tunnels: fix netns vs proto registration ordering
[safe/jmp/linux-2.6] / net / ipv6 / ip6_tunnel.c
index 78f4388..9b02492 100644 (file)
@@ -6,8 +6,6 @@
  *     Ville Nuorvala          <vnuorval@tcs.hut.fi>
  *     Yasuyuki Kozakai        <kozakai@linux-ipv6.org>
  *
- *     $Id$
- *
  *      Based on:
  *      linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
  *
@@ -52,6 +50,8 @@
 #include <net/xfrm.h>
 #include <net/dsfield.h>
 #include <net/inet_ecn.h>
+#include <net/net_namespace.h>
+#include <net/netns/generic.h>
 
 MODULE_AUTHOR("Ville Nuorvala");
 MODULE_DESCRIPTION("IPv6 tunneling device");
@@ -60,7 +60,7 @@ MODULE_LICENSE("GPL");
 #define IPV6_TLV_TEL_DST_SIZE 8
 
 #ifdef IP6_TNL_DEBUG
-#define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __FUNCTION__)
+#define IP6_TNL_TRACE(x...) printk(KERN_DEBUG "%s:" x "\n", __func__)
 #else
 #define IP6_TNL_TRACE(x...) do {;} while(0)
 #endif
@@ -74,21 +74,23 @@ MODULE_LICENSE("GPL");
                     (addr)->s6_addr32[2] ^ (addr)->s6_addr32[3]) & \
                    (HASH_SIZE - 1))
 
-static int ip6_fb_tnl_dev_init(struct net_device *dev);
-static int ip6_tnl_dev_init(struct net_device *dev);
+static void ip6_tnl_dev_init(struct net_device *dev);
 static void ip6_tnl_dev_setup(struct net_device *dev);
 
-/* the IPv6 tunnel fallback device */
-static struct net_device *ip6_fb_tnl_dev;
-
-
-/* lists for storing tunnels in use */
-static struct ip6_tnl *tnls_r_l[HASH_SIZE];
-static struct ip6_tnl *tnls_wc[1];
-static struct ip6_tnl **tnls[2] = { tnls_wc, tnls_r_l };
+static int ip6_tnl_net_id __read_mostly;
+struct ip6_tnl_net {
+       /* the IPv6 tunnel fallback device */
+       struct net_device *fb_tnl_dev;
+       /* lists for storing tunnels in use */
+       struct ip6_tnl *tnls_r_l[HASH_SIZE];
+       struct ip6_tnl *tnls_wc[1];
+       struct ip6_tnl **tnls[2];
+};
 
-/* lock for the tunnel lists */
-static DEFINE_RWLOCK(ip6_tnl_lock);
+/*
+ * Locking : hash tables are protected by RCU and a spinlock
+ */
+static DEFINE_SPINLOCK(ip6_tnl_lock);
 
 static inline struct dst_entry *ip6_tnl_dst_check(struct ip6_tnl *t)
 {
@@ -129,20 +131,25 @@ static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
  *   else %NULL
  **/
 
+#define for_each_ip6_tunnel_rcu(start) \
+       for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
+
 static struct ip6_tnl *
-ip6_tnl_lookup(struct in6_addr *remote, struct in6_addr *local)
+ip6_tnl_lookup(struct net *net, struct in6_addr *remote, struct in6_addr *local)
 {
        unsigned h0 = HASH(remote);
        unsigned h1 = HASH(local);
        struct ip6_tnl *t;
+       struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 
-       for (t = tnls_r_l[h0 ^ h1]; t; t = t->next) {
+       for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[h0 ^ h1]) {
                if (ipv6_addr_equal(local, &t->parms.laddr) &&
                    ipv6_addr_equal(remote, &t->parms.raddr) &&
                    (t->dev->flags & IFF_UP))
                        return t;
        }
-       if ((t = tnls_wc[0]) != NULL && (t->dev->flags & IFF_UP))
+       t = rcu_dereference(ip6n->tnls_wc[0]);
+       if (t && (t->dev->flags & IFF_UP))
                return t;
 
        return NULL;
@@ -160,7 +167,7 @@ ip6_tnl_lookup(struct in6_addr *remote, struct in6_addr *local)
  **/
 
 static struct ip6_tnl **
-ip6_tnl_bucket(struct ip6_tnl_parm *p)
+ip6_tnl_bucket(struct ip6_tnl_net *ip6n, struct ip6_tnl_parm *p)
 {
        struct in6_addr *remote = &p->raddr;
        struct in6_addr *local = &p->laddr;
@@ -171,7 +178,7 @@ ip6_tnl_bucket(struct ip6_tnl_parm *p)
                prio = 1;
                h = HASH(remote) ^ HASH(local);
        }
-       return &tnls[prio][h];
+       return &ip6n->tnls[prio][h];
 }
 
 /**
@@ -180,14 +187,14 @@ ip6_tnl_bucket(struct ip6_tnl_parm *p)
  **/
 
 static void
-ip6_tnl_link(struct ip6_tnl *t)
+ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 {
-       struct ip6_tnl **tp = ip6_tnl_bucket(&t->parms);
+       struct ip6_tnl **tp = ip6_tnl_bucket(ip6n, &t->parms);
 
+       spin_lock_bh(&ip6_tnl_lock);
        t->next = *tp;
-       write_lock_bh(&ip6_tnl_lock);
-       *tp = t;
-       write_unlock_bh(&ip6_tnl_lock);
+       rcu_assign_pointer(*tp, t);
+       spin_unlock_bh(&ip6_tnl_lock);
 }
 
 /**
@@ -196,15 +203,15 @@ ip6_tnl_link(struct ip6_tnl *t)
  **/
 
 static void
-ip6_tnl_unlink(struct ip6_tnl *t)
+ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
 {
        struct ip6_tnl **tp;
 
-       for (tp = ip6_tnl_bucket(&t->parms); *tp; tp = &(*tp)->next) {
+       for (tp = ip6_tnl_bucket(ip6n, &t->parms); *tp; tp = &(*tp)->next) {
                if (t == *tp) {
-                       write_lock_bh(&ip6_tnl_lock);
+                       spin_lock_bh(&ip6_tnl_lock);
                        *tp = t->next;
-                       write_unlock_bh(&ip6_tnl_lock);
+                       spin_unlock_bh(&ip6_tnl_lock);
                        break;
                }
        }
@@ -222,12 +229,13 @@ ip6_tnl_unlink(struct ip6_tnl *t)
  *   created tunnel or NULL
  **/
 
-static struct ip6_tnl *ip6_tnl_create(struct ip6_tnl_parm *p)
+static struct ip6_tnl *ip6_tnl_create(struct net *net, struct ip6_tnl_parm *p)
 {
        struct net_device *dev;
        struct ip6_tnl *t;
        char name[IFNAMSIZ];
        int err;
+       struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 
        if (p->name[0])
                strlcpy(name, p->name, IFNAMSIZ);
@@ -238,20 +246,22 @@ static struct ip6_tnl *ip6_tnl_create(struct ip6_tnl_parm *p)
        if (dev == NULL)
                goto failed;
 
+       dev_net_set(dev, net);
+
        if (strchr(name, '%')) {
                if (dev_alloc_name(dev, name) < 0)
                        goto failed_free;
        }
 
        t = netdev_priv(dev);
-       dev->init = ip6_tnl_dev_init;
        t->parms = *p;
+       ip6_tnl_dev_init(dev);
 
        if ((err = register_netdevice(dev)) < 0)
                goto failed_free;
 
        dev_hold(dev);
-       ip6_tnl_link(t);
+       ip6_tnl_link(ip6n, t);
        return t;
 
 failed_free:
@@ -274,20 +284,22 @@ failed:
  *   matching tunnel or NULL
  **/
 
-static struct ip6_tnl *ip6_tnl_locate(struct ip6_tnl_parm *p, int create)
+static struct ip6_tnl *ip6_tnl_locate(struct net *net,
+               struct ip6_tnl_parm *p, int create)
 {
        struct in6_addr *remote = &p->raddr;
        struct in6_addr *local = &p->laddr;
        struct ip6_tnl *t;
+       struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 
-       for (t = *ip6_tnl_bucket(p); t; t = t->next) {
+       for (t = *ip6_tnl_bucket(ip6n, p); t; t = t->next) {
                if (ipv6_addr_equal(local, &t->parms.laddr) &&
                    ipv6_addr_equal(remote, &t->parms.raddr))
                        return t;
        }
        if (!create)
                return NULL;
-       return ip6_tnl_create(p);
+       return ip6_tnl_create(net, p);
 }
 
 /**
@@ -302,13 +314,15 @@ static void
 ip6_tnl_dev_uninit(struct net_device *dev)
 {
        struct ip6_tnl *t = netdev_priv(dev);
+       struct net *net = dev_net(dev);
+       struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 
-       if (dev == ip6_fb_tnl_dev) {
-               write_lock_bh(&ip6_tnl_lock);
-               tnls_wc[0] = NULL;
-               write_unlock_bh(&ip6_tnl_lock);
+       if (dev == ip6n->fb_tnl_dev) {
+               spin_lock_bh(&ip6_tnl_lock);
+               ip6n->tnls_wc[0] = NULL;
+               spin_unlock_bh(&ip6_tnl_lock);
        } else {
-               ip6_tnl_unlink(t);
+               ip6_tnl_unlink(ip6n, t);
        }
        ip6_tnl_dst_reset(t);
        dev_put(dev);
@@ -385,13 +399,13 @@ parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
 
 static int
 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
-           int *type, int *code, int *msg, __u32 *info, int offset)
+           u8 *type, u8 *code, int *msg, __u32 *info, int offset)
 {
        struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
        struct ip6_tnl *t;
        int rel_msg = 0;
-       int rel_type = ICMPV6_DEST_UNREACH;
-       int rel_code = ICMPV6_ADDR_UNREACH;
+       u8 rel_type = ICMPV6_DEST_UNREACH;
+       u8 rel_code = ICMPV6_ADDR_UNREACH;
        __u32 rel_info = 0;
        __u16 len;
        int err = -ENOENT;
@@ -400,8 +414,9 @@ ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
           in trouble since we might need the source address for further
           processing of the error. */
 
-       read_lock(&ip6_tnl_lock);
-       if ((t = ip6_tnl_lookup(&ipv6h->daddr, &ipv6h->saddr)) == NULL)
+       rcu_read_lock();
+       if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr,
+                                       &ipv6h->saddr)) == NULL)
                goto out;
 
        if (t->parms.proto != ipproto && t->parms.proto != 0)
@@ -472,17 +487,17 @@ ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
        *msg = rel_msg;
 
 out:
-       read_unlock(&ip6_tnl_lock);
+       rcu_read_unlock();
        return err;
 }
 
 static int
 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
-          int type, int code, int offset, __be32 info)
+          u8 type, u8 code, int offset, __be32 info)
 {
        int rel_msg = 0;
-       int rel_type = type;
-       int rel_code = code;
+       u8 rel_type = type;
+       u8 rel_code = code;
        __u32 rel_info = ntohl(info);
        int err;
        struct sk_buff *skb2;
@@ -522,8 +537,8 @@ ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
        if (!skb2)
                return 0;
 
-       dst_release(skb2->dst);
-       skb2->dst = NULL;
+       skb_dst_drop(skb2);
+
        skb_pull(skb2, offset);
        skb_reset_network_header(skb2);
        eiph = ip_hdr(skb2);
@@ -533,7 +548,7 @@ ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
        fl.fl4_dst = eiph->saddr;
        fl.fl4_tos = RT_TOS(eiph->tos);
        fl.proto = IPPROTO_IPIP;
-       if (ip_route_output_key(&init_net, &rt, &fl))
+       if (ip_route_output_key(dev_net(skb->dev), &rt, &fl))
                goto out;
 
        skb2->dev = rt->u.dst.dev;
@@ -545,26 +560,26 @@ ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
                fl.fl4_dst = eiph->daddr;
                fl.fl4_src = eiph->saddr;
                fl.fl4_tos = eiph->tos;
-               if (ip_route_output_key(&init_net, &rt, &fl) ||
+               if (ip_route_output_key(dev_net(skb->dev), &rt, &fl) ||
                    rt->u.dst.dev->type != ARPHRD_TUNNEL) {
                        ip_rt_put(rt);
                        goto out;
                }
-               skb2->dst = (struct dst_entry *)rt;
+               skb_dst_set(skb2, (struct dst_entry *)rt);
        } else {
                ip_rt_put(rt);
                if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
                                   skb2->dev) ||
-                   skb2->dst->dev->type != ARPHRD_TUNNEL)
+                   skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
                        goto out;
        }
 
        /* change mtu on this route */
        if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
-               if (rel_info > dst_mtu(skb2->dst))
+               if (rel_info > dst_mtu(skb_dst(skb2)))
                        goto out;
 
-               skb2->dst->ops->update_pmtu(skb2->dst, rel_info);
+               skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), rel_info);
        }
 
        icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
@@ -576,11 +591,11 @@ out:
 
 static int
 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
-          int type, int code, int offset, __be32 info)
+          u8 type, u8 code, int offset, __be32 info)
 {
        int rel_msg = 0;
-       int rel_type = type;
-       int rel_code = code;
+       u8 rel_type = type;
+       u8 rel_code = code;
        __u32 rel_info = ntohl(info);
        int err;
 
@@ -596,13 +611,13 @@ ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
                if (!skb2)
                        return 0;
 
-               dst_release(skb2->dst);
-               skb2->dst = NULL;
+               skb_dst_drop(skb2);
                skb_pull(skb2, offset);
                skb_reset_network_header(skb2);
 
                /* Try to guess incoming interface */
-               rt = rt6_lookup(&ipv6_hdr(skb2)->saddr, NULL, 0, 0);
+               rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
+                               NULL, 0, 0);
 
                if (rt && rt->rt6i_dev)
                        skb2->dev = rt->rt6i_dev;
@@ -642,24 +657,24 @@ static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
                IP6_ECN_set_ce(ipv6_hdr(skb));
 }
 
+/* called with rcu_read_lock() */
 static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
 {
        struct ip6_tnl_parm *p = &t->parms;
        int ret = 0;
+       struct net *net = dev_net(t->dev);
 
        if (p->flags & IP6_TNL_F_CAP_RCV) {
                struct net_device *ldev = NULL;
 
                if (p->link)
-                       ldev = dev_get_by_index(&init_net, p->link);
+                       ldev = dev_get_by_index_rcu(net, p->link);
 
                if ((ipv6_addr_is_multicast(&p->laddr) ||
-                    likely(ipv6_chk_addr(&init_net, &p->laddr, ldev, 0))) &&
-                   likely(!ipv6_chk_addr(&init_net, &p->raddr, NULL, 0)))
+                    likely(ipv6_chk_addr(net, &p->laddr, ldev, 0))) &&
+                   likely(!ipv6_chk_addr(net, &p->raddr, NULL, 0)))
                        ret = 1;
 
-               if (ldev)
-                       dev_put(ldev);
        }
        return ret;
 }
@@ -682,22 +697,23 @@ static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
        struct ip6_tnl *t;
        struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 
-       read_lock(&ip6_tnl_lock);
+       rcu_read_lock();
 
-       if ((t = ip6_tnl_lookup(&ipv6h->saddr, &ipv6h->daddr)) != NULL) {
+       if ((t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
+                                       &ipv6h->daddr)) != NULL) {
                if (t->parms.proto != ipproto && t->parms.proto != 0) {
-                       read_unlock(&ip6_tnl_lock);
+                       rcu_read_unlock();
                        goto discard;
                }
 
                if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
-                       read_unlock(&ip6_tnl_lock);
+                       rcu_read_unlock();
                        goto discard;
                }
 
                if (!ip6_tnl_rcv_ctl(t)) {
-                       t->stat.rx_dropped++;
-                       read_unlock(&ip6_tnl_lock);
+                       t->dev->stats.rx_dropped++;
+                       rcu_read_unlock();
                        goto discard;
                }
                secpath_reset(skb);
@@ -707,19 +723,18 @@ static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
                skb->pkt_type = PACKET_HOST;
                memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
                skb->dev = t->dev;
-               dst_release(skb->dst);
-               skb->dst = NULL;
+               skb_dst_drop(skb);
                nf_reset(skb);
 
                dscp_ecn_decapsulate(t, ipv6h, skb);
 
-               t->stat.rx_packets++;
-               t->stat.rx_bytes += skb->len;
+               t->dev->stats.rx_packets++;
+               t->dev->stats.rx_bytes += skb->len;
                netif_rx(skb);
-               read_unlock(&ip6_tnl_lock);
+               rcu_read_unlock();
                return 0;
        }
-       read_unlock(&ip6_tnl_lock);
+       rcu_read_unlock();
        return 1;
 
 discard:
@@ -782,27 +797,28 @@ static inline int ip6_tnl_xmit_ctl(struct ip6_tnl *t)
 {
        struct ip6_tnl_parm *p = &t->parms;
        int ret = 0;
+       struct net *net = dev_net(t->dev);
 
        if (p->flags & IP6_TNL_F_CAP_XMIT) {
                struct net_device *ldev = NULL;
 
+               rcu_read_lock();
                if (p->link)
-                       ldev = dev_get_by_index(&init_net, p->link);
+                       ldev = dev_get_by_index_rcu(net, p->link);
 
-               if (unlikely(!ipv6_chk_addr(&init_net, &p->laddr, ldev, 0)))
+               if (unlikely(!ipv6_chk_addr(net, &p->laddr, ldev, 0)))
                        printk(KERN_WARNING
                               "%s xmit: Local address not yet configured!\n",
                               p->name);
                else if (!ipv6_addr_is_multicast(&p->raddr) &&
-                        unlikely(ipv6_chk_addr(&init_net, &p->raddr, NULL, 0)))
+                        unlikely(ipv6_chk_addr(net, &p->raddr, NULL, 0)))
                        printk(KERN_WARNING
                               "%s xmit: Routing loop! "
                               "Remote address found on this node!\n",
                               p->name);
                else
                        ret = 1;
-               if (ldev)
-                       dev_put(ldev);
+               rcu_read_unlock();
        }
        return ret;
 }
@@ -832,8 +848,9 @@ static int ip6_tnl_xmit2(struct sk_buff *skb,
                         int encap_limit,
                         __u32 *pmtu)
 {
+       struct net *net = dev_net(dev);
        struct ip6_tnl *t = netdev_priv(dev);
-       struct net_device_stats *stats = &t->stat;
+       struct net_device_stats *stats = &t->dev->stats;
        struct ipv6hdr *ipv6h = ipv6_hdr(skb);
        struct ipv6_tel_txoption opt;
        struct dst_entry *dst;
@@ -847,9 +864,9 @@ static int ip6_tnl_xmit2(struct sk_buff *skb,
        if ((dst = ip6_tnl_dst_check(t)) != NULL)
                dst_hold(dst);
        else {
-               dst = ip6_route_output(NULL, fl);
+               dst = ip6_route_output(net, NULL, fl);
 
-               if (dst->error || xfrm_lookup(&dst, fl, NULL, 0) < 0)
+               if (dst->error || xfrm_lookup(net, &dst, fl, NULL, 0) < 0)
                        goto tx_err_link_failure;
        }
 
@@ -870,8 +887,8 @@ static int ip6_tnl_xmit2(struct sk_buff *skb,
        }
        if (mtu < IPV6_MIN_MTU)
                mtu = IPV6_MIN_MTU;
-       if (skb->dst)
-               skb->dst->ops->update_pmtu(skb->dst, mtu);
+       if (skb_dst(skb))
+               skb_dst(skb)->ops->update_pmtu(skb_dst(skb), mtu);
        if (skb->len > mtu) {
                *pmtu = mtu;
                err = -EMSGSIZE;
@@ -895,8 +912,8 @@ static int ip6_tnl_xmit2(struct sk_buff *skb,
                kfree_skb(skb);
                skb = new_skb;
        }
-       dst_release(skb->dst);
-       skb->dst = dst_clone(dst);
+       skb_dst_drop(skb);
+       skb_dst_set(skb, dst_clone(dst));
 
        skb->transport_header = skb->network_header;
 
@@ -1023,23 +1040,18 @@ ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
        return 0;
 }
 
-static int
+static netdev_tx_t
 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
 {
        struct ip6_tnl *t = netdev_priv(dev);
-       struct net_device_stats *stats = &t->stat;
+       struct net_device_stats *stats = &t->dev->stats;
        int ret;
 
-       if (t->recursion++) {
-               t->stat.collisions++;
-               goto tx_err;
-       }
-
        switch (skb->protocol) {
-       case __constant_htons(ETH_P_IP):
+       case htons(ETH_P_IP):
                ret = ip4ip6_tnl_xmit(skb, dev);
                break;
-       case __constant_htons(ETH_P_IPV6):
+       case htons(ETH_P_IPV6):
                ret = ip6ip6_tnl_xmit(skb, dev);
                break;
        default:
@@ -1049,15 +1061,13 @@ ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
        if (ret < 0)
                goto tx_err;
 
-       t->recursion--;
-       return 0;
+       return NETDEV_TX_OK;
 
 tx_err:
        stats->tx_errors++;
        stats->tx_dropped++;
        kfree_skb(skb);
-       t->recursion--;
-       return 0;
+       return NETDEV_TX_OK;
 }
 
 static void ip6_tnl_set_cap(struct ip6_tnl *t)
@@ -1085,8 +1095,8 @@ static void ip6_tnl_link_config(struct ip6_tnl *t)
        struct ip6_tnl_parm *p = &t->parms;
        struct flowi *fl = &t->fl;
 
-       memcpy(&dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
-       memcpy(&dev->broadcast, &p->raddr, sizeof(struct in6_addr));
+       memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
+       memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
 
        /* Set up flowi template */
        ipv6_addr_copy(&fl->fl6_src, &p->laddr);
@@ -1112,7 +1122,8 @@ static void ip6_tnl_link_config(struct ip6_tnl *t)
                int strict = (ipv6_addr_type(&p->raddr) &
                              (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
 
-               struct rt6_info *rt = rt6_lookup(&p->raddr, &p->laddr,
+               struct rt6_info *rt = rt6_lookup(dev_net(dev),
+                                                &p->raddr, &p->laddr,
                                                 p->link, strict);
 
                if (rt == NULL)
@@ -1135,7 +1146,6 @@ static void ip6_tnl_link_config(struct ip6_tnl *t)
  * ip6_tnl_change - update the tunnel parameters
  *   @t: tunnel to be changed
  *   @p: tunnel configuration parameters
- *   @active: != 0 if tunnel is ready for use
  *
  * Description:
  *   ip6_tnl_change() updates the tunnel parameters
@@ -1191,15 +1201,17 @@ ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
        int err = 0;
        struct ip6_tnl_parm p;
        struct ip6_tnl *t = NULL;
+       struct net *net = dev_net(dev);
+       struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
 
        switch (cmd) {
        case SIOCGETTUNNEL:
-               if (dev == ip6_fb_tnl_dev) {
+               if (dev == ip6n->fb_tnl_dev) {
                        if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p))) {
                                err = -EFAULT;
                                break;
                        }
-                       t = ip6_tnl_locate(&p, 0);
+                       t = ip6_tnl_locate(net, &p, 0);
                }
                if (t == NULL)
                        t = netdev_priv(dev);
@@ -1220,8 +1232,8 @@ ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
                if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
                    p.proto != 0)
                        break;
-               t = ip6_tnl_locate(&p, cmd == SIOCADDTUNNEL);
-               if (dev != ip6_fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
+               t = ip6_tnl_locate(net, &p, cmd == SIOCADDTUNNEL);
+               if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
                        if (t != NULL) {
                                if (t->dev != dev) {
                                        err = -EEXIST;
@@ -1230,9 +1242,9 @@ ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
                        } else
                                t = netdev_priv(dev);
 
-                       ip6_tnl_unlink(t);
+                       ip6_tnl_unlink(ip6n, t);
                        err = ip6_tnl_change(t, &p);
-                       ip6_tnl_link(t);
+                       ip6_tnl_link(ip6n, t);
                        netdev_state_change(dev);
                }
                if (t) {
@@ -1248,15 +1260,15 @@ ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
                if (!capable(CAP_NET_ADMIN))
                        break;
 
-               if (dev == ip6_fb_tnl_dev) {
+               if (dev == ip6n->fb_tnl_dev) {
                        err = -EFAULT;
                        if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
                                break;
                        err = -ENOENT;
-                       if ((t = ip6_tnl_locate(&p, 0)) == NULL)
+                       if ((t = ip6_tnl_locate(net, &p, 0)) == NULL)
                                break;
                        err = -EPERM;
-                       if (t->dev == ip6_fb_tnl_dev)
+                       if (t->dev == ip6n->fb_tnl_dev)
                                break;
                        dev = t->dev;
                }
@@ -1270,19 +1282,6 @@ ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
 }
 
 /**
- * ip6_tnl_get_stats - return the stats for tunnel device
- *   @dev: virtual device associated with tunnel
- *
- * Return: stats for device
- **/
-
-static struct net_device_stats *
-ip6_tnl_get_stats(struct net_device *dev)
-{
-       return &(((struct ip6_tnl *)netdev_priv(dev))->stat);
-}
-
-/**
  * ip6_tnl_change_mtu - change mtu manually for tunnel device
  *   @dev: virtual device associated with tunnel
  *   @new_mtu: the new mtu
@@ -1302,6 +1301,14 @@ ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
        return 0;
 }
 
+
+static const struct net_device_ops ip6_tnl_netdev_ops = {
+       .ndo_uninit = ip6_tnl_dev_uninit,
+       .ndo_start_xmit = ip6_tnl_xmit,
+       .ndo_do_ioctl = ip6_tnl_ioctl,
+       .ndo_change_mtu = ip6_tnl_change_mtu,
+};
+
 /**
  * ip6_tnl_dev_setup - setup virtual tunnel device
  *   @dev: virtual device associated with tunnel
@@ -1312,18 +1319,15 @@ ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
 
 static void ip6_tnl_dev_setup(struct net_device *dev)
 {
-       dev->uninit = ip6_tnl_dev_uninit;
+       dev->netdev_ops = &ip6_tnl_netdev_ops;
        dev->destructor = free_netdev;
-       dev->hard_start_xmit = ip6_tnl_xmit;
-       dev->get_stats = ip6_tnl_get_stats;
-       dev->do_ioctl = ip6_tnl_ioctl;
-       dev->change_mtu = ip6_tnl_change_mtu;
 
        dev->type = ARPHRD_TUNNEL6;
        dev->hard_header_len = LL_MAX_HEADER + sizeof (struct ipv6hdr);
        dev->mtu = ETH_DATA_LEN - sizeof (struct ipv6hdr);
        dev->flags |= IFF_NOARP;
        dev->addr_len = sizeof(struct in6_addr);
+       dev->features |= NETIF_F_NETNS_LOCAL;
 }
 
 
@@ -1345,13 +1349,11 @@ ip6_tnl_dev_init_gen(struct net_device *dev)
  *   @dev: virtual device associated with tunnel
  **/
 
-static int
-ip6_tnl_dev_init(struct net_device *dev)
+static void ip6_tnl_dev_init(struct net_device *dev)
 {
        struct ip6_tnl *t = netdev_priv(dev);
        ip6_tnl_dev_init_gen(dev);
        ip6_tnl_link_config(t);
-       return 0;
 }
 
 /**
@@ -1361,15 +1363,16 @@ ip6_tnl_dev_init(struct net_device *dev)
  * Return: 0
  **/
 
-static int
-ip6_fb_tnl_dev_init(struct net_device *dev)
+static void __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
 {
        struct ip6_tnl *t = netdev_priv(dev);
+       struct net *net = dev_net(dev);
+       struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
+
        ip6_tnl_dev_init_gen(dev);
        t->parms.proto = IPPROTO_IPV6;
        dev_hold(dev);
-       tnls_wc[0] = t;
-       return 0;
+       ip6n->tnls_wc[0] = t;
 }
 
 static struct xfrm6_tunnel ip4ip6_handler = {
@@ -1384,6 +1387,70 @@ static struct xfrm6_tunnel ip6ip6_handler = {
        .priority       =       1,
 };
 
+static void __net_exit ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
+{
+       int h;
+       struct ip6_tnl *t;
+       LIST_HEAD(list);
+
+       for (h = 0; h < HASH_SIZE; h++) {
+               t = ip6n->tnls_r_l[h];
+               while (t != NULL) {
+                       unregister_netdevice_queue(t->dev, &list);
+                       t = t->next;
+               }
+       }
+
+       t = ip6n->tnls_wc[0];
+       unregister_netdevice_queue(t->dev, &list);
+       unregister_netdevice_many(&list);
+}
+
+static int __net_init ip6_tnl_init_net(struct net *net)
+{
+       struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
+       int err;
+
+       ip6n->tnls[0] = ip6n->tnls_wc;
+       ip6n->tnls[1] = ip6n->tnls_r_l;
+
+       err = -ENOMEM;
+       ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
+                                     ip6_tnl_dev_setup);
+
+       if (!ip6n->fb_tnl_dev)
+               goto err_alloc_dev;
+       dev_net_set(ip6n->fb_tnl_dev, net);
+
+       ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
+
+       err = register_netdev(ip6n->fb_tnl_dev);
+       if (err < 0)
+               goto err_register;
+       return 0;
+
+err_register:
+       free_netdev(ip6n->fb_tnl_dev);
+err_alloc_dev:
+       return err;
+}
+
+static void __net_exit ip6_tnl_exit_net(struct net *net)
+{
+       struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
+
+       rtnl_lock();
+       ip6_tnl_destroy_tunnels(ip6n);
+       rtnl_unlock();
+}
+
+static struct pernet_operations ip6_tnl_net_ops = {
+       .init = ip6_tnl_init_net,
+       .exit = ip6_tnl_exit_net,
+       .id   = &ip6_tnl_net_id,
+       .size = sizeof(struct ip6_tnl_net),
+};
+
 /**
  * ip6_tunnel_init - register protocol and reserve needed resources
  *
@@ -1394,53 +1461,32 @@ static int __init ip6_tunnel_init(void)
 {
        int  err;
 
-       if (xfrm6_tunnel_register(&ip4ip6_handler, AF_INET)) {
+       err = register_pernet_device(&ip6_tnl_net_ops);
+       if (err < 0)
+               goto out_pernet;
+
+       err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
+       if (err < 0) {
                printk(KERN_ERR "ip6_tunnel init: can't register ip4ip6\n");
-               err = -EAGAIN;
-               goto out;
+               goto out_ip4ip6;
        }
 
-       if (xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6)) {
+       err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
+       if (err < 0) {
                printk(KERN_ERR "ip6_tunnel init: can't register ip6ip6\n");
-               err = -EAGAIN;
-               goto unreg_ip4ip6;
-       }
-       ip6_fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
-                                     ip6_tnl_dev_setup);
-
-       if (!ip6_fb_tnl_dev) {
-               err = -ENOMEM;
-               goto fail;
+               goto out_ip6ip6;
        }
-       ip6_fb_tnl_dev->init = ip6_fb_tnl_dev_init;
 
-       if ((err = register_netdev(ip6_fb_tnl_dev))) {
-               free_netdev(ip6_fb_tnl_dev);
-               goto fail;
-       }
        return 0;
-fail:
-       xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
-unreg_ip4ip6:
+
+out_ip6ip6:
        xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
-out:
+out_ip4ip6:
+       unregister_pernet_device(&ip6_tnl_net_ops);
+out_pernet:
        return err;
 }
 
-static void __exit ip6_tnl_destroy_tunnels(void)
-{
-       int h;
-       struct ip6_tnl *t;
-
-       for (h = 0; h < HASH_SIZE; h++) {
-               while ((t = tnls_r_l[h]) != NULL)
-                       unregister_netdevice(t->dev);
-       }
-
-       t = tnls_wc[0];
-       unregister_netdevice(t->dev);
-}
-
 /**
  * ip6_tunnel_cleanup - free resources and unregister protocol
  **/
@@ -1453,9 +1499,7 @@ static void __exit ip6_tunnel_cleanup(void)
        if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
                printk(KERN_INFO "ip6_tunnel close: can't deregister ip6ip6\n");
 
-       rtnl_lock();
-       ip6_tnl_destroy_tunnels();
-       rtnl_unlock();
+       unregister_pernet_device(&ip6_tnl_net_ops);
 }
 
 module_init(ip6_tunnel_init);