst_cache)
+		return -ENOMEM;
+
+	for_each_possible_cpu(i)
+		seqlock_init(&per_cpu_ptr(t->dst_cache, i)->lock);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(ip6_tnl_dst_init);
 
 /**
  * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
@@ -271,6 +327,9 @@ ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 
 static void ip6_dev_free(struct net_device *dev)
 {
+	struct ip6_tnl *t = netdev_priv(dev);
+
+	ip6_tnl_dst_destroy(t);
 	free_percpu(dev->tstats);
 	free_netdev(dev);
 }
@@ -1010,23 +1069,23 @@ static int ip6_tnl_xmit2(struct sk_buff *skb,
 		memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
 		neigh_release(neigh);
 	} else if (!fl6->flowi6_mark)
-		dst = ip6_tnl_dst_check(t);
+		dst = ip6_tnl_dst_get(t);
 
 	if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
 		goto tx_err_link_failure;
 
 	if (!dst) {
-		ndst = ip6_route_output(net, NULL, fl6);
+		dst = ip6_route_output(net, NULL, fl6);
 
-		if (ndst->error)
+		if (dst->error)
 			goto tx_err_link_failure;
-		ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(fl6), NULL, 0);
-		if (IS_ERR(ndst)) {
-			err = PTR_ERR(ndst);
-			ndst = NULL;
+		dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
+		if (IS_ERR(dst)) {
+			err = PTR_ERR(dst);
+			dst = NULL;
 			goto tx_err_link_failure;
 		}
-		dst = ndst;
+		ndst = dst;
 	}
 
 	tdev = dst->dev;
@@ -1072,12 +1131,11 @@ static int ip6_tnl_xmit2(struct sk_buff *skb,
 		consume_skb(skb);
 		skb = new_skb;
 	}
-	if (fl6->flowi6_mark) {
-		skb_dst_set(skb, dst);
-		ndst = NULL;
-	} else {
-		skb_dst_set_noref(skb, dst);
-	}
+
+	if (!fl6->flowi6_mark && ndst)
+		ip6_tnl_dst_set(t, ndst);
+	skb_dst_set(skb, dst);
+
 	skb->transport_header = skb->network_header;
 
 	proto = fl6->flowi6_proto;
@@ -1101,14 +1159,12 @@ static int ip6_tnl_xmit2(struct sk_buff *skb,
 	ipv6h->saddr = fl6->saddr;
 	ipv6h->daddr = fl6->daddr;
 	ip6tunnel_xmit(NULL, skb, dev);
-	if (ndst)
-		ip6_tnl_dst_store(t, ndst);
 	return 0;
 tx_err_link_failure:
 	stats->tx_carrier_errors++;
 	dst_link_failure(skb);
 tx_err_dst_release:
-	dst_release(ndst);
+	dst_release(dst);
 	return err;
 }
 
@@ -1573,12 +1629,21 @@ static inline int
 ip6_tnl_dev_init_gen(struct net_device *dev)
 {
 	struct ip6_tnl *t = netdev_priv(dev);
+	int ret;
 
 	t->dev = dev;
 	t->net = dev_net(dev);
 	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 	if (!dev->tstats)
 		return -ENOMEM;
+
+	ret = ip6_tnl_dst_init(t);
+	if (ret) {
+		free_percpu(dev->tstats);
+		dev->tstats = NULL;
+		return ret;
+	}
+
 	return 0;
 }
 
diff --git a/net/ipv6/route.c b/net/ipv6/route.c
index 53617d715188..3d3c1b294725 100644
--- a/net/ipv6/route.c
+++ b/net/ipv6/route.c
@@ -1322,8 +1322,7 @@ static void ip6_link_failure(struct sk_buff *skb)
 	if (rt) {
 		if (rt->rt6i_flags & RTF_CACHE) {
 			dst_hold(&rt->dst);
-			if (ip6_del_rt(rt))
-				dst_free(&rt->dst);
+			ip6_del_rt(rt);
 		} else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
 			rt->rt6i_node->fn_sernum = -1;
 		}
@@ -2028,7 +2027,8 @@ static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
 	struct fib6_table *table;
 	struct net *net = dev_net(rt->dst.dev);
 
-	if (rt == net->ipv6.ip6_null_entry) {
+	if (rt == net->ipv6.ip6_null_entry ||
+	    rt->dst.flags & DST_NOCACHE) {
 		err = -ENOENT;
 		goto out;
 	}
@@ -2515,6 +2515,7 @@ struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
 	rt->rt6i_dst.addr = *addr;
 	rt->rt6i_dst.plen = 128;
 	rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
+	rt->dst.flags |= DST_NOCACHE;
 
 	atomic_set(&rt->dst.__refcnt, 1);