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);
199 vti6_tnl_link(ip6n, t);
207 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
209 struct net_device *dev;
215 if (!dev_valid_name(p->name))
217 strlcpy(name, p->name, IFNAMSIZ);
219 sprintf(name, "ip6_vti%%d");
222 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup);
226 dev_net_set(dev, net);
228 t = netdev_priv(dev);
230 t->net = dev_net(dev);
232 err = vti6_tnl_create2(dev);
245 * vti6_locate - find or create tunnel matching given parameters
246 * @net: network namespace
247 * @p: tunnel parameters
248 * @create: != 0 if allowed to create new tunnel if no match found
251 * vti6_locate() first tries to locate an existing tunnel
252 * based on @parms. If this is unsuccessful, but @create is set a new
253 * tunnel device is created and registered for use.
256 * matching tunnel or NULL
258 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
261 const struct in6_addr *remote = &p->raddr;
262 const struct in6_addr *local = &p->laddr;
263 struct ip6_tnl __rcu **tp;
265 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
267 for (tp = vti6_tnl_bucket(ip6n, p);
268 (t = rtnl_dereference(*tp)) != NULL;
270 if (ipv6_addr_equal(local, &t->parms.laddr) &&
271 ipv6_addr_equal(remote, &t->parms.raddr)) {
280 return vti6_tnl_create(net, p);
284 * vti6_dev_uninit - tunnel device uninitializer
285 * @dev: the device to be destroyed
288 * vti6_dev_uninit() removes tunnel from its list
290 static void vti6_dev_uninit(struct net_device *dev)
292 struct ip6_tnl *t = netdev_priv(dev);
293 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id);
295 if (dev == ip6n->fb_tnl_dev)
296 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
298 vti6_tnl_unlink(ip6n, t);
302 static int vti6_rcv(struct sk_buff *skb)
305 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
308 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
310 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
315 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
320 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
321 t->dev->stats.rx_dropped++;
328 return xfrm6_rcv_tnl(skb, t);
337 static int vti6_rcv_cb(struct sk_buff *skb, int err)
339 unsigned short family;
340 struct net_device *dev;
341 struct pcpu_sw_netstats *tstats;
342 struct xfrm_state *x;
343 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
344 u32 orig_mark = skb->mark;
353 dev->stats.rx_errors++;
354 dev->stats.rx_dropped++;
359 x = xfrm_input_state(skb);
360 family = x->inner_mode->afinfo->family;
362 skb->mark = be32_to_cpu(t->parms.i_key);
363 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
364 skb->mark = orig_mark;
369 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
372 tstats = this_cpu_ptr(dev->tstats);
373 u64_stats_update_begin(&tstats->syncp);
374 tstats->rx_packets++;
375 tstats->rx_bytes += skb->len;
376 u64_stats_update_end(&tstats->syncp);
382 * vti6_addr_conflict - compare packet addresses to tunnel's own
383 * @t: the outgoing tunnel device
384 * @hdr: IPv6 header from the incoming packet
387 * Avoid trivial tunneling loop by checking that tunnel exit-point
388 * doesn't match source of incoming packet.
395 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
397 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
400 static bool vti6_state_check(const struct xfrm_state *x,
401 const struct in6_addr *dst,
402 const struct in6_addr *src)
404 xfrm_address_t *daddr = (xfrm_address_t *)dst;
405 xfrm_address_t *saddr = (xfrm_address_t *)src;
407 /* if there is no transform then this tunnel is not functional.
408 * Or if the xfrm is not mode tunnel.
410 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
411 x->props.family != AF_INET6)
414 if (ipv6_addr_any(dst))
415 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
417 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
424 * vti6_xmit - send a packet
425 * @skb: the outgoing socket buffer
426 * @dev: the outgoing tunnel device
427 * @fl: the flow informations for the xfrm_lookup
430 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
432 struct ip6_tnl *t = netdev_priv(dev);
433 struct net_device_stats *stats = &t->dev->stats;
434 struct dst_entry *dst = skb_dst(skb);
435 struct net_device *tdev;
436 struct xfrm_state *x;
437 int pkt_len = skb->len;
442 switch (skb->protocol) {
443 case htons(ETH_P_IP): {
446 fl->u.ip4.flowi4_oif = dev->ifindex;
447 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
448 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4);
450 goto tx_err_link_failure;
452 skb_dst_set(skb, dst);
455 case htons(ETH_P_IPV6):
456 fl->u.ip6.flowi6_oif = dev->ifindex;
457 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
458 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
462 goto tx_err_link_failure;
464 skb_dst_set(skb, dst);
467 goto tx_err_link_failure;
472 dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
476 goto tx_err_link_failure;
480 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
481 goto tx_err_link_failure;
483 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
484 (const struct in6_addr *)&x->id.daddr))
485 goto tx_err_link_failure;
491 net_warn_ratelimited("%s: Local routing loop detected!\n",
493 goto tx_err_dst_release;
497 if (skb->len > mtu) {
498 skb_dst(skb)->ops->update_pmtu(dst, NULL, skb, mtu);
500 if (skb->protocol == htons(ETH_P_IPV6)) {
501 if (mtu < IPV6_MIN_MTU)
504 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
506 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
511 goto tx_err_dst_release;
514 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
515 skb_dst_set(skb, dst);
516 skb->dev = skb_dst(skb)->dev;
518 err = dst_output(t->net, skb->sk, skb);
519 if (net_xmit_eval(err) == 0) {
520 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
522 u64_stats_update_begin(&tstats->syncp);
523 tstats->tx_bytes += pkt_len;
524 tstats->tx_packets++;
525 u64_stats_update_end(&tstats->syncp);
528 stats->tx_aborted_errors++;
533 stats->tx_carrier_errors++;
534 dst_link_failure(skb);
541 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
543 struct ip6_tnl *t = netdev_priv(dev);
544 struct net_device_stats *stats = &t->dev->stats;
545 struct ipv6hdr *ipv6h;
549 memset(&fl, 0, sizeof(fl));
551 switch (skb->protocol) {
552 case htons(ETH_P_IPV6):
553 ipv6h = ipv6_hdr(skb);
555 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
556 vti6_addr_conflict(t, ipv6h))
559 xfrm_decode_session(skb, &fl, AF_INET6);
560 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
562 case htons(ETH_P_IP):
563 xfrm_decode_session(skb, &fl, AF_INET);
564 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
570 /* override mark with tunnel output key */
571 fl.flowi_mark = be32_to_cpu(t->parms.o_key);
573 ret = vti6_xmit(skb, dev, &fl);
586 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
587 u8 type, u8 code, int offset, __be32 info)
591 struct xfrm_state *x;
593 struct ip_esp_hdr *esph;
594 struct ip_auth_hdr *ah;
595 struct ip_comp_hdr *ipch;
596 struct net *net = dev_net(skb->dev);
597 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
598 int protocol = iph->nexthdr;
600 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
604 mark = be32_to_cpu(t->parms.o_key);
608 esph = (struct ip_esp_hdr *)(skb->data + offset);
612 ah = (struct ip_auth_hdr *)(skb->data + offset);
616 ipch = (struct ip_comp_hdr *)(skb->data + offset);
617 spi = htonl(ntohs(ipch->cpi));
623 if (type != ICMPV6_PKT_TOOBIG &&
624 type != NDISC_REDIRECT)
627 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
628 spi, protocol, AF_INET6);
632 if (type == NDISC_REDIRECT)
633 ip6_redirect(skb, net, skb->dev->ifindex, 0);
635 ip6_update_pmtu(skb, net, info, 0, 0);
641 static void vti6_link_config(struct ip6_tnl *t)
643 struct net_device *dev = t->dev;
644 struct __ip6_tnl_parm *p = &t->parms;
646 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
647 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
649 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
650 IP6_TNL_F_CAP_PER_PACKET);
651 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
653 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
654 dev->flags |= IFF_POINTOPOINT;
656 dev->flags &= ~IFF_POINTOPOINT;
660 * vti6_tnl_change - update the tunnel parameters
661 * @t: tunnel to be changed
662 * @p: tunnel configuration parameters
665 * vti6_tnl_change() updates the tunnel parameters
668 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
670 t->parms.laddr = p->laddr;
671 t->parms.raddr = p->raddr;
672 t->parms.link = p->link;
673 t->parms.i_key = p->i_key;
674 t->parms.o_key = p->o_key;
675 t->parms.proto = p->proto;
676 dst_cache_reset(&t->dst_cache);
681 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
683 struct net *net = dev_net(t->dev);
684 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
687 vti6_tnl_unlink(ip6n, t);
689 err = vti6_tnl_change(t, p);
690 vti6_tnl_link(ip6n, t);
691 netdev_state_change(t->dev);
696 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
705 memcpy(p->name, u->name, sizeof(u->name));
709 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
717 u->i_flags |= GRE_KEY;
719 u->o_flags |= GRE_KEY;
722 memcpy(u->name, p->name, sizeof(u->name));
726 * vti6_tnl_ioctl - configure vti6 tunnels from userspace
727 * @dev: virtual device associated with tunnel
728 * @ifr: parameters passed from userspace
729 * @cmd: command to be performed
732 * vti6_ioctl() is used for managing vti6 tunnels
735 * The possible commands are the following:
736 * %SIOCGETTUNNEL: get tunnel parameters for device
737 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
738 * %SIOCCHGTUNNEL: change tunnel parameters to those given
739 * %SIOCDELTUNNEL: delete tunnel
741 * The fallback device "ip6_vti0", created during module
742 * initialization, can be used for creating other tunnel devices.
746 * %-EFAULT if unable to copy data to or from userspace,
747 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
748 * %-EINVAL if passed tunnel parameters are invalid,
749 * %-EEXIST if changing a tunnel's parameters would cause a conflict
750 * %-ENODEV if attempting to change or delete a nonexisting device
753 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
756 struct ip6_tnl_parm2 p;
757 struct __ip6_tnl_parm p1;
758 struct ip6_tnl *t = NULL;
759 struct net *net = dev_net(dev);
760 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
764 if (dev == ip6n->fb_tnl_dev) {
765 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
769 vti6_parm_from_user(&p1, &p);
770 t = vti6_locate(net, &p1, 0);
772 memset(&p, 0, sizeof(p));
775 t = netdev_priv(dev);
776 vti6_parm_to_user(&p, &t->parms);
777 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
783 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
786 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
789 if (p.proto != IPPROTO_IPV6 && p.proto != 0)
791 vti6_parm_from_user(&p1, &p);
792 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
793 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
800 t = netdev_priv(dev);
802 err = vti6_update(t, &p1);
806 vti6_parm_to_user(&p, &t->parms);
807 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
811 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
815 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
818 if (dev == ip6n->fb_tnl_dev) {
820 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
823 vti6_parm_from_user(&p1, &p);
824 t = vti6_locate(net, &p1, 0);
828 if (t->dev == ip6n->fb_tnl_dev)
833 unregister_netdevice(dev);
842 * vti6_tnl_change_mtu - change mtu manually for tunnel device
843 * @dev: virtual device associated with tunnel
844 * @new_mtu: the new mtu
848 * %-EINVAL if mtu too small
850 static int vti6_change_mtu(struct net_device *dev, int new_mtu)
852 if (new_mtu < IPV6_MIN_MTU)
859 static const struct net_device_ops vti6_netdev_ops = {
860 .ndo_init = vti6_dev_init,
861 .ndo_uninit = vti6_dev_uninit,
862 .ndo_start_xmit = vti6_tnl_xmit,
863 .ndo_do_ioctl = vti6_ioctl,
864 .ndo_change_mtu = vti6_change_mtu,
865 .ndo_get_stats64 = ip_tunnel_get_stats64,
866 .ndo_get_iflink = ip6_tnl_get_iflink,
870 * vti6_dev_setup - setup virtual tunnel device
871 * @dev: virtual device associated with tunnel
874 * Initialize function pointers and device parameters
876 static void vti6_dev_setup(struct net_device *dev)
878 dev->netdev_ops = &vti6_netdev_ops;
879 dev->destructor = vti6_dev_free;
881 dev->type = ARPHRD_TUNNEL6;
882 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
883 dev->mtu = ETH_DATA_LEN;
884 dev->flags |= IFF_NOARP;
885 dev->addr_len = sizeof(struct in6_addr);
890 * vti6_dev_init_gen - general initializer for all tunnel devices
891 * @dev: virtual device associated with tunnel
893 static inline int vti6_dev_init_gen(struct net_device *dev)
895 struct ip6_tnl *t = netdev_priv(dev);
898 t->net = dev_net(dev);
899 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
907 * vti6_dev_init - initializer for all non fallback tunnel devices
908 * @dev: virtual device associated with tunnel
910 static int vti6_dev_init(struct net_device *dev)
912 struct ip6_tnl *t = netdev_priv(dev);
913 int err = vti6_dev_init_gen(dev);
922 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
923 * @dev: fallback device
927 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
929 struct ip6_tnl *t = netdev_priv(dev);
930 struct net *net = dev_net(dev);
931 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
933 t->parms.proto = IPPROTO_IPV6;
935 rcu_assign_pointer(ip6n->tnls_wc[0], t);
939 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[])
944 static void vti6_netlink_parms(struct nlattr *data[],
945 struct __ip6_tnl_parm *parms)
947 memset(parms, 0, sizeof(*parms));
952 if (data[IFLA_VTI_LINK])
953 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
955 if (data[IFLA_VTI_LOCAL])
956 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
958 if (data[IFLA_VTI_REMOTE])
959 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
961 if (data[IFLA_VTI_IKEY])
962 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
964 if (data[IFLA_VTI_OKEY])
965 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
968 static int vti6_newlink(struct net *src_net, struct net_device *dev,
969 struct nlattr *tb[], struct nlattr *data[])
971 struct net *net = dev_net(dev);
974 nt = netdev_priv(dev);
975 vti6_netlink_parms(data, &nt->parms);
977 nt->parms.proto = IPPROTO_IPV6;
979 if (vti6_locate(net, &nt->parms, 0))
982 return vti6_tnl_create2(dev);
985 static void vti6_dellink(struct net_device *dev, struct list_head *head)
987 struct net *net = dev_net(dev);
988 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
990 if (dev != ip6n->fb_tnl_dev)
991 unregister_netdevice_queue(dev, head);
994 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
995 struct nlattr *data[])
998 struct __ip6_tnl_parm p;
999 struct net *net = dev_net(dev);
1000 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1002 if (dev == ip6n->fb_tnl_dev)
1005 vti6_netlink_parms(data, &p);
1007 t = vti6_locate(net, &p, 0);
1013 t = netdev_priv(dev);
1015 return vti6_update(t, &p);
1018 static size_t vti6_get_size(const struct net_device *dev)
1023 /* IFLA_VTI_LOCAL */
1024 nla_total_size(sizeof(struct in6_addr)) +
1025 /* IFLA_VTI_REMOTE */
1026 nla_total_size(sizeof(struct in6_addr)) +
1034 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1036 struct ip6_tnl *tunnel = netdev_priv(dev);
1037 struct __ip6_tnl_parm *parm = &tunnel->parms;
1039 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1040 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1041 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1042 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1043 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key))
1044 goto nla_put_failure;
1051 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1052 [IFLA_VTI_LINK] = { .type = NLA_U32 },
1053 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) },
1054 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) },
1055 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
1056 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
1059 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1061 .maxtype = IFLA_VTI_MAX,
1062 .policy = vti6_policy,
1063 .priv_size = sizeof(struct ip6_tnl),
1064 .setup = vti6_dev_setup,
1065 .validate = vti6_validate,
1066 .newlink = vti6_newlink,
1067 .dellink = vti6_dellink,
1068 .changelink = vti6_changelink,
1069 .get_size = vti6_get_size,
1070 .fill_info = vti6_fill_info,
1071 .get_link_net = ip6_tnl_get_link_net,
1074 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
1080 for (h = 0; h < HASH_SIZE; h++) {
1081 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1083 unregister_netdevice_queue(t->dev, &list);
1084 t = rtnl_dereference(t->next);
1088 t = rtnl_dereference(ip6n->tnls_wc[0]);
1089 unregister_netdevice_queue(t->dev, &list);
1090 unregister_netdevice_many(&list);
1093 static int __net_init vti6_init_net(struct net *net)
1095 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1096 struct ip6_tnl *t = NULL;
1099 ip6n->tnls[0] = ip6n->tnls_wc;
1100 ip6n->tnls[1] = ip6n->tnls_r_l;
1103 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1104 NET_NAME_UNKNOWN, vti6_dev_setup);
1106 if (!ip6n->fb_tnl_dev)
1108 dev_net_set(ip6n->fb_tnl_dev, net);
1109 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1111 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1115 err = register_netdev(ip6n->fb_tnl_dev);
1119 t = netdev_priv(ip6n->fb_tnl_dev);
1121 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1125 vti6_dev_free(ip6n->fb_tnl_dev);
1130 static void __net_exit vti6_exit_net(struct net *net)
1132 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1135 vti6_destroy_tunnels(ip6n);
1139 static struct pernet_operations vti6_net_ops = {
1140 .init = vti6_init_net,
1141 .exit = vti6_exit_net,
1143 .size = sizeof(struct vti6_net),
1146 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1147 .handler = vti6_rcv,
1148 .cb_handler = vti6_rcv_cb,
1149 .err_handler = vti6_err,
1153 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1154 .handler = vti6_rcv,
1155 .cb_handler = vti6_rcv_cb,
1156 .err_handler = vti6_err,
1160 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1161 .handler = vti6_rcv,
1162 .cb_handler = vti6_rcv_cb,
1163 .err_handler = vti6_err,
1167 static bool is_vti6_tunnel(const struct net_device *dev)
1169 return dev->netdev_ops == &vti6_netdev_ops;
1172 static int vti6_device_event(struct notifier_block *unused,
1173 unsigned long event, void *ptr)
1175 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1176 struct ip6_tnl *t = netdev_priv(dev);
1178 if (!is_vti6_tunnel(dev))
1183 if (!net_eq(t->net, dev_net(dev)))
1184 xfrm_garbage_collect(t->net);
1190 static struct notifier_block vti6_notifier_block __read_mostly = {
1191 .notifier_call = vti6_device_event,
1195 * vti6_tunnel_init - register protocol and reserve needed resources
1197 * Return: 0 on success
1199 static int __init vti6_tunnel_init(void)
1204 register_netdevice_notifier(&vti6_notifier_block);
1206 msg = "tunnel device";
1207 err = register_pernet_device(&vti6_net_ops);
1209 goto pernet_dev_failed;
1211 msg = "tunnel protocols";
1212 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1214 goto xfrm_proto_esp_failed;
1215 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1217 goto xfrm_proto_ah_failed;
1218 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1220 goto xfrm_proto_comp_failed;
1222 msg = "netlink interface";
1223 err = rtnl_link_register(&vti6_link_ops);
1225 goto rtnl_link_failed;
1230 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1231 xfrm_proto_comp_failed:
1232 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1233 xfrm_proto_ah_failed:
1234 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1235 xfrm_proto_esp_failed:
1236 unregister_pernet_device(&vti6_net_ops);
1238 unregister_netdevice_notifier(&vti6_notifier_block);
1239 pr_err("vti6 init: failed to register %s\n", msg);
1244 * vti6_tunnel_cleanup - free resources and unregister protocol
1246 static void __exit vti6_tunnel_cleanup(void)
1248 rtnl_link_unregister(&vti6_link_ops);
1249 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1250 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1251 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1252 unregister_pernet_device(&vti6_net_ops);
1253 unregister_netdevice_notifier(&vti6_notifier_block);
1256 module_init(vti6_tunnel_init);
1257 module_exit(vti6_tunnel_cleanup);
1258 MODULE_LICENSE("GPL");
1259 MODULE_ALIAS_RTNL_LINK("vti6");
1260 MODULE_ALIAS_NETDEV("ip6_vti0");
1261 MODULE_AUTHOR("Steffen Klassert");
1262 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");