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 <linux/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 unsigned 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 ipv6_addr_any(&t->parms.raddr) &&
161 (t->dev->flags & IFF_UP))
165 hash = HASH(remote, &any);
166 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
167 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
168 ipv6_addr_any(&t->parms.laddr) &&
169 (t->dev->flags & IFF_UP))
173 t = rcu_dereference(ip6n->collect_md_tun);
174 if (t && t->dev->flags & IFF_UP)
177 t = rcu_dereference(ip6n->tnls_wc[0]);
178 if (t && (t->dev->flags & IFF_UP))
185 * ip6_tnl_bucket - get head of list matching given tunnel parameters
186 * @p: parameters containing tunnel end-points
189 * ip6_tnl_bucket() returns the head of the list matching the
190 * &struct in6_addr entries laddr and raddr in @p.
192 * Return: head of IPv6 tunnel list
195 static struct ip6_tnl __rcu **
196 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
198 const struct in6_addr *remote = &p->raddr;
199 const struct in6_addr *local = &p->laddr;
203 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
205 h = HASH(remote, local);
207 return &ip6n->tnls[prio][h];
211 * ip6_tnl_link - add tunnel to hash table
212 * @t: tunnel to be added
216 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
218 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
220 if (t->parms.collect_md)
221 rcu_assign_pointer(ip6n->collect_md_tun, t);
222 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
223 rcu_assign_pointer(*tp, t);
227 * ip6_tnl_unlink - remove tunnel from hash table
228 * @t: tunnel to be removed
232 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
234 struct ip6_tnl __rcu **tp;
235 struct ip6_tnl *iter;
237 if (t->parms.collect_md)
238 rcu_assign_pointer(ip6n->collect_md_tun, NULL);
240 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
241 (iter = rtnl_dereference(*tp)) != NULL;
244 rcu_assign_pointer(*tp, t->next);
250 static void ip6_dev_free(struct net_device *dev)
252 struct ip6_tnl *t = netdev_priv(dev);
254 gro_cells_destroy(&t->gro_cells);
255 dst_cache_destroy(&t->dst_cache);
256 free_percpu(dev->tstats);
259 static int ip6_tnl_create2(struct net_device *dev)
261 struct ip6_tnl *t = netdev_priv(dev);
262 struct net *net = dev_net(dev);
263 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
266 t = netdev_priv(dev);
268 dev->rtnl_link_ops = &ip6_link_ops;
269 err = register_netdevice(dev);
273 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 if (!dev_valid_name(p->name))
304 strlcpy(name, p->name, IFNAMSIZ);
306 sprintf(name, "ip6tnl%%d");
309 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
314 dev_net_set(dev, net);
316 t = netdev_priv(dev);
318 t->net = dev_net(dev);
319 err = ip6_tnl_create2(dev);
332 * ip6_tnl_locate - find or create tunnel matching given parameters
333 * @p: tunnel parameters
334 * @create: != 0 if allowed to create new tunnel if no match found
337 * ip6_tnl_locate() first tries to locate an existing tunnel
338 * based on @parms. If this is unsuccessful, but @create is set a new
339 * tunnel device is created and registered for use.
342 * matching tunnel or error pointer
345 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
346 struct __ip6_tnl_parm *p, int create)
348 const struct in6_addr *remote = &p->raddr;
349 const struct in6_addr *local = &p->laddr;
350 struct ip6_tnl __rcu **tp;
352 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
354 for (tp = ip6_tnl_bucket(ip6n, p);
355 (t = rtnl_dereference(*tp)) != NULL;
357 if (ipv6_addr_equal(local, &t->parms.laddr) &&
358 ipv6_addr_equal(remote, &t->parms.raddr)) {
360 return ERR_PTR(-EEXIST);
366 return ERR_PTR(-ENODEV);
367 return ip6_tnl_create(net, p);
371 * ip6_tnl_dev_uninit - tunnel device uninitializer
372 * @dev: the device to be destroyed
375 * ip6_tnl_dev_uninit() removes tunnel from its list
379 ip6_tnl_dev_uninit(struct net_device *dev)
381 struct ip6_tnl *t = netdev_priv(dev);
382 struct net *net = t->net;
383 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
385 if (dev == ip6n->fb_tnl_dev)
386 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
388 ip6_tnl_unlink(ip6n, t);
389 dst_cache_reset(&t->dst_cache);
394 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
395 * @skb: received socket buffer
398 * 0 if none was found,
399 * else index to encapsulation limit
402 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
404 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
405 unsigned int nhoff = raw - skb->data;
406 unsigned int off = nhoff + sizeof(*ipv6h);
407 u8 next, nexthdr = ipv6h->nexthdr;
409 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
410 struct ipv6_opt_hdr *hdr;
413 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
416 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
417 if (nexthdr == NEXTHDR_FRAGMENT) {
418 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
419 if (frag_hdr->frag_off)
422 } else if (nexthdr == NEXTHDR_AUTH) {
423 optlen = (hdr->hdrlen + 2) << 2;
425 optlen = ipv6_optlen(hdr);
427 /* cache hdr->nexthdr, since pskb_may_pull() might
431 if (nexthdr == NEXTHDR_DEST) {
434 /* Remember : hdr is no longer valid at this point. */
435 if (!pskb_may_pull(skb, off + optlen))
439 struct ipv6_tlv_tnl_enc_lim *tel;
441 /* No more room for encapsulation limit */
442 if (i + sizeof(*tel) > optlen)
445 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
446 /* return index of option if found and valid */
447 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
449 return i + off - nhoff;
450 /* else jump to next option */
452 i += tel->length + 2;
462 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
465 * ip6_tnl_err - tunnel error handler
468 * ip6_tnl_err() should handle errors in the tunnel according
469 * to the specifications in RFC 2473.
473 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
474 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
476 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
479 u8 rel_type = ICMPV6_DEST_UNREACH;
480 u8 rel_code = ICMPV6_ADDR_UNREACH;
486 /* If the packet doesn't contain the original IPv6 header we are
487 in trouble since we might need the source address for further
488 processing of the error. */
491 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
495 tproto = ACCESS_ONCE(t->parms.proto);
496 if (tproto != ipproto && tproto != 0)
503 struct ipv6_tlv_tnl_enc_lim *tel;
505 case ICMPV6_DEST_UNREACH:
506 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
510 case ICMPV6_TIME_EXCEED:
511 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
512 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
517 case ICMPV6_PARAMPROB:
519 if ((*code) == ICMPV6_HDR_FIELD)
520 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
522 if (teli && teli == *info - 2) {
523 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
524 if (tel->encap_limit == 0) {
525 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
530 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
534 case ICMPV6_PKT_TOOBIG:
535 mtu = *info - offset;
536 if (mtu < IPV6_MIN_MTU)
540 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
542 rel_type = ICMPV6_PKT_TOOBIG;
561 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
562 u8 type, u8 code, int offset, __be32 info)
567 __u32 rel_info = ntohl(info);
569 struct sk_buff *skb2;
570 const struct iphdr *eiph;
574 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
575 &rel_msg, &rel_info, offset);
583 case ICMPV6_DEST_UNREACH:
584 if (rel_code != ICMPV6_ADDR_UNREACH)
586 rel_type = ICMP_DEST_UNREACH;
587 rel_code = ICMP_HOST_UNREACH;
589 case ICMPV6_PKT_TOOBIG:
592 rel_type = ICMP_DEST_UNREACH;
593 rel_code = ICMP_FRAG_NEEDED;
596 rel_type = ICMP_REDIRECT;
597 rel_code = ICMP_REDIR_HOST;
602 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
605 skb2 = skb_clone(skb, GFP_ATOMIC);
611 skb_pull(skb2, offset);
612 skb_reset_network_header(skb2);
615 /* Try to guess incoming interface */
616 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
619 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
623 skb2->dev = rt->dst.dev;
625 /* route "incoming" packet */
626 if (rt->rt_flags & RTCF_LOCAL) {
629 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
630 eiph->daddr, eiph->saddr,
633 RT_TOS(eiph->tos), 0);
635 rt->dst.dev->type != ARPHRD_TUNNEL6) {
640 skb_dst_set(skb2, &rt->dst);
643 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
645 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL6)
649 /* change mtu on this route */
650 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
651 if (rel_info > dst_mtu(skb_dst(skb2)))
654 skb_dst_update_pmtu_no_confirm(skb2, rel_info);
656 if (rel_type == ICMP_REDIRECT)
657 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
659 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
667 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
668 u8 type, u8 code, int offset, __be32 info)
673 __u32 rel_info = ntohl(info);
676 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
677 &rel_msg, &rel_info, offset);
681 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
683 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
689 skb_pull(skb2, offset);
690 skb_reset_network_header(skb2);
692 /* Try to guess incoming interface */
693 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
696 if (rt && rt->dst.dev)
697 skb2->dev = rt->dst.dev;
699 icmpv6_send(skb2, rel_type, rel_code, rel_info);
709 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
710 const struct ipv6hdr *ipv6h,
713 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
715 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
716 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
718 return IP6_ECN_decapsulate(ipv6h, skb);
721 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
722 const struct ipv6hdr *ipv6h,
725 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
726 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
728 return IP6_ECN_decapsulate(ipv6h, skb);
731 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
732 const struct in6_addr *laddr,
733 const struct in6_addr *raddr)
735 struct __ip6_tnl_parm *p = &t->parms;
736 int ltype = ipv6_addr_type(laddr);
737 int rtype = ipv6_addr_type(raddr);
740 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
741 flags = IP6_TNL_F_CAP_PER_PACKET;
742 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
743 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
744 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
745 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
746 if (ltype&IPV6_ADDR_UNICAST)
747 flags |= IP6_TNL_F_CAP_XMIT;
748 if (rtype&IPV6_ADDR_UNICAST)
749 flags |= IP6_TNL_F_CAP_RCV;
753 EXPORT_SYMBOL(ip6_tnl_get_cap);
755 /* called with rcu_read_lock() */
756 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
757 const struct in6_addr *laddr,
758 const struct in6_addr *raddr)
760 struct __ip6_tnl_parm *p = &t->parms;
762 struct net *net = t->net;
764 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
765 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
766 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
767 struct net_device *ldev = NULL;
770 ldev = dev_get_by_index_rcu(net, p->link);
772 if ((ipv6_addr_is_multicast(laddr) ||
773 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
774 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
779 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
781 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
782 const struct tnl_ptk_info *tpi,
783 struct metadata_dst *tun_dst,
784 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
785 const struct ipv6hdr *ipv6h,
786 struct sk_buff *skb),
789 struct pcpu_sw_netstats *tstats;
790 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
793 if ((!(tpi->flags & TUNNEL_CSUM) &&
794 (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
795 ((tpi->flags & TUNNEL_CSUM) &&
796 !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
797 tunnel->dev->stats.rx_crc_errors++;
798 tunnel->dev->stats.rx_errors++;
802 if (tunnel->parms.i_flags & TUNNEL_SEQ) {
803 if (!(tpi->flags & TUNNEL_SEQ) ||
805 (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
806 tunnel->dev->stats.rx_fifo_errors++;
807 tunnel->dev->stats.rx_errors++;
810 tunnel->i_seqno = ntohl(tpi->seq) + 1;
813 skb->protocol = tpi->proto;
815 /* Warning: All skb pointers will be invalidated! */
816 if (tunnel->dev->type == ARPHRD_ETHER) {
817 if (!pskb_may_pull(skb, ETH_HLEN)) {
818 tunnel->dev->stats.rx_length_errors++;
819 tunnel->dev->stats.rx_errors++;
823 ipv6h = ipv6_hdr(skb);
824 skb->protocol = eth_type_trans(skb, tunnel->dev);
825 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
827 skb->dev = tunnel->dev;
830 skb_reset_network_header(skb);
831 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
833 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
835 err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
838 net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
840 ipv6_get_dsfield(ipv6h));
842 ++tunnel->dev->stats.rx_frame_errors;
843 ++tunnel->dev->stats.rx_errors;
848 tstats = this_cpu_ptr(tunnel->dev->tstats);
849 u64_stats_update_begin(&tstats->syncp);
850 tstats->rx_packets++;
851 tstats->rx_bytes += skb->len;
852 u64_stats_update_end(&tstats->syncp);
854 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
857 skb_dst_set(skb, (struct dst_entry *)tun_dst);
859 gro_cells_receive(&tunnel->gro_cells, skb);
864 dst_release((struct dst_entry *)tun_dst);
869 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
870 const struct tnl_ptk_info *tpi,
871 struct metadata_dst *tun_dst,
874 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
875 const struct ipv6hdr *ipv6h,
876 struct sk_buff *skb);
878 dscp_ecn_decapsulate = ip6ip6_dscp_ecn_decapsulate;
879 if (tpi->proto == htons(ETH_P_IP))
880 dscp_ecn_decapsulate = ip4ip6_dscp_ecn_decapsulate;
882 return __ip6_tnl_rcv(t, skb, tpi, NULL, dscp_ecn_decapsulate,
885 EXPORT_SYMBOL(ip6_tnl_rcv);
887 static const struct tnl_ptk_info tpi_v6 = {
888 /* no tunnel info required for ipxip6. */
889 .proto = htons(ETH_P_IPV6),
892 static const struct tnl_ptk_info tpi_v4 = {
893 /* no tunnel info required for ipxip6. */
894 .proto = htons(ETH_P_IP),
897 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
898 const struct tnl_ptk_info *tpi,
899 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
900 const struct ipv6hdr *ipv6h,
901 struct sk_buff *skb))
904 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
905 struct metadata_dst *tun_dst = NULL;
909 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
912 u8 tproto = ACCESS_ONCE(t->parms.proto);
914 if (tproto != ipproto && tproto != 0)
916 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
918 ipv6h = ipv6_hdr(skb);
919 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
921 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
923 if (t->parms.collect_md) {
924 tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
928 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
942 static int ip4ip6_rcv(struct sk_buff *skb)
944 return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
945 ip4ip6_dscp_ecn_decapsulate);
948 static int ip6ip6_rcv(struct sk_buff *skb)
950 return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
951 ip6ip6_dscp_ecn_decapsulate);
954 struct ipv6_tel_txoption {
955 struct ipv6_txoptions ops;
959 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
961 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
963 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
965 opt->dst_opt[4] = encap_limit;
966 opt->dst_opt[5] = IPV6_TLV_PADN;
969 opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
970 opt->ops.opt_nflen = 8;
974 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
975 * @t: the outgoing tunnel device
976 * @hdr: IPv6 header from the incoming packet
979 * Avoid trivial tunneling loop by checking that tunnel exit-point
980 * doesn't match source of incoming packet.
988 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
990 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
993 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
994 const struct in6_addr *laddr,
995 const struct in6_addr *raddr)
997 struct __ip6_tnl_parm *p = &t->parms;
999 struct net *net = t->net;
1001 if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
1002 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
1003 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
1004 struct net_device *ldev = NULL;
1008 ldev = dev_get_by_index_rcu(net, p->link);
1010 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
1011 pr_warn_ratelimited("%s xmit: Local address not yet configured!\n",
1013 else if (!ipv6_addr_is_multicast(raddr) &&
1014 unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
1015 pr_warn_ratelimited("%s xmit: Routing loop! Remote address found on this node!\n",
1023 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1026 * ip6_tnl_xmit - encapsulate packet and send
1027 * @skb: the outgoing socket buffer
1028 * @dev: the outgoing tunnel device
1029 * @dsfield: dscp code for outer header
1030 * @fl6: flow of tunneled packet
1031 * @encap_limit: encapsulation limit
1032 * @pmtu: Path MTU is stored if packet is too big
1033 * @proto: next header value
1036 * Build new header and do some sanity checks on the packet before sending
1042 * %-EMSGSIZE message too big. return mtu in this case.
1045 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1046 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
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;
1053 struct ipv6_tel_txoption opt;
1054 struct dst_entry *dst = NULL, *ndst = NULL;
1055 struct net_device *tdev;
1057 unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1058 unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1059 unsigned int max_headroom = psh_hlen;
1060 bool use_cache = false;
1064 if (t->parms.collect_md) {
1065 hop_limit = skb_tunnel_info(skb)->key.ttl;
1068 hop_limit = t->parms.hop_limit;
1072 if (ipv6_addr_any(&t->parms.raddr)) {
1073 if (skb->protocol == htons(ETH_P_IPV6)) {
1074 struct in6_addr *addr6;
1075 struct neighbour *neigh;
1079 goto tx_err_link_failure;
1081 neigh = dst_neigh_lookup(skb_dst(skb),
1082 &ipv6_hdr(skb)->daddr);
1084 goto tx_err_link_failure;
1086 addr6 = (struct in6_addr *)&neigh->primary_key;
1087 addr_type = ipv6_addr_type(addr6);
1089 if (addr_type == IPV6_ADDR_ANY)
1090 addr6 = &ipv6_hdr(skb)->daddr;
1092 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1093 neigh_release(neigh);
1095 } else if (t->parms.proto != 0 && !(t->parms.flags &
1096 (IP6_TNL_F_USE_ORIG_TCLASS |
1097 IP6_TNL_F_USE_ORIG_FWMARK))) {
1098 /* enable the cache only if neither the outer protocol nor the
1099 * routing decision depends on the current inner header value
1105 dst = dst_cache_get(&t->dst_cache);
1107 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1108 goto tx_err_link_failure;
1112 /* add dsfield to flowlabel for route lookup */
1113 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1115 dst = ip6_route_output(net, NULL, fl6);
1118 goto tx_err_link_failure;
1119 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1123 goto tx_err_link_failure;
1125 if (t->parms.collect_md &&
1126 ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1127 &fl6->daddr, 0, &fl6->saddr))
1128 goto tx_err_link_failure;
1135 stats->collisions++;
1136 net_warn_ratelimited("%s: Local routing loop detected!\n",
1138 goto tx_err_dst_release;
1140 mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1141 if (encap_limit >= 0) {
1145 mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1146 IPV6_MIN_MTU : IPV4_MIN_MTU);
1148 skb_dst_update_pmtu_no_confirm(skb, mtu);
1149 if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1152 goto tx_err_dst_release;
1155 if (t->err_count > 0) {
1156 if (time_before(jiffies,
1157 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1160 dst_link_failure(skb);
1166 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1169 * Okay, now see if we can stuff it in the buffer as-is.
1171 max_headroom += LL_RESERVED_SPACE(tdev);
1173 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1174 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1175 struct sk_buff *new_skb;
1177 new_skb = skb_realloc_headroom(skb, max_headroom);
1179 goto tx_err_dst_release;
1182 skb_set_owner_w(new_skb, skb->sk);
1187 if (t->parms.collect_md) {
1188 if (t->encap.type != TUNNEL_ENCAP_NONE)
1189 goto tx_err_dst_release;
1191 if (use_cache && ndst)
1192 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1194 skb_dst_set(skb, dst);
1196 hop_limit = hop_limit ? : ip6_dst_hoplimit(dst);
1198 /* Calculate max headroom for all the headers and adjust
1199 * needed_headroom if necessary.
1201 max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1202 + dst->header_len + t->hlen;
1203 if (max_headroom > dev->needed_headroom)
1204 dev->needed_headroom = max_headroom;
1206 err = ip6_tnl_encap(skb, t, &proto, fl6);
1210 if (encap_limit >= 0) {
1211 init_tel_txopt(&opt, encap_limit);
1212 ipv6_push_frag_opts(skb, &opt.ops, &proto);
1215 skb_push(skb, sizeof(struct ipv6hdr));
1216 skb_reset_network_header(skb);
1217 ipv6h = ipv6_hdr(skb);
1218 ip6_flow_hdr(ipv6h, dsfield,
1219 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1220 ipv6h->hop_limit = hop_limit;
1221 ipv6h->nexthdr = proto;
1222 ipv6h->saddr = fl6->saddr;
1223 ipv6h->daddr = fl6->daddr;
1224 ip6tunnel_xmit(NULL, skb, dev);
1226 tx_err_link_failure:
1227 stats->tx_carrier_errors++;
1228 dst_link_failure(skb);
1233 EXPORT_SYMBOL(ip6_tnl_xmit);
1236 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1238 struct ip6_tnl *t = netdev_priv(dev);
1239 const struct iphdr *iph;
1240 int encap_limit = -1;
1247 /* ensure we can access the full inner ip header */
1248 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1252 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1254 tproto = ACCESS_ONCE(t->parms.proto);
1255 if (tproto != IPPROTO_IPIP && tproto != 0)
1258 if (t->parms.collect_md) {
1259 struct ip_tunnel_info *tun_info;
1260 const struct ip_tunnel_key *key;
1262 tun_info = skb_tunnel_info(skb);
1263 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1264 ip_tunnel_info_af(tun_info) != AF_INET6))
1266 key = &tun_info->key;
1267 memset(&fl6, 0, sizeof(fl6));
1268 fl6.flowi6_proto = IPPROTO_IPIP;
1269 fl6.daddr = key->u.ipv6.dst;
1270 fl6.flowlabel = key->label;
1273 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1274 encap_limit = t->parms.encap_limit;
1276 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1277 fl6.flowi6_proto = IPPROTO_IPIP;
1279 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1280 dsfield = ipv4_get_dsfield(iph);
1282 dsfield = ip6_tclass(t->parms.flowinfo);
1283 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1284 fl6.flowi6_mark = skb->mark;
1286 fl6.flowi6_mark = t->parms.fwmark;
1289 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1290 dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1292 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1295 skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1297 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1300 /* XXX: send ICMP error even if DF is not set. */
1301 if (err == -EMSGSIZE)
1302 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1311 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1313 struct ip6_tnl *t = netdev_priv(dev);
1314 struct ipv6hdr *ipv6h;
1315 int encap_limit = -1;
1323 if (unlikely(!pskb_may_pull(skb, sizeof(*ipv6h))))
1326 ipv6h = ipv6_hdr(skb);
1327 tproto = ACCESS_ONCE(t->parms.proto);
1328 if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1329 ip6_tnl_addr_conflict(t, ipv6h))
1332 if (t->parms.collect_md) {
1333 struct ip_tunnel_info *tun_info;
1334 const struct ip_tunnel_key *key;
1336 tun_info = skb_tunnel_info(skb);
1337 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1338 ip_tunnel_info_af(tun_info) != AF_INET6))
1340 key = &tun_info->key;
1341 memset(&fl6, 0, sizeof(fl6));
1342 fl6.flowi6_proto = IPPROTO_IPV6;
1343 fl6.daddr = key->u.ipv6.dst;
1344 fl6.flowlabel = key->label;
1347 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1348 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1349 ipv6h = ipv6_hdr(skb);
1351 struct ipv6_tlv_tnl_enc_lim *tel;
1353 tel = (void *)&skb_network_header(skb)[offset];
1354 if (tel->encap_limit == 0) {
1355 icmpv6_send(skb, ICMPV6_PARAMPROB,
1356 ICMPV6_HDR_FIELD, offset + 2);
1359 encap_limit = tel->encap_limit - 1;
1360 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1361 encap_limit = t->parms.encap_limit;
1364 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1365 fl6.flowi6_proto = IPPROTO_IPV6;
1367 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1368 dsfield = ipv6_get_dsfield(ipv6h);
1370 dsfield = ip6_tclass(t->parms.flowinfo);
1371 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1372 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1373 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1374 fl6.flowi6_mark = skb->mark;
1376 fl6.flowi6_mark = t->parms.fwmark;
1379 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1380 dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1382 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1385 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1387 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1390 if (err == -EMSGSIZE)
1391 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1399 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1401 struct ip6_tnl *t = netdev_priv(dev);
1402 struct net_device_stats *stats = &t->dev->stats;
1405 switch (skb->protocol) {
1406 case htons(ETH_P_IP):
1407 ret = ip4ip6_tnl_xmit(skb, dev);
1409 case htons(ETH_P_IPV6):
1410 ret = ip6ip6_tnl_xmit(skb, dev);
1419 return NETDEV_TX_OK;
1423 stats->tx_dropped++;
1425 return NETDEV_TX_OK;
1428 static void ip6_tnl_link_config(struct ip6_tnl *t)
1430 struct net_device *dev = t->dev;
1431 struct __ip6_tnl_parm *p = &t->parms;
1432 struct flowi6 *fl6 = &t->fl.u.ip6;
1435 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1436 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1438 /* Set up flowi template */
1439 fl6->saddr = p->laddr;
1440 fl6->daddr = p->raddr;
1441 fl6->flowi6_oif = p->link;
1444 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1445 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1446 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1447 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1449 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1450 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1452 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1453 dev->flags |= IFF_POINTOPOINT;
1455 dev->flags &= ~IFF_POINTOPOINT;
1458 t->hlen = t->encap_hlen + t->tun_hlen;
1459 t_hlen = t->hlen + sizeof(struct ipv6hdr);
1461 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1462 int strict = (ipv6_addr_type(&p->raddr) &
1463 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1465 struct rt6_info *rt = rt6_lookup(t->net,
1466 &p->raddr, &p->laddr,
1473 dev->hard_header_len = rt->dst.dev->hard_header_len +
1476 dev->mtu = rt->dst.dev->mtu - t_hlen;
1477 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1480 if (dev->mtu < IPV6_MIN_MTU)
1481 dev->mtu = IPV6_MIN_MTU;
1488 * ip6_tnl_change - update the tunnel parameters
1489 * @t: tunnel to be changed
1490 * @p: tunnel configuration parameters
1493 * ip6_tnl_change() updates the tunnel parameters
1497 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1499 t->parms.laddr = p->laddr;
1500 t->parms.raddr = p->raddr;
1501 t->parms.flags = p->flags;
1502 t->parms.hop_limit = p->hop_limit;
1503 t->parms.encap_limit = p->encap_limit;
1504 t->parms.flowinfo = p->flowinfo;
1505 t->parms.link = p->link;
1506 t->parms.proto = p->proto;
1507 t->parms.fwmark = p->fwmark;
1508 dst_cache_reset(&t->dst_cache);
1509 ip6_tnl_link_config(t);
1513 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1515 struct net *net = t->net;
1516 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1519 ip6_tnl_unlink(ip6n, t);
1521 err = ip6_tnl_change(t, p);
1522 ip6_tnl_link(ip6n, t);
1523 netdev_state_change(t->dev);
1527 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1529 /* for default tnl0 device allow to change only the proto */
1530 t->parms.proto = p->proto;
1531 netdev_state_change(t->dev);
1536 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1538 p->laddr = u->laddr;
1539 p->raddr = u->raddr;
1540 p->flags = u->flags;
1541 p->hop_limit = u->hop_limit;
1542 p->encap_limit = u->encap_limit;
1543 p->flowinfo = u->flowinfo;
1545 p->proto = u->proto;
1546 memcpy(p->name, u->name, sizeof(u->name));
1550 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1552 u->laddr = p->laddr;
1553 u->raddr = p->raddr;
1554 u->flags = p->flags;
1555 u->hop_limit = p->hop_limit;
1556 u->encap_limit = p->encap_limit;
1557 u->flowinfo = p->flowinfo;
1559 u->proto = p->proto;
1560 memcpy(u->name, p->name, sizeof(u->name));
1564 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1565 * @dev: virtual device associated with tunnel
1566 * @ifr: parameters passed from userspace
1567 * @cmd: command to be performed
1570 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1573 * The possible commands are the following:
1574 * %SIOCGETTUNNEL: get tunnel parameters for device
1575 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1576 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1577 * %SIOCDELTUNNEL: delete tunnel
1579 * The fallback device "ip6tnl0", created during module
1580 * initialization, can be used for creating other tunnel devices.
1584 * %-EFAULT if unable to copy data to or from userspace,
1585 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1586 * %-EINVAL if passed tunnel parameters are invalid,
1587 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1588 * %-ENODEV if attempting to change or delete a nonexisting device
1592 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1595 struct ip6_tnl_parm p;
1596 struct __ip6_tnl_parm p1;
1597 struct ip6_tnl *t = netdev_priv(dev);
1598 struct net *net = t->net;
1599 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1601 memset(&p1, 0, sizeof(p1));
1605 if (dev == ip6n->fb_tnl_dev) {
1606 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1610 ip6_tnl_parm_from_user(&p1, &p);
1611 t = ip6_tnl_locate(net, &p1, 0);
1613 t = netdev_priv(dev);
1615 memset(&p, 0, sizeof(p));
1617 ip6_tnl_parm_to_user(&p, &t->parms);
1618 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1625 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1628 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1631 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1634 ip6_tnl_parm_from_user(&p1, &p);
1635 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1636 if (cmd == SIOCCHGTUNNEL) {
1638 if (t->dev != dev) {
1643 t = netdev_priv(dev);
1644 if (dev == ip6n->fb_tnl_dev)
1645 err = ip6_tnl0_update(t, &p1);
1647 err = ip6_tnl_update(t, &p1);
1651 ip6_tnl_parm_to_user(&p, &t->parms);
1652 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1661 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1664 if (dev == ip6n->fb_tnl_dev) {
1666 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1669 ip6_tnl_parm_from_user(&p1, &p);
1670 t = ip6_tnl_locate(net, &p1, 0);
1674 if (t->dev == ip6n->fb_tnl_dev)
1679 unregister_netdevice(dev);
1688 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1689 * @dev: virtual device associated with tunnel
1690 * @new_mtu: the new mtu
1694 * %-EINVAL if mtu too small
1697 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1699 struct ip6_tnl *tnl = netdev_priv(dev);
1701 if (tnl->parms.proto == IPPROTO_IPV6) {
1702 if (new_mtu < IPV6_MIN_MTU)
1705 if (new_mtu < ETH_MIN_MTU)
1708 if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1709 if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1712 if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1718 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1720 int ip6_tnl_get_iflink(const struct net_device *dev)
1722 struct ip6_tnl *t = netdev_priv(dev);
1724 return t->parms.link;
1726 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1728 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1731 if (num >= MAX_IPTUN_ENCAP_OPS)
1734 return !cmpxchg((const struct ip6_tnl_encap_ops **)
1735 &ip6tun_encaps[num],
1736 NULL, ops) ? 0 : -1;
1738 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1740 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1745 if (num >= MAX_IPTUN_ENCAP_OPS)
1748 ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1749 &ip6tun_encaps[num],
1750 ops, NULL) == ops) ? 0 : -1;
1756 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1758 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1759 struct ip_tunnel_encap *ipencap)
1763 memset(&t->encap, 0, sizeof(t->encap));
1765 hlen = ip6_encap_hlen(ipencap);
1769 t->encap.type = ipencap->type;
1770 t->encap.sport = ipencap->sport;
1771 t->encap.dport = ipencap->dport;
1772 t->encap.flags = ipencap->flags;
1774 t->encap_hlen = hlen;
1775 t->hlen = t->encap_hlen + t->tun_hlen;
1779 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1781 static const struct net_device_ops ip6_tnl_netdev_ops = {
1782 .ndo_init = ip6_tnl_dev_init,
1783 .ndo_uninit = ip6_tnl_dev_uninit,
1784 .ndo_start_xmit = ip6_tnl_start_xmit,
1785 .ndo_do_ioctl = ip6_tnl_ioctl,
1786 .ndo_change_mtu = ip6_tnl_change_mtu,
1787 .ndo_get_stats = ip6_get_stats,
1788 .ndo_get_iflink = ip6_tnl_get_iflink,
1791 #define IPXIPX_FEATURES (NETIF_F_SG | \
1792 NETIF_F_FRAGLIST | \
1794 NETIF_F_GSO_SOFTWARE | \
1798 * ip6_tnl_dev_setup - setup virtual tunnel device
1799 * @dev: virtual device associated with tunnel
1802 * Initialize function pointers and device parameters
1805 static void ip6_tnl_dev_setup(struct net_device *dev)
1807 dev->netdev_ops = &ip6_tnl_netdev_ops;
1808 dev->needs_free_netdev = true;
1809 dev->priv_destructor = ip6_dev_free;
1811 dev->type = ARPHRD_TUNNEL6;
1812 dev->flags |= IFF_NOARP;
1813 dev->addr_len = sizeof(struct in6_addr);
1814 dev->features |= NETIF_F_LLTX;
1815 netif_keep_dst(dev);
1817 dev->features |= IPXIPX_FEATURES;
1818 dev->hw_features |= IPXIPX_FEATURES;
1820 /* This perm addr will be used as interface identifier by IPv6 */
1821 dev->addr_assign_type = NET_ADDR_RANDOM;
1822 eth_random_addr(dev->perm_addr);
1827 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1828 * @dev: virtual device associated with tunnel
1832 ip6_tnl_dev_init_gen(struct net_device *dev)
1834 struct ip6_tnl *t = netdev_priv(dev);
1839 t->net = dev_net(dev);
1840 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1844 ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1848 ret = gro_cells_init(&t->gro_cells, dev);
1853 t->hlen = t->encap_hlen + t->tun_hlen;
1854 t_hlen = t->hlen + sizeof(struct ipv6hdr);
1856 dev->type = ARPHRD_TUNNEL6;
1857 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1858 dev->mtu = ETH_DATA_LEN - t_hlen;
1859 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1861 dev->min_mtu = ETH_MIN_MTU;
1862 dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1868 dst_cache_destroy(&t->dst_cache);
1870 free_percpu(dev->tstats);
1877 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1878 * @dev: virtual device associated with tunnel
1881 static int ip6_tnl_dev_init(struct net_device *dev)
1883 struct ip6_tnl *t = netdev_priv(dev);
1884 int err = ip6_tnl_dev_init_gen(dev);
1888 ip6_tnl_link_config(t);
1889 if (t->parms.collect_md)
1890 netif_keep_dst(dev);
1895 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1896 * @dev: fallback device
1901 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1903 struct ip6_tnl *t = netdev_priv(dev);
1904 struct net *net = dev_net(dev);
1905 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1907 t->parms.proto = IPPROTO_IPV6;
1909 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1913 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1914 struct netlink_ext_ack *extack)
1918 if (!data || !data[IFLA_IPTUN_PROTO])
1921 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1922 if (proto != IPPROTO_IPV6 &&
1923 proto != IPPROTO_IPIP &&
1930 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1931 struct __ip6_tnl_parm *parms)
1933 memset(parms, 0, sizeof(*parms));
1938 if (data[IFLA_IPTUN_LINK])
1939 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1941 if (data[IFLA_IPTUN_LOCAL])
1942 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1944 if (data[IFLA_IPTUN_REMOTE])
1945 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1947 if (data[IFLA_IPTUN_TTL])
1948 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1950 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1951 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1953 if (data[IFLA_IPTUN_FLOWINFO])
1954 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1956 if (data[IFLA_IPTUN_FLAGS])
1957 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1959 if (data[IFLA_IPTUN_PROTO])
1960 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1962 if (data[IFLA_IPTUN_COLLECT_METADATA])
1963 parms->collect_md = true;
1965 if (data[IFLA_IPTUN_FWMARK])
1966 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1969 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1970 struct ip_tunnel_encap *ipencap)
1974 memset(ipencap, 0, sizeof(*ipencap));
1979 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1981 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1984 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1986 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1989 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1991 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1994 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1996 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
2002 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
2003 struct nlattr *tb[], struct nlattr *data[],
2004 struct netlink_ext_ack *extack)
2006 struct net *net = dev_net(dev);
2007 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2008 struct ip_tunnel_encap ipencap;
2009 struct ip6_tnl *nt, *t;
2012 nt = netdev_priv(dev);
2014 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2015 err = ip6_tnl_encap_setup(nt, &ipencap);
2020 ip6_tnl_netlink_parms(data, &nt->parms);
2022 if (nt->parms.collect_md) {
2023 if (rtnl_dereference(ip6n->collect_md_tun))
2026 t = ip6_tnl_locate(net, &nt->parms, 0);
2031 err = ip6_tnl_create2(dev);
2032 if (!err && tb[IFLA_MTU])
2033 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2038 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2039 struct nlattr *data[],
2040 struct netlink_ext_ack *extack)
2042 struct ip6_tnl *t = netdev_priv(dev);
2043 struct __ip6_tnl_parm p;
2044 struct net *net = t->net;
2045 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2046 struct ip_tunnel_encap ipencap;
2048 if (dev == ip6n->fb_tnl_dev)
2051 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2052 int err = ip6_tnl_encap_setup(t, &ipencap);
2057 ip6_tnl_netlink_parms(data, &p);
2061 t = ip6_tnl_locate(net, &p, 0);
2066 t = netdev_priv(dev);
2068 return ip6_tnl_update(t, &p);
2071 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2073 struct net *net = dev_net(dev);
2074 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2076 if (dev != ip6n->fb_tnl_dev)
2077 unregister_netdevice_queue(dev, head);
2080 static size_t ip6_tnl_get_size(const struct net_device *dev)
2083 /* IFLA_IPTUN_LINK */
2085 /* IFLA_IPTUN_LOCAL */
2086 nla_total_size(sizeof(struct in6_addr)) +
2087 /* IFLA_IPTUN_REMOTE */
2088 nla_total_size(sizeof(struct in6_addr)) +
2089 /* IFLA_IPTUN_TTL */
2091 /* IFLA_IPTUN_ENCAP_LIMIT */
2093 /* IFLA_IPTUN_FLOWINFO */
2095 /* IFLA_IPTUN_FLAGS */
2097 /* IFLA_IPTUN_PROTO */
2099 /* IFLA_IPTUN_ENCAP_TYPE */
2101 /* IFLA_IPTUN_ENCAP_FLAGS */
2103 /* IFLA_IPTUN_ENCAP_SPORT */
2105 /* IFLA_IPTUN_ENCAP_DPORT */
2107 /* IFLA_IPTUN_COLLECT_METADATA */
2109 /* IFLA_IPTUN_FWMARK */
2114 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2116 struct ip6_tnl *tunnel = netdev_priv(dev);
2117 struct __ip6_tnl_parm *parm = &tunnel->parms;
2119 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2120 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2121 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2122 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2123 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2124 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2125 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2126 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2127 nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2128 goto nla_put_failure;
2130 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2131 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2132 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2133 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2134 goto nla_put_failure;
2136 if (parm->collect_md)
2137 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2138 goto nla_put_failure;
2146 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2148 struct ip6_tnl *tunnel = netdev_priv(dev);
2152 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2154 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2155 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
2156 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
2157 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
2158 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
2159 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
2160 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
2161 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
2162 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
2163 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
2164 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
2165 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
2166 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
2167 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG },
2168 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 },
2171 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2173 .maxtype = IFLA_IPTUN_MAX,
2174 .policy = ip6_tnl_policy,
2175 .priv_size = sizeof(struct ip6_tnl),
2176 .setup = ip6_tnl_dev_setup,
2177 .validate = ip6_tnl_validate,
2178 .newlink = ip6_tnl_newlink,
2179 .changelink = ip6_tnl_changelink,
2180 .dellink = ip6_tnl_dellink,
2181 .get_size = ip6_tnl_get_size,
2182 .fill_info = ip6_tnl_fill_info,
2183 .get_link_net = ip6_tnl_get_link_net,
2186 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2187 .handler = ip4ip6_rcv,
2188 .err_handler = ip4ip6_err,
2192 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2193 .handler = ip6ip6_rcv,
2194 .err_handler = ip6ip6_err,
2198 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
2200 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2201 struct net_device *dev, *aux;
2206 for_each_netdev_safe(net, dev, aux)
2207 if (dev->rtnl_link_ops == &ip6_link_ops)
2208 unregister_netdevice_queue(dev, &list);
2210 for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2211 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2213 /* If dev is in the same netns, it has already
2214 * been added to the list by the previous loop.
2216 if (!net_eq(dev_net(t->dev), net))
2217 unregister_netdevice_queue(t->dev, &list);
2218 t = rtnl_dereference(t->next);
2222 unregister_netdevice_many(&list);
2225 static int __net_init ip6_tnl_init_net(struct net *net)
2227 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2228 struct ip6_tnl *t = NULL;
2231 ip6n->tnls[0] = ip6n->tnls_wc;
2232 ip6n->tnls[1] = ip6n->tnls_r_l;
2235 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2236 NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2238 if (!ip6n->fb_tnl_dev)
2240 dev_net_set(ip6n->fb_tnl_dev, net);
2241 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2242 /* FB netdevice is special: we have one, and only one per netns.
2243 * Allowing to move it to another netns is clearly unsafe.
2245 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2247 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2251 err = register_netdev(ip6n->fb_tnl_dev);
2255 t = netdev_priv(ip6n->fb_tnl_dev);
2257 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2261 free_netdev(ip6n->fb_tnl_dev);
2266 static void __net_exit ip6_tnl_exit_net(struct net *net)
2269 ip6_tnl_destroy_tunnels(net);
2273 static struct pernet_operations ip6_tnl_net_ops = {
2274 .init = ip6_tnl_init_net,
2275 .exit = ip6_tnl_exit_net,
2276 .id = &ip6_tnl_net_id,
2277 .size = sizeof(struct ip6_tnl_net),
2281 * ip6_tunnel_init - register protocol and reserve needed resources
2283 * Return: 0 on success
2286 static int __init ip6_tunnel_init(void)
2290 if (!ipv6_mod_enabled())
2293 err = register_pernet_device(&ip6_tnl_net_ops);
2297 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2299 pr_err("%s: can't register ip4ip6\n", __func__);
2303 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2305 pr_err("%s: can't register ip6ip6\n", __func__);
2308 err = rtnl_link_register(&ip6_link_ops);
2310 goto rtnl_link_failed;
2315 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2317 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2319 unregister_pernet_device(&ip6_tnl_net_ops);
2325 * ip6_tunnel_cleanup - free resources and unregister protocol
2328 static void __exit ip6_tunnel_cleanup(void)
2330 rtnl_link_unregister(&ip6_link_ops);
2331 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2332 pr_info("%s: can't deregister ip4ip6\n", __func__);
2334 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2335 pr_info("%s: can't deregister ip6ip6\n", __func__);
2337 unregister_pernet_device(&ip6_tnl_net_ops);
2340 module_init(ip6_tunnel_init);
2341 module_exit(ip6_tunnel_cleanup);