ipv6: Add support for IPsec virtual tunnel interfaces
authorSteffen Klassert <steffen.klassert@secunet.com>
Mon, 19 Aug 2013 06:07:34 +0000 (08:07 +0200)
committerSteffen Klassert <steffen.klassert@secunet.com>
Thu, 10 Oct 2013 10:00:01 +0000 (12:00 +0200)
This patch adds IPv6  support for IPsec virtual tunnel interfaces
(vti). IPsec virtual tunnel interfaces provide a routable interface
for IPsec tunnel endpoints.

Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
net/ipv6/Kconfig
net/ipv6/Makefile
net/ipv6/ip6_vti.c [new file with mode: 0644]

index 11b13ea69db4e6c35c8b992de9055bf6a312d802..e1a8d903e3662cb1aa36def1ffa5208cd6d515a8 100644 (file)
@@ -153,6 +153,17 @@ config INET6_XFRM_MODE_ROUTEOPTIMIZATION
        ---help---
          Support for MIPv6 route optimization mode.
 
+config IPV6_VTI
+tristate "Virtual (secure) IPv6: tunneling"
+       select IPV6_TUNNEL
+       depends on INET6_XFRM_MODE_TUNNEL
+       ---help---
+       Tunneling means encapsulating data of one protocol type within
+       another protocol and sending it over a channel that understands the
+       encapsulating protocol. This can be used with xfrm mode tunnel to give
+       the notion of a secure tunnel for IPSEC and then use routing protocol
+       on top.
+
 config IPV6_SIT
        tristate "IPv6: IPv6-in-IPv4 tunnel (SIT driver)"
        select INET_TUNNEL
index 470a9c008e9b9e921045bb5b11b77569e53ac5c9..17bb830872db21e07080ed9f6a0ef93cbccee9b7 100644 (file)
@@ -36,6 +36,7 @@ obj-$(CONFIG_INET6_XFRM_MODE_BEET) += xfrm6_mode_beet.o
 obj-$(CONFIG_IPV6_MIP6) += mip6.o
 obj-$(CONFIG_NETFILTER)        += netfilter/
 
+obj-$(CONFIG_IPV6_VTI) += ip6_vti.o
 obj-$(CONFIG_IPV6_SIT) += sit.o
 obj-$(CONFIG_IPV6_TUNNEL) += ip6_tunnel.o
 obj-$(CONFIG_IPV6_GRE) += ip6_gre.o
diff --git a/net/ipv6/ip6_vti.c b/net/ipv6/ip6_vti.c
new file mode 100644 (file)
index 0000000..ed94ba6
--- /dev/null
@@ -0,0 +1,1056 @@
+/*
+ *     IPv6 virtual tunneling interface
+ *
+ *     Copyright (C) 2013 secunet Security Networks AG
+ *
+ *     Author:
+ *     Steffen Klassert <steffen.klassert@secunet.com>
+ *
+ *     Based on:
+ *     net/ipv6/ip6_tunnel.c
+ *
+ *     This program is free software; you can redistribute it and/or
+ *     modify it under the terms of the GNU General Public License
+ *     as published by the Free Software Foundation; either version
+ *     2 of the License, or (at your option) any later version.
+ */
+
+#include <linux/module.h>
+#include <linux/capability.h>
+#include <linux/errno.h>
+#include <linux/types.h>
+#include <linux/sockios.h>
+#include <linux/icmp.h>
+#include <linux/if.h>
+#include <linux/in.h>
+#include <linux/ip.h>
+#include <linux/if_tunnel.h>
+#include <linux/net.h>
+#include <linux/in6.h>
+#include <linux/netdevice.h>
+#include <linux/if_arp.h>
+#include <linux/icmpv6.h>
+#include <linux/init.h>
+#include <linux/route.h>
+#include <linux/rtnetlink.h>
+#include <linux/netfilter_ipv6.h>
+#include <linux/slab.h>
+#include <linux/hash.h>
+
+#include <linux/uaccess.h>
+#include <linux/atomic.h>
+
+#include <net/icmp.h>
+#include <net/ip.h>
+#include <net/ip_tunnels.h>
+#include <net/ipv6.h>
+#include <net/ip6_route.h>
+#include <net/addrconf.h>
+#include <net/ip6_tunnel.h>
+#include <net/xfrm.h>
+#include <net/net_namespace.h>
+#include <net/netns/generic.h>
+
+#define HASH_SIZE_SHIFT  5
+#define HASH_SIZE (1 << HASH_SIZE_SHIFT)
+
+static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
+{
+       u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
+
+       return hash_32(hash, HASH_SIZE_SHIFT);
+}
+
+static int vti6_dev_init(struct net_device *dev);
+static void vti6_dev_setup(struct net_device *dev);
+static struct rtnl_link_ops vti6_link_ops __read_mostly;
+
+static int vti6_net_id __read_mostly;
+struct vti6_net {
+       /* the vti6 tunnel fallback device */
+       struct net_device *fb_tnl_dev;
+       /* lists for storing tunnels in use */
+       struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
+       struct ip6_tnl __rcu *tnls_wc[1];
+       struct ip6_tnl __rcu **tnls[2];
+};
+
+static struct net_device_stats *vti6_get_stats(struct net_device *dev)
+{
+       struct pcpu_tstats sum = { 0 };
+       int i;
+
+       for_each_possible_cpu(i) {
+               const struct pcpu_tstats *tstats = per_cpu_ptr(dev->tstats, i);
+
+               sum.rx_packets += tstats->rx_packets;
+               sum.rx_bytes   += tstats->rx_bytes;
+               sum.tx_packets += tstats->tx_packets;
+               sum.tx_bytes   += tstats->tx_bytes;
+       }
+       dev->stats.rx_packets = sum.rx_packets;
+       dev->stats.rx_bytes   = sum.rx_bytes;
+       dev->stats.tx_packets = sum.tx_packets;
+       dev->stats.tx_bytes   = sum.tx_bytes;
+       return &dev->stats;
+}
+
+#define for_each_vti6_tunnel_rcu(start) \
+       for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
+
+/**
+ * vti6_tnl_lookup - fetch tunnel matching the end-point addresses
+ *   @net: network namespace
+ *   @remote: the address of the tunnel exit-point
+ *   @local: the address of the tunnel entry-point
+ *
+ * Return:
+ *   tunnel matching given end-points if found,
+ *   else fallback tunnel if its device is up,
+ *   else %NULL
+ **/
+static struct ip6_tnl *
+vti6_tnl_lookup(struct net *net, const struct in6_addr *remote,
+               const struct in6_addr *local)
+{
+       unsigned int hash = HASH(remote, local);
+       struct ip6_tnl *t;
+       struct vti6_net *ip6n = net_generic(net, vti6_net_id);
+
+       for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
+               if (ipv6_addr_equal(local, &t->parms.laddr) &&
+                   ipv6_addr_equal(remote, &t->parms.raddr) &&
+                   (t->dev->flags & IFF_UP))
+                       return t;
+       }
+       t = rcu_dereference(ip6n->tnls_wc[0]);
+       if (t && (t->dev->flags & IFF_UP))
+               return t;
+
+       return NULL;
+}
+
+/**
+ * vti6_tnl_bucket - get head of list matching given tunnel parameters
+ *   @p: parameters containing tunnel end-points
+ *
+ * Description:
+ *   vti6_tnl_bucket() returns the head of the list matching the
+ *   &struct in6_addr entries laddr and raddr in @p.
+ *
+ * Return: head of IPv6 tunnel list
+ **/
+static struct ip6_tnl __rcu **
+vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
+{
+       const struct in6_addr *remote = &p->raddr;
+       const struct in6_addr *local = &p->laddr;
+       unsigned int h = 0;
+       int prio = 0;
+
+       if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
+               prio = 1;
+               h = HASH(remote, local);
+       }
+       return &ip6n->tnls[prio][h];
+}
+
+static void
+vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
+{
+       struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
+
+       rcu_assign_pointer(t->next , rtnl_dereference(*tp));
+       rcu_assign_pointer(*tp, t);
+}
+
+static void
+vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
+{
+       struct ip6_tnl __rcu **tp;
+       struct ip6_tnl *iter;
+
+       for (tp = vti6_tnl_bucket(ip6n, &t->parms);
+            (iter = rtnl_dereference(*tp)) != NULL;
+            tp = &iter->next) {
+               if (t == iter) {
+                       rcu_assign_pointer(*tp, t->next);
+                       break;
+               }
+       }
+}
+
+static void vti6_dev_free(struct net_device *dev)
+{
+       free_percpu(dev->tstats);
+       free_netdev(dev);
+}
+
+static int vti6_tnl_create2(struct net_device *dev)
+{
+       struct ip6_tnl *t = netdev_priv(dev);
+       struct net *net = dev_net(dev);
+       struct vti6_net *ip6n = net_generic(net, vti6_net_id);
+       int err;
+
+       err = vti6_dev_init(dev);
+       if (err < 0)
+               goto out;
+
+       err = register_netdevice(dev);
+       if (err < 0)
+               goto out;
+
+       strcpy(t->parms.name, dev->name);
+       dev->rtnl_link_ops = &vti6_link_ops;
+
+       dev_hold(dev);
+       vti6_tnl_link(ip6n, t);
+
+       return 0;
+
+out:
+       return err;
+}
+
+static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
+{
+       struct net_device *dev;
+       struct ip6_tnl *t;
+       char name[IFNAMSIZ];
+       int err;
+
+       if (p->name[0])
+               strlcpy(name, p->name, IFNAMSIZ);
+       else
+               sprintf(name, "ip6_vti%%d");
+
+       dev = alloc_netdev(sizeof(*t), name, vti6_dev_setup);
+       if (dev == NULL)
+               goto failed;
+
+       dev_net_set(dev, net);
+
+       t = netdev_priv(dev);
+       t->parms = *p;
+       t->net = dev_net(dev);
+
+       err = vti6_tnl_create2(dev);
+       if (err < 0)
+               goto failed_free;
+
+       return t;
+
+failed_free:
+       vti6_dev_free(dev);
+failed:
+       return NULL;
+}
+
+/**
+ * vti6_locate - find or create tunnel matching given parameters
+ *   @net: network namespace
+ *   @p: tunnel parameters
+ *   @create: != 0 if allowed to create new tunnel if no match found
+ *
+ * Description:
+ *   vti6_locate() first tries to locate an existing tunnel
+ *   based on @parms. If this is unsuccessful, but @create is set a new
+ *   tunnel device is created and registered for use.
+ *
+ * Return:
+ *   matching tunnel or NULL
+ **/
+static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
+                                  int create)
+{
+       const struct in6_addr *remote = &p->raddr;
+       const struct in6_addr *local = &p->laddr;
+       struct ip6_tnl __rcu **tp;
+       struct ip6_tnl *t;
+       struct vti6_net *ip6n = net_generic(net, vti6_net_id);
+
+       for (tp = vti6_tnl_bucket(ip6n, p);
+            (t = rtnl_dereference(*tp)) != NULL;
+            tp = &t->next) {
+               if (ipv6_addr_equal(local, &t->parms.laddr) &&
+                   ipv6_addr_equal(remote, &t->parms.raddr))
+                       return t;
+       }
+       if (!create)
+               return NULL;
+       return vti6_tnl_create(net, p);
+}
+
+/**
+ * vti6_dev_uninit - tunnel device uninitializer
+ *   @dev: the device to be destroyed
+ *
+ * Description:
+ *   vti6_dev_uninit() removes tunnel from its list
+ **/
+static void vti6_dev_uninit(struct net_device *dev)
+{
+       struct ip6_tnl *t = netdev_priv(dev);
+       struct net *net = dev_net(dev);
+       struct vti6_net *ip6n = net_generic(net, vti6_net_id);
+
+       if (dev == ip6n->fb_tnl_dev)
+               RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
+       else
+               vti6_tnl_unlink(ip6n, t);
+       ip6_tnl_dst_reset(t);
+       dev_put(dev);
+}
+
+static int vti6_rcv(struct sk_buff *skb)
+{
+       struct ip6_tnl *t;
+       const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
+
+       rcu_read_lock();
+
+       if ((t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr,
+                                &ipv6h->daddr)) != NULL) {
+               struct pcpu_tstats *tstats;
+
+               if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
+                       rcu_read_unlock();
+                       goto discard;
+               }
+
+               if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
+                       rcu_read_unlock();
+                       return 0;
+               }
+
+               if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
+                       t->dev->stats.rx_dropped++;
+                       rcu_read_unlock();
+                       goto discard;
+               }
+
+               tstats = this_cpu_ptr(t->dev->tstats);
+               tstats->rx_packets++;
+               tstats->rx_bytes += skb->len;
+
+               skb->mark = 0;
+               secpath_reset(skb);
+               skb->dev = t->dev;
+
+               rcu_read_unlock();
+               return 0;
+       }
+       rcu_read_unlock();
+       return 1;
+
+discard:
+       kfree_skb(skb);
+       return 0;
+}
+
+/**
+ * vti6_addr_conflict - compare packet addresses to tunnel's own
+ *   @t: the outgoing tunnel device
+ *   @hdr: IPv6 header from the incoming packet
+ *
+ * Description:
+ *   Avoid trivial tunneling loop by checking that tunnel exit-point
+ *   doesn't match source of incoming packet.
+ *
+ * Return:
+ *   1 if conflict,
+ *   0 else
+ **/
+static inline bool
+vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
+{
+       return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
+}
+
+/**
+ * vti6_xmit - send a packet
+ *   @skb: the outgoing socket buffer
+ *   @dev: the outgoing tunnel device
+ **/
+static int vti6_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+       struct net *net = dev_net(dev);
+       struct ip6_tnl *t = netdev_priv(dev);
+       struct net_device_stats *stats = &t->dev->stats;
+       struct dst_entry *dst = NULL, *ndst = NULL;
+       struct flowi6 fl6;
+       struct ipv6hdr *ipv6h = ipv6_hdr(skb);
+       struct net_device *tdev;
+       int err = -1;
+
+       if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
+           !ip6_tnl_xmit_ctl(t) || vti6_addr_conflict(t, ipv6h))
+               return err;
+
+       dst = ip6_tnl_dst_check(t);
+       if (!dst) {
+               memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
+
+               ndst = ip6_route_output(net, NULL, &fl6);
+
+               if (ndst->error)
+                       goto tx_err_link_failure;
+               ndst = xfrm_lookup(net, ndst, flowi6_to_flowi(&fl6), NULL, 0);
+               if (IS_ERR(ndst)) {
+                       err = PTR_ERR(ndst);
+                       ndst = NULL;
+                       goto tx_err_link_failure;
+               }
+               dst = ndst;
+       }
+
+       if (!dst->xfrm || dst->xfrm->props.mode != XFRM_MODE_TUNNEL)
+               goto tx_err_link_failure;
+
+       tdev = dst->dev;
+
+       if (tdev == dev) {
+               stats->collisions++;
+               net_warn_ratelimited("%s: Local routing loop detected!\n",
+                                    t->parms.name);
+               goto tx_err_dst_release;
+       }
+
+
+       skb_dst_drop(skb);
+       skb_dst_set_noref(skb, dst);
+
+       ip6tunnel_xmit(skb, dev);
+       if (ndst) {
+               dev->mtu = dst_mtu(ndst);
+               ip6_tnl_dst_store(t, ndst);
+       }
+
+       return 0;
+tx_err_link_failure:
+       stats->tx_carrier_errors++;
+       dst_link_failure(skb);
+tx_err_dst_release:
+       dst_release(ndst);
+       return err;
+}
+
+static netdev_tx_t
+vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+       struct ip6_tnl *t = netdev_priv(dev);
+       struct net_device_stats *stats = &t->dev->stats;
+       int ret;
+
+       switch (skb->protocol) {
+       case htons(ETH_P_IPV6):
+               ret = vti6_xmit(skb, dev);
+               break;
+       default:
+               goto tx_err;
+       }
+
+       if (ret < 0)
+               goto tx_err;
+
+       return NETDEV_TX_OK;
+
+tx_err:
+       stats->tx_errors++;
+       stats->tx_dropped++;
+       kfree_skb(skb);
+       return NETDEV_TX_OK;
+}
+
+static void vti6_link_config(struct ip6_tnl *t)
+{
+       struct dst_entry *dst;
+       struct net_device *dev = t->dev;
+       struct __ip6_tnl_parm *p = &t->parms;
+       struct flowi6 *fl6 = &t->fl.u.ip6;
+
+       memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
+       memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
+
+       /* Set up flowi template */
+       fl6->saddr = p->laddr;
+       fl6->daddr = p->raddr;
+       fl6->flowi6_oif = p->link;
+       fl6->flowi6_mark = be32_to_cpu(p->i_key);
+       fl6->flowi6_proto = p->proto;
+       fl6->flowlabel = 0;
+
+       p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
+                     IP6_TNL_F_CAP_PER_PACKET);
+       p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
+
+       if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
+               dev->flags |= IFF_POINTOPOINT;
+       else
+               dev->flags &= ~IFF_POINTOPOINT;
+
+       dev->iflink = p->link;
+
+       if (p->flags & IP6_TNL_F_CAP_XMIT) {
+
+               dst = ip6_route_output(dev_net(dev), NULL, fl6);
+               if (dst->error)
+                       return;
+
+               dst = xfrm_lookup(dev_net(dev), dst, flowi6_to_flowi(fl6),
+                                 NULL, 0);
+               if (IS_ERR(dst))
+                       return;
+
+               if (dst->dev) {
+                       dev->hard_header_len = dst->dev->hard_header_len;
+
+                       dev->mtu = dst_mtu(dst);
+
+                       if (dev->mtu < IPV6_MIN_MTU)
+                               dev->mtu = IPV6_MIN_MTU;
+               }
+               dst_release(dst);
+       }
+}
+
+/**
+ * vti6_tnl_change - update the tunnel parameters
+ *   @t: tunnel to be changed
+ *   @p: tunnel configuration parameters
+ *
+ * Description:
+ *   vti6_tnl_change() updates the tunnel parameters
+ **/
+static int
+vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
+{
+       t->parms.laddr = p->laddr;
+       t->parms.raddr = p->raddr;
+       t->parms.link = p->link;
+       t->parms.i_key = p->i_key;
+       t->parms.o_key = p->o_key;
+       t->parms.proto = p->proto;
+       ip6_tnl_dst_reset(t);
+       vti6_link_config(t);
+       return 0;
+}
+
+static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
+{
+       struct net *net = dev_net(t->dev);
+       struct vti6_net *ip6n = net_generic(net, vti6_net_id);
+       int err;
+
+       vti6_tnl_unlink(ip6n, t);
+       synchronize_net();
+       err = vti6_tnl_change(t, p);
+       vti6_tnl_link(ip6n, t);
+       netdev_state_change(t->dev);
+       return err;
+}
+
+static void
+vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
+{
+       p->laddr = u->laddr;
+       p->raddr = u->raddr;
+       p->link = u->link;
+       p->i_key = u->i_key;
+       p->o_key = u->o_key;
+       p->proto = u->proto;
+
+       memcpy(p->name, u->name, sizeof(u->name));
+}
+
+static void
+vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
+{
+       u->laddr = p->laddr;
+       u->raddr = p->raddr;
+       u->link = p->link;
+       u->i_key = p->i_key;
+       u->o_key = p->o_key;
+       u->proto = p->proto;
+
+       memcpy(u->name, p->name, sizeof(u->name));
+}
+
+/**
+ * vti6_tnl_ioctl - configure vti6 tunnels from userspace
+ *   @dev: virtual device associated with tunnel
+ *   @ifr: parameters passed from userspace
+ *   @cmd: command to be performed
+ *
+ * Description:
+ *   vti6_ioctl() is used for managing vti6 tunnels
+ *   from userspace.
+ *
+ *   The possible commands are the following:
+ *     %SIOCGETTUNNEL: get tunnel parameters for device
+ *     %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
+ *     %SIOCCHGTUNNEL: change tunnel parameters to those given
+ *     %SIOCDELTUNNEL: delete tunnel
+ *
+ *   The fallback device "ip6_vti0", created during module
+ *   initialization, can be used for creating other tunnel devices.
+ *
+ * Return:
+ *   0 on success,
+ *   %-EFAULT if unable to copy data to or from userspace,
+ *   %-EPERM if current process hasn't %CAP_NET_ADMIN set
+ *   %-EINVAL if passed tunnel parameters are invalid,
+ *   %-EEXIST if changing a tunnel's parameters would cause a conflict
+ *   %-ENODEV if attempting to change or delete a nonexisting device
+ **/
+static int
+vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
+{
+       int err = 0;
+       struct ip6_tnl_parm2 p;
+       struct __ip6_tnl_parm p1;
+       struct ip6_tnl *t = NULL;
+       struct net *net = dev_net(dev);
+       struct vti6_net *ip6n = net_generic(net, vti6_net_id);
+
+       switch (cmd) {
+       case SIOCGETTUNNEL:
+               if (dev == ip6n->fb_tnl_dev) {
+                       if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
+                               err = -EFAULT;
+                               break;
+                       }
+                       vti6_parm_from_user(&p1, &p);
+                       t = vti6_locate(net, &p1, 0);
+               } else {
+                       memset(&p, 0, sizeof(p));
+               }
+               if (t == NULL)
+                       t = netdev_priv(dev);
+               vti6_parm_to_user(&p, &t->parms);
+               if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
+                       err = -EFAULT;
+               break;
+       case SIOCADDTUNNEL:
+       case SIOCCHGTUNNEL:
+               err = -EPERM;
+               if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
+                       break;
+               err = -EFAULT;
+               if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
+                       break;
+               err = -EINVAL;
+               if (p.proto != IPPROTO_IPV6  && p.proto != 0)
+                       break;
+               vti6_parm_from_user(&p1, &p);
+               t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
+               if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
+                       if (t != NULL) {
+                               if (t->dev != dev) {
+                                       err = -EEXIST;
+                                       break;
+                               }
+                       } else
+                               t = netdev_priv(dev);
+
+                       err = vti6_update(t, &p1);
+               }
+               if (t) {
+                       err = 0;
+                       vti6_parm_to_user(&p, &t->parms);
+                       if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
+                               err = -EFAULT;
+
+               } else
+                       err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
+               break;
+       case SIOCDELTUNNEL:
+               err = -EPERM;
+               if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
+                       break;
+
+               if (dev == ip6n->fb_tnl_dev) {
+                       err = -EFAULT;
+                       if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
+                               break;
+                       err = -ENOENT;
+                       vti6_parm_from_user(&p1, &p);
+                       t = vti6_locate(net, &p1, 0);
+                       if (t == NULL)
+                               break;
+                       err = -EPERM;
+                       if (t->dev == ip6n->fb_tnl_dev)
+                               break;
+                       dev = t->dev;
+               }
+               err = 0;
+               unregister_netdevice(dev);
+               break;
+       default:
+               err = -EINVAL;
+       }
+       return err;
+}
+
+/**
+ * vti6_tnl_change_mtu - change mtu manually for tunnel device
+ *   @dev: virtual device associated with tunnel
+ *   @new_mtu: the new mtu
+ *
+ * Return:
+ *   0 on success,
+ *   %-EINVAL if mtu too small
+ **/
+static int vti6_change_mtu(struct net_device *dev, int new_mtu)
+{
+       if (new_mtu < IPV6_MIN_MTU)
+               return -EINVAL;
+
+       dev->mtu = new_mtu;
+       return 0;
+}
+
+static const struct net_device_ops vti6_netdev_ops = {
+       .ndo_uninit     = vti6_dev_uninit,
+       .ndo_start_xmit = vti6_tnl_xmit,
+       .ndo_do_ioctl   = vti6_ioctl,
+       .ndo_change_mtu = vti6_change_mtu,
+       .ndo_get_stats  = vti6_get_stats,
+};
+
+/**
+ * vti6_dev_setup - setup virtual tunnel device
+ *   @dev: virtual device associated with tunnel
+ *
+ * Description:
+ *   Initialize function pointers and device parameters
+ **/
+static void vti6_dev_setup(struct net_device *dev)
+{
+       struct ip6_tnl *t;
+
+       dev->netdev_ops = &vti6_netdev_ops;
+       dev->destructor = vti6_dev_free;
+
+       dev->type = ARPHRD_TUNNEL6;
+       dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
+       dev->mtu = ETH_DATA_LEN;
+       t = netdev_priv(dev);
+       dev->flags |= IFF_NOARP;
+       dev->addr_len = sizeof(struct in6_addr);
+       dev->features |= NETIF_F_NETNS_LOCAL;
+       dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
+}
+
+/**
+ * vti6_dev_init_gen - general initializer for all tunnel devices
+ *   @dev: virtual device associated with tunnel
+ **/
+static inline int vti6_dev_init_gen(struct net_device *dev)
+{
+       struct ip6_tnl *t = netdev_priv(dev);
+
+       t->dev = dev;
+       t->net = dev_net(dev);
+       dev->tstats = alloc_percpu(struct pcpu_tstats);
+       if (!dev->tstats)
+               return -ENOMEM;
+       return 0;
+}
+
+/**
+ * vti6_dev_init - initializer for all non fallback tunnel devices
+ *   @dev: virtual device associated with tunnel
+ **/
+static int vti6_dev_init(struct net_device *dev)
+{
+       struct ip6_tnl *t = netdev_priv(dev);
+       int err = vti6_dev_init_gen(dev);
+
+       if (err)
+               return err;
+       vti6_link_config(t);
+       return 0;
+}
+
+/**
+ * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
+ *   @dev: fallback device
+ *
+ * Return: 0
+ **/
+static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
+{
+       struct ip6_tnl *t = netdev_priv(dev);
+       struct net *net = dev_net(dev);
+       struct vti6_net *ip6n = net_generic(net, vti6_net_id);
+       int err = vti6_dev_init_gen(dev);
+
+       if (err)
+               return err;
+
+       t->parms.proto = IPPROTO_IPV6;
+       dev_hold(dev);
+
+       vti6_link_config(t);
+
+       rcu_assign_pointer(ip6n->tnls_wc[0], t);
+       return 0;
+}
+
+static int vti6_validate(struct nlattr *tb[], struct nlattr *data[])
+{
+       return 0;
+}
+
+static void vti6_netlink_parms(struct nlattr *data[],
+                              struct __ip6_tnl_parm *parms)
+{
+       memset(parms, 0, sizeof(*parms));
+
+       if (!data)
+               return;
+
+       if (data[IFLA_VTI_LINK])
+               parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
+
+       if (data[IFLA_VTI_LOCAL])
+               nla_memcpy(&parms->laddr, data[IFLA_VTI_LOCAL],
+                          sizeof(struct in6_addr));
+
+       if (data[IFLA_VTI_REMOTE])
+               nla_memcpy(&parms->raddr, data[IFLA_VTI_REMOTE],
+                          sizeof(struct in6_addr));
+
+       if (data[IFLA_VTI_IKEY])
+               parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
+
+       if (data[IFLA_VTI_OKEY])
+               parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
+}
+
+static int vti6_newlink(struct net *src_net, struct net_device *dev,
+                       struct nlattr *tb[], struct nlattr *data[])
+{
+       struct net *net = dev_net(dev);
+       struct ip6_tnl *nt;
+
+       nt = netdev_priv(dev);
+       vti6_netlink_parms(data, &nt->parms);
+
+       nt->parms.proto = IPPROTO_IPV6;
+
+       if (vti6_locate(net, &nt->parms, 0))
+               return -EEXIST;
+
+       return vti6_tnl_create2(dev);
+}
+
+static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
+                          struct nlattr *data[])
+{
+       struct ip6_tnl *t;
+       struct __ip6_tnl_parm p;
+       struct net *net = dev_net(dev);
+       struct vti6_net *ip6n = net_generic(net, vti6_net_id);
+
+       if (dev == ip6n->fb_tnl_dev)
+               return -EINVAL;
+
+       vti6_netlink_parms(data, &p);
+
+       t = vti6_locate(net, &p, 0);
+
+       if (t) {
+               if (t->dev != dev)
+                       return -EEXIST;
+       } else
+               t = netdev_priv(dev);
+
+       return vti6_update(t, &p);
+}
+
+static size_t vti6_get_size(const struct net_device *dev)
+{
+       return
+               /* IFLA_VTI_LINK */
+               nla_total_size(4) +
+               /* IFLA_VTI_LOCAL */
+               nla_total_size(sizeof(struct in6_addr)) +
+               /* IFLA_VTI_REMOTE */
+               nla_total_size(sizeof(struct in6_addr)) +
+               /* IFLA_VTI_IKEY */
+               nla_total_size(4) +
+               /* IFLA_VTI_OKEY */
+               nla_total_size(4) +
+               0;
+}
+
+static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
+{
+       struct ip6_tnl *tunnel = netdev_priv(dev);
+       struct __ip6_tnl_parm *parm = &tunnel->parms;
+
+       if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
+           nla_put(skb, IFLA_VTI_LOCAL, sizeof(struct in6_addr),
+                   &parm->laddr) ||
+           nla_put(skb, IFLA_VTI_REMOTE, sizeof(struct in6_addr),
+                   &parm->raddr) ||
+           nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
+           nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key))
+               goto nla_put_failure;
+       return 0;
+
+nla_put_failure:
+       return -EMSGSIZE;
+}
+
+static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
+       [IFLA_VTI_LINK]         = { .type = NLA_U32 },
+       [IFLA_VTI_LOCAL]        = { .len = sizeof(struct in6_addr) },
+       [IFLA_VTI_REMOTE]       = { .len = sizeof(struct in6_addr) },
+       [IFLA_VTI_IKEY]         = { .type = NLA_U32 },
+       [IFLA_VTI_OKEY]         = { .type = NLA_U32 },
+};
+
+static struct rtnl_link_ops vti6_link_ops __read_mostly = {
+       .kind           = "vti6",
+       .maxtype        = IFLA_VTI_MAX,
+       .policy         = vti6_policy,
+       .priv_size      = sizeof(struct ip6_tnl),
+       .setup          = vti6_dev_setup,
+       .validate       = vti6_validate,
+       .newlink        = vti6_newlink,
+       .changelink     = vti6_changelink,
+       .get_size       = vti6_get_size,
+       .fill_info      = vti6_fill_info,
+};
+
+static struct xfrm_tunnel_notifier vti6_handler __read_mostly = {
+       .handler        = vti6_rcv,
+       .priority       =       1,
+};
+
+static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
+{
+       int h;
+       struct ip6_tnl *t;
+       LIST_HEAD(list);
+
+       for (h = 0; h < HASH_SIZE; h++) {
+               t = rtnl_dereference(ip6n->tnls_r_l[h]);
+               while (t != NULL) {
+                       unregister_netdevice_queue(t->dev, &list);
+                       t = rtnl_dereference(t->next);
+               }
+       }
+
+       t = rtnl_dereference(ip6n->tnls_wc[0]);
+       unregister_netdevice_queue(t->dev, &list);
+       unregister_netdevice_many(&list);
+}
+
+static int __net_init vti6_init_net(struct net *net)
+{
+       struct vti6_net *ip6n = net_generic(net, vti6_net_id);
+       struct ip6_tnl *t = NULL;
+       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), "ip6_vti0",
+                                       vti6_dev_setup);
+
+       if (!ip6n->fb_tnl_dev)
+               goto err_alloc_dev;
+       dev_net_set(ip6n->fb_tnl_dev, net);
+
+       err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
+       if (err < 0)
+               goto err_register;
+
+       err = register_netdev(ip6n->fb_tnl_dev);
+       if (err < 0)
+               goto err_register;
+
+       t = netdev_priv(ip6n->fb_tnl_dev);
+
+       strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
+       return 0;
+
+err_register:
+       vti6_dev_free(ip6n->fb_tnl_dev);
+err_alloc_dev:
+       return err;
+}
+
+static void __net_exit vti6_exit_net(struct net *net)
+{
+       struct vti6_net *ip6n = net_generic(net, vti6_net_id);
+
+       rtnl_lock();
+       vti6_destroy_tunnels(ip6n);
+       rtnl_unlock();
+}
+
+static struct pernet_operations vti6_net_ops = {
+       .init = vti6_init_net,
+       .exit = vti6_exit_net,
+       .id   = &vti6_net_id,
+       .size = sizeof(struct vti6_net),
+};
+
+/**
+ * vti6_tunnel_init - register protocol and reserve needed resources
+ *
+ * Return: 0 on success
+ **/
+static int __init vti6_tunnel_init(void)
+{
+       int  err;
+
+       err = register_pernet_device(&vti6_net_ops);
+       if (err < 0)
+               goto out_pernet;
+
+       err = xfrm6_mode_tunnel_input_register(&vti6_handler);
+       if (err < 0) {
+               pr_err("%s: can't register vti6\n", __func__);
+               goto out;
+       }
+       err = rtnl_link_register(&vti6_link_ops);
+       if (err < 0)
+               goto rtnl_link_failed;
+
+       return 0;
+
+rtnl_link_failed:
+       xfrm6_mode_tunnel_input_deregister(&vti6_handler);
+out:
+       unregister_pernet_device(&vti6_net_ops);
+out_pernet:
+       return err;
+}
+
+/**
+ * vti6_tunnel_cleanup - free resources and unregister protocol
+ **/
+static void __exit vti6_tunnel_cleanup(void)
+{
+       rtnl_link_unregister(&vti6_link_ops);
+       if (xfrm6_mode_tunnel_input_deregister(&vti6_handler))
+               pr_info("%s: can't deregister vti6\n", __func__);
+
+       unregister_pernet_device(&vti6_net_ops);
+}
+
+module_init(vti6_tunnel_init);
+module_exit(vti6_tunnel_cleanup);
+MODULE_LICENSE("GPL");
+MODULE_ALIAS_RTNL_LINK("vti6");
+MODULE_ALIAS_NETDEV("ip6_vti0");
+MODULE_AUTHOR("Steffen Klassert");
+MODULE_DESCRIPTION("IPv6 virtual tunnel interface");