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>
52 #include <linux/etherdevice.h>
54 #define IP6_VTI_HASH_SIZE_SHIFT 5
55 #define IP6_VTI_HASH_SIZE (1 << IP6_VTI_HASH_SIZE_SHIFT)
57 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
59 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
61 return hash_32(hash, IP6_VTI_HASH_SIZE_SHIFT);
64 static int vti6_dev_init(struct net_device *dev);
65 static void vti6_dev_setup(struct net_device *dev);
66 static struct rtnl_link_ops vti6_link_ops __read_mostly;
68 static unsigned int vti6_net_id __read_mostly;
70 /* the vti6 tunnel fallback device */
71 struct net_device *fb_tnl_dev;
72 /* lists for storing tunnels in use */
73 struct ip6_tnl __rcu *tnls_r_l[IP6_VTI_HASH_SIZE];
74 struct ip6_tnl __rcu *tnls_wc[1];
75 struct ip6_tnl __rcu **tnls[2];
78 #define for_each_vti6_tunnel_rcu(start) \
79 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
82 * vti6_tnl_lookup - fetch tunnel matching the end-point addresses
83 * @net: network namespace
84 * @remote: the address of the tunnel exit-point
85 * @local: the address of the tunnel entry-point
88 * tunnel matching given end-points if found,
89 * else fallback tunnel if its device is up,
92 static struct ip6_tnl *
93 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote,
94 const struct in6_addr *local)
96 unsigned int hash = HASH(remote, local);
98 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
101 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
102 if (ipv6_addr_equal(local, &t->parms.laddr) &&
103 ipv6_addr_equal(remote, &t->parms.raddr) &&
104 (t->dev->flags & IFF_UP))
108 memset(&any, 0, sizeof(any));
109 hash = HASH(&any, local);
110 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
111 if (ipv6_addr_equal(local, &t->parms.laddr) &&
112 (t->dev->flags & IFF_UP))
116 hash = HASH(remote, &any);
117 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
118 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
119 (t->dev->flags & IFF_UP))
123 t = rcu_dereference(ip6n->tnls_wc[0]);
124 if (t && (t->dev->flags & IFF_UP))
131 * vti6_tnl_bucket - get head of list matching given tunnel parameters
132 * @p: parameters containing tunnel end-points
135 * vti6_tnl_bucket() returns the head of the list matching the
136 * &struct in6_addr entries laddr and raddr in @p.
138 * Return: head of IPv6 tunnel list
140 static struct ip6_tnl __rcu **
141 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
143 const struct in6_addr *remote = &p->raddr;
144 const struct in6_addr *local = &p->laddr;
148 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
150 h = HASH(remote, local);
152 return &ip6n->tnls[prio][h];
156 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
158 struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
160 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
161 rcu_assign_pointer(*tp, t);
165 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
167 struct ip6_tnl __rcu **tp;
168 struct ip6_tnl *iter;
170 for (tp = vti6_tnl_bucket(ip6n, &t->parms);
171 (iter = rtnl_dereference(*tp)) != NULL;
174 rcu_assign_pointer(*tp, t->next);
180 static void vti6_dev_free(struct net_device *dev)
182 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 ipv6h = ipv6_hdr(skb);
322 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
323 t->dev->stats.rx_dropped++;
330 return xfrm6_rcv_tnl(skb, t);
339 static int vti6_rcv_cb(struct sk_buff *skb, int err)
341 unsigned short family;
342 struct net_device *dev;
343 struct pcpu_sw_netstats *tstats;
344 struct xfrm_state *x;
345 struct xfrm_mode *inner_mode;
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);
364 inner_mode = x->inner_mode;
366 if (x->sel.family == AF_UNSPEC) {
367 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
368 if (inner_mode == NULL) {
369 XFRM_INC_STATS(dev_net(skb->dev),
370 LINUX_MIB_XFRMINSTATEMODEERROR);
375 family = inner_mode->afinfo->family;
377 skb->mark = be32_to_cpu(t->parms.i_key);
378 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
379 skb->mark = orig_mark;
384 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
387 tstats = this_cpu_ptr(dev->tstats);
388 u64_stats_update_begin(&tstats->syncp);
389 tstats->rx_packets++;
390 tstats->rx_bytes += skb->len;
391 u64_stats_update_end(&tstats->syncp);
397 * vti6_addr_conflict - compare packet addresses to tunnel's own
398 * @t: the outgoing tunnel device
399 * @hdr: IPv6 header from the incoming packet
402 * Avoid trivial tunneling loop by checking that tunnel exit-point
403 * doesn't match source of incoming packet.
410 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
412 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
415 static bool vti6_state_check(const struct xfrm_state *x,
416 const struct in6_addr *dst,
417 const struct in6_addr *src)
419 xfrm_address_t *daddr = (xfrm_address_t *)dst;
420 xfrm_address_t *saddr = (xfrm_address_t *)src;
422 /* if there is no transform then this tunnel is not functional.
423 * Or if the xfrm is not mode tunnel.
425 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
426 x->props.family != AF_INET6)
429 if (ipv6_addr_any(dst))
430 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
432 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
439 * vti6_xmit - send a packet
440 * @skb: the outgoing socket buffer
441 * @dev: the outgoing tunnel device
442 * @fl: the flow informations for the xfrm_lookup
445 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
447 struct ip6_tnl *t = netdev_priv(dev);
448 struct net_device_stats *stats = &t->dev->stats;
449 struct dst_entry *dst = skb_dst(skb);
450 struct net_device *tdev;
451 struct xfrm_state *x;
452 int pkt_len = skb->len;
457 fl->u.ip6.flowi6_oif = dev->ifindex;
458 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
459 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
463 goto tx_err_link_failure;
465 skb_dst_set(skb, dst);
469 dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
473 goto tx_err_link_failure;
477 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
478 goto tx_err_link_failure;
480 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
481 (const struct in6_addr *)&x->id.daddr))
482 goto tx_err_link_failure;
488 net_warn_ratelimited("%s: Local routing loop detected!\n",
490 goto tx_err_dst_release;
494 if (skb->len > mtu) {
495 skb_dst_update_pmtu_no_confirm(skb, mtu);
497 if (skb->protocol == htons(ETH_P_IPV6)) {
498 if (mtu < IPV6_MIN_MTU)
501 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
503 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
508 goto tx_err_dst_release;
511 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
512 skb_dst_set(skb, dst);
513 skb->dev = skb_dst(skb)->dev;
515 err = dst_output(t->net, skb->sk, skb);
516 if (net_xmit_eval(err) == 0)
518 iptunnel_xmit_stats(dev, err);
522 stats->tx_carrier_errors++;
523 dst_link_failure(skb);
530 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
532 struct ip6_tnl *t = netdev_priv(dev);
533 struct net_device_stats *stats = &t->dev->stats;
537 if (!pskb_inet_may_pull(skb))
540 memset(&fl, 0, sizeof(fl));
542 switch (skb->protocol) {
543 case htons(ETH_P_IPV6):
544 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
545 vti6_addr_conflict(t, ipv6_hdr(skb)))
548 xfrm_decode_session(skb, &fl, AF_INET6);
549 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
551 case htons(ETH_P_IP):
552 xfrm_decode_session(skb, &fl, AF_INET);
553 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
559 /* override mark with tunnel output key */
560 fl.flowi_mark = be32_to_cpu(t->parms.o_key);
562 ret = vti6_xmit(skb, dev, &fl);
575 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
576 u8 type, u8 code, int offset, __be32 info)
580 struct xfrm_state *x;
582 struct ip_esp_hdr *esph;
583 struct ip_auth_hdr *ah;
584 struct ip_comp_hdr *ipch;
585 struct net *net = dev_net(skb->dev);
586 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
587 int protocol = iph->nexthdr;
589 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
593 mark = be32_to_cpu(t->parms.o_key);
597 esph = (struct ip_esp_hdr *)(skb->data + offset);
601 ah = (struct ip_auth_hdr *)(skb->data + offset);
605 ipch = (struct ip_comp_hdr *)(skb->data + offset);
606 spi = htonl(ntohs(ipch->cpi));
612 if (type != ICMPV6_PKT_TOOBIG &&
613 type != NDISC_REDIRECT)
616 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
617 spi, protocol, AF_INET6);
621 if (type == NDISC_REDIRECT)
622 ip6_redirect(skb, net, skb->dev->ifindex, 0,
623 sock_net_uid(net, NULL));
625 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
631 static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu)
633 struct net_device *dev = t->dev;
634 struct __ip6_tnl_parm *p = &t->parms;
635 struct net_device *tdev = NULL;
638 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
639 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
641 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
642 IP6_TNL_F_CAP_PER_PACKET);
643 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
645 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
646 dev->flags |= IFF_POINTOPOINT;
648 dev->flags &= ~IFF_POINTOPOINT;
650 if (keep_mtu && dev->mtu) {
651 dev->mtu = clamp(dev->mtu, dev->min_mtu, dev->max_mtu);
655 if (p->flags & IP6_TNL_F_CAP_XMIT) {
656 int strict = (ipv6_addr_type(&p->raddr) &
657 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
658 struct rt6_info *rt = rt6_lookup(t->net,
659 &p->raddr, &p->laddr,
660 p->link, NULL, strict);
667 if (!tdev && p->link)
668 tdev = __dev_get_by_index(t->net, p->link);
671 mtu = tdev->mtu - sizeof(struct ipv6hdr);
673 mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr);
675 dev->mtu = max_t(int, mtu, IPV4_MIN_MTU);
679 * vti6_tnl_change - update the tunnel parameters
680 * @t: tunnel to be changed
681 * @p: tunnel configuration parameters
682 * @keep_mtu: MTU was set from userspace, don't re-compute it
685 * vti6_tnl_change() updates the tunnel parameters
688 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
691 t->parms.laddr = p->laddr;
692 t->parms.raddr = p->raddr;
693 t->parms.link = p->link;
694 t->parms.i_key = p->i_key;
695 t->parms.o_key = p->o_key;
696 t->parms.proto = p->proto;
697 t->parms.fwmark = p->fwmark;
698 dst_cache_reset(&t->dst_cache);
699 vti6_link_config(t, keep_mtu);
703 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p,
706 struct net *net = dev_net(t->dev);
707 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
710 vti6_tnl_unlink(ip6n, t);
712 err = vti6_tnl_change(t, p, keep_mtu);
713 vti6_tnl_link(ip6n, t);
714 netdev_state_change(t->dev);
719 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
728 memcpy(p->name, u->name, sizeof(u->name));
732 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
740 u->i_flags |= GRE_KEY;
742 u->o_flags |= GRE_KEY;
745 memcpy(u->name, p->name, sizeof(u->name));
749 * vti6_ioctl - configure vti6 tunnels from userspace
750 * @dev: virtual device associated with tunnel
751 * @ifr: parameters passed from userspace
752 * @cmd: command to be performed
755 * vti6_ioctl() is used for managing vti6 tunnels
758 * The possible commands are the following:
759 * %SIOCGETTUNNEL: get tunnel parameters for device
760 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
761 * %SIOCCHGTUNNEL: change tunnel parameters to those given
762 * %SIOCDELTUNNEL: delete tunnel
764 * The fallback device "ip6_vti0", created during module
765 * initialization, can be used for creating other tunnel devices.
769 * %-EFAULT if unable to copy data to or from userspace,
770 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
771 * %-EINVAL if passed tunnel parameters are invalid,
772 * %-EEXIST if changing a tunnel's parameters would cause a conflict
773 * %-ENODEV if attempting to change or delete a nonexisting device
776 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
779 struct ip6_tnl_parm2 p;
780 struct __ip6_tnl_parm p1;
781 struct ip6_tnl *t = NULL;
782 struct net *net = dev_net(dev);
783 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
787 if (dev == ip6n->fb_tnl_dev) {
788 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
792 vti6_parm_from_user(&p1, &p);
793 t = vti6_locate(net, &p1, 0);
795 memset(&p, 0, sizeof(p));
798 t = netdev_priv(dev);
799 vti6_parm_to_user(&p, &t->parms);
800 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
806 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
809 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
812 if (p.proto != IPPROTO_IPV6 && p.proto != 0)
814 vti6_parm_from_user(&p1, &p);
815 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
816 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
823 t = netdev_priv(dev);
825 err = vti6_update(t, &p1, false);
829 vti6_parm_to_user(&p, &t->parms);
830 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
834 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
838 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
841 if (dev == ip6n->fb_tnl_dev) {
843 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
846 vti6_parm_from_user(&p1, &p);
847 t = vti6_locate(net, &p1, 0);
851 if (t->dev == ip6n->fb_tnl_dev)
856 unregister_netdevice(dev);
864 static const struct net_device_ops vti6_netdev_ops = {
865 .ndo_init = vti6_dev_init,
866 .ndo_uninit = vti6_dev_uninit,
867 .ndo_start_xmit = vti6_tnl_xmit,
868 .ndo_do_ioctl = vti6_ioctl,
869 .ndo_get_stats64 = ip_tunnel_get_stats64,
870 .ndo_get_iflink = ip6_tnl_get_iflink,
874 * vti6_dev_setup - setup virtual tunnel device
875 * @dev: virtual device associated with tunnel
878 * Initialize function pointers and device parameters
880 static void vti6_dev_setup(struct net_device *dev)
882 dev->netdev_ops = &vti6_netdev_ops;
883 dev->needs_free_netdev = true;
884 dev->priv_destructor = vti6_dev_free;
886 dev->type = ARPHRD_TUNNEL6;
887 dev->min_mtu = IPV4_MIN_MTU;
888 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr);
889 dev->flags |= IFF_NOARP;
890 dev->addr_len = sizeof(struct in6_addr);
892 /* This perm addr will be used as interface identifier by IPv6 */
893 dev->addr_assign_type = NET_ADDR_RANDOM;
894 eth_random_addr(dev->perm_addr);
898 * vti6_dev_init_gen - general initializer for all tunnel devices
899 * @dev: virtual device associated with tunnel
901 static inline int vti6_dev_init_gen(struct net_device *dev)
903 struct ip6_tnl *t = netdev_priv(dev);
906 t->net = dev_net(dev);
907 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
914 * vti6_dev_init - initializer for all non fallback tunnel devices
915 * @dev: virtual device associated with tunnel
917 static int vti6_dev_init(struct net_device *dev)
919 struct ip6_tnl *t = netdev_priv(dev);
920 int err = vti6_dev_init_gen(dev);
924 vti6_link_config(t, true);
929 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
930 * @dev: fallback device
934 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
936 struct ip6_tnl *t = netdev_priv(dev);
937 struct net *net = dev_net(dev);
938 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
940 t->parms.proto = IPPROTO_IPV6;
943 rcu_assign_pointer(ip6n->tnls_wc[0], t);
947 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[],
948 struct netlink_ext_ack *extack)
953 static void vti6_netlink_parms(struct nlattr *data[],
954 struct __ip6_tnl_parm *parms)
956 memset(parms, 0, sizeof(*parms));
961 if (data[IFLA_VTI_LINK])
962 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
964 if (data[IFLA_VTI_LOCAL])
965 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
967 if (data[IFLA_VTI_REMOTE])
968 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
970 if (data[IFLA_VTI_IKEY])
971 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
973 if (data[IFLA_VTI_OKEY])
974 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
976 if (data[IFLA_VTI_FWMARK])
977 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
980 static int vti6_newlink(struct net *src_net, struct net_device *dev,
981 struct nlattr *tb[], struct nlattr *data[],
982 struct netlink_ext_ack *extack)
984 struct net *net = dev_net(dev);
987 nt = netdev_priv(dev);
988 vti6_netlink_parms(data, &nt->parms);
990 nt->parms.proto = IPPROTO_IPV6;
992 if (vti6_locate(net, &nt->parms, 0))
995 return vti6_tnl_create2(dev);
998 static void vti6_dellink(struct net_device *dev, struct list_head *head)
1000 struct net *net = dev_net(dev);
1001 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1003 if (dev != ip6n->fb_tnl_dev)
1004 unregister_netdevice_queue(dev, head);
1007 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
1008 struct nlattr *data[],
1009 struct netlink_ext_ack *extack)
1012 struct __ip6_tnl_parm p;
1013 struct net *net = dev_net(dev);
1014 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1016 if (dev == ip6n->fb_tnl_dev)
1019 vti6_netlink_parms(data, &p);
1021 t = vti6_locate(net, &p, 0);
1027 t = netdev_priv(dev);
1029 return vti6_update(t, &p, tb && tb[IFLA_MTU]);
1032 static size_t vti6_get_size(const struct net_device *dev)
1037 /* IFLA_VTI_LOCAL */
1038 nla_total_size(sizeof(struct in6_addr)) +
1039 /* IFLA_VTI_REMOTE */
1040 nla_total_size(sizeof(struct in6_addr)) +
1045 /* IFLA_VTI_FWMARK */
1050 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1052 struct ip6_tnl *tunnel = netdev_priv(dev);
1053 struct __ip6_tnl_parm *parm = &tunnel->parms;
1055 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1056 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1057 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1058 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1059 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) ||
1060 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark))
1061 goto nla_put_failure;
1068 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1069 [IFLA_VTI_LINK] = { .type = NLA_U32 },
1070 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) },
1071 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) },
1072 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
1073 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
1074 [IFLA_VTI_FWMARK] = { .type = NLA_U32 },
1077 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1079 .maxtype = IFLA_VTI_MAX,
1080 .policy = vti6_policy,
1081 .priv_size = sizeof(struct ip6_tnl),
1082 .setup = vti6_dev_setup,
1083 .validate = vti6_validate,
1084 .newlink = vti6_newlink,
1085 .dellink = vti6_dellink,
1086 .changelink = vti6_changelink,
1087 .get_size = vti6_get_size,
1088 .fill_info = vti6_fill_info,
1089 .get_link_net = ip6_tnl_get_link_net,
1092 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n,
1093 struct list_head *list)
1098 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1099 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1101 unregister_netdevice_queue(t->dev, list);
1102 t = rtnl_dereference(t->next);
1106 t = rtnl_dereference(ip6n->tnls_wc[0]);
1108 unregister_netdevice_queue(t->dev, list);
1111 static int __net_init vti6_init_net(struct net *net)
1113 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1114 struct ip6_tnl *t = NULL;
1117 ip6n->tnls[0] = ip6n->tnls_wc;
1118 ip6n->tnls[1] = ip6n->tnls_r_l;
1120 if (!net_has_fallback_tunnels(net))
1123 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1124 NET_NAME_UNKNOWN, vti6_dev_setup);
1126 if (!ip6n->fb_tnl_dev)
1128 dev_net_set(ip6n->fb_tnl_dev, net);
1129 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1131 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1135 err = register_netdev(ip6n->fb_tnl_dev);
1139 t = netdev_priv(ip6n->fb_tnl_dev);
1141 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1145 free_netdev(ip6n->fb_tnl_dev);
1150 static void __net_exit vti6_exit_batch_net(struct list_head *net_list)
1152 struct vti6_net *ip6n;
1157 list_for_each_entry(net, net_list, exit_list) {
1158 ip6n = net_generic(net, vti6_net_id);
1159 vti6_destroy_tunnels(ip6n, &list);
1161 unregister_netdevice_many(&list);
1165 static struct pernet_operations vti6_net_ops = {
1166 .init = vti6_init_net,
1167 .exit_batch = vti6_exit_batch_net,
1169 .size = sizeof(struct vti6_net),
1172 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1173 .handler = vti6_rcv,
1174 .cb_handler = vti6_rcv_cb,
1175 .err_handler = vti6_err,
1179 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1180 .handler = vti6_rcv,
1181 .cb_handler = vti6_rcv_cb,
1182 .err_handler = vti6_err,
1186 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1187 .handler = vti6_rcv,
1188 .cb_handler = vti6_rcv_cb,
1189 .err_handler = vti6_err,
1194 * vti6_tunnel_init - register protocol and reserve needed resources
1196 * Return: 0 on success
1198 static int __init vti6_tunnel_init(void)
1203 msg = "tunnel device";
1204 err = register_pernet_device(&vti6_net_ops);
1206 goto pernet_dev_failed;
1208 msg = "tunnel protocols";
1209 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1211 goto xfrm_proto_esp_failed;
1212 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1214 goto xfrm_proto_ah_failed;
1215 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1217 goto xfrm_proto_comp_failed;
1219 msg = "netlink interface";
1220 err = rtnl_link_register(&vti6_link_ops);
1222 goto rtnl_link_failed;
1227 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1228 xfrm_proto_comp_failed:
1229 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1230 xfrm_proto_ah_failed:
1231 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1232 xfrm_proto_esp_failed:
1233 unregister_pernet_device(&vti6_net_ops);
1235 pr_err("vti6 init: failed to register %s\n", msg);
1240 * vti6_tunnel_cleanup - free resources and unregister protocol
1242 static void __exit vti6_tunnel_cleanup(void)
1244 rtnl_link_unregister(&vti6_link_ops);
1245 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1246 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1247 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1248 unregister_pernet_device(&vti6_net_ops);
1251 module_init(vti6_tunnel_init);
1252 module_exit(vti6_tunnel_cleanup);
1253 MODULE_LICENSE("GPL");
1254 MODULE_ALIAS_RTNL_LINK("vti6");
1255 MODULE_ALIAS_NETDEV("ip6_vti0");
1256 MODULE_AUTHOR("Steffen Klassert");
1257 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");