1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * IPv6 virtual tunneling interface
5 * Copyright (C) 2013 secunet Security Networks AG
11 * net/ipv6/ip6_tunnel.c
14 #include <linux/module.h>
15 #include <linux/capability.h>
16 #include <linux/errno.h>
17 #include <linux/types.h>
18 #include <linux/sockios.h>
19 #include <linux/icmp.h>
23 #include <linux/net.h>
24 #include <linux/in6.h>
25 #include <linux/netdevice.h>
26 #include <linux/if_arp.h>
27 #include <linux/icmpv6.h>
28 #include <linux/init.h>
29 #include <linux/route.h>
30 #include <linux/rtnetlink.h>
31 #include <linux/netfilter_ipv6.h>
32 #include <linux/slab.h>
33 #include <linux/hash.h>
35 #include <linux/uaccess.h>
36 #include <linux/atomic.h>
40 #include <net/ip_tunnels.h>
42 #include <net/ip6_route.h>
43 #include <net/addrconf.h>
44 #include <net/ip6_tunnel.h>
46 #include <net/net_namespace.h>
47 #include <net/netns/generic.h>
48 #include <linux/etherdevice.h>
50 #define IP6_VTI_HASH_SIZE_SHIFT 5
51 #define IP6_VTI_HASH_SIZE (1 << IP6_VTI_HASH_SIZE_SHIFT)
53 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
55 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
57 return hash_32(hash, IP6_VTI_HASH_SIZE_SHIFT);
60 static int vti6_dev_init(struct net_device *dev);
61 static void vti6_dev_setup(struct net_device *dev);
62 static struct rtnl_link_ops vti6_link_ops __read_mostly;
64 static unsigned int vti6_net_id __read_mostly;
66 /* the vti6 tunnel fallback device */
67 struct net_device *fb_tnl_dev;
68 /* lists for storing tunnels in use */
69 struct ip6_tnl __rcu *tnls_r_l[IP6_VTI_HASH_SIZE];
70 struct ip6_tnl __rcu *tnls_wc[1];
71 struct ip6_tnl __rcu **tnls[2];
74 #define for_each_vti6_tunnel_rcu(start) \
75 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
78 * vti6_tnl_lookup - fetch tunnel matching the end-point addresses
79 * @net: network namespace
80 * @remote: the address of the tunnel exit-point
81 * @local: the address of the tunnel entry-point
84 * tunnel matching given end-points if found,
85 * else fallback tunnel if its device is up,
88 static struct ip6_tnl *
89 vti6_tnl_lookup(struct net *net, const struct in6_addr *remote,
90 const struct in6_addr *local)
92 unsigned int hash = HASH(remote, local);
94 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
97 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
98 if (ipv6_addr_equal(local, &t->parms.laddr) &&
99 ipv6_addr_equal(remote, &t->parms.raddr) &&
100 (t->dev->flags & IFF_UP))
104 memset(&any, 0, sizeof(any));
105 hash = HASH(&any, local);
106 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
107 if (ipv6_addr_equal(local, &t->parms.laddr) &&
108 (t->dev->flags & IFF_UP))
112 hash = HASH(remote, &any);
113 for_each_vti6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
114 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
115 (t->dev->flags & IFF_UP))
119 t = rcu_dereference(ip6n->tnls_wc[0]);
120 if (t && (t->dev->flags & IFF_UP))
127 * vti6_tnl_bucket - get head of list matching given tunnel parameters
128 * @ip6n: the private data for ip6_vti in the netns
129 * @p: parameters containing tunnel end-points
132 * vti6_tnl_bucket() returns the head of the list matching the
133 * &struct in6_addr entries laddr and raddr in @p.
135 * Return: head of IPv6 tunnel list
137 static struct ip6_tnl __rcu **
138 vti6_tnl_bucket(struct vti6_net *ip6n, const struct __ip6_tnl_parm *p)
140 const struct in6_addr *remote = &p->raddr;
141 const struct in6_addr *local = &p->laddr;
145 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
147 h = HASH(remote, local);
149 return &ip6n->tnls[prio][h];
153 vti6_tnl_link(struct vti6_net *ip6n, struct ip6_tnl *t)
155 struct ip6_tnl __rcu **tp = vti6_tnl_bucket(ip6n, &t->parms);
157 rcu_assign_pointer(t->next, rtnl_dereference(*tp));
158 rcu_assign_pointer(*tp, t);
162 vti6_tnl_unlink(struct vti6_net *ip6n, struct ip6_tnl *t)
164 struct ip6_tnl __rcu **tp;
165 struct ip6_tnl *iter;
167 for (tp = vti6_tnl_bucket(ip6n, &t->parms);
168 (iter = rtnl_dereference(*tp)) != NULL;
171 rcu_assign_pointer(*tp, t->next);
177 static int vti6_tnl_create2(struct net_device *dev)
179 struct ip6_tnl *t = netdev_priv(dev);
180 struct net *net = dev_net(dev);
181 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
184 dev->rtnl_link_ops = &vti6_link_ops;
185 err = register_netdevice(dev);
189 strcpy(t->parms.name, dev->name);
191 vti6_tnl_link(ip6n, t);
199 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
201 struct net_device *dev;
207 if (!dev_valid_name(p->name))
209 strscpy(name, p->name, IFNAMSIZ);
211 sprintf(name, "ip6_vti%%d");
214 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup);
218 dev_net_set(dev, net);
220 t = netdev_priv(dev);
222 t->net = dev_net(dev);
224 err = vti6_tnl_create2(dev);
237 * vti6_locate - find or create tunnel matching given parameters
238 * @net: network namespace
239 * @p: tunnel parameters
240 * @create: != 0 if allowed to create new tunnel if no match found
243 * vti6_locate() first tries to locate an existing tunnel
244 * based on @parms. If this is unsuccessful, but @create is set a new
245 * tunnel device is created and registered for use.
248 * matching tunnel or NULL
250 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
253 const struct in6_addr *remote = &p->raddr;
254 const struct in6_addr *local = &p->laddr;
255 struct ip6_tnl __rcu **tp;
257 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
259 for (tp = vti6_tnl_bucket(ip6n, p);
260 (t = rtnl_dereference(*tp)) != NULL;
262 if (ipv6_addr_equal(local, &t->parms.laddr) &&
263 ipv6_addr_equal(remote, &t->parms.raddr)) {
272 return vti6_tnl_create(net, p);
276 * vti6_dev_uninit - tunnel device uninitializer
277 * @dev: the device to be destroyed
280 * vti6_dev_uninit() removes tunnel from its list
282 static void vti6_dev_uninit(struct net_device *dev)
284 struct ip6_tnl *t = netdev_priv(dev);
285 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id);
287 if (dev == ip6n->fb_tnl_dev)
288 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
290 vti6_tnl_unlink(ip6n, t);
291 netdev_put(dev, &t->dev_tracker);
294 static int vti6_input_proto(struct sk_buff *skb, int nexthdr, __be32 spi,
298 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
301 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
303 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
308 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
313 ipv6h = ipv6_hdr(skb);
314 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
315 DEV_STATS_INC(t->dev, rx_dropped);
322 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t;
323 XFRM_SPI_SKB_CB(skb)->family = AF_INET6;
324 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct ipv6hdr, daddr);
325 return xfrm_input(skb, nexthdr, spi, encap_type);
334 static int vti6_rcv(struct sk_buff *skb)
336 int nexthdr = skb_network_header(skb)[IP6CB(skb)->nhoff];
338 return vti6_input_proto(skb, nexthdr, 0, 0);
341 static int vti6_rcv_cb(struct sk_buff *skb, int err)
343 unsigned short family;
344 struct net_device *dev;
345 struct xfrm_state *x;
346 const struct xfrm_mode *inner_mode;
347 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
348 u32 orig_mark = skb->mark;
357 DEV_STATS_INC(dev, rx_errors);
358 DEV_STATS_INC(dev, rx_dropped);
363 x = xfrm_input_state(skb);
365 inner_mode = &x->inner_mode;
367 if (x->sel.family == AF_UNSPEC) {
368 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
369 if (inner_mode == NULL) {
370 XFRM_INC_STATS(dev_net(skb->dev),
371 LINUX_MIB_XFRMINSTATEMODEERROR);
376 family = inner_mode->family;
378 skb->mark = be32_to_cpu(t->parms.i_key);
379 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
380 skb->mark = orig_mark;
385 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
387 dev_sw_netstats_rx_add(dev, skb->len);
393 * vti6_addr_conflict - compare packet addresses to tunnel's own
394 * @t: the outgoing tunnel device
395 * @hdr: IPv6 header from the incoming packet
398 * Avoid trivial tunneling loop by checking that tunnel exit-point
399 * doesn't match source of incoming packet.
406 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
408 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
411 static bool vti6_state_check(const struct xfrm_state *x,
412 const struct in6_addr *dst,
413 const struct in6_addr *src)
415 xfrm_address_t *daddr = (xfrm_address_t *)dst;
416 xfrm_address_t *saddr = (xfrm_address_t *)src;
418 /* if there is no transform then this tunnel is not functional.
419 * Or if the xfrm is not mode tunnel.
421 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
422 x->props.family != AF_INET6)
425 if (ipv6_addr_any(dst))
426 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
428 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
435 * vti6_xmit - send a packet
436 * @skb: the outgoing socket buffer
437 * @dev: the outgoing tunnel device
438 * @fl: the flow informations for the xfrm_lookup
441 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
443 struct ip6_tnl *t = netdev_priv(dev);
444 struct dst_entry *dst = skb_dst(skb);
445 struct net_device *tdev;
446 struct xfrm_state *x;
447 int pkt_len = skb->len;
452 switch (skb->protocol) {
453 case htons(ETH_P_IP): {
456 fl->u.ip4.flowi4_oif = dev->ifindex;
457 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
458 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4);
460 goto tx_err_link_failure;
462 skb_dst_set(skb, dst);
465 case htons(ETH_P_IPV6):
466 fl->u.ip6.flowi6_oif = dev->ifindex;
467 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
468 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
472 goto tx_err_link_failure;
474 skb_dst_set(skb, dst);
477 goto tx_err_link_failure;
482 dst = xfrm_lookup_route(t->net, dst, fl, NULL, 0);
486 goto tx_err_link_failure;
489 if (dst->flags & DST_XFRM_QUEUE)
493 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
494 goto tx_err_link_failure;
496 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
497 (const struct in6_addr *)&x->id.daddr))
498 goto tx_err_link_failure;
503 DEV_STATS_INC(dev, collisions);
504 net_warn_ratelimited("%s: Local routing loop detected!\n",
506 goto tx_err_dst_release;
510 if (skb->len > mtu) {
511 skb_dst_update_pmtu_no_confirm(skb, mtu);
513 if (skb->protocol == htons(ETH_P_IPV6)) {
514 if (mtu < IPV6_MIN_MTU)
517 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
519 if (!(ip_hdr(skb)->frag_off & htons(IP_DF)))
521 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
526 goto tx_err_dst_release;
530 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
531 skb_dst_set(skb, dst);
532 skb->dev = skb_dst(skb)->dev;
534 err = dst_output(t->net, skb->sk, skb);
535 if (net_xmit_eval(err) == 0)
537 iptunnel_xmit_stats(dev, err);
541 DEV_STATS_INC(dev, tx_carrier_errors);
542 dst_link_failure(skb);
549 vti6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
551 struct ip6_tnl *t = netdev_priv(dev);
555 if (!pskb_inet_may_pull(skb))
558 memset(&fl, 0, sizeof(fl));
560 switch (skb->protocol) {
561 case htons(ETH_P_IPV6):
562 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
563 vti6_addr_conflict(t, ipv6_hdr(skb)))
566 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
567 xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET6);
569 case htons(ETH_P_IP):
570 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
571 xfrm_decode_session(dev_net(dev), skb, &fl, AF_INET);
577 /* override mark with tunnel output key */
578 fl.flowi_mark = be32_to_cpu(t->parms.o_key);
580 ret = vti6_xmit(skb, dev, &fl);
587 DEV_STATS_INC(dev, tx_errors);
588 DEV_STATS_INC(dev, tx_dropped);
593 static int vti6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
594 u8 type, u8 code, int offset, __be32 info)
598 struct xfrm_state *x;
600 struct ip_esp_hdr *esph;
601 struct ip_auth_hdr *ah;
602 struct ip_comp_hdr *ipch;
603 struct net *net = dev_net(skb->dev);
604 const struct ipv6hdr *iph = (const struct ipv6hdr *)skb->data;
605 int protocol = iph->nexthdr;
607 t = vti6_tnl_lookup(dev_net(skb->dev), &iph->daddr, &iph->saddr);
611 mark = be32_to_cpu(t->parms.o_key);
615 esph = (struct ip_esp_hdr *)(skb->data + offset);
619 ah = (struct ip_auth_hdr *)(skb->data + offset);
623 ipch = (struct ip_comp_hdr *)(skb->data + offset);
624 spi = htonl(ntohs(ipch->cpi));
630 if (type != ICMPV6_PKT_TOOBIG &&
631 type != NDISC_REDIRECT)
634 x = xfrm_state_lookup(net, mark, (const xfrm_address_t *)&iph->daddr,
635 spi, protocol, AF_INET6);
639 if (type == NDISC_REDIRECT)
640 ip6_redirect(skb, net, skb->dev->ifindex, 0,
641 sock_net_uid(net, NULL));
643 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
649 static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu)
651 struct net_device *dev = t->dev;
652 struct __ip6_tnl_parm *p = &t->parms;
653 struct net_device *tdev = NULL;
656 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
657 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
659 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
660 IP6_TNL_F_CAP_PER_PACKET);
661 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
663 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
664 dev->flags |= IFF_POINTOPOINT;
666 dev->flags &= ~IFF_POINTOPOINT;
668 if (keep_mtu && dev->mtu) {
670 clamp(dev->mtu, dev->min_mtu, dev->max_mtu));
674 if (p->flags & IP6_TNL_F_CAP_XMIT) {
675 int strict = (ipv6_addr_type(&p->raddr) &
676 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
677 struct rt6_info *rt = rt6_lookup(t->net,
678 &p->raddr, &p->laddr,
679 p->link, NULL, strict);
686 if (!tdev && p->link)
687 tdev = __dev_get_by_index(t->net, p->link);
690 mtu = tdev->mtu - sizeof(struct ipv6hdr);
692 mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr);
694 dev->mtu = max_t(int, mtu, IPV4_MIN_MTU);
698 * vti6_tnl_change - update the tunnel parameters
699 * @t: tunnel to be changed
700 * @p: tunnel configuration parameters
701 * @keep_mtu: MTU was set from userspace, don't re-compute it
704 * vti6_tnl_change() updates the tunnel parameters
707 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
710 t->parms.laddr = p->laddr;
711 t->parms.raddr = p->raddr;
712 t->parms.link = p->link;
713 t->parms.i_key = p->i_key;
714 t->parms.o_key = p->o_key;
715 t->parms.proto = p->proto;
716 t->parms.fwmark = p->fwmark;
717 dst_cache_reset(&t->dst_cache);
718 vti6_link_config(t, keep_mtu);
722 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p,
725 struct net *net = dev_net(t->dev);
726 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
729 vti6_tnl_unlink(ip6n, t);
731 err = vti6_tnl_change(t, p, keep_mtu);
732 vti6_tnl_link(ip6n, t);
733 netdev_state_change(t->dev);
738 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
747 memcpy(p->name, u->name, sizeof(u->name));
751 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
759 u->i_flags |= GRE_KEY;
761 u->o_flags |= GRE_KEY;
764 memcpy(u->name, p->name, sizeof(u->name));
768 * vti6_siocdevprivate - configure vti6 tunnels from userspace
769 * @dev: virtual device associated with tunnel
771 * @data: parameters passed from userspace
772 * @cmd: command to be performed
775 * vti6_siocdevprivate() is used for managing vti6 tunnels
778 * The possible commands are the following:
779 * %SIOCGETTUNNEL: get tunnel parameters for device
780 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
781 * %SIOCCHGTUNNEL: change tunnel parameters to those given
782 * %SIOCDELTUNNEL: delete tunnel
784 * The fallback device "ip6_vti0", created during module
785 * initialization, can be used for creating other tunnel devices.
789 * %-EFAULT if unable to copy data to or from userspace,
790 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
791 * %-EINVAL if passed tunnel parameters are invalid,
792 * %-EEXIST if changing a tunnel's parameters would cause a conflict
793 * %-ENODEV if attempting to change or delete a nonexisting device
796 vti6_siocdevprivate(struct net_device *dev, struct ifreq *ifr, void __user *data, int cmd)
799 struct ip6_tnl_parm2 p;
800 struct __ip6_tnl_parm p1;
801 struct ip6_tnl *t = NULL;
802 struct net *net = dev_net(dev);
803 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
805 memset(&p1, 0, sizeof(p1));
809 if (dev == ip6n->fb_tnl_dev) {
810 if (copy_from_user(&p, data, sizeof(p))) {
814 vti6_parm_from_user(&p1, &p);
815 t = vti6_locate(net, &p1, 0);
817 memset(&p, 0, sizeof(p));
820 t = netdev_priv(dev);
821 vti6_parm_to_user(&p, &t->parms);
822 if (copy_to_user(data, &p, sizeof(p)))
828 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
831 if (copy_from_user(&p, data, sizeof(p)))
834 if (p.proto != IPPROTO_IPV6 && p.proto != 0)
836 vti6_parm_from_user(&p1, &p);
837 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
838 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
845 t = netdev_priv(dev);
847 err = vti6_update(t, &p1, false);
851 vti6_parm_to_user(&p, &t->parms);
852 if (copy_to_user(data, &p, sizeof(p)))
856 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
860 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
863 if (dev == ip6n->fb_tnl_dev) {
865 if (copy_from_user(&p, data, sizeof(p)))
868 vti6_parm_from_user(&p1, &p);
869 t = vti6_locate(net, &p1, 0);
873 if (t->dev == ip6n->fb_tnl_dev)
878 unregister_netdevice(dev);
886 static const struct net_device_ops vti6_netdev_ops = {
887 .ndo_init = vti6_dev_init,
888 .ndo_uninit = vti6_dev_uninit,
889 .ndo_start_xmit = vti6_tnl_xmit,
890 .ndo_siocdevprivate = vti6_siocdevprivate,
891 .ndo_get_iflink = ip6_tnl_get_iflink,
895 * vti6_dev_setup - setup virtual tunnel device
896 * @dev: virtual device associated with tunnel
899 * Initialize function pointers and device parameters
901 static void vti6_dev_setup(struct net_device *dev)
903 dev->netdev_ops = &vti6_netdev_ops;
904 dev->header_ops = &ip_tunnel_header_ops;
905 dev->needs_free_netdev = true;
907 dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
908 dev->type = ARPHRD_TUNNEL6;
909 dev->min_mtu = IPV4_MIN_MTU;
910 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr);
911 dev->flags |= IFF_NOARP;
912 dev->addr_len = sizeof(struct in6_addr);
914 /* This perm addr will be used as interface identifier by IPv6 */
915 dev->addr_assign_type = NET_ADDR_RANDOM;
916 eth_random_addr(dev->perm_addr);
920 * vti6_dev_init_gen - general initializer for all tunnel devices
921 * @dev: virtual device associated with tunnel
923 static inline int vti6_dev_init_gen(struct net_device *dev)
925 struct ip6_tnl *t = netdev_priv(dev);
928 t->net = dev_net(dev);
929 netdev_hold(dev, &t->dev_tracker, GFP_KERNEL);
930 netdev_lockdep_set_classes(dev);
935 * vti6_dev_init - initializer for all non fallback tunnel devices
936 * @dev: virtual device associated with tunnel
938 static int vti6_dev_init(struct net_device *dev)
940 struct ip6_tnl *t = netdev_priv(dev);
941 int err = vti6_dev_init_gen(dev);
945 vti6_link_config(t, true);
950 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
951 * @dev: fallback device
955 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
957 struct ip6_tnl *t = netdev_priv(dev);
958 struct net *net = dev_net(dev);
959 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
961 t->parms.proto = IPPROTO_IPV6;
963 rcu_assign_pointer(ip6n->tnls_wc[0], t);
967 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[],
968 struct netlink_ext_ack *extack)
973 static void vti6_netlink_parms(struct nlattr *data[],
974 struct __ip6_tnl_parm *parms)
976 memset(parms, 0, sizeof(*parms));
981 if (data[IFLA_VTI_LINK])
982 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
984 if (data[IFLA_VTI_LOCAL])
985 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
987 if (data[IFLA_VTI_REMOTE])
988 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
990 if (data[IFLA_VTI_IKEY])
991 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
993 if (data[IFLA_VTI_OKEY])
994 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
996 if (data[IFLA_VTI_FWMARK])
997 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
1000 static int vti6_newlink(struct net *src_net, struct net_device *dev,
1001 struct nlattr *tb[], struct nlattr *data[],
1002 struct netlink_ext_ack *extack)
1004 struct net *net = dev_net(dev);
1007 nt = netdev_priv(dev);
1008 vti6_netlink_parms(data, &nt->parms);
1010 nt->parms.proto = IPPROTO_IPV6;
1012 if (vti6_locate(net, &nt->parms, 0))
1015 return vti6_tnl_create2(dev);
1018 static void vti6_dellink(struct net_device *dev, struct list_head *head)
1020 struct net *net = dev_net(dev);
1021 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1023 if (dev != ip6n->fb_tnl_dev)
1024 unregister_netdevice_queue(dev, head);
1027 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
1028 struct nlattr *data[],
1029 struct netlink_ext_ack *extack)
1032 struct __ip6_tnl_parm p;
1033 struct net *net = dev_net(dev);
1034 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1036 if (dev == ip6n->fb_tnl_dev)
1039 vti6_netlink_parms(data, &p);
1041 t = vti6_locate(net, &p, 0);
1047 t = netdev_priv(dev);
1049 return vti6_update(t, &p, tb && tb[IFLA_MTU]);
1052 static size_t vti6_get_size(const struct net_device *dev)
1057 /* IFLA_VTI_LOCAL */
1058 nla_total_size(sizeof(struct in6_addr)) +
1059 /* IFLA_VTI_REMOTE */
1060 nla_total_size(sizeof(struct in6_addr)) +
1065 /* IFLA_VTI_FWMARK */
1070 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1072 struct ip6_tnl *tunnel = netdev_priv(dev);
1073 struct __ip6_tnl_parm *parm = &tunnel->parms;
1075 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1076 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1077 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1078 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1079 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) ||
1080 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark))
1081 goto nla_put_failure;
1088 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1089 [IFLA_VTI_LINK] = { .type = NLA_U32 },
1090 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) },
1091 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) },
1092 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
1093 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
1094 [IFLA_VTI_FWMARK] = { .type = NLA_U32 },
1097 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1099 .maxtype = IFLA_VTI_MAX,
1100 .policy = vti6_policy,
1101 .priv_size = sizeof(struct ip6_tnl),
1102 .setup = vti6_dev_setup,
1103 .validate = vti6_validate,
1104 .newlink = vti6_newlink,
1105 .dellink = vti6_dellink,
1106 .changelink = vti6_changelink,
1107 .get_size = vti6_get_size,
1108 .fill_info = vti6_fill_info,
1109 .get_link_net = ip6_tnl_get_link_net,
1112 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n,
1113 struct list_head *list)
1118 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1119 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1121 unregister_netdevice_queue(t->dev, list);
1122 t = rtnl_dereference(t->next);
1126 t = rtnl_dereference(ip6n->tnls_wc[0]);
1128 unregister_netdevice_queue(t->dev, list);
1131 static int __net_init vti6_init_net(struct net *net)
1133 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1134 struct ip6_tnl *t = NULL;
1137 ip6n->tnls[0] = ip6n->tnls_wc;
1138 ip6n->tnls[1] = ip6n->tnls_r_l;
1140 if (!net_has_fallback_tunnels(net))
1143 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1144 NET_NAME_UNKNOWN, vti6_dev_setup);
1146 if (!ip6n->fb_tnl_dev)
1148 dev_net_set(ip6n->fb_tnl_dev, net);
1149 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1151 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1155 err = register_netdev(ip6n->fb_tnl_dev);
1159 t = netdev_priv(ip6n->fb_tnl_dev);
1161 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1165 free_netdev(ip6n->fb_tnl_dev);
1170 static void __net_exit vti6_exit_batch_rtnl(struct list_head *net_list,
1171 struct list_head *dev_to_kill)
1173 struct vti6_net *ip6n;
1177 list_for_each_entry(net, net_list, exit_list) {
1178 ip6n = net_generic(net, vti6_net_id);
1179 vti6_destroy_tunnels(ip6n, dev_to_kill);
1183 static struct pernet_operations vti6_net_ops = {
1184 .init = vti6_init_net,
1185 .exit_batch_rtnl = vti6_exit_batch_rtnl,
1187 .size = sizeof(struct vti6_net),
1190 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1191 .handler = vti6_rcv,
1192 .input_handler = vti6_input_proto,
1193 .cb_handler = vti6_rcv_cb,
1194 .err_handler = vti6_err,
1198 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1199 .handler = vti6_rcv,
1200 .input_handler = vti6_input_proto,
1201 .cb_handler = vti6_rcv_cb,
1202 .err_handler = vti6_err,
1206 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1207 .handler = vti6_rcv,
1208 .input_handler = vti6_input_proto,
1209 .cb_handler = vti6_rcv_cb,
1210 .err_handler = vti6_err,
1214 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1215 static int vti6_rcv_tunnel(struct sk_buff *skb)
1217 const xfrm_address_t *saddr;
1220 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr;
1221 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr);
1223 return vti6_input_proto(skb, IPPROTO_IPV6, spi, 0);
1226 static struct xfrm6_tunnel vti_ipv6_handler __read_mostly = {
1227 .handler = vti6_rcv_tunnel,
1228 .cb_handler = vti6_rcv_cb,
1229 .err_handler = vti6_err,
1233 static struct xfrm6_tunnel vti_ip6ip_handler __read_mostly = {
1234 .handler = vti6_rcv_tunnel,
1235 .cb_handler = vti6_rcv_cb,
1236 .err_handler = vti6_err,
1242 * vti6_tunnel_init - register protocol and reserve needed resources
1244 * Return: 0 on success
1246 static int __init vti6_tunnel_init(void)
1251 msg = "tunnel device";
1252 err = register_pernet_device(&vti6_net_ops);
1254 goto pernet_dev_failed;
1256 msg = "tunnel protocols";
1257 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1259 goto xfrm_proto_esp_failed;
1260 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1262 goto xfrm_proto_ah_failed;
1263 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1265 goto xfrm_proto_comp_failed;
1266 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1267 msg = "ipv6 tunnel";
1268 err = xfrm6_tunnel_register(&vti_ipv6_handler, AF_INET6);
1270 goto vti_tunnel_ipv6_failed;
1271 err = xfrm6_tunnel_register(&vti_ip6ip_handler, AF_INET);
1273 goto vti_tunnel_ip6ip_failed;
1276 msg = "netlink interface";
1277 err = rtnl_link_register(&vti6_link_ops);
1279 goto rtnl_link_failed;
1284 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1285 err = xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET);
1286 vti_tunnel_ip6ip_failed:
1287 err = xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6);
1288 vti_tunnel_ipv6_failed:
1290 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1291 xfrm_proto_comp_failed:
1292 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1293 xfrm_proto_ah_failed:
1294 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1295 xfrm_proto_esp_failed:
1296 unregister_pernet_device(&vti6_net_ops);
1298 pr_err("vti6 init: failed to register %s\n", msg);
1303 * vti6_tunnel_cleanup - free resources and unregister protocol
1305 static void __exit vti6_tunnel_cleanup(void)
1307 rtnl_link_unregister(&vti6_link_ops);
1308 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1309 xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET);
1310 xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6);
1312 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1313 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1314 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1315 unregister_pernet_device(&vti6_net_ops);
1318 module_init(vti6_tunnel_init);
1319 module_exit(vti6_tunnel_cleanup);
1320 MODULE_LICENSE("GPL");
1321 MODULE_ALIAS_RTNL_LINK("vti6");
1322 MODULE_ALIAS_NETDEV("ip6_vti0");
1323 MODULE_AUTHOR("Steffen Klassert");
1324 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");