2 * IPv6 tunneling device
3 * Linux INET6 implementation
6 * Ville Nuorvala <vnuorval@tcs.hut.fi>
7 * Yasuyuki Kozakai <kozakai@linux-ipv6.org>
10 * linux/net/ipv6/sit.c and linux/net/ipv4/ipip.c
14 * This program is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU General Public License
16 * as published by the Free Software Foundation; either version
17 * 2 of the License, or (at your option) any later version.
21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
23 #include <linux/module.h>
24 #include <linux/capability.h>
25 #include <linux/errno.h>
26 #include <linux/types.h>
27 #include <linux/sockios.h>
28 #include <linux/icmp.h>
32 #include <linux/net.h>
33 #include <linux/in6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/icmpv6.h>
37 #include <linux/init.h>
38 #include <linux/route.h>
39 #include <linux/rtnetlink.h>
40 #include <linux/netfilter_ipv6.h>
41 #include <linux/slab.h>
42 #include <linux/hash.h>
43 #include <linux/etherdevice.h>
45 #include <asm/uaccess.h>
46 #include <linux/atomic.h>
50 #include <net/ip_tunnels.h>
52 #include <net/ip6_route.h>
53 #include <net/addrconf.h>
54 #include <net/ip6_tunnel.h>
56 #include <net/dsfield.h>
57 #include <net/inet_ecn.h>
58 #include <net/net_namespace.h>
59 #include <net/netns/generic.h>
61 MODULE_AUTHOR("Ville Nuorvala");
62 MODULE_DESCRIPTION("IPv6 tunneling device");
63 MODULE_LICENSE("GPL");
64 MODULE_ALIAS_RTNL_LINK("ip6tnl");
65 MODULE_ALIAS_NETDEV("ip6tnl0");
67 #define HASH_SIZE_SHIFT 5
68 #define HASH_SIZE (1 << HASH_SIZE_SHIFT)
70 static bool log_ecn_error = true;
71 module_param(log_ecn_error, bool, 0644);
72 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
74 static u32 HASH(const struct in6_addr *addr1, const struct in6_addr *addr2)
76 u32 hash = ipv6_addr_hash(addr1) ^ ipv6_addr_hash(addr2);
78 return hash_32(hash, HASH_SIZE_SHIFT);
81 static int ip6_tnl_dev_init(struct net_device *dev);
82 static void ip6_tnl_dev_setup(struct net_device *dev);
83 static struct rtnl_link_ops ip6_link_ops __read_mostly;
85 static int ip6_tnl_net_id __read_mostly;
87 /* the IPv6 tunnel fallback device */
88 struct net_device *fb_tnl_dev;
89 /* lists for storing tunnels in use */
90 struct ip6_tnl __rcu *tnls_r_l[HASH_SIZE];
91 struct ip6_tnl __rcu *tnls_wc[1];
92 struct ip6_tnl __rcu **tnls[2];
95 static struct net_device_stats *ip6_get_stats(struct net_device *dev)
97 struct pcpu_sw_netstats tmp, sum = { 0 };
100 for_each_possible_cpu(i) {
102 const struct pcpu_sw_netstats *tstats =
103 per_cpu_ptr(dev->tstats, i);
106 start = u64_stats_fetch_begin_irq(&tstats->syncp);
107 tmp.rx_packets = tstats->rx_packets;
108 tmp.rx_bytes = tstats->rx_bytes;
109 tmp.tx_packets = tstats->tx_packets;
110 tmp.tx_bytes = tstats->tx_bytes;
111 } while (u64_stats_fetch_retry_irq(&tstats->syncp, start));
113 sum.rx_packets += tmp.rx_packets;
114 sum.rx_bytes += tmp.rx_bytes;
115 sum.tx_packets += tmp.tx_packets;
116 sum.tx_bytes += tmp.tx_bytes;
118 dev->stats.rx_packets = sum.rx_packets;
119 dev->stats.rx_bytes = sum.rx_bytes;
120 dev->stats.tx_packets = sum.tx_packets;
121 dev->stats.tx_bytes = sum.tx_bytes;
126 * ip6_tnl_lookup - fetch tunnel matching the end-point addresses
127 * @remote: the address of the tunnel exit-point
128 * @local: the address of the tunnel entry-point
131 * tunnel matching given end-points if found,
132 * else fallback tunnel if its device is up,
136 #define for_each_ip6_tunnel_rcu(start) \
137 for (t = rcu_dereference(start); t; t = rcu_dereference(t->next))
139 static struct ip6_tnl *
140 ip6_tnl_lookup(struct net *net, const struct in6_addr *remote, const struct in6_addr *local)
142 unsigned int hash = HASH(remote, local);
144 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
147 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
148 if (ipv6_addr_equal(local, &t->parms.laddr) &&
149 ipv6_addr_equal(remote, &t->parms.raddr) &&
150 (t->dev->flags & IFF_UP))
154 memset(&any, 0, sizeof(any));
155 hash = HASH(&any, local);
156 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
157 if (ipv6_addr_equal(local, &t->parms.laddr) &&
158 (t->dev->flags & IFF_UP))
162 hash = HASH(remote, &any);
163 for_each_ip6_tunnel_rcu(ip6n->tnls_r_l[hash]) {
164 if (ipv6_addr_equal(remote, &t->parms.raddr) &&
165 (t->dev->flags & IFF_UP))
169 t = rcu_dereference(ip6n->tnls_wc[0]);
170 if (t && (t->dev->flags & IFF_UP))
177 * ip6_tnl_bucket - get head of list matching given tunnel parameters
178 * @p: parameters containing tunnel end-points
181 * ip6_tnl_bucket() returns the head of the list matching the
182 * &struct in6_addr entries laddr and raddr in @p.
184 * Return: head of IPv6 tunnel list
187 static struct ip6_tnl __rcu **
188 ip6_tnl_bucket(struct ip6_tnl_net *ip6n, const struct __ip6_tnl_parm *p)
190 const struct in6_addr *remote = &p->raddr;
191 const struct in6_addr *local = &p->laddr;
195 if (!ipv6_addr_any(remote) || !ipv6_addr_any(local)) {
197 h = HASH(remote, local);
199 return &ip6n->tnls[prio][h];
203 * ip6_tnl_link - add tunnel to hash table
204 * @t: tunnel to be added
208 ip6_tnl_link(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
210 struct ip6_tnl __rcu **tp = ip6_tnl_bucket(ip6n, &t->parms);
212 rcu_assign_pointer(t->next , rtnl_dereference(*tp));
213 rcu_assign_pointer(*tp, t);
217 * ip6_tnl_unlink - remove tunnel from hash table
218 * @t: tunnel to be removed
222 ip6_tnl_unlink(struct ip6_tnl_net *ip6n, struct ip6_tnl *t)
224 struct ip6_tnl __rcu **tp;
225 struct ip6_tnl *iter;
227 for (tp = ip6_tnl_bucket(ip6n, &t->parms);
228 (iter = rtnl_dereference(*tp)) != NULL;
231 rcu_assign_pointer(*tp, t->next);
237 static void ip6_dev_free(struct net_device *dev)
239 struct ip6_tnl *t = netdev_priv(dev);
241 gro_cells_destroy(&t->gro_cells);
242 dst_cache_destroy(&t->dst_cache);
243 free_percpu(dev->tstats);
247 static int ip6_tnl_create2(struct net_device *dev)
249 struct ip6_tnl *t = netdev_priv(dev);
250 struct net *net = dev_net(dev);
251 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
254 t = netdev_priv(dev);
256 dev->rtnl_link_ops = &ip6_link_ops;
257 err = register_netdevice(dev);
261 strcpy(t->parms.name, dev->name);
264 ip6_tnl_link(ip6n, t);
272 * ip6_tnl_create - create a new tunnel
273 * @p: tunnel parameters
274 * @pt: pointer to new tunnel
277 * Create tunnel matching given parameters.
280 * created tunnel or error pointer
283 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
285 struct net_device *dev;
291 strlcpy(name, p->name, IFNAMSIZ);
293 sprintf(name, "ip6tnl%%d");
295 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
300 dev_net_set(dev, net);
302 t = netdev_priv(dev);
304 t->net = dev_net(dev);
305 err = ip6_tnl_create2(dev);
318 * ip6_tnl_locate - find or create tunnel matching given parameters
319 * @p: tunnel parameters
320 * @create: != 0 if allowed to create new tunnel if no match found
323 * ip6_tnl_locate() first tries to locate an existing tunnel
324 * based on @parms. If this is unsuccessful, but @create is set a new
325 * tunnel device is created and registered for use.
328 * matching tunnel or error pointer
331 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
332 struct __ip6_tnl_parm *p, int create)
334 const struct in6_addr *remote = &p->raddr;
335 const struct in6_addr *local = &p->laddr;
336 struct ip6_tnl __rcu **tp;
338 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
340 for (tp = ip6_tnl_bucket(ip6n, p);
341 (t = rtnl_dereference(*tp)) != NULL;
343 if (ipv6_addr_equal(local, &t->parms.laddr) &&
344 ipv6_addr_equal(remote, &t->parms.raddr)) {
346 return ERR_PTR(-EEXIST);
352 return ERR_PTR(-ENODEV);
353 return ip6_tnl_create(net, p);
357 * ip6_tnl_dev_uninit - tunnel device uninitializer
358 * @dev: the device to be destroyed
361 * ip6_tnl_dev_uninit() removes tunnel from its list
365 ip6_tnl_dev_uninit(struct net_device *dev)
367 struct ip6_tnl *t = netdev_priv(dev);
368 struct net *net = t->net;
369 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
371 if (dev == ip6n->fb_tnl_dev)
372 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
374 ip6_tnl_unlink(ip6n, t);
375 dst_cache_reset(&t->dst_cache);
380 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
381 * @skb: received socket buffer
384 * 0 if none was found,
385 * else index to encapsulation limit
388 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
390 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) raw;
391 __u8 nexthdr = ipv6h->nexthdr;
392 __u16 off = sizeof(*ipv6h);
394 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
396 struct ipv6_opt_hdr *hdr;
397 if (raw + off + sizeof(*hdr) > skb->data &&
398 !pskb_may_pull(skb, raw - skb->data + off + sizeof (*hdr)))
401 hdr = (struct ipv6_opt_hdr *) (raw + off);
402 if (nexthdr == NEXTHDR_FRAGMENT) {
403 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
404 if (frag_hdr->frag_off)
407 } else if (nexthdr == NEXTHDR_AUTH) {
408 optlen = (hdr->hdrlen + 2) << 2;
410 optlen = ipv6_optlen(hdr);
412 if (nexthdr == NEXTHDR_DEST) {
415 struct ipv6_tlv_tnl_enc_lim *tel;
417 /* No more room for encapsulation limit */
418 if (i + sizeof (*tel) > off + optlen)
421 tel = (struct ipv6_tlv_tnl_enc_lim *) &raw[i];
422 /* return index of option if found and valid */
423 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
426 /* else jump to next option */
428 i += tel->length + 2;
433 nexthdr = hdr->nexthdr;
438 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
441 * ip6_tnl_err - tunnel error handler
444 * ip6_tnl_err() should handle errors in the tunnel according
445 * to the specifications in RFC 2473.
449 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
450 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
452 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *) skb->data;
455 u8 rel_type = ICMPV6_DEST_UNREACH;
456 u8 rel_code = ICMPV6_ADDR_UNREACH;
462 /* If the packet doesn't contain the original IPv6 header we are
463 in trouble since we might need the source address for further
464 processing of the error. */
467 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
471 tproto = ACCESS_ONCE(t->parms.proto);
472 if (tproto != ipproto && tproto != 0)
479 struct ipv6_tlv_tnl_enc_lim *tel;
481 case ICMPV6_DEST_UNREACH:
482 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
486 case ICMPV6_TIME_EXCEED:
487 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
488 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
493 case ICMPV6_PARAMPROB:
495 if ((*code) == ICMPV6_HDR_FIELD)
496 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
498 if (teli && teli == *info - 2) {
499 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
500 if (tel->encap_limit == 0) {
501 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
506 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
510 case ICMPV6_PKT_TOOBIG:
511 mtu = *info - offset;
512 if (mtu < IPV6_MIN_MTU)
516 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
518 rel_type = ICMPV6_PKT_TOOBIG;
537 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
538 u8 type, u8 code, int offset, __be32 info)
543 __u32 rel_info = ntohl(info);
545 struct sk_buff *skb2;
546 const struct iphdr *eiph;
550 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
551 &rel_msg, &rel_info, offset);
559 case ICMPV6_DEST_UNREACH:
560 if (rel_code != ICMPV6_ADDR_UNREACH)
562 rel_type = ICMP_DEST_UNREACH;
563 rel_code = ICMP_HOST_UNREACH;
565 case ICMPV6_PKT_TOOBIG:
568 rel_type = ICMP_DEST_UNREACH;
569 rel_code = ICMP_FRAG_NEEDED;
572 rel_type = ICMP_REDIRECT;
573 rel_code = ICMP_REDIR_HOST;
578 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
581 skb2 = skb_clone(skb, GFP_ATOMIC);
587 skb_pull(skb2, offset);
588 skb_reset_network_header(skb2);
591 /* Try to guess incoming interface */
592 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
595 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
599 skb2->dev = rt->dst.dev;
601 /* route "incoming" packet */
602 if (rt->rt_flags & RTCF_LOCAL) {
605 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
606 eiph->daddr, eiph->saddr,
609 RT_TOS(eiph->tos), 0);
611 rt->dst.dev->type != ARPHRD_TUNNEL) {
616 skb_dst_set(skb2, &rt->dst);
619 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
621 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
625 /* change mtu on this route */
626 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
627 if (rel_info > dst_mtu(skb_dst(skb2)))
630 skb_dst(skb2)->ops->update_pmtu(skb_dst(skb2), NULL, skb2, rel_info);
632 if (rel_type == ICMP_REDIRECT)
633 skb_dst(skb2)->ops->redirect(skb_dst(skb2), NULL, skb2);
635 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
643 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
644 u8 type, u8 code, int offset, __be32 info)
649 __u32 rel_info = ntohl(info);
652 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
653 &rel_msg, &rel_info, offset);
657 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
659 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
665 skb_pull(skb2, offset);
666 skb_reset_network_header(skb2);
668 /* Try to guess incoming interface */
669 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
672 if (rt && rt->dst.dev)
673 skb2->dev = rt->dst.dev;
675 icmpv6_send(skb2, rel_type, rel_code, rel_info);
685 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
686 const struct ipv6hdr *ipv6h,
689 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
691 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
692 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
694 return IP6_ECN_decapsulate(ipv6h, skb);
697 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
698 const struct ipv6hdr *ipv6h,
701 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
702 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
704 return IP6_ECN_decapsulate(ipv6h, skb);
707 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
708 const struct in6_addr *laddr,
709 const struct in6_addr *raddr)
711 struct __ip6_tnl_parm *p = &t->parms;
712 int ltype = ipv6_addr_type(laddr);
713 int rtype = ipv6_addr_type(raddr);
716 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
717 flags = IP6_TNL_F_CAP_PER_PACKET;
718 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
719 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
720 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
721 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
722 if (ltype&IPV6_ADDR_UNICAST)
723 flags |= IP6_TNL_F_CAP_XMIT;
724 if (rtype&IPV6_ADDR_UNICAST)
725 flags |= IP6_TNL_F_CAP_RCV;
729 EXPORT_SYMBOL(ip6_tnl_get_cap);
731 /* called with rcu_read_lock() */
732 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
733 const struct in6_addr *laddr,
734 const struct in6_addr *raddr)
736 struct __ip6_tnl_parm *p = &t->parms;
738 struct net *net = t->net;
740 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
741 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
742 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
743 struct net_device *ldev = NULL;
746 ldev = dev_get_by_index_rcu(net, p->link);
748 if ((ipv6_addr_is_multicast(laddr) ||
749 likely(ipv6_chk_addr(net, laddr, ldev, 0))) &&
750 likely(!ipv6_chk_addr(net, raddr, NULL, 0)))
755 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
757 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
758 const struct tnl_ptk_info *tpi,
759 struct metadata_dst *tun_dst,
760 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
761 const struct ipv6hdr *ipv6h,
762 struct sk_buff *skb),
765 struct pcpu_sw_netstats *tstats;
766 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
769 if ((!(tpi->flags & TUNNEL_CSUM) &&
770 (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
771 ((tpi->flags & TUNNEL_CSUM) &&
772 !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
773 tunnel->dev->stats.rx_crc_errors++;
774 tunnel->dev->stats.rx_errors++;
778 if (tunnel->parms.i_flags & TUNNEL_SEQ) {
779 if (!(tpi->flags & TUNNEL_SEQ) ||
781 (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
782 tunnel->dev->stats.rx_fifo_errors++;
783 tunnel->dev->stats.rx_errors++;
786 tunnel->i_seqno = ntohl(tpi->seq) + 1;
789 skb->protocol = tpi->proto;
791 /* Warning: All skb pointers will be invalidated! */
792 if (tunnel->dev->type == ARPHRD_ETHER) {
793 if (!pskb_may_pull(skb, ETH_HLEN)) {
794 tunnel->dev->stats.rx_length_errors++;
795 tunnel->dev->stats.rx_errors++;
799 ipv6h = ipv6_hdr(skb);
800 skb->protocol = eth_type_trans(skb, tunnel->dev);
801 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
803 skb->dev = tunnel->dev;
806 skb_reset_network_header(skb);
807 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
809 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
811 err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
814 net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
816 ipv6_get_dsfield(ipv6h));
818 ++tunnel->dev->stats.rx_frame_errors;
819 ++tunnel->dev->stats.rx_errors;
824 tstats = this_cpu_ptr(tunnel->dev->tstats);
825 u64_stats_update_begin(&tstats->syncp);
826 tstats->rx_packets++;
827 tstats->rx_bytes += skb->len;
828 u64_stats_update_end(&tstats->syncp);
830 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
832 gro_cells_receive(&tunnel->gro_cells, skb);
840 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
841 const struct tnl_ptk_info *tpi,
842 struct metadata_dst *tun_dst,
845 return __ip6_tnl_rcv(t, skb, tpi, NULL, ip6ip6_dscp_ecn_decapsulate,
848 EXPORT_SYMBOL(ip6_tnl_rcv);
850 static const struct tnl_ptk_info tpi_v6 = {
851 /* no tunnel info required for ipxip6. */
852 .proto = htons(ETH_P_IPV6),
855 static const struct tnl_ptk_info tpi_v4 = {
856 /* no tunnel info required for ipxip6. */
857 .proto = htons(ETH_P_IP),
860 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
861 const struct tnl_ptk_info *tpi,
862 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
863 const struct ipv6hdr *ipv6h,
864 struct sk_buff *skb))
867 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
871 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
874 u8 tproto = ACCESS_ONCE(t->parms.proto);
876 if (tproto != ipproto && tproto != 0)
878 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
880 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
882 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
884 ret = __ip6_tnl_rcv(t, skb, tpi, NULL, dscp_ecn_decapsulate,
898 static int ip4ip6_rcv(struct sk_buff *skb)
900 return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
901 ip4ip6_dscp_ecn_decapsulate);
904 static int ip6ip6_rcv(struct sk_buff *skb)
906 return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
907 ip6ip6_dscp_ecn_decapsulate);
910 struct ipv6_tel_txoption {
911 struct ipv6_txoptions ops;
915 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
917 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
919 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
921 opt->dst_opt[4] = encap_limit;
922 opt->dst_opt[5] = IPV6_TLV_PADN;
925 opt->ops.dst0opt = (struct ipv6_opt_hdr *) opt->dst_opt;
926 opt->ops.opt_nflen = 8;
930 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
931 * @t: the outgoing tunnel device
932 * @hdr: IPv6 header from the incoming packet
935 * Avoid trivial tunneling loop by checking that tunnel exit-point
936 * doesn't match source of incoming packet.
944 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
946 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
949 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
950 const struct in6_addr *laddr,
951 const struct in6_addr *raddr)
953 struct __ip6_tnl_parm *p = &t->parms;
955 struct net *net = t->net;
957 if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
958 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
959 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
960 struct net_device *ldev = NULL;
964 ldev = dev_get_by_index_rcu(net, p->link);
966 if (unlikely(!ipv6_chk_addr(net, laddr, ldev, 0)))
967 pr_warn("%s xmit: Local address not yet configured!\n",
969 else if (!ipv6_addr_is_multicast(raddr) &&
970 unlikely(ipv6_chk_addr(net, raddr, NULL, 0)))
971 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
979 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
982 * ip6_tnl_xmit - encapsulate packet and send
983 * @skb: the outgoing socket buffer
984 * @dev: the outgoing tunnel device
985 * @dsfield: dscp code for outer header
986 * @fl6: flow of tunneled packet
987 * @encap_limit: encapsulation limit
988 * @pmtu: Path MTU is stored if packet is too big
989 * @proto: next header value
992 * Build new header and do some sanity checks on the packet before sending
998 * %-EMSGSIZE message too big. return mtu in this case.
1001 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1002 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1005 struct ip6_tnl *t = netdev_priv(dev);
1006 struct net *net = t->net;
1007 struct net_device_stats *stats = &t->dev->stats;
1008 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1009 struct ipv6_tel_txoption opt;
1010 struct dst_entry *dst = NULL, *ndst = NULL;
1011 struct net_device *tdev;
1013 unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1014 unsigned int max_headroom = psh_hlen;
1018 if (ipv6_addr_any(&t->parms.raddr)) {
1019 struct in6_addr *addr6;
1020 struct neighbour *neigh;
1024 goto tx_err_link_failure;
1026 neigh = dst_neigh_lookup(skb_dst(skb),
1027 &ipv6_hdr(skb)->daddr);
1029 goto tx_err_link_failure;
1031 addr6 = (struct in6_addr *)&neigh->primary_key;
1032 addr_type = ipv6_addr_type(addr6);
1034 if (addr_type == IPV6_ADDR_ANY)
1035 addr6 = &ipv6_hdr(skb)->daddr;
1037 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1038 neigh_release(neigh);
1039 } else if (!fl6->flowi6_mark)
1040 dst = dst_cache_get(&t->dst_cache);
1042 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1043 goto tx_err_link_failure;
1046 dst = ip6_route_output(net, NULL, fl6);
1049 goto tx_err_link_failure;
1050 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1054 goto tx_err_link_failure;
1062 stats->collisions++;
1063 net_warn_ratelimited("%s: Local routing loop detected!\n",
1065 goto tx_err_dst_release;
1067 mtu = dst_mtu(dst) - psh_hlen;
1068 if (encap_limit >= 0) {
1072 if (mtu < IPV6_MIN_MTU)
1075 skb_dst(skb)->ops->update_pmtu(skb_dst(skb), NULL, skb, mtu);
1076 if (skb->len > mtu && !skb_is_gso(skb)) {
1079 goto tx_err_dst_release;
1082 if (t->err_count > 0) {
1083 if (time_before(jiffies,
1084 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1087 dst_link_failure(skb);
1093 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1096 * Okay, now see if we can stuff it in the buffer as-is.
1098 max_headroom += LL_RESERVED_SPACE(tdev);
1100 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1101 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1102 struct sk_buff *new_skb;
1104 new_skb = skb_realloc_headroom(skb, max_headroom);
1106 goto tx_err_dst_release;
1109 skb_set_owner_w(new_skb, skb->sk);
1114 if (!fl6->flowi6_mark && ndst)
1115 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1116 skb_dst_set(skb, dst);
1118 if (encap_limit >= 0) {
1119 init_tel_txopt(&opt, encap_limit);
1120 ipv6_push_nfrag_opts(skb, &opt.ops, &proto, NULL);
1123 /* Calculate max headroom for all the headers and adjust
1124 * needed_headroom if necessary.
1126 max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1127 + dst->header_len + t->hlen;
1128 if (max_headroom > dev->needed_headroom)
1129 dev->needed_headroom = max_headroom;
1131 err = ip6_tnl_encap(skb, t, &proto, fl6);
1135 skb_push(skb, sizeof(struct ipv6hdr));
1136 skb_reset_network_header(skb);
1137 ipv6h = ipv6_hdr(skb);
1138 ip6_flow_hdr(ipv6h, INET_ECN_encapsulate(0, dsfield),
1139 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1140 ipv6h->hop_limit = t->parms.hop_limit;
1141 ipv6h->nexthdr = proto;
1142 ipv6h->saddr = fl6->saddr;
1143 ipv6h->daddr = fl6->daddr;
1144 ip6tunnel_xmit(NULL, skb, dev);
1146 tx_err_link_failure:
1147 stats->tx_carrier_errors++;
1148 dst_link_failure(skb);
1153 EXPORT_SYMBOL(ip6_tnl_xmit);
1156 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1158 struct ip6_tnl *t = netdev_priv(dev);
1159 const struct iphdr *iph = ip_hdr(skb);
1160 int encap_limit = -1;
1167 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1169 tproto = ACCESS_ONCE(t->parms.proto);
1170 if (tproto != IPPROTO_IPIP && tproto != 0)
1173 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1174 encap_limit = t->parms.encap_limit;
1176 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1177 fl6.flowi6_proto = IPPROTO_IPIP;
1179 dsfield = ipv4_get_dsfield(iph);
1181 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1182 fl6.flowlabel |= htonl((__u32)iph->tos << IPV6_TCLASS_SHIFT)
1184 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1185 fl6.flowi6_mark = skb->mark;
1187 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1190 skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1192 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1195 /* XXX: send ICMP error even if DF is not set. */
1196 if (err == -EMSGSIZE)
1197 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1206 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1208 struct ip6_tnl *t = netdev_priv(dev);
1209 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1210 int encap_limit = -1;
1218 tproto = ACCESS_ONCE(t->parms.proto);
1219 if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1220 ip6_tnl_addr_conflict(t, ipv6h))
1223 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1225 struct ipv6_tlv_tnl_enc_lim *tel;
1226 tel = (struct ipv6_tlv_tnl_enc_lim *)&skb_network_header(skb)[offset];
1227 if (tel->encap_limit == 0) {
1228 icmpv6_send(skb, ICMPV6_PARAMPROB,
1229 ICMPV6_HDR_FIELD, offset + 2);
1232 encap_limit = tel->encap_limit - 1;
1233 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1234 encap_limit = t->parms.encap_limit;
1236 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1237 fl6.flowi6_proto = IPPROTO_IPV6;
1239 dsfield = ipv6_get_dsfield(ipv6h);
1240 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1241 fl6.flowlabel |= (*(__be32 *) ipv6h & IPV6_TCLASS_MASK);
1242 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1243 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1244 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1245 fl6.flowi6_mark = skb->mark;
1247 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1250 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1252 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1255 if (err == -EMSGSIZE)
1256 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1264 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1266 struct ip6_tnl *t = netdev_priv(dev);
1267 struct net_device_stats *stats = &t->dev->stats;
1270 switch (skb->protocol) {
1271 case htons(ETH_P_IP):
1272 ret = ip4ip6_tnl_xmit(skb, dev);
1274 case htons(ETH_P_IPV6):
1275 ret = ip6ip6_tnl_xmit(skb, dev);
1284 return NETDEV_TX_OK;
1288 stats->tx_dropped++;
1290 return NETDEV_TX_OK;
1293 static void ip6_tnl_link_config(struct ip6_tnl *t)
1295 struct net_device *dev = t->dev;
1296 struct __ip6_tnl_parm *p = &t->parms;
1297 struct flowi6 *fl6 = &t->fl.u.ip6;
1300 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1301 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1303 /* Set up flowi template */
1304 fl6->saddr = p->laddr;
1305 fl6->daddr = p->raddr;
1306 fl6->flowi6_oif = p->link;
1309 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1310 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1311 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1312 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1314 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1315 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1317 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1318 dev->flags |= IFF_POINTOPOINT;
1320 dev->flags &= ~IFF_POINTOPOINT;
1323 t->hlen = t->encap_hlen + t->tun_hlen;
1324 t_hlen = t->hlen + sizeof(struct ipv6hdr);
1326 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1327 int strict = (ipv6_addr_type(&p->raddr) &
1328 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1330 struct rt6_info *rt = rt6_lookup(t->net,
1331 &p->raddr, &p->laddr,
1338 dev->hard_header_len = rt->dst.dev->hard_header_len +
1341 dev->mtu = rt->dst.dev->mtu - t_hlen;
1342 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1345 if (dev->mtu < IPV6_MIN_MTU)
1346 dev->mtu = IPV6_MIN_MTU;
1353 * ip6_tnl_change - update the tunnel parameters
1354 * @t: tunnel to be changed
1355 * @p: tunnel configuration parameters
1358 * ip6_tnl_change() updates the tunnel parameters
1362 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1364 t->parms.laddr = p->laddr;
1365 t->parms.raddr = p->raddr;
1366 t->parms.flags = p->flags;
1367 t->parms.hop_limit = p->hop_limit;
1368 t->parms.encap_limit = p->encap_limit;
1369 t->parms.flowinfo = p->flowinfo;
1370 t->parms.link = p->link;
1371 t->parms.proto = p->proto;
1372 dst_cache_reset(&t->dst_cache);
1373 ip6_tnl_link_config(t);
1377 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1379 struct net *net = t->net;
1380 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1383 ip6_tnl_unlink(ip6n, t);
1385 err = ip6_tnl_change(t, p);
1386 ip6_tnl_link(ip6n, t);
1387 netdev_state_change(t->dev);
1391 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1393 /* for default tnl0 device allow to change only the proto */
1394 t->parms.proto = p->proto;
1395 netdev_state_change(t->dev);
1400 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1402 p->laddr = u->laddr;
1403 p->raddr = u->raddr;
1404 p->flags = u->flags;
1405 p->hop_limit = u->hop_limit;
1406 p->encap_limit = u->encap_limit;
1407 p->flowinfo = u->flowinfo;
1409 p->proto = u->proto;
1410 memcpy(p->name, u->name, sizeof(u->name));
1414 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1416 u->laddr = p->laddr;
1417 u->raddr = p->raddr;
1418 u->flags = p->flags;
1419 u->hop_limit = p->hop_limit;
1420 u->encap_limit = p->encap_limit;
1421 u->flowinfo = p->flowinfo;
1423 u->proto = p->proto;
1424 memcpy(u->name, p->name, sizeof(u->name));
1428 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1429 * @dev: virtual device associated with tunnel
1430 * @ifr: parameters passed from userspace
1431 * @cmd: command to be performed
1434 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1437 * The possible commands are the following:
1438 * %SIOCGETTUNNEL: get tunnel parameters for device
1439 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1440 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1441 * %SIOCDELTUNNEL: delete tunnel
1443 * The fallback device "ip6tnl0", created during module
1444 * initialization, can be used for creating other tunnel devices.
1448 * %-EFAULT if unable to copy data to or from userspace,
1449 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1450 * %-EINVAL if passed tunnel parameters are invalid,
1451 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1452 * %-ENODEV if attempting to change or delete a nonexisting device
1456 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1459 struct ip6_tnl_parm p;
1460 struct __ip6_tnl_parm p1;
1461 struct ip6_tnl *t = netdev_priv(dev);
1462 struct net *net = t->net;
1463 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1465 memset(&p1, 0, sizeof(p1));
1469 if (dev == ip6n->fb_tnl_dev) {
1470 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1474 ip6_tnl_parm_from_user(&p1, &p);
1475 t = ip6_tnl_locate(net, &p1, 0);
1477 t = netdev_priv(dev);
1479 memset(&p, 0, sizeof(p));
1481 ip6_tnl_parm_to_user(&p, &t->parms);
1482 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1489 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1492 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1495 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1498 ip6_tnl_parm_from_user(&p1, &p);
1499 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1500 if (cmd == SIOCCHGTUNNEL) {
1502 if (t->dev != dev) {
1507 t = netdev_priv(dev);
1508 if (dev == ip6n->fb_tnl_dev)
1509 err = ip6_tnl0_update(t, &p1);
1511 err = ip6_tnl_update(t, &p1);
1515 ip6_tnl_parm_to_user(&p, &t->parms);
1516 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1525 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1528 if (dev == ip6n->fb_tnl_dev) {
1530 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1533 ip6_tnl_parm_from_user(&p1, &p);
1534 t = ip6_tnl_locate(net, &p1, 0);
1538 if (t->dev == ip6n->fb_tnl_dev)
1543 unregister_netdevice(dev);
1552 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1553 * @dev: virtual device associated with tunnel
1554 * @new_mtu: the new mtu
1558 * %-EINVAL if mtu too small
1561 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1563 struct ip6_tnl *tnl = netdev_priv(dev);
1565 if (tnl->parms.proto == IPPROTO_IPIP) {
1569 if (new_mtu < IPV6_MIN_MTU)
1572 if (new_mtu > 0xFFF8 - dev->hard_header_len)
1577 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1579 int ip6_tnl_get_iflink(const struct net_device *dev)
1581 struct ip6_tnl *t = netdev_priv(dev);
1583 return t->parms.link;
1585 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1587 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1590 if (num >= MAX_IPTUN_ENCAP_OPS)
1593 return !cmpxchg((const struct ip6_tnl_encap_ops **)
1594 &ip6tun_encaps[num],
1595 NULL, ops) ? 0 : -1;
1597 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1599 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1604 if (num >= MAX_IPTUN_ENCAP_OPS)
1607 ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1608 &ip6tun_encaps[num],
1609 ops, NULL) == ops) ? 0 : -1;
1615 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1617 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1618 struct ip_tunnel_encap *ipencap)
1622 memset(&t->encap, 0, sizeof(t->encap));
1624 hlen = ip6_encap_hlen(ipencap);
1628 t->encap.type = ipencap->type;
1629 t->encap.sport = ipencap->sport;
1630 t->encap.dport = ipencap->dport;
1631 t->encap.flags = ipencap->flags;
1633 t->encap_hlen = hlen;
1634 t->hlen = t->encap_hlen + t->tun_hlen;
1638 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1640 static const struct net_device_ops ip6_tnl_netdev_ops = {
1641 .ndo_init = ip6_tnl_dev_init,
1642 .ndo_uninit = ip6_tnl_dev_uninit,
1643 .ndo_start_xmit = ip6_tnl_start_xmit,
1644 .ndo_do_ioctl = ip6_tnl_ioctl,
1645 .ndo_change_mtu = ip6_tnl_change_mtu,
1646 .ndo_get_stats = ip6_get_stats,
1647 .ndo_get_iflink = ip6_tnl_get_iflink,
1650 #define IPXIPX_FEATURES (NETIF_F_SG | \
1651 NETIF_F_FRAGLIST | \
1653 NETIF_F_GSO_SOFTWARE | \
1657 * ip6_tnl_dev_setup - setup virtual tunnel device
1658 * @dev: virtual device associated with tunnel
1661 * Initialize function pointers and device parameters
1664 static void ip6_tnl_dev_setup(struct net_device *dev)
1666 dev->netdev_ops = &ip6_tnl_netdev_ops;
1667 dev->destructor = ip6_dev_free;
1669 dev->type = ARPHRD_TUNNEL6;
1670 dev->flags |= IFF_NOARP;
1671 dev->addr_len = sizeof(struct in6_addr);
1672 dev->features |= NETIF_F_LLTX;
1673 netif_keep_dst(dev);
1675 dev->features |= IPXIPX_FEATURES;
1676 dev->hw_features |= IPXIPX_FEATURES;
1678 /* This perm addr will be used as interface identifier by IPv6 */
1679 dev->addr_assign_type = NET_ADDR_RANDOM;
1680 eth_random_addr(dev->perm_addr);
1685 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1686 * @dev: virtual device associated with tunnel
1690 ip6_tnl_dev_init_gen(struct net_device *dev)
1692 struct ip6_tnl *t = netdev_priv(dev);
1697 t->net = dev_net(dev);
1698 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1702 ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1706 ret = gro_cells_init(&t->gro_cells, dev);
1711 t->hlen = t->encap_hlen + t->tun_hlen;
1712 t_hlen = t->hlen + sizeof(struct ipv6hdr);
1714 dev->type = ARPHRD_TUNNEL6;
1715 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1716 dev->mtu = ETH_DATA_LEN - t_hlen;
1717 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1723 dst_cache_destroy(&t->dst_cache);
1725 free_percpu(dev->tstats);
1732 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1733 * @dev: virtual device associated with tunnel
1736 static int ip6_tnl_dev_init(struct net_device *dev)
1738 struct ip6_tnl *t = netdev_priv(dev);
1739 int err = ip6_tnl_dev_init_gen(dev);
1743 ip6_tnl_link_config(t);
1748 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1749 * @dev: fallback device
1754 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1756 struct ip6_tnl *t = netdev_priv(dev);
1757 struct net *net = dev_net(dev);
1758 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1760 t->parms.proto = IPPROTO_IPV6;
1763 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1767 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[])
1771 if (!data || !data[IFLA_IPTUN_PROTO])
1774 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1775 if (proto != IPPROTO_IPV6 &&
1776 proto != IPPROTO_IPIP &&
1783 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1784 struct __ip6_tnl_parm *parms)
1786 memset(parms, 0, sizeof(*parms));
1791 if (data[IFLA_IPTUN_LINK])
1792 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1794 if (data[IFLA_IPTUN_LOCAL])
1795 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1797 if (data[IFLA_IPTUN_REMOTE])
1798 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1800 if (data[IFLA_IPTUN_TTL])
1801 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1803 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1804 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1806 if (data[IFLA_IPTUN_FLOWINFO])
1807 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1809 if (data[IFLA_IPTUN_FLAGS])
1810 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1812 if (data[IFLA_IPTUN_PROTO])
1813 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1816 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1817 struct ip_tunnel_encap *ipencap)
1821 memset(ipencap, 0, sizeof(*ipencap));
1826 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1828 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1831 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1833 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1836 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1838 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1841 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1843 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1849 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1850 struct nlattr *tb[], struct nlattr *data[])
1852 struct net *net = dev_net(dev);
1853 struct ip6_tnl *nt, *t;
1854 struct ip_tunnel_encap ipencap;
1856 nt = netdev_priv(dev);
1858 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1859 int err = ip6_tnl_encap_setup(nt, &ipencap);
1865 ip6_tnl_netlink_parms(data, &nt->parms);
1867 t = ip6_tnl_locate(net, &nt->parms, 0);
1871 return ip6_tnl_create2(dev);
1874 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
1875 struct nlattr *data[])
1877 struct ip6_tnl *t = netdev_priv(dev);
1878 struct __ip6_tnl_parm p;
1879 struct net *net = t->net;
1880 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1881 struct ip_tunnel_encap ipencap;
1883 if (dev == ip6n->fb_tnl_dev)
1886 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
1887 int err = ip6_tnl_encap_setup(t, &ipencap);
1892 ip6_tnl_netlink_parms(data, &p);
1894 t = ip6_tnl_locate(net, &p, 0);
1899 t = netdev_priv(dev);
1901 return ip6_tnl_update(t, &p);
1904 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
1906 struct net *net = dev_net(dev);
1907 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1909 if (dev != ip6n->fb_tnl_dev)
1910 unregister_netdevice_queue(dev, head);
1913 static size_t ip6_tnl_get_size(const struct net_device *dev)
1916 /* IFLA_IPTUN_LINK */
1918 /* IFLA_IPTUN_LOCAL */
1919 nla_total_size(sizeof(struct in6_addr)) +
1920 /* IFLA_IPTUN_REMOTE */
1921 nla_total_size(sizeof(struct in6_addr)) +
1922 /* IFLA_IPTUN_TTL */
1924 /* IFLA_IPTUN_ENCAP_LIMIT */
1926 /* IFLA_IPTUN_FLOWINFO */
1928 /* IFLA_IPTUN_FLAGS */
1930 /* IFLA_IPTUN_PROTO */
1932 /* IFLA_IPTUN_ENCAP_TYPE */
1934 /* IFLA_IPTUN_ENCAP_FLAGS */
1936 /* IFLA_IPTUN_ENCAP_SPORT */
1938 /* IFLA_IPTUN_ENCAP_DPORT */
1943 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
1945 struct ip6_tnl *tunnel = netdev_priv(dev);
1946 struct __ip6_tnl_parm *parm = &tunnel->parms;
1948 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
1949 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
1950 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
1951 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
1952 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
1953 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
1954 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
1955 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto))
1956 goto nla_put_failure;
1958 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE,
1959 tunnel->encap.type) ||
1960 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT,
1961 tunnel->encap.sport) ||
1962 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT,
1963 tunnel->encap.dport) ||
1964 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS,
1965 tunnel->encap.flags))
1966 goto nla_put_failure;
1974 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
1976 struct ip6_tnl *tunnel = netdev_priv(dev);
1980 EXPORT_SYMBOL(ip6_tnl_get_link_net);
1982 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
1983 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
1984 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
1985 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
1986 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
1987 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
1988 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
1989 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
1990 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
1991 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
1992 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
1993 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
1994 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
1997 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
1999 .maxtype = IFLA_IPTUN_MAX,
2000 .policy = ip6_tnl_policy,
2001 .priv_size = sizeof(struct ip6_tnl),
2002 .setup = ip6_tnl_dev_setup,
2003 .validate = ip6_tnl_validate,
2004 .newlink = ip6_tnl_newlink,
2005 .changelink = ip6_tnl_changelink,
2006 .dellink = ip6_tnl_dellink,
2007 .get_size = ip6_tnl_get_size,
2008 .fill_info = ip6_tnl_fill_info,
2009 .get_link_net = ip6_tnl_get_link_net,
2012 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2013 .handler = ip4ip6_rcv,
2014 .err_handler = ip4ip6_err,
2018 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2019 .handler = ip6ip6_rcv,
2020 .err_handler = ip6ip6_err,
2024 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net)
2026 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2027 struct net_device *dev, *aux;
2032 for_each_netdev_safe(net, dev, aux)
2033 if (dev->rtnl_link_ops == &ip6_link_ops)
2034 unregister_netdevice_queue(dev, &list);
2036 for (h = 0; h < HASH_SIZE; h++) {
2037 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2039 /* If dev is in the same netns, it has already
2040 * been added to the list by the previous loop.
2042 if (!net_eq(dev_net(t->dev), net))
2043 unregister_netdevice_queue(t->dev, &list);
2044 t = rtnl_dereference(t->next);
2048 unregister_netdevice_many(&list);
2051 static int __net_init ip6_tnl_init_net(struct net *net)
2053 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2054 struct ip6_tnl *t = NULL;
2057 ip6n->tnls[0] = ip6n->tnls_wc;
2058 ip6n->tnls[1] = ip6n->tnls_r_l;
2061 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2062 NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2064 if (!ip6n->fb_tnl_dev)
2066 dev_net_set(ip6n->fb_tnl_dev, net);
2067 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2068 /* FB netdevice is special: we have one, and only one per netns.
2069 * Allowing to move it to another netns is clearly unsafe.
2071 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2073 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2077 err = register_netdev(ip6n->fb_tnl_dev);
2081 t = netdev_priv(ip6n->fb_tnl_dev);
2083 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2087 ip6_dev_free(ip6n->fb_tnl_dev);
2092 static void __net_exit ip6_tnl_exit_net(struct net *net)
2095 ip6_tnl_destroy_tunnels(net);
2099 static struct pernet_operations ip6_tnl_net_ops = {
2100 .init = ip6_tnl_init_net,
2101 .exit = ip6_tnl_exit_net,
2102 .id = &ip6_tnl_net_id,
2103 .size = sizeof(struct ip6_tnl_net),
2107 * ip6_tunnel_init - register protocol and reserve needed resources
2109 * Return: 0 on success
2112 static int __init ip6_tunnel_init(void)
2116 err = register_pernet_device(&ip6_tnl_net_ops);
2120 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2122 pr_err("%s: can't register ip4ip6\n", __func__);
2126 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2128 pr_err("%s: can't register ip6ip6\n", __func__);
2131 err = rtnl_link_register(&ip6_link_ops);
2133 goto rtnl_link_failed;
2138 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2140 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2142 unregister_pernet_device(&ip6_tnl_net_ops);
2148 * ip6_tunnel_cleanup - free resources and unregister protocol
2151 static void __exit ip6_tunnel_cleanup(void)
2153 rtnl_link_unregister(&ip6_link_ops);
2154 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2155 pr_info("%s: can't deregister ip4ip6\n", __func__);
2157 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2158 pr_info("%s: can't deregister ip6ip6\n", __func__);
2160 unregister_pernet_device(&ip6_tnl_net_ops);
2163 module_init(ip6_tunnel_init);
2164 module_exit(ip6_tunnel_cleanup);