2 * IPv6 virtual tunneling interface
4 * Copyright (C) 2013 secunet Security Networks AG
7 * Steffen Klassert <steffen.klassert@secunet.com>
10 * net/ipv6/ip6_tunnel.c
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
18 #include <linux/module.h>
19 #include <linux/capability.h>
20 #include <linux/errno.h>
21 #include <linux/types.h>
22 #include <linux/sockios.h>
23 #include <linux/icmp.h>
27 #include <linux/net.h>
28 #include <linux/in6.h>
29 #include <linux/netdevice.h>
30 #include <linux/if_arp.h>
31 #include <linux/icmpv6.h>
32 #include <linux/init.h>
33 #include <linux/route.h>
34 #include <linux/rtnetlink.h>
35 #include <linux/netfilter_ipv6.h>
36 #include <linux/slab.h>
37 #include <linux/hash.h>
39 #include <linux/uaccess.h>
40 #include <linux/atomic.h>
44 #include <net/ip_tunnels.h>
46 #include <net/ip6_route.h>
47 #include <net/addrconf.h>
48 #include <net/ip6_tunnel.h>
50 #include <net/net_namespace.h>
51 #include <net/netns/generic.h>
53 #define HASH_SIZE_SHIFT 5
54 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
56 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
58 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
60 return hash_32(hash, HASH_SIZE_SHIFT);
63 static int vti6_dev_init(struct net_device *dev);
64 static void vti6_dev_setup(struct net_device *dev);
65 static struct rtnl_link_ops vti6_link_ops __read_mostly;
67 static int vti6_net_id __read_mostly;
69 /* the vti6 tunnel fallback device */
70 struct net_device *fb_tnl_dev;
71 /* lists for storing tunnels in use */
72 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
73 struct ip6_tnl __rcu *tnls_wc[1];
74 struct ip6_tnl __rcu **tnls[2];
77 #define for_each_vti6_tunnel_rcu(start) \
78 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
81 * vti6_tnl_lookup - fetch tunnel matching the end-point addresses
82 * @net: network namespace
83 * @remote: the address of the tunnel exit-point
84 * @local: the address of the tunnel entry-point
87 * tunnel matching given end-points if found,
88 * else fallback tunnel if its device is up,
91 static struct ip6_tnl *
92 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote,
93 const struct in6_addr *local)
95 unsigned int hash = HASH(remote, local);
97 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
100 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
101 if (ipv6_addr_equal(local, &t->parms.laddr) &&
102 ipv6_addr_equal(remote, &t->parms.raddr) &&
103 (t->dev->flags & IFF_UP))
107 memset(&any, 0, sizeof(any));
108 hash = HASH(&any, local);
109 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
110 if (ipv6_addr_equal(local, &t->parms.laddr) &&
111 (t->dev->flags & IFF_UP))
115 hash = HASH(remote, &any);
116 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
117 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
118 (t->dev->flags & IFF_UP))
122 t = rcu_dereference(ip6n->tnls_wc[0]);
123 if (t && (t->dev->flags & IFF_UP))
130 * vti6_tnl_bucket - get head of list matching given tunnel parameters
131 * @p: parameters containing tunnel end-points
134 * vti6_tnl_bucket() returns the head of the list matching the
135 * &struct in6_addr entries laddr and raddr in @p.
137 * Return: head of IPv6 tunnel list
139 static struct ip6_tnl __rcu **
140 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
142 const struct in6_addr *remote = &p->raddr;
143 const struct in6_addr *local = &p->laddr;
147 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
149 h = HASH(remote, local);
151 return &ip6n->tnls[prio][h];
155 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
157 struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
159 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
160 rcu_assign_pointer(*tp, t);
164 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
166 struct ip6_tnl __rcu **tp;
167 struct ip6_tnl *iter;
169 for (tp = vti6_tnl_bucket(ip6n, &t->parms);
170 (iter = rtnl_dereference(*tp)) != NULL;
173 rcu_assign_pointer(*tp, t->next);
179 static void vti6_dev_free(struct net_device *dev)
181 free_percpu(dev->tstats);
185 static int vti6_tnl_create2(struct net_device *dev)
187 struct ip6_tnl *t = netdev_priv(dev);
188 struct net *net = dev_net(dev);
189 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
192 dev->rtnl_link_ops = &vti6_link_ops;
193 err = register_netdevice(dev);
197 strcpy(t->parms.name, dev->name);
200 vti6_tnl_link(ip6n, t);
208 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
210 struct net_device *dev;
216 if (!dev_valid_name(p->name))
218 strlcpy(name, p->name, IFNAMSIZ);
220 sprintf(name, "ip6_vti%%d");
223 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup);
227 dev_net_set(dev, net);
229 t = netdev_priv(dev);
231 t->net = dev_net(dev);
233 err = vti6_tnl_create2(dev);
246 * vti6_locate - find or create tunnel matching given parameters
247 * @net: network namespace
248 * @p: tunnel parameters
249 * @create: != 0 if allowed to create new tunnel if no match found
252 * vti6_locate() first tries to locate an existing tunnel
253 * based on @parms. If this is unsuccessful, but @create is set a new
254 * tunnel device is created and registered for use.
257 * matching tunnel or NULL
259 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
262 const struct in6_addr *remote = &p->raddr;
263 const struct in6_addr *local = &p->laddr;
264 struct ip6_tnl __rcu **tp;
266 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
268 for (tp = vti6_tnl_bucket(ip6n, p);
269 (t = rtnl_dereference(*tp)) != NULL;
271 if (ipv6_addr_equal(local, &t->parms.laddr) &&
272 ipv6_addr_equal(remote, &t->parms.raddr)) {
281 return vti6_tnl_create(net, p);
285 * vti6_dev_uninit - tunnel device uninitializer
286 * @dev: the device to be destroyed
289 * vti6_dev_uninit() removes tunnel from its list
291 static void vti6_dev_uninit(struct net_device *dev)
293 struct ip6_tnl *t = netdev_priv(dev);
294 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id);
296 if (dev == ip6n->fb_tnl_dev)
297 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
299 vti6_tnl_unlink(ip6n, t);
303 static int vti6_rcv(struct sk_buff *skb)
306 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
309 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
311 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
316 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
321 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
322 t->dev->stats.rx_dropped++;
327 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t;
331 return xfrm6_rcv(skb);
340 static int vti6_rcv_cb(struct sk_buff *skb, int err)
342 unsigned short family;
343 struct net_device *dev;
344 struct pcpu_sw_netstats *tstats;
345 struct xfrm_state *x;
346 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
347 u32 orig_mark = skb->mark;
356 dev->stats.rx_errors++;
357 dev->stats.rx_dropped++;
362 x = xfrm_input_state(skb);
363 family = x->inner_mode->afinfo->family;
365 skb->mark = be32_to_cpu(t->parms.i_key);
366 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
367 skb->mark = orig_mark;
372 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
375 tstats = this_cpu_ptr(dev->tstats);
376 u64_stats_update_begin(&tstats->syncp);
377 tstats->rx_packets++;
378 tstats->rx_bytes += skb->len;
379 u64_stats_update_end(&tstats->syncp);
385 * vti6_addr_conflict - compare packet addresses to tunnel's own
386 * @t: the outgoing tunnel device
387 * @hdr: IPv6 header from the incoming packet
390 * Avoid trivial tunneling loop by checking that tunnel exit-point
391 * doesn't match source of incoming packet.
398 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
400 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
403 static bool vti6_state_check(const struct xfrm_state *x,
404 const struct in6_addr *dst,
405 const struct in6_addr *src)
407 xfrm_address_t *daddr = (xfrm_address_t *)dst;
408 xfrm_address_t *saddr = (xfrm_address_t *)src;
410 /* if there is no transform then this tunnel is not functional.
411 * Or if the xfrm is not mode tunnel.
413 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
414 x->props.family != AF_INET6)
417 if (ipv6_addr_any(dst))
418 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
420 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
427 * vti6_xmit - send a packet
428 * @skb: the outgoing socket buffer
429 * @dev: the outgoing tunnel device
430 * @fl: the flow informations for the xfrm_lookup
433 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
435 struct ip6_tnl *t = netdev_priv(dev);
436 struct net_device_stats *stats = &t->dev->stats;
437 struct dst_entry *dst = skb_dst(skb);
438 struct net_device *tdev;
439 struct xfrm_state *x;
440 int pkt_len = skb->len;
445 goto tx_err_link_failure;
448 dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
452 goto tx_err_link_failure;
456 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
457 goto tx_err_link_failure;
459 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
460 (const struct in6_addr *)&x->id.daddr))
461 goto tx_err_link_failure;
467 net_warn_ratelimited("%s: Local routing loop detected!\n",
469 goto tx_err_dst_release;
473 if (skb->len > mtu) {
474 skb_dst(skb)->ops->update_pmtu(dst, NULL, skb, mtu);
476 if (skb->protocol == htons(ETH_P_IPV6)) {
477 if (mtu < IPV6_MIN_MTU)
480 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
482 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
487 goto tx_err_dst_release;
490 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
491 skb_dst_set(skb, dst);
492 skb->dev = skb_dst(skb)->dev;
494 err = dst_output(t->net, skb->sk, skb);
495 if (net_xmit_eval(err) == 0) {
496 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
498 u64_stats_update_begin(&tstats->syncp);
499 tstats->tx_bytes += pkt_len;
500 tstats->tx_packets++;
501 u64_stats_update_end(&tstats->syncp);
504 stats->tx_aborted_errors++;
509 stats->tx_carrier_errors++;
510 dst_link_failure(skb);
517 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
519 struct ip6_tnl *t = netdev_priv(dev);
520 struct net_device_stats *stats = &t->dev->stats;
521 struct ipv6hdr *ipv6h;
525 memset(&fl, 0, sizeof(fl));
527 switch (skb->protocol) {
528 case htons(ETH_P_IPV6):
529 ipv6h = ipv6_hdr(skb);
531 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
532 vti6_addr_conflict(t, ipv6h))
535 xfrm_decode_session(skb, &fl, AF_INET6);
536 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
538 case htons(ETH_P_IP):
539 xfrm_decode_session(skb, &fl, AF_INET);
540 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
546 /* override mark with tunnel output key */
547 fl.flowi_mark = be32_to_cpu(t->parms.o_key);
549 ret = vti6_xmit(skb, dev, &fl);
562 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
563 u8 type, u8 code, int offset, __be32 info)
567 struct xfrm_state *x;
569 struct ip_esp_hdr *esph;
570 struct ip_auth_hdr *ah;
571 struct ip_comp_hdr *ipch;
572 struct net *net = dev_net(skb->dev);
573 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
574 int protocol = iph->nexthdr;
576 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
580 mark = be32_to_cpu(t->parms.o_key);
584 esph = (struct ip_esp_hdr *)(skb->data + offset);
588 ah = (struct ip_auth_hdr *)(skb->data + offset);
592 ipch = (struct ip_comp_hdr *)(skb->data + offset);
593 spi = htonl(ntohs(ipch->cpi));
599 if (type != ICMPV6_PKT_TOOBIG &&
600 type != NDISC_REDIRECT)
603 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
604 spi, protocol, AF_INET6);
608 if (type == NDISC_REDIRECT)
609 ip6_redirect(skb, net, skb->dev->ifindex, 0,
610 sock_net_uid(net, NULL));
612 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
618 static void vti6_link_config(struct ip6_tnl *t)
620 struct net_device *dev = t->dev;
621 struct __ip6_tnl_parm *p = &t->parms;
623 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
624 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
626 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
627 IP6_TNL_F_CAP_PER_PACKET);
628 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
630 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
631 dev->flags |= IFF_POINTOPOINT;
633 dev->flags &= ~IFF_POINTOPOINT;
637 * vti6_tnl_change - update the tunnel parameters
638 * @t: tunnel to be changed
639 * @p: tunnel configuration parameters
642 * vti6_tnl_change() updates the tunnel parameters
645 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
647 t->parms.laddr = p->laddr;
648 t->parms.raddr = p->raddr;
649 t->parms.link = p->link;
650 t->parms.i_key = p->i_key;
651 t->parms.o_key = p->o_key;
652 t->parms.proto = p->proto;
653 dst_cache_reset(&t->dst_cache);
658 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
660 struct net *net = dev_net(t->dev);
661 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
664 vti6_tnl_unlink(ip6n, t);
666 err = vti6_tnl_change(t, p);
667 vti6_tnl_link(ip6n, t);
668 netdev_state_change(t->dev);
673 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
682 memcpy(p->name, u->name, sizeof(u->name));
686 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
694 u->i_flags |= GRE_KEY;
696 u->o_flags |= GRE_KEY;
699 memcpy(u->name, p->name, sizeof(u->name));
703 * vti6_tnl_ioctl - configure vti6 tunnels from userspace
704 * @dev: virtual device associated with tunnel
705 * @ifr: parameters passed from userspace
706 * @cmd: command to be performed
709 * vti6_ioctl() is used for managing vti6 tunnels
712 * The possible commands are the following:
713 * %SIOCGETTUNNEL: get tunnel parameters for device
714 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
715 * %SIOCCHGTUNNEL: change tunnel parameters to those given
716 * %SIOCDELTUNNEL: delete tunnel
718 * The fallback device "ip6_vti0", created during module
719 * initialization, can be used for creating other tunnel devices.
723 * %-EFAULT if unable to copy data to or from userspace,
724 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
725 * %-EINVAL if passed tunnel parameters are invalid,
726 * %-EEXIST if changing a tunnel's parameters would cause a conflict
727 * %-ENODEV if attempting to change or delete a nonexisting device
730 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
733 struct ip6_tnl_parm2 p;
734 struct __ip6_tnl_parm p1;
735 struct ip6_tnl *t = NULL;
736 struct net *net = dev_net(dev);
737 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
741 if (dev == ip6n->fb_tnl_dev) {
742 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
746 vti6_parm_from_user(&p1, &p);
747 t = vti6_locate(net, &p1, 0);
749 memset(&p, 0, sizeof(p));
752 t = netdev_priv(dev);
753 vti6_parm_to_user(&p, &t->parms);
754 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
760 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
763 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
766 if (p.proto != IPPROTO_IPV6 && p.proto != 0)
768 vti6_parm_from_user(&p1, &p);
769 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
770 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
777 t = netdev_priv(dev);
779 err = vti6_update(t, &p1);
783 vti6_parm_to_user(&p, &t->parms);
784 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
788 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
792 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
795 if (dev == ip6n->fb_tnl_dev) {
797 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
800 vti6_parm_from_user(&p1, &p);
801 t = vti6_locate(net, &p1, 0);
805 if (t->dev == ip6n->fb_tnl_dev)
810 unregister_netdevice(dev);
819 * vti6_tnl_change_mtu - change mtu manually for tunnel device
820 * @dev: virtual device associated with tunnel
821 * @new_mtu: the new mtu
825 * %-EINVAL if mtu too small
827 static int vti6_change_mtu(struct net_device *dev, int new_mtu)
829 if (new_mtu < IPV6_MIN_MTU)
836 static const struct net_device_ops vti6_netdev_ops = {
837 .ndo_init = vti6_dev_init,
838 .ndo_uninit = vti6_dev_uninit,
839 .ndo_start_xmit = vti6_tnl_xmit,
840 .ndo_do_ioctl = vti6_ioctl,
841 .ndo_change_mtu = vti6_change_mtu,
842 .ndo_get_stats64 = ip_tunnel_get_stats64,
843 .ndo_get_iflink = ip6_tnl_get_iflink,
847 * vti6_dev_setup - setup virtual tunnel device
848 * @dev: virtual device associated with tunnel
851 * Initialize function pointers and device parameters
853 static void vti6_dev_setup(struct net_device *dev)
855 dev->netdev_ops = &vti6_netdev_ops;
856 dev->destructor = vti6_dev_free;
858 dev->type = ARPHRD_TUNNEL6;
859 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
860 dev->mtu = ETH_DATA_LEN;
861 dev->flags |= IFF_NOARP;
862 dev->addr_len = sizeof(struct in6_addr);
867 * vti6_dev_init_gen - general initializer for all tunnel devices
868 * @dev: virtual device associated with tunnel
870 static inline int vti6_dev_init_gen(struct net_device *dev)
872 struct ip6_tnl *t = netdev_priv(dev);
875 t->net = dev_net(dev);
876 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
883 * vti6_dev_init - initializer for all non fallback tunnel devices
884 * @dev: virtual device associated with tunnel
886 static int vti6_dev_init(struct net_device *dev)
888 struct ip6_tnl *t = netdev_priv(dev);
889 int err = vti6_dev_init_gen(dev);
898 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
899 * @dev: fallback device
903 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
905 struct ip6_tnl *t = netdev_priv(dev);
906 struct net *net = dev_net(dev);
907 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
909 t->parms.proto = IPPROTO_IPV6;
912 rcu_assign_pointer(ip6n->tnls_wc[0], t);
916 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[])
921 static void vti6_netlink_parms(struct nlattr *data[],
922 struct __ip6_tnl_parm *parms)
924 memset(parms, 0, sizeof(*parms));
929 if (data[IFLA_VTI_LINK])
930 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
932 if (data[IFLA_VTI_LOCAL])
933 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
935 if (data[IFLA_VTI_REMOTE])
936 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
938 if (data[IFLA_VTI_IKEY])
939 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
941 if (data[IFLA_VTI_OKEY])
942 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
945 static int vti6_newlink(struct net *src_net, struct net_device *dev,
946 struct nlattr *tb[], struct nlattr *data[])
948 struct net *net = dev_net(dev);
951 nt = netdev_priv(dev);
952 vti6_netlink_parms(data, &nt->parms);
954 nt->parms.proto = IPPROTO_IPV6;
956 if (vti6_locate(net, &nt->parms, 0))
959 return vti6_tnl_create2(dev);
962 static void vti6_dellink(struct net_device *dev, struct list_head *head)
964 struct net *net = dev_net(dev);
965 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
967 if (dev != ip6n->fb_tnl_dev)
968 unregister_netdevice_queue(dev, head);
971 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
972 struct nlattr *data[])
975 struct __ip6_tnl_parm p;
976 struct net *net = dev_net(dev);
977 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
979 if (dev == ip6n->fb_tnl_dev)
982 vti6_netlink_parms(data, &p);
984 t = vti6_locate(net, &p, 0);
990 t = netdev_priv(dev);
992 return vti6_update(t, &p);
995 static size_t vti6_get_size(const struct net_device *dev)
1000 /* IFLA_VTI_LOCAL */
1001 nla_total_size(sizeof(struct in6_addr)) +
1002 /* IFLA_VTI_REMOTE */
1003 nla_total_size(sizeof(struct in6_addr)) +
1011 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1013 struct ip6_tnl *tunnel = netdev_priv(dev);
1014 struct __ip6_tnl_parm *parm = &tunnel->parms;
1016 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1017 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1018 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1019 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1020 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key))
1021 goto nla_put_failure;
1028 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1029 [IFLA_VTI_LINK] = { .type = NLA_U32 },
1030 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) },
1031 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) },
1032 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
1033 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
1036 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1038 .maxtype = IFLA_VTI_MAX,
1039 .policy = vti6_policy,
1040 .priv_size = sizeof(struct ip6_tnl),
1041 .setup = vti6_dev_setup,
1042 .validate = vti6_validate,
1043 .newlink = vti6_newlink,
1044 .dellink = vti6_dellink,
1045 .changelink = vti6_changelink,
1046 .get_size = vti6_get_size,
1047 .fill_info = vti6_fill_info,
1048 .get_link_net = ip6_tnl_get_link_net,
1051 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
1057 for (h = 0; h < HASH_SIZE; h++) {
1058 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1060 unregister_netdevice_queue(t->dev, &list);
1061 t = rtnl_dereference(t->next);
1065 t = rtnl_dereference(ip6n->tnls_wc[0]);
1066 unregister_netdevice_queue(t->dev, &list);
1067 unregister_netdevice_many(&list);
1070 static int __net_init vti6_init_net(struct net *net)
1072 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1073 struct ip6_tnl *t = NULL;
1076 ip6n->tnls[0] = ip6n->tnls_wc;
1077 ip6n->tnls[1] = ip6n->tnls_r_l;
1080 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1081 NET_NAME_UNKNOWN, vti6_dev_setup);
1083 if (!ip6n->fb_tnl_dev)
1085 dev_net_set(ip6n->fb_tnl_dev, net);
1086 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1088 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1092 err = register_netdev(ip6n->fb_tnl_dev);
1096 t = netdev_priv(ip6n->fb_tnl_dev);
1098 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1102 vti6_dev_free(ip6n->fb_tnl_dev);
1107 static void __net_exit vti6_exit_net(struct net *net)
1109 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1112 vti6_destroy_tunnels(ip6n);
1116 static struct pernet_operations vti6_net_ops = {
1117 .init = vti6_init_net,
1118 .exit = vti6_exit_net,
1120 .size = sizeof(struct vti6_net),
1123 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1124 .handler = vti6_rcv,
1125 .cb_handler = vti6_rcv_cb,
1126 .err_handler = vti6_err,
1130 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1131 .handler = vti6_rcv,
1132 .cb_handler = vti6_rcv_cb,
1133 .err_handler = vti6_err,
1137 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1138 .handler = vti6_rcv,
1139 .cb_handler = vti6_rcv_cb,
1140 .err_handler = vti6_err,
1144 static bool is_vti6_tunnel(const struct net_device *dev)
1146 return dev->netdev_ops == &vti6_netdev_ops;
1149 static int vti6_device_event(struct notifier_block *unused,
1150 unsigned long event, void *ptr)
1152 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1153 struct ip6_tnl *t = netdev_priv(dev);
1155 if (!is_vti6_tunnel(dev))
1160 if (!net_eq(t->net, dev_net(dev)))
1161 xfrm_garbage_collect(t->net);
1167 static struct notifier_block vti6_notifier_block __read_mostly = {
1168 .notifier_call = vti6_device_event,
1172 * vti6_tunnel_init - register protocol and reserve needed resources
1174 * Return: 0 on success
1176 static int __init vti6_tunnel_init(void)
1181 register_netdevice_notifier(&vti6_notifier_block);
1183 msg = "tunnel device";
1184 err = register_pernet_device(&vti6_net_ops);
1186 goto pernet_dev_failed;
1188 msg = "tunnel protocols";
1189 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1191 goto xfrm_proto_esp_failed;
1192 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1194 goto xfrm_proto_ah_failed;
1195 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1197 goto xfrm_proto_comp_failed;
1199 msg = "netlink interface";
1200 err = rtnl_link_register(&vti6_link_ops);
1202 goto rtnl_link_failed;
1207 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1208 xfrm_proto_comp_failed:
1209 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1210 xfrm_proto_ah_failed:
1211 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1212 xfrm_proto_esp_failed:
1213 unregister_pernet_device(&vti6_net_ops);
1215 unregister_netdevice_notifier(&vti6_notifier_block);
1216 pr_err("vti6 init: failed to register %s\n", msg);
1221 * vti6_tunnel_cleanup - free resources and unregister protocol
1223 static void __exit vti6_tunnel_cleanup(void)
1225 rtnl_link_unregister(&vti6_link_ops);
1226 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1227 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1228 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1229 unregister_pernet_device(&vti6_net_ops);
1230 unregister_netdevice_notifier(&vti6_notifier_block);
1233 module_init(vti6_tunnel_init);
1234 module_exit(vti6_tunnel_cleanup);
1235 MODULE_LICENSE("GPL");
1236 MODULE_ALIAS_RTNL_LINK("vti6");
1237 MODULE_ALIAS_NETDEV("ip6_vti0");
1238 MODULE_AUTHOR("Steffen Klassert");
1239 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");