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 IP6_VTI_HASH_SIZE_SHIFT 5
54 #define IP6_VTI_HASH_SIZE (1 << IP6_VTI_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, IP6_VTI_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[IP6_VTI_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 ipv6h = ipv6_hdr(skb);
321 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
322 t->dev->stats.rx_dropped++;
329 return xfrm6_rcv_tnl(skb, t);
338 static int vti6_rcv_cb(struct sk_buff *skb, int err)
340 unsigned short family;
341 struct net_device *dev;
342 struct pcpu_sw_netstats *tstats;
343 struct xfrm_state *x;
344 struct xfrm_mode *inner_mode;
345 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
346 u32 orig_mark = skb->mark;
355 dev->stats.rx_errors++;
356 dev->stats.rx_dropped++;
361 x = xfrm_input_state(skb);
363 inner_mode = x->inner_mode;
365 if (x->sel.family == AF_UNSPEC) {
366 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
367 if (inner_mode == NULL) {
368 XFRM_INC_STATS(dev_net(skb->dev),
369 LINUX_MIB_XFRMINSTATEMODEERROR);
374 family = inner_mode->afinfo->family;
376 skb->mark = be32_to_cpu(t->parms.i_key);
377 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
378 skb->mark = orig_mark;
383 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
386 tstats = this_cpu_ptr(dev->tstats);
387 u64_stats_update_begin(&tstats->syncp);
388 tstats->rx_packets++;
389 tstats->rx_bytes += skb->len;
390 u64_stats_update_end(&tstats->syncp);
396 * vti6_addr_conflict - compare packet addresses to tunnel's own
397 * @t: the outgoing tunnel device
398 * @hdr: IPv6 header from the incoming packet
401 * Avoid trivial tunneling loop by checking that tunnel exit-point
402 * doesn't match source of incoming packet.
409 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
411 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
414 static bool vti6_state_check(const struct xfrm_state *x,
415 const struct in6_addr *dst,
416 const struct in6_addr *src)
418 xfrm_address_t *daddr = (xfrm_address_t *)dst;
419 xfrm_address_t *saddr = (xfrm_address_t *)src;
421 /* if there is no transform then this tunnel is not functional.
422 * Or if the xfrm is not mode tunnel.
424 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
425 x->props.family != AF_INET6)
428 if (ipv6_addr_any(dst))
429 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
431 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
438 * vti6_xmit - send a packet
439 * @skb: the outgoing socket buffer
440 * @dev: the outgoing tunnel device
441 * @fl: the flow informations for the xfrm_lookup
444 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
446 struct ip6_tnl *t = netdev_priv(dev);
447 struct net_device_stats *stats = &t->dev->stats;
448 struct dst_entry *dst = skb_dst(skb);
449 struct net_device *tdev;
450 struct xfrm_state *x;
451 int pkt_len = skb->len;
456 switch (skb->protocol) {
457 case htons(ETH_P_IP): {
460 fl->u.ip4.flowi4_oif = dev->ifindex;
461 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
462 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4);
464 goto tx_err_link_failure;
466 skb_dst_set(skb, dst);
469 case htons(ETH_P_IPV6):
470 fl->u.ip6.flowi6_oif = dev->ifindex;
471 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
472 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
476 goto tx_err_link_failure;
478 skb_dst_set(skb, dst);
481 goto tx_err_link_failure;
486 dst = xfrm_lookup(t->net, dst, fl, NULL, 0);
490 goto tx_err_link_failure;
494 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
495 goto tx_err_link_failure;
497 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
498 (const struct in6_addr *)&x->id.daddr))
499 goto tx_err_link_failure;
505 net_warn_ratelimited("%s: Local routing loop detected!\n",
507 goto tx_err_dst_release;
511 if (skb->len > mtu) {
512 skb_dst(skb)->ops->update_pmtu(dst, NULL, skb, mtu);
514 if (skb->protocol == htons(ETH_P_IPV6)) {
515 if (mtu < IPV6_MIN_MTU)
518 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
520 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
525 goto tx_err_dst_release;
528 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
529 skb_dst_set(skb, dst);
530 skb->dev = skb_dst(skb)->dev;
532 err = dst_output(t->net, skb->sk, skb);
533 if (net_xmit_eval(err) == 0) {
534 struct pcpu_sw_netstats *tstats = this_cpu_ptr(dev->tstats);
536 u64_stats_update_begin(&tstats->syncp);
537 tstats->tx_bytes += pkt_len;
538 tstats->tx_packets++;
539 u64_stats_update_end(&tstats->syncp);
542 stats->tx_aborted_errors++;
547 stats->tx_carrier_errors++;
548 dst_link_failure(skb);
555 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
557 struct ip6_tnl *t = netdev_priv(dev);
558 struct net_device_stats *stats = &t->dev->stats;
559 struct ipv6hdr *ipv6h;
563 memset(&fl, 0, sizeof(fl));
565 switch (skb->protocol) {
566 case htons(ETH_P_IPV6):
567 ipv6h = ipv6_hdr(skb);
569 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
570 vti6_addr_conflict(t, ipv6h))
573 xfrm_decode_session(skb, &fl, AF_INET6);
574 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
576 case htons(ETH_P_IP):
577 xfrm_decode_session(skb, &fl, AF_INET);
578 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
584 /* override mark with tunnel output key */
585 fl.flowi_mark = be32_to_cpu(t->parms.o_key);
587 ret = vti6_xmit(skb, dev, &fl);
600 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
601 u8 type, u8 code, int offset, __be32 info)
605 struct xfrm_state *x;
607 struct ip_esp_hdr *esph;
608 struct ip_auth_hdr *ah;
609 struct ip_comp_hdr *ipch;
610 struct net *net = dev_net(skb->dev);
611 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
612 int protocol = iph->nexthdr;
614 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
618 mark = be32_to_cpu(t->parms.o_key);
622 esph = (struct ip_esp_hdr *)(skb->data + offset);
626 ah = (struct ip_auth_hdr *)(skb->data + offset);
630 ipch = (struct ip_comp_hdr *)(skb->data + offset);
631 spi = htonl(ntohs(ipch->cpi));
637 if (type != ICMPV6_PKT_TOOBIG &&
638 type != NDISC_REDIRECT)
641 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
642 spi, protocol, AF_INET6);
646 if (type == NDISC_REDIRECT)
647 ip6_redirect(skb, net, skb->dev->ifindex, 0);
649 ip6_update_pmtu(skb, net, info, 0, 0);
655 static void vti6_link_config(struct ip6_tnl *t)
657 struct net_device *dev = t->dev;
658 struct __ip6_tnl_parm *p = &t->parms;
660 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
661 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
663 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
664 IP6_TNL_F_CAP_PER_PACKET);
665 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
667 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
668 dev->flags |= IFF_POINTOPOINT;
670 dev->flags &= ~IFF_POINTOPOINT;
674 * vti6_tnl_change - update the tunnel parameters
675 * @t: tunnel to be changed
676 * @p: tunnel configuration parameters
679 * vti6_tnl_change() updates the tunnel parameters
682 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
684 t->parms.laddr = p->laddr;
685 t->parms.raddr = p->raddr;
686 t->parms.link = p->link;
687 t->parms.i_key = p->i_key;
688 t->parms.o_key = p->o_key;
689 t->parms.proto = p->proto;
690 dst_cache_reset(&t->dst_cache);
695 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
697 struct net *net = dev_net(t->dev);
698 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
701 vti6_tnl_unlink(ip6n, t);
703 err = vti6_tnl_change(t, p);
704 vti6_tnl_link(ip6n, t);
705 netdev_state_change(t->dev);
710 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
719 memcpy(p->name, u->name, sizeof(u->name));
723 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
731 u->i_flags |= GRE_KEY;
733 u->o_flags |= GRE_KEY;
736 memcpy(u->name, p->name, sizeof(u->name));
740 * vti6_tnl_ioctl - configure vti6 tunnels from userspace
741 * @dev: virtual device associated with tunnel
742 * @ifr: parameters passed from userspace
743 * @cmd: command to be performed
746 * vti6_ioctl() is used for managing vti6 tunnels
749 * The possible commands are the following:
750 * %SIOCGETTUNNEL: get tunnel parameters for device
751 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
752 * %SIOCCHGTUNNEL: change tunnel parameters to those given
753 * %SIOCDELTUNNEL: delete tunnel
755 * The fallback device "ip6_vti0", created during module
756 * initialization, can be used for creating other tunnel devices.
760 * %-EFAULT if unable to copy data to or from userspace,
761 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
762 * %-EINVAL if passed tunnel parameters are invalid,
763 * %-EEXIST if changing a tunnel's parameters would cause a conflict
764 * %-ENODEV if attempting to change or delete a nonexisting device
767 vti6_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
770 struct ip6_tnl_parm2 p;
771 struct __ip6_tnl_parm p1;
772 struct ip6_tnl *t = NULL;
773 struct net *net = dev_net(dev);
774 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
776 memset(&p1, 0, sizeof(p1));
780 if (dev == ip6n->fb_tnl_dev) {
781 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
785 vti6_parm_from_user(&p1, &p);
786 t = vti6_locate(net, &p1, 0);
788 memset(&p, 0, sizeof(p));
791 t = netdev_priv(dev);
792 vti6_parm_to_user(&p, &t->parms);
793 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
799 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
802 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
805 if (p.proto != IPPROTO_IPV6 && p.proto != 0)
807 vti6_parm_from_user(&p1, &p);
808 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
809 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
816 t = netdev_priv(dev);
818 err = vti6_update(t, &p1);
822 vti6_parm_to_user(&p, &t->parms);
823 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
827 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
831 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
834 if (dev == ip6n->fb_tnl_dev) {
836 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
839 vti6_parm_from_user(&p1, &p);
840 t = vti6_locate(net, &p1, 0);
844 if (t->dev == ip6n->fb_tnl_dev)
849 unregister_netdevice(dev);
858 * vti6_tnl_change_mtu - change mtu manually for tunnel device
859 * @dev: virtual device associated with tunnel
860 * @new_mtu: the new mtu
864 * %-EINVAL if mtu too small
866 static int vti6_change_mtu(struct net_device *dev, int new_mtu)
868 if (new_mtu < IPV6_MIN_MTU)
875 static const struct net_device_ops vti6_netdev_ops = {
876 .ndo_init = vti6_dev_init,
877 .ndo_uninit = vti6_dev_uninit,
878 .ndo_start_xmit = vti6_tnl_xmit,
879 .ndo_do_ioctl = vti6_ioctl,
880 .ndo_change_mtu = vti6_change_mtu,
881 .ndo_get_stats64 = ip_tunnel_get_stats64,
882 .ndo_get_iflink = ip6_tnl_get_iflink,
886 * vti6_dev_setup - setup virtual tunnel device
887 * @dev: virtual device associated with tunnel
890 * Initialize function pointers and device parameters
892 static void vti6_dev_setup(struct net_device *dev)
894 dev->netdev_ops = &vti6_netdev_ops;
895 dev->destructor = vti6_dev_free;
897 dev->type = ARPHRD_TUNNEL6;
898 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
899 dev->mtu = ETH_DATA_LEN;
900 dev->flags |= IFF_NOARP;
901 dev->addr_len = sizeof(struct in6_addr);
906 * vti6_dev_init_gen - general initializer for all tunnel devices
907 * @dev: virtual device associated with tunnel
909 static inline int vti6_dev_init_gen(struct net_device *dev)
911 struct ip6_tnl *t = netdev_priv(dev);
914 t->net = dev_net(dev);
915 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
923 * vti6_dev_init - initializer for all non fallback tunnel devices
924 * @dev: virtual device associated with tunnel
926 static int vti6_dev_init(struct net_device *dev)
928 struct ip6_tnl *t = netdev_priv(dev);
929 int err = vti6_dev_init_gen(dev);
938 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
939 * @dev: fallback device
943 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
945 struct ip6_tnl *t = netdev_priv(dev);
946 struct net *net = dev_net(dev);
947 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
949 t->parms.proto = IPPROTO_IPV6;
951 rcu_assign_pointer(ip6n->tnls_wc[0], t);
955 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[])
960 static void vti6_netlink_parms(struct nlattr *data[],
961 struct __ip6_tnl_parm *parms)
963 memset(parms, 0, sizeof(*parms));
968 if (data[IFLA_VTI_LINK])
969 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
971 if (data[IFLA_VTI_LOCAL])
972 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
974 if (data[IFLA_VTI_REMOTE])
975 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
977 if (data[IFLA_VTI_IKEY])
978 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
980 if (data[IFLA_VTI_OKEY])
981 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
984 static int vti6_newlink(struct net *src_net, struct net_device *dev,
985 struct nlattr *tb[], struct nlattr *data[])
987 struct net *net = dev_net(dev);
990 nt = netdev_priv(dev);
991 vti6_netlink_parms(data, &nt->parms);
993 nt->parms.proto = IPPROTO_IPV6;
995 if (vti6_locate(net, &nt->parms, 0))
998 return vti6_tnl_create2(dev);
1001 static void vti6_dellink(struct net_device *dev, struct list_head *head)
1003 struct net *net = dev_net(dev);
1004 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1006 if (dev != ip6n->fb_tnl_dev)
1007 unregister_netdevice_queue(dev, head);
1010 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
1011 struct nlattr *data[])
1014 struct __ip6_tnl_parm p;
1015 struct net *net = dev_net(dev);
1016 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1018 if (dev == ip6n->fb_tnl_dev)
1021 vti6_netlink_parms(data, &p);
1023 t = vti6_locate(net, &p, 0);
1029 t = netdev_priv(dev);
1031 return vti6_update(t, &p);
1034 static size_t vti6_get_size(const struct net_device *dev)
1039 /* IFLA_VTI_LOCAL */
1040 nla_total_size(sizeof(struct in6_addr)) +
1041 /* IFLA_VTI_REMOTE */
1042 nla_total_size(sizeof(struct in6_addr)) +
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 goto nla_put_failure;
1067 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1068 [IFLA_VTI_LINK] = { .type = NLA_U32 },
1069 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) },
1070 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) },
1071 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
1072 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
1075 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1077 .maxtype = IFLA_VTI_MAX,
1078 .policy = vti6_policy,
1079 .priv_size = sizeof(struct ip6_tnl),
1080 .setup = vti6_dev_setup,
1081 .validate = vti6_validate,
1082 .newlink = vti6_newlink,
1083 .dellink = vti6_dellink,
1084 .changelink = vti6_changelink,
1085 .get_size = vti6_get_size,
1086 .fill_info = vti6_fill_info,
1087 .get_link_net = ip6_tnl_get_link_net,
1090 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n)
1096 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1097 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1099 unregister_netdevice_queue(t->dev, &list);
1100 t = rtnl_dereference(t->next);
1104 t = rtnl_dereference(ip6n->tnls_wc[0]);
1105 unregister_netdevice_queue(t->dev, &list);
1106 unregister_netdevice_many(&list);
1109 static int __net_init vti6_init_net(struct net *net)
1111 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1112 struct ip6_tnl *t = NULL;
1115 ip6n->tnls[0] = ip6n->tnls_wc;
1116 ip6n->tnls[1] = ip6n->tnls_r_l;
1119 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1120 NET_NAME_UNKNOWN, vti6_dev_setup);
1122 if (!ip6n->fb_tnl_dev)
1124 dev_net_set(ip6n->fb_tnl_dev, net);
1125 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1127 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1131 err = register_netdev(ip6n->fb_tnl_dev);
1135 t = netdev_priv(ip6n->fb_tnl_dev);
1137 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1141 vti6_dev_free(ip6n->fb_tnl_dev);
1146 static void __net_exit vti6_exit_net(struct net *net)
1148 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1151 vti6_destroy_tunnels(ip6n);
1155 static struct pernet_operations vti6_net_ops = {
1156 .init = vti6_init_net,
1157 .exit = vti6_exit_net,
1159 .size = sizeof(struct vti6_net),
1162 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1163 .handler = vti6_rcv,
1164 .cb_handler = vti6_rcv_cb,
1165 .err_handler = vti6_err,
1169 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1170 .handler = vti6_rcv,
1171 .cb_handler = vti6_rcv_cb,
1172 .err_handler = vti6_err,
1176 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1177 .handler = vti6_rcv,
1178 .cb_handler = vti6_rcv_cb,
1179 .err_handler = vti6_err,
1183 static bool is_vti6_tunnel(const struct net_device *dev)
1185 return dev->netdev_ops == &vti6_netdev_ops;
1188 static int vti6_device_event(struct notifier_block *unused,
1189 unsigned long event, void *ptr)
1191 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1192 struct ip6_tnl *t = netdev_priv(dev);
1194 if (!is_vti6_tunnel(dev))
1199 if (!net_eq(t->net, dev_net(dev)))
1200 xfrm_garbage_collect(t->net);
1206 static struct notifier_block vti6_notifier_block __read_mostly = {
1207 .notifier_call = vti6_device_event,
1211 * vti6_tunnel_init - register protocol and reserve needed resources
1213 * Return: 0 on success
1215 static int __init vti6_tunnel_init(void)
1220 register_netdevice_notifier(&vti6_notifier_block);
1222 msg = "tunnel device";
1223 err = register_pernet_device(&vti6_net_ops);
1225 goto pernet_dev_failed;
1227 msg = "tunnel protocols";
1228 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1230 goto xfrm_proto_esp_failed;
1231 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1233 goto xfrm_proto_ah_failed;
1234 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1236 goto xfrm_proto_comp_failed;
1238 msg = "netlink interface";
1239 err = rtnl_link_register(&vti6_link_ops);
1241 goto rtnl_link_failed;
1246 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1247 xfrm_proto_comp_failed:
1248 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1249 xfrm_proto_ah_failed:
1250 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1251 xfrm_proto_esp_failed:
1252 unregister_pernet_device(&vti6_net_ops);
1254 unregister_netdevice_notifier(&vti6_notifier_block);
1255 pr_err("vti6 init: failed to register %s\n", msg);
1260 * vti6_tunnel_cleanup - free resources and unregister protocol
1262 static void __exit vti6_tunnel_cleanup(void)
1264 rtnl_link_unregister(&vti6_link_ops);
1265 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1266 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1267 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1268 unregister_pernet_device(&vti6_net_ops);
1269 unregister_netdevice_notifier(&vti6_notifier_block);
1272 module_init(vti6_tunnel_init);
1273 module_exit(vti6_tunnel_cleanup);
1274 MODULE_LICENSE("GPL");
1275 MODULE_ALIAS_RTNL_LINK("vti6");
1276 MODULE_ALIAS_NETDEV("ip6_vti0");
1277 MODULE_AUTHOR("Steffen Klassert");
1278 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");