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>
60 #include <net/dst_metadata.h>
62 MODULE_AUTHOR("Ville Nuorvala");
63 MODULE_DESCRIPTION("IPv6 tunneling device");
64 MODULE_LICENSE("GPL");
65 MODULE_ALIAS_RTNL_LINK("ip6tnl");
66 MODULE_ALIAS_NETDEV("ip6tnl0");
68 #define IP6_TUNNEL_HASH_SIZE_SHIFT 5
69 #define IP6_TUNNEL_HASH_SIZE (1 << IP6_TUNNEL_HASH_SIZE_SHIFT)
71 static bool log_ecn_error = true;
72 module_param(log_ecn_error, bool, 0644);
73 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
75 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
77 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
79 return hash_32(hash, IP6_TUNNEL_HASH_SIZE_SHIFT);
82 static int ip6_tnl_dev_init(struct net_device *dev);
83 static void ip6_tnl_dev_setup(struct net_device *dev);
84 static struct rtnl_link_ops ip6_link_ops __read_mostly;
86 static int ip6_tnl_net_id __read_mostly;
88 /* the IPv6 tunnel fallback device */
89 struct net_device *fb_tnl_dev;
90 /* lists for storing tunnels in use */
91 struct ip6_tnl __rcu *tnls_r_l[IP6_TUNNEL_HASH_SIZE];
92 struct ip6_tnl __rcu *tnls_wc[1];
93 struct ip6_tnl __rcu **tnls[2];
94 struct ip6_tnl __rcu *collect_md_tun;
97 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
99 struct pcpu_sw_netstats tmp, sum = { 0 };
102 for_each_possible_cpu(i) {
104 const struct pcpu_sw_netstats *tstats =
105 per_cpu_ptr(dev->tstats, i);
108 start = u64_stats_fetch_begin_irq(&tstats->syncp);
109 tmp.rx_packets = tstats->rx_packets;
110 tmp.rx_bytes = tstats->rx_bytes;
111 tmp.tx_packets = tstats->tx_packets;
112 tmp.tx_bytes = tstats->tx_bytes;
113 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
115 sum.rx_packets += tmp.rx_packets;
116 sum.rx_bytes += tmp.rx_bytes;
117 sum.tx_packets += tmp.tx_packets;
118 sum.tx_bytes += tmp.tx_bytes;
120 dev->stats.rx_packets = sum.rx_packets;
121 dev->stats.rx_bytes = sum.rx_bytes;
122 dev->stats.tx_packets = sum.tx_packets;
123 dev->stats.tx_bytes = sum.tx_bytes;
128 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
129 * @remote: the address of the tunnel exit-point
130 * @local: the address of the tunnel entry-point
133 * tunnel matching given end-points if found,
134 * else fallback tunnel if its device is up,
138 #define for_each_ip6_tunnel_rcu(start) \
139 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
141 static struct ip6_tnl *
142 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
144 unsigned int hash = HASH(remote, local);
146 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
149 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
150 if (ipv6_addr_equal(local, &t->parms.laddr) &&
151 ipv6_addr_equal(remote, &t->parms.raddr) &&
152 (t->dev->flags & IFF_UP))
156 memset(&any, 0, sizeof(any));
157 hash = HASH(&any, local);
158 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
159 if (ipv6_addr_equal(local, &t->parms.laddr) &&
160 (t->dev->flags & IFF_UP))
164 hash = HASH(remote, &any);
165 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
166 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
167 (t->dev->flags & IFF_UP))
171 t = rcu_dereference(ip6n->collect_md_tun);
175 t = rcu_dereference(ip6n->tnls_wc[0]);
176 if (t && (t->dev->flags & IFF_UP))
183 * ip6_tnl_bucket - get head of list matching given tunnel parameters
184 * @p: parameters containing tunnel end-points
187 * ip6_tnl_bucket() returns the head of the list matching the
188 * &struct in6_addr entries laddr and raddr in @p.
190 * Return: head of IPv6 tunnel list
193 static struct ip6_tnl __rcu **
194 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
196 const struct in6_addr *remote = &p->raddr;
197 const struct in6_addr *local = &p->laddr;
201 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
203 h = HASH(remote, local);
205 return &ip6n->tnls[prio][h];
209 * ip6_tnl_link - add tunnel to hash table
210 * @t: tunnel to be added
214 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
216 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
218 if (t->parms.collect_md)
219 rcu_assign_pointer(ip6n->collect_md_tun, t);
220 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
221 rcu_assign_pointer(*tp, t);
225 * ip6_tnl_unlink - remove tunnel from hash table
226 * @t: tunnel to be removed
230 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
232 struct ip6_tnl __rcu **tp;
233 struct ip6_tnl *iter;
235 if (t->parms.collect_md)
236 rcu_assign_pointer(ip6n->collect_md_tun, NULL);
238 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
239 (iter = rtnl_dereference(*tp)) != NULL;
242 rcu_assign_pointer(*tp, t->next);
248 static void ip6_dev_free(struct net_device *dev)
250 struct ip6_tnl *t = netdev_priv(dev);
252 gro_cells_destroy(&t->gro_cells);
253 dst_cache_destroy(&t->dst_cache);
254 free_percpu(dev->tstats);
258 static int ip6_tnl_create2(struct net_device *dev)
260 struct ip6_tnl *t = netdev_priv(dev);
261 struct net *net = dev_net(dev);
262 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
265 t = netdev_priv(dev);
267 dev->rtnl_link_ops = &ip6_link_ops;
268 err = register_netdevice(dev);
272 strcpy(t->parms.name, dev->name);
275 ip6_tnl_link(ip6n, t);
283 * ip6_tnl_create - create a new tunnel
284 * @p: tunnel parameters
285 * @pt: pointer to new tunnel
288 * Create tunnel matching given parameters.
291 * created tunnel or error pointer
294 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
296 struct net_device *dev;
302 strlcpy(name, p->name, IFNAMSIZ);
304 sprintf(name, "ip6tnl%%d");
306 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
311 dev_net_set(dev, net);
313 t = netdev_priv(dev);
315 t->net = dev_net(dev);
316 err = ip6_tnl_create2(dev);
329 * ip6_tnl_locate - find or create tunnel matching given parameters
330 * @p: tunnel parameters
331 * @create: != 0 if allowed to create new tunnel if no match found
334 * ip6_tnl_locate() first tries to locate an existing tunnel
335 * based on @parms. If this is unsuccessful, but @create is set a new
336 * tunnel device is created and registered for use.
339 * matching tunnel or error pointer
342 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
343 struct __ip6_tnl_parm *p, int create)
345 const struct in6_addr *remote = &p->raddr;
346 const struct in6_addr *local = &p->laddr;
347 struct ip6_tnl __rcu **tp;
349 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
351 for (tp = ip6_tnl_bucket(ip6n, p);
352 (t = rtnl_dereference(*tp)) != NULL;
354 if (ipv6_addr_equal(local, &t->parms.laddr) &&
355 ipv6_addr_equal(remote, &t->parms.raddr)) {
357 return ERR_PTR(-EEXIST);
363 return ERR_PTR(-ENODEV);
364 return ip6_tnl_create(net, p);
368 * ip6_tnl_dev_uninit - tunnel device uninitializer
369 * @dev: the device to be destroyed
372 * ip6_tnl_dev_uninit() removes tunnel from its list
376 ip6_tnl_dev_uninit(struct net_device *dev)
378 struct ip6_tnl *t = netdev_priv(dev);
379 struct net *net = t->net;
380 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
382 if (dev == ip6n->fb_tnl_dev)
383 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
385 ip6_tnl_unlink(ip6n, t);
386 dst_cache_reset(&t->dst_cache);
391 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
392 * @skb: received socket buffer
395 * 0 if none was found,
396 * else index to encapsulation limit
399 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
401 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
402 __u8 nexthdr = ipv6h->nexthdr;
403 __u16 off = sizeof(*ipv6h);
405 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
407 struct ipv6_opt_hdr *hdr;
408 if (raw + off + sizeof(*hdr) > skb->data &&
409 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
412 hdr = (struct ipv6_opt_hdr *) (raw + off);
413 if (nexthdr == NEXTHDR_FRAGMENT) {
414 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
415 if (frag_hdr->frag_off)
418 } else if (nexthdr == NEXTHDR_AUTH) {
419 optlen = (hdr->hdrlen + 2) << 2;
421 optlen = ipv6_optlen(hdr);
423 if (nexthdr == NEXTHDR_DEST) {
426 struct ipv6_tlv_tnl_enc_lim *tel;
428 /* No more room for encapsulation limit */
429 if (i + sizeof (*tel) > off + optlen)
432 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
433 /* return index of option if found and valid */
434 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
437 /* else jump to next option */
439 i += tel->length + 2;
444 nexthdr = hdr->nexthdr;
449 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
452 * ip6_tnl_err - tunnel error handler
455 * ip6_tnl_err() should handle errors in the tunnel according
456 * to the specifications in RFC 2473.
460 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
461 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
463 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
466 u8 rel_type = ICMPV6_DEST_UNREACH;
467 u8 rel_code = ICMPV6_ADDR_UNREACH;
473 /* If the packet doesn't contain the original IPv6 header we are
474 in trouble since we might need the source address for further
475 processing of the error. */
478 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
482 tproto = ACCESS_ONCE(t->parms.proto);
483 if (tproto != ipproto && tproto != 0)
490 struct ipv6_tlv_tnl_enc_lim *tel;
492 case ICMPV6_DEST_UNREACH:
493 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
497 case ICMPV6_TIME_EXCEED:
498 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
499 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
504 case ICMPV6_PARAMPROB:
506 if ((*code) == ICMPV6_HDR_FIELD)
507 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
509 if (teli && teli == *info - 2) {
510 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
511 if (tel->encap_limit == 0) {
512 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
517 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
521 case ICMPV6_PKT_TOOBIG:
522 mtu = *info - offset;
523 if (mtu < IPV6_MIN_MTU)
527 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
529 rel_type = ICMPV6_PKT_TOOBIG;
548 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
549 u8 type, u8 code, int offset, __be32 info)
554 __u32 rel_info = ntohl(info);
556 struct sk_buff *skb2;
557 const struct iphdr *eiph;
561 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
562 &rel_msg, &rel_info, offset);
570 case ICMPV6_DEST_UNREACH:
571 if (rel_code != ICMPV6_ADDR_UNREACH)
573 rel_type = ICMP_DEST_UNREACH;
574 rel_code = ICMP_HOST_UNREACH;
576 case ICMPV6_PKT_TOOBIG:
579 rel_type = ICMP_DEST_UNREACH;
580 rel_code = ICMP_FRAG_NEEDED;
583 rel_type = ICMP_REDIRECT;
584 rel_code = ICMP_REDIR_HOST;
589 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
592 skb2 = skb_clone(skb, GFP_ATOMIC);
598 skb_pull(skb2, offset);
599 skb_reset_network_header(skb2);
602 /* Try to guess incoming interface */
603 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
606 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
610 skb2->dev = rt->dst.dev;
612 /* route "incoming" packet */
613 if (rt->rt_flags & RTCF_LOCAL) {
616 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
617 eiph->daddr, eiph->saddr,
620 RT_TOS(eiph->tos), 0);
622 rt->dst.dev->type != ARPHRD_TUNNEL) {
627 skb_dst_set(skb2, &rt->dst);
630 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
632 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
636 /* change mtu on this route */
637 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
638 if (rel_info > dst_mtu(skb_dst(skb2)))
641 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
643 if (rel_type == ICMP_REDIRECT)
644 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
646 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
654 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
655 u8 type, u8 code, int offset, __be32 info)
660 __u32 rel_info = ntohl(info);
663 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
664 &rel_msg, &rel_info, offset);
668 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
670 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
676 skb_pull(skb2, offset);
677 skb_reset_network_header(skb2);
679 /* Try to guess incoming interface */
680 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
683 if (rt && rt->dst.dev)
684 skb2->dev = rt->dst.dev;
686 icmpv6_send(skb2, rel_type, rel_code, rel_info);
696 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
697 const struct ipv6hdr *ipv6h,
700 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
702 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
703 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
705 return IP6_ECN_decapsulate(ipv6h, skb);
708 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
709 const struct ipv6hdr *ipv6h,
712 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
713 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
715 return IP6_ECN_decapsulate(ipv6h, skb);
718 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
719 const struct in6_addr *laddr,
720 const struct in6_addr *raddr)
722 struct __ip6_tnl_parm *p = &t->parms;
723 int ltype = ipv6_addr_type(laddr);
724 int rtype = ipv6_addr_type(raddr);
727 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
728 flags = IP6_TNL_F_CAP_PER_PACKET;
729 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
730 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
731 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
732 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
733 if (ltype&IPV6_ADDR_UNICAST)
734 flags |= IP6_TNL_F_CAP_XMIT;
735 if (rtype&IPV6_ADDR_UNICAST)
736 flags |= IP6_TNL_F_CAP_RCV;
740 EXPORT_SYMBOL(ip6_tnl_get_cap);
742 /* called with rcu_read_lock() */
743 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
744 const struct in6_addr *laddr,
745 const struct in6_addr *raddr)
747 struct __ip6_tnl_parm *p = &t->parms;
749 struct net *net = t->net;
751 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
752 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
753 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
754 struct net_device *ldev = NULL;
757 ldev = dev_get_by_index_rcu(net, p->link);
759 if ((ipv6_addr_is_multicast(laddr) ||
760 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
761 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
766 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
768 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
769 const struct tnl_ptk_info *tpi,
770 struct metadata_dst *tun_dst,
771 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
772 const struct ipv6hdr *ipv6h,
773 struct sk_buff *skb),
776 struct pcpu_sw_netstats *tstats;
777 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
780 if ((!(tpi->flags & TUNNEL_CSUM) &&
781 (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
782 ((tpi->flags & TUNNEL_CSUM) &&
783 !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
784 tunnel->dev->stats.rx_crc_errors++;
785 tunnel->dev->stats.rx_errors++;
789 if (tunnel->parms.i_flags & TUNNEL_SEQ) {
790 if (!(tpi->flags & TUNNEL_SEQ) ||
792 (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
793 tunnel->dev->stats.rx_fifo_errors++;
794 tunnel->dev->stats.rx_errors++;
797 tunnel->i_seqno = ntohl(tpi->seq) + 1;
800 skb->protocol = tpi->proto;
802 /* Warning: All skb pointers will be invalidated! */
803 if (tunnel->dev->type == ARPHRD_ETHER) {
804 if (!pskb_may_pull(skb, ETH_HLEN)) {
805 tunnel->dev->stats.rx_length_errors++;
806 tunnel->dev->stats.rx_errors++;
810 ipv6h = ipv6_hdr(skb);
811 skb->protocol = eth_type_trans(skb, tunnel->dev);
812 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
814 skb->dev = tunnel->dev;
817 skb_reset_network_header(skb);
818 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
820 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
822 err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
825 net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
827 ipv6_get_dsfield(ipv6h));
829 ++tunnel->dev->stats.rx_frame_errors;
830 ++tunnel->dev->stats.rx_errors;
835 tstats = this_cpu_ptr(tunnel->dev->tstats);
836 u64_stats_update_begin(&tstats->syncp);
837 tstats->rx_packets++;
838 tstats->rx_bytes += skb->len;
839 u64_stats_update_end(&tstats->syncp);
841 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
844 skb_dst_set(skb, (struct dst_entry *)tun_dst);
846 gro_cells_receive(&tunnel->gro_cells, skb);
854 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
855 const struct tnl_ptk_info *tpi,
856 struct metadata_dst *tun_dst,
859 return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate,
862 EXPORT_SYMBOL(ip6_tnl_rcv);
864 static const struct tnl_ptk_info tpi_v6 = {
865 /* no tunnel info required for ipxip6. */
866 .proto = htons(ETH_P_IPV6),
869 static const struct tnl_ptk_info tpi_v4 = {
870 /* no tunnel info required for ipxip6. */
871 .proto = htons(ETH_P_IP),
874 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
875 const struct tnl_ptk_info *tpi,
876 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
877 const struct ipv6hdr *ipv6h,
878 struct sk_buff *skb))
881 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
882 struct metadata_dst *tun_dst = NULL;
886 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
889 u8 tproto = ACCESS_ONCE(t->parms.proto);
891 if (tproto != ipproto && tproto != 0)
893 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
895 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
897 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
899 if (t->parms.collect_md) {
900 tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
904 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
918 static int ip4ip6_rcv(struct sk_buff *skb)
920 return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
921 ip4ip6_dscp_ecn_decapsulate);
924 static int ip6ip6_rcv(struct sk_buff *skb)
926 return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
927 ip6ip6_dscp_ecn_decapsulate);
930 struct ipv6_tel_txoption {
931 struct ipv6_txoptions ops;
935 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
937 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
939 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
941 opt->dst_opt[4] = encap_limit;
942 opt->dst_opt[5] = IPV6_TLV_PADN;
945 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
946 opt->ops.opt_nflen = 8;
950 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
951 * @t: the outgoing tunnel device
952 * @hdr: IPv6 header from the incoming packet
955 * Avoid trivial tunneling loop by checking that tunnel exit-point
956 * doesn't match source of incoming packet.
964 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
966 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
969 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
970 const struct in6_addr *laddr,
971 const struct in6_addr *raddr)
973 struct __ip6_tnl_parm *p = &t->parms;
975 struct net *net = t->net;
977 if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
978 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
979 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
980 struct net_device *ldev = NULL;
984 ldev = dev_get_by_index_rcu(net, p->link);
986 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
987 pr_warn("%s xmit: Local address not yet configured!\n",
989 else if (!ipv6_addr_is_multicast(raddr) &&
990 unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
991 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
999 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1002 * ip6_tnl_xmit - encapsulate packet and send
1003 * @skb: the outgoing socket buffer
1004 * @dev: the outgoing tunnel device
1005 * @dsfield: dscp code for outer header
1006 * @fl6: flow of tunneled packet
1007 * @encap_limit: encapsulation limit
1008 * @pmtu: Path MTU is stored if packet is too big
1009 * @proto: next header value
1012 * Build new header and do some sanity checks on the packet before sending
1018 * %-EMSGSIZE message too big. return mtu in this case.
1021 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1022 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1025 struct ip6_tnl *t = netdev_priv(dev);
1026 struct net *net = t->net;
1027 struct net_device_stats *stats = &t->dev->stats;
1028 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1029 struct ipv6_tel_txoption opt;
1030 struct dst_entry *dst = NULL, *ndst = NULL;
1031 struct net_device *tdev;
1033 unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1034 unsigned int max_headroom = psh_hlen;
1038 if (t->parms.collect_md) {
1039 hop_limit = skb_tunnel_info(skb)->key.ttl;
1042 hop_limit = t->parms.hop_limit;
1046 if (ipv6_addr_any(&t->parms.raddr)) {
1047 struct in6_addr *addr6;
1048 struct neighbour *neigh;
1052 goto tx_err_link_failure;
1054 neigh = dst_neigh_lookup(skb_dst(skb),
1055 &ipv6_hdr(skb)->daddr);
1057 goto tx_err_link_failure;
1059 addr6 = (struct in6_addr *)&neigh->primary_key;
1060 addr_type = ipv6_addr_type(addr6);
1062 if (addr_type == IPV6_ADDR_ANY)
1063 addr6 = &ipv6_hdr(skb)->daddr;
1065 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1066 neigh_release(neigh);
1067 } else if (!fl6->flowi6_mark)
1068 dst = dst_cache_get(&t->dst_cache);
1070 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1071 goto tx_err_link_failure;
1075 dst = ip6_route_output(net, NULL, fl6);
1078 goto tx_err_link_failure;
1079 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1083 goto tx_err_link_failure;
1085 if (t->parms.collect_md &&
1086 ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1087 &fl6->daddr, 0, &fl6->saddr))
1088 goto tx_err_link_failure;
1095 stats->collisions++;
1096 net_warn_ratelimited("%s: Local routing loop detected!\n",
1098 goto tx_err_dst_release;
1100 mtu = dst_mtu(dst) - psh_hlen;
1101 if (encap_limit >= 0) {
1105 if (mtu < IPV6_MIN_MTU)
1107 if (skb_dst(skb) && !t->parms.collect_md)
1108 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1109 if (skb->len > mtu && !skb_is_gso(skb)) {
1112 goto tx_err_dst_release;
1115 if (t->err_count > 0) {
1116 if (time_before(jiffies,
1117 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1120 dst_link_failure(skb);
1126 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1129 * Okay, now see if we can stuff it in the buffer as-is.
1131 max_headroom += LL_RESERVED_SPACE(tdev);
1133 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1134 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1135 struct sk_buff *new_skb;
1137 new_skb = skb_realloc_headroom(skb, max_headroom);
1139 goto tx_err_dst_release;
1142 skb_set_owner_w(new_skb, skb->sk);
1147 if (t->parms.collect_md) {
1148 if (t->encap.type != TUNNEL_ENCAP_NONE)
1149 goto tx_err_dst_release;
1151 if (!fl6->flowi6_mark && ndst)
1152 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1154 skb_dst_set(skb, dst);
1156 if (encap_limit >= 0) {
1157 init_tel_txopt(&opt, encap_limit);
1158 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1161 /* Calculate max headroom for all the headers and adjust
1162 * needed_headroom if necessary.
1164 max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1165 + dst->header_len + t->hlen;
1166 if (max_headroom > dev->needed_headroom)
1167 dev->needed_headroom = max_headroom;
1169 err = ip6_tnl_encap(skb, t, &proto, fl6);
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 = 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);
1191 EXPORT_SYMBOL(ip6_tnl_xmit);
1194 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1196 struct ip6_tnl *t = netdev_priv(dev);
1197 const struct iphdr *iph = ip_hdr(skb);
1198 int encap_limit = -1;
1205 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1207 tproto = ACCESS_ONCE(t->parms.proto);
1208 if (tproto != IPPROTO_IPIP && tproto != 0)
1211 dsfield = ipv4_get_dsfield(iph);
1213 if (t->parms.collect_md) {
1214 struct ip_tunnel_info *tun_info;
1215 const struct ip_tunnel_key *key;
1217 tun_info = skb_tunnel_info(skb);
1218 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1219 ip_tunnel_info_af(tun_info) != AF_INET6))
1221 key = &tun_info->key;
1222 memset(&fl6, 0, sizeof(fl6));
1223 fl6.flowi6_proto = IPPROTO_IPIP;
1224 fl6.daddr = key->u.ipv6.dst;
1225 fl6.flowlabel = key->label;
1227 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1228 encap_limit = t->parms.encap_limit;
1230 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1231 fl6.flowi6_proto = IPPROTO_IPIP;
1233 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1234 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1236 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1237 fl6.flowi6_mark = skb->mark;
1240 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1243 skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1245 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1248 /* XXX: send ICMP error even if DF is not set. */
1249 if (err == -EMSGSIZE)
1250 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1259 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1261 struct ip6_tnl *t = netdev_priv(dev);
1262 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1263 int encap_limit = -1;
1271 tproto = ACCESS_ONCE(t->parms.proto);
1272 if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1273 ip6_tnl_addr_conflict(t, ipv6h))
1276 dsfield = ipv6_get_dsfield(ipv6h);
1278 if (t->parms.collect_md) {
1279 struct ip_tunnel_info *tun_info;
1280 const struct ip_tunnel_key *key;
1282 tun_info = skb_tunnel_info(skb);
1283 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1284 ip_tunnel_info_af(tun_info) != AF_INET6))
1286 key = &tun_info->key;
1287 memset(&fl6, 0, sizeof(fl6));
1288 fl6.flowi6_proto = IPPROTO_IPV6;
1289 fl6.daddr = key->u.ipv6.dst;
1290 fl6.flowlabel = key->label;
1292 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1294 struct ipv6_tlv_tnl_enc_lim *tel;
1296 tel = (void *)&skb_network_header(skb)[offset];
1297 if (tel->encap_limit == 0) {
1298 icmpv6_send(skb, ICMPV6_PARAMPROB,
1299 ICMPV6_HDR_FIELD, offset + 2);
1302 encap_limit = tel->encap_limit - 1;
1303 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1304 encap_limit = t->parms.encap_limit;
1307 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1308 fl6.flowi6_proto = IPPROTO_IPV6;
1310 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1311 fl6.flowlabel |= (*(__be32 *)ipv6h & IPV6_TCLASS_MASK);
1312 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1313 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1314 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1315 fl6.flowi6_mark = skb->mark;
1318 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1321 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1323 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1326 if (err == -EMSGSIZE)
1327 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1335 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1337 struct ip6_tnl *t = netdev_priv(dev);
1338 struct net_device_stats *stats = &t->dev->stats;
1341 switch (skb->protocol) {
1342 case htons(ETH_P_IP):
1343 ret = ip4ip6_tnl_xmit(skb, dev);
1345 case htons(ETH_P_IPV6):
1346 ret = ip6ip6_tnl_xmit(skb, dev);
1355 return NETDEV_TX_OK;
1359 stats->tx_dropped++;
1361 return NETDEV_TX_OK;
1364 static void ip6_tnl_link_config(struct ip6_tnl *t)
1366 struct net_device *dev = t->dev;
1367 struct __ip6_tnl_parm *p = &t->parms;
1368 struct flowi6 *fl6 = &t->fl.u.ip6;
1371 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1372 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1374 /* Set up flowi template */
1375 fl6->saddr = p->laddr;
1376 fl6->daddr = p->raddr;
1377 fl6->flowi6_oif = p->link;
1380 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1381 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1382 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1383 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1385 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1386 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1388 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1389 dev->flags |= IFF_POINTOPOINT;
1391 dev->flags &= ~IFF_POINTOPOINT;
1394 t->hlen = t->encap_hlen + t->tun_hlen;
1395 t_hlen = t->hlen + sizeof(struct ipv6hdr);
1397 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1398 int strict = (ipv6_addr_type(&p->raddr) &
1399 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1401 struct rt6_info *rt = rt6_lookup(t->net,
1402 &p->raddr, &p->laddr,
1409 dev->hard_header_len = rt->dst.dev->hard_header_len +
1412 dev->mtu = rt->dst.dev->mtu - t_hlen;
1413 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1416 if (dev->mtu < IPV6_MIN_MTU)
1417 dev->mtu = IPV6_MIN_MTU;
1424 * ip6_tnl_change - update the tunnel parameters
1425 * @t: tunnel to be changed
1426 * @p: tunnel configuration parameters
1429 * ip6_tnl_change() updates the tunnel parameters
1433 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1435 t->parms.laddr = p->laddr;
1436 t->parms.raddr = p->raddr;
1437 t->parms.flags = p->flags;
1438 t->parms.hop_limit = p->hop_limit;
1439 t->parms.encap_limit = p->encap_limit;
1440 t->parms.flowinfo = p->flowinfo;
1441 t->parms.link = p->link;
1442 t->parms.proto = p->proto;
1443 dst_cache_reset(&t->dst_cache);
1444 ip6_tnl_link_config(t);
1448 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1450 struct net *net = t->net;
1451 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1454 ip6_tnl_unlink(ip6n, t);
1456 err = ip6_tnl_change(t, p);
1457 ip6_tnl_link(ip6n, t);
1458 netdev_state_change(t->dev);
1462 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1464 /* for default tnl0 device allow to change only the proto */
1465 t->parms.proto = p->proto;
1466 netdev_state_change(t->dev);
1471 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1473 p->laddr = u->laddr;
1474 p->raddr = u->raddr;
1475 p->flags = u->flags;
1476 p->hop_limit = u->hop_limit;
1477 p->encap_limit = u->encap_limit;
1478 p->flowinfo = u->flowinfo;
1480 p->proto = u->proto;
1481 memcpy(p->name, u->name, sizeof(u->name));
1485 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1487 u->laddr = p->laddr;
1488 u->raddr = p->raddr;
1489 u->flags = p->flags;
1490 u->hop_limit = p->hop_limit;
1491 u->encap_limit = p->encap_limit;
1492 u->flowinfo = p->flowinfo;
1494 u->proto = p->proto;
1495 memcpy(u->name, p->name, sizeof(u->name));
1499 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1500 * @dev: virtual device associated with tunnel
1501 * @ifr: parameters passed from userspace
1502 * @cmd: command to be performed
1505 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1508 * The possible commands are the following:
1509 * %SIOCGETTUNNEL: get tunnel parameters for device
1510 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1511 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1512 * %SIOCDELTUNNEL: delete tunnel
1514 * The fallback device "ip6tnl0", created during module
1515 * initialization, can be used for creating other tunnel devices.
1519 * %-EFAULT if unable to copy data to or from userspace,
1520 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1521 * %-EINVAL if passed tunnel parameters are invalid,
1522 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1523 * %-ENODEV if attempting to change or delete a nonexisting device
1527 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1530 struct ip6_tnl_parm p;
1531 struct __ip6_tnl_parm p1;
1532 struct ip6_tnl *t = netdev_priv(dev);
1533 struct net *net = t->net;
1534 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1536 memset(&p1, 0, sizeof(p1));
1540 if (dev == ip6n->fb_tnl_dev) {
1541 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1545 ip6_tnl_parm_from_user(&p1, &p);
1546 t = ip6_tnl_locate(net, &p1, 0);
1548 t = netdev_priv(dev);
1550 memset(&p, 0, sizeof(p));
1552 ip6_tnl_parm_to_user(&p, &t->parms);
1553 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1560 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1563 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1566 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1569 ip6_tnl_parm_from_user(&p1, &p);
1570 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1571 if (cmd == SIOCCHGTUNNEL) {
1573 if (t->dev != dev) {
1578 t = netdev_priv(dev);
1579 if (dev == ip6n->fb_tnl_dev)
1580 err = ip6_tnl0_update(t, &p1);
1582 err = ip6_tnl_update(t, &p1);
1586 ip6_tnl_parm_to_user(&p, &t->parms);
1587 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1596 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1599 if (dev == ip6n->fb_tnl_dev) {
1601 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1604 ip6_tnl_parm_from_user(&p1, &p);
1605 t = ip6_tnl_locate(net, &p1, 0);
1609 if (t->dev == ip6n->fb_tnl_dev)
1614 unregister_netdevice(dev);
1623 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1624 * @dev: virtual device associated with tunnel
1625 * @new_mtu: the new mtu
1629 * %-EINVAL if mtu too small
1632 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1634 struct ip6_tnl *tnl = netdev_priv(dev);
1636 if (tnl->parms.proto == IPPROTO_IPIP) {
1640 if (new_mtu < IPV6_MIN_MTU)
1643 if (new_mtu > 0xFFF8 - dev->hard_header_len)
1648 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1650 int ip6_tnl_get_iflink(const struct net_device *dev)
1652 struct ip6_tnl *t = netdev_priv(dev);
1654 return t->parms.link;
1656 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1658 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1661 if (num >= MAX_IPTUN_ENCAP_OPS)
1664 return !cmpxchg((const struct ip6_tnl_encap_ops **)
1665 &ip6tun_encaps[num],
1666 NULL, ops) ? 0 : -1;
1668 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1670 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1675 if (num >= MAX_IPTUN_ENCAP_OPS)
1678 ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1679 &ip6tun_encaps[num],
1680 ops, NULL) == ops) ? 0 : -1;
1686 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1688 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1689 struct ip_tunnel_encap *ipencap)
1693 memset(&t->encap, 0, sizeof(t->encap));
1695 hlen = ip6_encap_hlen(ipencap);
1699 t->encap.type = ipencap->type;
1700 t->encap.sport = ipencap->sport;
1701 t->encap.dport = ipencap->dport;
1702 t->encap.flags = ipencap->flags;
1704 t->encap_hlen = hlen;
1705 t->hlen = t->encap_hlen + t->tun_hlen;
1709 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1711 static const struct net_device_ops ip6_tnl_netdev_ops = {
1712 .ndo_init = ip6_tnl_dev_init,
1713 .ndo_uninit = ip6_tnl_dev_uninit,
1714 .ndo_start_xmit = ip6_tnl_start_xmit,
1715 .ndo_do_ioctl = ip6_tnl_ioctl,
1716 .ndo_change_mtu = ip6_tnl_change_mtu,
1717 .ndo_get_stats = ip6_get_stats,
1718 .ndo_get_iflink = ip6_tnl_get_iflink,
1721 #define IPXIPX_FEATURES (NETIF_F_SG | \
1722 NETIF_F_FRAGLIST | \
1724 NETIF_F_GSO_SOFTWARE | \
1728 * ip6_tnl_dev_setup - setup virtual tunnel device
1729 * @dev: virtual device associated with tunnel
1732 * Initialize function pointers and device parameters
1735 static void ip6_tnl_dev_setup(struct net_device *dev)
1737 dev->netdev_ops = &ip6_tnl_netdev_ops;
1738 dev->destructor = ip6_dev_free;
1740 dev->type = ARPHRD_TUNNEL6;
1741 dev->flags |= IFF_NOARP;
1742 dev->addr_len = sizeof(struct in6_addr);
1743 dev->features |= NETIF_F_LLTX;
1744 netif_keep_dst(dev);
1746 dev->features |= IPXIPX_FEATURES;
1747 dev->hw_features |= IPXIPX_FEATURES;
1749 /* This perm addr will be used as interface identifier by IPv6 */
1750 dev->addr_assign_type = NET_ADDR_RANDOM;
1751 eth_random_addr(dev->perm_addr);
1756 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1757 * @dev: virtual device associated with tunnel
1761 ip6_tnl_dev_init_gen(struct net_device *dev)
1763 struct ip6_tnl *t = netdev_priv(dev);
1768 t->net = dev_net(dev);
1769 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1773 ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1777 ret = gro_cells_init(&t->gro_cells, dev);
1782 t->hlen = t->encap_hlen + t->tun_hlen;
1783 t_hlen = t->hlen + sizeof(struct ipv6hdr);
1785 dev->type = ARPHRD_TUNNEL6;
1786 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1787 dev->mtu = ETH_DATA_LEN - t_hlen;
1788 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1794 dst_cache_destroy(&t->dst_cache);
1796 free_percpu(dev->tstats);
1803 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1804 * @dev: virtual device associated with tunnel
1807 static int ip6_tnl_dev_init(struct net_device *dev)
1809 struct ip6_tnl *t = netdev_priv(dev);
1810 int err = ip6_tnl_dev_init_gen(dev);
1814 ip6_tnl_link_config(t);
1815 if (t->parms.collect_md) {
1816 dev->features |= NETIF_F_NETNS_LOCAL;
1817 netif_keep_dst(dev);
1823 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1824 * @dev: fallback device
1829 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1831 struct ip6_tnl *t = netdev_priv(dev);
1832 struct net *net = dev_net(dev);
1833 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1835 t->parms.proto = IPPROTO_IPV6;
1838 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1842 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1846 if (!data || !data[IFLA_IPTUN_PROTO])
1849 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1850 if (proto != IPPROTO_IPV6 &&
1851 proto != IPPROTO_IPIP &&
1858 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1859 struct __ip6_tnl_parm *parms)
1861 memset(parms, 0, sizeof(*parms));
1866 if (data[IFLA_IPTUN_LINK])
1867 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1869 if (data[IFLA_IPTUN_LOCAL])
1870 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1872 if (data[IFLA_IPTUN_REMOTE])
1873 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1875 if (data[IFLA_IPTUN_TTL])
1876 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1878 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1879 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1881 if (data[IFLA_IPTUN_FLOWINFO])
1882 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1884 if (data[IFLA_IPTUN_FLAGS])
1885 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1887 if (data[IFLA_IPTUN_PROTO])
1888 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1890 if (data[IFLA_IPTUN_COLLECT_METADATA])
1891 parms->collect_md = true;
1894 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1895 struct ip_tunnel_encap *ipencap)
1899 memset(ipencap, 0, sizeof(*ipencap));
1904 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1906 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1909 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1911 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1914 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1916 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1919 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1921 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1927 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1928 struct nlattr *tb[], struct nlattr *data[])
1930 struct net *net = dev_net(dev);
1931 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1932 struct ip6_tnl *nt, *t;
1933 struct ip_tunnel_encap ipencap;
1935 nt = netdev_priv(dev);
1937 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1938 int err = ip6_tnl_encap_setup(nt, &ipencap);
1944 ip6_tnl_netlink_parms(data, &nt->parms);
1946 if (nt->parms.collect_md) {
1947 if (rtnl_dereference(ip6n->collect_md_tun))
1950 t = ip6_tnl_locate(net, &nt->parms, 0);
1955 return ip6_tnl_create2(dev);
1958 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1959 struct nlattr *data[])
1961 struct ip6_tnl *t = netdev_priv(dev);
1962 struct __ip6_tnl_parm p;
1963 struct net *net = t->net;
1964 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1965 struct ip_tunnel_encap ipencap;
1967 if (dev == ip6n->fb_tnl_dev)
1970 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1971 int err = ip6_tnl_encap_setup(t, &ipencap);
1976 ip6_tnl_netlink_parms(data, &p);
1980 t = ip6_tnl_locate(net, &p, 0);
1985 t = netdev_priv(dev);
1987 return ip6_tnl_update(t, &p);
1990 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1992 struct net *net = dev_net(dev);
1993 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1995 if (dev != ip6n->fb_tnl_dev)
1996 unregister_netdevice_queue(dev, head);
1999 static size_t ip6_tnl_get_size(const struct net_device *dev)
2002 /* IFLA_IPTUN_LINK */
2004 /* IFLA_IPTUN_LOCAL */
2005 nla_total_size(sizeof(struct in6_addr)) +
2006 /* IFLA_IPTUN_REMOTE */
2007 nla_total_size(sizeof(struct in6_addr)) +
2008 /* IFLA_IPTUN_TTL */
2010 /* IFLA_IPTUN_ENCAP_LIMIT */
2012 /* IFLA_IPTUN_FLOWINFO */
2014 /* IFLA_IPTUN_FLAGS */
2016 /* IFLA_IPTUN_PROTO */
2018 /* IFLA_IPTUN_ENCAP_TYPE */
2020 /* IFLA_IPTUN_ENCAP_FLAGS */
2022 /* IFLA_IPTUN_ENCAP_SPORT */
2024 /* IFLA_IPTUN_ENCAP_DPORT */
2026 /* IFLA_IPTUN_COLLECT_METADATA */
2031 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2033 struct ip6_tnl *tunnel = netdev_priv(dev);
2034 struct __ip6_tnl_parm *parm = &tunnel->parms;
2036 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2037 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2038 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2039 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2040 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2041 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2042 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2043 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
2044 goto nla_put_failure;
2046 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2047 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2048 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2049 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2050 goto nla_put_failure;
2052 if (parm->collect_md)
2053 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2054 goto nla_put_failure;
2061 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2063 struct ip6_tnl *tunnel = netdev_priv(dev);
2067 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2069 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2070 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
2071 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
2072 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
2073 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
2074 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
2075 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
2076 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
2077 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
2078 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
2079 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
2080 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
2081 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
2082 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG },
2085 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2087 .maxtype = IFLA_IPTUN_MAX,
2088 .policy = ip6_tnl_policy,
2089 .priv_size = sizeof(struct ip6_tnl),
2090 .setup = ip6_tnl_dev_setup,
2091 .validate = ip6_tnl_validate,
2092 .newlink = ip6_tnl_newlink,
2093 .changelink = ip6_tnl_changelink,
2094 .dellink = ip6_tnl_dellink,
2095 .get_size = ip6_tnl_get_size,
2096 .fill_info = ip6_tnl_fill_info,
2097 .get_link_net = ip6_tnl_get_link_net,
2100 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2101 .handler = ip4ip6_rcv,
2102 .err_handler = ip4ip6_err,
2106 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2107 .handler = ip6ip6_rcv,
2108 .err_handler = ip6ip6_err,
2112 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
2114 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2115 struct net_device *dev, *aux;
2120 for_each_netdev_safe(net, dev, aux)
2121 if (dev->rtnl_link_ops == &ip6_link_ops)
2122 unregister_netdevice_queue(dev, &list);
2124 for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2125 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2127 /* If dev is in the same netns, it has already
2128 * been added to the list by the previous loop.
2130 if (!net_eq(dev_net(t->dev), net))
2131 unregister_netdevice_queue(t->dev, &list);
2132 t = rtnl_dereference(t->next);
2136 unregister_netdevice_many(&list);
2139 static int __net_init ip6_tnl_init_net(struct net *net)
2141 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2142 struct ip6_tnl *t = NULL;
2145 ip6n->tnls[0] = ip6n->tnls_wc;
2146 ip6n->tnls[1] = ip6n->tnls_r_l;
2149 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2150 NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2152 if (!ip6n->fb_tnl_dev)
2154 dev_net_set(ip6n->fb_tnl_dev, net);
2155 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2156 /* FB netdevice is special: we have one, and only one per netns.
2157 * Allowing to move it to another netns is clearly unsafe.
2159 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2161 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2165 err = register_netdev(ip6n->fb_tnl_dev);
2169 t = netdev_priv(ip6n->fb_tnl_dev);
2171 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2175 ip6_dev_free(ip6n->fb_tnl_dev);
2180 static void __net_exit ip6_tnl_exit_net(struct net *net)
2183 ip6_tnl_destroy_tunnels(net);
2187 static struct pernet_operations ip6_tnl_net_ops = {
2188 .init = ip6_tnl_init_net,
2189 .exit = ip6_tnl_exit_net,
2190 .id = &ip6_tnl_net_id,
2191 .size = sizeof(struct ip6_tnl_net),
2195 * ip6_tunnel_init - register protocol and reserve needed resources
2197 * Return: 0 on success
2200 static int __init ip6_tunnel_init(void)
2204 err = register_pernet_device(&ip6_tnl_net_ops);
2208 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2210 pr_err("%s: can't register ip4ip6\n", __func__);
2214 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2216 pr_err("%s: can't register ip6ip6\n", __func__);
2219 err = rtnl_link_register(&ip6_link_ops);
2221 goto rtnl_link_failed;
2226 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2228 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2230 unregister_pernet_device(&ip6_tnl_net_ops);
2236 * ip6_tunnel_cleanup - free resources and unregister protocol
2239 static void __exit ip6_tunnel_cleanup(void)
2241 rtnl_link_unregister(&ip6_link_ops);
2242 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2243 pr_info("%s: can't deregister ip4ip6\n", __func__);
2245 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2246 pr_info("%s: can't deregister ip6ip6\n", __func__);
2248 unregister_pernet_device(&ip6_tnl_net_ops);
2251 module_init(ip6_tunnel_init);
2252 module_exit(ip6_tunnel_cleanup);