ipvs: optimize dst usage for real server

Currently when forwarding requests to real servers we use dst_lock and atomic operations when cloning the dst_cache value. As the dst_cache value does not change most of the time it is better to use RCU and to lock dst_lock only when we need to replace the obsoleted dst. For this to work we keep dst_cache in new structure protected by RCU. For packets to remote real servers we will use noref version of dst_cache, it will be valid while we are in RCU read-side critical section because now dst_release for replaced dsts will be invoked after the grace period. Packets to local real servers that are passed to local stack with NF_ACCEPT need a dst clone. Signed-off-by: Julian Anastasov <ja@ssi.bg> Signed-off by: Hans Schillstrom <hans@schillstrom.com> Signed-off-by: Simon Horman <horms@verge.net.au>

ipvs: optimize dst usage for real server
Currently when forwarding requests to real servers we use dst_lock and atomic operations when cloning the dst_cache value. As the dst_cache value does not change most of the time it is better to use RCU and to lock dst_lock only when we need to replace the obsoleted dst. For this to work we keep dst_cache in new structure protected by RCU. For packets to remote real servers we will use noref version of dst_cache, it will be valid while we are in RCU read-side critical section because now dst_release for replaced dsts will be invoked after the grace period. Packets to local real servers that are passed to local stack with NF_ACCEPT need a dst clone. Signed-off-by: Julian Anastasov <ja@ssi.bg> Signed-off by: Hans Schillstrom <hans@schillstrom.com> Signed-off-by: Simon Horman <horms@verge.net.au>
Julian Anastasov · Pablo Neira Ayuso
1 parent 4115ded131
Showing 4 changed files with 177 additions and 59 deletions Side-by-side Diff
include/net/ip_vs.h
net/netfilter/ipvs/ip_vs_core.c
net/netfilter/ipvs/ip_vs_ctl.c
net/netfilter/ipvs/ip_vs_xmit.c
@@ -724,6 +724,13 @@
 	struct ip_vs_pe		*pe;
 };
  
+/* Information for cached dst */
+struct ip_vs_dest_dst {
+	struct dst_entry	*dst_cache;	/* destination cache entry */
+	u32			dst_cookie;
+	union nf_inet_addr	dst_saddr;
+	struct rcu_head		rcu_head;
+};
  
 /*
  *	The real server destination forwarding entry
@@ -752,9 +759,7 @@
  
 	/* for destination cache */
 	spinlock_t		dst_lock;	/* lock of dst_cache */
-	struct dst_entry	*dst_cache;	/* destination cache entry */
-	u32			dst_cookie;
-	union nf_inet_addr	dst_saddr;
+	struct ip_vs_dest_dst __rcu *dest_dst;	/* cached dst info */
  
 	/* for virtual service */
 	struct ip_vs_service	*svc;		/* service it belongs to */
@@ -1427,6 +1432,7 @@
 extern int ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
 			   struct ip_vs_protocol *pp, int offset,
 			   unsigned int hooknum, struct ip_vs_iphdr *iph);
+extern void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
  
 #ifdef CONFIG_IP_VS_IPV6
 extern int ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
@@ -1395,10 +1395,13 @@
 				goto ignore_ipip;
 			/* Prefer the resulting PMTU */
 			if (dest) {
-				spin_lock(&dest->dst_lock);
-				if (dest->dst_cache)
-					mtu = dst_mtu(dest->dst_cache);
-				spin_unlock(&dest->dst_lock);
+				struct ip_vs_dest_dst *dest_dst;
+
+				rcu_read_lock();
+				dest_dst = rcu_dereference(dest->dest_dst);
+				if (dest_dst)
+					mtu = dst_mtu(dest_dst->dst_cache);
+				rcu_read_unlock();
 			}
 			if (mtu > 68 + sizeof(struct iphdr))
 				mtu -= sizeof(struct iphdr);
@@ -641,15 +641,26 @@
 	return dest;
 }
  
-/* Release dst_cache for dest in user context */
+void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
+{
+	struct ip_vs_dest_dst *dest_dst = container_of(head,
+						       struct ip_vs_dest_dst,
+						       rcu_head);
+
+	dst_release(dest_dst->dst_cache);
+	kfree(dest_dst);
+}
+
+/* Release dest_dst and dst_cache for dest in user context */
 static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
 {
-	struct dst_entry *old_dst;
+	struct ip_vs_dest_dst *old;
  
-	old_dst = dest->dst_cache;
-	dest->dst_cache = NULL;
-	dst_release(old_dst);
-	dest->dst_saddr.ip = 0;
+	old = rcu_dereference_protected(dest->dest_dst, 1);
+	if (old) {
+		RCU_INIT_POINTER(dest->dest_dst, NULL);
+		call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
+	}
 }
  
 /*
@@ -1513,7 +1524,7 @@
 ip_vs_forget_dev(struct ip_vs_dest *dest, struct net_device *dev)
 {
 	spin_lock_bh(&dest->dst_lock);
-	if (dest->dst_cache && dest->dst_cache->dev == dev) {
+	if (dest->dest_dst && dest->dest_dst->dst_cache->dev == dev) {
 		IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
 			      dev->name,
 			      IP_VS_DBG_ADDR(dest->af, &dest->addr),
@@ -17,6 +17,8 @@
  * - not all connections have destination server, for example,
  * connections in backup server when fwmark is used
  * - bypass connections use daddr from packet
+ * - we can use dst without ref while sending in RCU section, we use
+ * ref when returning NF_ACCEPT for NAT-ed packet via loopback
  * LOCAL_OUT rules:
  * - skb->dev is NULL, skb->protocol is not set (both are set in POST_ROUTING)
  * - skb->pkt_type is not set yet
  
  
  
  
  
  
  
  
@@ -54,34 +56,51 @@
 	IP_VS_RT_MODE_TUNNEL	= 32,/* Tunnel mode */
 };
  
+static inline struct ip_vs_dest_dst *ip_vs_dest_dst_alloc(void)
+{
+	return kmalloc(sizeof(struct ip_vs_dest_dst), GFP_ATOMIC);
+}
+
+static inline void ip_vs_dest_dst_free(struct ip_vs_dest_dst *dest_dst)
+{
+	kfree(dest_dst);
+}
+
 /*
  *      Destination cache to speed up outgoing route lookup
  */
 static inline void
-__ip_vs_dst_set(struct ip_vs_dest *dest, struct dst_entry *dst, u32 dst_cookie)
+__ip_vs_dst_set(struct ip_vs_dest *dest, struct ip_vs_dest_dst *dest_dst,
+		struct dst_entry *dst, u32 dst_cookie)
 {
-	struct dst_entry *old_dst;
+	struct ip_vs_dest_dst *old;
  
-	old_dst = dest->dst_cache;
-	dest->dst_cache = dst;
-	dest->dst_cookie = dst_cookie;
-	dst_release(old_dst);
+	old = rcu_dereference_protected(dest->dest_dst,
+					lockdep_is_held(&dest->dst_lock));
+
+	if (dest_dst) {
+		dest_dst->dst_cache = dst;
+		dest_dst->dst_cookie = dst_cookie;
+	}
+	rcu_assign_pointer(dest->dest_dst, dest_dst);
+
+	if (old)
+		call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
 }
  
-static inline struct dst_entry *
+static inline struct ip_vs_dest_dst *
 __ip_vs_dst_check(struct ip_vs_dest *dest)
 {
-	struct dst_entry *dst = dest->dst_cache;
+	struct ip_vs_dest_dst *dest_dst = rcu_dereference(dest->dest_dst);
+	struct dst_entry *dst;
  
-	if (!dst)
+	if (!dest_dst)
 		return NULL;
-	if (dst->obsolete && dst->ops->check(dst, dest->dst_cookie) == NULL) {
-		dest->dst_cache = NULL;
-		dst_release(dst);
+	dst = dest_dst->dst_cache;
+	if (dst->obsolete &&
+	    dst->ops->check(dst, dest_dst->dst_cookie) == NULL)
 		return NULL;
-	}
-	dst_hold(dst);
-	return dst;
+	return dest_dst;
 }
  
 static inline bool
  
  
  
  
  
  
  
  
  
@@ -144,35 +163,48 @@
 {
 	struct net *net = dev_net(skb_dst(skb)->dev);
 	struct netns_ipvs *ipvs = net_ipvs(net);
+	struct ip_vs_dest_dst *dest_dst;
 	struct rtable *rt;			/* Route to the other host */
 	struct rtable *ort;			/* Original route */
 	struct iphdr *iph;
 	__be16 df;
 	int mtu;
-	int local;
+	int local, noref = 1;
  
 	if (dest) {
-		spin_lock(&dest->dst_lock);
-		rt = (struct rtable *) __ip_vs_dst_check(dest);
-		if (!rt) {
+		dest_dst = __ip_vs_dst_check(dest);
+		if (likely(dest_dst))
+			rt = (struct rtable *) dest_dst->dst_cache;
+		else {
+			dest_dst = ip_vs_dest_dst_alloc();
+			spin_lock(&dest->dst_lock);
+			if (!dest_dst) {
+				__ip_vs_dst_set(dest, NULL, NULL, 0);
+				spin_unlock(&dest->dst_lock);
+				goto err_unreach;
+			}
 			rt = do_output_route4(net, dest->addr.ip, rt_mode,
-					      &dest->dst_saddr.ip);
+					      &dest_dst->dst_saddr.ip);
 			if (!rt) {
+				__ip_vs_dst_set(dest, NULL, NULL, 0);
 				spin_unlock(&dest->dst_lock);
+				ip_vs_dest_dst_free(dest_dst);
 				goto err_unreach;
 			}
-			__ip_vs_dst_set(dest, dst_clone(&rt->dst), 0);
+			__ip_vs_dst_set(dest, dest_dst, &rt->dst, 0);
+			spin_unlock(&dest->dst_lock);
 			IP_VS_DBG(10, "new dst %pI4, src %pI4, refcnt=%d\n",
-				  &dest->addr.ip, &dest->dst_saddr.ip,
+				  &dest->addr.ip, &dest_dst->dst_saddr.ip,
 				  atomic_read(&rt->dst.__refcnt));
 		}
 		daddr = dest->addr.ip;
 		if (ret_saddr)
-			*ret_saddr = dest->dst_saddr.ip;
-		spin_unlock(&dest->dst_lock);
+			*ret_saddr = dest_dst->dst_saddr.ip;
 	} else {
 		__be32 saddr = htonl(INADDR_ANY);
  
+		noref = 0;
+
 		/* For such unconfigured boxes avoid many route lookups
 		 * for performance reasons because we do not remember saddr
 		 */
@@ -210,7 +242,8 @@
 			goto err_put;
 		}
 		/* skb to local stack, preserve old route */
-		ip_rt_put(rt);
+		if (!noref)
+			ip_rt_put(rt);
 		return local;
 	}
  
  
@@ -240,12 +273,19 @@
 	}
  
 	skb_dst_drop(skb);
-	skb_dst_set(skb, &rt->dst);
+	if (noref) {
+		if (!local)
+			skb_dst_set_noref_force(skb, &rt->dst);
+		else
+			skb_dst_set(skb, dst_clone(&rt->dst));
+	} else
+		skb_dst_set(skb, &rt->dst);
  
 	return local;
  
 err_put:
-	ip_rt_put(rt);
+	if (!noref)
+		ip_rt_put(rt);
 	return -1;
  
 err_unreach:
  
  
  
  
  
  
  
  
  
  
@@ -303,36 +343,48 @@
 		      struct ip_vs_iphdr *ipvsh, int do_xfrm, int rt_mode)
 {
 	struct net *net = dev_net(skb_dst(skb)->dev);
+	struct ip_vs_dest_dst *dest_dst;
 	struct rt6_info *rt;			/* Route to the other host */
 	struct rt6_info *ort;			/* Original route */
 	struct dst_entry *dst;
 	int mtu;
-	int local;
+	int local, noref = 1;
  
 	if (dest) {
-		spin_lock(&dest->dst_lock);
-		rt = (struct rt6_info *)__ip_vs_dst_check(dest);
-		if (!rt) {
+		dest_dst = __ip_vs_dst_check(dest);
+		if (likely(dest_dst))
+			rt = (struct rt6_info *) dest_dst->dst_cache;
+		else {
 			u32 cookie;
  
+			dest_dst = ip_vs_dest_dst_alloc();
+			spin_lock(&dest->dst_lock);
+			if (!dest_dst) {
+				__ip_vs_dst_set(dest, NULL, NULL, 0);
+				spin_unlock(&dest->dst_lock);
+				goto err_unreach;
+			}
 			dst = __ip_vs_route_output_v6(net, &dest->addr.in6,
-						      &dest->dst_saddr.in6,
+						      &dest_dst->dst_saddr.in6,
 						      do_xfrm);
 			if (!dst) {
+				__ip_vs_dst_set(dest, NULL, NULL, 0);
 				spin_unlock(&dest->dst_lock);
+				ip_vs_dest_dst_free(dest_dst);
 				goto err_unreach;
 			}
 			rt = (struct rt6_info *) dst;
 			cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
-			__ip_vs_dst_set(dest, dst_clone(&rt->dst), cookie);
+			__ip_vs_dst_set(dest, dest_dst, &rt->dst, cookie);
+			spin_unlock(&dest->dst_lock);
 			IP_VS_DBG(10, "new dst %pI6, src %pI6, refcnt=%d\n",
-				  &dest->addr.in6, &dest->dst_saddr.in6,
+				  &dest->addr.in6, &dest_dst->dst_saddr.in6,
 				  atomic_read(&rt->dst.__refcnt));
 		}
 		if (ret_saddr)
-			*ret_saddr = dest->dst_saddr.in6;
-		spin_unlock(&dest->dst_lock);
+			*ret_saddr = dest_dst->dst_saddr.in6;
 	} else {
+		noref = 0;
 		dst = __ip_vs_route_output_v6(net, daddr, ret_saddr, do_xfrm);
 		if (!dst)
 			goto err_unreach;
@@ -367,7 +419,8 @@
 			goto err_put;
 		}
 		/* skb to local stack, preserve old route */
-		dst_release(&rt->dst);
+		if (!noref)
+			dst_release(&rt->dst);
 		return local;
 	}
  
  
@@ -399,12 +452,19 @@
 	}
  
 	skb_dst_drop(skb);
-	skb_dst_set(skb, &rt->dst);
+	if (noref) {
+		if (!local)
+			skb_dst_set_noref_force(skb, &rt->dst);
+		else
+			skb_dst_set(skb, dst_clone(&rt->dst));
+	} else
+		skb_dst_set(skb, &rt->dst);
  
 	return local;
  
 err_put:
-	dst_release(&rt->dst);
+	if (!noref)
+		dst_release(&rt->dst);
 	return -1;
  
 err_unreach:
@@ -494,6 +554,7 @@
  
 	EnterFunction(10);
  
+	rcu_read_lock();
 	if (__ip_vs_get_out_rt(skb, NULL, iph->daddr, IP_VS_RT_MODE_NON_LOCAL,
 			       NULL) < 0)
 		goto tx_error;
  
@@ -504,12 +565,14 @@
 	skb->local_df = 1;
  
 	ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 0);
+	rcu_read_unlock();
  
 	LeaveFunction(10);
 	return NF_STOLEN;
  
  tx_error:
 	kfree_skb(skb);
+	rcu_read_unlock();
 	LeaveFunction(10);
 	return NF_STOLEN;
 }
@@ -521,6 +584,7 @@
 {
 	EnterFunction(10);
  
+	rcu_read_lock();
 	if (__ip_vs_get_out_rt_v6(skb, NULL, &ipvsh->daddr.in6, NULL,
 				  ipvsh, 0, IP_VS_RT_MODE_NON_LOCAL) < 0)
 		goto tx_error;
  
@@ -529,12 +593,14 @@
 	skb->local_df = 1;
  
 	ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 0);
+	rcu_read_unlock();
  
 	LeaveFunction(10);
 	return NF_STOLEN;
  
  tx_error:
 	kfree_skb(skb);
+	rcu_read_unlock();
 	LeaveFunction(10);
 	return NF_STOLEN;
 }
@@ -553,6 +619,7 @@
  
 	EnterFunction(10);
  
+	rcu_read_lock();
 	/* check if it is a connection of no-client-port */
 	if (unlikely(cp->flags & IP_VS_CONN_F_NO_CPORT)) {
 		__be16 _pt, *p;
  
@@ -620,12 +687,14 @@
 	skb->local_df = 1;
  
 	rc = ip_vs_nat_send_or_cont(NFPROTO_IPV4, skb, cp, local);
+	rcu_read_unlock();
  
 	LeaveFunction(10);
 	return rc;
  
   tx_error:
 	kfree_skb(skb);
+	rcu_read_unlock();
 	LeaveFunction(10);
 	return NF_STOLEN;
 }
@@ -640,6 +709,7 @@
  
 	EnterFunction(10);
  
+	rcu_read_lock();
 	/* check if it is a connection of no-client-port */
 	if (unlikely(cp->flags & IP_VS_CONN_F_NO_CPORT && !ipvsh->fragoffs)) {
 		__be16 _pt, *p;
@@ -707,6 +777,7 @@
 	skb->local_df = 1;
  
 	rc = ip_vs_nat_send_or_cont(NFPROTO_IPV6, skb, cp, local);
+	rcu_read_unlock();
  
 	LeaveFunction(10);
 	return rc;
@@ -714,6 +785,7 @@
 tx_error:
 	LeaveFunction(10);
 	kfree_skb(skb);
+	rcu_read_unlock();
 	return NF_STOLEN;
 }
 #endif
@@ -755,6 +827,7 @@
  
 	EnterFunction(10);
  
+	rcu_read_lock();
 	local = __ip_vs_get_out_rt(skb, cp->dest, cp->daddr.ip,
 				   IP_VS_RT_MODE_LOCAL |
 				   IP_VS_RT_MODE_NON_LOCAL |
  
@@ -762,8 +835,10 @@
 				   IP_VS_RT_MODE_TUNNEL, &saddr);
 	if (local < 0)
 		goto tx_error;
-	if (local)
+	if (local) {
+		rcu_read_unlock();
 		return ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 1);
+	}
  
 	rt = skb_rtable(skb);
 	tdev = rt->dst.dev;
@@ -818,6 +893,7 @@
 		ip_local_out(skb);
 	else if (ret == NF_DROP)
 		kfree_skb(skb);
+	rcu_read_unlock();
  
 	LeaveFunction(10);
  
@@ -825,6 +901,7 @@
  
   tx_error:
 	kfree_skb(skb);
+	rcu_read_unlock();
 	LeaveFunction(10);
 	return NF_STOLEN;
 }
@@ -844,6 +921,7 @@
  
 	EnterFunction(10);
  
+	rcu_read_lock();
 	local = __ip_vs_get_out_rt_v6(skb, cp->dest, &cp->daddr.in6,
 				      &saddr, ipvsh, 1,
 				      IP_VS_RT_MODE_LOCAL |
  
@@ -851,8 +929,10 @@
 				      IP_VS_RT_MODE_TUNNEL);
 	if (local < 0)
 		goto tx_error;
-	if (local)
+	if (local) {
+		rcu_read_unlock();
 		return ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 1);
+	}
  
 	rt = (struct rt6_info *) skb_dst(skb);
 	tdev = rt->dst.dev;
@@ -901,6 +981,7 @@
 		ip6_local_out(skb);
 	else if (ret == NF_DROP)
 		kfree_skb(skb);
+	rcu_read_unlock();
  
 	LeaveFunction(10);
  
@@ -908,6 +989,7 @@
  
 tx_error:
 	kfree_skb(skb);
+	rcu_read_unlock();
 	LeaveFunction(10);
 	return NF_STOLEN;
 }
  
  
@@ -926,14 +1008,17 @@
  
 	EnterFunction(10);
  
+	rcu_read_lock();
 	local = __ip_vs_get_out_rt(skb, cp->dest, cp->daddr.ip,
 				   IP_VS_RT_MODE_LOCAL |
 				   IP_VS_RT_MODE_NON_LOCAL |
 				   IP_VS_RT_MODE_KNOWN_NH, NULL);
 	if (local < 0)
 		goto tx_error;
-	if (local)
+	if (local) {
+		rcu_read_unlock();
 		return ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 1);
+	}
  
 	ip_send_check(ip_hdr(skb));
  
  
@@ -941,12 +1026,14 @@
 	skb->local_df = 1;
  
 	ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 0);
+	rcu_read_unlock();
  
 	LeaveFunction(10);
 	return NF_STOLEN;
  
   tx_error:
 	kfree_skb(skb);
+	rcu_read_unlock();
 	LeaveFunction(10);
 	return NF_STOLEN;
 }
  
  
  
  
@@ -960,25 +1047,30 @@
  
 	EnterFunction(10);
  
+	rcu_read_lock();
 	local = __ip_vs_get_out_rt_v6(skb, cp->dest, &cp->daddr.in6, NULL,
 				      ipvsh, 0,
 				      IP_VS_RT_MODE_LOCAL |
 				      IP_VS_RT_MODE_NON_LOCAL);
 	if (local < 0)
 		goto tx_error;
-	if (local)
+	if (local) {
+		rcu_read_unlock();
 		return ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 1);
+	}
  
 	/* Another hack: avoid icmp_send in ip_fragment */
 	skb->local_df = 1;
  
 	ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 0);
+	rcu_read_unlock();
  
 	LeaveFunction(10);
 	return NF_STOLEN;
  
 tx_error:
 	kfree_skb(skb);
+	rcu_read_unlock();
 	LeaveFunction(10);
 	return NF_STOLEN;
 }
@@ -1023,6 +1115,7 @@
 	rt_mode = (hooknum != NF_INET_FORWARD) ?
 		  IP_VS_RT_MODE_LOCAL | IP_VS_RT_MODE_NON_LOCAL |
 		  IP_VS_RT_MODE_RDR : IP_VS_RT_MODE_NON_LOCAL;
+	rcu_read_lock();
 	local = __ip_vs_get_out_rt(skb, cp->dest, cp->daddr.ip, rt_mode, NULL);
 	if (local < 0)
 		goto tx_error;
  
@@ -1067,10 +1160,12 @@
 	skb->local_df = 1;
  
 	rc = ip_vs_nat_send_or_cont(NFPROTO_IPV4, skb, cp, local);
+	rcu_read_unlock();
 	goto out;
  
   tx_error:
-	dev_kfree_skb(skb);
+	kfree_skb(skb);
+	rcu_read_unlock();
 	rc = NF_STOLEN;
   out:
 	LeaveFunction(10);
@@ -1111,6 +1206,7 @@
 	rt_mode = (hooknum != NF_INET_FORWARD) ?
 		  IP_VS_RT_MODE_LOCAL | IP_VS_RT_MODE_NON_LOCAL |
 		  IP_VS_RT_MODE_RDR : IP_VS_RT_MODE_NON_LOCAL;
+	rcu_read_lock();
 	local = __ip_vs_get_out_rt_v6(skb, cp->dest, &cp->daddr.in6, NULL,
 				      ipvsh, 0, rt_mode);
 	if (local < 0)
  
@@ -1156,10 +1252,12 @@
 	skb->local_df = 1;
  
 	rc = ip_vs_nat_send_or_cont(NFPROTO_IPV6, skb, cp, local);
+	rcu_read_unlock();
 	goto out;
  
 tx_error:
-	dev_kfree_skb(skb);
+	kfree_skb(skb);
+	rcu_read_unlock();
 	rc = NF_STOLEN;
 out:
 	LeaveFunction(10);
...	...	@@ -724,6 +724,13 @@
724	724	struct ip_vs_pe *pe;
725	725	};
726	726
	727	+/* Information for cached dst */
	728	+struct ip_vs_dest_dst {
	729	+ struct dst_entry dst_cache; / destination cache entry */
	730	+ u32 dst_cookie;
	731	+ union nf_inet_addr dst_saddr;
	732	+ struct rcu_head rcu_head;
	733	+};
727	734
728	735	/*
729	736	* The real server destination forwarding entry
...	...	@@ -752,9 +759,7 @@
752	759
753	760	/* for destination cache */
754	761	spinlock_t dst_lock; /* lock of dst_cache */
755		- struct dst_entry dst_cache; / destination cache entry */
756		- u32 dst_cookie;
757		- union nf_inet_addr dst_saddr;
	762	+ struct ip_vs_dest_dst __rcu dest_dst; / cached dst info */
758	763
759	764	/* for virtual service */
760	765	struct ip_vs_service svc; / service it belongs to */
...	...	@@ -1427,6 +1432,7 @@
1427	1432	extern int ip_vs_icmp_xmit(struct sk_buff skb, struct ip_vs_conn cp,
1428	1433	struct ip_vs_protocol *pp, int offset,
1429	1434	unsigned int hooknum, struct ip_vs_iphdr *iph);
	1435	+extern void ip_vs_dest_dst_rcu_free(struct rcu_head *head);
1430	1436
1431	1437	#ifdef CONFIG_IP_VS_IPV6
1432	1438	extern int ip_vs_bypass_xmit_v6(struct sk_buff skb, struct ip_vs_conn cp,
...	...	@@ -1395,10 +1395,13 @@
1395	1395	goto ignore_ipip;
1396	1396	/* Prefer the resulting PMTU */
1397	1397	if (dest) {
1398		- spin_lock(&dest->dst_lock);
1399		- if (dest->dst_cache)
1400		- mtu = dst_mtu(dest->dst_cache);
1401		- spin_unlock(&dest->dst_lock);
	1398	+ struct ip_vs_dest_dst *dest_dst;
	1399	+
	1400	+ rcu_read_lock();
	1401	+ dest_dst = rcu_dereference(dest->dest_dst);
	1402	+ if (dest_dst)
	1403	+ mtu = dst_mtu(dest_dst->dst_cache);
	1404	+ rcu_read_unlock();
1402	1405	}
1403	1406	if (mtu > 68 + sizeof(struct iphdr))
1404	1407	mtu -= sizeof(struct iphdr);
...	...	@@ -641,15 +641,26 @@
641	641	return dest;
642	642	}
643	643
644		-/* Release dst_cache for dest in user context */
	644	+void ip_vs_dest_dst_rcu_free(struct rcu_head *head)
	645	+{
	646	+ struct ip_vs_dest_dst *dest_dst = container_of(head,
	647	+ struct ip_vs_dest_dst,
	648	+ rcu_head);
	649	+
	650	+ dst_release(dest_dst->dst_cache);
	651	+ kfree(dest_dst);
	652	+}
	653	+
	654	+/* Release dest_dst and dst_cache for dest in user context */
645	655	static void __ip_vs_dst_cache_reset(struct ip_vs_dest *dest)
646	656	{
647		- struct dst_entry *old_dst;
	657	+ struct ip_vs_dest_dst *old;
648	658
649		- old_dst = dest->dst_cache;
650		- dest->dst_cache = NULL;
651		- dst_release(old_dst);
652		- dest->dst_saddr.ip = 0;
	659	+ old = rcu_dereference_protected(dest->dest_dst, 1);
	660	+ if (old) {
	661	+ RCU_INIT_POINTER(dest->dest_dst, NULL);
	662	+ call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
	663	+ }
653	664	}
654	665
655	666	/*
...	...	@@ -1513,7 +1524,7 @@
1513	1524	ip_vs_forget_dev(struct ip_vs_dest dest, struct net_device dev)
1514	1525	{
1515	1526	spin_lock_bh(&dest->dst_lock);
1516		- if (dest->dst_cache && dest->dst_cache->dev == dev) {
	1527	+ if (dest->dest_dst && dest->dest_dst->dst_cache->dev == dev) {
1517	1528	IP_VS_DBG_BUF(3, "Reset dev:%s dest %s:%u ,dest->refcnt=%d\n",
1518	1529	dev->name,
1519	1530	IP_VS_DBG_ADDR(dest->af, &dest->addr),
...	...	@@ -17,6 +17,8 @@
17	17	* - not all connections have destination server, for example,
18	18	* connections in backup server when fwmark is used
19	19	* - bypass connections use daddr from packet
	20	+ * - we can use dst without ref while sending in RCU section, we use
	21	+ * ref when returning NF_ACCEPT for NAT-ed packet via loopback
20	22	* LOCAL_OUT rules:
21	23	* - skb->dev is NULL, skb->protocol is not set (both are set in POST_ROUTING)
22	24	* - skb->pkt_type is not set yet
23	25
24	26
25	27
26	28
27	29
28	30
29	31
30	32
...	...	@@ -54,34 +56,51 @@
54	56	IP_VS_RT_MODE_TUNNEL = 32,/* Tunnel mode */
55	57	};
56	58
	59	+static inline struct ip_vs_dest_dst *ip_vs_dest_dst_alloc(void)
	60	+{
	61	+ return kmalloc(sizeof(struct ip_vs_dest_dst), GFP_ATOMIC);
	62	+}
	63	+
	64	+static inline void ip_vs_dest_dst_free(struct ip_vs_dest_dst *dest_dst)
	65	+{
	66	+ kfree(dest_dst);
	67	+}
	68	+
57	69	/*
58	70	* Destination cache to speed up outgoing route lookup
59	71	*/
60	72	static inline void
61		-__ip_vs_dst_set(struct ip_vs_dest dest, struct dst_entry dst, u32 dst_cookie)
	73	+__ip_vs_dst_set(struct ip_vs_dest dest, struct ip_vs_dest_dst dest_dst,
	74	+ struct dst_entry *dst, u32 dst_cookie)
62	75	{
63		- struct dst_entry *old_dst;
	76	+ struct ip_vs_dest_dst *old;
64	77
65		- old_dst = dest->dst_cache;
66		- dest->dst_cache = dst;
67		- dest->dst_cookie = dst_cookie;
68		- dst_release(old_dst);
	78	+ old = rcu_dereference_protected(dest->dest_dst,
	79	+ lockdep_is_held(&dest->dst_lock));
	80	+
	81	+ if (dest_dst) {
	82	+ dest_dst->dst_cache = dst;
	83	+ dest_dst->dst_cookie = dst_cookie;
	84	+ }
	85	+ rcu_assign_pointer(dest->dest_dst, dest_dst);
	86	+
	87	+ if (old)
	88	+ call_rcu(&old->rcu_head, ip_vs_dest_dst_rcu_free);
69	89	}
70	90
71		-static inline struct dst_entry *
	91	+static inline struct ip_vs_dest_dst *
72	92	__ip_vs_dst_check(struct ip_vs_dest *dest)
73	93	{
74		- struct dst_entry *dst = dest->dst_cache;
	94	+ struct ip_vs_dest_dst *dest_dst = rcu_dereference(dest->dest_dst);
	95	+ struct dst_entry *dst;
75	96
76		- if (!dst)
	97	+ if (!dest_dst)
77	98	return NULL;
78		- if (dst->obsolete && dst->ops->check(dst, dest->dst_cookie) == NULL) {
79		- dest->dst_cache = NULL;
80		- dst_release(dst);
	99	+ dst = dest_dst->dst_cache;
	100	+ if (dst->obsolete &&
	101	+ dst->ops->check(dst, dest_dst->dst_cookie) == NULL)
81	102	return NULL;
82		- }
83		- dst_hold(dst);
84		- return dst;
	103	+ return dest_dst;
85	104	}
86	105
87	106	static inline bool
88	107
89	108
90	109
91	110
92	111
93	112
94	113
95	114
96	115
...	...	@@ -144,35 +163,48 @@
144	163	{
145	164	struct net *net = dev_net(skb_dst(skb)->dev);
146	165	struct netns_ipvs *ipvs = net_ipvs(net);
	166	+ struct ip_vs_dest_dst *dest_dst;
147	167	struct rtable rt; / Route to the other host */
148	168	struct rtable ort; / Original route */
149	169	struct iphdr *iph;
150	170	__be16 df;
151	171	int mtu;
152		- int local;
	172	+ int local, noref = 1;
153	173
154	174	if (dest) {
155		- spin_lock(&dest->dst_lock);
156		- rt = (struct rtable *) __ip_vs_dst_check(dest);
157		- if (!rt) {
	175	+ dest_dst = __ip_vs_dst_check(dest);
	176	+ if (likely(dest_dst))
	177	+ rt = (struct rtable *) dest_dst->dst_cache;
	178	+ else {
	179	+ dest_dst = ip_vs_dest_dst_alloc();
	180	+ spin_lock(&dest->dst_lock);
	181	+ if (!dest_dst) {
	182	+ __ip_vs_dst_set(dest, NULL, NULL, 0);
	183	+ spin_unlock(&dest->dst_lock);
	184	+ goto err_unreach;
	185	+ }
158	186	rt = do_output_route4(net, dest->addr.ip, rt_mode,
159		- &dest->dst_saddr.ip);
	187	+ &dest_dst->dst_saddr.ip);
160	188	if (!rt) {
	189	+ __ip_vs_dst_set(dest, NULL, NULL, 0);
161	190	spin_unlock(&dest->dst_lock);
	191	+ ip_vs_dest_dst_free(dest_dst);
162	192	goto err_unreach;
163	193	}
164		- __ip_vs_dst_set(dest, dst_clone(&rt->dst), 0);
	194	+ __ip_vs_dst_set(dest, dest_dst, &rt->dst, 0);
	195	+ spin_unlock(&dest->dst_lock);
165	196	IP_VS_DBG(10, "new dst %pI4, src %pI4, refcnt=%d\n",
166		- &dest->addr.ip, &dest->dst_saddr.ip,
	197	+ &dest->addr.ip, &dest_dst->dst_saddr.ip,
167	198	atomic_read(&rt->dst.__refcnt));
168	199	}
169	200	daddr = dest->addr.ip;
170	201	if (ret_saddr)
171		- *ret_saddr = dest->dst_saddr.ip;
172		- spin_unlock(&dest->dst_lock);
	202	+ *ret_saddr = dest_dst->dst_saddr.ip;
173	203	} else {
174	204	__be32 saddr = htonl(INADDR_ANY);
175	205
	206	+ noref = 0;
	207	+
176	208	/* For such unconfigured boxes avoid many route lookups
177	209	* for performance reasons because we do not remember saddr
178	210	*/
...	...	@@ -210,7 +242,8 @@
210	242	goto err_put;
211	243	}
212	244	/* skb to local stack, preserve old route */
213		- ip_rt_put(rt);
	245	+ if (!noref)
	246	+ ip_rt_put(rt);
214	247	return local;
215	248	}
216	249
217	250
...	...	@@ -240,12 +273,19 @@
240	273	}
241	274
242	275	skb_dst_drop(skb);
243		- skb_dst_set(skb, &rt->dst);
	276	+ if (noref) {
	277	+ if (!local)
	278	+ skb_dst_set_noref_force(skb, &rt->dst);
	279	+ else
	280	+ skb_dst_set(skb, dst_clone(&rt->dst));
	281	+ } else
	282	+ skb_dst_set(skb, &rt->dst);
244	283
245	284	return local;
246	285
247	286	err_put:
248		- ip_rt_put(rt);
	287	+ if (!noref)
	288	+ ip_rt_put(rt);
249	289	return -1;
250	290
251	291	err_unreach:
252	292
253	293
254	294
255	295
256	296
257	297
258	298
259	299
260	300
261	301
...	...	@@ -303,36 +343,48 @@
303	343	struct ip_vs_iphdr *ipvsh, int do_xfrm, int rt_mode)
304	344	{
305	345	struct net *net = dev_net(skb_dst(skb)->dev);
	346	+ struct ip_vs_dest_dst *dest_dst;
306	347	struct rt6_info rt; / Route to the other host */
307	348	struct rt6_info ort; / Original route */
308	349	struct dst_entry *dst;
309	350	int mtu;
310		- int local;
	351	+ int local, noref = 1;
311	352
312	353	if (dest) {
313		- spin_lock(&dest->dst_lock);
314		- rt = (struct rt6_info *)__ip_vs_dst_check(dest);
315		- if (!rt) {
	354	+ dest_dst = __ip_vs_dst_check(dest);
	355	+ if (likely(dest_dst))
	356	+ rt = (struct rt6_info *) dest_dst->dst_cache;
	357	+ else {
316	358	u32 cookie;
317	359
	360	+ dest_dst = ip_vs_dest_dst_alloc();
	361	+ spin_lock(&dest->dst_lock);
	362	+ if (!dest_dst) {
	363	+ __ip_vs_dst_set(dest, NULL, NULL, 0);
	364	+ spin_unlock(&dest->dst_lock);
	365	+ goto err_unreach;
	366	+ }
318	367	dst = __ip_vs_route_output_v6(net, &dest->addr.in6,
319		- &dest->dst_saddr.in6,
	368	+ &dest_dst->dst_saddr.in6,
320	369	do_xfrm);
321	370	if (!dst) {
	371	+ __ip_vs_dst_set(dest, NULL, NULL, 0);
322	372	spin_unlock(&dest->dst_lock);
	373	+ ip_vs_dest_dst_free(dest_dst);
323	374	goto err_unreach;
324	375	}
325	376	rt = (struct rt6_info *) dst;
326	377	cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
327		- __ip_vs_dst_set(dest, dst_clone(&rt->dst), cookie);
	378	+ __ip_vs_dst_set(dest, dest_dst, &rt->dst, cookie);
	379	+ spin_unlock(&dest->dst_lock);
328	380	IP_VS_DBG(10, "new dst %pI6, src %pI6, refcnt=%d\n",
329		- &dest->addr.in6, &dest->dst_saddr.in6,
	381	+ &dest->addr.in6, &dest_dst->dst_saddr.in6,
330	382	atomic_read(&rt->dst.__refcnt));
331	383	}
332	384	if (ret_saddr)
333		- *ret_saddr = dest->dst_saddr.in6;
334		- spin_unlock(&dest->dst_lock);
	385	+ *ret_saddr = dest_dst->dst_saddr.in6;
335	386	} else {
	387	+ noref = 0;
336	388	dst = __ip_vs_route_output_v6(net, daddr, ret_saddr, do_xfrm);
337	389	if (!dst)
338	390	goto err_unreach;
...	...	@@ -367,7 +419,8 @@
367	419	goto err_put;
368	420	}
369	421	/* skb to local stack, preserve old route */
370		- dst_release(&rt->dst);
	422	+ if (!noref)
	423	+ dst_release(&rt->dst);
371	424	return local;
372	425	}
373	426
374	427
...	...	@@ -399,12 +452,19 @@
399	452	}
400	453
401	454	skb_dst_drop(skb);
402		- skb_dst_set(skb, &rt->dst);
	455	+ if (noref) {
	456	+ if (!local)
	457	+ skb_dst_set_noref_force(skb, &rt->dst);
	458	+ else
	459	+ skb_dst_set(skb, dst_clone(&rt->dst));
	460	+ } else
	461	+ skb_dst_set(skb, &rt->dst);
403	462
404	463	return local;
405	464
406	465	err_put:
407		- dst_release(&rt->dst);
	466	+ if (!noref)
	467	+ dst_release(&rt->dst);
408	468	return -1;
409	469
410	470	err_unreach:
...	...	@@ -494,6 +554,7 @@
494	554
495	555	EnterFunction(10);
496	556
	557	+ rcu_read_lock();
497	558	if (__ip_vs_get_out_rt(skb, NULL, iph->daddr, IP_VS_RT_MODE_NON_LOCAL,
498	559	NULL) < 0)
499	560	goto tx_error;
500	561
...	...	@@ -504,12 +565,14 @@
504	565	skb->local_df = 1;
505	566
506	567	ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 0);
	568	+ rcu_read_unlock();
507	569
508	570	LeaveFunction(10);
509	571	return NF_STOLEN;
510	572
511	573	tx_error:
512	574	kfree_skb(skb);
	575	+ rcu_read_unlock();
513	576	LeaveFunction(10);
514	577	return NF_STOLEN;
515	578	}
...	...	@@ -521,6 +584,7 @@
521	584	{
522	585	EnterFunction(10);
523	586
	587	+ rcu_read_lock();
524	588	if (__ip_vs_get_out_rt_v6(skb, NULL, &ipvsh->daddr.in6, NULL,
525	589	ipvsh, 0, IP_VS_RT_MODE_NON_LOCAL) < 0)
526	590	goto tx_error;
527	591
...	...	@@ -529,12 +593,14 @@
529	593	skb->local_df = 1;
530	594
531	595	ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 0);
	596	+ rcu_read_unlock();
532	597
533	598	LeaveFunction(10);
534	599	return NF_STOLEN;
535	600
536	601	tx_error:
537	602	kfree_skb(skb);
	603	+ rcu_read_unlock();
538	604	LeaveFunction(10);
539	605	return NF_STOLEN;
540	606	}
...	...	@@ -553,6 +619,7 @@
553	619
554	620	EnterFunction(10);
555	621
	622	+ rcu_read_lock();
556	623	/* check if it is a connection of no-client-port */
557	624	if (unlikely(cp->flags & IP_VS_CONN_F_NO_CPORT)) {
558	625	__be16 _pt, *p;
559	626
...	...	@@ -620,12 +687,14 @@
620	687	skb->local_df = 1;
621	688
622	689	rc = ip_vs_nat_send_or_cont(NFPROTO_IPV4, skb, cp, local);
	690	+ rcu_read_unlock();
623	691
624	692	LeaveFunction(10);
625	693	return rc;
626	694
627	695	tx_error:
628	696	kfree_skb(skb);
	697	+ rcu_read_unlock();
629	698	LeaveFunction(10);
630	699	return NF_STOLEN;
631	700	}
...	...	@@ -640,6 +709,7 @@
640	709
641	710	EnterFunction(10);
642	711
	712	+ rcu_read_lock();
643	713	/* check if it is a connection of no-client-port */
644	714	if (unlikely(cp->flags & IP_VS_CONN_F_NO_CPORT && !ipvsh->fragoffs)) {
645	715	__be16 _pt, *p;
...	...	@@ -707,6 +777,7 @@
707	777	skb->local_df = 1;
708	778
709	779	rc = ip_vs_nat_send_or_cont(NFPROTO_IPV6, skb, cp, local);
	780	+ rcu_read_unlock();
710	781
711	782	LeaveFunction(10);
712	783	return rc;
...	...	@@ -714,6 +785,7 @@
714	785	tx_error:
715	786	LeaveFunction(10);
716	787	kfree_skb(skb);
	788	+ rcu_read_unlock();
717	789	return NF_STOLEN;
718	790	}
719	791	#endif
...	...	@@ -755,6 +827,7 @@
755	827
756	828	EnterFunction(10);
757	829
	830	+ rcu_read_lock();
758	831	local = __ip_vs_get_out_rt(skb, cp->dest, cp->daddr.ip,
759	832	IP_VS_RT_MODE_LOCAL \|
760	833	IP_VS_RT_MODE_NON_LOCAL \|
761	834
...	...	@@ -762,8 +835,10 @@
762	835	IP_VS_RT_MODE_TUNNEL, &saddr);
763	836	if (local < 0)
764	837	goto tx_error;
765		- if (local)
	838	+ if (local) {
	839	+ rcu_read_unlock();
766	840	return ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 1);
	841	+ }
767	842
768	843	rt = skb_rtable(skb);
769	844	tdev = rt->dst.dev;
...	...	@@ -818,6 +893,7 @@
818	893	ip_local_out(skb);
819	894	else if (ret == NF_DROP)
820	895	kfree_skb(skb);
	896	+ rcu_read_unlock();
821	897
822	898	LeaveFunction(10);
823	899
...	...	@@ -825,6 +901,7 @@
825	901
826	902	tx_error:
827	903	kfree_skb(skb);
	904	+ rcu_read_unlock();
828	905	LeaveFunction(10);
829	906	return NF_STOLEN;
830	907	}
...	...	@@ -844,6 +921,7 @@
844	921
845	922	EnterFunction(10);
846	923
	924	+ rcu_read_lock();
847	925	local = __ip_vs_get_out_rt_v6(skb, cp->dest, &cp->daddr.in6,
848	926	&saddr, ipvsh, 1,
849	927	IP_VS_RT_MODE_LOCAL \|
850	928
...	...	@@ -851,8 +929,10 @@
851	929	IP_VS_RT_MODE_TUNNEL);
852	930	if (local < 0)
853	931	goto tx_error;
854		- if (local)
	932	+ if (local) {
	933	+ rcu_read_unlock();
855	934	return ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 1);
	935	+ }
856	936
857	937	rt = (struct rt6_info *) skb_dst(skb);
858	938	tdev = rt->dst.dev;
...	...	@@ -901,6 +981,7 @@
901	981	ip6_local_out(skb);
902	982	else if (ret == NF_DROP)
903	983	kfree_skb(skb);
	984	+ rcu_read_unlock();
904	985
905	986	LeaveFunction(10);
906	987
...	...	@@ -908,6 +989,7 @@
908	989
909	990	tx_error:
910	991	kfree_skb(skb);
	992	+ rcu_read_unlock();
911	993	LeaveFunction(10);
912	994	return NF_STOLEN;
913	995	}
914	996
915	997
...	...	@@ -926,14 +1008,17 @@
926	1008
927	1009	EnterFunction(10);
928	1010
	1011	+ rcu_read_lock();
929	1012	local = __ip_vs_get_out_rt(skb, cp->dest, cp->daddr.ip,
930	1013	IP_VS_RT_MODE_LOCAL \|
931	1014	IP_VS_RT_MODE_NON_LOCAL \|
932	1015	IP_VS_RT_MODE_KNOWN_NH, NULL);
933	1016	if (local < 0)
934	1017	goto tx_error;
935		- if (local)
	1018	+ if (local) {
	1019	+ rcu_read_unlock();
936	1020	return ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 1);
	1021	+ }
937	1022
938	1023	ip_send_check(ip_hdr(skb));
939	1024
940	1025
...	...	@@ -941,12 +1026,14 @@
941	1026	skb->local_df = 1;
942	1027
943	1028	ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 0);
	1029	+ rcu_read_unlock();
944	1030
945	1031	LeaveFunction(10);
946	1032	return NF_STOLEN;
947	1033
948	1034	tx_error:
949	1035	kfree_skb(skb);
	1036	+ rcu_read_unlock();
950	1037	LeaveFunction(10);
951	1038	return NF_STOLEN;
952	1039	}
953	1040
954	1041
955	1042
956	1043
...	...	@@ -960,25 +1047,30 @@
960	1047
961	1048	EnterFunction(10);
962	1049
	1050	+ rcu_read_lock();
963	1051	local = __ip_vs_get_out_rt_v6(skb, cp->dest, &cp->daddr.in6, NULL,
964	1052	ipvsh, 0,
965	1053	IP_VS_RT_MODE_LOCAL \|
966	1054	IP_VS_RT_MODE_NON_LOCAL);
967	1055	if (local < 0)
968	1056	goto tx_error;
969		- if (local)
	1057	+ if (local) {
	1058	+ rcu_read_unlock();
970	1059	return ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 1);
	1060	+ }
971	1061
972	1062	/* Another hack: avoid icmp_send in ip_fragment */
973	1063	skb->local_df = 1;
974	1064
975	1065	ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 0);
	1066	+ rcu_read_unlock();
976	1067
977	1068	LeaveFunction(10);
978	1069	return NF_STOLEN;
979	1070
980	1071	tx_error:
981	1072	kfree_skb(skb);
	1073	+ rcu_read_unlock();
982	1074	LeaveFunction(10);
983	1075	return NF_STOLEN;
984	1076	}
...	...	@@ -1023,6 +1115,7 @@
1023	1115	rt_mode = (hooknum != NF_INET_FORWARD) ?
1024	1116	IP_VS_RT_MODE_LOCAL \| IP_VS_RT_MODE_NON_LOCAL \|
1025	1117	IP_VS_RT_MODE_RDR : IP_VS_RT_MODE_NON_LOCAL;
	1118	+ rcu_read_lock();
1026	1119	local = __ip_vs_get_out_rt(skb, cp->dest, cp->daddr.ip, rt_mode, NULL);
1027	1120	if (local < 0)
1028	1121	goto tx_error;
1029	1122
...	...	@@ -1067,10 +1160,12 @@
1067	1160	skb->local_df = 1;
1068	1161
1069	1162	rc = ip_vs_nat_send_or_cont(NFPROTO_IPV4, skb, cp, local);
	1163	+ rcu_read_unlock();
1070	1164	goto out;
1071	1165
1072	1166	tx_error:
1073		- dev_kfree_skb(skb);
	1167	+ kfree_skb(skb);
	1168	+ rcu_read_unlock();
1074	1169	rc = NF_STOLEN;
1075	1170	out:
1076	1171	LeaveFunction(10);
...	...	@@ -1111,6 +1206,7 @@
1111	1206	rt_mode = (hooknum != NF_INET_FORWARD) ?
1112	1207	IP_VS_RT_MODE_LOCAL \| IP_VS_RT_MODE_NON_LOCAL \|
1113	1208	IP_VS_RT_MODE_RDR : IP_VS_RT_MODE_NON_LOCAL;
	1209	+ rcu_read_lock();
1114	1210	local = __ip_vs_get_out_rt_v6(skb, cp->dest, &cp->daddr.in6, NULL,
1115	1211	ipvsh, 0, rt_mode);
1116	1212	if (local < 0)
1117	1213
...	...	@@ -1156,10 +1252,12 @@
1156	1252	skb->local_df = 1;
1157	1253
1158	1254	rc = ip_vs_nat_send_or_cont(NFPROTO_IPV6, skb, cp, local);
	1255	+ rcu_read_unlock();
1159	1256	goto out;
1160	1257
1161	1258	tx_error:
1162		- dev_kfree_skb(skb);
	1259	+ kfree_skb(skb);
	1260	+ rcu_read_unlock();
1163	1261	rc = NF_STOLEN;
1164	1262	out:
1165	1263	LeaveFunction(10);