1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * IPv6 virtual tunneling interface
5 * Copyright (C) 2013 secunet Security Networks AG
8 * Steffen Klassert <steffen.klassert@secunet.com>
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 void vti6_dev_free(struct net_device *dev)
179 free_percpu(dev->tstats);
182 static int vti6_tnl_create2(struct net_device *dev)
184 struct ip6_tnl *t = netdev_priv(dev);
185 struct net *net = dev_net(dev);
186 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
189 dev->rtnl_link_ops = &vti6_link_ops;
190 err = register_netdevice(dev);
194 strcpy(t->parms.name, dev->name);
196 vti6_tnl_link(ip6n, t);
204 static struct ip6_tnl *vti6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
206 struct net_device *dev;
212 if (!dev_valid_name(p->name))
214 strlcpy(name, p->name, IFNAMSIZ);
216 sprintf(name, "ip6_vti%%d");
219 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN, vti6_dev_setup);
223 dev_net_set(dev, net);
225 t = netdev_priv(dev);
227 t->net = dev_net(dev);
229 err = vti6_tnl_create2(dev);
242 * vti6_locate - find or create tunnel matching given parameters
243 * @net: network namespace
244 * @p: tunnel parameters
245 * @create: != 0 if allowed to create new tunnel if no match found
248 * vti6_locate() first tries to locate an existing tunnel
249 * based on @parms. If this is unsuccessful, but @create is set a new
250 * tunnel device is created and registered for use.
253 * matching tunnel or NULL
255 static struct ip6_tnl *vti6_locate(struct net *net, struct __ip6_tnl_parm *p,
258 const struct in6_addr *remote = &p->raddr;
259 const struct in6_addr *local = &p->laddr;
260 struct ip6_tnl __rcu **tp;
262 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
264 for (tp = vti6_tnl_bucket(ip6n, p);
265 (t = rtnl_dereference(*tp)) != NULL;
267 if (ipv6_addr_equal(local, &t->parms.laddr) &&
268 ipv6_addr_equal(remote, &t->parms.raddr)) {
277 return vti6_tnl_create(net, p);
281 * vti6_dev_uninit - tunnel device uninitializer
282 * @dev: the device to be destroyed
285 * vti6_dev_uninit() removes tunnel from its list
287 static void vti6_dev_uninit(struct net_device *dev)
289 struct ip6_tnl *t = netdev_priv(dev);
290 struct vti6_net *ip6n = net_generic(t->net, vti6_net_id);
292 if (dev == ip6n->fb_tnl_dev)
293 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
295 vti6_tnl_unlink(ip6n, t);
296 dev_put_track(dev, &t->dev_tracker);
299 static int vti6_input_proto(struct sk_buff *skb, int nexthdr, __be32 spi,
303 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
306 t = vti6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
308 if (t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) {
313 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
318 ipv6h = ipv6_hdr(skb);
319 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
320 t->dev->stats.rx_dropped++;
327 XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6 = t;
328 XFRM_SPI_SKB_CB(skb)->family = AF_INET6;
329 XFRM_SPI_SKB_CB(skb)->daddroff = offsetof(struct ipv6hdr, daddr);
330 return xfrm_input(skb, nexthdr, spi, encap_type);
339 static int vti6_rcv(struct sk_buff *skb)
341 int nexthdr = skb_network_header(skb)[IP6CB(skb)->nhoff];
343 return vti6_input_proto(skb, nexthdr, 0, 0);
346 static int vti6_rcv_cb(struct sk_buff *skb, int err)
348 unsigned short family;
349 struct net_device *dev;
350 struct xfrm_state *x;
351 const struct xfrm_mode *inner_mode;
352 struct ip6_tnl *t = XFRM_TUNNEL_SKB_CB(skb)->tunnel.ip6;
353 u32 orig_mark = skb->mark;
362 dev->stats.rx_errors++;
363 dev->stats.rx_dropped++;
368 x = xfrm_input_state(skb);
370 inner_mode = &x->inner_mode;
372 if (x->sel.family == AF_UNSPEC) {
373 inner_mode = xfrm_ip2inner_mode(x, XFRM_MODE_SKB_CB(skb)->protocol);
374 if (inner_mode == NULL) {
375 XFRM_INC_STATS(dev_net(skb->dev),
376 LINUX_MIB_XFRMINSTATEMODEERROR);
381 family = inner_mode->family;
383 skb->mark = be32_to_cpu(t->parms.i_key);
384 ret = xfrm_policy_check(NULL, XFRM_POLICY_IN, skb, family);
385 skb->mark = orig_mark;
390 skb_scrub_packet(skb, !net_eq(t->net, dev_net(skb->dev)));
392 dev_sw_netstats_rx_add(dev, skb->len);
398 * vti6_addr_conflict - compare packet addresses to tunnel's own
399 * @t: the outgoing tunnel device
400 * @hdr: IPv6 header from the incoming packet
403 * Avoid trivial tunneling loop by checking that tunnel exit-point
404 * doesn't match source of incoming packet.
411 vti6_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
413 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
416 static bool vti6_state_check(const struct xfrm_state *x,
417 const struct in6_addr *dst,
418 const struct in6_addr *src)
420 xfrm_address_t *daddr = (xfrm_address_t *)dst;
421 xfrm_address_t *saddr = (xfrm_address_t *)src;
423 /* if there is no transform then this tunnel is not functional.
424 * Or if the xfrm is not mode tunnel.
426 if (!x || x->props.mode != XFRM_MODE_TUNNEL ||
427 x->props.family != AF_INET6)
430 if (ipv6_addr_any(dst))
431 return xfrm_addr_equal(saddr, &x->props.saddr, AF_INET6);
433 if (!xfrm_state_addr_check(x, daddr, saddr, AF_INET6))
440 * vti6_xmit - send a packet
441 * @skb: the outgoing socket buffer
442 * @dev: the outgoing tunnel device
443 * @fl: the flow informations for the xfrm_lookup
446 vti6_xmit(struct sk_buff *skb, struct net_device *dev, struct flowi *fl)
448 struct ip6_tnl *t = netdev_priv(dev);
449 struct net_device_stats *stats = &t->dev->stats;
450 struct dst_entry *dst = skb_dst(skb);
451 struct net_device *tdev;
452 struct xfrm_state *x;
453 int pkt_len = skb->len;
458 switch (skb->protocol) {
459 case htons(ETH_P_IP): {
462 fl->u.ip4.flowi4_oif = dev->ifindex;
463 fl->u.ip4.flowi4_flags |= FLOWI_FLAG_ANYSRC;
464 rt = __ip_route_output_key(dev_net(dev), &fl->u.ip4);
466 goto tx_err_link_failure;
468 skb_dst_set(skb, dst);
471 case htons(ETH_P_IPV6):
472 fl->u.ip6.flowi6_oif = dev->ifindex;
473 fl->u.ip6.flowi6_flags |= FLOWI_FLAG_ANYSRC;
474 dst = ip6_route_output(dev_net(dev), NULL, &fl->u.ip6);
478 goto tx_err_link_failure;
480 skb_dst_set(skb, dst);
483 goto tx_err_link_failure;
488 dst = xfrm_lookup_route(t->net, dst, fl, NULL, 0);
492 goto tx_err_link_failure;
495 if (dst->flags & DST_XFRM_QUEUE)
499 if (!vti6_state_check(x, &t->parms.raddr, &t->parms.laddr))
500 goto tx_err_link_failure;
502 if (!ip6_tnl_xmit_ctl(t, (const struct in6_addr *)&x->props.saddr,
503 (const struct in6_addr *)&x->id.daddr))
504 goto tx_err_link_failure;
510 net_warn_ratelimited("%s: Local routing loop detected!\n",
512 goto tx_err_dst_release;
516 if (skb->len > mtu) {
517 skb_dst_update_pmtu_no_confirm(skb, mtu);
519 if (skb->protocol == htons(ETH_P_IPV6)) {
520 if (mtu < IPV6_MIN_MTU)
523 icmpv6_ndo_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
525 if (!(ip_hdr(skb)->frag_off & htons(IP_DF)))
527 icmp_ndo_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
532 goto tx_err_dst_release;
536 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
537 skb_dst_set(skb, dst);
538 skb->dev = skb_dst(skb)->dev;
540 err = dst_output(t->net, skb->sk, skb);
541 if (net_xmit_eval(err) == 0)
543 iptunnel_xmit_stats(dev, err);
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;
562 if (!pskb_inet_may_pull(skb))
565 memset(&fl, 0, sizeof(fl));
567 switch (skb->protocol) {
568 case htons(ETH_P_IPV6):
569 if ((t->parms.proto != IPPROTO_IPV6 && t->parms.proto != 0) ||
570 vti6_addr_conflict(t, ipv6_hdr(skb)))
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,
648 sock_net_uid(net, NULL));
650 ip6_update_pmtu(skb, net, info, 0, 0, sock_net_uid(net, NULL));
656 static void vti6_link_config(struct ip6_tnl *t, bool keep_mtu)
658 struct net_device *dev = t->dev;
659 struct __ip6_tnl_parm *p = &t->parms;
660 struct net_device *tdev = NULL;
663 __dev_addr_set(dev, &p->laddr, sizeof(struct in6_addr));
664 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
666 p->flags &= ~(IP6_TNL_F_CAP_XMIT | IP6_TNL_F_CAP_RCV |
667 IP6_TNL_F_CAP_PER_PACKET);
668 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
670 if (p->flags & IP6_TNL_F_CAP_XMIT && p->flags & IP6_TNL_F_CAP_RCV)
671 dev->flags |= IFF_POINTOPOINT;
673 dev->flags &= ~IFF_POINTOPOINT;
675 if (keep_mtu && dev->mtu) {
676 dev->mtu = clamp(dev->mtu, dev->min_mtu, dev->max_mtu);
680 if (p->flags & IP6_TNL_F_CAP_XMIT) {
681 int strict = (ipv6_addr_type(&p->raddr) &
682 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
683 struct rt6_info *rt = rt6_lookup(t->net,
684 &p->raddr, &p->laddr,
685 p->link, NULL, strict);
692 if (!tdev && p->link)
693 tdev = __dev_get_by_index(t->net, p->link);
696 mtu = tdev->mtu - sizeof(struct ipv6hdr);
698 mtu = ETH_DATA_LEN - LL_MAX_HEADER - sizeof(struct ipv6hdr);
700 dev->mtu = max_t(int, mtu, IPV4_MIN_MTU);
704 * vti6_tnl_change - update the tunnel parameters
705 * @t: tunnel to be changed
706 * @p: tunnel configuration parameters
707 * @keep_mtu: MTU was set from userspace, don't re-compute it
710 * vti6_tnl_change() updates the tunnel parameters
713 vti6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p,
716 t->parms.laddr = p->laddr;
717 t->parms.raddr = p->raddr;
718 t->parms.link = p->link;
719 t->parms.i_key = p->i_key;
720 t->parms.o_key = p->o_key;
721 t->parms.proto = p->proto;
722 t->parms.fwmark = p->fwmark;
723 dst_cache_reset(&t->dst_cache);
724 vti6_link_config(t, keep_mtu);
728 static int vti6_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p,
731 struct net *net = dev_net(t->dev);
732 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
735 vti6_tnl_unlink(ip6n, t);
737 err = vti6_tnl_change(t, p, keep_mtu);
738 vti6_tnl_link(ip6n, t);
739 netdev_state_change(t->dev);
744 vti6_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm2 *u)
753 memcpy(p->name, u->name, sizeof(u->name));
757 vti6_parm_to_user(struct ip6_tnl_parm2 *u, const struct __ip6_tnl_parm *p)
765 u->i_flags |= GRE_KEY;
767 u->o_flags |= GRE_KEY;
770 memcpy(u->name, p->name, sizeof(u->name));
774 * vti6_siocdevprivate - configure vti6 tunnels from userspace
775 * @dev: virtual device associated with tunnel
777 * @data: parameters passed from userspace
778 * @cmd: command to be performed
781 * vti6_siocdevprivate() is used for managing vti6 tunnels
784 * The possible commands are the following:
785 * %SIOCGETTUNNEL: get tunnel parameters for device
786 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
787 * %SIOCCHGTUNNEL: change tunnel parameters to those given
788 * %SIOCDELTUNNEL: delete tunnel
790 * The fallback device "ip6_vti0", created during module
791 * initialization, can be used for creating other tunnel devices.
795 * %-EFAULT if unable to copy data to or from userspace,
796 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
797 * %-EINVAL if passed tunnel parameters are invalid,
798 * %-EEXIST if changing a tunnel's parameters would cause a conflict
799 * %-ENODEV if attempting to change or delete a nonexisting device
802 vti6_siocdevprivate(struct net_device *dev, struct ifreq *ifr, void __user *data, int cmd)
805 struct ip6_tnl_parm2 p;
806 struct __ip6_tnl_parm p1;
807 struct ip6_tnl *t = NULL;
808 struct net *net = dev_net(dev);
809 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
811 memset(&p1, 0, sizeof(p1));
815 if (dev == ip6n->fb_tnl_dev) {
816 if (copy_from_user(&p, data, sizeof(p))) {
820 vti6_parm_from_user(&p1, &p);
821 t = vti6_locate(net, &p1, 0);
823 memset(&p, 0, sizeof(p));
826 t = netdev_priv(dev);
827 vti6_parm_to_user(&p, &t->parms);
828 if (copy_to_user(data, &p, sizeof(p)))
834 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
837 if (copy_from_user(&p, data, sizeof(p)))
840 if (p.proto != IPPROTO_IPV6 && p.proto != 0)
842 vti6_parm_from_user(&p1, &p);
843 t = vti6_locate(net, &p1, cmd == SIOCADDTUNNEL);
844 if (dev != ip6n->fb_tnl_dev && cmd == SIOCCHGTUNNEL) {
851 t = netdev_priv(dev);
853 err = vti6_update(t, &p1, false);
857 vti6_parm_to_user(&p, &t->parms);
858 if (copy_to_user(data, &p, sizeof(p)))
862 err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
866 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
869 if (dev == ip6n->fb_tnl_dev) {
871 if (copy_from_user(&p, data, sizeof(p)))
874 vti6_parm_from_user(&p1, &p);
875 t = vti6_locate(net, &p1, 0);
879 if (t->dev == ip6n->fb_tnl_dev)
884 unregister_netdevice(dev);
892 static const struct net_device_ops vti6_netdev_ops = {
893 .ndo_init = vti6_dev_init,
894 .ndo_uninit = vti6_dev_uninit,
895 .ndo_start_xmit = vti6_tnl_xmit,
896 .ndo_siocdevprivate = vti6_siocdevprivate,
897 .ndo_get_stats64 = dev_get_tstats64,
898 .ndo_get_iflink = ip6_tnl_get_iflink,
902 * vti6_dev_setup - setup virtual tunnel device
903 * @dev: virtual device associated with tunnel
906 * Initialize function pointers and device parameters
908 static void vti6_dev_setup(struct net_device *dev)
910 dev->netdev_ops = &vti6_netdev_ops;
911 dev->header_ops = &ip_tunnel_header_ops;
912 dev->needs_free_netdev = true;
913 dev->priv_destructor = vti6_dev_free;
915 dev->type = ARPHRD_TUNNEL6;
916 dev->min_mtu = IPV4_MIN_MTU;
917 dev->max_mtu = IP_MAX_MTU - sizeof(struct ipv6hdr);
918 dev->flags |= IFF_NOARP;
919 dev->addr_len = sizeof(struct in6_addr);
921 /* This perm addr will be used as interface identifier by IPv6 */
922 dev->addr_assign_type = NET_ADDR_RANDOM;
923 eth_random_addr(dev->perm_addr);
927 * vti6_dev_init_gen - general initializer for all tunnel devices
928 * @dev: virtual device associated with tunnel
930 static inline int vti6_dev_init_gen(struct net_device *dev)
932 struct ip6_tnl *t = netdev_priv(dev);
935 t->net = dev_net(dev);
936 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
939 dev_hold_track(dev, &t->dev_tracker, GFP_KERNEL);
944 * vti6_dev_init - initializer for all non fallback tunnel devices
945 * @dev: virtual device associated with tunnel
947 static int vti6_dev_init(struct net_device *dev)
949 struct ip6_tnl *t = netdev_priv(dev);
950 int err = vti6_dev_init_gen(dev);
954 vti6_link_config(t, true);
959 * vti6_fb_tnl_dev_init - initializer for fallback tunnel device
960 * @dev: fallback device
964 static int __net_init vti6_fb_tnl_dev_init(struct net_device *dev)
966 struct ip6_tnl *t = netdev_priv(dev);
967 struct net *net = dev_net(dev);
968 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
970 t->parms.proto = IPPROTO_IPV6;
972 rcu_assign_pointer(ip6n->tnls_wc[0], t);
976 static int vti6_validate(struct nlattr *tb[], struct nlattr *data[],
977 struct netlink_ext_ack *extack)
982 static void vti6_netlink_parms(struct nlattr *data[],
983 struct __ip6_tnl_parm *parms)
985 memset(parms, 0, sizeof(*parms));
990 if (data[IFLA_VTI_LINK])
991 parms->link = nla_get_u32(data[IFLA_VTI_LINK]);
993 if (data[IFLA_VTI_LOCAL])
994 parms->laddr = nla_get_in6_addr(data[IFLA_VTI_LOCAL]);
996 if (data[IFLA_VTI_REMOTE])
997 parms->raddr = nla_get_in6_addr(data[IFLA_VTI_REMOTE]);
999 if (data[IFLA_VTI_IKEY])
1000 parms->i_key = nla_get_be32(data[IFLA_VTI_IKEY]);
1002 if (data[IFLA_VTI_OKEY])
1003 parms->o_key = nla_get_be32(data[IFLA_VTI_OKEY]);
1005 if (data[IFLA_VTI_FWMARK])
1006 parms->fwmark = nla_get_u32(data[IFLA_VTI_FWMARK]);
1009 static int vti6_newlink(struct net *src_net, struct net_device *dev,
1010 struct nlattr *tb[], struct nlattr *data[],
1011 struct netlink_ext_ack *extack)
1013 struct net *net = dev_net(dev);
1016 nt = netdev_priv(dev);
1017 vti6_netlink_parms(data, &nt->parms);
1019 nt->parms.proto = IPPROTO_IPV6;
1021 if (vti6_locate(net, &nt->parms, 0))
1024 return vti6_tnl_create2(dev);
1027 static void vti6_dellink(struct net_device *dev, struct list_head *head)
1029 struct net *net = dev_net(dev);
1030 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1032 if (dev != ip6n->fb_tnl_dev)
1033 unregister_netdevice_queue(dev, head);
1036 static int vti6_changelink(struct net_device *dev, struct nlattr *tb[],
1037 struct nlattr *data[],
1038 struct netlink_ext_ack *extack)
1041 struct __ip6_tnl_parm p;
1042 struct net *net = dev_net(dev);
1043 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1045 if (dev == ip6n->fb_tnl_dev)
1048 vti6_netlink_parms(data, &p);
1050 t = vti6_locate(net, &p, 0);
1056 t = netdev_priv(dev);
1058 return vti6_update(t, &p, tb && tb[IFLA_MTU]);
1061 static size_t vti6_get_size(const struct net_device *dev)
1066 /* IFLA_VTI_LOCAL */
1067 nla_total_size(sizeof(struct in6_addr)) +
1068 /* IFLA_VTI_REMOTE */
1069 nla_total_size(sizeof(struct in6_addr)) +
1074 /* IFLA_VTI_FWMARK */
1079 static int vti6_fill_info(struct sk_buff *skb, const struct net_device *dev)
1081 struct ip6_tnl *tunnel = netdev_priv(dev);
1082 struct __ip6_tnl_parm *parm = &tunnel->parms;
1084 if (nla_put_u32(skb, IFLA_VTI_LINK, parm->link) ||
1085 nla_put_in6_addr(skb, IFLA_VTI_LOCAL, &parm->laddr) ||
1086 nla_put_in6_addr(skb, IFLA_VTI_REMOTE, &parm->raddr) ||
1087 nla_put_be32(skb, IFLA_VTI_IKEY, parm->i_key) ||
1088 nla_put_be32(skb, IFLA_VTI_OKEY, parm->o_key) ||
1089 nla_put_u32(skb, IFLA_VTI_FWMARK, parm->fwmark))
1090 goto nla_put_failure;
1097 static const struct nla_policy vti6_policy[IFLA_VTI_MAX + 1] = {
1098 [IFLA_VTI_LINK] = { .type = NLA_U32 },
1099 [IFLA_VTI_LOCAL] = { .len = sizeof(struct in6_addr) },
1100 [IFLA_VTI_REMOTE] = { .len = sizeof(struct in6_addr) },
1101 [IFLA_VTI_IKEY] = { .type = NLA_U32 },
1102 [IFLA_VTI_OKEY] = { .type = NLA_U32 },
1103 [IFLA_VTI_FWMARK] = { .type = NLA_U32 },
1106 static struct rtnl_link_ops vti6_link_ops __read_mostly = {
1108 .maxtype = IFLA_VTI_MAX,
1109 .policy = vti6_policy,
1110 .priv_size = sizeof(struct ip6_tnl),
1111 .setup = vti6_dev_setup,
1112 .validate = vti6_validate,
1113 .newlink = vti6_newlink,
1114 .dellink = vti6_dellink,
1115 .changelink = vti6_changelink,
1116 .get_size = vti6_get_size,
1117 .fill_info = vti6_fill_info,
1118 .get_link_net = ip6_tnl_get_link_net,
1121 static void __net_exit vti6_destroy_tunnels(struct vti6_net *ip6n,
1122 struct list_head *list)
1127 for (h = 0; h < IP6_VTI_HASH_SIZE; h++) {
1128 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1130 unregister_netdevice_queue(t->dev, list);
1131 t = rtnl_dereference(t->next);
1135 t = rtnl_dereference(ip6n->tnls_wc[0]);
1137 unregister_netdevice_queue(t->dev, list);
1140 static int __net_init vti6_init_net(struct net *net)
1142 struct vti6_net *ip6n = net_generic(net, vti6_net_id);
1143 struct ip6_tnl *t = NULL;
1146 ip6n->tnls[0] = ip6n->tnls_wc;
1147 ip6n->tnls[1] = ip6n->tnls_r_l;
1149 if (!net_has_fallback_tunnels(net))
1152 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6_vti0",
1153 NET_NAME_UNKNOWN, vti6_dev_setup);
1155 if (!ip6n->fb_tnl_dev)
1157 dev_net_set(ip6n->fb_tnl_dev, net);
1158 ip6n->fb_tnl_dev->rtnl_link_ops = &vti6_link_ops;
1160 err = vti6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1164 err = register_netdev(ip6n->fb_tnl_dev);
1168 t = netdev_priv(ip6n->fb_tnl_dev);
1170 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1174 free_netdev(ip6n->fb_tnl_dev);
1179 static void __net_exit vti6_exit_batch_net(struct list_head *net_list)
1181 struct vti6_net *ip6n;
1186 list_for_each_entry(net, net_list, exit_list) {
1187 ip6n = net_generic(net, vti6_net_id);
1188 vti6_destroy_tunnels(ip6n, &list);
1190 unregister_netdevice_many(&list);
1194 static struct pernet_operations vti6_net_ops = {
1195 .init = vti6_init_net,
1196 .exit_batch = vti6_exit_batch_net,
1198 .size = sizeof(struct vti6_net),
1201 static struct xfrm6_protocol vti_esp6_protocol __read_mostly = {
1202 .handler = vti6_rcv,
1203 .input_handler = vti6_input_proto,
1204 .cb_handler = vti6_rcv_cb,
1205 .err_handler = vti6_err,
1209 static struct xfrm6_protocol vti_ah6_protocol __read_mostly = {
1210 .handler = vti6_rcv,
1211 .input_handler = vti6_input_proto,
1212 .cb_handler = vti6_rcv_cb,
1213 .err_handler = vti6_err,
1217 static struct xfrm6_protocol vti_ipcomp6_protocol __read_mostly = {
1218 .handler = vti6_rcv,
1219 .input_handler = vti6_input_proto,
1220 .cb_handler = vti6_rcv_cb,
1221 .err_handler = vti6_err,
1225 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1226 static int vti6_rcv_tunnel(struct sk_buff *skb)
1228 const xfrm_address_t *saddr;
1231 saddr = (const xfrm_address_t *)&ipv6_hdr(skb)->saddr;
1232 spi = xfrm6_tunnel_spi_lookup(dev_net(skb->dev), saddr);
1234 return vti6_input_proto(skb, IPPROTO_IPV6, spi, 0);
1237 static struct xfrm6_tunnel vti_ipv6_handler __read_mostly = {
1238 .handler = vti6_rcv_tunnel,
1239 .cb_handler = vti6_rcv_cb,
1240 .err_handler = vti6_err,
1244 static struct xfrm6_tunnel vti_ip6ip_handler __read_mostly = {
1245 .handler = vti6_rcv_tunnel,
1246 .cb_handler = vti6_rcv_cb,
1247 .err_handler = vti6_err,
1253 * vti6_tunnel_init - register protocol and reserve needed resources
1255 * Return: 0 on success
1257 static int __init vti6_tunnel_init(void)
1262 msg = "tunnel device";
1263 err = register_pernet_device(&vti6_net_ops);
1265 goto pernet_dev_failed;
1267 msg = "tunnel protocols";
1268 err = xfrm6_protocol_register(&vti_esp6_protocol, IPPROTO_ESP);
1270 goto xfrm_proto_esp_failed;
1271 err = xfrm6_protocol_register(&vti_ah6_protocol, IPPROTO_AH);
1273 goto xfrm_proto_ah_failed;
1274 err = xfrm6_protocol_register(&vti_ipcomp6_protocol, IPPROTO_COMP);
1276 goto xfrm_proto_comp_failed;
1277 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1278 msg = "ipv6 tunnel";
1279 err = xfrm6_tunnel_register(&vti_ipv6_handler, AF_INET6);
1281 goto vti_tunnel_ipv6_failed;
1282 err = xfrm6_tunnel_register(&vti_ip6ip_handler, AF_INET);
1284 goto vti_tunnel_ip6ip_failed;
1287 msg = "netlink interface";
1288 err = rtnl_link_register(&vti6_link_ops);
1290 goto rtnl_link_failed;
1295 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1296 err = xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET);
1297 vti_tunnel_ip6ip_failed:
1298 err = xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6);
1299 vti_tunnel_ipv6_failed:
1301 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1302 xfrm_proto_comp_failed:
1303 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1304 xfrm_proto_ah_failed:
1305 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1306 xfrm_proto_esp_failed:
1307 unregister_pernet_device(&vti6_net_ops);
1309 pr_err("vti6 init: failed to register %s\n", msg);
1314 * vti6_tunnel_cleanup - free resources and unregister protocol
1316 static void __exit vti6_tunnel_cleanup(void)
1318 rtnl_link_unregister(&vti6_link_ops);
1319 #if IS_REACHABLE(CONFIG_INET6_XFRM_TUNNEL)
1320 xfrm6_tunnel_deregister(&vti_ip6ip_handler, AF_INET);
1321 xfrm6_tunnel_deregister(&vti_ipv6_handler, AF_INET6);
1323 xfrm6_protocol_deregister(&vti_ipcomp6_protocol, IPPROTO_COMP);
1324 xfrm6_protocol_deregister(&vti_ah6_protocol, IPPROTO_AH);
1325 xfrm6_protocol_deregister(&vti_esp6_protocol, IPPROTO_ESP);
1326 unregister_pernet_device(&vti6_net_ops);
1329 module_init(vti6_tunnel_init);
1330 module_exit(vti6_tunnel_cleanup);
1331 MODULE_LICENSE("GPL");
1332 MODULE_ALIAS_RTNL_LINK("vti6");
1333 MODULE_ALIAS_NETDEV("ip6_vti0");
1334 MODULE_AUTHOR("Steffen Klassert");
1335 MODULE_DESCRIPTION("IPv6 virtual tunnel interface");