2 * IPv6 tunneling device
3 * Linux INET6 implementation
6 * Ville Nuorvala <vnuorval@tcs.hut.fi>
7 * Yasuyuki Kozakai <kozakai@linux-ipv6.org>
10 * linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
14 * This program is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU General Public License
16 * as published by the Free Software Foundation; either version
17 * 2 of the License, or (at your option) any later version.
21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
23 #include <linux/module.h>
24 #include <linux/capability.h>
25 #include <linux/errno.h>
26 #include <linux/types.h>
27 #include <linux/sockios.h>
28 #include <linux/icmp.h>
32 #include <linux/net.h>
33 #include <linux/in6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/icmpv6.h>
37 #include <linux/init.h>
38 #include <linux/route.h>
39 #include <linux/rtnetlink.h>
40 #include <linux/netfilter_ipv6.h>
41 #include <linux/slab.h>
42 #include <linux/hash.h>
43 #include <linux/etherdevice.h>
45 #include <asm/uaccess.h>
46 #include <linux/atomic.h>
50 #include <net/ip_tunnels.h>
52 #include <net/ip6_route.h>
53 #include <net/addrconf.h>
54 #include <net/ip6_tunnel.h>
56 #include <net/dsfield.h>
57 #include <net/inet_ecn.h>
58 #include <net/net_namespace.h>
59 #include <net/netns/generic.h>
61 MODULE_AUTHOR("Ville Nuorvala");
62 MODULE_DESCRIPTION("IPv6 tunneling device");
63 MODULE_LICENSE("GPL");
64 MODULE_ALIAS_RTNL_LINK("ip6tnl");
65 MODULE_ALIAS_NETDEV("ip6tnl0");
67 #define HASH_SIZE_SHIFT 5
68 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
70 static bool log_ecn_error = true;
71 module_param(log_ecn_error, bool, 0644);
72 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
74 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
76 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
78 return hash_32(hash, HASH_SIZE_SHIFT);
81 static int ip6_tnl_dev_init(struct net_device *dev);
82 static void ip6_tnl_dev_setup(struct net_device *dev);
83 static struct rtnl_link_ops ip6_link_ops __read_mostly;
85 static int ip6_tnl_net_id __read_mostly;
87 /* the IPv6 tunnel fallback device */
88 struct net_device *fb_tnl_dev;
89 /* lists for storing tunnels in use */
90 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
91 struct ip6_tnl __rcu *tnls_wc[1];
92 struct ip6_tnl __rcu **tnls[2];
95 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
97 struct pcpu_sw_netstats tmp, sum = { 0 };
100 for_each_possible_cpu(i) {
102 const struct pcpu_sw_netstats *tstats =
103 per_cpu_ptr(dev->tstats, i);
106 start = u64_stats_fetch_begin_irq(&tstats->syncp);
107 tmp.rx_packets = tstats->rx_packets;
108 tmp.rx_bytes = tstats->rx_bytes;
109 tmp.tx_packets = tstats->tx_packets;
110 tmp.tx_bytes = tstats->tx_bytes;
111 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
113 sum.rx_packets += tmp.rx_packets;
114 sum.rx_bytes += tmp.rx_bytes;
115 sum.tx_packets += tmp.tx_packets;
116 sum.tx_bytes += tmp.tx_bytes;
118 dev->stats.rx_packets = sum.rx_packets;
119 dev->stats.rx_bytes = sum.rx_bytes;
120 dev->stats.tx_packets = sum.tx_packets;
121 dev->stats.tx_bytes = sum.tx_bytes;
126 * Locking : hash tables are protected by RCU and RTNL
129 static void ip6_tnl_per_cpu_dst_set(struct ip6_tnl_dst *idst,
130 struct dst_entry *dst)
132 write_seqlock_bh(&idst->lock);
133 dst_release(rcu_dereference_protected(
135 lockdep_is_held(&idst->lock.lock)));
138 idst->cookie = rt6_get_cookie((struct rt6_info *)dst);
142 rcu_assign_pointer(idst->dst, dst);
143 write_sequnlock_bh(&idst->lock);
146 struct dst_entry *ip6_tnl_dst_get(struct ip6_tnl *t)
148 struct ip6_tnl_dst *idst;
149 struct dst_entry *dst;
153 idst = raw_cpu_ptr(t->dst_cache);
157 seq = read_seqbegin(&idst->lock);
158 dst = rcu_dereference(idst->dst);
159 cookie = idst->cookie;
160 } while (read_seqretry(&idst->lock, seq));
162 if (dst && !atomic_inc_not_zero(&dst->__refcnt))
166 if (dst && dst->obsolete && !dst->ops->check(dst, cookie)) {
167 ip6_tnl_per_cpu_dst_set(idst, NULL);
173 EXPORT_SYMBOL_GPL(ip6_tnl_dst_get);
175 void ip6_tnl_dst_reset(struct ip6_tnl *t)
179 for_each_possible_cpu(i)
180 ip6_tnl_per_cpu_dst_set(per_cpu_ptr(t->dst_cache, i), NULL);
182 EXPORT_SYMBOL_GPL(ip6_tnl_dst_reset);
184 void ip6_tnl_dst_set(struct ip6_tnl *t, struct dst_entry *dst)
186 ip6_tnl_per_cpu_dst_set(raw_cpu_ptr(t->dst_cache), dst);
189 EXPORT_SYMBOL_GPL(ip6_tnl_dst_set);
191 void ip6_tnl_dst_destroy(struct ip6_tnl *t)
196 ip6_tnl_dst_reset(t);
197 free_percpu(t->dst_cache);
199 EXPORT_SYMBOL_GPL(ip6_tnl_dst_destroy);
201 int ip6_tnl_dst_init(struct ip6_tnl *t)
205 t->dst_cache = alloc_percpu(struct ip6_tnl_dst);
209 for_each_possible_cpu(i)
210 seqlock_init(&per_cpu_ptr(t->dst_cache, i)->lock);
214 EXPORT_SYMBOL_GPL(ip6_tnl_dst_init);
217 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
218 * @remote: the address of the tunnel exit-point
219 * @local: the address of the tunnel entry-point
222 * tunnel matching given end-points if found,
223 * else fallback tunnel if its device is up,
227 #define for_each_ip6_tunnel_rcu(start) \
228 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
230 static struct ip6_tnl *
231 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
233 unsigned int hash = HASH(remote, local);
235 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
238 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
239 if (ipv6_addr_equal(local, &t->parms.laddr) &&
240 ipv6_addr_equal(remote, &t->parms.raddr) &&
241 (t->dev->flags & IFF_UP))
245 memset(&any, 0, sizeof(any));
246 hash = HASH(&any, local);
247 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
248 if (ipv6_addr_equal(local, &t->parms.laddr) &&
249 ipv6_addr_any(&t->parms.raddr) &&
250 (t->dev->flags & IFF_UP))
254 hash = HASH(remote, &any);
255 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
256 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
257 ipv6_addr_any(&t->parms.laddr) &&
258 (t->dev->flags & IFF_UP))
262 t = rcu_dereference(ip6n->tnls_wc[0]);
263 if (t && (t->dev->flags & IFF_UP))
270 * ip6_tnl_bucket - get head of list matching given tunnel parameters
271 * @p: parameters containing tunnel end-points
274 * ip6_tnl_bucket() returns the head of the list matching the
275 * &struct in6_addr entries laddr and raddr in @p.
277 * Return: head of IPv6 tunnel list
280 static struct ip6_tnl __rcu **
281 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
283 const struct in6_addr *remote = &p->raddr;
284 const struct in6_addr *local = &p->laddr;
288 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
290 h = HASH(remote, local);
292 return &ip6n->tnls[prio][h];
296 * ip6_tnl_link - add tunnel to hash table
297 * @t: tunnel to be added
301 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
303 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
305 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
306 rcu_assign_pointer(*tp, t);
310 * ip6_tnl_unlink - remove tunnel from hash table
311 * @t: tunnel to be removed
315 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
317 struct ip6_tnl __rcu **tp;
318 struct ip6_tnl *iter;
320 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
321 (iter = rtnl_dereference(*tp)) != NULL;
324 rcu_assign_pointer(*tp, t->next);
330 static void ip6_dev_free(struct net_device *dev)
332 struct ip6_tnl *t = netdev_priv(dev);
334 ip6_tnl_dst_destroy(t);
335 free_percpu(dev->tstats);
339 static int ip6_tnl_create2(struct net_device *dev)
341 struct ip6_tnl *t = netdev_priv(dev);
342 struct net *net = dev_net(dev);
343 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
346 t = netdev_priv(dev);
348 dev->rtnl_link_ops = &ip6_link_ops;
349 err = register_netdevice(dev);
353 strcpy(t->parms.name, dev->name);
356 ip6_tnl_link(ip6n, t);
364 * ip6_tnl_create - create a new tunnel
365 * @p: tunnel parameters
366 * @pt: pointer to new tunnel
369 * Create tunnel matching given parameters.
372 * created tunnel or error pointer
375 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
377 struct net_device *dev;
383 strlcpy(name, p->name, IFNAMSIZ);
385 sprintf(name, "ip6tnl%%d");
387 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
392 dev_net_set(dev, net);
394 t = netdev_priv(dev);
396 t->net = dev_net(dev);
397 err = ip6_tnl_create2(dev);
410 * ip6_tnl_locate - find or create tunnel matching given parameters
411 * @p: tunnel parameters
412 * @create: != 0 if allowed to create new tunnel if no match found
415 * ip6_tnl_locate() first tries to locate an existing tunnel
416 * based on @parms. If this is unsuccessful, but @create is set a new
417 * tunnel device is created and registered for use.
420 * matching tunnel or error pointer
423 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
424 struct __ip6_tnl_parm *p, int create)
426 const struct in6_addr *remote = &p->raddr;
427 const struct in6_addr *local = &p->laddr;
428 struct ip6_tnl __rcu **tp;
430 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
432 for (tp = ip6_tnl_bucket(ip6n, p);
433 (t = rtnl_dereference(*tp)) != NULL;
435 if (ipv6_addr_equal(local, &t->parms.laddr) &&
436 ipv6_addr_equal(remote, &t->parms.raddr)) {
438 return ERR_PTR(-EEXIST);
444 return ERR_PTR(-ENODEV);
445 return ip6_tnl_create(net, p);
449 * ip6_tnl_dev_uninit - tunnel device uninitializer
450 * @dev: the device to be destroyed
453 * ip6_tnl_dev_uninit() removes tunnel from its list
457 ip6_tnl_dev_uninit(struct net_device *dev)
459 struct ip6_tnl *t = netdev_priv(dev);
460 struct net *net = t->net;
461 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
463 if (dev == ip6n->fb_tnl_dev)
464 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
466 ip6_tnl_unlink(ip6n, t);
467 ip6_tnl_dst_reset(t);
472 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
473 * @skb: received socket buffer
476 * 0 if none was found,
477 * else index to encapsulation limit
480 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
482 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
483 unsigned int nhoff = raw - skb->data;
484 unsigned int off = nhoff + sizeof(*ipv6h);
485 u8 next, nexthdr = ipv6h->nexthdr;
487 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
488 struct ipv6_opt_hdr *hdr;
491 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
494 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
495 if (nexthdr == NEXTHDR_FRAGMENT) {
496 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
497 if (frag_hdr->frag_off)
500 } else if (nexthdr == NEXTHDR_AUTH) {
501 optlen = (hdr->hdrlen + 2) << 2;
503 optlen = ipv6_optlen(hdr);
505 /* cache hdr->nexthdr, since pskb_may_pull() might
509 if (nexthdr == NEXTHDR_DEST) {
512 /* Remember : hdr is no longer valid at this point. */
513 if (!pskb_may_pull(skb, off + optlen))
517 struct ipv6_tlv_tnl_enc_lim *tel;
519 /* No more room for encapsulation limit */
520 if (i + sizeof(*tel) > optlen)
523 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
524 /* return index of option if found and valid */
525 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
527 return i + off - nhoff;
528 /* else jump to next option */
530 i += tel->length + 2;
540 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
543 * ip6_tnl_err - tunnel error handler
546 * ip6_tnl_err() should handle errors in the tunnel according
547 * to the specifications in RFC 2473.
551 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
552 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
554 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
557 u8 rel_type = ICMPV6_DEST_UNREACH;
558 u8 rel_code = ICMPV6_ADDR_UNREACH;
564 /* If the packet doesn't contain the original IPv6 header we are
565 in trouble since we might need the source address for further
566 processing of the error. */
569 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
573 tproto = ACCESS_ONCE(t->parms.proto);
574 if (tproto != ipproto && tproto != 0)
581 struct ipv6_tlv_tnl_enc_lim *tel;
583 case ICMPV6_DEST_UNREACH:
584 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
588 case ICMPV6_TIME_EXCEED:
589 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
590 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
595 case ICMPV6_PARAMPROB:
597 if ((*code) == ICMPV6_HDR_FIELD)
598 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
600 if (teli && teli == *info - 2) {
601 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
602 if (tel->encap_limit == 0) {
603 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
608 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
612 case ICMPV6_PKT_TOOBIG:
613 mtu = *info - offset;
614 if (mtu < IPV6_MIN_MTU)
618 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
620 rel_type = ICMPV6_PKT_TOOBIG;
639 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
640 u8 type, u8 code, int offset, __be32 info)
645 __u32 rel_info = ntohl(info);
647 struct sk_buff *skb2;
648 const struct iphdr *eiph;
652 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
653 &rel_msg, &rel_info, offset);
661 case ICMPV6_DEST_UNREACH:
662 if (rel_code != ICMPV6_ADDR_UNREACH)
664 rel_type = ICMP_DEST_UNREACH;
665 rel_code = ICMP_HOST_UNREACH;
667 case ICMPV6_PKT_TOOBIG:
670 rel_type = ICMP_DEST_UNREACH;
671 rel_code = ICMP_FRAG_NEEDED;
674 rel_type = ICMP_REDIRECT;
675 rel_code = ICMP_REDIR_HOST;
680 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
683 skb2 = skb_clone(skb, GFP_ATOMIC);
689 skb_pull(skb2, offset);
690 skb_reset_network_header(skb2);
693 /* Try to guess incoming interface */
694 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
697 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
701 skb2->dev = rt->dst.dev;
703 /* route "incoming" packet */
704 if (rt->rt_flags & RTCF_LOCAL) {
707 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
708 eiph->daddr, eiph->saddr,
711 RT_TOS(eiph->tos), 0);
713 rt->dst.dev->type != ARPHRD_TUNNEL) {
718 skb_dst_set(skb2, &rt->dst);
721 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
723 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
727 /* change mtu on this route */
728 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
729 if (rel_info > dst_mtu(skb_dst(skb2)))
732 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
734 if (rel_type == ICMP_REDIRECT)
735 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
737 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
745 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
746 u8 type, u8 code, int offset, __be32 info)
751 __u32 rel_info = ntohl(info);
754 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
755 &rel_msg, &rel_info, offset);
759 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
761 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
767 skb_pull(skb2, offset);
768 skb_reset_network_header(skb2);
770 /* Try to guess incoming interface */
771 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
774 if (rt && rt->dst.dev)
775 skb2->dev = rt->dst.dev;
777 icmpv6_send(skb2, rel_type, rel_code, rel_info);
787 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
788 const struct ipv6hdr *ipv6h,
791 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
793 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
794 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
796 return IP6_ECN_decapsulate(ipv6h, skb);
799 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
800 const struct ipv6hdr *ipv6h,
803 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
804 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
806 return IP6_ECN_decapsulate(ipv6h, skb);
809 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
810 const struct in6_addr *laddr,
811 const struct in6_addr *raddr)
813 struct __ip6_tnl_parm *p = &t->parms;
814 int ltype = ipv6_addr_type(laddr);
815 int rtype = ipv6_addr_type(raddr);
818 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
819 flags = IP6_TNL_F_CAP_PER_PACKET;
820 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
821 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
822 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
823 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
824 if (ltype&IPV6_ADDR_UNICAST)
825 flags |= IP6_TNL_F_CAP_XMIT;
826 if (rtype&IPV6_ADDR_UNICAST)
827 flags |= IP6_TNL_F_CAP_RCV;
831 EXPORT_SYMBOL(ip6_tnl_get_cap);
833 /* called with rcu_read_lock() */
834 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
835 const struct in6_addr *laddr,
836 const struct in6_addr *raddr)
838 struct __ip6_tnl_parm *p = &t->parms;
840 struct net *net = t->net;
842 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
843 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
844 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
845 struct net_device *ldev = NULL;
848 ldev = dev_get_by_index_rcu(net, p->link);
850 if ((ipv6_addr_is_multicast(laddr) ||
851 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
852 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
857 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
860 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
861 * @skb: received socket buffer
862 * @protocol: ethernet protocol ID
863 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
868 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
870 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
871 const struct ipv6hdr *ipv6h,
872 struct sk_buff *skb))
875 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
880 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
882 struct pcpu_sw_netstats *tstats;
884 tproto = ACCESS_ONCE(t->parms.proto);
885 if (tproto != ipproto && tproto != 0) {
890 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
895 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
896 t->dev->stats.rx_dropped++;
900 skb->mac_header = skb->network_header;
901 skb_reset_network_header(skb);
902 skb->protocol = htons(protocol);
903 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
905 __skb_tunnel_rx(skb, t->dev, t->net);
907 err = dscp_ecn_decapsulate(t, ipv6h, skb);
910 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
912 ipv6_get_dsfield(ipv6h));
914 ++t->dev->stats.rx_frame_errors;
915 ++t->dev->stats.rx_errors;
921 tstats = this_cpu_ptr(t->dev->tstats);
922 u64_stats_update_begin(&tstats->syncp);
923 tstats->rx_packets++;
924 tstats->rx_bytes += skb->len;
925 u64_stats_update_end(&tstats->syncp);
940 static int ip4ip6_rcv(struct sk_buff *skb)
942 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
943 ip4ip6_dscp_ecn_decapsulate);
946 static int ip6ip6_rcv(struct sk_buff *skb)
948 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
949 ip6ip6_dscp_ecn_decapsulate);
952 struct ipv6_tel_txoption {
953 struct ipv6_txoptions ops;
957 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
959 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
961 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
963 opt->dst_opt[4] = encap_limit;
964 opt->dst_opt[5] = IPV6_TLV_PADN;
967 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
968 opt->ops.opt_nflen = 8;
972 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
973 * @t: the outgoing tunnel device
974 * @hdr: IPv6 header from the incoming packet
977 * Avoid trivial tunneling loop by checking that tunnel exit-point
978 * doesn't match source of incoming packet.
986 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
988 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
991 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
992 const struct in6_addr *laddr,
993 const struct in6_addr *raddr)
995 struct __ip6_tnl_parm *p = &t->parms;
997 struct net *net = t->net;
999 if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1000 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1001 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1002 struct net_device *ldev = NULL;
1006 ldev = dev_get_by_index_rcu(net, p->link);
1008 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
1009 pr_warn("%s xmit: Local address not yet configured!\n",
1011 else if (!ipv6_addr_is_multicast(raddr) &&
1012 unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
1013 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1021 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1024 * ip6_tnl_xmit2 - encapsulate packet and send
1025 * @skb: the outgoing socket buffer
1026 * @dev: the outgoing tunnel device
1027 * @dsfield: dscp code for outer header
1028 * @fl: flow of tunneled packet
1029 * @encap_limit: encapsulation limit
1030 * @pmtu: Path MTU is stored if packet is too big
1033 * Build new header and do some sanity checks on the packet before sending
1039 * %-EMSGSIZE message too big. return mtu in this case.
1042 static int ip6_tnl_xmit2(struct sk_buff *skb,
1043 struct net_device *dev,
1049 struct ip6_tnl *t = netdev_priv(dev);
1050 struct net *net = t->net;
1051 struct net_device_stats *stats = &t->dev->stats;
1052 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1053 struct ipv6_tel_txoption opt;
1054 struct dst_entry *dst = NULL, *ndst = NULL;
1055 struct net_device *tdev;
1056 bool use_cache = false;
1058 unsigned int max_headroom = sizeof(struct ipv6hdr);
1063 if (ipv6_addr_any(&t->parms.raddr)) {
1064 struct in6_addr *addr6;
1065 struct neighbour *neigh;
1069 goto tx_err_link_failure;
1071 neigh = dst_neigh_lookup(skb_dst(skb),
1072 &ipv6_hdr(skb)->daddr);
1074 goto tx_err_link_failure;
1076 addr6 = (struct in6_addr *)&neigh->primary_key;
1077 addr_type = ipv6_addr_type(addr6);
1079 if (addr_type == IPV6_ADDR_ANY)
1080 addr6 = &ipv6_hdr(skb)->daddr;
1082 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1083 neigh_release(neigh);
1084 } else if (!(t->parms.flags &
1085 (IP6_TNL_F_USE_ORIG_TCLASS | IP6_TNL_F_USE_ORIG_FWMARK))) {
1086 /* enable the cache only only if the routing decision does
1087 * not depend on the current inner header value
1093 dst = ip6_tnl_dst_get(t);
1095 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1096 goto tx_err_link_failure;
1099 dst = ip6_route_output(net, NULL, fl6);
1102 goto tx_err_link_failure;
1103 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1107 goto tx_err_link_failure;
1115 stats->collisions++;
1116 net_warn_ratelimited("%s: Local routing loop detected!\n",
1118 goto tx_err_dst_release;
1120 mtu = dst_mtu(dst) - sizeof(*ipv6h);
1121 if (encap_limit >= 0) {
1125 if (mtu < IPV6_MIN_MTU)
1128 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1129 if (skb->len > mtu) {
1132 goto tx_err_dst_release;
1135 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1138 * Okay, now see if we can stuff it in the buffer as-is.
1140 max_headroom += LL_RESERVED_SPACE(tdev);
1142 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1143 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1144 struct sk_buff *new_skb;
1146 new_skb = skb_realloc_headroom(skb, max_headroom);
1148 goto tx_err_dst_release;
1151 skb_set_owner_w(new_skb, skb->sk);
1156 if (use_cache && ndst)
1157 ip6_tnl_dst_set(t, ndst);
1158 skb_dst_set(skb, dst);
1160 skb->transport_header = skb->network_header;
1162 proto = fl6->flowi6_proto;
1163 if (encap_limit >= 0) {
1164 init_tel_txopt(&opt, encap_limit);
1165 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1168 if (likely(!skb->encapsulation)) {
1169 skb_reset_inner_headers(skb);
1170 skb->encapsulation = 1;
1173 skb_push(skb, sizeof(struct ipv6hdr));
1174 skb_reset_network_header(skb);
1175 ipv6h = ipv6_hdr(skb);
1176 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1177 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1178 ipv6h->hop_limit = t->parms.hop_limit;
1179 ipv6h->nexthdr = proto;
1180 ipv6h->saddr = fl6->saddr;
1181 ipv6h->daddr = fl6->daddr;
1182 ip6tunnel_xmit(NULL, skb, dev);
1184 tx_err_link_failure:
1185 stats->tx_carrier_errors++;
1186 dst_link_failure(skb);
1193 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1195 struct ip6_tnl *t = netdev_priv(dev);
1196 const struct iphdr *iph = ip_hdr(skb);
1197 int encap_limit = -1;
1204 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1206 tproto = ACCESS_ONCE(t->parms.proto);
1207 if (tproto != IPPROTO_IPIP && tproto != 0)
1210 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1211 encap_limit = t->parms.encap_limit;
1213 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1214 fl6.flowi6_proto = IPPROTO_IPIP;
1216 dsfield = ipv4_get_dsfield(iph);
1218 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1219 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1221 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1222 fl6.flowi6_mark = skb->mark;
1224 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1226 /* XXX: send ICMP error even if DF is not set. */
1227 if (err == -EMSGSIZE)
1228 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1237 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1239 struct ip6_tnl *t = netdev_priv(dev);
1240 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1241 int encap_limit = -1;
1249 tproto = ACCESS_ONCE(t->parms.proto);
1250 if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1251 ip6_tnl_addr_conflict(t, ipv6h))
1254 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1256 struct ipv6_tlv_tnl_enc_lim *tel;
1257 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1258 if (tel->encap_limit == 0) {
1259 icmpv6_send(skb, ICMPV6_PARAMPROB,
1260 ICMPV6_HDR_FIELD, offset + 2);
1263 encap_limit = tel->encap_limit - 1;
1264 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1265 encap_limit = t->parms.encap_limit;
1267 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1268 fl6.flowi6_proto = IPPROTO_IPV6;
1270 dsfield = ipv6_get_dsfield(ipv6h);
1271 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1272 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1273 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1274 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1275 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1276 fl6.flowi6_mark = skb->mark;
1278 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1280 if (err == -EMSGSIZE)
1281 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1289 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1291 struct ip6_tnl *t = netdev_priv(dev);
1292 struct net_device_stats *stats = &t->dev->stats;
1295 switch (skb->protocol) {
1296 case htons(ETH_P_IP):
1297 ret = ip4ip6_tnl_xmit(skb, dev);
1299 case htons(ETH_P_IPV6):
1300 ret = ip6ip6_tnl_xmit(skb, dev);
1309 return NETDEV_TX_OK;
1313 stats->tx_dropped++;
1315 return NETDEV_TX_OK;
1318 static void ip6_tnl_link_config(struct ip6_tnl *t)
1320 struct net_device *dev = t->dev;
1321 struct __ip6_tnl_parm *p = &t->parms;
1322 struct flowi6 *fl6 = &t->fl.u.ip6;
1324 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1325 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1327 /* Set up flowi template */
1328 fl6->saddr = p->laddr;
1329 fl6->daddr = p->raddr;
1330 fl6->flowi6_oif = p->link;
1333 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1334 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1335 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1336 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1338 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1339 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1341 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1342 dev->flags |= IFF_POINTOPOINT;
1344 dev->flags &= ~IFF_POINTOPOINT;
1346 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1347 int strict = (ipv6_addr_type(&p->raddr) &
1348 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1350 struct rt6_info *rt = rt6_lookup(t->net,
1351 &p->raddr, &p->laddr,
1358 dev->hard_header_len = rt->dst.dev->hard_header_len +
1359 sizeof(struct ipv6hdr);
1361 dev->mtu = rt->dst.dev->mtu - sizeof(struct ipv6hdr);
1362 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1365 if (dev->mtu < IPV6_MIN_MTU)
1366 dev->mtu = IPV6_MIN_MTU;
1373 * ip6_tnl_change - update the tunnel parameters
1374 * @t: tunnel to be changed
1375 * @p: tunnel configuration parameters
1378 * ip6_tnl_change() updates the tunnel parameters
1382 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1384 t->parms.laddr = p->laddr;
1385 t->parms.raddr = p->raddr;
1386 t->parms.flags = p->flags;
1387 t->parms.hop_limit = p->hop_limit;
1388 t->parms.encap_limit = p->encap_limit;
1389 t->parms.flowinfo = p->flowinfo;
1390 t->parms.link = p->link;
1391 t->parms.proto = p->proto;
1392 ip6_tnl_dst_reset(t);
1393 ip6_tnl_link_config(t);
1397 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1399 struct net *net = t->net;
1400 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1403 ip6_tnl_unlink(ip6n, t);
1405 err = ip6_tnl_change(t, p);
1406 ip6_tnl_link(ip6n, t);
1407 netdev_state_change(t->dev);
1411 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1413 /* for default tnl0 device allow to change only the proto */
1414 t->parms.proto = p->proto;
1415 netdev_state_change(t->dev);
1420 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1422 p->laddr = u->laddr;
1423 p->raddr = u->raddr;
1424 p->flags = u->flags;
1425 p->hop_limit = u->hop_limit;
1426 p->encap_limit = u->encap_limit;
1427 p->flowinfo = u->flowinfo;
1429 p->proto = u->proto;
1430 memcpy(p->name, u->name, sizeof(u->name));
1434 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1436 u->laddr = p->laddr;
1437 u->raddr = p->raddr;
1438 u->flags = p->flags;
1439 u->hop_limit = p->hop_limit;
1440 u->encap_limit = p->encap_limit;
1441 u->flowinfo = p->flowinfo;
1443 u->proto = p->proto;
1444 memcpy(u->name, p->name, sizeof(u->name));
1448 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1449 * @dev: virtual device associated with tunnel
1450 * @ifr: parameters passed from userspace
1451 * @cmd: command to be performed
1454 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1457 * The possible commands are the following:
1458 * %SIOCGETTUNNEL: get tunnel parameters for device
1459 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1460 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1461 * %SIOCDELTUNNEL: delete tunnel
1463 * The fallback device "ip6tnl0", created during module
1464 * initialization, can be used for creating other tunnel devices.
1468 * %-EFAULT if unable to copy data to or from userspace,
1469 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1470 * %-EINVAL if passed tunnel parameters are invalid,
1471 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1472 * %-ENODEV if attempting to change or delete a nonexisting device
1476 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1479 struct ip6_tnl_parm p;
1480 struct __ip6_tnl_parm p1;
1481 struct ip6_tnl *t = netdev_priv(dev);
1482 struct net *net = t->net;
1483 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1487 if (dev == ip6n->fb_tnl_dev) {
1488 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1492 ip6_tnl_parm_from_user(&p1, &p);
1493 t = ip6_tnl_locate(net, &p1, 0);
1495 t = netdev_priv(dev);
1497 memset(&p, 0, sizeof(p));
1499 ip6_tnl_parm_to_user(&p, &t->parms);
1500 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1507 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1510 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1513 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1516 ip6_tnl_parm_from_user(&p1, &p);
1517 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1518 if (cmd == SIOCCHGTUNNEL) {
1520 if (t->dev != dev) {
1525 t = netdev_priv(dev);
1526 if (dev == ip6n->fb_tnl_dev)
1527 err = ip6_tnl0_update(t, &p1);
1529 err = ip6_tnl_update(t, &p1);
1533 ip6_tnl_parm_to_user(&p, &t->parms);
1534 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1543 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1546 if (dev == ip6n->fb_tnl_dev) {
1548 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1551 ip6_tnl_parm_from_user(&p1, &p);
1552 t = ip6_tnl_locate(net, &p1, 0);
1556 if (t->dev == ip6n->fb_tnl_dev)
1561 unregister_netdevice(dev);
1570 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1571 * @dev: virtual device associated with tunnel
1572 * @new_mtu: the new mtu
1576 * %-EINVAL if mtu too small
1580 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1582 struct ip6_tnl *tnl = netdev_priv(dev);
1584 if (tnl->parms.proto == IPPROTO_IPIP) {
1588 if (new_mtu < IPV6_MIN_MTU)
1591 if (new_mtu > 0xFFF8 - dev->hard_header_len)
1597 int ip6_tnl_get_iflink(const struct net_device *dev)
1599 struct ip6_tnl *t = netdev_priv(dev);
1601 return t->parms.link;
1603 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1605 static const struct net_device_ops ip6_tnl_netdev_ops = {
1606 .ndo_init = ip6_tnl_dev_init,
1607 .ndo_uninit = ip6_tnl_dev_uninit,
1608 .ndo_start_xmit = ip6_tnl_xmit,
1609 .ndo_do_ioctl = ip6_tnl_ioctl,
1610 .ndo_change_mtu = ip6_tnl_change_mtu,
1611 .ndo_get_stats = ip6_get_stats,
1612 .ndo_get_iflink = ip6_tnl_get_iflink,
1617 * ip6_tnl_dev_setup - setup virtual tunnel device
1618 * @dev: virtual device associated with tunnel
1621 * Initialize function pointers and device parameters
1624 static void ip6_tnl_dev_setup(struct net_device *dev)
1628 dev->netdev_ops = &ip6_tnl_netdev_ops;
1629 dev->destructor = ip6_dev_free;
1631 dev->type = ARPHRD_TUNNEL6;
1632 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
1633 dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr);
1634 t = netdev_priv(dev);
1635 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1637 dev->flags |= IFF_NOARP;
1638 dev->addr_len = sizeof(struct in6_addr);
1639 netif_keep_dst(dev);
1640 /* This perm addr will be used as interface identifier by IPv6 */
1641 dev->addr_assign_type = NET_ADDR_RANDOM;
1642 eth_random_addr(dev->perm_addr);
1647 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1648 * @dev: virtual device associated with tunnel
1652 ip6_tnl_dev_init_gen(struct net_device *dev)
1654 struct ip6_tnl *t = netdev_priv(dev);
1658 t->net = dev_net(dev);
1659 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1663 ret = ip6_tnl_dst_init(t);
1665 free_percpu(dev->tstats);
1674 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1675 * @dev: virtual device associated with tunnel
1678 static int ip6_tnl_dev_init(struct net_device *dev)
1680 struct ip6_tnl *t = netdev_priv(dev);
1681 int err = ip6_tnl_dev_init_gen(dev);
1685 ip6_tnl_link_config(t);
1690 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1691 * @dev: fallback device
1696 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1698 struct ip6_tnl *t = netdev_priv(dev);
1699 struct net *net = dev_net(dev);
1700 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1702 t->parms.proto = IPPROTO_IPV6;
1705 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1709 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1713 if (!data || !data[IFLA_IPTUN_PROTO])
1716 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1717 if (proto != IPPROTO_IPV6 &&
1718 proto != IPPROTO_IPIP &&
1725 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1726 struct __ip6_tnl_parm *parms)
1728 memset(parms, 0, sizeof(*parms));
1733 if (data[IFLA_IPTUN_LINK])
1734 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1736 if (data[IFLA_IPTUN_LOCAL])
1737 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1739 if (data[IFLA_IPTUN_REMOTE])
1740 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1742 if (data[IFLA_IPTUN_TTL])
1743 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1745 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1746 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1748 if (data[IFLA_IPTUN_FLOWINFO])
1749 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1751 if (data[IFLA_IPTUN_FLAGS])
1752 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1754 if (data[IFLA_IPTUN_PROTO])
1755 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1758 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1759 struct nlattr *tb[], struct nlattr *data[])
1761 struct net *net = dev_net(dev);
1762 struct ip6_tnl *nt, *t;
1764 nt = netdev_priv(dev);
1765 ip6_tnl_netlink_parms(data, &nt->parms);
1767 t = ip6_tnl_locate(net, &nt->parms, 0);
1771 return ip6_tnl_create2(dev);
1774 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1775 struct nlattr *data[])
1777 struct ip6_tnl *t = netdev_priv(dev);
1778 struct __ip6_tnl_parm p;
1779 struct net *net = t->net;
1780 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1782 if (dev == ip6n->fb_tnl_dev)
1785 ip6_tnl_netlink_parms(data, &p);
1787 t = ip6_tnl_locate(net, &p, 0);
1792 t = netdev_priv(dev);
1794 return ip6_tnl_update(t, &p);
1797 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1799 struct net *net = dev_net(dev);
1800 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1802 if (dev != ip6n->fb_tnl_dev)
1803 unregister_netdevice_queue(dev, head);
1806 static size_t ip6_tnl_get_size(const struct net_device *dev)
1809 /* IFLA_IPTUN_LINK */
1811 /* IFLA_IPTUN_LOCAL */
1812 nla_total_size(sizeof(struct in6_addr)) +
1813 /* IFLA_IPTUN_REMOTE */
1814 nla_total_size(sizeof(struct in6_addr)) +
1815 /* IFLA_IPTUN_TTL */
1817 /* IFLA_IPTUN_ENCAP_LIMIT */
1819 /* IFLA_IPTUN_FLOWINFO */
1821 /* IFLA_IPTUN_FLAGS */
1823 /* IFLA_IPTUN_PROTO */
1828 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1830 struct ip6_tnl *tunnel = netdev_priv(dev);
1831 struct __ip6_tnl_parm *parm = &tunnel->parms;
1833 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1834 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
1835 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
1836 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1837 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1838 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1839 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1840 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1841 goto nla_put_failure;
1848 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
1850 struct ip6_tnl *tunnel = netdev_priv(dev);
1854 EXPORT_SYMBOL(ip6_tnl_get_link_net);
1856 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1857 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1858 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
1859 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
1860 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1861 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
1862 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
1863 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
1864 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1867 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1869 .maxtype = IFLA_IPTUN_MAX,
1870 .policy = ip6_tnl_policy,
1871 .priv_size = sizeof(struct ip6_tnl),
1872 .setup = ip6_tnl_dev_setup,
1873 .validate = ip6_tnl_validate,
1874 .newlink = ip6_tnl_newlink,
1875 .changelink = ip6_tnl_changelink,
1876 .dellink = ip6_tnl_dellink,
1877 .get_size = ip6_tnl_get_size,
1878 .fill_info = ip6_tnl_fill_info,
1879 .get_link_net = ip6_tnl_get_link_net,
1882 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1883 .handler = ip4ip6_rcv,
1884 .err_handler = ip4ip6_err,
1888 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1889 .handler = ip6ip6_rcv,
1890 .err_handler = ip6ip6_err,
1894 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
1896 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1897 struct net_device *dev, *aux;
1902 for_each_netdev_safe(net, dev, aux)
1903 if (dev->rtnl_link_ops == &ip6_link_ops)
1904 unregister_netdevice_queue(dev, &list);
1906 for (h = 0; h < HASH_SIZE; h++) {
1907 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1909 /* If dev is in the same netns, it has already
1910 * been added to the list by the previous loop.
1912 if (!net_eq(dev_net(t->dev), net))
1913 unregister_netdevice_queue(t->dev, &list);
1914 t = rtnl_dereference(t->next);
1918 unregister_netdevice_many(&list);
1921 static int __net_init ip6_tnl_init_net(struct net *net)
1923 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1924 struct ip6_tnl *t = NULL;
1927 ip6n->tnls[0] = ip6n->tnls_wc;
1928 ip6n->tnls[1] = ip6n->tnls_r_l;
1931 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1932 NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
1934 if (!ip6n->fb_tnl_dev)
1936 dev_net_set(ip6n->fb_tnl_dev, net);
1937 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1938 /* FB netdevice is special: we have one, and only one per netns.
1939 * Allowing to move it to another netns is clearly unsafe.
1941 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1943 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1947 err = register_netdev(ip6n->fb_tnl_dev);
1951 t = netdev_priv(ip6n->fb_tnl_dev);
1953 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1957 ip6_dev_free(ip6n->fb_tnl_dev);
1962 static void __net_exit ip6_tnl_exit_net(struct net *net)
1965 ip6_tnl_destroy_tunnels(net);
1969 static struct pernet_operations ip6_tnl_net_ops = {
1970 .init = ip6_tnl_init_net,
1971 .exit = ip6_tnl_exit_net,
1972 .id = &ip6_tnl_net_id,
1973 .size = sizeof(struct ip6_tnl_net),
1977 * ip6_tunnel_init - register protocol and reserve needed resources
1979 * Return: 0 on success
1982 static int __init ip6_tunnel_init(void)
1986 err = register_pernet_device(&ip6_tnl_net_ops);
1990 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1992 pr_err("%s: can't register ip4ip6\n", __func__);
1996 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1998 pr_err("%s: can't register ip6ip6\n", __func__);
2001 err = rtnl_link_register(&ip6_link_ops);
2003 goto rtnl_link_failed;
2008 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2010 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2012 unregister_pernet_device(&ip6_tnl_net_ops);
2018 * ip6_tunnel_cleanup - free resources and unregister protocol
2021 static void __exit ip6_tunnel_cleanup(void)
2023 rtnl_link_unregister(&ip6_link_ops);
2024 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2025 pr_info("%s: can't deregister ip4ip6\n", __func__);
2027 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2028 pr_info("%s: can't deregister ip6ip6\n", __func__);
2030 unregister_pernet_device(&ip6_tnl_net_ops);
2033 module_init(ip6_tunnel_init);
2034 module_exit(ip6_tunnel_cleanup);