ipvs: consolidate all dst checks on transmit in one place
authorJulian Anastasov <ja@ssi.bg>
Thu, 21 Mar 2013 09:58:05 +0000 (11:58 +0200)
committerPablo Neira Ayuso <pablo@netfilter.org>
Mon, 1 Apr 2013 22:23:41 +0000 (00:23 +0200)
Consolidate the PMTU checks, ICMP sending and
skb_dst modification in __ip_vs_get_out_rt and
__ip_vs_get_out_rt_v6. Now skb_dst is changed early
to simplify the transmitters.

Make sure update_pmtu is called only for local clients.

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>
net/netfilter/ipvs/ip_vs_xmit.c

index e6e2d3f..603eb8a 100644 (file)
@@ -51,6 +51,7 @@ enum {
                                      */
        IP_VS_RT_MODE_CONNECT   = 8, /* Always bind route to saddr */
        IP_VS_RT_MODE_KNOWN_NH  = 16,/* Route via remote addr */
+       IP_VS_RT_MODE_TUNNEL    = 32,/* Tunnel mode */
 };
 
 /*
@@ -137,13 +138,17 @@ retry:
 }
 
 /* Get route to destination or remote server */
-static struct rtable *
+static int
 __ip_vs_get_out_rt(struct sk_buff *skb, struct ip_vs_dest *dest,
                   __be32 daddr, int rt_mode, __be32 *ret_saddr)
 {
        struct net *net = dev_net(skb_dst(skb)->dev);
+       struct netns_ipvs *ipvs = net_ipvs(net);
        struct rtable *rt;                      /* Route to the other host */
        struct rtable *ort;                     /* Original route */
+       struct iphdr *iph;
+       __be16 df;
+       int mtu;
        int local;
 
        if (dest) {
@@ -154,7 +159,7 @@ __ip_vs_get_out_rt(struct sk_buff *skb, struct ip_vs_dest *dest,
                                              &dest->dst_saddr.ip);
                        if (!rt) {
                                spin_unlock(&dest->dst_lock);
-                               return NULL;
+                               goto err_unreach;
                        }
                        __ip_vs_dst_set(dest, dst_clone(&rt->dst), 0);
                        IP_VS_DBG(10, "new dst %pI4, src %pI4, refcnt=%d\n",
@@ -174,37 +179,78 @@ __ip_vs_get_out_rt(struct sk_buff *skb, struct ip_vs_dest *dest,
                rt_mode &= ~IP_VS_RT_MODE_CONNECT;
                rt = do_output_route4(net, daddr, rt_mode, &saddr);
                if (!rt)
-                       return NULL;
+                       goto err_unreach;
                if (ret_saddr)
                        *ret_saddr = saddr;
        }
 
-       local = rt->rt_flags & RTCF_LOCAL;
+       local = (rt->rt_flags & RTCF_LOCAL) ? 1 : 0;
        if (!((local ? IP_VS_RT_MODE_LOCAL : IP_VS_RT_MODE_NON_LOCAL) &
              rt_mode)) {
                IP_VS_DBG_RL("Stopping traffic to %s address, dest: %pI4\n",
                             (rt->rt_flags & RTCF_LOCAL) ?
                             "local":"non-local", &daddr);
-               ip_rt_put(rt);
-               return NULL;
+               goto err_put;
        }
-       if (local && !(rt_mode & IP_VS_RT_MODE_RDR) &&
-           !((ort = skb_rtable(skb)) && ort->rt_flags & RTCF_LOCAL)) {
-               IP_VS_DBG_RL("Redirect from non-local address %pI4 to local "
-                            "requires NAT method, dest: %pI4\n",
-                            &ip_hdr(skb)->daddr, &daddr);
+       iph = ip_hdr(skb);
+       if (likely(!local)) {
+               if (unlikely(ipv4_is_loopback(iph->saddr))) {
+                       IP_VS_DBG_RL("Stopping traffic from loopback address "
+                                    "%pI4 to non-local address, dest: %pI4\n",
+                                    &iph->saddr, &daddr);
+                       goto err_put;
+               }
+       } else {
+               ort = skb_rtable(skb);
+               if (!(rt_mode & IP_VS_RT_MODE_RDR) &&
+                   !(ort->rt_flags & RTCF_LOCAL)) {
+                       IP_VS_DBG_RL("Redirect from non-local address %pI4 to "
+                                    "local requires NAT method, dest: %pI4\n",
+                                    &iph->daddr, &daddr);
+                       goto err_put;
+               }
+               /* skb to local stack, preserve old route */
                ip_rt_put(rt);
-               return NULL;
+               return local;
        }
-       if (unlikely(!local && ipv4_is_loopback(ip_hdr(skb)->saddr))) {
-               IP_VS_DBG_RL("Stopping traffic from loopback address %pI4 "
-                            "to non-local address, dest: %pI4\n",
-                            &ip_hdr(skb)->saddr, &daddr);
-               ip_rt_put(rt);
-               return NULL;
+
+       if (likely(!(rt_mode & IP_VS_RT_MODE_TUNNEL))) {
+               mtu = dst_mtu(&rt->dst);
+               df = iph->frag_off & htons(IP_DF);
+       } else {
+               struct sock *sk = skb->sk;
+
+               mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
+               if (mtu < 68) {
+                       IP_VS_DBG_RL("%s(): mtu less than 68\n", __func__);
+                       goto err_put;
+               }
+               ort = skb_rtable(skb);
+               if (!skb->dev && sk && sk->sk_state != TCP_TIME_WAIT)
+                       ort->dst.ops->update_pmtu(&ort->dst, sk, NULL, mtu);
+               /* MTU check allowed? */
+               df = sysctl_pmtu_disc(ipvs) ? iph->frag_off & htons(IP_DF) : 0;
        }
 
-       return rt;
+       /* MTU checking */
+       if (unlikely(df && skb->len > mtu && !skb_is_gso(skb))) {
+               icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu));
+               IP_VS_DBG(1, "frag needed for %pI4\n", &iph->saddr);
+               goto err_put;
+       }
+
+       skb_dst_drop(skb);
+       skb_dst_set(skb, &rt->dst);
+
+       return local;
+
+err_put:
+       ip_rt_put(rt);
+       return -1;
+
+err_unreach:
+       dst_link_failure(skb);
+       return -1;
 }
 
 #ifdef CONFIG_IP_VS_IPV6
@@ -251,15 +297,16 @@ out_err:
 /*
  * Get route to destination or remote server
  */
-static struct rt6_info *
+static int
 __ip_vs_get_out_rt_v6(struct sk_buff *skb, struct ip_vs_dest *dest,
                      struct in6_addr *daddr, struct in6_addr *ret_saddr,
-                     int do_xfrm, int rt_mode)
+                     struct ip_vs_iphdr *ipvsh, int do_xfrm, int rt_mode)
 {
        struct net *net = dev_net(skb_dst(skb)->dev);
        struct rt6_info *rt;                    /* Route to the other host */
        struct rt6_info *ort;                   /* Original route */
        struct dst_entry *dst;
+       int mtu;
        int local;
 
        if (dest) {
@@ -273,7 +320,7 @@ __ip_vs_get_out_rt_v6(struct sk_buff *skb, struct ip_vs_dest *dest,
                                                      do_xfrm);
                        if (!dst) {
                                spin_unlock(&dest->dst_lock);
-                               return NULL;
+                               goto err_unreach;
                        }
                        rt = (struct rt6_info *) dst;
                        cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
@@ -288,7 +335,7 @@ __ip_vs_get_out_rt_v6(struct sk_buff *skb, struct ip_vs_dest *dest,
        } else {
                dst = __ip_vs_route_output_v6(net, daddr, ret_saddr, do_xfrm);
                if (!dst)
-                       return NULL;
+                       goto err_unreach;
                rt = (struct rt6_info *) dst;
        }
 
@@ -297,29 +344,72 @@ __ip_vs_get_out_rt_v6(struct sk_buff *skb, struct ip_vs_dest *dest,
              rt_mode)) {
                IP_VS_DBG_RL("Stopping traffic to %s address, dest: %pI6c\n",
                             local ? "local":"non-local", daddr);
-               dst_release(&rt->dst);
-               return NULL;
+               goto err_put;
        }
-       if (local && !(rt_mode & IP_VS_RT_MODE_RDR) &&
-           !((ort = (struct rt6_info *) skb_dst(skb)) &&
-             __ip_vs_is_local_route6(ort))) {
-               IP_VS_DBG_RL("Redirect from non-local address %pI6c to local "
-                            "requires NAT method, dest: %pI6c\n",
-                            &ipv6_hdr(skb)->daddr, daddr);
+       if (likely(!local)) {
+               if (unlikely((!skb->dev || skb->dev->flags & IFF_LOOPBACK) &&
+                            ipv6_addr_type(&ipv6_hdr(skb)->saddr) &
+                                           IPV6_ADDR_LOOPBACK)) {
+                       IP_VS_DBG_RL("Stopping traffic from loopback address "
+                                    "%pI6c to non-local address, "
+                                    "dest: %pI6c\n",
+                                    &ipv6_hdr(skb)->saddr, daddr);
+                       goto err_put;
+               }
+       } else {
+               ort = (struct rt6_info *) skb_dst(skb);
+               if (!(rt_mode & IP_VS_RT_MODE_RDR) &&
+                   !__ip_vs_is_local_route6(ort)) {
+                       IP_VS_DBG_RL("Redirect from non-local address %pI6c "
+                                    "to local requires NAT method, "
+                                    "dest: %pI6c\n",
+                                    &ipv6_hdr(skb)->daddr, daddr);
+                       goto err_put;
+               }
+               /* skb to local stack, preserve old route */
                dst_release(&rt->dst);
-               return NULL;
+               return local;
        }
-       if (unlikely(!local && (!skb->dev || skb->dev->flags & IFF_LOOPBACK) &&
-                    ipv6_addr_type(&ipv6_hdr(skb)->saddr) &
-                                   IPV6_ADDR_LOOPBACK)) {
-               IP_VS_DBG_RL("Stopping traffic from loopback address %pI6c "
-                            "to non-local address, dest: %pI6c\n",
-                            &ipv6_hdr(skb)->saddr, daddr);
-               dst_release(&rt->dst);
-               return NULL;
+
+       /* MTU checking */
+       if (likely(!(rt_mode & IP_VS_RT_MODE_TUNNEL)))
+               mtu = dst_mtu(&rt->dst);
+       else {
+               struct sock *sk = skb->sk;
+
+               mtu = dst_mtu(&rt->dst) - sizeof(struct ipv6hdr);
+               if (mtu < IPV6_MIN_MTU) {
+                       IP_VS_DBG_RL("%s(): mtu less than %d\n", __func__,
+                                    IPV6_MIN_MTU);
+                       goto err_put;
+               }
+               ort = (struct rt6_info *) skb_dst(skb);
+               if (!skb->dev && sk && sk->sk_state != TCP_TIME_WAIT)
+                       ort->dst.ops->update_pmtu(&ort->dst, sk, NULL, mtu);
        }
 
-       return rt;
+       if (unlikely(__mtu_check_toobig_v6(skb, mtu))) {
+               if (!skb->dev)
+                       skb->dev = net->loopback_dev;
+               /* only send ICMP too big on first fragment */
+               if (!ipvsh->fragoffs)
+                       icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
+               IP_VS_DBG(1, "frag needed for %pI6c\n", &ipv6_hdr(skb)->saddr);
+               goto err_put;
+       }
+
+       skb_dst_drop(skb);
+       skb_dst_set(skb, &rt->dst);
+
+       return local;
+
+err_put:
+       dst_release(&rt->dst);
+       return -1;
+
+err_unreach:
+       dst_link_failure(skb);
+       return -1;
 }
 #endif
 
@@ -400,32 +490,15 @@ int
 ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
                  struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh)
 {
-       struct rtable *rt;                      /* Route to the other host */
        struct iphdr  *iph = ip_hdr(skb);
-       int    mtu;
 
        EnterFunction(10);
 
-       rt = __ip_vs_get_out_rt(skb, NULL, iph->daddr, IP_VS_RT_MODE_NON_LOCAL,
-                               NULL);
-       if (!rt)
-               goto tx_error_icmp;
-
-       /* MTU checking */
-       mtu = dst_mtu(&rt->dst);
-       if ((skb->len > mtu) && (iph->frag_off & htons(IP_DF)) &&
-           !skb_is_gso(skb)) {
-               ip_rt_put(rt);
-               icmp_send(skb, ICMP_DEST_UNREACH,ICMP_FRAG_NEEDED, htonl(mtu));
-               IP_VS_DBG_RL("%s(): frag needed\n", __func__);
+       if (__ip_vs_get_out_rt(skb, NULL, iph->daddr, IP_VS_RT_MODE_NON_LOCAL,
+                              NULL) < 0)
                goto tx_error;
-       }
 
-       ip_send_check(ip_hdr(skb));
-
-       /* drop old route */
-       skb_dst_drop(skb);
-       skb_dst_set(skb, &rt->dst);
+       ip_send_check(iph);
 
        /* Another hack: avoid icmp_send in ip_fragment */
        skb->local_df = 1;
@@ -435,8 +508,6 @@ ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
        LeaveFunction(10);
        return NF_STOLEN;
 
- tx_error_icmp:
-       dst_link_failure(skb);
  tx_error:
        kfree_skb(skb);
        LeaveFunction(10);
@@ -446,37 +517,13 @@ ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
 #ifdef CONFIG_IP_VS_IPV6
 int
 ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
-                    struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph)
+                    struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh)
 {
-       struct rt6_info *rt;                    /* Route to the other host */
-       int    mtu;
-
        EnterFunction(10);
 
-       rt = __ip_vs_get_out_rt_v6(skb, NULL, &iph->daddr.in6, NULL, 0,
-                                  IP_VS_RT_MODE_NON_LOCAL);
-       if (!rt)
-               goto tx_error_icmp;
-
-       /* MTU checking */
-       mtu = dst_mtu(&rt->dst);
-       if (__mtu_check_toobig_v6(skb, mtu)) {
-               if (!skb->dev) {
-                       struct net *net = dev_net(skb_dst(skb)->dev);
-
-                       skb->dev = net->loopback_dev;
-               }
-               /* only send ICMP too big on first fragment */
-               if (!iph->fragoffs)
-                       icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
-               dst_release(&rt->dst);
-               IP_VS_DBG_RL("%s(): frag needed\n", __func__);
+       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;
-       }
-
-       /* drop old route */
-       skb_dst_drop(skb);
-       skb_dst_set(skb, &rt->dst);
 
        /* Another hack: avoid icmp_send in ip_fragment */
        skb->local_df = 1;
@@ -486,8 +533,6 @@ ip_vs_bypass_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
        LeaveFunction(10);
        return NF_STOLEN;
 
- tx_error_icmp:
-       dst_link_failure(skb);
  tx_error:
        kfree_skb(skb);
        LeaveFunction(10);
@@ -504,28 +549,29 @@ ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
               struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh)
 {
        struct rtable *rt;              /* Route to the other host */
-       int mtu;
-       struct iphdr *iph = ip_hdr(skb);
-       int local, rc;
+       int local, rc, was_input;
 
        EnterFunction(10);
 
        /* check if it is a connection of no-client-port */
        if (unlikely(cp->flags & IP_VS_CONN_F_NO_CPORT)) {
                __be16 _pt, *p;
-               p = skb_header_pointer(skb, iph->ihl*4, sizeof(_pt), &_pt);
+
+               p = skb_header_pointer(skb, ipvsh->len, sizeof(_pt), &_pt);
                if (p == NULL)
                        goto tx_error;
                ip_vs_conn_fill_cport(cp, *p);
                IP_VS_DBG(10, "filled cport=%d\n", ntohs(*p));
        }
 
-       if (!(rt = __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_RDR, NULL)))
-               goto tx_error_icmp;
-       local = rt->rt_flags & RTCF_LOCAL;
+       was_input = rt_is_input_route(skb_rtable(skb));
+       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_RDR, NULL);
+       if (local < 0)
+               goto tx_error;
+       rt = skb_rtable(skb);
        /*
         * Avoid duplicate tuple in reply direction for NAT traffic
         * to local address when connection is sync-ed
@@ -539,49 +585,31 @@ ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
                        IP_VS_DBG_RL_PKT(10, AF_INET, pp, skb, 0,
                                         "ip_vs_nat_xmit(): "
                                         "stopping DNAT to local address");
-                       goto tx_error_put;
+                       goto tx_error;
                }
        }
 #endif
 
        /* From world but DNAT to loopback address? */
-       if (local && ipv4_is_loopback(cp->daddr.ip) &&
-           rt_is_input_route(skb_rtable(skb))) {
+       if (local && ipv4_is_loopback(cp->daddr.ip) && was_input) {
                IP_VS_DBG_RL_PKT(1, AF_INET, pp, skb, 0, "ip_vs_nat_xmit(): "
                                 "stopping DNAT to loopback address");
-               goto tx_error_put;
-       }
-
-       /* MTU checking */
-       mtu = dst_mtu(&rt->dst);
-       if ((skb->len > mtu) && (iph->frag_off & htons(IP_DF)) &&
-           !skb_is_gso(skb)) {
-               icmp_send(skb, ICMP_DEST_UNREACH,ICMP_FRAG_NEEDED, htonl(mtu));
-               IP_VS_DBG_RL_PKT(0, AF_INET, pp, skb, 0,
-                                "ip_vs_nat_xmit(): frag needed for");
-               goto tx_error_put;
+               goto tx_error;
        }
 
        /* copy-on-write the packet before mangling it */
        if (!skb_make_writable(skb, sizeof(struct iphdr)))
-               goto tx_error_put;
+               goto tx_error;
 
        if (skb_cow(skb, rt->dst.dev->hard_header_len))
-               goto tx_error_put;
+               goto tx_error;
 
        /* mangle the packet */
        if (pp->dnat_handler && !pp->dnat_handler(skb, pp, cp, ipvsh))
-               goto tx_error_put;
+               goto tx_error;
        ip_hdr(skb)->daddr = cp->daddr.ip;
        ip_send_check(ip_hdr(skb));
 
-       if (!local) {
-               /* drop old route */
-               skb_dst_drop(skb);
-               skb_dst_set(skb, &rt->dst);
-       } else
-               ip_rt_put(rt);
-
        IP_VS_DBG_PKT(10, AF_INET, pp, skb, 0, "After DNAT");
 
        /* FIXME: when application helper enlarges the packet and the length
@@ -596,44 +624,40 @@ ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
        LeaveFunction(10);
        return rc;
 
-  tx_error_icmp:
-       dst_link_failure(skb);
   tx_error:
        kfree_skb(skb);
        LeaveFunction(10);
        return NF_STOLEN;
-  tx_error_put:
-       ip_rt_put(rt);
-       goto tx_error;
 }
 
 #ifdef CONFIG_IP_VS_IPV6
 int
 ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
-                 struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph)
+                 struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh)
 {
        struct rt6_info *rt;            /* Route to the other host */
-       int mtu;
        int local, rc;
 
        EnterFunction(10);
 
        /* check if it is a connection of no-client-port */
-       if (unlikely(cp->flags & IP_VS_CONN_F_NO_CPORT && !iph->fragoffs)) {
+       if (unlikely(cp->flags & IP_VS_CONN_F_NO_CPORT && !ipvsh->fragoffs)) {
                __be16 _pt, *p;
-               p = skb_header_pointer(skb, iph->len, sizeof(_pt), &_pt);
+               p = skb_header_pointer(skb, ipvsh->len, sizeof(_pt), &_pt);
                if (p == NULL)
                        goto tx_error;
                ip_vs_conn_fill_cport(cp, *p);
                IP_VS_DBG(10, "filled cport=%d\n", ntohs(*p));
        }
 
-       if (!(rt = __ip_vs_get_out_rt_v6(skb, cp->dest, &cp->daddr.in6, NULL,
-                                        0, (IP_VS_RT_MODE_LOCAL |
-                                            IP_VS_RT_MODE_NON_LOCAL |
-                                            IP_VS_RT_MODE_RDR))))
-               goto tx_error_icmp;
-       local = __ip_vs_is_local_route6(rt);
+       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 |
+                                     IP_VS_RT_MODE_RDR);
+       if (local < 0)
+               goto tx_error;
+       rt = (struct rt6_info *) skb_dst(skb);
        /*
         * Avoid duplicate tuple in reply direction for NAT traffic
         * to local address when connection is sync-ed
@@ -647,7 +671,7 @@ ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
                        IP_VS_DBG_RL_PKT(10, AF_INET6, pp, skb, 0,
                                         "ip_vs_nat_xmit_v6(): "
                                         "stopping DNAT to local address");
-                       goto tx_error_put;
+                       goto tx_error;
                }
        }
 #endif
@@ -658,45 +682,21 @@ ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
                IP_VS_DBG_RL_PKT(1, AF_INET6, pp, skb, 0,
                                 "ip_vs_nat_xmit_v6(): "
                                 "stopping DNAT to loopback address");
-               goto tx_error_put;
-       }
-
-       /* MTU checking */
-       mtu = dst_mtu(&rt->dst);
-       if (__mtu_check_toobig_v6(skb, mtu)) {
-               if (!skb->dev) {
-                       struct net *net = dev_net(skb_dst(skb)->dev);
-
-                       skb->dev = net->loopback_dev;
-               }
-               /* only send ICMP too big on first fragment */
-               if (!iph->fragoffs)
-                       icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
-               IP_VS_DBG_RL_PKT(0, AF_INET6, pp, skb, 0,
-                                "ip_vs_nat_xmit_v6(): frag needed for");
-               goto tx_error_put;
+               goto tx_error;
        }
 
        /* copy-on-write the packet before mangling it */
        if (!skb_make_writable(skb, sizeof(struct ipv6hdr)))
-               goto tx_error_put;
+               goto tx_error;
 
        if (skb_cow(skb, rt->dst.dev->hard_header_len))
-               goto tx_error_put;
+               goto tx_error;
 
        /* mangle the packet */
-       if (pp->dnat_handler && !pp->dnat_handler(skb, pp, cp, iph))
+       if (pp->dnat_handler && !pp->dnat_handler(skb, pp, cp, ipvsh))
                goto tx_error;
        ipv6_hdr(skb)->daddr = cp->daddr.in6;
 
-       if (!local || !skb->dev) {
-               skb_dst_drop(skb);
-               skb_dst_set(skb, &rt->dst);
-       } else {
-               /* destined to loopback, do we need to change route? */
-               dst_release(&rt->dst);
-       }
-
        IP_VS_DBG_PKT(10, AF_INET6, pp, skb, 0, "After DNAT");
 
        /* FIXME: when application helper enlarges the packet and the length
@@ -711,15 +711,10 @@ ip_vs_nat_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
        LeaveFunction(10);
        return rc;
 
-tx_error_icmp:
-       dst_link_failure(skb);
 tx_error:
        LeaveFunction(10);
        kfree_skb(skb);
        return NF_STOLEN;
-tx_error_put:
-       dst_release(&rt->dst);
-       goto tx_error;
 }
 #endif
 
@@ -756,40 +751,26 @@ ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
        __be16 df;
        struct iphdr  *iph;                     /* Our new IP header */
        unsigned int max_headroom;              /* The extra header space needed */
-       int    mtu;
-       int ret;
+       int ret, local;
 
        EnterFunction(10);
 
-       if (!(rt = __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_CONNECT, &saddr)))
-               goto tx_error_icmp;
-       if (rt->rt_flags & RTCF_LOCAL) {
-               ip_rt_put(rt);
+       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_CONNECT |
+                                  IP_VS_RT_MODE_TUNNEL, &saddr);
+       if (local < 0)
+               goto tx_error;
+       if (local)
                return ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 1);
-       }
 
+       rt = skb_rtable(skb);
        tdev = rt->dst.dev;
 
-       mtu = dst_mtu(&rt->dst) - sizeof(struct iphdr);
-       if (mtu < 68) {
-               IP_VS_DBG_RL("%s(): mtu less than 68\n", __func__);
-               goto tx_error_put;
-       }
-       if (rt_is_output_route(skb_rtable(skb)))
-               skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
-
        /* Copy DF, reset fragment offset and MF */
        df = sysctl_pmtu_disc(ipvs) ? old_iph->frag_off & htons(IP_DF) : 0;
 
-       if (df && mtu < ntohs(old_iph->tot_len) && !skb_is_gso(skb)) {
-               icmp_send(skb, ICMP_DEST_UNREACH,ICMP_FRAG_NEEDED, htonl(mtu));
-               IP_VS_DBG_RL("%s(): frag needed\n", __func__);
-               goto tx_error_put;
-       }
-
        /*
         * Okay, now see if we can stuff it in the buffer as-is.
         */
@@ -798,12 +779,9 @@ ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
        if (skb_headroom(skb) < max_headroom || skb_cloned(skb)) {
                struct sk_buff *new_skb =
                        skb_realloc_headroom(skb, max_headroom);
-               if (!new_skb) {
-                       ip_rt_put(rt);
-                       kfree_skb(skb);
-                       IP_VS_ERR_RL("%s(): no memory\n", __func__);
-                       return NF_STOLEN;
-               }
+
+               if (!new_skb)
+                       goto tx_error;
                consume_skb(skb);
                skb = new_skb;
                old_iph = ip_hdr(skb);
@@ -818,10 +796,6 @@ ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
        skb_reset_network_header(skb);
        memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 
-       /* drop old route */
-       skb_dst_drop(skb);
-       skb_dst_set(skb, &rt->dst);
-
        /*
         *      Push down and install the IPIP header.
         */
@@ -849,15 +823,10 @@ ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
 
        return NF_STOLEN;
 
-  tx_error_icmp:
-       dst_link_failure(skb);
   tx_error:
        kfree_skb(skb);
        LeaveFunction(10);
        return NF_STOLEN;
-tx_error_put:
-       ip_rt_put(rt);
-       goto tx_error;
 }
 
 #ifdef CONFIG_IP_VS_IPV6
@@ -871,45 +840,23 @@ ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
        struct ipv6hdr  *old_iph = ipv6_hdr(skb);
        struct ipv6hdr  *iph;           /* Our new IP header */
        unsigned int max_headroom;      /* The extra header space needed */
-       int    mtu;
-       int ret;
+       int ret, local;
 
        EnterFunction(10);
 
-       if (!(rt = __ip_vs_get_out_rt_v6(skb, cp->dest, &cp->daddr.in6,
-                                        &saddr, 1, (IP_VS_RT_MODE_LOCAL |
-                                                    IP_VS_RT_MODE_NON_LOCAL))))
-               goto tx_error_icmp;
-       if (__ip_vs_is_local_route6(rt)) {
-               dst_release(&rt->dst);
+       local = __ip_vs_get_out_rt_v6(skb, cp->dest, &cp->daddr.in6,
+                                     &saddr, ipvsh, 1,
+                                     IP_VS_RT_MODE_LOCAL |
+                                     IP_VS_RT_MODE_NON_LOCAL |
+                                     IP_VS_RT_MODE_TUNNEL);
+       if (local < 0)
+               goto tx_error;
+       if (local)
                return ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 1);
-       }
 
+       rt = (struct rt6_info *) skb_dst(skb);
        tdev = rt->dst.dev;
 
-       mtu = dst_mtu(&rt->dst) - sizeof(struct ipv6hdr);
-       if (mtu < IPV6_MIN_MTU) {
-               IP_VS_DBG_RL("%s(): mtu less than %d\n", __func__,
-                            IPV6_MIN_MTU);
-               goto tx_error_put;
-       }
-       if (skb_dst(skb))
-               skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
-
-       /* MTU checking: Notice that 'mtu' have been adjusted before hand */
-       if (__mtu_check_toobig_v6(skb, mtu)) {
-               if (!skb->dev) {
-                       struct net *net = dev_net(skb_dst(skb)->dev);
-
-                       skb->dev = net->loopback_dev;
-               }
-               /* only send ICMP too big on first fragment */
-               if (!ipvsh->fragoffs)
-                       icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
-               IP_VS_DBG_RL("%s(): frag needed\n", __func__);
-               goto tx_error_put;
-       }
-
        /*
         * Okay, now see if we can stuff it in the buffer as-is.
         */
@@ -918,12 +865,9 @@ ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
        if (skb_headroom(skb) < max_headroom || skb_cloned(skb)) {
                struct sk_buff *new_skb =
                        skb_realloc_headroom(skb, max_headroom);
-               if (!new_skb) {
-                       dst_release(&rt->dst);
-                       kfree_skb(skb);
-                       IP_VS_ERR_RL("%s(): no memory\n", __func__);
-                       return NF_STOLEN;
-               }
+
+               if (!new_skb)
+                       goto tx_error;
                consume_skb(skb);
                skb = new_skb;
                old_iph = ipv6_hdr(skb);
@@ -935,10 +879,6 @@ ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
        skb_reset_network_header(skb);
        memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 
-       /* drop old route */
-       skb_dst_drop(skb);
-       skb_dst_set(skb, &rt->dst);
-
        /*
         *      Push down and install the IPIP header.
         */
@@ -966,15 +906,10 @@ ip_vs_tunnel_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
 
        return NF_STOLEN;
 
-tx_error_icmp:
-       dst_link_failure(skb);
 tx_error:
        kfree_skb(skb);
        LeaveFunction(10);
        return NF_STOLEN;
-tx_error_put:
-       dst_release(&rt->dst);
-       goto tx_error;
 }
 #endif
 
@@ -987,38 +922,21 @@ int
 ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
              struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh)
 {
-       struct rtable *rt;                      /* Route to the other host */
-       struct iphdr  *iph = ip_hdr(skb);
-       int    mtu;
+       int local;
 
        EnterFunction(10);
 
-       if (!(rt = __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)))
-               goto tx_error_icmp;
-       if (rt->rt_flags & RTCF_LOCAL) {
-               ip_rt_put(rt);
-               return ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 1);
-       }
-
-       /* MTU checking */
-       mtu = dst_mtu(&rt->dst);
-       if ((iph->frag_off & htons(IP_DF)) && skb->len > mtu &&
-           !skb_is_gso(skb)) {
-               icmp_send(skb, ICMP_DEST_UNREACH,ICMP_FRAG_NEEDED, htonl(mtu));
-               ip_rt_put(rt);
-               IP_VS_DBG_RL("%s(): frag needed\n", __func__);
+       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)
+               return ip_vs_send_or_cont(NFPROTO_IPV4, skb, cp, 1);
 
        ip_send_check(ip_hdr(skb));
 
-       /* drop old route */
-       skb_dst_drop(skb);
-       skb_dst_set(skb, &rt->dst);
-
        /* Another hack: avoid icmp_send in ip_fragment */
        skb->local_df = 1;
 
@@ -1027,8 +945,6 @@ ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
        LeaveFunction(10);
        return NF_STOLEN;
 
-  tx_error_icmp:
-       dst_link_failure(skb);
   tx_error:
        kfree_skb(skb);
        LeaveFunction(10);
@@ -1038,41 +954,20 @@ ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
 #ifdef CONFIG_IP_VS_IPV6
 int
 ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
-                struct ip_vs_protocol *pp, struct ip_vs_iphdr *iph)
+                struct ip_vs_protocol *pp, struct ip_vs_iphdr *ipvsh)
 {
-       struct rt6_info *rt;                    /* Route to the other host */
-       int    mtu;
+       int local;
 
        EnterFunction(10);
 
-       if (!(rt = __ip_vs_get_out_rt_v6(skb, cp->dest, &cp->daddr.in6, NULL,
-                                        0, (IP_VS_RT_MODE_LOCAL |
-                                            IP_VS_RT_MODE_NON_LOCAL))))
-               goto tx_error_icmp;
-       if (__ip_vs_is_local_route6(rt)) {
-               dst_release(&rt->dst);
-               return ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 1);
-       }
-
-       /* MTU checking */
-       mtu = dst_mtu(&rt->dst);
-       if (__mtu_check_toobig_v6(skb, mtu)) {
-               if (!skb->dev) {
-                       struct net *net = dev_net(skb_dst(skb)->dev);
-
-                       skb->dev = net->loopback_dev;
-               }
-               /* only send ICMP too big on first fragment */
-               if (!iph->fragoffs)
-                       icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
-               dst_release(&rt->dst);
-               IP_VS_DBG_RL("%s(): frag needed\n", __func__);
+       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;
-       }
-
-       /* drop old route */
-       skb_dst_drop(skb);
-       skb_dst_set(skb, &rt->dst);
+       if (local)
+               return ip_vs_send_or_cont(NFPROTO_IPV6, skb, cp, 1);
 
        /* Another hack: avoid icmp_send in ip_fragment */
        skb->local_df = 1;
@@ -1082,8 +977,6 @@ ip_vs_dr_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
        LeaveFunction(10);
        return NF_STOLEN;
 
-tx_error_icmp:
-       dst_link_failure(skb);
 tx_error:
        kfree_skb(skb);
        LeaveFunction(10);
@@ -1102,10 +995,9 @@ ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
                struct ip_vs_iphdr *iph)
 {
        struct rtable   *rt;    /* Route to the other host */
-       int mtu;
        int rc;
        int local;
-       int rt_mode;
+       int rt_mode, was_input;
 
        EnterFunction(10);
 
@@ -1125,15 +1017,16 @@ ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
        /*
         * mangle and send the packet here (only for VS/NAT)
         */
+       was_input = rt_is_input_route(skb_rtable(skb));
 
        /* LOCALNODE from FORWARD hook is not supported */
        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;
-       if (!(rt = __ip_vs_get_out_rt(skb, cp->dest, cp->daddr.ip,
-                                     rt_mode, NULL)))
-               goto tx_error_icmp;
-       local = rt->rt_flags & RTCF_LOCAL;
+       local = __ip_vs_get_out_rt(skb, cp->dest, cp->daddr.ip, rt_mode, NULL);
+       if (local < 0)
+               goto tx_error;
+       rt = skb_rtable(skb);
 
        /*
         * Avoid duplicate tuple in reply direction for NAT traffic
@@ -1148,71 +1041,49 @@ ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
                        IP_VS_DBG(10, "%s(): "
                                  "stopping DNAT to local address %pI4\n",
                                  __func__, &cp->daddr.ip);
-                       goto tx_error_put;
+                       goto tx_error;
                }
        }
 #endif
 
        /* From world but DNAT to loopback address? */
-       if (local && ipv4_is_loopback(cp->daddr.ip) &&
-           rt_is_input_route(skb_rtable(skb))) {
+       if (local && ipv4_is_loopback(cp->daddr.ip) && was_input) {
                IP_VS_DBG(1, "%s(): "
                          "stopping DNAT to loopback %pI4\n",
                          __func__, &cp->daddr.ip);
-               goto tx_error_put;
-       }
-
-       /* MTU checking */
-       mtu = dst_mtu(&rt->dst);
-       if ((skb->len > mtu) && (ip_hdr(skb)->frag_off & htons(IP_DF)) &&
-           !skb_is_gso(skb)) {
-               icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu));
-               IP_VS_DBG_RL("%s(): frag needed\n", __func__);
-               goto tx_error_put;
+               goto tx_error;
        }
 
        /* copy-on-write the packet before mangling it */
        if (!skb_make_writable(skb, offset))
-               goto tx_error_put;
+               goto tx_error;
 
        if (skb_cow(skb, rt->dst.dev->hard_header_len))
-               goto tx_error_put;
+               goto tx_error;
 
        ip_vs_nat_icmp(skb, pp, cp, 0);
 
-       if (!local) {
-               skb_dst_drop(skb);
-               skb_dst_set(skb, &rt->dst);
-       } else
-               ip_rt_put(rt);
-
        /* Another hack: avoid icmp_send in ip_fragment */
        skb->local_df = 1;
 
        rc = ip_vs_nat_send_or_cont(NFPROTO_IPV4, skb, cp, local);
        goto out;
 
-  tx_error_icmp:
-       dst_link_failure(skb);
   tx_error:
        dev_kfree_skb(skb);
        rc = NF_STOLEN;
   out:
        LeaveFunction(10);
        return rc;
-  tx_error_put:
-       ip_rt_put(rt);
-       goto tx_error;
 }
 
 #ifdef CONFIG_IP_VS_IPV6
 int
 ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
                struct ip_vs_protocol *pp, int offset, unsigned int hooknum,
-               struct ip_vs_iphdr *iph)
+               struct ip_vs_iphdr *ipvsh)
 {
        struct rt6_info *rt;    /* Route to the other host */
-       int mtu;
        int rc;
        int local;
        int rt_mode;
@@ -1224,7 +1095,7 @@ ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
           translate address/port back */
        if (IP_VS_FWD_METHOD(cp) != IP_VS_CONN_F_MASQ) {
                if (cp->packet_xmit)
-                       rc = cp->packet_xmit(skb, cp, pp, iph);
+                       rc = cp->packet_xmit(skb, cp, pp, ipvsh);
                else
                        rc = NF_ACCEPT;
                /* do not touch skb anymore */
@@ -1240,11 +1111,11 @@ ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
        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;
-       if (!(rt = __ip_vs_get_out_rt_v6(skb, cp->dest, &cp->daddr.in6, NULL,
-                                        0, rt_mode)))
-               goto tx_error_icmp;
-
-       local = __ip_vs_is_local_route6(rt);
+       local = __ip_vs_get_out_rt_v6(skb, cp->dest, &cp->daddr.in6, NULL,
+                                     ipvsh, 0, rt_mode);
+       if (local < 0)
+               goto tx_error;
+       rt = (struct rt6_info *) skb_dst(skb);
        /*
         * Avoid duplicate tuple in reply direction for NAT traffic
         * to local address when connection is sync-ed
@@ -1258,7 +1129,7 @@ ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
                        IP_VS_DBG(10, "%s(): "
                                  "stopping DNAT to local address %pI6\n",
                                  __func__, &cp->daddr.in6);
-                       goto tx_error_put;
+                       goto tx_error;
                }
        }
 #endif
@@ -1269,57 +1140,29 @@ ip_vs_icmp_xmit_v6(struct sk_buff *skb, struct ip_vs_conn *cp,
                IP_VS_DBG(1, "%s(): "
                          "stopping DNAT to loopback %pI6\n",
                          __func__, &cp->daddr.in6);
-               goto tx_error_put;
-       }
-
-       /* MTU checking */
-       mtu = dst_mtu(&rt->dst);
-       if (__mtu_check_toobig_v6(skb, mtu)) {
-               if (!skb->dev) {
-                       struct net *net = dev_net(skb_dst(skb)->dev);
-
-                       skb->dev = net->loopback_dev;
-               }
-               /* only send ICMP too big on first fragment */
-               if (!iph->fragoffs)
-                       icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
-               IP_VS_DBG_RL("%s(): frag needed\n", __func__);
-               goto tx_error_put;
+               goto tx_error;
        }
 
        /* copy-on-write the packet before mangling it */
        if (!skb_make_writable(skb, offset))
-               goto tx_error_put;
+               goto tx_error;
 
        if (skb_cow(skb, rt->dst.dev->hard_header_len))
-               goto tx_error_put;
+               goto tx_error;
 
        ip_vs_nat_icmp_v6(skb, pp, cp, 0);
 
-       if (!local || !skb->dev) {
-               skb_dst_drop(skb);
-               skb_dst_set(skb, &rt->dst);
-       } else {
-               /* destined to loopback, do we need to change route? */
-               dst_release(&rt->dst);
-       }
-
        /* Another hack: avoid icmp_send in ip_fragment */
        skb->local_df = 1;
 
        rc = ip_vs_nat_send_or_cont(NFPROTO_IPV6, skb, cp, local);
        goto out;
 
-tx_error_icmp:
-       dst_link_failure(skb);
 tx_error:
        dev_kfree_skb(skb);
        rc = NF_STOLEN;
 out:
        LeaveFunction(10);
        return rc;
-tx_error_put:
-       dst_release(&rt->dst);
-       goto tx_error;
 }
 #endif