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 (t->dev->flags & IFF_UP))
253 hash = HASH(remote, &any);
254 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
255 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
256 (t->dev->flags & IFF_UP))
260 t = rcu_dereference(ip6n->tnls_wc[0]);
261 if (t && (t->dev->flags & IFF_UP))
268 * ip6_tnl_bucket - get head of list matching given tunnel parameters
269 * @p: parameters containing tunnel end-points
272 * ip6_tnl_bucket() returns the head of the list matching the
273 * &struct in6_addr entries laddr and raddr in @p.
275 * Return: head of IPv6 tunnel list
278 static struct ip6_tnl __rcu **
279 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
281 const struct in6_addr *remote = &p->raddr;
282 const struct in6_addr *local = &p->laddr;
286 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
288 h = HASH(remote, local);
290 return &ip6n->tnls[prio][h];
294 * ip6_tnl_link - add tunnel to hash table
295 * @t: tunnel to be added
299 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
301 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
303 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
304 rcu_assign_pointer(*tp, t);
308 * ip6_tnl_unlink - remove tunnel from hash table
309 * @t: tunnel to be removed
313 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
315 struct ip6_tnl __rcu **tp;
316 struct ip6_tnl *iter;
318 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
319 (iter = rtnl_dereference(*tp)) != NULL;
322 rcu_assign_pointer(*tp, t->next);
328 static void ip6_dev_free(struct net_device *dev)
330 struct ip6_tnl *t = netdev_priv(dev);
332 ip6_tnl_dst_destroy(t);
333 free_percpu(dev->tstats);
337 static int ip6_tnl_create2(struct net_device *dev)
339 struct ip6_tnl *t = netdev_priv(dev);
340 struct net *net = dev_net(dev);
341 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
344 t = netdev_priv(dev);
346 dev->rtnl_link_ops = &ip6_link_ops;
347 err = register_netdevice(dev);
351 strcpy(t->parms.name, dev->name);
354 ip6_tnl_link(ip6n, t);
362 * ip6_tnl_create - create a new tunnel
363 * @p: tunnel parameters
364 * @pt: pointer to new tunnel
367 * Create tunnel matching given parameters.
370 * created tunnel or error pointer
373 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
375 struct net_device *dev;
381 strlcpy(name, p->name, IFNAMSIZ);
383 sprintf(name, "ip6tnl%%d");
385 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
390 dev_net_set(dev, net);
392 t = netdev_priv(dev);
394 t->net = dev_net(dev);
395 err = ip6_tnl_create2(dev);
408 * ip6_tnl_locate - find or create tunnel matching given parameters
409 * @p: tunnel parameters
410 * @create: != 0 if allowed to create new tunnel if no match found
413 * ip6_tnl_locate() first tries to locate an existing tunnel
414 * based on @parms. If this is unsuccessful, but @create is set a new
415 * tunnel device is created and registered for use.
418 * matching tunnel or error pointer
421 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
422 struct __ip6_tnl_parm *p, int create)
424 const struct in6_addr *remote = &p->raddr;
425 const struct in6_addr *local = &p->laddr;
426 struct ip6_tnl __rcu **tp;
428 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
430 for (tp = ip6_tnl_bucket(ip6n, p);
431 (t = rtnl_dereference(*tp)) != NULL;
433 if (ipv6_addr_equal(local, &t->parms.laddr) &&
434 ipv6_addr_equal(remote, &t->parms.raddr)) {
436 return ERR_PTR(-EEXIST);
442 return ERR_PTR(-ENODEV);
443 return ip6_tnl_create(net, p);
447 * ip6_tnl_dev_uninit - tunnel device uninitializer
448 * @dev: the device to be destroyed
451 * ip6_tnl_dev_uninit() removes tunnel from its list
455 ip6_tnl_dev_uninit(struct net_device *dev)
457 struct ip6_tnl *t = netdev_priv(dev);
458 struct net *net = t->net;
459 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
461 if (dev == ip6n->fb_tnl_dev)
462 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
464 ip6_tnl_unlink(ip6n, t);
465 ip6_tnl_dst_reset(t);
470 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
471 * @skb: received socket buffer
474 * 0 if none was found,
475 * else index to encapsulation limit
478 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
480 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
481 __u8 nexthdr = ipv6h->nexthdr;
482 __u16 off = sizeof(*ipv6h);
484 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
486 struct ipv6_opt_hdr *hdr;
487 if (raw + off + sizeof(*hdr) > skb->data &&
488 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
491 hdr = (struct ipv6_opt_hdr *) (raw + off);
492 if (nexthdr == NEXTHDR_FRAGMENT) {
493 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
494 if (frag_hdr->frag_off)
497 } else if (nexthdr == NEXTHDR_AUTH) {
498 optlen = (hdr->hdrlen + 2) << 2;
500 optlen = ipv6_optlen(hdr);
502 if (nexthdr == NEXTHDR_DEST) {
505 struct ipv6_tlv_tnl_enc_lim *tel;
507 /* No more room for encapsulation limit */
508 if (i + sizeof (*tel) > off + optlen)
511 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
512 /* return index of option if found and valid */
513 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
516 /* else jump to next option */
518 i += tel->length + 2;
523 nexthdr = hdr->nexthdr;
528 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
531 * ip6_tnl_err - tunnel error handler
534 * ip6_tnl_err() should handle errors in the tunnel according
535 * to the specifications in RFC 2473.
539 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
540 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
542 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
545 u8 rel_type = ICMPV6_DEST_UNREACH;
546 u8 rel_code = ICMPV6_ADDR_UNREACH;
552 /* If the packet doesn't contain the original IPv6 header we are
553 in trouble since we might need the source address for further
554 processing of the error. */
557 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
561 tproto = ACCESS_ONCE(t->parms.proto);
562 if (tproto != ipproto && tproto != 0)
569 struct ipv6_tlv_tnl_enc_lim *tel;
571 case ICMPV6_DEST_UNREACH:
572 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
576 case ICMPV6_TIME_EXCEED:
577 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
578 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
583 case ICMPV6_PARAMPROB:
585 if ((*code) == ICMPV6_HDR_FIELD)
586 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
588 if (teli && teli == *info - 2) {
589 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
590 if (tel->encap_limit == 0) {
591 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
596 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
600 case ICMPV6_PKT_TOOBIG:
601 mtu = *info - offset;
602 if (mtu < IPV6_MIN_MTU)
606 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
608 rel_type = ICMPV6_PKT_TOOBIG;
627 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
628 u8 type, u8 code, int offset, __be32 info)
633 __u32 rel_info = ntohl(info);
635 struct sk_buff *skb2;
636 const struct iphdr *eiph;
640 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
641 &rel_msg, &rel_info, offset);
649 case ICMPV6_DEST_UNREACH:
650 if (rel_code != ICMPV6_ADDR_UNREACH)
652 rel_type = ICMP_DEST_UNREACH;
653 rel_code = ICMP_HOST_UNREACH;
655 case ICMPV6_PKT_TOOBIG:
658 rel_type = ICMP_DEST_UNREACH;
659 rel_code = ICMP_FRAG_NEEDED;
662 rel_type = ICMP_REDIRECT;
663 rel_code = ICMP_REDIR_HOST;
668 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
671 skb2 = skb_clone(skb, GFP_ATOMIC);
677 skb_pull(skb2, offset);
678 skb_reset_network_header(skb2);
681 /* Try to guess incoming interface */
682 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
685 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
689 skb2->dev = rt->dst.dev;
691 /* route "incoming" packet */
692 if (rt->rt_flags & RTCF_LOCAL) {
695 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
696 eiph->daddr, eiph->saddr,
699 RT_TOS(eiph->tos), 0);
701 rt->dst.dev->type != ARPHRD_TUNNEL) {
706 skb_dst_set(skb2, &rt->dst);
709 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
711 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
715 /* change mtu on this route */
716 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
717 if (rel_info > dst_mtu(skb_dst(skb2)))
720 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
722 if (rel_type == ICMP_REDIRECT)
723 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
725 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
733 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
734 u8 type, u8 code, int offset, __be32 info)
739 __u32 rel_info = ntohl(info);
742 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
743 &rel_msg, &rel_info, offset);
747 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
749 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
755 skb_pull(skb2, offset);
756 skb_reset_network_header(skb2);
758 /* Try to guess incoming interface */
759 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
762 if (rt && rt->dst.dev)
763 skb2->dev = rt->dst.dev;
765 icmpv6_send(skb2, rel_type, rel_code, rel_info);
775 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
776 const struct ipv6hdr *ipv6h,
779 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
781 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
782 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
784 return IP6_ECN_decapsulate(ipv6h, skb);
787 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
788 const struct ipv6hdr *ipv6h,
791 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
792 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
794 return IP6_ECN_decapsulate(ipv6h, skb);
797 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
798 const struct in6_addr *laddr,
799 const struct in6_addr *raddr)
801 struct __ip6_tnl_parm *p = &t->parms;
802 int ltype = ipv6_addr_type(laddr);
803 int rtype = ipv6_addr_type(raddr);
806 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
807 flags = IP6_TNL_F_CAP_PER_PACKET;
808 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
809 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
810 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
811 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
812 if (ltype&IPV6_ADDR_UNICAST)
813 flags |= IP6_TNL_F_CAP_XMIT;
814 if (rtype&IPV6_ADDR_UNICAST)
815 flags |= IP6_TNL_F_CAP_RCV;
819 EXPORT_SYMBOL(ip6_tnl_get_cap);
821 /* called with rcu_read_lock() */
822 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
823 const struct in6_addr *laddr,
824 const struct in6_addr *raddr)
826 struct __ip6_tnl_parm *p = &t->parms;
828 struct net *net = t->net;
830 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
831 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
832 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
833 struct net_device *ldev = NULL;
836 ldev = dev_get_by_index_rcu(net, p->link);
838 if ((ipv6_addr_is_multicast(laddr) ||
839 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
840 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
845 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
848 * ip6_tnl_rcv - decapsulate IPv6 packet and retransmit it locally
849 * @skb: received socket buffer
850 * @protocol: ethernet protocol ID
851 * @dscp_ecn_decapsulate: the function to decapsulate DSCP code and ECN
856 static int ip6_tnl_rcv(struct sk_buff *skb, __u16 protocol,
858 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
859 const struct ipv6hdr *ipv6h,
860 struct sk_buff *skb))
863 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
868 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
870 struct pcpu_sw_netstats *tstats;
872 tproto = ACCESS_ONCE(t->parms.proto);
873 if (tproto != ipproto && tproto != 0) {
878 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) {
883 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr)) {
884 t->dev->stats.rx_dropped++;
888 skb->mac_header = skb->network_header;
889 skb_reset_network_header(skb);
890 skb->protocol = htons(protocol);
891 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
893 __skb_tunnel_rx(skb, t->dev, t->net);
895 err = dscp_ecn_decapsulate(t, ipv6h, skb);
898 net_info_ratelimited("non-ECT from %pI6 with dsfield=%#x\n",
900 ipv6_get_dsfield(ipv6h));
902 ++t->dev->stats.rx_frame_errors;
903 ++t->dev->stats.rx_errors;
909 tstats = this_cpu_ptr(t->dev->tstats);
910 u64_stats_update_begin(&tstats->syncp);
911 tstats->rx_packets++;
912 tstats->rx_bytes += skb->len;
913 u64_stats_update_end(&tstats->syncp);
928 static int ip4ip6_rcv(struct sk_buff *skb)
930 return ip6_tnl_rcv(skb, ETH_P_IP, IPPROTO_IPIP,
931 ip4ip6_dscp_ecn_decapsulate);
934 static int ip6ip6_rcv(struct sk_buff *skb)
936 return ip6_tnl_rcv(skb, ETH_P_IPV6, IPPROTO_IPV6,
937 ip6ip6_dscp_ecn_decapsulate);
940 struct ipv6_tel_txoption {
941 struct ipv6_txoptions ops;
945 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
947 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
949 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
951 opt->dst_opt[4] = encap_limit;
952 opt->dst_opt[5] = IPV6_TLV_PADN;
955 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
956 opt->ops.opt_nflen = 8;
960 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
961 * @t: the outgoing tunnel device
962 * @hdr: IPv6 header from the incoming packet
965 * Avoid trivial tunneling loop by checking that tunnel exit-point
966 * doesn't match source of incoming packet.
974 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
976 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
979 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
980 const struct in6_addr *laddr,
981 const struct in6_addr *raddr)
983 struct __ip6_tnl_parm *p = &t->parms;
985 struct net *net = t->net;
987 if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
988 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
989 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
990 struct net_device *ldev = NULL;
994 ldev = dev_get_by_index_rcu(net, p->link);
996 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
997 pr_warn("%s xmit: Local address not yet configured!\n",
999 else if (!ipv6_addr_is_multicast(raddr) &&
1000 unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
1001 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1009 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1012 * ip6_tnl_xmit2 - encapsulate packet and send
1013 * @skb: the outgoing socket buffer
1014 * @dev: the outgoing tunnel device
1015 * @dsfield: dscp code for outer header
1016 * @fl: flow of tunneled packet
1017 * @encap_limit: encapsulation limit
1018 * @pmtu: Path MTU is stored if packet is too big
1021 * Build new header and do some sanity checks on the packet before sending
1027 * %-EMSGSIZE message too big. return mtu in this case.
1030 static int ip6_tnl_xmit2(struct sk_buff *skb,
1031 struct net_device *dev,
1037 struct ip6_tnl *t = netdev_priv(dev);
1038 struct net *net = t->net;
1039 struct net_device_stats *stats = &t->dev->stats;
1040 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1041 struct ipv6_tel_txoption opt;
1042 struct dst_entry *dst = NULL, *ndst = NULL;
1043 struct net_device *tdev;
1045 unsigned int max_headroom = sizeof(struct ipv6hdr);
1050 if (ipv6_addr_any(&t->parms.raddr)) {
1051 struct in6_addr *addr6;
1052 struct neighbour *neigh;
1056 goto tx_err_link_failure;
1058 neigh = dst_neigh_lookup(skb_dst(skb),
1059 &ipv6_hdr(skb)->daddr);
1061 goto tx_err_link_failure;
1063 addr6 = (struct in6_addr *)&neigh->primary_key;
1064 addr_type = ipv6_addr_type(addr6);
1066 if (addr_type == IPV6_ADDR_ANY)
1067 addr6 = &ipv6_hdr(skb)->daddr;
1069 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1070 neigh_release(neigh);
1071 } else if (!fl6->flowi6_mark)
1072 dst = ip6_tnl_dst_get(t);
1074 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1075 goto tx_err_link_failure;
1078 dst = ip6_route_output(net, NULL, fl6);
1081 goto tx_err_link_failure;
1082 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1086 goto tx_err_link_failure;
1094 stats->collisions++;
1095 net_warn_ratelimited("%s: Local routing loop detected!\n",
1097 goto tx_err_dst_release;
1099 mtu = dst_mtu(dst) - sizeof(*ipv6h);
1100 if (encap_limit >= 0) {
1104 if (mtu < IPV6_MIN_MTU)
1107 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1108 if (skb->len > mtu) {
1111 goto tx_err_dst_release;
1114 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1117 * Okay, now see if we can stuff it in the buffer as-is.
1119 max_headroom += LL_RESERVED_SPACE(tdev);
1121 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1122 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1123 struct sk_buff *new_skb;
1125 new_skb = skb_realloc_headroom(skb, max_headroom);
1127 goto tx_err_dst_release;
1130 skb_set_owner_w(new_skb, skb->sk);
1135 if (!fl6->flowi6_mark && ndst)
1136 ip6_tnl_dst_set(t, ndst);
1137 skb_dst_set(skb, dst);
1139 skb->transport_header = skb->network_header;
1141 proto = fl6->flowi6_proto;
1142 if (encap_limit >= 0) {
1143 init_tel_txopt(&opt, encap_limit);
1144 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1147 if (likely(!skb->encapsulation)) {
1148 skb_reset_inner_headers(skb);
1149 skb->encapsulation = 1;
1152 skb_push(skb, sizeof(struct ipv6hdr));
1153 skb_reset_network_header(skb);
1154 ipv6h = ipv6_hdr(skb);
1155 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1156 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1157 ipv6h->hop_limit = t->parms.hop_limit;
1158 ipv6h->nexthdr = proto;
1159 ipv6h->saddr = fl6->saddr;
1160 ipv6h->daddr = fl6->daddr;
1161 ip6tunnel_xmit(NULL, skb, dev);
1163 tx_err_link_failure:
1164 stats->tx_carrier_errors++;
1165 dst_link_failure(skb);
1172 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1174 struct ip6_tnl *t = netdev_priv(dev);
1175 const struct iphdr *iph = ip_hdr(skb);
1176 int encap_limit = -1;
1183 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1185 tproto = ACCESS_ONCE(t->parms.proto);
1186 if (tproto != IPPROTO_IPIP && tproto != 0)
1189 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1190 encap_limit = t->parms.encap_limit;
1192 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1193 fl6.flowi6_proto = IPPROTO_IPIP;
1195 dsfield = ipv4_get_dsfield(iph);
1197 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1198 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1200 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1201 fl6.flowi6_mark = skb->mark;
1203 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1205 /* XXX: send ICMP error even if DF is not set. */
1206 if (err == -EMSGSIZE)
1207 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1216 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1218 struct ip6_tnl *t = netdev_priv(dev);
1219 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1220 int encap_limit = -1;
1228 tproto = ACCESS_ONCE(t->parms.proto);
1229 if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1230 ip6_tnl_addr_conflict(t, ipv6h))
1233 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1235 struct ipv6_tlv_tnl_enc_lim *tel;
1236 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1237 if (tel->encap_limit == 0) {
1238 icmpv6_send(skb, ICMPV6_PARAMPROB,
1239 ICMPV6_HDR_FIELD, offset + 2);
1242 encap_limit = tel->encap_limit - 1;
1243 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1244 encap_limit = t->parms.encap_limit;
1246 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1247 fl6.flowi6_proto = IPPROTO_IPV6;
1249 dsfield = ipv6_get_dsfield(ipv6h);
1250 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1251 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1252 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1253 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1254 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1255 fl6.flowi6_mark = skb->mark;
1257 err = ip6_tnl_xmit2(skb, dev, dsfield, &fl6, encap_limit, &mtu);
1259 if (err == -EMSGSIZE)
1260 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1268 ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1270 struct ip6_tnl *t = netdev_priv(dev);
1271 struct net_device_stats *stats = &t->dev->stats;
1274 switch (skb->protocol) {
1275 case htons(ETH_P_IP):
1276 ret = ip4ip6_tnl_xmit(skb, dev);
1278 case htons(ETH_P_IPV6):
1279 ret = ip6ip6_tnl_xmit(skb, dev);
1288 return NETDEV_TX_OK;
1292 stats->tx_dropped++;
1294 return NETDEV_TX_OK;
1297 static void ip6_tnl_link_config(struct ip6_tnl *t)
1299 struct net_device *dev = t->dev;
1300 struct __ip6_tnl_parm *p = &t->parms;
1301 struct flowi6 *fl6 = &t->fl.u.ip6;
1303 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1304 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1306 /* Set up flowi template */
1307 fl6->saddr = p->laddr;
1308 fl6->daddr = p->raddr;
1309 fl6->flowi6_oif = p->link;
1312 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1313 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1314 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1315 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1317 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1318 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1320 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1321 dev->flags |= IFF_POINTOPOINT;
1323 dev->flags &= ~IFF_POINTOPOINT;
1325 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1326 int strict = (ipv6_addr_type(&p->raddr) &
1327 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1329 struct rt6_info *rt = rt6_lookup(t->net,
1330 &p->raddr, &p->laddr,
1337 dev->hard_header_len = rt->dst.dev->hard_header_len +
1338 sizeof(struct ipv6hdr);
1340 dev->mtu = rt->dst.dev->mtu - sizeof(struct ipv6hdr);
1341 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1344 if (dev->mtu < IPV6_MIN_MTU)
1345 dev->mtu = IPV6_MIN_MTU;
1352 * ip6_tnl_change - update the tunnel parameters
1353 * @t: tunnel to be changed
1354 * @p: tunnel configuration parameters
1357 * ip6_tnl_change() updates the tunnel parameters
1361 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1363 t->parms.laddr = p->laddr;
1364 t->parms.raddr = p->raddr;
1365 t->parms.flags = p->flags;
1366 t->parms.hop_limit = p->hop_limit;
1367 t->parms.encap_limit = p->encap_limit;
1368 t->parms.flowinfo = p->flowinfo;
1369 t->parms.link = p->link;
1370 t->parms.proto = p->proto;
1371 ip6_tnl_dst_reset(t);
1372 ip6_tnl_link_config(t);
1376 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1378 struct net *net = t->net;
1379 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1382 ip6_tnl_unlink(ip6n, t);
1384 err = ip6_tnl_change(t, p);
1385 ip6_tnl_link(ip6n, t);
1386 netdev_state_change(t->dev);
1390 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1392 /* for default tnl0 device allow to change only the proto */
1393 t->parms.proto = p->proto;
1394 netdev_state_change(t->dev);
1399 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1401 p->laddr = u->laddr;
1402 p->raddr = u->raddr;
1403 p->flags = u->flags;
1404 p->hop_limit = u->hop_limit;
1405 p->encap_limit = u->encap_limit;
1406 p->flowinfo = u->flowinfo;
1408 p->proto = u->proto;
1409 memcpy(p->name, u->name, sizeof(u->name));
1413 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1415 u->laddr = p->laddr;
1416 u->raddr = p->raddr;
1417 u->flags = p->flags;
1418 u->hop_limit = p->hop_limit;
1419 u->encap_limit = p->encap_limit;
1420 u->flowinfo = p->flowinfo;
1422 u->proto = p->proto;
1423 memcpy(u->name, p->name, sizeof(u->name));
1427 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1428 * @dev: virtual device associated with tunnel
1429 * @ifr: parameters passed from userspace
1430 * @cmd: command to be performed
1433 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1436 * The possible commands are the following:
1437 * %SIOCGETTUNNEL: get tunnel parameters for device
1438 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1439 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1440 * %SIOCDELTUNNEL: delete tunnel
1442 * The fallback device "ip6tnl0", created during module
1443 * initialization, can be used for creating other tunnel devices.
1447 * %-EFAULT if unable to copy data to or from userspace,
1448 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1449 * %-EINVAL if passed tunnel parameters are invalid,
1450 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1451 * %-ENODEV if attempting to change or delete a nonexisting device
1455 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1458 struct ip6_tnl_parm p;
1459 struct __ip6_tnl_parm p1;
1460 struct ip6_tnl *t = netdev_priv(dev);
1461 struct net *net = t->net;
1462 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1466 if (dev == ip6n->fb_tnl_dev) {
1467 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1471 ip6_tnl_parm_from_user(&p1, &p);
1472 t = ip6_tnl_locate(net, &p1, 0);
1474 t = netdev_priv(dev);
1476 memset(&p, 0, sizeof(p));
1478 ip6_tnl_parm_to_user(&p, &t->parms);
1479 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1486 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1489 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1492 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1495 ip6_tnl_parm_from_user(&p1, &p);
1496 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1497 if (cmd == SIOCCHGTUNNEL) {
1499 if (t->dev != dev) {
1504 t = netdev_priv(dev);
1505 if (dev == ip6n->fb_tnl_dev)
1506 err = ip6_tnl0_update(t, &p1);
1508 err = ip6_tnl_update(t, &p1);
1512 ip6_tnl_parm_to_user(&p, &t->parms);
1513 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1522 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1525 if (dev == ip6n->fb_tnl_dev) {
1527 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1530 ip6_tnl_parm_from_user(&p1, &p);
1531 t = ip6_tnl_locate(net, &p1, 0);
1535 if (t->dev == ip6n->fb_tnl_dev)
1540 unregister_netdevice(dev);
1549 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1550 * @dev: virtual device associated with tunnel
1551 * @new_mtu: the new mtu
1555 * %-EINVAL if mtu too small
1559 ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1561 struct ip6_tnl *tnl = netdev_priv(dev);
1563 if (tnl->parms.proto == IPPROTO_IPIP) {
1567 if (new_mtu < IPV6_MIN_MTU)
1570 if (new_mtu > 0xFFF8 - dev->hard_header_len)
1576 int ip6_tnl_get_iflink(const struct net_device *dev)
1578 struct ip6_tnl *t = netdev_priv(dev);
1580 return t->parms.link;
1582 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1584 static const struct net_device_ops ip6_tnl_netdev_ops = {
1585 .ndo_init = ip6_tnl_dev_init,
1586 .ndo_uninit = ip6_tnl_dev_uninit,
1587 .ndo_start_xmit = ip6_tnl_xmit,
1588 .ndo_do_ioctl = ip6_tnl_ioctl,
1589 .ndo_change_mtu = ip6_tnl_change_mtu,
1590 .ndo_get_stats = ip6_get_stats,
1591 .ndo_get_iflink = ip6_tnl_get_iflink,
1596 * ip6_tnl_dev_setup - setup virtual tunnel device
1597 * @dev: virtual device associated with tunnel
1600 * Initialize function pointers and device parameters
1603 static void ip6_tnl_dev_setup(struct net_device *dev)
1607 dev->netdev_ops = &ip6_tnl_netdev_ops;
1608 dev->destructor = ip6_dev_free;
1610 dev->type = ARPHRD_TUNNEL6;
1611 dev->hard_header_len = LL_MAX_HEADER + sizeof(struct ipv6hdr);
1612 dev->mtu = ETH_DATA_LEN - sizeof(struct ipv6hdr);
1613 t = netdev_priv(dev);
1614 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1616 dev->flags |= IFF_NOARP;
1617 dev->addr_len = sizeof(struct in6_addr);
1618 netif_keep_dst(dev);
1619 /* This perm addr will be used as interface identifier by IPv6 */
1620 dev->addr_assign_type = NET_ADDR_RANDOM;
1621 eth_random_addr(dev->perm_addr);
1626 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1627 * @dev: virtual device associated with tunnel
1631 ip6_tnl_dev_init_gen(struct net_device *dev)
1633 struct ip6_tnl *t = netdev_priv(dev);
1637 t->net = dev_net(dev);
1638 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1642 ret = ip6_tnl_dst_init(t);
1644 free_percpu(dev->tstats);
1653 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1654 * @dev: virtual device associated with tunnel
1657 static int ip6_tnl_dev_init(struct net_device *dev)
1659 struct ip6_tnl *t = netdev_priv(dev);
1660 int err = ip6_tnl_dev_init_gen(dev);
1664 ip6_tnl_link_config(t);
1669 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1670 * @dev: fallback device
1675 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1677 struct ip6_tnl *t = netdev_priv(dev);
1678 struct net *net = dev_net(dev);
1679 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1681 t->parms.proto = IPPROTO_IPV6;
1684 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1688 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1692 if (!data || !data[IFLA_IPTUN_PROTO])
1695 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1696 if (proto != IPPROTO_IPV6 &&
1697 proto != IPPROTO_IPIP &&
1704 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1705 struct __ip6_tnl_parm *parms)
1707 memset(parms, 0, sizeof(*parms));
1712 if (data[IFLA_IPTUN_LINK])
1713 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1715 if (data[IFLA_IPTUN_LOCAL])
1716 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1718 if (data[IFLA_IPTUN_REMOTE])
1719 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1721 if (data[IFLA_IPTUN_TTL])
1722 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1724 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1725 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1727 if (data[IFLA_IPTUN_FLOWINFO])
1728 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1730 if (data[IFLA_IPTUN_FLAGS])
1731 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1733 if (data[IFLA_IPTUN_PROTO])
1734 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1737 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1738 struct nlattr *tb[], struct nlattr *data[])
1740 struct net *net = dev_net(dev);
1741 struct ip6_tnl *nt, *t;
1743 nt = netdev_priv(dev);
1744 ip6_tnl_netlink_parms(data, &nt->parms);
1746 t = ip6_tnl_locate(net, &nt->parms, 0);
1750 return ip6_tnl_create2(dev);
1753 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1754 struct nlattr *data[])
1756 struct ip6_tnl *t = netdev_priv(dev);
1757 struct __ip6_tnl_parm p;
1758 struct net *net = t->net;
1759 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1761 if (dev == ip6n->fb_tnl_dev)
1764 ip6_tnl_netlink_parms(data, &p);
1766 t = ip6_tnl_locate(net, &p, 0);
1771 t = netdev_priv(dev);
1773 return ip6_tnl_update(t, &p);
1776 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1778 struct net *net = dev_net(dev);
1779 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1781 if (dev != ip6n->fb_tnl_dev)
1782 unregister_netdevice_queue(dev, head);
1785 static size_t ip6_tnl_get_size(const struct net_device *dev)
1788 /* IFLA_IPTUN_LINK */
1790 /* IFLA_IPTUN_LOCAL */
1791 nla_total_size(sizeof(struct in6_addr)) +
1792 /* IFLA_IPTUN_REMOTE */
1793 nla_total_size(sizeof(struct in6_addr)) +
1794 /* IFLA_IPTUN_TTL */
1796 /* IFLA_IPTUN_ENCAP_LIMIT */
1798 /* IFLA_IPTUN_FLOWINFO */
1800 /* IFLA_IPTUN_FLAGS */
1802 /* IFLA_IPTUN_PROTO */
1807 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1809 struct ip6_tnl *tunnel = netdev_priv(dev);
1810 struct __ip6_tnl_parm *parm = &tunnel->parms;
1812 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1813 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
1814 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
1815 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1816 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1817 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1818 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1819 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1820 goto nla_put_failure;
1827 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
1829 struct ip6_tnl *tunnel = netdev_priv(dev);
1833 EXPORT_SYMBOL(ip6_tnl_get_link_net);
1835 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1836 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1837 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
1838 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
1839 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1840 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
1841 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
1842 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
1843 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1846 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1848 .maxtype = IFLA_IPTUN_MAX,
1849 .policy = ip6_tnl_policy,
1850 .priv_size = sizeof(struct ip6_tnl),
1851 .setup = ip6_tnl_dev_setup,
1852 .validate = ip6_tnl_validate,
1853 .newlink = ip6_tnl_newlink,
1854 .changelink = ip6_tnl_changelink,
1855 .dellink = ip6_tnl_dellink,
1856 .get_size = ip6_tnl_get_size,
1857 .fill_info = ip6_tnl_fill_info,
1858 .get_link_net = ip6_tnl_get_link_net,
1861 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
1862 .handler = ip4ip6_rcv,
1863 .err_handler = ip4ip6_err,
1867 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
1868 .handler = ip6ip6_rcv,
1869 .err_handler = ip6ip6_err,
1873 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
1875 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1876 struct net_device *dev, *aux;
1881 for_each_netdev_safe(net, dev, aux)
1882 if (dev->rtnl_link_ops == &ip6_link_ops)
1883 unregister_netdevice_queue(dev, &list);
1885 for (h = 0; h < HASH_SIZE; h++) {
1886 t = rtnl_dereference(ip6n->tnls_r_l[h]);
1888 /* If dev is in the same netns, it has already
1889 * been added to the list by the previous loop.
1891 if (!net_eq(dev_net(t->dev), net))
1892 unregister_netdevice_queue(t->dev, &list);
1893 t = rtnl_dereference(t->next);
1897 unregister_netdevice_many(&list);
1900 static int __net_init ip6_tnl_init_net(struct net *net)
1902 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1903 struct ip6_tnl *t = NULL;
1906 ip6n->tnls[0] = ip6n->tnls_wc;
1907 ip6n->tnls[1] = ip6n->tnls_r_l;
1910 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
1911 NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
1913 if (!ip6n->fb_tnl_dev)
1915 dev_net_set(ip6n->fb_tnl_dev, net);
1916 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
1917 /* FB netdevice is special: we have one, and only one per netns.
1918 * Allowing to move it to another netns is clearly unsafe.
1920 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
1922 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
1926 err = register_netdev(ip6n->fb_tnl_dev);
1930 t = netdev_priv(ip6n->fb_tnl_dev);
1932 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
1936 ip6_dev_free(ip6n->fb_tnl_dev);
1941 static void __net_exit ip6_tnl_exit_net(struct net *net)
1944 ip6_tnl_destroy_tunnels(net);
1948 static struct pernet_operations ip6_tnl_net_ops = {
1949 .init = ip6_tnl_init_net,
1950 .exit = ip6_tnl_exit_net,
1951 .id = &ip6_tnl_net_id,
1952 .size = sizeof(struct ip6_tnl_net),
1956 * ip6_tunnel_init - register protocol and reserve needed resources
1958 * Return: 0 on success
1961 static int __init ip6_tunnel_init(void)
1965 err = register_pernet_device(&ip6_tnl_net_ops);
1969 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
1971 pr_err("%s: can't register ip4ip6\n", __func__);
1975 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
1977 pr_err("%s: can't register ip6ip6\n", __func__);
1980 err = rtnl_link_register(&ip6_link_ops);
1982 goto rtnl_link_failed;
1987 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
1989 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
1991 unregister_pernet_device(&ip6_tnl_net_ops);
1997 * ip6_tunnel_cleanup - free resources and unregister protocol
2000 static void __exit ip6_tunnel_cleanup(void)
2002 rtnl_link_unregister(&ip6_link_ops);
2003 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2004 pr_info("%s: can't deregister ip4ip6\n", __func__);
2006 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2007 pr_info("%s: can't deregister ip6ip6\n", __func__);
2009 unregister_pernet_device(&ip6_tnl_net_ops);
2012 module_init(ip6_tunnel_init);
2013 module_exit(ip6_tunnel_cleanup);