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);
276 ip6_tnl_link(ip6n, t);
284 * ip6_tnl_create - create a new tunnel
285 * @p: tunnel parameters
286 * @pt: pointer to new tunnel
289 * Create tunnel matching given parameters.
292 * created tunnel or error pointer
295 static struct ip6_tnl *ip6_tnl_create(struct net *net, struct __ip6_tnl_parm *p)
297 struct net_device *dev;
303 if (!dev_valid_name(p->name))
305 strlcpy(name, p->name, IFNAMSIZ);
307 sprintf(name, "ip6tnl%%d");
310 dev = alloc_netdev(sizeof(*t), name, NET_NAME_UNKNOWN,
315 dev_net_set(dev, net);
317 t = netdev_priv(dev);
319 t->net = dev_net(dev);
320 err = ip6_tnl_create2(dev);
333 * ip6_tnl_locate - find or create tunnel matching given parameters
334 * @p: tunnel parameters
335 * @create: != 0 if allowed to create new tunnel if no match found
338 * ip6_tnl_locate() first tries to locate an existing tunnel
339 * based on @parms. If this is unsuccessful, but @create is set a new
340 * tunnel device is created and registered for use.
343 * matching tunnel or error pointer
346 static struct ip6_tnl *ip6_tnl_locate(struct net *net,
347 struct __ip6_tnl_parm *p, int create)
349 const struct in6_addr *remote = &p->raddr;
350 const struct in6_addr *local = &p->laddr;
351 struct ip6_tnl __rcu **tp;
353 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
355 for (tp = ip6_tnl_bucket(ip6n, p);
356 (t = rtnl_dereference(*tp)) != NULL;
358 if (ipv6_addr_equal(local, &t->parms.laddr) &&
359 ipv6_addr_equal(remote, &t->parms.raddr)) {
361 return ERR_PTR(-EEXIST);
367 return ERR_PTR(-ENODEV);
368 return ip6_tnl_create(net, p);
372 * ip6_tnl_dev_uninit - tunnel device uninitializer
373 * @dev: the device to be destroyed
376 * ip6_tnl_dev_uninit() removes tunnel from its list
380 ip6_tnl_dev_uninit(struct net_device *dev)
382 struct ip6_tnl *t = netdev_priv(dev);
383 struct net *net = t->net;
384 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
386 if (dev == ip6n->fb_tnl_dev)
387 RCU_INIT_POINTER(ip6n->tnls_wc[0], NULL);
389 ip6_tnl_unlink(ip6n, t);
390 dst_cache_reset(&t->dst_cache);
395 * parse_tvl_tnl_enc_lim - handle encapsulation limit option
396 * @skb: received socket buffer
399 * 0 if none was found,
400 * else index to encapsulation limit
403 __u16 ip6_tnl_parse_tlv_enc_lim(struct sk_buff *skb, __u8 *raw)
405 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)raw;
406 unsigned int nhoff = raw - skb->data;
407 unsigned int off = nhoff + sizeof(*ipv6h);
408 u8 next, nexthdr = ipv6h->nexthdr;
410 while (ipv6_ext_hdr(nexthdr) && nexthdr != NEXTHDR_NONE) {
411 struct ipv6_opt_hdr *hdr;
414 if (!pskb_may_pull(skb, off + sizeof(*hdr)))
417 hdr = (struct ipv6_opt_hdr *)(skb->data + off);
418 if (nexthdr == NEXTHDR_FRAGMENT) {
419 struct frag_hdr *frag_hdr = (struct frag_hdr *) hdr;
420 if (frag_hdr->frag_off)
423 } else if (nexthdr == NEXTHDR_AUTH) {
424 optlen = (hdr->hdrlen + 2) << 2;
426 optlen = ipv6_optlen(hdr);
428 /* cache hdr->nexthdr, since pskb_may_pull() might
432 if (nexthdr == NEXTHDR_DEST) {
435 /* Remember : hdr is no longer valid at this point. */
436 if (!pskb_may_pull(skb, off + optlen))
440 struct ipv6_tlv_tnl_enc_lim *tel;
442 /* No more room for encapsulation limit */
443 if (i + sizeof(*tel) > optlen)
446 tel = (struct ipv6_tlv_tnl_enc_lim *)(skb->data + off + i);
447 /* return index of option if found and valid */
448 if (tel->type == IPV6_TLV_TNL_ENCAP_LIMIT &&
450 return i + off - nhoff;
451 /* else jump to next option */
453 i += tel->length + 2;
463 EXPORT_SYMBOL(ip6_tnl_parse_tlv_enc_lim);
466 * ip6_tnl_err - tunnel error handler
469 * ip6_tnl_err() should handle errors in the tunnel according
470 * to the specifications in RFC 2473.
474 ip6_tnl_err(struct sk_buff *skb, __u8 ipproto, struct inet6_skb_parm *opt,
475 u8 *type, u8 *code, int *msg, __u32 *info, int offset)
477 const struct ipv6hdr *ipv6h = (const struct ipv6hdr *)skb->data;
478 struct net *net = dev_net(skb->dev);
479 u8 rel_type = ICMPV6_DEST_UNREACH;
480 u8 rel_code = ICMPV6_ADDR_UNREACH;
488 /* If the packet doesn't contain the original IPv6 header we are
489 in trouble since we might need the source address for further
490 processing of the error. */
493 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->daddr, &ipv6h->saddr);
497 tproto = READ_ONCE(t->parms.proto);
498 if (tproto != ipproto && tproto != 0)
504 struct ipv6_tlv_tnl_enc_lim *tel;
506 case ICMPV6_DEST_UNREACH:
507 net_dbg_ratelimited("%s: Path to destination invalid or inactive!\n",
511 case ICMPV6_TIME_EXCEED:
512 if ((*code) == ICMPV6_EXC_HOPLIMIT) {
513 net_dbg_ratelimited("%s: Too small hop limit or routing loop in tunnel!\n",
518 case ICMPV6_PARAMPROB:
520 if ((*code) == ICMPV6_HDR_FIELD)
521 teli = ip6_tnl_parse_tlv_enc_lim(skb, skb->data);
523 if (teli && teli == *info - 2) {
524 tel = (struct ipv6_tlv_tnl_enc_lim *) &skb->data[teli];
525 if (tel->encap_limit == 0) {
526 net_dbg_ratelimited("%s: Too small encapsulation limit or routing loop in tunnel!\n",
531 net_dbg_ratelimited("%s: Recipient unable to parse tunneled packet!\n",
535 case ICMPV6_PKT_TOOBIG:
536 ip6_update_pmtu(skb, net, htonl(*info), 0, 0,
537 sock_net_uid(net, NULL));
538 mtu = *info - offset;
539 if (mtu < IPV6_MIN_MTU)
541 len = sizeof(*ipv6h) + ntohs(ipv6h->payload_len);
543 rel_type = ICMPV6_PKT_TOOBIG;
550 ip6_redirect(skb, net, skb->dev->ifindex, 0,
551 sock_net_uid(net, NULL));
566 ip4ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
567 u8 type, u8 code, int offset, __be32 info)
569 __u32 rel_info = ntohl(info);
570 const struct iphdr *eiph;
571 struct sk_buff *skb2;
572 int err, rel_msg = 0;
578 err = ip6_tnl_err(skb, IPPROTO_IPIP, opt, &rel_type, &rel_code,
579 &rel_msg, &rel_info, offset);
587 case ICMPV6_DEST_UNREACH:
588 if (rel_code != ICMPV6_ADDR_UNREACH)
590 rel_type = ICMP_DEST_UNREACH;
591 rel_code = ICMP_HOST_UNREACH;
593 case ICMPV6_PKT_TOOBIG:
596 rel_type = ICMP_DEST_UNREACH;
597 rel_code = ICMP_FRAG_NEEDED;
603 if (!pskb_may_pull(skb, offset + sizeof(struct iphdr)))
606 skb2 = skb_clone(skb, GFP_ATOMIC);
612 skb_pull(skb2, offset);
613 skb_reset_network_header(skb2);
616 /* Try to guess incoming interface */
617 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL, eiph->saddr,
618 0, 0, 0, IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
622 skb2->dev = rt->dst.dev;
625 /* route "incoming" packet */
626 if (rt->rt_flags & RTCF_LOCAL) {
627 rt = ip_route_output_ports(dev_net(skb->dev), &fl4, NULL,
628 eiph->daddr, eiph->saddr, 0, 0,
629 IPPROTO_IPIP, RT_TOS(eiph->tos), 0);
630 if (IS_ERR(rt) || rt->dst.dev->type != ARPHRD_TUNNEL) {
635 skb_dst_set(skb2, &rt->dst);
637 if (ip_route_input(skb2, eiph->daddr, eiph->saddr, eiph->tos,
639 skb_dst(skb2)->dev->type != ARPHRD_TUNNEL)
643 /* change mtu on this route */
644 if (rel_type == ICMP_DEST_UNREACH && rel_code == ICMP_FRAG_NEEDED) {
645 if (rel_info > dst_mtu(skb_dst(skb2)))
648 skb_dst_update_pmtu(skb2, rel_info);
651 icmp_send(skb2, rel_type, rel_code, htonl(rel_info));
659 ip6ip6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
660 u8 type, u8 code, int offset, __be32 info)
662 __u32 rel_info = ntohl(info);
663 int err, rel_msg = 0;
667 err = ip6_tnl_err(skb, IPPROTO_IPV6, opt, &rel_type, &rel_code,
668 &rel_msg, &rel_info, offset);
672 if (rel_msg && pskb_may_pull(skb, offset + sizeof(struct ipv6hdr))) {
674 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
680 skb_pull(skb2, offset);
681 skb_reset_network_header(skb2);
683 /* Try to guess incoming interface */
684 rt = rt6_lookup(dev_net(skb->dev), &ipv6_hdr(skb2)->saddr,
687 if (rt && rt->dst.dev)
688 skb2->dev = rt->dst.dev;
690 icmpv6_send(skb2, rel_type, rel_code, rel_info);
700 static int ip4ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
701 const struct ipv6hdr *ipv6h,
704 __u8 dsfield = ipv6_get_dsfield(ipv6h) & ~INET_ECN_MASK;
706 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
707 ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, dsfield);
709 return IP6_ECN_decapsulate(ipv6h, skb);
712 static int ip6ip6_dscp_ecn_decapsulate(const struct ip6_tnl *t,
713 const struct ipv6hdr *ipv6h,
716 if (t->parms.flags & IP6_TNL_F_RCV_DSCP_COPY)
717 ipv6_copy_dscp(ipv6_get_dsfield(ipv6h), ipv6_hdr(skb));
719 return IP6_ECN_decapsulate(ipv6h, skb);
722 __u32 ip6_tnl_get_cap(struct ip6_tnl *t,
723 const struct in6_addr *laddr,
724 const struct in6_addr *raddr)
726 struct __ip6_tnl_parm *p = &t->parms;
727 int ltype = ipv6_addr_type(laddr);
728 int rtype = ipv6_addr_type(raddr);
731 if (ltype == IPV6_ADDR_ANY || rtype == IPV6_ADDR_ANY) {
732 flags = IP6_TNL_F_CAP_PER_PACKET;
733 } else if (ltype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
734 rtype & (IPV6_ADDR_UNICAST|IPV6_ADDR_MULTICAST) &&
735 !((ltype|rtype) & IPV6_ADDR_LOOPBACK) &&
736 (!((ltype|rtype) & IPV6_ADDR_LINKLOCAL) || p->link)) {
737 if (ltype&IPV6_ADDR_UNICAST)
738 flags |= IP6_TNL_F_CAP_XMIT;
739 if (rtype&IPV6_ADDR_UNICAST)
740 flags |= IP6_TNL_F_CAP_RCV;
744 EXPORT_SYMBOL(ip6_tnl_get_cap);
746 /* called with rcu_read_lock() */
747 int ip6_tnl_rcv_ctl(struct ip6_tnl *t,
748 const struct in6_addr *laddr,
749 const struct in6_addr *raddr)
751 struct __ip6_tnl_parm *p = &t->parms;
753 struct net *net = t->net;
755 if ((p->flags & IP6_TNL_F_CAP_RCV) ||
756 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
757 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_RCV))) {
758 struct net_device *ldev = NULL;
761 ldev = dev_get_by_index_rcu(net, p->link);
763 if ((ipv6_addr_is_multicast(laddr) ||
764 likely(ipv6_chk_addr_and_flags(net, laddr, ldev, false,
765 0, IFA_F_TENTATIVE))) &&
766 ((p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) ||
767 likely(!ipv6_chk_addr_and_flags(net, raddr, ldev, true,
768 0, IFA_F_TENTATIVE))))
773 EXPORT_SYMBOL_GPL(ip6_tnl_rcv_ctl);
775 static int __ip6_tnl_rcv(struct ip6_tnl *tunnel, struct sk_buff *skb,
776 const struct tnl_ptk_info *tpi,
777 struct metadata_dst *tun_dst,
778 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
779 const struct ipv6hdr *ipv6h,
780 struct sk_buff *skb),
783 struct pcpu_sw_netstats *tstats;
784 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
787 if ((!(tpi->flags & TUNNEL_CSUM) &&
788 (tunnel->parms.i_flags & TUNNEL_CSUM)) ||
789 ((tpi->flags & TUNNEL_CSUM) &&
790 !(tunnel->parms.i_flags & TUNNEL_CSUM))) {
791 tunnel->dev->stats.rx_crc_errors++;
792 tunnel->dev->stats.rx_errors++;
796 if (tunnel->parms.i_flags & TUNNEL_SEQ) {
797 if (!(tpi->flags & TUNNEL_SEQ) ||
799 (s32)(ntohl(tpi->seq) - tunnel->i_seqno) < 0)) {
800 tunnel->dev->stats.rx_fifo_errors++;
801 tunnel->dev->stats.rx_errors++;
804 tunnel->i_seqno = ntohl(tpi->seq) + 1;
807 skb->protocol = tpi->proto;
809 /* Warning: All skb pointers will be invalidated! */
810 if (tunnel->dev->type == ARPHRD_ETHER) {
811 if (!pskb_may_pull(skb, ETH_HLEN)) {
812 tunnel->dev->stats.rx_length_errors++;
813 tunnel->dev->stats.rx_errors++;
817 ipv6h = ipv6_hdr(skb);
818 skb->protocol = eth_type_trans(skb, tunnel->dev);
819 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
821 skb->dev = tunnel->dev;
824 skb_reset_network_header(skb);
825 memset(skb->cb, 0, sizeof(struct inet6_skb_parm));
827 __skb_tunnel_rx(skb, tunnel->dev, tunnel->net);
829 err = dscp_ecn_decapsulate(tunnel, ipv6h, skb);
832 net_info_ratelimited("non-ECT from %pI6 with DS=%#x\n",
834 ipv6_get_dsfield(ipv6h));
836 ++tunnel->dev->stats.rx_frame_errors;
837 ++tunnel->dev->stats.rx_errors;
842 tstats = this_cpu_ptr(tunnel->dev->tstats);
843 u64_stats_update_begin(&tstats->syncp);
844 tstats->rx_packets++;
845 tstats->rx_bytes += skb->len;
846 u64_stats_update_end(&tstats->syncp);
848 skb_scrub_packet(skb, !net_eq(tunnel->net, dev_net(tunnel->dev)));
851 skb_dst_set(skb, (struct dst_entry *)tun_dst);
853 gro_cells_receive(&tunnel->gro_cells, skb);
858 dst_release((struct dst_entry *)tun_dst);
863 int ip6_tnl_rcv(struct ip6_tnl *t, struct sk_buff *skb,
864 const struct tnl_ptk_info *tpi,
865 struct metadata_dst *tun_dst,
868 return __ip6_tnl_rcv(t, skb, tpi, tun_dst, ip6ip6_dscp_ecn_decapsulate,
871 EXPORT_SYMBOL(ip6_tnl_rcv);
873 static const struct tnl_ptk_info tpi_v6 = {
874 /* no tunnel info required for ipxip6. */
875 .proto = htons(ETH_P_IPV6),
878 static const struct tnl_ptk_info tpi_v4 = {
879 /* no tunnel info required for ipxip6. */
880 .proto = htons(ETH_P_IP),
883 static int ipxip6_rcv(struct sk_buff *skb, u8 ipproto,
884 const struct tnl_ptk_info *tpi,
885 int (*dscp_ecn_decapsulate)(const struct ip6_tnl *t,
886 const struct ipv6hdr *ipv6h,
887 struct sk_buff *skb))
890 const struct ipv6hdr *ipv6h = ipv6_hdr(skb);
891 struct metadata_dst *tun_dst = NULL;
895 t = ip6_tnl_lookup(dev_net(skb->dev), &ipv6h->saddr, &ipv6h->daddr);
898 u8 tproto = READ_ONCE(t->parms.proto);
900 if (tproto != ipproto && tproto != 0)
902 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
904 if (!ip6_tnl_rcv_ctl(t, &ipv6h->daddr, &ipv6h->saddr))
906 if (iptunnel_pull_header(skb, 0, tpi->proto, false))
908 if (t->parms.collect_md) {
909 tun_dst = ipv6_tun_rx_dst(skb, 0, 0, 0);
913 ret = __ip6_tnl_rcv(t, skb, tpi, tun_dst, dscp_ecn_decapsulate,
927 static int ip4ip6_rcv(struct sk_buff *skb)
929 return ipxip6_rcv(skb, IPPROTO_IPIP, &tpi_v4,
930 ip4ip6_dscp_ecn_decapsulate);
933 static int ip6ip6_rcv(struct sk_buff *skb)
935 return ipxip6_rcv(skb, IPPROTO_IPV6, &tpi_v6,
936 ip6ip6_dscp_ecn_decapsulate);
939 struct ipv6_tel_txoption {
940 struct ipv6_txoptions ops;
944 static void init_tel_txopt(struct ipv6_tel_txoption *opt, __u8 encap_limit)
946 memset(opt, 0, sizeof(struct ipv6_tel_txoption));
948 opt->dst_opt[2] = IPV6_TLV_TNL_ENCAP_LIMIT;
950 opt->dst_opt[4] = encap_limit;
951 opt->dst_opt[5] = IPV6_TLV_PADN;
954 opt->ops.dst1opt = (struct ipv6_opt_hdr *) opt->dst_opt;
955 opt->ops.opt_nflen = 8;
959 * ip6_tnl_addr_conflict - compare packet addresses to tunnel's own
960 * @t: the outgoing tunnel device
961 * @hdr: IPv6 header from the incoming packet
964 * Avoid trivial tunneling loop by checking that tunnel exit-point
965 * doesn't match source of incoming packet.
973 ip6_tnl_addr_conflict(const struct ip6_tnl *t, const struct ipv6hdr *hdr)
975 return ipv6_addr_equal(&t->parms.raddr, &hdr->saddr);
978 int ip6_tnl_xmit_ctl(struct ip6_tnl *t,
979 const struct in6_addr *laddr,
980 const struct in6_addr *raddr)
982 struct __ip6_tnl_parm *p = &t->parms;
984 struct net *net = t->net;
986 if (t->parms.collect_md)
989 if ((p->flags & IP6_TNL_F_CAP_XMIT) ||
990 ((p->flags & IP6_TNL_F_CAP_PER_PACKET) &&
991 (ip6_tnl_get_cap(t, laddr, raddr) & IP6_TNL_F_CAP_XMIT))) {
992 struct net_device *ldev = NULL;
996 ldev = dev_get_by_index_rcu(net, p->link);
998 if (unlikely(!ipv6_chk_addr_and_flags(net, laddr, ldev, false,
999 0, IFA_F_TENTATIVE)))
1000 pr_warn("%s xmit: Local address not yet configured!\n",
1002 else if (!(p->flags & IP6_TNL_F_ALLOW_LOCAL_REMOTE) &&
1003 !ipv6_addr_is_multicast(raddr) &&
1004 unlikely(ipv6_chk_addr_and_flags(net, raddr, ldev,
1005 true, 0, IFA_F_TENTATIVE)))
1006 pr_warn("%s xmit: Routing loop! Remote address found on this node!\n",
1014 EXPORT_SYMBOL_GPL(ip6_tnl_xmit_ctl);
1017 * ip6_tnl_xmit - encapsulate packet and send
1018 * @skb: the outgoing socket buffer
1019 * @dev: the outgoing tunnel device
1020 * @dsfield: dscp code for outer header
1021 * @fl6: flow of tunneled packet
1022 * @encap_limit: encapsulation limit
1023 * @pmtu: Path MTU is stored if packet is too big
1024 * @proto: next header value
1027 * Build new header and do some sanity checks on the packet before sending
1033 * %-EMSGSIZE message too big. return mtu in this case.
1036 int ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev, __u8 dsfield,
1037 struct flowi6 *fl6, int encap_limit, __u32 *pmtu,
1040 struct ip6_tnl *t = netdev_priv(dev);
1041 struct net *net = t->net;
1042 struct net_device_stats *stats = &t->dev->stats;
1043 struct ipv6hdr *ipv6h;
1044 struct ipv6_tel_txoption opt;
1045 struct dst_entry *dst = NULL, *ndst = NULL;
1046 struct net_device *tdev;
1048 unsigned int eth_hlen = t->dev->type == ARPHRD_ETHER ? ETH_HLEN : 0;
1049 unsigned int psh_hlen = sizeof(struct ipv6hdr) + t->encap_hlen;
1050 unsigned int max_headroom = psh_hlen;
1051 bool use_cache = false;
1055 if (t->parms.collect_md) {
1056 hop_limit = skb_tunnel_info(skb)->key.ttl;
1059 hop_limit = t->parms.hop_limit;
1063 if (ipv6_addr_any(&t->parms.raddr)) {
1064 if (skb->protocol == htons(ETH_P_IPV6)) {
1065 struct in6_addr *addr6;
1066 struct neighbour *neigh;
1070 goto tx_err_link_failure;
1072 neigh = dst_neigh_lookup(skb_dst(skb),
1073 &ipv6_hdr(skb)->daddr);
1075 goto tx_err_link_failure;
1077 addr6 = (struct in6_addr *)&neigh->primary_key;
1078 addr_type = ipv6_addr_type(addr6);
1080 if (addr_type == IPV6_ADDR_ANY)
1081 addr6 = &ipv6_hdr(skb)->daddr;
1083 memcpy(&fl6->daddr, addr6, sizeof(fl6->daddr));
1084 neigh_release(neigh);
1086 } else if (t->parms.proto != 0 && !(t->parms.flags &
1087 (IP6_TNL_F_USE_ORIG_TCLASS |
1088 IP6_TNL_F_USE_ORIG_FWMARK))) {
1089 /* enable the cache only if neither the outer protocol nor the
1090 * routing decision depends on the current inner header value
1096 dst = dst_cache_get(&t->dst_cache);
1098 if (!ip6_tnl_xmit_ctl(t, &fl6->saddr, &fl6->daddr))
1099 goto tx_err_link_failure;
1103 /* add dsfield to flowlabel for route lookup */
1104 fl6->flowlabel = ip6_make_flowinfo(dsfield, fl6->flowlabel);
1106 dst = ip6_route_output(net, NULL, fl6);
1109 goto tx_err_link_failure;
1110 dst = xfrm_lookup(net, dst, flowi6_to_flowi(fl6), NULL, 0);
1114 goto tx_err_link_failure;
1116 if (t->parms.collect_md && ipv6_addr_any(&fl6->saddr) &&
1117 ipv6_dev_get_saddr(net, ip6_dst_idev(dst)->dev,
1118 &fl6->daddr, 0, &fl6->saddr))
1119 goto tx_err_link_failure;
1126 stats->collisions++;
1127 net_warn_ratelimited("%s: Local routing loop detected!\n",
1129 goto tx_err_dst_release;
1131 mtu = dst_mtu(dst) - eth_hlen - psh_hlen - t->tun_hlen;
1132 if (encap_limit >= 0) {
1136 mtu = max(mtu, skb->protocol == htons(ETH_P_IPV6) ?
1137 IPV6_MIN_MTU : IPV4_MIN_MTU);
1139 skb_dst_update_pmtu(skb, mtu);
1140 if (skb->len - t->tun_hlen - eth_hlen > mtu && !skb_is_gso(skb)) {
1143 goto tx_err_dst_release;
1146 if (t->err_count > 0) {
1147 if (time_before(jiffies,
1148 t->err_time + IP6TUNNEL_ERR_TIMEO)) {
1151 dst_link_failure(skb);
1157 skb_scrub_packet(skb, !net_eq(t->net, dev_net(dev)));
1160 * Okay, now see if we can stuff it in the buffer as-is.
1162 max_headroom += LL_RESERVED_SPACE(tdev);
1164 if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
1165 (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
1166 struct sk_buff *new_skb;
1168 new_skb = skb_realloc_headroom(skb, max_headroom);
1170 goto tx_err_dst_release;
1173 skb_set_owner_w(new_skb, skb->sk);
1178 if (t->parms.collect_md) {
1179 if (t->encap.type != TUNNEL_ENCAP_NONE)
1180 goto tx_err_dst_release;
1182 if (use_cache && ndst)
1183 dst_cache_set_ip6(&t->dst_cache, ndst, &fl6->saddr);
1185 skb_dst_set(skb, dst);
1187 if (encap_limit >= 0) {
1188 init_tel_txopt(&opt, encap_limit);
1189 ipv6_push_frag_opts(skb, &opt.ops, &proto);
1191 hop_limit = hop_limit ? : ip6_dst_hoplimit(dst);
1193 /* Calculate max headroom for all the headers and adjust
1194 * needed_headroom if necessary.
1196 max_headroom = LL_RESERVED_SPACE(dst->dev) + sizeof(struct ipv6hdr)
1197 + dst->header_len + t->hlen;
1198 if (max_headroom > dev->needed_headroom)
1199 dev->needed_headroom = max_headroom;
1201 err = ip6_tnl_encap(skb, t, &proto, fl6);
1205 skb_push(skb, sizeof(struct ipv6hdr));
1206 skb_reset_network_header(skb);
1207 ipv6h = ipv6_hdr(skb);
1208 ip6_flow_hdr(ipv6h, dsfield,
1209 ip6_make_flowlabel(net, skb, fl6->flowlabel, true, fl6));
1210 ipv6h->hop_limit = hop_limit;
1211 ipv6h->nexthdr = proto;
1212 ipv6h->saddr = fl6->saddr;
1213 ipv6h->daddr = fl6->daddr;
1214 ip6tunnel_xmit(NULL, skb, dev);
1216 tx_err_link_failure:
1217 stats->tx_carrier_errors++;
1218 dst_link_failure(skb);
1223 EXPORT_SYMBOL(ip6_tnl_xmit);
1226 ip4ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1228 struct ip6_tnl *t = netdev_priv(dev);
1229 const struct iphdr *iph = ip_hdr(skb);
1230 int encap_limit = -1;
1237 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1239 tproto = READ_ONCE(t->parms.proto);
1240 if (tproto != IPPROTO_IPIP && tproto != 0)
1243 if (t->parms.collect_md) {
1244 struct ip_tunnel_info *tun_info;
1245 const struct ip_tunnel_key *key;
1247 tun_info = skb_tunnel_info(skb);
1248 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1249 ip_tunnel_info_af(tun_info) != AF_INET6))
1251 key = &tun_info->key;
1252 memset(&fl6, 0, sizeof(fl6));
1253 fl6.flowi6_proto = IPPROTO_IPIP;
1254 fl6.saddr = key->u.ipv6.src;
1255 fl6.daddr = key->u.ipv6.dst;
1256 fl6.flowlabel = key->label;
1259 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1260 encap_limit = t->parms.encap_limit;
1262 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1263 fl6.flowi6_proto = IPPROTO_IPIP;
1265 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1266 dsfield = ipv4_get_dsfield(iph);
1268 dsfield = ip6_tclass(t->parms.flowinfo);
1269 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1270 fl6.flowi6_mark = skb->mark;
1272 fl6.flowi6_mark = t->parms.fwmark;
1275 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1277 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1280 dsfield = INET_ECN_encapsulate(dsfield, ipv4_get_dsfield(iph));
1282 skb_set_inner_ipproto(skb, IPPROTO_IPIP);
1284 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1287 /* XXX: send ICMP error even if DF is not set. */
1288 if (err == -EMSGSIZE)
1289 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
1298 ip6ip6_tnl_xmit(struct sk_buff *skb, struct net_device *dev)
1300 struct ip6_tnl *t = netdev_priv(dev);
1301 struct ipv6hdr *ipv6h = ipv6_hdr(skb);
1302 int encap_limit = -1;
1310 tproto = READ_ONCE(t->parms.proto);
1311 if ((tproto != IPPROTO_IPV6 && tproto != 0) ||
1312 ip6_tnl_addr_conflict(t, ipv6h))
1315 if (t->parms.collect_md) {
1316 struct ip_tunnel_info *tun_info;
1317 const struct ip_tunnel_key *key;
1319 tun_info = skb_tunnel_info(skb);
1320 if (unlikely(!tun_info || !(tun_info->mode & IP_TUNNEL_INFO_TX) ||
1321 ip_tunnel_info_af(tun_info) != AF_INET6))
1323 key = &tun_info->key;
1324 memset(&fl6, 0, sizeof(fl6));
1325 fl6.flowi6_proto = IPPROTO_IPV6;
1326 fl6.saddr = key->u.ipv6.src;
1327 fl6.daddr = key->u.ipv6.dst;
1328 fl6.flowlabel = key->label;
1331 offset = ip6_tnl_parse_tlv_enc_lim(skb, skb_network_header(skb));
1332 /* ip6_tnl_parse_tlv_enc_lim() might have reallocated skb->head */
1333 ipv6h = ipv6_hdr(skb);
1335 struct ipv6_tlv_tnl_enc_lim *tel;
1337 tel = (void *)&skb_network_header(skb)[offset];
1338 if (tel->encap_limit == 0) {
1339 icmpv6_send(skb, ICMPV6_PARAMPROB,
1340 ICMPV6_HDR_FIELD, offset + 2);
1343 encap_limit = tel->encap_limit - 1;
1344 } else if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT)) {
1345 encap_limit = t->parms.encap_limit;
1348 memcpy(&fl6, &t->fl.u.ip6, sizeof(fl6));
1349 fl6.flowi6_proto = IPPROTO_IPV6;
1351 if (t->parms.flags & IP6_TNL_F_USE_ORIG_TCLASS)
1352 dsfield = ipv6_get_dsfield(ipv6h);
1354 dsfield = ip6_tclass(t->parms.flowinfo);
1355 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FLOWLABEL)
1356 fl6.flowlabel |= ip6_flowlabel(ipv6h);
1357 if (t->parms.flags & IP6_TNL_F_USE_ORIG_FWMARK)
1358 fl6.flowi6_mark = skb->mark;
1360 fl6.flowi6_mark = t->parms.fwmark;
1363 fl6.flowi6_uid = sock_net_uid(dev_net(dev), NULL);
1365 if (iptunnel_handle_offloads(skb, SKB_GSO_IPXIP6))
1368 dsfield = INET_ECN_encapsulate(dsfield, ipv6_get_dsfield(ipv6h));
1370 skb_set_inner_ipproto(skb, IPPROTO_IPV6);
1372 err = ip6_tnl_xmit(skb, dev, dsfield, &fl6, encap_limit, &mtu,
1375 if (err == -EMSGSIZE)
1376 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu);
1384 ip6_tnl_start_xmit(struct sk_buff *skb, struct net_device *dev)
1386 struct ip6_tnl *t = netdev_priv(dev);
1387 struct net_device_stats *stats = &t->dev->stats;
1390 switch (skb->protocol) {
1391 case htons(ETH_P_IP):
1392 ret = ip4ip6_tnl_xmit(skb, dev);
1394 case htons(ETH_P_IPV6):
1395 ret = ip6ip6_tnl_xmit(skb, dev);
1404 return NETDEV_TX_OK;
1408 stats->tx_dropped++;
1410 return NETDEV_TX_OK;
1413 static void ip6_tnl_link_config(struct ip6_tnl *t)
1415 struct net_device *dev = t->dev;
1416 struct __ip6_tnl_parm *p = &t->parms;
1417 struct flowi6 *fl6 = &t->fl.u.ip6;
1420 memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
1421 memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
1423 /* Set up flowi template */
1424 fl6->saddr = p->laddr;
1425 fl6->daddr = p->raddr;
1426 fl6->flowi6_oif = p->link;
1429 if (!(p->flags&IP6_TNL_F_USE_ORIG_TCLASS))
1430 fl6->flowlabel |= IPV6_TCLASS_MASK & p->flowinfo;
1431 if (!(p->flags&IP6_TNL_F_USE_ORIG_FLOWLABEL))
1432 fl6->flowlabel |= IPV6_FLOWLABEL_MASK & p->flowinfo;
1434 p->flags &= ~(IP6_TNL_F_CAP_XMIT|IP6_TNL_F_CAP_RCV|IP6_TNL_F_CAP_PER_PACKET);
1435 p->flags |= ip6_tnl_get_cap(t, &p->laddr, &p->raddr);
1437 if (p->flags&IP6_TNL_F_CAP_XMIT && p->flags&IP6_TNL_F_CAP_RCV)
1438 dev->flags |= IFF_POINTOPOINT;
1440 dev->flags &= ~IFF_POINTOPOINT;
1443 t->hlen = t->encap_hlen + t->tun_hlen;
1444 t_hlen = t->hlen + sizeof(struct ipv6hdr);
1446 if (p->flags & IP6_TNL_F_CAP_XMIT) {
1447 int strict = (ipv6_addr_type(&p->raddr) &
1448 (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL));
1450 struct rt6_info *rt = rt6_lookup(t->net,
1451 &p->raddr, &p->laddr,
1452 p->link, NULL, strict);
1458 dev->hard_header_len = rt->dst.dev->hard_header_len +
1461 dev->mtu = rt->dst.dev->mtu - t_hlen;
1462 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1465 if (dev->mtu < IPV6_MIN_MTU)
1466 dev->mtu = IPV6_MIN_MTU;
1473 * ip6_tnl_change - update the tunnel parameters
1474 * @t: tunnel to be changed
1475 * @p: tunnel configuration parameters
1478 * ip6_tnl_change() updates the tunnel parameters
1482 ip6_tnl_change(struct ip6_tnl *t, const struct __ip6_tnl_parm *p)
1484 t->parms.laddr = p->laddr;
1485 t->parms.raddr = p->raddr;
1486 t->parms.flags = p->flags;
1487 t->parms.hop_limit = p->hop_limit;
1488 t->parms.encap_limit = p->encap_limit;
1489 t->parms.flowinfo = p->flowinfo;
1490 t->parms.link = p->link;
1491 t->parms.proto = p->proto;
1492 t->parms.fwmark = p->fwmark;
1493 dst_cache_reset(&t->dst_cache);
1494 ip6_tnl_link_config(t);
1498 static int ip6_tnl_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1500 struct net *net = t->net;
1501 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1504 ip6_tnl_unlink(ip6n, t);
1506 err = ip6_tnl_change(t, p);
1507 ip6_tnl_link(ip6n, t);
1508 netdev_state_change(t->dev);
1512 static int ip6_tnl0_update(struct ip6_tnl *t, struct __ip6_tnl_parm *p)
1514 /* for default tnl0 device allow to change only the proto */
1515 t->parms.proto = p->proto;
1516 netdev_state_change(t->dev);
1521 ip6_tnl_parm_from_user(struct __ip6_tnl_parm *p, const struct ip6_tnl_parm *u)
1523 p->laddr = u->laddr;
1524 p->raddr = u->raddr;
1525 p->flags = u->flags;
1526 p->hop_limit = u->hop_limit;
1527 p->encap_limit = u->encap_limit;
1528 p->flowinfo = u->flowinfo;
1530 p->proto = u->proto;
1531 memcpy(p->name, u->name, sizeof(u->name));
1535 ip6_tnl_parm_to_user(struct ip6_tnl_parm *u, const struct __ip6_tnl_parm *p)
1537 u->laddr = p->laddr;
1538 u->raddr = p->raddr;
1539 u->flags = p->flags;
1540 u->hop_limit = p->hop_limit;
1541 u->encap_limit = p->encap_limit;
1542 u->flowinfo = p->flowinfo;
1544 u->proto = p->proto;
1545 memcpy(u->name, p->name, sizeof(u->name));
1549 * ip6_tnl_ioctl - configure ipv6 tunnels from userspace
1550 * @dev: virtual device associated with tunnel
1551 * @ifr: parameters passed from userspace
1552 * @cmd: command to be performed
1555 * ip6_tnl_ioctl() is used for managing IPv6 tunnels
1558 * The possible commands are the following:
1559 * %SIOCGETTUNNEL: get tunnel parameters for device
1560 * %SIOCADDTUNNEL: add tunnel matching given tunnel parameters
1561 * %SIOCCHGTUNNEL: change tunnel parameters to those given
1562 * %SIOCDELTUNNEL: delete tunnel
1564 * The fallback device "ip6tnl0", created during module
1565 * initialization, can be used for creating other tunnel devices.
1569 * %-EFAULT if unable to copy data to or from userspace,
1570 * %-EPERM if current process hasn't %CAP_NET_ADMIN set
1571 * %-EINVAL if passed tunnel parameters are invalid,
1572 * %-EEXIST if changing a tunnel's parameters would cause a conflict
1573 * %-ENODEV if attempting to change or delete a nonexisting device
1577 ip6_tnl_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
1580 struct ip6_tnl_parm p;
1581 struct __ip6_tnl_parm p1;
1582 struct ip6_tnl *t = netdev_priv(dev);
1583 struct net *net = t->net;
1584 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1586 memset(&p1, 0, sizeof(p1));
1590 if (dev == ip6n->fb_tnl_dev) {
1591 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
1595 ip6_tnl_parm_from_user(&p1, &p);
1596 t = ip6_tnl_locate(net, &p1, 0);
1598 t = netdev_priv(dev);
1600 memset(&p, 0, sizeof(p));
1602 ip6_tnl_parm_to_user(&p, &t->parms);
1603 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p))) {
1610 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1613 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1616 if (p.proto != IPPROTO_IPV6 && p.proto != IPPROTO_IPIP &&
1619 ip6_tnl_parm_from_user(&p1, &p);
1620 t = ip6_tnl_locate(net, &p1, cmd == SIOCADDTUNNEL);
1621 if (cmd == SIOCCHGTUNNEL) {
1623 if (t->dev != dev) {
1628 t = netdev_priv(dev);
1629 if (dev == ip6n->fb_tnl_dev)
1630 err = ip6_tnl0_update(t, &p1);
1632 err = ip6_tnl_update(t, &p1);
1636 ip6_tnl_parm_to_user(&p, &t->parms);
1637 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
1646 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1649 if (dev == ip6n->fb_tnl_dev) {
1651 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
1654 ip6_tnl_parm_from_user(&p1, &p);
1655 t = ip6_tnl_locate(net, &p1, 0);
1659 if (t->dev == ip6n->fb_tnl_dev)
1664 unregister_netdevice(dev);
1673 * ip6_tnl_change_mtu - change mtu manually for tunnel device
1674 * @dev: virtual device associated with tunnel
1675 * @new_mtu: the new mtu
1679 * %-EINVAL if mtu too small
1682 int ip6_tnl_change_mtu(struct net_device *dev, int new_mtu)
1684 struct ip6_tnl *tnl = netdev_priv(dev);
1686 if (tnl->parms.proto == IPPROTO_IPV6) {
1687 if (new_mtu < IPV6_MIN_MTU)
1690 if (new_mtu < ETH_MIN_MTU)
1693 if (tnl->parms.proto == IPPROTO_IPV6 || tnl->parms.proto == 0) {
1694 if (new_mtu > IP6_MAX_MTU - dev->hard_header_len)
1697 if (new_mtu > IP_MAX_MTU - dev->hard_header_len)
1703 EXPORT_SYMBOL(ip6_tnl_change_mtu);
1705 int ip6_tnl_get_iflink(const struct net_device *dev)
1707 struct ip6_tnl *t = netdev_priv(dev);
1709 return t->parms.link;
1711 EXPORT_SYMBOL(ip6_tnl_get_iflink);
1713 int ip6_tnl_encap_add_ops(const struct ip6_tnl_encap_ops *ops,
1716 if (num >= MAX_IPTUN_ENCAP_OPS)
1719 return !cmpxchg((const struct ip6_tnl_encap_ops **)
1720 &ip6tun_encaps[num],
1721 NULL, ops) ? 0 : -1;
1723 EXPORT_SYMBOL(ip6_tnl_encap_add_ops);
1725 int ip6_tnl_encap_del_ops(const struct ip6_tnl_encap_ops *ops,
1730 if (num >= MAX_IPTUN_ENCAP_OPS)
1733 ret = (cmpxchg((const struct ip6_tnl_encap_ops **)
1734 &ip6tun_encaps[num],
1735 ops, NULL) == ops) ? 0 : -1;
1741 EXPORT_SYMBOL(ip6_tnl_encap_del_ops);
1743 int ip6_tnl_encap_setup(struct ip6_tnl *t,
1744 struct ip_tunnel_encap *ipencap)
1748 memset(&t->encap, 0, sizeof(t->encap));
1750 hlen = ip6_encap_hlen(ipencap);
1754 t->encap.type = ipencap->type;
1755 t->encap.sport = ipencap->sport;
1756 t->encap.dport = ipencap->dport;
1757 t->encap.flags = ipencap->flags;
1759 t->encap_hlen = hlen;
1760 t->hlen = t->encap_hlen + t->tun_hlen;
1764 EXPORT_SYMBOL_GPL(ip6_tnl_encap_setup);
1766 static const struct net_device_ops ip6_tnl_netdev_ops = {
1767 .ndo_init = ip6_tnl_dev_init,
1768 .ndo_uninit = ip6_tnl_dev_uninit,
1769 .ndo_start_xmit = ip6_tnl_start_xmit,
1770 .ndo_do_ioctl = ip6_tnl_ioctl,
1771 .ndo_change_mtu = ip6_tnl_change_mtu,
1772 .ndo_get_stats = ip6_get_stats,
1773 .ndo_get_iflink = ip6_tnl_get_iflink,
1776 #define IPXIPX_FEATURES (NETIF_F_SG | \
1777 NETIF_F_FRAGLIST | \
1779 NETIF_F_GSO_SOFTWARE | \
1783 * ip6_tnl_dev_setup - setup virtual tunnel device
1784 * @dev: virtual device associated with tunnel
1787 * Initialize function pointers and device parameters
1790 static void ip6_tnl_dev_setup(struct net_device *dev)
1792 dev->netdev_ops = &ip6_tnl_netdev_ops;
1793 dev->needs_free_netdev = true;
1794 dev->priv_destructor = ip6_dev_free;
1796 dev->type = ARPHRD_TUNNEL6;
1797 dev->flags |= IFF_NOARP;
1798 dev->addr_len = sizeof(struct in6_addr);
1799 dev->features |= NETIF_F_LLTX;
1800 netif_keep_dst(dev);
1802 dev->features |= IPXIPX_FEATURES;
1803 dev->hw_features |= IPXIPX_FEATURES;
1805 /* This perm addr will be used as interface identifier by IPv6 */
1806 dev->addr_assign_type = NET_ADDR_RANDOM;
1807 eth_random_addr(dev->perm_addr);
1812 * ip6_tnl_dev_init_gen - general initializer for all tunnel devices
1813 * @dev: virtual device associated with tunnel
1817 ip6_tnl_dev_init_gen(struct net_device *dev)
1819 struct ip6_tnl *t = netdev_priv(dev);
1824 t->net = dev_net(dev);
1825 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
1829 ret = dst_cache_init(&t->dst_cache, GFP_KERNEL);
1833 ret = gro_cells_init(&t->gro_cells, dev);
1838 t->hlen = t->encap_hlen + t->tun_hlen;
1839 t_hlen = t->hlen + sizeof(struct ipv6hdr);
1841 dev->type = ARPHRD_TUNNEL6;
1842 dev->hard_header_len = LL_MAX_HEADER + t_hlen;
1843 dev->mtu = ETH_DATA_LEN - t_hlen;
1844 if (!(t->parms.flags & IP6_TNL_F_IGN_ENCAP_LIMIT))
1846 dev->min_mtu = ETH_MIN_MTU;
1847 dev->max_mtu = IP6_MAX_MTU - dev->hard_header_len;
1852 dst_cache_destroy(&t->dst_cache);
1854 free_percpu(dev->tstats);
1861 * ip6_tnl_dev_init - initializer for all non fallback tunnel devices
1862 * @dev: virtual device associated with tunnel
1865 static int ip6_tnl_dev_init(struct net_device *dev)
1867 struct ip6_tnl *t = netdev_priv(dev);
1868 int err = ip6_tnl_dev_init_gen(dev);
1872 ip6_tnl_link_config(t);
1873 if (t->parms.collect_md) {
1874 dev->features |= NETIF_F_NETNS_LOCAL;
1875 netif_keep_dst(dev);
1881 * ip6_fb_tnl_dev_init - initializer for fallback tunnel device
1882 * @dev: fallback device
1887 static int __net_init ip6_fb_tnl_dev_init(struct net_device *dev)
1889 struct ip6_tnl *t = netdev_priv(dev);
1890 struct net *net = dev_net(dev);
1891 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1893 t->parms.proto = IPPROTO_IPV6;
1896 rcu_assign_pointer(ip6n->tnls_wc[0], t);
1900 static int ip6_tnl_validate(struct nlattr *tb[], struct nlattr *data[],
1901 struct netlink_ext_ack *extack)
1905 if (!data || !data[IFLA_IPTUN_PROTO])
1908 proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1909 if (proto != IPPROTO_IPV6 &&
1910 proto != IPPROTO_IPIP &&
1917 static void ip6_tnl_netlink_parms(struct nlattr *data[],
1918 struct __ip6_tnl_parm *parms)
1920 memset(parms, 0, sizeof(*parms));
1925 if (data[IFLA_IPTUN_LINK])
1926 parms->link = nla_get_u32(data[IFLA_IPTUN_LINK]);
1928 if (data[IFLA_IPTUN_LOCAL])
1929 parms->laddr = nla_get_in6_addr(data[IFLA_IPTUN_LOCAL]);
1931 if (data[IFLA_IPTUN_REMOTE])
1932 parms->raddr = nla_get_in6_addr(data[IFLA_IPTUN_REMOTE]);
1934 if (data[IFLA_IPTUN_TTL])
1935 parms->hop_limit = nla_get_u8(data[IFLA_IPTUN_TTL]);
1937 if (data[IFLA_IPTUN_ENCAP_LIMIT])
1938 parms->encap_limit = nla_get_u8(data[IFLA_IPTUN_ENCAP_LIMIT]);
1940 if (data[IFLA_IPTUN_FLOWINFO])
1941 parms->flowinfo = nla_get_be32(data[IFLA_IPTUN_FLOWINFO]);
1943 if (data[IFLA_IPTUN_FLAGS])
1944 parms->flags = nla_get_u32(data[IFLA_IPTUN_FLAGS]);
1946 if (data[IFLA_IPTUN_PROTO])
1947 parms->proto = nla_get_u8(data[IFLA_IPTUN_PROTO]);
1949 if (data[IFLA_IPTUN_COLLECT_METADATA])
1950 parms->collect_md = true;
1952 if (data[IFLA_IPTUN_FWMARK])
1953 parms->fwmark = nla_get_u32(data[IFLA_IPTUN_FWMARK]);
1956 static bool ip6_tnl_netlink_encap_parms(struct nlattr *data[],
1957 struct ip_tunnel_encap *ipencap)
1961 memset(ipencap, 0, sizeof(*ipencap));
1966 if (data[IFLA_IPTUN_ENCAP_TYPE]) {
1968 ipencap->type = nla_get_u16(data[IFLA_IPTUN_ENCAP_TYPE]);
1971 if (data[IFLA_IPTUN_ENCAP_FLAGS]) {
1973 ipencap->flags = nla_get_u16(data[IFLA_IPTUN_ENCAP_FLAGS]);
1976 if (data[IFLA_IPTUN_ENCAP_SPORT]) {
1978 ipencap->sport = nla_get_be16(data[IFLA_IPTUN_ENCAP_SPORT]);
1981 if (data[IFLA_IPTUN_ENCAP_DPORT]) {
1983 ipencap->dport = nla_get_be16(data[IFLA_IPTUN_ENCAP_DPORT]);
1989 static int ip6_tnl_newlink(struct net *src_net, struct net_device *dev,
1990 struct nlattr *tb[], struct nlattr *data[],
1991 struct netlink_ext_ack *extack)
1993 struct net *net = dev_net(dev);
1994 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
1995 struct ip_tunnel_encap ipencap;
1996 struct ip6_tnl *nt, *t;
1999 nt = netdev_priv(dev);
2001 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2002 err = ip6_tnl_encap_setup(nt, &ipencap);
2007 ip6_tnl_netlink_parms(data, &nt->parms);
2009 if (nt->parms.collect_md) {
2010 if (rtnl_dereference(ip6n->collect_md_tun))
2013 t = ip6_tnl_locate(net, &nt->parms, 0);
2018 err = ip6_tnl_create2(dev);
2019 if (!err && tb[IFLA_MTU])
2020 ip6_tnl_change_mtu(dev, nla_get_u32(tb[IFLA_MTU]));
2025 static int ip6_tnl_changelink(struct net_device *dev, struct nlattr *tb[],
2026 struct nlattr *data[],
2027 struct netlink_ext_ack *extack)
2029 struct ip6_tnl *t = netdev_priv(dev);
2030 struct __ip6_tnl_parm p;
2031 struct net *net = t->net;
2032 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2033 struct ip_tunnel_encap ipencap;
2035 if (dev == ip6n->fb_tnl_dev)
2038 if (ip6_tnl_netlink_encap_parms(data, &ipencap)) {
2039 int err = ip6_tnl_encap_setup(t, &ipencap);
2044 ip6_tnl_netlink_parms(data, &p);
2048 t = ip6_tnl_locate(net, &p, 0);
2053 t = netdev_priv(dev);
2055 return ip6_tnl_update(t, &p);
2058 static void ip6_tnl_dellink(struct net_device *dev, struct list_head *head)
2060 struct net *net = dev_net(dev);
2061 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2063 if (dev != ip6n->fb_tnl_dev)
2064 unregister_netdevice_queue(dev, head);
2067 static size_t ip6_tnl_get_size(const struct net_device *dev)
2070 /* IFLA_IPTUN_LINK */
2072 /* IFLA_IPTUN_LOCAL */
2073 nla_total_size(sizeof(struct in6_addr)) +
2074 /* IFLA_IPTUN_REMOTE */
2075 nla_total_size(sizeof(struct in6_addr)) +
2076 /* IFLA_IPTUN_TTL */
2078 /* IFLA_IPTUN_ENCAP_LIMIT */
2080 /* IFLA_IPTUN_FLOWINFO */
2082 /* IFLA_IPTUN_FLAGS */
2084 /* IFLA_IPTUN_PROTO */
2086 /* IFLA_IPTUN_ENCAP_TYPE */
2088 /* IFLA_IPTUN_ENCAP_FLAGS */
2090 /* IFLA_IPTUN_ENCAP_SPORT */
2092 /* IFLA_IPTUN_ENCAP_DPORT */
2094 /* IFLA_IPTUN_COLLECT_METADATA */
2096 /* IFLA_IPTUN_FWMARK */
2101 static int ip6_tnl_fill_info(struct sk_buff *skb, const struct net_device *dev)
2103 struct ip6_tnl *tunnel = netdev_priv(dev);
2104 struct __ip6_tnl_parm *parm = &tunnel->parms;
2106 if (nla_put_u32(skb, IFLA_IPTUN_LINK, parm->link) ||
2107 nla_put_in6_addr(skb, IFLA_IPTUN_LOCAL, &parm->laddr) ||
2108 nla_put_in6_addr(skb, IFLA_IPTUN_REMOTE, &parm->raddr) ||
2109 nla_put_u8(skb, IFLA_IPTUN_TTL, parm->hop_limit) ||
2110 nla_put_u8(skb, IFLA_IPTUN_ENCAP_LIMIT, parm->encap_limit) ||
2111 nla_put_be32(skb, IFLA_IPTUN_FLOWINFO, parm->flowinfo) ||
2112 nla_put_u32(skb, IFLA_IPTUN_FLAGS, parm->flags) ||
2113 nla_put_u8(skb, IFLA_IPTUN_PROTO, parm->proto) ||
2114 nla_put_u32(skb, IFLA_IPTUN_FWMARK, parm->fwmark))
2115 goto nla_put_failure;
2117 if (nla_put_u16(skb, IFLA_IPTUN_ENCAP_TYPE, tunnel->encap.type) ||
2118 nla_put_be16(skb, IFLA_IPTUN_ENCAP_SPORT, tunnel->encap.sport) ||
2119 nla_put_be16(skb, IFLA_IPTUN_ENCAP_DPORT, tunnel->encap.dport) ||
2120 nla_put_u16(skb, IFLA_IPTUN_ENCAP_FLAGS, tunnel->encap.flags))
2121 goto nla_put_failure;
2123 if (parm->collect_md)
2124 if (nla_put_flag(skb, IFLA_IPTUN_COLLECT_METADATA))
2125 goto nla_put_failure;
2133 struct net *ip6_tnl_get_link_net(const struct net_device *dev)
2135 struct ip6_tnl *tunnel = netdev_priv(dev);
2139 EXPORT_SYMBOL(ip6_tnl_get_link_net);
2141 static const struct nla_policy ip6_tnl_policy[IFLA_IPTUN_MAX + 1] = {
2142 [IFLA_IPTUN_LINK] = { .type = NLA_U32 },
2143 [IFLA_IPTUN_LOCAL] = { .len = sizeof(struct in6_addr) },
2144 [IFLA_IPTUN_REMOTE] = { .len = sizeof(struct in6_addr) },
2145 [IFLA_IPTUN_TTL] = { .type = NLA_U8 },
2146 [IFLA_IPTUN_ENCAP_LIMIT] = { .type = NLA_U8 },
2147 [IFLA_IPTUN_FLOWINFO] = { .type = NLA_U32 },
2148 [IFLA_IPTUN_FLAGS] = { .type = NLA_U32 },
2149 [IFLA_IPTUN_PROTO] = { .type = NLA_U8 },
2150 [IFLA_IPTUN_ENCAP_TYPE] = { .type = NLA_U16 },
2151 [IFLA_IPTUN_ENCAP_FLAGS] = { .type = NLA_U16 },
2152 [IFLA_IPTUN_ENCAP_SPORT] = { .type = NLA_U16 },
2153 [IFLA_IPTUN_ENCAP_DPORT] = { .type = NLA_U16 },
2154 [IFLA_IPTUN_COLLECT_METADATA] = { .type = NLA_FLAG },
2155 [IFLA_IPTUN_FWMARK] = { .type = NLA_U32 },
2158 static struct rtnl_link_ops ip6_link_ops __read_mostly = {
2160 .maxtype = IFLA_IPTUN_MAX,
2161 .policy = ip6_tnl_policy,
2162 .priv_size = sizeof(struct ip6_tnl),
2163 .setup = ip6_tnl_dev_setup,
2164 .validate = ip6_tnl_validate,
2165 .newlink = ip6_tnl_newlink,
2166 .changelink = ip6_tnl_changelink,
2167 .dellink = ip6_tnl_dellink,
2168 .get_size = ip6_tnl_get_size,
2169 .fill_info = ip6_tnl_fill_info,
2170 .get_link_net = ip6_tnl_get_link_net,
2173 static struct xfrm6_tunnel ip4ip6_handler __read_mostly = {
2174 .handler = ip4ip6_rcv,
2175 .err_handler = ip4ip6_err,
2179 static struct xfrm6_tunnel ip6ip6_handler __read_mostly = {
2180 .handler = ip6ip6_rcv,
2181 .err_handler = ip6ip6_err,
2185 static void __net_exit ip6_tnl_destroy_tunnels(struct net *net, struct list_head *list)
2187 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2188 struct net_device *dev, *aux;
2192 for_each_netdev_safe(net, dev, aux)
2193 if (dev->rtnl_link_ops == &ip6_link_ops)
2194 unregister_netdevice_queue(dev, list);
2196 for (h = 0; h < IP6_TUNNEL_HASH_SIZE; h++) {
2197 t = rtnl_dereference(ip6n->tnls_r_l[h]);
2199 /* If dev is in the same netns, it has already
2200 * been added to the list by the previous loop.
2202 if (!net_eq(dev_net(t->dev), net))
2203 unregister_netdevice_queue(t->dev, list);
2204 t = rtnl_dereference(t->next);
2209 static int __net_init ip6_tnl_init_net(struct net *net)
2211 struct ip6_tnl_net *ip6n = net_generic(net, ip6_tnl_net_id);
2212 struct ip6_tnl *t = NULL;
2215 ip6n->tnls[0] = ip6n->tnls_wc;
2216 ip6n->tnls[1] = ip6n->tnls_r_l;
2218 if (!net_has_fallback_tunnels(net))
2221 ip6n->fb_tnl_dev = alloc_netdev(sizeof(struct ip6_tnl), "ip6tnl0",
2222 NET_NAME_UNKNOWN, ip6_tnl_dev_setup);
2224 if (!ip6n->fb_tnl_dev)
2226 dev_net_set(ip6n->fb_tnl_dev, net);
2227 ip6n->fb_tnl_dev->rtnl_link_ops = &ip6_link_ops;
2228 /* FB netdevice is special: we have one, and only one per netns.
2229 * Allowing to move it to another netns is clearly unsafe.
2231 ip6n->fb_tnl_dev->features |= NETIF_F_NETNS_LOCAL;
2233 err = ip6_fb_tnl_dev_init(ip6n->fb_tnl_dev);
2237 err = register_netdev(ip6n->fb_tnl_dev);
2241 t = netdev_priv(ip6n->fb_tnl_dev);
2243 strcpy(t->parms.name, ip6n->fb_tnl_dev->name);
2247 free_netdev(ip6n->fb_tnl_dev);
2252 static void __net_exit ip6_tnl_exit_batch_net(struct list_head *net_list)
2258 list_for_each_entry(net, net_list, exit_list)
2259 ip6_tnl_destroy_tunnels(net, &list);
2260 unregister_netdevice_many(&list);
2264 static struct pernet_operations ip6_tnl_net_ops = {
2265 .init = ip6_tnl_init_net,
2266 .exit_batch = ip6_tnl_exit_batch_net,
2267 .id = &ip6_tnl_net_id,
2268 .size = sizeof(struct ip6_tnl_net),
2272 * ip6_tunnel_init - register protocol and reserve needed resources
2274 * Return: 0 on success
2277 static int __init ip6_tunnel_init(void)
2281 if (!ipv6_mod_enabled())
2284 err = register_pernet_device(&ip6_tnl_net_ops);
2288 err = xfrm6_tunnel_register(&ip4ip6_handler, AF_INET);
2290 pr_err("%s: can't register ip4ip6\n", __func__);
2294 err = xfrm6_tunnel_register(&ip6ip6_handler, AF_INET6);
2296 pr_err("%s: can't register ip6ip6\n", __func__);
2299 err = rtnl_link_register(&ip6_link_ops);
2301 goto rtnl_link_failed;
2306 xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6);
2308 xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET);
2310 unregister_pernet_device(&ip6_tnl_net_ops);
2316 * ip6_tunnel_cleanup - free resources and unregister protocol
2319 static void __exit ip6_tunnel_cleanup(void)
2321 rtnl_link_unregister(&ip6_link_ops);
2322 if (xfrm6_tunnel_deregister(&ip4ip6_handler, AF_INET))
2323 pr_info("%s: can't deregister ip4ip6\n", __func__);
2325 if (xfrm6_tunnel_deregister(&ip6ip6_handler, AF_INET6))
2326 pr_info("%s: can't deregister ip6ip6\n", __func__);
2328 unregister_pernet_device(&ip6_tnl_net_ops);
2331 module_init(ip6_tunnel_init);
2332 module_exit(ip6_tunnel_cleanup);