netfilter: ip6t_ipv6header: fix match on packets ending with NEXTHDR_NONE
[linux-2.6.git] / net / ipv6 / ip6_tunnel.c
index 70684e0..d994c55 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
@@ -74,18 +74,19 @@ 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_fb_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;
+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);
@@ -130,19 +131,20 @@ static inline void ip6_tnl_dst_store(struct ip6_tnl *t, struct dst_entry *dst)
  **/
 
 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 (t = ip6n->tnls_r_l[h0 ^ h1]; t; t = t->next) {
                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))
+       if ((t = ip6n->tnls_wc[0]) != NULL && (t->dev->flags & IFF_UP))
                return t;
 
        return NULL;
@@ -160,7 +162,7 @@ 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 +173,7 @@ ip6_tnl_bucket(struct ip6_tnl_parm *p)
                prio = 1;
                h = HASH(remote) ^ HASH(local);
        }
-       return &tnls[prio][h];
+       return &ip6n->tnls[prio][h];
 }
 
 /**
@@ -180,9 +182,9 @@ 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);
 
        t->next = *tp;
        write_lock_bh(&ip6_tnl_lock);
@@ -196,11 +198,11 @@ 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);
                        *tp = t->next;
@@ -222,40 +224,43 @@ 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]) {
+       if (p->name[0])
                strlcpy(name, p->name, IFNAMSIZ);
-       } else {
-               int i;
-               for (i = 1; i < IP6_TNL_MAX; i++) {
-                       sprintf(name, "ip6tnl%d", i);
-                       if (__dev_get_by_name(name) == NULL)
-                               break;
-               }
-               if (i == IP6_TNL_MAX)
-                       goto failed;
-       }
+       else
+               sprintf(name, "ip6tnl%%d");
+
        dev = alloc_netdev(sizeof (*t), name, ip6_tnl_dev_setup);
        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;
 
-       if ((err = register_netdevice(dev)) < 0) {
-               free_netdev(dev);
-               goto failed;
-       }
        dev_hold(dev);
-       ip6_tnl_link(t);
+       ip6_tnl_link(ip6n, t);
        return t;
+
+failed_free:
+       free_netdev(dev);
 failed:
        return NULL;
 }
@@ -274,20 +279,22 @@ 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 +309,15 @@ 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) {
+       if (dev == ip6n->fb_tnl_dev) {
                write_lock_bh(&ip6_tnl_lock);
-               tnls_wc[0] = NULL;
+               ip6n->tnls_wc[0] = NULL;
                write_unlock_bh(&ip6_tnl_lock);
        } else {
-               ip6_tnl_unlink(t);
+               ip6_tnl_unlink(ip6n, t);
        }
        ip6_tnl_dst_reset(t);
        dev_put(dev);
@@ -384,8 +393,8 @@ parse_tlv_tnl_enc_lim(struct sk_buff *skb, __u8 * raw)
  **/
 
 static int
-ip6_tnl_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
-           int *type, int *code, int *msg, __be32 *info, int offset)
+ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
+           int *type, int *code, int *msg, __u32 *info, int offset)
 {
        struct ipv6hdr *ipv6h = (struct ipv6hdr *) skb->data;
        struct ip6_tnl *t;
@@ -401,7 +410,11 @@ ip6_tnl_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
           processing of the error. */
 
        read_lock(&ip6_tnl_lock);
-       if ((t = ip6_tnl_lookup(&ipv6h->daddr, &ipv6h->saddr)) == NULL)
+       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)
                goto out;
 
        err = 0;
@@ -432,7 +445,7 @@ ip6_tnl_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
                if ((*code) == ICMPV6_HDR_FIELD)
                        teli = parse_tlv_tnl_enc_lim(skb, skb->data);
 
-               if (teli && teli == ntohl(*info) - 2) {
+               if (teli && teli == *info - 2) {
                        tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
                        if (tel->encap_limit == 0) {
                                if (net_ratelimit())
@@ -449,7 +462,7 @@ ip6_tnl_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
                }
                break;
        case ICMPV6_PKT_TOOBIG:
-               mtu = ntohl(*info) - offset;
+               mtu = *info - offset;
                if (mtu < IPV6_MIN_MTU)
                        mtu = IPV6_MIN_MTU;
                t->dev->mtu = mtu;
@@ -475,20 +488,20 @@ out:
 
 static int
 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
-          int type, int code, int offset, __u32 info)
+          int type, int code, int offset, __be32 info)
 {
        int rel_msg = 0;
        int rel_type = type;
        int rel_code = code;
-       __u32 rel_info = info;
+       __u32 rel_info = ntohl(info);
        int err;
        struct sk_buff *skb2;
        struct iphdr *eiph;
        struct flowi fl;
        struct rtable *rt;
 
-       err = ip6_tnl_err(skb, opt, &rel_type, &rel_code, &rel_msg, &rel_info,
-                         offset);
+       err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
+                         &rel_msg, &rel_info, offset);
        if (err < 0)
                return err;
 
@@ -522,15 +535,15 @@ ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
        dst_release(skb2->dst);
        skb2->dst = NULL;
        skb_pull(skb2, offset);
-       skb2->nh.raw = skb2->data;
-       eiph = skb2->nh.iph;
+       skb_reset_network_header(skb2);
+       eiph = ip_hdr(skb2);
 
        /* Try to guess incoming interface */
        memset(&fl, 0, sizeof(fl));
        fl.fl4_dst = eiph->saddr;
        fl.fl4_tos = RT_TOS(eiph->tos);
        fl.proto = IPPROTO_IPIP;
-       if (ip_route_output_key(&rt, &fl))
+       if (ip_route_output_key(dev_net(skb->dev), &rt, &fl))
                goto out;
 
        skb2->dev = rt->u.dst.dev;
@@ -542,11 +555,12 @@ 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(&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;
        } else {
                ip_rt_put(rt);
                if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
@@ -561,10 +575,9 @@ ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
                        goto out;
 
                skb2->dst->ops->update_pmtu(skb2->dst, rel_info);
-               rel_info = htonl(rel_info);
        }
 
-       icmp_send(skb2, rel_type, rel_code, rel_info);
+       icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
 
 out:
        kfree_skb(skb2);
@@ -573,16 +586,16 @@ out:
 
 static int
 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
-          int type, int code, int offset, __u32 info)
+          int type, int code, int offset, __be32 info)
 {
        int rel_msg = 0;
        int rel_type = type;
        int rel_code = code;
-       __u32 rel_info = info;
+       __u32 rel_info = ntohl(info);
        int err;
 
-       err = ip6_tnl_err(skb, opt, &rel_type, &rel_code, &rel_msg, &rel_info,
-                         offset);
+       err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
+                         &rel_msg, &rel_info, offset);
        if (err < 0)
                return err;
 
@@ -596,10 +609,11 @@ ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
                dst_release(skb2->dst);
                skb2->dst = NULL;
                skb_pull(skb2, offset);
-               skb2->nh.raw = skb2->data;
+               skb_reset_network_header(skb2);
 
                /* Try to guess incoming interface */
-               rt = rt6_lookup(&skb2->nh.ipv6h->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;
@@ -622,10 +636,10 @@ static void ip4ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
        __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
 
        if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
-               ipv4_change_dsfield(skb->nh.iph, INET_ECN_MASK, dsfield);
+               ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
 
        if (INET_ECN_is_ce(dsfield))
-               IP_ECN_set_ce(skb->nh.iph);
+               IP_ECN_set_ce(ip_hdr(skb));
 }
 
 static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
@@ -633,26 +647,27 @@ static void ip6ip6_dscp_ecn_decapsulate(struct ip6_tnl *t,
                                        struct sk_buff *skb)
 {
        if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
-               ipv6_copy_dscp(ipv6h, skb->nh.ipv6h);
+               ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
 
        if (INET_ECN_is_ce(ipv6_get_dsfield(ipv6h)))
-               IP6_ECN_set_ce(skb->nh.ipv6h);
+               IP6_ECN_set_ce(ipv6_hdr(skb));
 }
 
 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(p->link);
+                       ldev = dev_get_by_index(net, p->link);
 
                if ((ipv6_addr_is_multicast(&p->laddr) ||
-                    likely(ipv6_chk_addr(&p->laddr, ldev, 0))) &&
-                   likely(!ipv6_chk_addr(&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)
@@ -671,31 +686,36 @@ static inline int ip6_tnl_rcv_ctl(struct ip6_tnl *t)
  **/
 
 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
+                      __u8 ipproto,
                       void (*dscp_ecn_decapsulate)(struct ip6_tnl *t,
                                                    struct ipv6hdr *ipv6h,
                                                    struct sk_buff *skb))
 {
-       struct ipv6hdr *ipv6h;
        struct ip6_tnl *t;
-
-       ipv6h = skb->nh.ipv6h;
+       struct ipv6hdr *ipv6h = ipv6_hdr(skb);
 
        read_lock(&ip6_tnl_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);
+                       goto discard;
+               }
+
                if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
                        read_unlock(&ip6_tnl_lock);
                        goto discard;
                }
 
                if (!ip6_tnl_rcv_ctl(t)) {
-                       t->stat.rx_dropped++;
+                       t->dev->stats.rx_dropped++;
                        read_unlock(&ip6_tnl_lock);
                        goto discard;
                }
                secpath_reset(skb);
-               skb->mac.raw = skb->nh.raw;
-               skb->nh.raw = skb->data;
+               skb->mac_header = skb->network_header;
+               skb_reset_network_header(skb);
                skb->protocol = htons(protocol);
                skb->pkt_type = PACKET_HOST;
                memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
@@ -706,8 +726,8 @@ static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
 
                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);
                return 0;
@@ -722,12 +742,14 @@ discard:
 
 static int ip4ip6_rcv(struct sk_buff *skb)
 {
-       return ip6_tnl_rcv(skb, ETH_P_IP, ip4ip6_dscp_ecn_decapsulate);
+       return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
+                          ip4ip6_dscp_ecn_decapsulate);
 }
 
 static int ip6ip6_rcv(struct sk_buff *skb)
 {
-       return ip6_tnl_rcv(skb, ETH_P_IPV6, ip6ip6_dscp_ecn_decapsulate);
+       return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
+                          ip6ip6_dscp_ecn_decapsulate);
 }
 
 struct ipv6_tel_txoption {
@@ -773,19 +795,20 @@ 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;
 
                if (p->link)
-                       ldev = dev_get_by_index(p->link);
+                       ldev = dev_get_by_index(net, p->link);
 
-               if (unlikely(!ipv6_chk_addr(&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(&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",
@@ -823,14 +846,15 @@ 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 ipv6hdr *ipv6h = skb->nh.ipv6h;
+       struct net_device_stats *stats = &t->dev->stats;
+       struct ipv6hdr *ipv6h = ipv6_hdr(skb);
        struct ipv6_tel_txoption opt;
        struct dst_entry *dst;
        struct net_device *tdev;
        int mtu;
-       int max_headroom = sizeof(struct ipv6hdr);
+       unsigned int max_headroom = sizeof(struct ipv6hdr);
        u8 proto;
        int err = -1;
        int pkt_len;
@@ -838,9 +862,9 @@ 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;
        }
 
@@ -874,8 +898,8 @@ static int ip6_tnl_xmit2(struct sk_buff *skb,
         */
        max_headroom += LL_RESERVED_SPACE(tdev);
 
-       if (skb_headroom(skb) < max_headroom ||
-           skb_cloned(skb) || skb_shared(skb)) {
+       if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
+           (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
                struct sk_buff *new_skb;
 
                if (!(new_skb = skb_realloc_headroom(skb, max_headroom)))
@@ -889,27 +913,26 @@ static int ip6_tnl_xmit2(struct sk_buff *skb,
        dst_release(skb->dst);
        skb->dst = dst_clone(dst);
 
-       skb->h.raw = skb->nh.raw;
+       skb->transport_header = skb->network_header;
 
        proto = fl->proto;
        if (encap_limit >= 0) {
                init_tel_txopt(&opt, encap_limit);
                ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
        }
-       skb->nh.raw = skb_push(skb, sizeof(struct ipv6hdr));
-       ipv6h = skb->nh.ipv6h;
+       skb_push(skb, sizeof(struct ipv6hdr));
+       skb_reset_network_header(skb);
+       ipv6h = ipv6_hdr(skb);
        *(__be32*)ipv6h = fl->fl6_flowlabel | htonl(0x60000000);
        dsfield = INET_ECN_encapsulate(0, dsfield);
        ipv6_change_dsfield(ipv6h, ~INET_ECN_MASK, dsfield);
-       ipv6h->payload_len = htons(skb->len - sizeof(struct ipv6hdr));
        ipv6h->hop_limit = t->parms.hop_limit;
        ipv6h->nexthdr = proto;
        ipv6_addr_copy(&ipv6h->saddr, &fl->fl6_src);
        ipv6_addr_copy(&ipv6h->daddr, &fl->fl6_dst);
        nf_reset(skb);
        pkt_len = skb->len;
-       err = NF_HOOK(PF_INET6, NF_IP6_LOCAL_OUT, skb, NULL,
-                     skb->dst->dev, dst_output);
+       err = ip6_local_out(skb);
 
        if (net_xmit_eval(err) == 0) {
                stats->tx_bytes += pkt_len;
@@ -932,14 +955,15 @@ static inline int
 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
 {
        struct ip6_tnl *t = netdev_priv(dev);
-       struct iphdr  *iph = skb->nh.iph;
+       struct iphdr  *iph = ip_hdr(skb);
        int encap_limit = -1;
        struct flowi fl;
        __u8 dsfield;
        __u32 mtu;
        int err;
 
-       if (!ip6_tnl_xmit_ctl(t))
+       if ((t->parms.proto != IPPROTO_IPIP && t->parms.proto != 0) ||
+           !ip6_tnl_xmit_ctl(t))
                return -1;
 
        if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
@@ -951,8 +975,8 @@ ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
        dsfield = ipv4_get_dsfield(iph);
 
        if ((t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS))
-               fl.fl6_flowlabel |= ntohl(((__u32)iph->tos << IPV6_TCLASS_SHIFT)
-                                         & IPV6_TCLASS_MASK);
+               fl.fl6_flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
+                                         & IPV6_TCLASS_MASK;
 
        err = ip6_tnl_xmit2(skb, dev, dsfield, &fl, encap_limit, &mtu);
        if (err != 0) {
@@ -970,7 +994,7 @@ static inline int
 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
 {
        struct ip6_tnl *t = netdev_priv(dev);
-       struct ipv6hdr *ipv6h = skb->nh.ipv6h;
+       struct ipv6hdr *ipv6h = ipv6_hdr(skb);
        int encap_limit = -1;
        __u16 offset;
        struct flowi fl;
@@ -978,12 +1002,14 @@ ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
        __u32 mtu;
        int err;
 
-       if (!ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
+       if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
+           !ip6_tnl_xmit_ctl(t) || ip6_tnl_addr_conflict(t, ipv6h))
                return -1;
 
-       if ((offset = parse_tlv_tnl_enc_lim(skb, skb->nh.raw)) > 0) {
+       offset = parse_tlv_tnl_enc_lim(skb, skb_network_header(skb));
+       if (offset > 0) {
                struct ipv6_tlv_tnl_enc_lim *tel;
-               tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->nh.raw[offset];
+               tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
                if (tel->encap_limit == 0) {
                        icmpv6_send(skb, ICMPV6_PARAMPROB,
                                    ICMPV6_HDR_FIELD, offset + 2, skb->dev);
@@ -1016,19 +1042,19 @@ static int
 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++;
+               stats->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:
@@ -1101,7 +1127,8 @@ 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)
@@ -1124,7 +1151,6 @@ 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
@@ -1140,6 +1166,7 @@ ip6_tnl_change(struct ip6_tnl *t, struct ip6_tnl_parm *p)
        t->parms.encap_limit = p->encap_limit;
        t->parms.flowinfo = p->flowinfo;
        t->parms.link = p->link;
+       t->parms.proto = p->proto;
        ip6_tnl_dst_reset(t);
        ip6_tnl_link_config(t);
        return 0;
@@ -1179,15 +1206,17 @@ 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);
@@ -1205,10 +1234,11 @@ ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
                if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof (p)))
                        break;
                err = -EINVAL;
-               if (p.proto != IPPROTO_IPV6)
+               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;
@@ -1217,9 +1247,9 @@ 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) {
@@ -1235,15 +1265,15 @@ 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;
                }
@@ -1257,19 +1287,6 @@ 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
@@ -1289,6 +1306,14 @@ 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
@@ -1299,19 +1324,15 @@ ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
 
 static void ip6_tnl_dev_setup(struct net_device *dev)
 {
-       SET_MODULE_OWNER(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;
 }
 
 
@@ -1324,7 +1345,6 @@ static inline void
 ip6_tnl_dev_init_gen(struct net_device *dev)
 {
        struct ip6_tnl *t = netdev_priv(dev);
-       t->fl.proto = IPPROTO_IPV6;
        t->dev = dev;
        strcpy(t->parms.name, dev->name);
 }
@@ -1334,13 +1354,11 @@ 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;
 }
 
 /**
@@ -1350,14 +1368,16 @@ ip6_tnl_dev_init(struct net_device *dev)
  * Return: 0
  **/
 
-static int
-ip6_fb_tnl_dev_init(struct net_device *dev)
+static void 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 = {
@@ -1372,6 +1392,78 @@ static struct xfrm6_tunnel ip6ip6_handler = {
        .priority       =       1,
 };
 
+static void ip6_tnl_destroy_tunnels(struct ip6_tnl_net *ip6n)
+{
+       int h;
+       struct ip6_tnl *t;
+
+       for (h = 0; h < HASH_SIZE; h++) {
+               while ((t = ip6n->tnls_r_l[h]) != NULL)
+                       unregister_netdevice(t->dev);
+       }
+
+       t = ip6n->tnls_wc[0];
+       unregister_netdevice(t->dev);
+}
+
+static int ip6_tnl_init_net(struct net *net)
+{
+       int err;
+       struct ip6_tnl_net *ip6n;
+
+       err = -ENOMEM;
+       ip6n = kzalloc(sizeof(struct ip6_tnl_net), GFP_KERNEL);
+       if (ip6n == NULL)
+               goto err_alloc;
+
+       err = net_assign_generic(net, ip6_tnl_net_id, ip6n);
+       if (err < 0)
+               goto err_assign;
+
+       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:
+       /* nothing */
+err_assign:
+       kfree(ip6n);
+err_alloc:
+       return err;
+}
+
+static void ip6_tnl_exit_net(struct net *net)
+{
+       struct ip6_tnl_net *ip6n;
+
+       ip6n = net_generic(net, ip6_tnl_net_id);
+       rtnl_lock();
+       ip6_tnl_destroy_tunnels(ip6n);
+       rtnl_unlock();
+       kfree(ip6n);
+}
+
+static struct pernet_operations ip6_tnl_net_ops = {
+       .init = ip6_tnl_init_net,
+       .exit = ip6_tnl_exit_net,
+};
+
 /**
  * ip6_tunnel_init - register protocol and reserve needed resources
  *
@@ -1393,21 +1485,12 @@ static int __init ip6_tunnel_init(void)
                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;
-       }
-       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;
-       }
+       err = register_pernet_gen_device(&ip6_tnl_net_id, &ip6_tnl_net_ops);
+       if (err < 0)
+               goto err_pernet;
        return 0;
-fail:
+err_pernet:
        xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
 unreg_ip4ip6:
        xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
@@ -1415,20 +1498,6 @@ out:
        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
  **/
@@ -1441,9 +1510,7 @@ 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_gen_device(ip6_tnl_net_id, &ip6_tnl_net_ops);
 }
 
 module_init(ip6_tunnel_init);